<%BANNER%>

Depression, Disease Knowledge, And Epilepsy

Permanent Link: http://ufdc.ufl.edu/UFE0024774/00001

Material Information

Title: Depression, Disease Knowledge, And Epilepsy Measuring The Impact On Adherence
Physical Description: 1 online resource (101 p.)
Language: english
Creator: Krishnan, Mohan
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: adherence, compliance, epilepsy, medication, seizures
Clinical and Health Psychology -- Dissertations, Academic -- UF
Genre: Psychology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The effectiveness of medication treatments for chronic illnesses depends on the ability of patients to use these medications consistently and as prescribed. Research has shown that people who suffer from depression have more difficulty in succeeding in medication adherence than individuals who are not depressed. The present study attempted to extend this research to people with epilepsy, a chronic neurological illness. In a sample of people with epilepsy, no relationship between adherence and depression was found. However, other factors, including race, knowledge about epilepsy, the perception that health status is primarily determined by chance, fate, or other supernatural processes, and a measure of verbal memory ability were found to predict self-reported adherence. This study suggests that healthcare knowledge and attitudes play a complex role in reported adherence and highlights the need for more detailed investigation of measuring medication adherence and factors that underlie patients? success or failure in achieving adherence.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Mohan Krishnan.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Bauer, Russell M.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-02-28

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024774:00001

Permanent Link: http://ufdc.ufl.edu/UFE0024774/00001

Material Information

Title: Depression, Disease Knowledge, And Epilepsy Measuring The Impact On Adherence
Physical Description: 1 online resource (101 p.)
Language: english
Creator: Krishnan, Mohan
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: adherence, compliance, epilepsy, medication, seizures
Clinical and Health Psychology -- Dissertations, Academic -- UF
Genre: Psychology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The effectiveness of medication treatments for chronic illnesses depends on the ability of patients to use these medications consistently and as prescribed. Research has shown that people who suffer from depression have more difficulty in succeeding in medication adherence than individuals who are not depressed. The present study attempted to extend this research to people with epilepsy, a chronic neurological illness. In a sample of people with epilepsy, no relationship between adherence and depression was found. However, other factors, including race, knowledge about epilepsy, the perception that health status is primarily determined by chance, fate, or other supernatural processes, and a measure of verbal memory ability were found to predict self-reported adherence. This study suggests that healthcare knowledge and attitudes play a complex role in reported adherence and highlights the need for more detailed investigation of measuring medication adherence and factors that underlie patients? success or failure in achieving adherence.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Mohan Krishnan.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Bauer, Russell M.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-02-28

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024774:00001


This item has the following downloads:


Full Text

PAGE 1

DEPRE SSION, DISEASE KNOWLEDGE, AND EPILEPSY: MEASURING THE IMPACT ON ADHERENCE By MOHAN KRISHNAN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009 1

PAGE 2

2009 Mohan Krishnan 2

PAGE 3

To all those who have taught m e what it means to be ill, and what it means to become well. To my mother and father. To my partners in crime, Calvin, Wei, and Chris, you are always there for me. 3

PAGE 4

ACKNOWL EDGMENTS The author would like to acknowledge Davi d Loring, Lori Waxenberg, Elena Andresen, and Kimford Meador for their invaluable advice and support in the process of developing and executing this dissertation. The author would furt her like to thank Ste phan Eisenschenk, Denise Riley, Donna Lilly, and many others at Shands at the University of Florida, as well as Ramon Bautista, John DeCerce, Juan Ochoa, and many others at Shands Jacksonville for support in recruitment of study participants. 4

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........7 LIST OF ABBREVIATIONS.......................................................................................................... 8 ABSTRACT...................................................................................................................................10 CHAPTER 1 INTRODUCTION................................................................................................................. .12 2 REVIEW OF LITERATURE.................................................................................................13 Diagnosis, Treatment and Outcomes for Epilepsy.................................................................13 Role of Anti-Epileptic Drugs and Efficacy............................................................................14 Treatment Resistance........................................................................................................... ...15 The Role of Adherence...........................................................................................................16 Measuring Adherence............................................................................................................ .18 Why Do People Adhere or Not Adhere?................................................................................23 Depression in the Context of Epilepsy...................................................................................26 Mechanisms Behind the Impact of Depression on Adherence...............................................29 Rationale for the Present Study..............................................................................................33 Statement of the Problem....................................................................................................... .33 3 METHODS...................................................................................................................... .......37 Population..................................................................................................................... ..........37 Recruitment Procedure...........................................................................................................37 Assessment Procedure........................................................................................................... .38 Assessment Instruments......................................................................................................... .38 Epilepsy Knowledge Instruments....................................................................................38 Epilepsy Attitudes Instruments........................................................................................39 Apathy Instrument...........................................................................................................41 Adherence Instruments....................................................................................................41 Clinical Variables............................................................................................................42 Depression Instruments...................................................................................................43 Cognitive Screening........................................................................................................43 Power Analysis and Statistical Methods.................................................................................44 Participants.............................................................................................................................45 Modifications to Methods.......................................................................................................46 4 RESULTS...................................................................................................................... .........48 5

PAGE 6

Characterization of Sample .....................................................................................................48 Demographics..................................................................................................................48 Clinical Characteristics....................................................................................................48 Psychiatric Characteristics...............................................................................................49 Cognitive Characteristics.................................................................................................50 Healthcare Attitudes and Beliefs............................................................................................52 Illness Perception.............................................................................................................52 Medications.....................................................................................................................52 Locus of Control..............................................................................................................52 Apathy.............................................................................................................................53 Epilepsy Knowledge........................................................................................................53 Adherence Characteristics...................................................................................................... 54 Morisky........................................................................................................................ ....54 RAM................................................................................................................................54 Associations among Morisky and RAM Measures of Adherence..................................55 Adherence and Depression.....................................................................................................55 Bivariate Relationships among Predictors of Adherence.......................................................56 Demographics..................................................................................................................56 Clinical Characteristics....................................................................................................57 Cognitive Characteristics.................................................................................................58 Healthcare Attitudes and Beliefs.....................................................................................59 The Effect of Depression on Relationships among Adherence and Health Behaviors...60 Adherence and Health Behaviors....................................................................................61 Adherence, Depression, and Seizure Control..................................................................63 5 DISCUSSION................................................................................................................... ......70 Review of study findings........................................................................................................70 Aim 1: Relationship between Depression and Adherence..............................................70 Aim 2: Relationships among Health K nowledge, Attitudes, and Adherence.................71 Aim 3: Effects of Knowledge and Attitudes in Depressed Individuals...........................72 Aim 4: Depression, Adheren ce, and Seizure Control......................................................73 Implications of the Study...................................................................................................... ..73 Depressions Role in Epilepsy Adherence......................................................................73 Possible Roles of Knowledge a nd Attitudes in Adherence.............................................76 Implications for Seizure Control.....................................................................................78 The Challenge in Measuring Adherence.........................................................................80 Characteristics of the Study Sample................................................................................83 Study Limitations.............................................................................................................. ......86 Conclusion..............................................................................................................................88 LIST OF REFERENCES...............................................................................................................90 BIOGRAPHICAL SKETCH .......................................................................................................101 6

PAGE 7

LIST OF TABLES Table page 2-1 Studies of adherence in epilepsy........................................................................................35 3-1 Summary of Study Measures.............................................................................................47 4-1 Demographic and clinical ch aracteristics of participants..................................................65 4-2 Psychiatric comorbidities observed in the study sample...................................................65 4-3 Cognitive characteristics of study participants..................................................................66 4-4 Correlations among CES-D Total Sc ore and measures of adherence................................66 4-5 Associations among measures of de pression and measures of adherence.........................67 4-6 Relationships among health knowledge, attitudes, and adherence....................................68 4-7 Effects of attitudes and knowledge on depressed and non-depressed participants............69 4-8 Logistic regression model predicting inclusive RAM.......................................................69 4-9 Logistic regression model predicting exclusive RAM.......................................................69 7

PAGE 8

LIST OF ABBRE VIATIONS AED Anti-epileptic drug AES Apathy Evaluation Scale AUC Area under the curve BMQ Beliefs About Medicine Questionnaire CBT Cognitive-behavioral therapy CES-D Center for Epidemiological Studies Scale for Depression EKS Epilepsy Knowledge Scale EPKQ Epilepsy Patient Knowledge Questionnaire FSIQ Full scale intelligence quotient GABA Gamma-aminobutyric acid HAART Highly active anti -retroviral therapy IPQ Illness Perception Questionnaire MDD Major depressive disorder MDE Major depressive episode MHLC Multidimensional Health Locus of Control Questionnaire MINI Mini International Neuropsychiatric Interview MPR Medication possession ratio RAM Retrospective adherence measurement RAVLT Rey Auditory Verbal Learning Test ROC Receiver operating characteristic SD Standard Deviation SEK Standard error of kurtosis SES Standard error of skewness WAIS-III Wechsler Adult Intel ligence Scale Third Edition 8

PAGE 9

W AIS-R Wechsler Adult In telligence Scale Revised 9

PAGE 10

Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy DEPRESSION, DISEASE KNOWLEDGE, AND EPILEPSY: MEASURING THE IMPACT ON ADHERENCE By Mohan Krishnan August 2009 Chair: Russell Bauer Major: Psychology Adherence to medication is cr ucial to the success of treatment for chronic illness. Research investigating adherence behaviors in individuals with chronic illnesses has identified comorbid depression as a major factor decreas ing the likelihood of adherence. While this research has shown that depre ssed individuals are less likely to adhere successfully to their medication regimens, the mechanisms for this eff ect remain unclear. Research has also examined a limited set of factors predicting adherence behaviors for people with ep ilepsy, a chronic illness in which depression is a known, common comorbidity, but has not evalua ted the relationship between depression and adherence in people w ith epilepsy. The present study proposes several candidate mechanisms through whic h depression may affect comple x healthcare behaviors such as medication adherence: diffi culty acquiring and using diseas e knowledge, increased perception of illness severity, difficulty maintaining positiv e attitudes towards medical care, lack of an internal locus of control with respect to achie ving healthcare outcomes, and lack of general motivation to engage in complex behaviors. The effects of depression and the possible ro les of these mechanisms on adherence were investigated in a sample of 56 participants recruited from outpatient epilepsy clinics. Surprisingly, no association between depres sion and adherence was found. Using several 10

PAGE 11

11 different self-report measures of adherence, di fferent healthcare beha viors as well as other variables were found to predict adherence, with Caucasian race a nd epilepsy knowledge predicting one measure of adherence, a Chance locus of control predicting a second measure of adherence, and total learning on the Rey Auditory Verbal Learning Test, a measure of word list learning, predicting a third measure of adherence. No differences in predictor patte rns were seen between depressed and non-depres sed individuals; it was found, how ever, that the measures of adherence were not significantly correlated in non-depressed individu als but became strongly correlated in depressed individuals. Finally, no evidence of an im pact on self-report adherence on seizure control was observed. The study suggests that healthcare knowle dge and attitudes play a complex role in reported adherence. Better un derstanding the role of various aspects of knowledge and attitudes about epilepsy and epilepsy care may allow clinicians to improve medication adherence in the future.

PAGE 12

CHAP TER 1 INTRODUCTION Epilepsy is a chronic illness that affect s a significant portion of the population and contributes substantially to im pairment in many domains of lif e (Duncan et al., 2006). Advances in medical management of epilepsy have allowe d many individuals with epilepsy to return to important occupational, social, and recreational activi ties (Sander, 2004). Nonetheless, a number of individuals do not become seizure-free with pharmacotherapy for epilepsy (French, 2007). Some of these cases may represent forms of epil epsy that do not respond for reasons related to pathology at the neurochemical or neuroanatomi cal level (Winawer, 2006). Others may represent cases of refusal or inability to make use of anti-epileptic drugs, or comp lete non-compliance with the pharmacotherapy regimen (Cramer, Glassman, & Rienzi, 2002). Based on knowledge of health behaviors in ot her chronic illness popul ations, it is likely that there also exist cases of people with epilepsy who do not ach ieve adequate seizure control because of marginal adherence. Relatively little is known about the effects of marginal adherence on seizure control, or about psychological or he alth education factors that represent barriers between marginal and adequate adherence. The pr esent study aims to identify factors such as comorbid depression, poor disease knowledge or negative attitudes towards epilepsy and epilepsy management, and overall motivation levels that may affect adherence to anti-epileptic drug regimens in this population. Understanding these issues may assist in the development of targeted, brief psychoeducational interventions designed to increase dis ease knowledge, modify attitudes toward self-management, or addres s specific motivation issues that affect AED adherence in this population. Such interventions may have the potential to add to the number of people who achieve adequate seizure control via pharmacotherapy. 12

PAGE 13

CHAP TER 2 REVIEW OF LITERATURE Diagnosis, Treatment and Outcomes for Epilepsy Epilepsy is a disorder of the brain in whic h abnormalities in neural activity and abnormal propagation of neuroelectrical signals predispose the individua l to occurrence of seizures. Unusual neuronal discharge patter ns during a seizure impair no rmal neurocognitive functioning. Epilepsy affects children and adults of all ages, although some forms of epilepsy occur predominantly during certain neurodevelopmenta l periods (Duncan, 2006). Initial care for 80% of individuals experiencing seizures is delivered via a primary care physician, although ultimately, most people with ongoing treatment for epilepsy see a neurologist as well (Trost et al., 2005). As apparent seizure events occur in a variety of neurologic conditions (strokes, tumors, substance intoxication and withdr awal, and neuropsychiatric conditions), characterization of the seizure disorder is a primar y goal of initial care (T rost et al., 2005; Beghi et al., 2006). Standards for this vary widel y, with varying recommendations regarding the interpretation of patient histor ical information and the use of diagnostic techniques such as structural brain imaging, electroencephalogra phy, and neuropsychological examinations, which all identify some causes of seizures but not others and have imper fect concordance (Trost et al., 2005). Once an epilepsy diagnosis is made, reco mmendations on continuing management by a primary care physician or specialty care by a neurologist also vary depending upon the underlying nature of the condition (Payakachat, Summers, & Barbuto, 2006). However, for the majority of adolescent and adult epilepsies, firs t line treatment with anti-epileptic drugs (AEDs) is indicated, with alternatives such as neurosurgical interventi ons considered if AED treatment does not succeed and in some cases where surger y has particular efficacy (Trost et al., 2005). 13

PAGE 14

Role of Anti-Epilept ic Drugs and Efficacy At least fifteen different me dications are currently approve d in the United States for treatment of epilepsy (Trost et al., 2005). Th e four old AEDs (phenytoin, phenobarbitol, carbamazepine, and valproic acid) are often di stinguished from new AEDs (lamotrogine, oxcarbazepine, tiagabine, gabapentin, and a num ber of others) (Brunb ech & Sabers, 2004). Phenytoin and phenobarbitol are less commonly used at present due to serious side effects, but carbamazepine and valproic acid are still common ly used as effective first line medications (Brunbech & Sabers, 2004). The n ew AEDs generally have a bene ficial side effect profile, although not all are approved as monotherapy in the United States (Brunbech & Sabers, 2004; Loring & Meador, 2001; Meador et al., 2003). AEDs are often categorized based on primary site of pharmacological action glutamate antagon ists are differentiated from GABA agonists (Chengappa, Gershon, & Levine, 2001). This categor ization can be beneficial in terms of primary and secondary symptom management: GABA agonists are generally anxiolytic but increase fatigue and sedation and cause weight gain, while glutamatergic drugs typically are activating and cause weight lo ss, but can be anxiogenic (Chengappa, Gershon, & Levine, 2001). Overall, treatment with AEDs achieves adequa te seizure control in 60-70% of people with epilepsy (Duncan, 2006). Approximately 50% of pa tients achieve adequate seizure control on their first medication trial (French, 2007). When the initial therapy does not work, physicians have the option of trying an additional monot herapy or using add-on therapy, in which an additional medication is added with the intention of being complementary to the initial therapy (Sander, 2004). In particular, a number of newe r AEDs are often used in this way, including lamotrogine, levetiracetam, and oxcarbazepine (S ander, 2004). Interestingly, while 10-20% of the remainder eventually achieve adequate cont rol via pharmacotherapy, t hose who tolerate their first anti-epileptic drug but do not achieve seiz ure control are unlikely to achieve adequate 14

PAGE 15

seizure control on subsequent tria ls, in com parison to those for wh om failure occurs in the first trial due to side effects (French, 2007). Treatment Resistance There are many possible causes of failure to achieve seizure control with AEDs. One possible cause is that certain epilepsy pathophysiologies are less effectively controlled by the presently existing anti-epileptic drugs. Certain neuropathological ch aracteristics, such as mesial temporal sclerosis, in the case of temporal lobe epilepsy, appear to increase risk of treatment resistance (French, 2007). Another possible cause is that techniques of matching patients to medications are insufficient to allow for ad equate testing of al l potentially efficacious combinations of available AEDs. Because of th is, treatment success in all individual patient cases may actually be possible, but is not reached for logistical reasons. A third possibility is an inability to balance response and iatrogenesis ; AEDs may be partially efficacious but achieve toxicity or excessive adverse e ffects at doses sufficient for adequate seizure control. The most common side effects reported by patients include fatigue, tremors, wei ght gain, memory and attentional problems, agitation and irritability, although certain AEDs also carry small risks for serious, acute adverse reactions (Carpay, Alde nkamp, & van Donselaar, 2005). An additional possibility is that suboptimal adhe rence to AEDs appears to reflect treatment resistance when in fact the medications being taken are capable of achieving success. Notably, pharmacoresistance does not occur only during the initial process of di agnosing epilepsy or during initial attempts to achieve seizure control. A large proportion of the epilepsy population becomes medically intractable many years after diagnosis, following years of adequate management; and a subset of these individuals actually again achieve adequate control at a later time, suggesting that some dynamic physiological or be havioral aspect of the patients presentation is at least partly 15

PAGE 16

responsib le for changing efficacy (Berg, 2006). While many processes may play a role in this, adherence to medicatio ns is certainly one. The Role of Adherence Healthcare professionals pres cribe medications and other th erapies with the expectation and belief that these therapies have the poten tial to reduce symptoms improve functioning, or otherwise support, heal, or cure their patients. Medications in pa rticular are rigorously tested before they can be prescribed, to insure that they deliver benefits to patients without doing excessive harm. All of this assumes that patients use medications in the way in which they were designed and prescribed to be used. However, th is is often not the case. Broad research has shown that as many as 50% of individuals across disease populati ons do not take their medications as prescribed; and one in five does not even get their prescription filled at all (Marinker, 1997). In addition, this research suggests that healthcare providers may not consistently screen for adherence in a valid way or even inquire about th eir patients medication taking practices. Maladherence has been shown to impact a numbe r of healthcare outcomes. In the context of epilepsy, Manjunath et al. ( 2009) demonstrated that indivi duals who were identified as maladherent have a higher near-term likeli hood of having seizures. Faught et al. (2009) demonstrated that individuals who showed eviden ce of non-adherence to anti-epileptic drugs had a 1.4x rate of hospitalization for seizure-rela ted complications. While this study could not demonstrate that maladherence caused these excess hospitalizations the individuals who were found to be non-adherence had longer hospital stays, and also had a higher number of emergency department visits. In terms of healthcare util ization, non-adherence in the Faught et al. (2009) study was associated with more than $4000 per quarter in additional emergency and inpatient 16

PAGE 17

healthcare costs. Ettinger et al. (2008) in a sam ple of only older adults with seizure disorders, also found a total increase in healthcare utilization costs of more than $2600 per annum. Jones et al. (2006) demonstrated that adhere nce may be related to seizure control; the rates of poor seizure control in this study were significantly higher for those identified as maladherent. Hovinga et al. (2008) likewise de monstrated that individuals identified as maladherent had poorer seizure control, were more likely to have experienced a past loss of seizure control, and were more likely to have ha d seizures in the past year. Faught et al. (2008) also demonstrated that maladherence to anti-ep ileptic drugs was associated with an all-cause mortality risk increase of nearly five times as well as increased risks of motor vehicle accidents and fractures. In summary, maladherence to anti-ep ileptic drugs has been shown to have a wide range of effects ranging from increase seizure risk, increased but less efficient healthcare utilization, and increased risk of morbidity and mortality. It should also be noted that two choices of no menclature are made here. First, the extent to which a patient is faithful to a prescribed me dication regimen has been variously referred to as compliance, adherence, and concordance (Vermeire et al., 2001; Bissonnette, 2008; Lehane & McCarthy, 2009). Each of these terms de notes a different notion of the responsibilities of patient and prescribing hea lthcare provider in management of the medication regimen (Horne, 2006). Compliance is sometimes used to connote a relatively authoritarian relationship in which doctors prescribe medication wit hout significant input from patient s. In contrast, adherence and concordance are viewed as recogniz ing the significant role of patie nts in this process, such as indicating to their doctors that their medicati ons are unacceptable because of any one of a number of reasons including side -effect profile, complexity of administration, or cost (Vermeire et al., 2001). While these terms ar e sometimes taken to have subtle distinctions in meaning, they 17

PAGE 18

are often used interchangeably, pa rticularly in the context of em pirical res earch into medicationtaking behavior (e.g., Bissonnette, 2008). The term adherence has been presently chosen as being reflective largely of the current practice of healthcare on the side of both professionals and patients. While patients provide feedback and guidance on their preferred medication prescriptions, the prescription ultimately serves as a form of contra ct between patient and doctor, explicitly denoting a set of behaviors to which both have committed. A second choice of notation is made to use the te rm maladherent to refer to behavior that is inconsistent with, or violates this contract between practitioner and patient. This term is not chosen here to invoke blame or intentionality, but merely to encompass the entire theoretical range of imperfect adherence behaviors; whil e a person who never fill s a prescription could accurately be called non-adherent, likewise ca lling a person who misses a medication dose once every two weeks or once every month non-adherent has the potential to be confusing due to the significant fraction of the time in which they are adherent. Therefore, maladherence will be used throughout to refer to behaviors that deviate from the prescription plan created by a patient and their doctor. Measuring Adherence Studies inside and outside of the epilepsy lite rature have used a variety of techniques to assess adherence to medication regimens. The methods most commonly used can be loosely categorized into self-report questionnaires, meth ods that make use of drug level measurements (e.g. via a blood assay), and methods that track dispensation or use rates of medications. Each of these methods is generally considered to have advantages and disadva ntages, with no method providing a perfect assessment of adherence behaviors (Dunbar, 1984; Vitolins et al., 2000; Berg, 2006). In this section, a brief review of the most common techniques for assessing adherence will be provided. 18

PAGE 19

Self-report questionnaires have a long histor y in adherence assessm ent, with published studies using self-reported adherence as a measurement as early as 1984 (Mattson, 1984). They have the advantages of being low-cost and havi ng extremely short administration times, allowing for wider use (Morisky, Green, & Levine, 1986). One of the most widely used assessments in this category is a four-item questionnaire developed by Morisky and colleagues (1986). The questionnaire uses four yes / no questions to asse ss the likelihood of adherence, with more than one affirmative response being considered an indicator of likely maladherence. Although this method is very simple, it has been shown in a variet y of studies to have predictive validity in the form of effective prediction of future overall health status related to an illness as well as specific clinical variables being targeted by a medicati on, such as blood pressure control (Morisky, Green, & Levine, 1986; Morisky et al., 2008; Jerant et al., 2008). One major limitation of this method is its subjectivity no specific definitions of adherence behaviors are utilized, and little is done to differentiate or quantif y maladherence. To address this concern, most ot her self-report methods make use of some kind of process to as sist a participant or patient in identifying the number of deviations from their medication schedule over a given time period. These methods commonly involve cueing them to guide recollec tion of medication-taking over different set intervals such as the past 48 hours, past week, or past month. These methods have the advantage of quantifying the type (i.e., missed or extra doses or deviations from dosing quantity or time) and frequency of maladherence events. Although such methods may not reflect a long-term or trait characteristic of maladherence, they appear to predict near-term a dherence reasonably well (Jerant et al., 2008). Nonetheless, even these adap tations do not remove the issue of subjectivity entirely. In addition to the question of subjec tivity in the form of defining maladherence, patients may be understandably inclined to respond to perceived consequences of endorsing 19

PAGE 20

m aladherence. For instance, patients may fear the disapproval of their physician, offending or insulting their physicians by admitting lapses in adhe rence, or even a refusal to continue treating them if they admit to adherence issues. Patients ma y also be disinclined to report lapses in their adherence based on a perception that their physic ian may be less likely to trust them in the future, because the maladherence may be perceived as telling a lie. Research participants may also respond differently to self-report questionnaire s in a research study based on the way in which the researchers relationship to the cl inician is presented, in cluding how research confidentiality is explained, whether the study take s place in or near the epilepsy clinic setting, and so on. Drug level measurements became popular in adherence assessment because of their appeal as objective measures of adherence, wh ich do not rely on patient reports that may be limited by recollection, demand charact eristics of the clinical inte rview, willingness to disclose maladherence behavior, or other factors (Vitolin s et al., 2000). Indee d, this technique has demonstrated that many patients do present for clin ical visits with either insufficient or toxic levels of their epilepsy medicati ons (Shakya et al., 2008). One comple xity of this method is that drug levels are unique to each medication, with di fferent anti-epileptic drugs having different half-lives, different therapeutic windows, and many other characteristics that are different (Gomes Mda, Maia Filho Hde, & Noe, 1998; Tros t et al., 2005). Also, such measurements are not applicable as measurements for adherence in all anti-epileptic drugs (Walters et al., 2004). Further complicating the picture is that blood levels of medicati ons are not singly determined by adherence. Interactions with other medications or even othe r substances can attenuate or potentiate metabolic consumption of a medica tion. For instance, besides other medications, grapefruit juice is a well-know n example of a common consumable that affects the blood levels 20

PAGE 21

of som e epilepsy medications (Garg et al., 1998). Blood levels can also be affected by factors such as renal or hepatic compromise, and even in the absence of specific interactions with other consumed substances or renal or hepatic malf unction, can vary merely by phenotypic variability in body metabolism of medications (Goldste in et al., 2007; Johannessen & Landmark, 2008). Finally, drug level measurements are highly timedependent not only do they depend heavily on short-term adherence to recent medication dosage s, but they also vary over the course of the day based on the time between the last dosage and the blood assay, particular ly in short half-life medications (Glauser & Pippenger, 2000). A second category of objective measures of adherence includes various methods to assess the rate at which medicati ons are dispensed or used (Kar ve et al., 2008; Cooper et al., 2009). The logic of these measurements is that me dications are prescribed at the rate at which they are intended to be used (e.g., if a person is to use sixty pills of a medication that is taken twice daily over the course of a mo nth, they are prescribed sixty pills every month). Therefore, if an individual is found, perhaps, to refill this prescription only on ce every 40 days instead of once a month, or alternatively is found to have more of the given pill in their possession than they should at a given point in time, they are pr esumed to be under-using the medication (missing doses). Similarly, if they refill the prescription more frequently than indicated or have less remaining supply than expected, they are pr esumed to be over-using the medication. There are several ways this kind of assessm ent can be made. Reviews of pharmacy or insurance data are often able to reveal the re fill dates and quantities of medications, which can then be used to compute a medication possession ratio (MPR), a measure of the actual prescription rate in comparison to the expect ed rate (Cooper et al., 2009). This method has strength in that it is typically able to access a large cross-section of patients. It may create 21

PAGE 22

problem s during medication transitions for instance, in handling tapering of doses or mid-refill changes in prescription that may cause a patient to have more or less medication than expected (for instance, patients are sometimes allowed to use multiple pills from a smaller dose to complete already prescribed medications before filling a prescription for a new, higher dose). Another method of monitoring pill consumption is to conduct a pill audit by having patients present for a research study or clinical visit w ith all of the medications they have in their possession, including the bottles in which they were originally dispensed. Based on the time that has passed since the prescription was filled, the expected number of remaining pills can be computed and compared to the actual number of pills the patient possesses. Both this method and the MPR method potentially are vulnerable to error in situations when patients engage in both overand under-dosing or who take the correct dosage overall but do not follow the dosing schedule correctly. For instance, a patient who st ops taking their medication for a week and then doubles up to compensate for this would appe ar adherent according to these methods. One more recent development in medication use rate monitoring of adherence that addresses this limitation is the advent of elect ronic monitoring techniques (Vaur et al., 1999). In these methods, patients are dispensed medications in special containers that electronically record pill dispensation and the data from these dispensers is used to compute adherence. This method still supposes that pills, once dispensed, are actually consumed but it does have the potential to address variability in adherence that is mask ed by MPR calculations or pill audits. One major strength of electronic monitoring is that it can be used to es sentially continuously sample adherence behaviors over a longe r period of time, rather than estimating adherence based on a snapshot obtained at the time of the clinical or re search assessment. This can be very important one study determined that adherence measured in this way peaks around the time of a clinical 22

PAGE 23

assessm ent and then drops significantly in the mo nth following the assessment (Cramer, Scheyer, and Mattson, 1990). Another study established that electronic monitori ng identified adherence lapses that were missed by other techniques, in particular demons trating that adherence measured in this way had a non-significant correlation wi th blood level monitori ng (Cramer et al., 1989). Thus, while there are many established methods of assessing adherence both clinically and in research, it has become gene rally accepted that no measurement of adherence can take the place of a gold standard, either in the sense of providing a completely reliable snapshot of actual adherence behaviors or of providing a measurement that is demonstrably more accurate than most other measurements. Dunbar (1984) advocated this position, for instance, noting, Each measurement procedure offers somewhat di fferent information and has unique advantages and disadvantages. Why Do People Adhere or Not Adhere? Given that adherence is necessar y, at least at some level, for medications to be beneficial to patients, and given that adherence can be measured, it is reasonable to then ask whether a patients ongoing adherence behaviors might be pr edicted. This would allow for identification of patients who might be at higher risk for adherence problems and for whom additional interventions could be used to maximize the like lihood of their benefit fro m medications for their disease. There are a number of reasons, both inten tional and unintentional, why patients do not adhere to prescribed medication. In the fram ework of adherence propos ed here, intentional maladherence consists of situ ations in which patients choose to deviate from adherence, whereas unintentional maladherence involves situations in which patients have chosen to follow the medication regimen devised with their provider but may not do so due to other barriers. In the area of intentional maladherence, research indicates that the form ation of subjective theories is 23

PAGE 24

very im portant for patients (Wagner, 2003). These subjective theories are rational and sophisticated in the same sense th at scientific theories are, but are nave in the sense that they benefit only from personal experience and not from available scientific and clinical evidence. However, patients take them very seriously, and they can have very important consequences. For example, Remien et al. (2003) found that individuals taking highly active anti-retroviral therapy (HAART) for HIV formed s ophisticated theories about medication effects. One such theory was that an individual could train their body to resist HIV by reducing or skipping HAART doses. This has some surface valid ity without medical education individuals are often encouraged to wean themselves off pai nkillers, cigarettes, or alcohol in exactly this way. However, it may not be true in the case of managing other kinds of chronic illness such as HIV. Other aspects of intentional non-adhe rence include independe nt response to and management of side effects by the patient (R emien et al., 2003). Again, patients engage in problem solving using locally available inform ation (their own experiences and those of influential others). This problem solving often occurs in a comp lex social ecology; and research has indicated that significant variance in medicat ion adherence is accoun ted for by interactions with important others as well as internal cognitive modeling of disease (Naar-King et al., 2006; Remien et al., 2003; Johnson et al., 2006). This suggests the importance of assessing the roles of important others in making medical decisions as well as assessing disease knowledge in the context of understanding th e motivation or process which underlies maladherence. In the area of unintentional maladherence, factors cited include such concerns as motivational level and regimen complexity. Depre ssion is often examined in this context. A meta-analysis conducted by DiMatteo et al. (2000) found that individuals with chronic health problems including cancer, cardiova scular disease, and renal di sease had substantially lower 24

PAGE 25

rates of adherence to m edicati ons if they suffered from como rbid depression. Aggregating the results of 12 studies with more than 650 participants, these researchers found th at clinically significant depression, was highly pr edictive of adherence across disease groups, with an overall odds ratio of 3.03 (individuals with comorbid de pression are three times more likely to be nonadherent than their non-depressed peers). Interestingly, this study found that, while depression predicted adherence, anxiety di sorders and/or symptoms of anxiety did not. Apathy, which Marin defined simply as the lack of motivation seen in many neuropsychiatric disorders, and which can occur in the absence or pres ence of depression (Marin, Biedrz ycki, & Firinciogullaari, 1991), has also been independently examined to some extent. Rabkin et al. (2000) investigated the symptoms of apathy as a predictor of adherenc e to HIV regimens. These researchers found that apathy did not remain predictive once control fo r the highly correlated symptoms of depression was added to the model. However, as the rela tionship between apathy and depression can vary greatly based on the neuropathology of a disorder more work is necessary to understand the independent role of apathy (Litvan, Cummings, & Mega, 1996). As might be expected, regimen complexity has also been investigated as a cont ributor to adherence. The majority of studies indicate that adherenc e decreases with increases in regi men complexity, although this is not found in all instances (Yeager et al., 2005). Several researchers have noted that informati onal, emotional, motivational, and logistical aspects of adherence can in teract with each other dynami cally, changing based on other contributions to emotional functioning, new info rmation and experience, and so on (Remien et al., 2003; Fisher et al., 2006). In fact, this dynamic process underscores the notion that demographically identifying the likely maladhe rent patient in some general way is not particularly likely to meet with success. Thus, we are not yet at a stage where adherence behavior 25

PAGE 26

can be accu rately predicted based on knowledge of the informational, emotional, motivational, and logistical context in which the individual patient exists. Depression in the Context of Epilepsy Comorbidity of Major Depressive Disorder is high in many different symptomatic chronic illnesses, although rates vary from illness to illness for reasons that are not yet fully known (Evans et al., 2005). In a review of the literature on depression in epilepsy, Kanner (2003) notes that reported rates of de pression vary between 3-9% in patients whose epilepsy is considered well-controlled and between 20-55% in patients experiencing recurrent seizures, making seizure control a major predictor of depression in this population. Psychosocial explanations appear to be very important in understanding depression in the context of chronic illnesses, including epilepsy. In particular, coping style and self-efficacy (belief that one is capable of planning and executing actions that can lead to desired outcomes) appear to have a substantial relationship to the formation and/or maintenance of depression (Goldstein et al., 2005; DiIorio et al., 2006). Some factors suggest general biological disease process mechanisms for comorbid depression as well. For instance, there are differential gender distributions in primary and comorbid depression, with men at greater risk for secondary, or comorbid depression, and women at greater ri sk for primary depression, in spite of similar characteristics of depression as a disease pro cess in both cases (Harden, 2002). This suggests a potential role for biological factors in the comorbid depression disease process. While depression occurs at elevated rates in a variety of chronic illnesses, rates of depression in people with epilepsy are particul arly high, with some studies suggesting higher rates of comorbidity with epilepsy than with other chronic illnesses (Torta & Keller, 1999; Ettinger et al., 2004). Some studies have indicated that rates are pa rticularly high in intractable temporal lobe epilepsy, a common refractory epilep sy syndrome, with stud ies reporting rates as 26

PAGE 27

high as 88% (Edeh & Toone, 1987; Gaitatzis, Trim ble, & Sander, 2004). There are many possible explanations for this. It is conceivable that the differences may be related to different forms in which impacts on quality of life occur in refractory epilepsy such as the disruption of work, driving, and other basic life activities. On the other hand, since limbic system structures commonly involved in refractory ep ilepsy syndromes are also important in the pathogenesis of depression (Kennedy, Javanmard, & Vaccarino, 1997; Drevets, 2001), it might be inferred that particularly high rates of depr ession in people with refractory epilepsy may be due to a common pathway whereby an epilepsy disease process temporal lobe pathol ogy leads to seizures, cognitive problems, and depression. Supporting this are seve ral studies that identified higher rates of depression in temporal lobe epilepsy than in othe r epilepsy syndromes (Rodin, 1976; Gurege, 1991). On the other hand, at least two la rger studies examined depression rates and failed to find this pattern. Although rates were elev ated in patients with refractory epilepsy, they were not higher in patients with temporal lobe epilepsy in pa rticular (Swinkels et al., 2006; Adams et al., 2008). Comorbid depression is not onl y common but serious. It negatively affects health-related quality of life (HRQoL) not just within areas such as emotional functioning but broadly across all areas (Cramer et al., 2003; Cramer, Brandenburg, & Xu, 2005; Johnson et al., 2004). Zeber et al. (2007) showed that individua ls with epilepsy and comorb id psychiatric disorders had significantly lower health-related quality of lif e, with depression having the second greatest impact on quality of life after posttraumatic stress disorder (PTSD). Indeed, comorbid depression has been shown to be a better predictor of HRQoL th an clinical severity variables such as seizure type or frequency; this has itself been interpre ted as evidence that depression is not just a reaction to the psychosocial experience of ha ving seizures (Kanner & Balabanov, 2002). A 27

PAGE 28

vicious cycle has been proposed in epile psy, in which com orbid depression and poorly regulated stress responses might contribute to negative outcomes such as higher seizure frequencies. Thapar, Roland, & Harold (2005) dem onstrated not only that individuals with high seizure frequency are at greater risk for experiencing depression but also that individuals who experience depression are likely to experience highe r seizure frequencies in the future than nondepressed peers. However, studies that have i nvestigated the efficacy of empirically supported psychotherapeutic techniques for stress manage ment have thus far had poor methodological quality and have not been able to demonstr ate an ability to reduce seizure frequency by controlling stress responses (Ramaratnam, Baker, & Goldstein, 2005). Finally, it may also be that other symptoms of depression such as insomnia (and resulting sleep depr ivation) are indirect contributors to seizure frequency. One more aspect of the serious nature of comorbid depression is that it may be less amenable to resolu tion than primary depression. On the one hand, depression comorbid with epilepsy has been shown in a number of studies to be amenable to a variety of pharmacological interventions and psyc hotherapeutic interventi ons, although, of note, randomized, controlled trials of antidepressants specifically in the population of individuals with epilepsy and comorbid depression have not b een published, nor have efficacy studies for psychotherapeutic treatments such as cogni tive behavioral therapy in this population (Krishnamoorthy, 2003; Garcia-Morales, de la Pena Mayor, & Kanner, 2008; Mula, Schmitz, & Sander, 2008). On the other hand, seizure contro l has been found, even in large, community studies, to be among the largest determinants of point prevalen ce of depression in individuals with epilepsy, with dramatically lower rates of depression in indivi duals who have fully controlled seizures when compared to those who have incomplete seizure control (Evans et al., 2005). This suggests that depr ession is not only common but likely persistent in many 28

PAGE 29

individu als who have active seizures, even given the broad availability and accessibility of treatments. Comorbid depression negatively and seri ously impacts other aspects of epilepsy management including healthcar e utilization and hospitalization. Individuals with comorbid depression indicate greater severity of seizures and describe thei r seizures as more bothersome than people with epilepsy who do not have comorb id depression (Cramer et al., 2003). Cramer et al. (2004) indicated that indi viduals with epilepsy and co morbid depression had higher healthcare utilization costs whether or not their depres sion was being treated with antidepressants. In a revi ew of the impacts of comorbid i llnesses on management of refractory epilepsy, Lee et al. (2005) found that, while comorbidity in general increased healthcare utilization, depression particularly had the strongest effect on both the likelihood for hospitalization and increase in healthcare costs (costs were 83 % greater in refractory epilepsy patients who were also depressed). Mechanisms Behind the Impact of Depression on Adherence Depression has a strong relationship with a dherence and is frequently comorbid with chronic illness. Not only do indivi duals with chronic i llnesses broadly have a rate of depression on the order of 25-40%, but 60-70% of individuals who are depre ssed suffer from one or more chronic illness (Cassano & Fava 2002). Individuals who have comorbid depression have frequently been observed to have both poorer ad herence and poorer health outcomes, and yet few studies have explicitly shown adherence to be a mediator of this process or have addressed why depression impacts adherence (Wing, Phelan, & Tate, 2002). For this reason, understanding the mechanisms behind the large effect of de pression on adherence has begun to receive considerable attention. Research ers investigating the comorbidity of depression with asthma have shown that this combination of illnesses is associated both with poorer knowledge and 29

PAGE 30

attitudes regarding asthm a and with poorer disease-monitoring ab ility (Baiardini et al., 2006). Motivational or perseverative aspects of health behavior have also b een implicated some researchers have demonstrated that sensitivity to treatment side-eff ects, leading to higher rates of apparent treatment resistance due to adverse effects of medication, is a mechanism by which depression affects adherence (Magai et al., 2007). Attitudes underlying coping styles may vary qualitatively between individuals who do and do not experience comorbid depression (B arton et al., 2003). Sacco et al. (2005) demonstrated that, in the context of Type II Diabetes, the relationship between adherence and depression is fully mediated by an attitudinal variable perceived self-efficacy (in this model, adherence was a predictor of depression, the effect of which was co mpletely mediated when selfefficacy was taken into account). Based on existing research outside of epilepsy, it does therefore appear that depression exacerbates many of the ma jor causes of maladherence already identified. Safren et al. (2001), in compar ing interventions to improve a dherence, noted that, when HIV positive patients had comorbid depression, they c ould improve adherence when administered a cognitive-behavioral therapy (CBT) intervention that worked not only on adherence barriers but also addressed depression. In this study, th e CBT intervention was compared to a purely adherence-skills-oriented intervention that was approximately equally effective in non-depressed participants. Treating depression by other means such as the use of antidepressants has also been shown to improve adherence rates for treatmen t of comorbid illness (Dalessandro et al., 2007). On the other hand, Wang & Li (2003) examined the ro le of health education in individuals with comorbid depression and hypertension. Health edu cation was found to be sp ecifically effective in modifying hypertension health behavior in this study, improving both hypertension medication adherence and physiological outcomes without re quiring the need for successful treatment of 30

PAGE 31

depression. In fact, although individuals in this intervention w ho had co morbid depression did not show any decrease in depressive symptoms, they showed a larger magnitude of interventionrelated improvement compared to controls than did non-depressed partic ipants. This suggests that identifying specific health behavior and health education interventions for chronically ill populations may possibly be an av enue through which health stat us can be improved for these patients independently of success in treating secondary depression. Identifying and Addressing A dherence Behavior in Epilepsy A literature search was performed in order to identify existing studies that have attempted to determine predictors of adherence to anti-ep ileptic drugs. Using keywor d-driven searches of the Pubmed database as well as individual review of cited ar ticles within these and other publications1, 19 studies were identified. These studies are summarized in Table 2-1. Studies were included if they employed any specific meas ure of adherence to medications and examined one or more possible predictor vari ables quantitatively. Studies were included if they used either cross-sectional or longitudinal techniques, and case-control st udies in which adherence was a group characteristic rather than an outcome meas ure were included if st udy dependent variables contained one or more variable th at could be construed as a predic tor of adherence. Studies were not included if adherence was used primarily as a predictor or dependent variable, such as when adherence was used to predict seizure control or h ealthcare utilization costs. Studies that reported relationships between predictors of adherence and adherence but di d not appear to make use of statistical tests of signifi cance were not included. Twelve of these studies focused on adults (w ith one also includi ng adolescents), while 1 Keywords used in this search for adherence were: adherence, compliance, maladherence, non-adherence, and noncompliance; epilepsy and depression were also added as se arch terms. Each article was individually screened to ensure that it met the criteria described above. 31

PAGE 32

seven studies focused exclusivel y on children. Twelve studies used a variety of self-reported adherence measures, adm inistered primarily in questionnaire format, and one more study used a clinician rating system. Four used an assay method such as a phenobarbitol challenge or a serum level check (including one of the previously me ntioned studies which us ed both self-report and assay methods). Three studies used consumption monitoring two of th ese studies used the medication prescription ratio and the third used electronic monitoring of pill dispensation. The studies found somewhat conflicting results with resp ect to epilepsy clinical variables but fairly consistent results with respect to psychosocial variables. Variables that fell into the category of disease attitudes were the most frequently significant predictors. These variables included any variable that measured beliefs, attitudes, or other affective labeling of medical treatment, symptoms, diagnosis, or prognosis in the context of epilepsy. Variables in this category were found to be significant in eight of the studies and only found to be non-significant in one study. On the other hand, clinical variables such as duration of epilepsy and treatment complexity were strong but inconsistent predictors in so me studies and non-significant in others. For instance, two studies examined dur ation of epilepsy or epilepsy treatment and found this to be significant in predicting adhere nce, while three more found this to be non-significant. Demographics were supported as predictors only occasionally, with conflicting results in different studies. For instance, Faught et al (2009) found higher rates of adherence based on medication prescription ratio in Caucasian pa rticipants and McAule y et al. (2008) found lower rates of adherence based on the Morisky measure in Caucasian participants. Finally, of note, no study examined the presence of depression as a predictor of adherence, although one study (McAuley et al., 2008) did examine a past history of depression treatment as a predictor of adherence. These studies ar e summarized in Table 2-1. 32

PAGE 33

Rationale for the Present Study W hile it seems generally true that patients w ith epilepsy have adherence characteristics similar to those seen in other chronic illnesses, th e literature in this area has not addressed some major concerns. First, the literature does not pr ovide an adequate answer as to why some people with epilepsy do not maintain adequate adherenc e. This is a barrier to the development of specific, streamlined interventions that target actual factors relate d to adherence issues in this population. Second, depression, a condition that is frequently comorbid with symptomatic epilepsy, has rarely been specifica lly considered as a contributor to adherence. In fact, only one study was identified that assesse d any relationship between adhe rence and depression in people with epilepsy, and that study (McAuley et al ., 2008) only examined the relationship between adherence a reported history of being treated for depression. As depression is likely to exacerbate many processes implicated in poor adhe rence, understanding its specific contribution to adherence in epilepsy is important for two r easons. First, such an understanding may assist in identifying a subset of epilepsy patients who are at increased risk for recurrent seizures. As well, it may help in targeting a subset of patients in whom direct treatment of depression would have secondary benefits on adherence a nd subsequent seizure control. Statement of the Problem The present study seeks determine, in an epilepsy sample, how disease knowledge, attitudes, and motivation contri bute to adherence to anti-epile ptic drugs, and to determine whether comorbid depression plays a specific role in adherence and seizure control. Specifically, the aims of this study are: 1. To determine whether comorbid depression reduces the likelihood that people with epilepsy will maintain adequate adherence to epilepsy medications 2. To determine whether epilepsy knowledge and at titudes towards medical care affect rates of adherence to antiepileptic medications 33

PAGE 34

34 3. To ascertain whether knowledge and attitudes have differential contributions to adherence rates in depressed individuals 4. To determine whether depression may negativ ely affect seizure control (e.g., through the mechanism of poorer adherence)

PAGE 35

Table 2-1. Studies of adherence in epilepsy Publication Method Sample Measure of Adherence Supported Predictors of Adherence* Predictors of Adherence Not Supported** Asadi-Pooya et al. (2005) Cross sectional design, interview Children with epilepsy (n = 181) Self-report via interview Younger mother, family history of epilepsy Therapy complexity, parental education Buck et al. (1997) Cross sectional design, postal questionnaires Adolescents and adults with epilepsy (n = 696) Self-report (frequency of missed doses) Disease attitudes, treatment complexity, side effects, practitioner quality, age Duration of epilepsy, seizure severity Briesacher et al. (2008) Longitudinal design, retrospective claims data analysis Adults with epilepsy (n = 4984) and other chronic illnesses Medication prescription ratio Higher comorbidity, polytherapy, previous familiarity with epilepsy medications Demographics Cramer et al. (2002) Cross sectional design, postal questionnaires Adults with epilepsy (n = 661) Self-report (frequency of missed doses) Duration of epilepsy treatment, some treatment complexity variables (trend level) Total number of capsules taken for either epilepsy medication or all medication DiIorio et al. (2003) Mixed longitudinal / cross sectional design, questionnaires Adults with epilepsy (n = 314) Self-report via questionnaire (Morisky) Disease attitudes (stigma) Enriquez-Caceres et al. (2006) Cross sectional design, retrospective chart review Adults with epilepsy (n = 114) Clinician-report adherence 35 Disease knowledge and attitudes, access to treatment, clinician relationship Side-effects Faught et al. (2009) Longitudinal design, retrospective claims data analysis Adults with epilepsy (n = 33,658) Medication prescription ratio Younger age, male sex, Caucasian race, higher comorbidity Gomes et al. (1998) Cross sectional design, questionnaire Adults with epilepsy (n = 45) Self-report via questionnaire (missed doses in last week) Disease knowledge and attitudes Demographics, SES, treatment complexity Hazzard et al. (1990) Cross sectional design Children with epilepsy (n = 35) Serum level Disease attitudes, psychosocial risk factors Hovinga et al. (2008) Cross sectional design, online study Adults with epilepsy (n = 408) Self-report (frequency of missed doses) Several measures of illness / disability severity, past history of loss of seizure control Jones et al. (2006) Cross sectional design, questionnaire Adults with epilepsy (n = 54) Self report via questionnaire (Morisky) Disease knowledge and attitudes

PAGE 36

36Publication Method Sample Measure of Adherence Supported Predictors of Adherence* Predictors of Adherence Not Supported** Kemp et al. (2007) Cross sectional design, questionnaire Adults with epilepsy (n = 37) Phenobarbitol challenge Time since last seizure Age at onset and duration of epilepsy diagnosis, disease attitudes Kyngs et al. (2001) Cross sectional design, postal questionnaire Adolescents with epilepsy (n = 232) Self-report via questionnaire Psychosocial risk factors, disease attitudes Lugo Gonzalez et al. (2001) Cross sectional design, questionnaire Children and adolescents with epilepsy (n = 54) Self-report via questionnaire Subjective memory complaints, time between clinical follow-up visits Lusic et al. (2005) Cross sectional design, questionnaire Adults with epilepsy (n = 146) Self report via questionnaire Substance abuse, duration of treatment, treatment complexity McAuley et al. (2008) Cross sectional design, questionnaire Adults with epilepsy (n = 50) Self report (Morisky) Non-Cau casian race Other demographics, history of seizure freedom, treatment for depression Mitchell et al. (2000) Longitudinal design, questionnaire Children with epilepsy (n = 119) Adherence to follow-up schedule, self-reported via questionnaire, serum level Psychosocial risk factors Seizure severity, disease knowledge Modi et al. (2008) Longitudinal design, prospective Children with epilepsy (n = 35) Electronic monitoring system Married parents, higher socioeconomic status Other demographics, type of epilepsy, specific medications, seizure frequency, duration Snodgrass et al. (2001) Cross sectional casecontrol design Children with epilepsy (n = 200) selected for serumlevel-determined adherence or nonadherence Serum level Race, insurance status Treatment complexity Notes: (*) A predictor was considered supported if it was a st atistically significant predictor of adherence based on the stu dys chosen criteria or if it was statistically significantly predicted by epilepsy in the model of choice but represented a pre-st anding risk factor theoretical ly. (**) A predicted was considered not supported if it was entered into a model in the study an d was statistically non-significant or failed to enter / was drop ped from a model in a stepwise process. Again, the original authors crite rion for significance was used in all cases.

PAGE 37

CHAP TER 3 METHODS Population Although epilepsy represents a complex constellation of syndromes with different etiologies, prognoses, and neurocognitive impacts, individuals with different seizure disorder variations generally receive similar care and neurologists who work with epilepsy patients similarly provide their array of services to patients with a br oad range of seizure disorder characteristics. Therefore, the present study s ought to recruit a samp le that is generally representative of the broad a dult clinical epilepsy population. A focus on adult patients only was selected due to the relatively substantial differences in clinical care, family involvement, and established research instruments between pediatric adult patient populations. As such, epilepsy patients age 20 years or older we re recruited for the current st udy. Two key inclusion criteria were (a) stable diagnosis of epilepsy for at least tw o years, and (b) at least one seizure in the past 6 months. These inclusion criteria were imposed in order to derive a samp le of individuals with epilepsy for whom medication adherence may be a relevant issue. Recruitment Procedure Two recruitment sites were used in order to produce a demographically representative sample comprised of individuals living in an urban / suburban environment and individuals living in a rural environment. Individuals were recruited during their visits to outpatient neurology clinics at Shands at the University of Florida (Gainesville, FL) and Shands Jacksonville (Jacksonville, FL), two tertiary-care, academic medical centers in Florida. Patients who arrived for appointments at epilepsy clinics at these two sites were screened to identify individuals who met the recruitment criteria. Eligib le individuals were then approached and were given information about the study and obtain inform ed consent. Participants received a nominal 37

PAGE 38

financial compensation ($20) for their participation. Institutiona l Review Board approval was obtained for all procedures. Assessment Procedure Participan ts were asked to complete self-re port questionnaires after consent was obtained and prior to being seen by their physician. Th ey were then given a brief neurocognitive and diagnostic assessment by a Masters level graduate student in clinical psychology and completed additional self-report questionnaires after their appointment with their neurologist. Participants were also asked to participate in a follow-up session, consisting of a brief telephone interview in which knowledge and adherence was re-assessed along with seizure frequency over the month following the initial assessment. Assessment Instruments Epilepsy Knowledge Instruments Two instruments were used to assess patient-centered knowledge of epilepsy as a disorder, appropriate management of epilepsy, an d the relationship between epilepsy and daily activities such as driving and employment. The Epilepsy Patient Knowledge Questionnaire (EPKQ ) is a brief, 13-item instrument consisting of a mixture of true/false, multiple choice, and free response items addressing each of these areas (Long et al., 2000). This instrument can be scored and interpreted on the basis of the to tal percent of items an swered correctly. In a validation study (Long et al., 2000 ) with 175 adult patie nts with epilepsy, the authors found moderate mean response accuracy (58%). The items on this instrument range in difficulty from items answered correctly by only 14% of the vali dation sample to items answered correctly by 92% of the validation sample. Resulting scores are relatively free of effects of age, education, or length of diagnosis but unfortunately has not been formally assessed for validity or reliability to our knowledge. The Epilepsy Knowledge Scale (EKS) is an additional brief instrument 38

PAGE 39

consisting of either 10 or 19 yes/no item s (see be low) assessing the major domains of epilepsy knowledge discussed above (Ried et al., 2001). It was originally developed as a component of an intervention, Modular Service Package Epile psy (MOSES), designed to improve health education for patients with epilepsy in Germany. Th is instrument is designed so that 10 items can be given at a baseline assessment, with an additi onal nine items delivered at follow-up to control for practice effects. The dependent measure is th e percentage of items answered correctly. As this is a brief instrument assessing multiple dom ains of knowledge, the re liability of each subsegment is moderate, although it improved when a ll 19 items were considered (Cronbach alpha for the initial 10 items, CR = 0.45, for the additional nine items, CR = 0.0.57, and for the entire 19 item scale, CR = 0.72). This instrument was also show n to be sensitive to increases in epilepsy knowledge as delivered by the MOSE S intervention (May et al., 2002). The 19 item version of this scale was used in this study. Epilepsy Attitudes Instruments Three different, but overlapping, aspects of attitudes towards epilepsy were measured with three different instrument s. The first instrument, the Illness Perception Questionnaire (Brief IPQ ), assesses a patients attit udes about having the diagnosis of epilepsy and the impact of epilepsy on their life. This instrument consists of nine items rated by the participant on a 0-10 Likert scale with anchors rele vant to each individual item (e.g. How much does your illness affect your life? where 0 = no affect at all and 10 = sever ely affects my life). This instrument has been broadly used in many illn ess groups and was found to have good test-retest reliability (item reliability between r = 0.6-0.7 for most items) (Broadbent et al., 2006). It also has good concurrent validity (both with respect to the original long-form Illness Perception Questionnaire Revised and to some other meas ures of disease-specif ic self-efficacy, with 39

PAGE 40

which it had correlations ranging from r = 0.260.61) (2006). The Brief IPQ can also reliably discrim inate between individuals who experience chronic illness and t hose who experience truly episodic illnesses (Broad bent et al., 2006). The second instrument, the Beliefs About Medicine Questionnaire (BMQ) assesses an individuals attitudes towards the use of medical t echniques, including pharmacotherapy, to manage or improve their health functioning. It c onsists of ten items asse ssing attitudes about the medications used to treat their illness (e.g., Hav ing to take medicine for epilepsy worries me) and beliefs about medications in general (e.g., Doctors use too many me dicines). BMQ items appear to cluster on four factors, consisting of beliefs about the need for medication to treat the patients illness, concerns over dependency and other adverse impacts of medication, general belief that medications are harmful or danger ous, and general beliefs that they are overprescribed (Horne & Weinman, 1999). These four dimensions have good internal reliability (CR = 0.51-0.86) when tested in several different populations, justifying the use of subscale total scores computed by summation of items associated with each dimension, and the instrument also shows good measurement stability in the form of two-week test-re test reliability ( CR = 0.600.78). The final instrument, the Multidimensional Health Locus of Control (MHLC), assesses a patients beliefs about the extent to which their health functioning is under their own control, is controlled by healthcare providers, or is not amen able to control by either themselves or their care providers (Wallston, Wallston, & DeVellis, 1978). These three dimensions of locus of control are commonly referred to as Internal, Powerful Others, and Chance loci of control, respectively, and these abbreviati ons will be used throughout. It consists of 18 Likert scale items (values of 1-6 corresponding to strongly disagree, moderately disagr ee, slightly disagree, slightly 40

PAGE 41

agree, etc.), with six item s falling in each of th ese three areas. Scores are summed within each area to determine the strength of each aspect of lo cus of control. For example, an item indicative of Internal locus is, I am directly responsib le for my condition getting better or worse. An example of an item indicative of Chance locus is, If my condition worsens, its a matter of fate. Scores on these dimensions have been specificall y related to adjustment and health functioning in people with epilepsy (Gramstad, Iversen, & Englesen, 2001; Spector, Cull, & Goldstein, 2001). Apathy Instrument The Apathy Evaluation Scale (AES) was used to measure symptoms of clinical apathy as a proxy for general patient motivational level (Mar in, Biedrzycki, & Firinc iogullaari, 1991). This instrument is also sometimes referred to as the Marin Apathy Inventory. Apathy has been shown to be at least somewhat related to de pression in clinical st udies of neurological populations, with some populations presenting relatively low correlations and apathy and depression as two relatively dis tinguishable syndromes, and othe rs showing moderately strong correlations between the two. It is generally thought that apathy presents to varying degrees a component of depression but can also occur inde pendently of depression in certain neurological disorders (Hama et al., 2006). At least one study ha s previously used this measure to assess the concept of general motivational level (Resnick et al., 1998). It has been shown to have good internal consistency and test-retest reliability and good validity when comp ared to a structured interview assessing Marins res earch criteria for syndromal apathy (Marin, Biedrzycki, & Firinciogullaari, 1991). Adherence Instruments Adherence to AEDs was determined using two alternative self-report methods that have been validated in various patient populati ons. The measure developed by Morisky, Green, & 41

PAGE 42

Levine consists of four questions (e.g., S om etimes if you feel worse when you take the medicine, do you stop taking it?), to which a participant responds with either a yes or a no. (1986). More than one yes response is consider ed a sensitive indicator of poor to marginal adherence. This measure has prev iously been used successfully w ith epilepsy patients (Jones et al., 2006). Gao and Nau (2000) examined this scale in comparison to self-reports of the number of missed doses (a direct retrospective adherence measurement, or RAM) over the past two days and the past two weeks; they found that th e latter measures produced consistently higher apparent adherence than the Morisky scale. As it is not yet clear how th is difference arises, and few studies have compared these two commonly used methods, the present study will use the methods of both Morisky et al. and Gao & Nau. These measures of adherence will be referred to simply as the Morisky measurement and the RAM. Clinical Variables As part of the questionnaire pa cket administered to partic ipants, a number of clinical history variables were elucidated. Participants were asked to provide the age at which they first experienced seizures (age at onset ), the duration of time that they have had epilepsy, if they have ever experienced a period of seizure remission of six months or longer, and their seizure frequency in the last 30 days. Medication regime n complexity was also estimated by determining the daily frequency of epilepsy medication dosages (i.e., whether a partic ipant took AEDs at one, two, three, four, etc., different times per day). Several basic healthcare behaviors were also assessed, including whether the participants use a pillbox to organize thei r medications, whether they have family members or other loved ones ta ke part in their health care by participating in clinic visits, and whether they receive assistance at home with their medication-taking. Finally, a list of AEDs was also collected. 42

PAGE 43

Depression Instruments Participan ts were administered the 20-item version of the Center for Epidemiological Studies Scale for Depression (CES-D) prior to their clinical appointment and the Mini International Neuropsychiatric Interview after their clinical appoint ment. Both measures have been validated for use with people with epileps y (Jones et al., 2005). The CES-D scale allows for investigation of depression as a multidimensiona l construct with quantitative measurement of severity (Radloff, 1977). The Mini Internationa l Neuropsychiatric Interv iew is a structured clinical interview that identifie s 17 major Axis I psychiatric dia gnoses; it has comparable validity to the DSM-IV Structured Clinical Inte rview Diagnostic (Sheeh an et al., 1998). Cognitive Screening Cognitive screening consisted of a brief battery of neuropsychological tests. The Shipley Institute of Living Scale (SILS) wa s administered to gauge overall intellectual functioning; this brief measure consists of a Vocabulary subtest in which a participant must chose one of four words that most closely matches the meaning of a target word, a nd an Abstraction subtest, which involves tests of inductive reasoning (Shipley, 1940). This test is able to provide an estimation of the likely Full Scale Intelligence Quotient th at would be obtained using the Wechsler Adult Intelligence Scale (WAIS-R). The Stroop Test was used to measure both cognitive processing speed and freedom from interf erence, an aspect of executive functioning (Golden, 1978). The Digit Symbol subtest of the WAIS-III was also used to measure cognitive processing speed (Wechsler, 1997). Finally, the Rey Auditory Verbal Learning Test (RAVLT), a test of word list learning, was used to assess memory (Rey, 1964). Table 3-1 briefly summarizes the instruments listed above. 43

PAGE 44

Po wer Analysis and Statistical Methods Chi-squared tests were selected for use to determine if depression is associated with adherence. For this analysis, sample size was es timated using the method of Fleiss, Tytan, & Ury (1980). In order to do this, the Morisky et al. (1986) questionnaire was used. The measured adherence rate in epilepsy ( 41% indicating good adherence and 59% indicating poor adherence, using the Morisky questionnaire) in Jones et al. (2006) was taken to be the population base rate of adherence in epilepsy. The widely reported approximation of the depression base rate in epilepsy as 30% was used as an estimate of the prevalence of depression (Jones et al., 2005). Finally, the meta-analysis by DiMatteo et al. i ndicated an odds ratio of 3.0 for the increased likelihood of poor adherence in individuals with depression comorbid with a chronic illness (2000). Using these figures, the expected rate of good adherence in non-depr essed individuals is 51% and in depressed individuals is 17%. Given these figures, the sample size required for a power of 0.80 is 72 (50 non-depres sed, 22 depressed) and the sample size required for a power of 0.90 is 93 (65 non-depressed, 28 depressed). A proj ected sample size of 100 was therefore likely to recruit both sufficient depr essed and non-depressed participan ts to be adequately powered without the need for over-samp ling either of the two groups. In order to determine best predictors in the domains of epilepsy knowledge and attitudes, initial analysis were planned to be completed by examining distributional properties to select the most psychometrically sound instruments and then using bivariate correlations between these potential predictors and adherence. Distribu tional properties could then be assessed by investigating skewness and kurto sis. Skewness was considered acceptable if the absolute magnitude of skewness was less than the standard error of skewness (SES), mildly problematic if it was between 1-2x the SES, and problematic if it was >2x the SES. The magnitude of kurtosis was compared to the standard er ror of kurtosis (SEK) according to the same criteria, again with 44

PAGE 45

acceptable k urtosis being smaller than the SEK, mild kurtosis problems being 1-2x the SEK, and problematic kurtosis >2x the SEK. For cognitive va riables, the decision was also made to make use of raw scores rather than demographically corrected scores. This decision was based on the rationale that adherence is a specific behavior that does or does not occur, and is not itself adjusted in interpretation based on the demographi cs of patients. Raw scores, as they reflect actual ability in various cognitive areas, are therefore preferable on the presumption that adherence may require a certain level of cognitiv e ability, rather than a certain degree of superiority or inferiority compared to other individuals with th e same age, sex, race, or education of the participant. Logistic regression was chosen to determine the impact of factors such as knowledge and attitudes on adherence rates, as well as the diffe rential importance of these factors with comorbid depression. Linear regression was chosen to predict the impact of adherence on subsequent seizure frequency. A regression model was chosen to determine if comorbid depression reduces seizure control through the mechanism of poorer adherence to epilepsy medications or through some other mechanism. These analyses were esti mated to have between three and six predictors; with an estimate of 15 particip ants per predictor va riable, the study would also be adequately sized to allow for evaluation of these models at the desired sample size. Participants Recruitment for the study began in the Summ er of 2007 and ended in the Summer of 2008. Although a significant number of pr e-screened clinic patients met criteria for the study, many declined or were unable (e.g. because of trans portation issues) to participate, and the study recruited participants in smaller numbers than expected. A total of 56 participants were successfully recruited an d participated in the study, although some of these individuals were unable to complete significant por tions of the study su ccessfully (n = 7), generally because they 45

PAGE 46

46 were found to have excessive difficulty answering questionnaire items. This lead to a smaller sample of 49, who completed the assessmen t without substantial missing components. Modifications to Methods Due to the smaller than anticipated enrollment in the study, analyses were simplified to be appropriate for the obtained sample size. In a ddition to addressing multicollinearity, the numbers of predictors used in multiple variable models such as the regressions were reduced by considering only those variables th at had significant uni variate associations. As the number of individuals who were successfully contacted for follow-up was al so very small, longitudinal analyses were deferred. Finall y, rather than determining the impact of depression on seizure control via linear regression, sepa rate chi-square analyses were completed for the depressed and non-depressed groups.

PAGE 47

Table 3-1. S ummary of Study Measures Measure Domain Format Epilepsy Patient Knowledge Questionnaire (EPKQ) Disease Knowledge Questionnaire Epilepsy Knowledge Scale (EKS) Disease Knowledge Questionnaire Illness Perception Questionnaire (Brief IPQ) Disease Attitudes Questionnaire Beliefs about Medicine Questionnaire (BMQ) Disease Attitudes Questionnaire Multidimensional Health Locus of Control (MHLC) Disease Attitudes Questionnaire Apathy Evaluation Scale (AES) Emotional Functioning Questionnaire Morisky Questionnaire Adherence Questionnaire Reported Adherence Measure (R AM) Adherence Questionnaire Questionnaire Items to Assess Clinical History Clinical History Questionnaire Center for Epidemiological Studies Depression Scale (CES-D) Emotional Functioning Questionnaire Mini International Neuropsychiatric Interview (MINI) Emotional Functioning Interview Shipley Institute of Living Scale (SILS) Cognitive Functioning Individual Testing Stroop Test Cognitive Functioning Individual Testing Digit Symbol subtest (WAIS-III) Cognitive Functioning Individual Testing Rey Auditory Verbal Learning Test (RAVLT) Cognitive Functioning Individual Testing 47

PAGE 48

CHAP TER 4 RESULTS Characterization of Sample Demographics Of the enrolled participants, 79% were female and the average age of study participants was 43 years (SD = 12.5 years), with a range of 20-78 years. The most commonly represented racial group was Caucasian (63%), followed by African Americans (29%). Two participants identified themselves as Hispanic (4%) and one each as Native American and multi-racial. The most commonly reported level of education was completion of high school (29%). 28% of the sample did not complete 12 years of educati on, however, and almost 44% had some college education or more. Almost half of the participants (49%) identified themselves as currently married or in another kind of committed relatio nship, while the remainder were approximately evenly split between those who were divorced or separated (28%) and those never married (24%). Only 10% reported living alone; 8% report ed only children living with them, while the remainder reported having other adults in the home. Clinical Characteristics With respect to seizure disorder onset, slightly more than half of the participants had onset after the age of 21 years (51%), with anothe r 33% having onset between 7-21 years and the remainder having onset earlier than 7 years of age. The majority of the sample had duration of illness of longer than five years (71%). Approxima tely half the population (53%) reported having a period of at least six months of seizure freedom at some point since their first seizure, while the remainder had never been seizure free for such a period of time. 28% had been seizure free in the past 30 days, 50% reported one or less seizures in the past month, and the remainder reported a larger number of seizures, with 7 individuals repor ting 10 or more seizures in the past thirty days 48

PAGE 49

(14%). In term s of medication regimen complexity, participants on average took seizure medications twice daily (M = 2.0, SD = 0.6, Range = 1-4 times per day). Most participants received polytherapy, with 49% of participants for whom medicat ion information was available receiving two medications and 22% receiving three or more. The most commonly prescribed AEDs were levetiracetam (Keppra; 19 particip ants), lamotrigine (Lamictal; 13), phenytoin (Dilantin; 12), and topiramate (Topamax; 11). In terms of basic healthcare behaviors, 31% reported that they attend doctors visits unaccompanied. 43% indicated that they are accompanied by someone to their appointments who communicates with their doctors for purposes of their care while the remainder indicated that they are accompanied by someone who does not take part in their clinic visits (e.g. for assistance with transportation). 45% of responde nts indicated that they use a pillbox with separate spaces for different days and/or times of day for their medication, while 51% keep their medication in the containers received from th e pharmacy and the remainder indicate keeping their medications in some other way. In this samp le, most participants (77%) indicated that they manage their own medications, while the remainder indicated that they receive help in of some kind. Demographic and clinical ch aracteristics of part icipants are summarized in Table 4-1. Psychiatric Characteristics Psychiatric characteristics of participants were determined using the Mini International Neuropsychiatric Interview (MINI). Table 4-2 below summarizes the rates of various psychiatric co-morbidities observed in the sample populatio n. Based on responses to the MINI, the most common psychiatric co-morbidity was a curren t major depressive ep isode (MDE; 19 of 48 participants who were able to complete the MINI, or 40%). Four of the participants with a current major depressive episode also met criter ia for a past manic episode, and one more for a past hypomanic episode, however, indicating that these individuals meet criteria for Bipolar 49

PAGE 50

Disorder rather than M ajor Depressive Disorder. Fourteen participants met the additional criteria for recurrent major depressive episodes (29%), and thr ee met criteria for psychotic features of depression. An additional four individuals who did not meet criteria for a major depressive episode met criteria for current dysthymia. Th e second most common current co-morbidity was generalized anxiety disorder (21%). In comparison to the structured clinical inte rview, self-reported rate s of depression on the CES-D were somewhat higher. The range of respondent scores on the CES-D was from 0 to 49 out of a possible 60 points, with a mean score of 21.7 (SD = 11.7). In terms of distributional properties, the skewness and kurtosis of the CES-D were both acceptable. If the traditional clinical cut-off of 16 points is used, 70% of participants endors ed depressive symptoms at the level of Major Depressive Disorder on the CE S-D. A receiver operating characteristics (ROC) analysis of the CES-D scale, using the MINI dia gnosis of current major depressive episode as the referent standard, indicated onl y fair detectabilit y (AUC = 0.736), with no cut-off achieving simultaneously strong sensitivity and specificit y. The non-parametric association between the CES-D and the MINI MDE criteria was likewise modest (Spearmans = 0.40, p = 0.005). Cognitive Characteristics Cognitive characteristics are summarized in Table 4-3. The mean raw score on the Shipley Institute of Living Scale Vocabulary subtest, w ith blank items imputed at chance level, was 23.2 (SD = 5.9, Range = 11-36). The mean Abstract ion subtest raw score was 15.2 (SD = 8.4, Range = 2-36), and the mean Total Raw Score was 38.4 (SD = 11.5, Range = 18-60). All three scores had acceptable skewness and kurtosis. These scores were also age corrected to provide a basic estimation of participants general intellect ual functioning. When age-corrected, the mean Vocabulary T-score was 37.7 (SD = 10.2, Range = 15-60). In particular it was notable that only two participants obtained a Vocabulary T-score above 50 (that is, above the 50th %ile compared 50

PAGE 51

to age peers). The m ean Abstraction T-scor e was 44.4 (SD = 7.6, Range = 31-62). Interestingly, participants Abstraction T-scores were signif icantly higher than th eir Vocabulary T-scores (t(46) = -4.50, p < 0.001). Finally, the Shipley Institute of Living Total Raw Score can also be used to estimate the likely WAIS-R Full Scale IQ (FSIQ); when this was computed, the average estimated FSIQ was 96.6 (SD = 9.3, Range = 76-112) suggesting overall average intellectual functioning. All individuals in th e study are likely to fall into the Borderline Intellectual Functioning classificat ion or higher. On the Stroop Test, participants comp leted on average 76.1 responses on the Word condition (SD = 17.3) and 59.4 on the Color condition (SD = 15.0). Their Color-Word Interference was on average -2.1 (SD = 9.2, Range = -17.6 to +20.6). When these scores were age-corrected, the mean Interference T score (whe re T scores have a mean of 50 and standard deviation of 10) was 49.1 (SD = 8.7). Using a 5th percentile cutoff for impairment (i.e. a T score of 34 or less), two participants were impaired on this measure. On the Digit Symbol subtest from the WA IS-III, their mean raw score was 53.5 (SD = 21.2, Range = 14-89). When Digit Symbol raw scores were age-correcte d, participants had a mean scaled score (where scaled scores have a mean of 10 and standard deviation of 3) of 7.1 (SD = 3.0). Using a 5th percentile cutoff for impairment (i.e ., a scaled score of 5 or lower), 18 participants were impaired on this measure. On the RAVLT, average Total Learning was 36.7 (SD = 10.2) with a range of 14-58. Delayed Recall (Trial VII) was 6.3 (SD = 3.4) with a range of 0-12. When RAVLT Delayed Recall was corrected for age and sex (Savage & Gouvier, 1992), the mean T score was 39.8 (SD = 13.3). Using a 5th %ile cutoff, 16 participants were impaired on this measure. 51

PAGE 52

Overall, the distr ibutional properties of cogn itive variables were acceptable except that the Digit Symbol raw score showed mild problem s with kurtosis (kurtosis = -1.2, 1.8x SEK). Healthcare Attitudes and Beliefs Illness Perception On the Illness Perceptions Questionnaire, part icipants endorsed relatively high levels of perceived illness related to th eir epilepsy, with scor es ranging from 10-64, out of a possible theoretical range of 0-80 (Mean = 46.1, SD = 10.4). The highest average levels of severity were reported for items related to how long the condition is likely to persist and the level of concern over the condition (M = 8.3 and 8.4, respectively). In contrast, patients reported feeling fairly strongly that treatment can help them with their illness (M = 2.2, SD = 2.0) and that they understand their illness fairly well (M = 2.4, SD = 2.9). This instrument had significant negative skew (skewness = -1.1, 3.3x SES) and was hi ghly leptokurtic (kurtosis = 2.7, 3.9x SEK). Medications On the Beliefs About Medicines Questionnaire, the average reported Specific Necessity was 2.1 (SD = 0.8) and the average reported Specific Concerns was 3.0 (SD = 0.8). The average reported General Harm was 3.3 (SD = 0.8) a nd General Overuse was 3.8 (SD = 0.7). The Specific Necessity scale had significant positiv e skew (skewness = 1.0, 3.0x SES) as well as some issues with kurtosis (kurtosis = 0.9, 1.4x SE K). The General scales both had mild problems with skew, with the General Harm scale having a slight, positive skew and the General Overuse scale having a slight, negative skew (Both <2x SES). Locus of Control On the Multidimensional Health Locus of Cont rol instrument, the mean Internal scale was 25.4 (SD = 4.2), the mean Chance scale was 18.6 (SD = 6.1), and the mean Powerful Others 52

PAGE 53

scale was 23.5 (SD = 5.2). The Powerful Othe rs scale had m ild negative skew but the distributional properties of the three subscales were otherwise acceptable. Apathy On the AES, respondent scores ranged from 18 to 57 out of a theoretically possible range of 18-72. The mean response was 32.8 (SD = 10.0), and both the skewness and kurtosis of this measure were acceptable. Epilepsy Knowledge On the 19-item Epilepsy Knowledge Scale, sc ores ranged from 6-16 correct, with an average of 11.1 correct responses (SD = 2.4). The most frequently correct responses were an item asking whether blood samples could determine the level of anti-epileptic drugs in the body and an item asking whether doctors can achieve seizure control through medication in most cases (both 94% correct). The least fr equently correct question was one asking if a person with epilepsy who drives is legally re quired to inform authorities a bout their condition (8% correct). This measure showed acceptable distributional properties. On the 13-item Epilepsy Patient Knowledge Questionnaire (EPKQ), participant scores ranged from 4-12 correct, with an average of 8.2 correct answers (SD = 2.3). The most frequently correct answers were the item related to appropriate actions to take when one stops having seizures while taking seiz ure medications (88% correct), situations under which a car can be driven by a person who still has seizures (84%), and abilit y to name seizure medications (84%). The least commonly correct response wa s for an item probing for the ability of people with epilepsy to do various activities such as swimming under supervision, exercising, and consuming a limited amount of alcohol with dinner (only 18% answ ered correctly). The skewness of the EPKQ was acceptable; it was slig htly platykurtotic (kurtosis = -1.0, or 1.5x the SEK). 53

PAGE 54

Adherence Characteris tics Morisky On the 4-item Morisky measure of adherence, 33% endorsed none of the four items, and 43% endorsed one item, for a total of 76% of pa rticipants indicating good adherence. Of the remainder, most (20% of sample) endorsed 2/4 it ems, while only one participant each endorsed 3/4 or 4/4 items. The most frequently endorsed item was the first item (Do you ever forget to take your medicine?), which was endorsed by 49% of participants. RAM Using the Gao & Nau (2000) measure of a dherence, 33% of pa rticipants endorsed maladherence on the two day retrospective and 53% on the 14-day retrospective. The association between 2-day and 14-day RAM was strong (Spearmans = 0.66, p < 0.001). If any reported maladherence during either time period is taken as a sign of likely mala dherence, then only 47% of participants report ed good adherence using this measure. However, it was noted that many participants endorsed taking their medications either early or late As no guideline was given to participants regarding what constitutes a signific ant deviation from the prescribed time schedule for medication, reports of early or late doses were recorded as maladherence events whenever participants mentioned them. When the retrospe ctive adherence measure was recalculated using only recollections of missed doses, extra doses, or taking medications at a different dosage than prescribed, then the adherence rates using this measure improved. With the alternate methodology, 69% reported likely adherence, more in line with the resu lts obtained with the Morisky technique. These two alternate calculati ons of RAM will be described herein as the Inclusive RAM (including reports of early or la te dosing) and the Exclusive RAM (not including early or late dosing). 54

PAGE 55

Association s among Morisky and RAM Measures of Adherence Surprisingly, the association between the Morisky measurement and the RAM measurement was not significant either with th e inclusion of early / late doses (Spearmans = 0.23, p = 0.12) or without them (Spearmans = 0.17, p = 0.26). The association between the RAM measure of adherence including and exclud ing time deviations from the dosing schedule was strong (Spearmans = 0.63, p < 0.001). Adherence and Depression The association between depression and repor ted adherence was calculated in several different ways using depression as assessed by the MINI and th e CES-D separately, since the concordance between these measures was poor in the sample. Individuals meeting current criteria for a Major Depressive Episode on the MINI were not found to be more likely to report maladherence than individuals using the Morisky measure, the Inclusive RAM, or the Exclusive RAM (Morisky: 2[1] = 0.15, p = 0.70; RAM inclusive: 2[1] = 0.03, p = 0.86; RAM exclusive: 2[1] = 0.36; p = 0.55). When individua ls with a history of mania or hypomania were excluded to consider only individuals who meet criteria on the MINI for Major Depressive Disorder, the results were unchanged. Symptoms of depression as a ssessed by the CES-D Total Score were likewise unrelated to any of the three measures of adherence, either when used as continuous variables (i.e., Spearman correlations with the dichotomous adherence variables) or when us ed to classify individuals as depressed or non-depressed usi ng the established cut-off scor e of 16. Correlations among the CES-D Total Score and the measures of adhere nce are provided in Tabl e 4-4, and dichotomous tests of association among depression and the measures of adherence, with unadjusted odds ratios, are provided in Table 4-5. 55

PAGE 56

In summ ary, no evidence was found in support of an association between depression and self-reported maladherence in this sample of epilepsy patients. Bivariate Relationships among Predictors of Adherence Bivariate relationships among the remaining pr oposed predictors of adherence behavior and the adherence measurements were considered next. Since most of these variables take on wide ranges and have interval or ratio-level data properties, a nd assessment of issues such as multicolinearity for a regression model rely on Pearson correlations, Pearson correlations were used for this analysis, excepting for a few variab les which did not lend themselves well to use as continuous variables, in which case data was reduced into dichotomous form and chi-squared tests of association were used. These results are summarized in Table 4-6. Demographics No sex differences were obser ved in reported rates of adhe rence using any of the three measures (RAM Inclusive: 2[1] = 0.18; p = 0.67; RAM Exclusive: 2[1] = 0.24; p = 0.63; Morisky: 2[1] = 1.28; p = 0.26). Age was not correlat ed with any of th e three adherence measures. (all p > 0.50). The effect s of race were considered using the two largest racial groups in the sample, Caucasians and African Americans. Using the inclusive RAM measure, Caucasian participants were more likely to endorse maladherence than African Americans (65% vs. 23%; RAM Inclusive: 2[1] = 6.30; p = 0.01). This effect was nonsignificant for the other measures of adherence (RAM Exclusive: 2[1] = 2.30; p = 0.13; Morisky: 2[1] = 0.14; p = 0.71). When Caucasians were compared to all other racial groups in the sample, the same results were obtained. Those who had completed high school were not more or less likely to report maladherence using any of the measures than t hose who had not completed this much schooling (RAM Inclusive: 2[1] = 2.10; p = 0.15; RAM Exclusive: 2[1] = 0.00; p = 0.97; Morisky: 2[1] = 0.27; p = 0.60). In summary, among demographic variables, only race was significantly 56

PAGE 57

associated w ith any of the measures of adherence, with Caucasians reporting higher rates of maladherence on one measure. Clinical Characteristics With respect to basic healthca re behaviors and clinical variables, individuals who had partners, family members, or frie nds participate in their doctors appointments were not more or less likely to report maladherence than those who did not receive this kind of assistance, according to any measure (RAM Inclusive: 2[1] = 0.41; p = 0.52; RAM Exclusive: 2[1] = 1.58; p = 0.21; Morisky: 2[1] = 0.41; p = 0.52). The number of AE Ds an individual was prescribed was not correlated with any m easure of adherence, and when those who were prescribed monotherapy were compared to all indivi duals receiving polytherapy, there was also no association between monotherapy and adherence with any measur e. Individuals who used a pillbox were likewise not more or less likely to report adherence than those who did not (RAM Inclusive: 2[1] = 0.07; p = 0.79; RAM Exclusive: 2[1] = 1.37; p = 0.24; Morisky: 2[1] = 0.88; p = 0.35). Age of onset for the seizure disorder achieved trend-leve l significance in its association with adherence for the inclusive RAM only ( 2[1] = 2.97; p = 0.09), with a trend towards higher maladherence with those who had onset in childhood (64% reporting maladherence vs. 39% in those with adult onset). The same pattern was seen for the exclusive RAM, but the effect was not significant ( 2[1] = 1.86; p = 0.17); no difference was seen using the Morisky measure ( 2[1] = 0.01; p = 0.94). Those who had seizur es for at least five years were more likely to report maladherence according to the inclusive RAM than those with shorter illness durations (62% vs. 29%; 2[1] = 4.38; p = 0.04), but this effect lost significance for the exclusive RAM (2[1] = 0.89; p = 0.35) and for the Morisky measure ( 2[1] = 0.12; p = 0.73). As an approximation of quality of seizure control, th ose who reported no seizures in the past 30 days were compared to those who had reported one or more seizure in that time period. Those who 57

PAGE 58

reported no seizures in the last 30 days were not significantly m ore or less likely to report adherence according to any of the measures (RAM Inclusive: 2[1] = 2.12; p = 0.15; RAM Exclusive: 2[1] = 2.65, p = 0.10; Morisky: 2[1] = 0.05; p = 0.83). In su mmary, among clinical variables, earlier age at onset a nd longer duration of seizures were associated with maladherence. Cognitive Characteristics The Shipley Verbal subtest raw score was positively associated with reported maladherence (i.e., higher measured verbal intell ectual abilities were associated with a lower likelihood of adherence) using the inclusive RAM (r = 0.29, p = 0.048), but not to either other measure. The relationship with the inclus ive RAM dropped to trend significance when demographically corrected (r = 0.26, p = 0.86) and the relationsh ips with the other measures remained non-significant. The Abstraction subt est raw score was not associated with any measure of adherence, nor was the corresponding demographically corrected T score. The SILS WAIS-R IQ estimate was also not significantly associated with any measure of adherence. The Stroop Word raw score was associated at the tre nd level, again, positively, to the exclusive RAM (r = 0.28, p = 0.054) but not to the other adheren ce measures. The Stroop Color raw score was positively associated with reported maladhere nce by the exclusive RAM (r = 0.32, p = 0.03) only. The Stroop Color-Word raw score was positiv ely associated with the exclusive RAM (r = 0.32, p = 0.03) and at the trend level with the inclusive RAM (r = 0.28, p = 0.06). The calculated interference was not associated w ith any of the measures, but wh en the interference score was demographically adjusted, the T score was associ ated at a trend level with the inclusive RAM only (r = 0.25, p = 0.09). Neither raw score on the Digit Symbol subtest of the WAIS-III nor the age-corrected scaled score were significantly associated with any measure of adherence. RAVLT Total Learning was negatively associated with maladherence measured by the Morisky measure (r = -0.32, p = 0.03) only; similarl y, the RAVLT Delayed Recall was negatively associated with 58

PAGE 59

Morisky m aladherence at the trend level (r = 0.29, p = 0.052) only; the latter relationship was likewise trend significant only with the Morisky measure (r = -0.28, p = 0.059). In summary, among cognitive variables, higher Shipley Verbal raw scores, higher Stroop Word, Color, and Color-Word raw scores, and lower RAVLT Total Learning and Delayed Recall were associated with maladherence using one or more measure. Healthcare Attitudes and Beliefs Scores on the IPQ were not associated with re ported adherence using any of the measures. On the MHLC, Internal scale scores were unrelat ed to reported adherence. In contrast, higher Chance scale scores were associated with lower reported adherence on both the inclusive and exclusive RAMs (r = -0.34, p = 0.02; r = -0.48, p = 0.001, respectively). The Powerful Others scale score (i.e., the extent to which locus of control is attrib uted to healthcare providers) was negatively associated with th e exclusive RAM maladherence (r = -0.37, p = 0.01) and at the trend level also with the incl usive RAM (r = -0.24, p = 0.10) but not with the Morisky measure. The BMQ Specific Necessity subscale was not signifi cantly associated with any of the measures, but the Specific Concerns scal e was negatively associated wi th inclusive RAM (r = -0.28, p = 0.05) only. Neither the General Harm nor the General Overuse subscales were associated with any of the measures of maladherence. Epileps y knowledge as measured with the EKS was positively associated with inclusive RAM (r = 0.37, p = 0.01) but not with the other measures. The EPKQ was not significantly associated with any of the measures. Finally, the AES was associated at the trend level with both the inclusive and exclusive RAMs (r = 0.26, p = 0.08; r = 0.26, p = 0.08, respectively). In summary, higher Chan ce and Powerful Others locus of control, higher levels of Specific Concerns with respec t to epilepsy medications, and greater epilepsy knowledge were associated with reduced adherence. 59

PAGE 60

The Effect of Depression on Relationsh ips among Adherence and Health Behaviors To assess w hether any of the health behavior s, clinical, or cognitive characteristics that predicted adherence had different effects in depressed individuals than in non-depressed individuals, correlations for the significant variables mentioned above were repeated separately for those individuals who met MI NI criteria for current Major Depressive Disorder and those who did not. In those who did not meet curren t criteria for MDD, the MHLC Chance scale continued to be negatively associated with exclusive RAM (r = -0.51, p = 0.003) and at a trend level with the inclusive RAM as well (r = -0.33, p = 0.07). The BMQ Specific Necessity scale was significantly positively associated with in clusive RAM (r = -0.36, p = 0.04) and at a trend level with the exclusive RAM (r = -0.35, p = 0.052). The EKS was positively correlated with inclusive RAM only (r = 0.38, p = 0.04). The RAVLT Total Learning was negatively associated with Morisky maladherence at the trend level only (r = 0.33, p = 0.07). Caucasian race continued to be associated w ith the inclusive RAM (r = 0.47, p = 0.01). Duration of seizures was no longer a significant predictor. When these correlations were repeated in the depressed group (it should be noted that this group was very small, leading to a very large requirement for the magnitude of correlations in order for them to reach statistical signif icance), the MHLC Chance scale was no longer significantly associated with the exclusive RAM but continued at the tre nd level of association with the inclusive RAM (r = -0.51, p = 0.06). Neit her the BMQ Specific Necessity subscale nor the EKS continued to have significant correlatio n with any measure of adherence. The RAVLT Total Learning score likewise was no longer significan t as a predictor, even at the trend level. Likewise, Caucasian race was also non-significa nt. Duration of seizures continued to be nonsignificant. As well, while the Morisky measure of adherence was not significantly correlated with either the inclusive or exclusive RAM in the non-depressed subgroup, it was moderately to 60

PAGE 61

strongly correlated with both RA M m easures in the depressed group (with the inclusive RAM: r = 0.60, p = 0.02; with the excl usive RAM: r = 0.83, p < 0.001). The correlation between the Morisky measure and the inclusive RAM was signifi cantly greater in the depressed group than in the non-depressed group (Z = -1.7, p = 0.049), as was the correlation between the Morisky measure and the exclusive RAM (Z = -3.6, p < 0 .001). Finally, these associations were also tested with the Kappa statistic, an assessment of association between dichotomous variables, with similar results. In the non-depressed group, the incl usive RAM was likewise not significantly associated with the Morisky measure ( = 0.09, p = 0.54) but this association became significant in the depressed group ( = 0.53, p = 0.02). The exclusive RAM was similarly not significantly associated with th e Morisky measure in the non-depressed group = 0.11, p = 0.52) but this association again became significant in the depressed group = 0.81, p = 0.002). In the non-depressed group, only 53% of participants showed concordance on the exclusive RAM and the Morisky measure of adhere nce. In contrast, the concordance rate was 93% in the depressed group. Results of the se parate analyses for the depressed and nondepressed groups of participants are summarized in Table 4-7. Adherence and Health Behaviors Because of the small sample size, variables with larger bivariate associations with at least one of the adherence measures as well as good distributional properties were chosen for the regression analysis. The set of pr edictors was fixed for regressi ons against both the Morisky and RAM assessments of adherence, in order to allow for comparis on between models. First considering variables significant for any measure of adherence, race, duration, Shipley Verbal raw score, Stroop task scores RAVLT Total Learning, Chance and Powerful Others loci of control, Specific Concerns fr om the BMQ, and knowledge as measured by the EKS were all significantly associated with at least one measure of a dherence at the p < 0.05 61

PAGE 62

level. S ince a number of intercor related cognitive and healthcare attitude variables met this criteria, further rationalization in these areas was undertaken. Numerous strong correlations among the cognitive variables existed; to avoid mu lticollinearity, the cognitive variable with the most robust univariate relationship (by magnitude of both the correlation itself and of the significance, or p value), the RAVLT Total Lear ning Score, was chosen. Among the healthcare attitude variables, since the Chance and Powerful Others scales of the MHLC were similar and positively associated, the Chance scale was chose n, as it likewise had the most robust univariate relationship with one of the adherence measures (again, by both magnitude of the correlation and by significance). This left a total of six predic tors: race, duration, RAVLT Total Learning, MHLC Chance, BMQ Specific Concerns, and the EKS. When bivariate correlations among this remaining group were considered, no strong corr elations (all r < 0.6) remained. This set of predictors was then used for regression models ag ainst each of the three m easures of adherence the inclusive and exclusive RAMs and the Morisky measure. The model assessing the six predictors ability to predict maladherence with the inclusive RAM was significant ( 2[6] = 23.13; p = 0.001) and correctly classified 78% of participants (compared to a correct classificat ion of 52% of partic ipants with a constant model). Caucasian race and the EKS were significant predictors and the MHLC Chance subscale and the BMQ Specific Concerns subscale were significant at th e trend levels. The results of this regression are provided in Table 4-8. When the same model was a pplied to the exclusive RAM, it was also significant ( 2[6] = 13.78; p = 0.03) and correctly clas sified 74% of participants (compared to 67% for this null model). Only the MHLC Chance subscale was signi ficant as a predictor. The results of this regression are provided in Table 4-9. 62

PAGE 63

Next, the model was applied to the Moris ky adherence m easure. This model was not significant ( 2[6] = 8.41; p = 0.21); while it correctly cl assified 80% of participants, the null model classified 76% of particip ant correctly due to the lower ba se rate of maladherence reported by this measure. Since the individual significan ce of the RAVLT Total Learning score was still significant, this model was re-attempted with all other predictors removed. The model predicting the Morisky adherence using only the R AVLT Total Learning was significant ( 2[1] = 5.10; p = 0.024) and correctly classified 81% of participants. The odds ratio associated with each point change in the RAVLT Total Learni ng was 0.921 (95% CI = 0.853 0.995). Adherence, Depression, and Seizure Control Finally, to assess the relationships among reported adherenc e, seizure control, and depression, chi-square tests for the association between seizure contro l and adherence were repeated separately for the participants who did and did not meet cr iteria for current Major Depressive Disorder on the MINI. For these tests, seizure control was approximated by comparing participants with no repo rted seizures in the past 30 days with those who had one or more reported seizures in the last 30 days. On the inclusive RAM measure, the association between seizure control and adherence was not significant for eith er the depressed ( 2[1] = 1.75; p = 0.19) or non-depressed group ( 2[1] = 1.01, p = 0.31). Similarly, on the exclusive RAM, no significant association was found for either the depressed ( 2[1] = 1.75; p = 0.19) or the nondepressed ( 2 [1] = 0.93; p = 0.33) group. Finally, for the Morisky measure as well, no significant association was found for either the depressed ( 2 [1] = 0.01; p = 0.94) or the nondepressed group ( 2 [1] = 0.64; p = 0.43). To evaluate the possibility that depression might have an association to seizure control without taki ng adherence into account, the association between current Major Depressive Disord er and seizure control was al so considered. Those who met 63

PAGE 64

64 criteria for current MDD were not significantly more or less likel y to report being seizure free over the past 30 days ( 2 [1] = 2.29; p = 0.13).

PAGE 65

Table 4-1. D emographic and clinical characteristics of participants Characteristic Frequency (%) or Mean (SD in parenthesis) Female 79% Age 42 y (12.5 y) Race: Caucasian 63% African American 29% Education: <12 y 28% 12 y 28% >12 y 44% Married 49% Seizure onset at 21 y old or older 51% Illness duration > 5 y 71% Period of seizure freedom >6 months (lifetime since seizure onset) 53% Seizure freedom in the last 30 days 28% Number of medication doses per day 2.0 (0.6) Receiving monotherapy 19% Table 4-2. Psychiatric comorbidit ies observed in the study sample Comorbidity* Number Meeting Criteria Percentage Major Depressive Episode 19 40 Dysthymia 4 8 Mania 1 2 Hypomania 1 2 Panic Disorder 6 13 Agoraphobia without Panic Disorder (AWOPD) 2 4 Social Phobia 4 8 Obsessive-Compulsive Disorder 2 4 Posttraumatic Stress Disorder 7 15 Alcohol Dependence 3 6 Substance Dependence 2 4 Anorexia Nervosa 0 0 Bulimia Nervosa 2 4 Generalized Anxiety Disorder 10 21 Notes: All comorbidities listed were assessed for current fulfillment of criteria. 65

PAGE 66

Table 4-3. C ognitive characteristics of study participants Test Average Score* Minimum Maximum SD SILS Vocabulary Raw Score 23.2 11 36 5.9 SILS Vocabulary T score 37.7 15 60 10.2 SILS Abstraction Raw Score 15.2 2 36 8.4 SILS Abstraction T score 44.4 31 62 7.6 SILS Estimated FSIQ (StS) 96.6 76 112 9.3 Stroop Word Raw 76.1 38 113 17.3 Stroop Color Raw 59.4 29 89 15.0 Stroop Color-Word Raw 30.9 0 64 13.7 Stroop Color-Word Interference -2.1 -17.6 20.6 9.2 RAVLT Total Learning 36.7 14 58 10.2 RAVLT Delayed Recall 6.3 0 12 3.4 Digit Symbol Raw Score 53.5 14 89 21.2 Notes: Scores are listed as raw scores unless otherwise indicated. T scores and Standard Scores represent demographically corr ected performance in comparison to published population norms. Where T scores are indicated, these scores have a mean of 50 and a standa rd deviation of 10 in the normative sample. Where Standard Scores (StS ) are listed, these have a mean of 100 and a standard deviation of 15 in the normative sample. Table 4-4. Correlations among CES-D Tota l Score and measures of adherence Correlation with CES-D Total Score sig Inclusive RAM 0.04 0.77 Exclusive RAM 0.00 0.99 Morisky 0.12 0.40 66

PAGE 67

Table 4-5. A ssociations among measures of depression and measures of adherence Measure of Depression Measure of Adherence Percentage of depressed individuals reporting maladherence Percentage of non-depressed individuals reporting maladherence Risk Ratio Test Statistic sig CES-D Inclusive RAM 59% 38% 1.6 2 [1] = 2.05 0.15 Exclusive RAM 34% 25% 1.4 2 [1] = 0.44 0.51 Morisky 24% 19% 1.3 2 [1] = 0.15 0.70 MINI Inclusive RAM 53% 50% 1.1 2 [1] = 0.03 0.86 Exclusive RAM 37% 29% 1.3 2 [1] = 0.36 0.55 Morisky 26% 21% 1.2 2 [1] = 0.15 0.70 67

PAGE 68

Table 4-6. Relationships among health knowledge, attitudes, and adherence Inclusive RAM Exclusive RAM Morisky IPQ MHLC Internal MHLC Chance MHLC Others BMQ Specific Necessity BMQ Specific Concerns BMQ General Harm BMQ General Overuse EKS EPKQ AES Inclusive RAM 1.00 0.62** 0.23 0.13 0.01 -0.34* -0.24 0.16 -0.28* 0.08 0.09 0.37* 0.20 0.26 Exclusive RAM 1.00 0.17 0.04 0.09 -0.48** -0.37* 0.13 -0.04 0.03 0.13 0.14 0.05 0.26 Morisky 1.00 0.12 0.15 -0.01 0.03 -0.03 0.14 -0.04 0.09 0.02 -0.13 0.01 IPQ 1.00 0.01 -0.05 -0.14 -0.14 -0.46** -0.22 -0.19 -0.14 0.02 0.29* MHLC Internal 1.00 0.04 0.12 -0.20 0.22 0.39** 0.10 0.08 0.03 -0.21 MHLC Chance 1.00 0.35* -0.06 -0.06 -0.16 -0.38** 0.31* -0.15 0.00 MHLC Others 1.00 -0.21 -0.03 0.29* 0.02 -0.15 -0.08 0.10 BMQ Specific Necessity 1.00 -0.01 -0.11 -0.11 0.33* -0.11 0.34* BMQ Specific Concerns 68 1.00 0.37* 0.40** 0.09 -0.04 -0.21 BMQ General Harm 1.00 0.64** -0.02 0.15 -0.02 BMQ General Overuse 1.00 0.17 0.25 0.11 EKS 1.00 0.32* -0.10 EPKQ 1.00 0.06 AES 1.00 Note: Correlations marked with () are tr end-significant at the p < 0.10 level; (*) ar e significant at the p < 0.05 level, and (**) at the p < 0.01 level.

PAGE 69

Table 4-7. E ffects of attitudes and knowledge on depressed and non-depressed participants Inclusive RAM Exclusive RAM Morisky Not Depressed DepressedNot Depressed Depressed Not Depressed Depressed Inclusive RAM 1.00 1.00 0.65** 0.73** 0.11 0.60* Exclusive RAM 0.65** 0.73** 1.00 1.00 -0.11 0.83** Morisky 0.11 0.60** -0.11 0.83** 1.00 1.00 MHLC Chance -0.33 -0.51 -0.51** -0.36 0.11 -0.20 BMQ Specific Necessity 0.36* -0.30 0.35 -0.27 0.04 -0.20 EKS 0.38* 0.39 0.20 0.00 0.04 -0.07 Note: Correlations marked with () are trend-sign ificant at the p < 0.10 level; (*) are significant at the p < 0.05 level, and (**) at the p < 0.01 level. Table 4-8. Logistic regression model predicting inclusive RAM Predictor B SE Wald Df Sig Risk Ratio 95% CI Constant 0.52 3.44 0.02 1.00 0.88 1.69 Caucasian race 2.02 0.99 4.21 1.00 0.04 7.57 (1.10, 52.32) Duration of epilepsy >5 yrs 1.21 0.90 1.83 1.00 0.18 3.36 (0.58, 19.47) RAVLT Total Learning Score -0.050.051.21 1.000.270.95 (0.87, 1.04) MHLC Total Chance Scale -0.13 0.08 2.75 1.00 0.10 0.87 (0.75, 1.02) BMQ Specific Concerns -1.09 0.61 3.22 1.00 0.07 0.34 (0.10, 1.11) EKS Total Score 0.43 0.21 4.18 1.00 0.04 1.54 (1.02, 2.34) Table 4-9. Logistic regression model predicting exclusive RAM Predictor B SE Walddf sig Risk Ratio 95% CI Constant 2.93 3.45 0.72 1.00 0.40 Caucasian race 0.83 0.87 0.91 1.00 0.34 2.30 (0.41, 12.77) Duration of epilepsy >5 yrs 0.66 0.90 0.54 1.00 0.46 1.94 (0.33, 11.32) RAVLT Total Learning Score -0.020.040.31 1.000.580.98 (0.90, 1.06) MHLC Total Chance Scale -0.23 0.08 7.18 1.00 0.01 0.80 (0.68, 0.94) BMQ Specific Concerns -0.11 0.50 0.05 1.00 0.83 0.90 (0.33, 2.39) EKS Total Score 0.03 0.18 0.02 1.00 0.89 1.03 (0.71, 1.47) 69

PAGE 70

CHAP TER 5 DISCUSSION Review of study findings Aim 1: Relationship between Depression and Adherence It was hypothesized that, consiste nt with findings in studies of adherence in other diseases, co-morbid depression would be significantly associated with decreased adherence. Despite the use of two independent measures of self-reported adherence as well as two measures of depression, no relationship between depression and adherence wa s observed, nor even a trend towards such a relationship. This was quite unex pected, since this finding has been widely reported and replicated in other disease populatio ns. On the one hand, the present study had an insufficiently large sample size to conclusively ru le out the possibility of a relationship between depression and adherence in epilepsy patients. On the other hand, the la ck of even a trend towards such a pattern is surprising. Adding to the surprising nature of this finding are the observati ons that the base rates of reported maladherence and depression are both genera lly consistent with expectations from other studies, limiting the possibility that this finding arose from an atypical sample. The rates of maladherence, ranging from 24% for the Morisky measure to 31% for the exclusive RAM and 53% for the inclusive RAM, were generally consis tent with reports of maladherence rates both across chronic illness groups and specifically when comparing the obtained Morisky rate of adherence in comparison to with studies of epilepsy patients us ing this same measure (Jones et al., 2006; McAuley et al., 2008). For instance, Jone s et al. (2006) reported that 59% of their sample endorsed one or more Morisky items, whic h is comparable to the 67% observed in the present study (this study classified all individuals who endorsed at least one Morisky item as maladherent). McAuley et al. ( 2008) reported that 69% of partic ipants endorsed 0-1 Morisky 70

PAGE 71

item s, which compares favorably with a rate of 76% observed in the present study. Likewise, while the rate of reported depression according to a traditional cut-off score for the CES-D was higher than expected (70%), the percentage of individuals meeti ng criteria for a current Major Depressive Episode on the MINI (40%) was very comparable to the observed rates of depression of 20-55% in people with incompletely controlled epilepsy (Kanner, 2003). Aim 2: Relationships among Health Knowledge, Attitudes, and Adherence It was hypothesized that health knowledge, be liefs, and attitudes woul d predict rates of adherence, and this hypothesis was partially validated. Measures of healthcare knowledge and attitudes in several different areas were assessed: perception of the impact of illness, beliefs about medications, beliefs about locus of cont rol in the context of health management, knowledge about epilepsy, and apathy as a measure of ability to engage in motivated behavior. None of these measures was found to be significa ntly associated with Morisky reports of adherence. In contrast, several of them were found to be associated with RAM reports of adherence. Higher levels of Chance or Powerf ul Others locus of control on the MHLC instrument were found to be associated with more reported maladherence on the RAMs. Higher levels of endorsed concerns about epilepsy me dications on the BMQ Specific Concerns subscale were also found to be associated with more reported maladherence on the RAMs. Finally, higher knowledge as measured by one of two knowledge instruments had the same association. While the first two findings are generally consistent with expectations, the finding that greater levels of epilepsy knowledge predicted lowe r adherence was surprising. Additionally, several other variables were found to si gnificantly predict reported adherence. Among demographic variables, Ca ucasian race was found to be a significant predictor of lower adherence rates. This finding ha s been replicated by at least one other study in epilepsy (McAuley, 2008). The relatively weak im portance of demographic factors in predicting 71

PAGE 72

adherence is consistent with studi es outside of epilepsy that exam ine adherence to m edications in chronic illnesses have generall y suggested that demographic variables are not very strong predictors of adherence (e.g., Curtis et al., 2009). Among basic clinical variables, only the related characteristics of age at onset a nd duration of the seizure disorder were predictors of adherence. Finally, among cognitive variable s, two separate patterns were observed. A number of cognitive variables performance on the Ship ley Institute of Living Scales Verbal subtest, a measure of verbal reasoning ability, as well as speed of performance (but not level of inte rference) on the Stroop Test were positively associated with the RAM meas ures of adherence. That is, individuals who performed bette r on these instruments were more likely to report maladherence. The second pattern observed was that performance on the RAVLT was negatively associated with the Morisky measure that is, that indi viduals with poorer verb al memory performance were more likely to report maladherence. Thus, in terestingly, individuals with stronger cognitive performance were likely to report maladherence on one instrument, and individuals with weaker cognitive functioning (albeit, in different domains) were likely to report maladherence on another. Aim 3: Effects of Knowledge and A ttitudes in Depressed Individuals It was hypothesized that knowledge, belief, and attitude measures would be more strongly associated with adherence reports in depressed individuals because of the effect of depression on these cognitive and emotional processes. Surpri singly, while different re lationships were found between these variables and adherence in depressed and non-depressed individuals, the opposite pattern was largely found, with these variables having generally more predictive power in nondepressed individuals, falling mostly to nonsignificance in depressed individuals. The effect of depression on the relationships between other predictors of adherence took a somewhat complex form. In non-depressed individua ls, the relationships tended to be similar to 72

PAGE 73

the relationships in the whole sam ple, with different health behavioral and c ognitive variables predicting the RAM and Morisky m easures of adherence, which were themselves not strongly associated. In contrast, in the depressed group, many of these predictors became non-significant, but the relationship between the Morisky and RA M measures of adherence became quite strong. In fact, in the depressed group, only a single person had different assessments of adherence by the Morisky and exclusive RAM methods. Aim 4: Depression, Adherence, and Seizure Control Surprisingly, no significant relationships were found be tween seizure control and adherence using any of the adherence measures, either in the full sample or in either the depressed or non-depressed groups. This is in contrast to the findings of Jones (2006) who found an association between the Morisky measure and seizure control, using a similar measure of seizure control (patient report of less than one seizure per month). Likewise, no relationship was found between depression and seizure control even without c onsidering adherence. Implications of the Study Depressions Role in Epilepsy Adherence The surprising lack of an a ssociation between adherence an d depression in this study, using two different measures of adherence as we ll as two different measures of depression has several possible explanations. First, the study was nonetheless under-powered to detect adherence differences between populations at the average rate observed in other disease populations. Given this and the absence of other published studies assessing the relationship between adherence and depression in epilepsy patients, interpreting the findings as indicative that there truly is not the strong relationship between these variables in epilepsy that exists outside of epilepsy is not yet justified. Although this may be the case, more work would be necessary to confirm this hypothesis. As adhere nce studies do not exist in every chronic or 73

PAGE 74

serious illness, however, it has not fu lly been esta blished whether there are significant differences in medication adherence rates from dis ease to disease. One notable issue is that much of the research on adherence and depression ha s been done with populations that have nonneurological diseases. While resu lts associating adhere nce and depression have been reported in neurological diseases, including multiple sclero sis (Mohr et al., 1996) and Parkinsons disease (Grosset, Bone, & Grosset, 2005) these studies did not report data that could readily be converted into an odds or risk ratio as has typically been done in other studies of adherence and depression, and it is difficult to determine, based on these studies, whether the expectation for the risk for maladherence conferred by co-morbid depr ession should be of the same size observed in other studies or not. Should the risk be smaller for some reason in patients with neurological conditions, then a substantially larger sa mple might be needed to detect it. One aspect that differentiates neurological and psychiatric diseases from diseases of other systems of the body is that many of them have a direct effect on cogni tion and/or emotion, by virtue of disrupting brain func tion, in addition to the indirect effects on thes e areas through biological (e.g. stress response) a nd/or other psychosocial processes (e.g. stigma, loss of roles, etc.). In some neurological diso rders, depression is less a co-morbi d feature and nearly a part of the typical disease process in the case of Parkinsons disease, emotional disturbances associated with depression can even precede clearly identifiable motor symptoms (Mller et al., 2006; Reisberg et al., 2008). In th is way, depression may possibly be thought of as a marker of the underlying neural disruption of epilepsy itself, rather than as a secondary and separate disease process. Indeed, some authors have suggest ed that, as in Parkinsons disease, depression may sometimes be a prodromal feature of epile psy (Kanner, 2008). This would argue that epilepsy-related depression is in some way a different disorder than observed in other chronic 74

PAGE 75

illn ess populations not involving brain dysfunctio n. Arguing against this, however, are the observations that depression happens at a similar base rate in epilepsy as in other symptomatic chronic illnesses, and that the clinical presentati on of depression in epilepsy is by and large very similar to that in other chronic illness populations as well as in depre ssed individuals without other medical co-morbidities (Evans et al., 2005 ). The neurological basis of epilepsy may also make health behaviors different in epilepsy in ot her ways, perhaps by eliciting different kinds of supportive services from family members and other loved ones. The effects of epilepsy on memory might also distort either adherence or ab ility to report adherence. Arguing against these last two hypotheses is the finding that reported rates of adhere nce and adherence measurement tools generally appear similar in people with epil epsy and other individuals, as well, in the latter case, as the finding that at least one of the adherence measures in this study was not associated with an objective measure of memory. Nonethel ess, given the strong body of literature finding adherence issues across diseases to be related to depression, and the similarities observed in adherence behaviors themselves in people with epilepsy and people with other chr onic illnesses, multiple replications or very large samples would be necessary before we can conclude that depression does not affect adherence in people with epilepsy. Additionally, while DiMatteo et al. (2000) demo nstrated effects on adherence that were isolated to depression, it is notable that the popula tion in the present study had significant psychiatric comorbidity outside of depression, including substant ial rates of panic disorder, generalized anxiety disorder, a nd posttraumatic stress disorder. Even if these disorders do not directly affect adherence, this study included in dividuals as depressed participants who often had other psychiatric como rbidities, and individuals as not depressed who may likewise have had psychiatric comorbidities other than depressi on. While this increases generalizability to the 75

PAGE 76

clinical setting, where psyc hiatric disorders are often not seen in isolation, un m easured effects of these comorbidities may also confound attempts to detect the effect of de pression on adherence. Possible Roles of Knowledge and Attitudes in Adherence It was hypothesized that health care knowledge and attitudes might be mechanisms through which the cognitive and emotional sequelae of depression affect adherence behaviors. Surprisingly, it was instead found that, not only di d adherence not depend on depression in this sample, but that adherence in depressed individu als was more uniformly described by different adherence measurements and had few associatio ns with healthcare knowledge and attitudes, whereas, in contrast, these aspects of thought and belief were modestly strong predictors of adherence in non-depressed individuals. Although healthcare beliefs, attitudes, and knowledge are certainly acknowledged to vary in non-depressed indi viduals, there is no theoretical basis to expect problems in these areas in the non-depr essed individuals inst ead of the depressed individuals. There are a number of possible interpretations of this. One possibility is that depression either reduces the range or reliability of report on these other in struments, artificially making them less capable as predictors of adhere nce. Another possibility is that depression acts on adherence through completely different mechanis ms that were not examined in this study. It is also possible that depression truly does not have a significan t impact on adherence rates in people with epilepsy, but that the proposed mode l might describe its mechanism of impact in other populations where it does ha ve the generally established impact of reducing adherence rates. Finally, it is possible that, while the tw o adherence measures more closely measure the same thing in people with epilepsy and co-morbid depression, they do not actually measure adherence effectively, thus lead ing to a spurious lack of associ ation. However, if this were the case, it would seem quite odd that these meas ures would produce such comparable rates of adherence in depressed a nd non-depressed participants. 76

PAGE 77

In term s of the independent impact of knowledge and attitudes on adherence, some of this studys findings were consistent with the idea th at maladherence is, in part, the by-product of subjective theories developed by patients through which they attempt to understand their relationship with their illness and their ability to take actions to maintain their health and wellbeing (Wagner, 2003; Remien et al., 2003). Fo r instance, it was found that individuals who had specific concerns about their epilepsy medi cations were less likely to be adherent. One possible explanation for this is that when patients identify concerns about their medications, they may develop subjective models for how their medications may negatively impact them, and based upon these models, either more frequently choose to engage in maladherence or else be more passively tolerant of thei r maladherence and less motivated to change it. Supporting this is the finding in another study that both adherenc e and the BMQ Specific Concerns and Specific Necessity subscales were associat ed with seizure control (Jones et al., 2006). Similarly, it was found that individuals who had higher Chance and Powerful Others loci of control were less likely to be adherent. It seems reasonable that individuals who pe rceive that their health is determined by forces outside their control are un likely to see their own actions as being strong determinants on their health status. Therefore, these individuals may place a low importance on the consistency of their actions, for instance be ing relatively indifferent to missing medication doses. At least one other study of adherence behaviors outside of epilepsy has found that adherence was associated with higher internal locus of control and lower chance and powerful other loci of control (Ubbiali et al., 2008). The findings that stronger levels of epilepsy knowledge on at least one measure of this construct led to lower reported ra tes of adherence is more surpri sing, especially considering that it was found in the same measures (the RAMs) in which better cognitive performance was also 77

PAGE 78

found to be associated w ith lower reported adhere nce. One possible explanation is that the adage that a little bit of knowledge can be dangerous holds true in this case, and that certain levels or kinds of knowledge about seizure disorders or about health and medicine in general may fuel the subjective theories described by Remien et al (2003) and cause individuals to engage in intentional maladherence. Anot her possibility is that both th e higher levels of epilepsy knowledge and higher cognitive functioning suggest th at individuals reported more events of maladherence on the RAM not because they actually were more maladherent, but because they either better remembered these events or were more sensitive to reports of subtle maladherence, raising their internal standard for good adhe rence to their medication regimen or more conscientiously reporting their adherence behaviors. Arguing in favor of this is the finding that the relationship between epilepsy knowledge and adherence was found only in the inclusive RAM and not in the exclusive RAM, as the inclus ion in the former measure of early or late dosing by the participants ow n definition allowed more r oom for report of subtle maladherence. On the other hand, the relationships between health attitude or belief measures on instruments such as the MHLC and BMQ and these same RAM measures of adherence seem harder to explain if the adhere nce reports are primarily driven by this conscientiousness, and if both mechanisms are at play, disentangling them will be challenging. Implications for Seizure Control The surprising failure to find a relationship between seizure control and adherence may be attributable to a number of cause s. Some non-significant associati ons did approach significance, and it is possible these might become significant in a larger sample. In add ition to the effects of sample size, this result may reflect choices made in the selection of the sample. While the sample of individuals who have a stable diagnosis but continue to have had recent seizures was intended to generate a sample in which the question of adherence would be particularly relevant, it is 78

PAGE 79

possible that the opposite result was obtained by excluding a large num ber of adherent and seizure free individuals who had not had a seizure in the past six m onths, as in the study by Jones et al. (2006). Although this is possi ble, the fact that observed a dherence rates are generally in keeping with those reported elsewhere does make this seem somewhat less likely. Finally, there are inherent limitations in considering only the last 30 days as a measure of the quality of seizure control an individual who typi cally has one seizure in six mont hs would be classified, quite arbitrarily, by this system into the good seiz ure control cate gory if the most recent seizure occurred more than 30 days ago, and into the poor seizure control category if it happened to occur in the last 30 days, even if this did not represent a deviation from the historical seizure frequency. As for the relationship between depression and seizure control, the failure to find this relationship was also surprising. On the one ha nd, this study selected only individuals with relatively recent seizures. As a result of this in clusion criteria, the eff ect of individuals who do not have recent seizures, who in other studies (as discussed by Kanner, 2003) have been found much less likely to be depressed, were not considered. It is po ssible that depression may not have an impact on short-term seizure frequency. Fo r instance, Haut, Shinnar, & Mosh (2005) found that there was no significant relationship between seizure clustering and de pression. On the other hand, since the majority of partic ipants in this study (50%) had 0-1 seizures in the past month, with a small number having a much larger number of seizures, it is also certainly possible that this effect is due primarily to the difficulty in measuring changes in se izure frequency in a group that has a relatively low base seiz ure frequency. Also not considered in the present study is that many of the participants who were depressed were actively being treated with antidepressants. While those classified as depressed for study pur poses still met full criteria for Major Depressive 79

PAGE 80

Disorder, it is also possible that any partial antidepressant res ponse obtained by participants had a confounding im pact on seizure frequency. The Challenge in Measuring Adherence Of the three chief constructs under study in this dissertation, ep ilepsy, depression, and adherence, adherence is unique in that there is no establis hed reference measurement or assessment of adherence that is considered essentially fully accura te (Dunbar, 1984). Each technique (self-report pill counting, electroni c monitoring of adherence, and blood assay monitoring of medication levels) has significant limitations. In a ddition, there is no consensuallyagreed upon definition of a maladherence event, part icularly in the case of a deviation from the time at which a dose is prescribed to be taken. Wh ile researchers have atte mpted to define this construct (e.g., Cramer, Vachon, Desforges, & Su ssman, 1995), these definitions have not been universally adopted. One might imagine a continuous video monitoring of a patient in thei r natural environment (although even this might result in measurement reactivity). Were this referential standard to exist, then all other, more economical measur es of adherence could be assessed for efficacy against it. Unfortunately, in the absence of such a circumstance, no existing measure of adherence is able to play this ro le just as the self-report measur e of adherence is susceptible to the possibility of forgetting or intentionally distor ting the recollection of a dherence behaviors, so too the blood level is susceptible to mechanisms other than adherence that alter observed levels of a medication. Thus, significant work still needs to be done to establish more acceptable operational definitions of adherence and maladherence throug h the process of convergent and discriminant validation. Work by the Internat ional Society of Pharmacoecono mics and Outcomes Research Economics of Medication Compliance Working Group is encouraging in this regard, as this body 80

PAGE 81

m ay have the ability to standardize practice within research as well as clinical settings (Hughes et al., 2007). Even if an even tual standard lacks complete reliability, the use of common protocols is likely to extend the interpretabil ity and comparability of adherence research. While the two measures of adherence used in this study were both within the self-report domain, they had relatively poor concordance, except when considered in depressed individuals separately. The finding of poor concordance be tween the Morisky measure and the RAMs is consistent with results reported outside of ep ilepsy (Gao & Nau, 2000). However, in the absence of a referential standard, it is hard to interpret the disconcordant results of the instruments. In judging the two measures of adherence used in the present study, then, it becomes necessary to attempt to understand their behavi or in terms of their relationshi ps with other study variables. One striking finding is that only one of the two measures, the Mori sky measure, correlated with a test of memory and learning, the Rey Auditory Verbal Learning Test In fact, where the retrospective adherence measure (RAM) correlated with cognitive variables, it had the unexpected pattern of generally indicating more reported maladherence in cognitively stronger participants. There are many possibl e interpretations of this. Certai nly, one possibility is that the relationship between the RAVLT and the Morisk y measure is an indicator that memory functioning plays a role in adhe rence and the Morisky measure, since it is correlated with memory functioning, is more likely to concord with the true adherence behaviors. At the same time, it might be argued that the task in the RAM is not truly a retrospective memory task (as the RAVLT is) but is actually a test of an aspect of prospective memory (i.e., memory that a task that was to be done, was in fact done; G ould, McDonald-Midzczak, & King, 1997). The body of literature investigating prospective memory is much smaller than the body of literature investigating retrospective memory, with some st udies in other illnesses suggesting that they 81

PAGE 82

m ay be incompletely correlated (e.g., Mart ins & Damasceno, 2008). Only two studies were found investigating prospective memory in peop le with epilepsy (Lpez-Gngora et al., 2008; Adda et al., 2008). One of these studies (Adda et al., 2008) did sugge st, however, that prospective memory may be impaired via pathol ogy of the same mesial temporal structures through which retrospective memory is impaired in individuals with te mporal lobe epilepsy. Nonetheless, this research is still limited, and one might not expect the RAM, if it measures an aspect of prospective memory, to correlate strong ly with a test of retrospective memory. Still another possibility is that individuals with weak memo ry performance may be less aware of their adherence behaviors, and that their report mi ght represent a systematic bias to endorsing maladherence in the absence of me mories that contradict it, al though there does not seem to be an obvious reason to expect a syst ematic bias towards reporting maladherence instead of adherence. Another striking finding was that, while the m easures of adherence were poorly correlated in the non-depressed portion of the sample, they became very highly concordant in the depressed portion of the sample. While st udies have looked at the eff ect of depression on assessed adherence, we are not aware of othe r studies that have looked at the performance of self-report adherence measurements in de pressed and non-depressed indi viduals. One possibility for explaining this finding is that individuals who are not depresse d use disparate processes for responding to the two types of questionnaires. Su pporting this is the obser vation that only the Morisky measurement was strongly associated wi th measures of memory. As discussed above, memory may play a different role in the Morisky measure, which does not require identification of specific instances of maladherence, and the RA M, which does. In qualitative analysis of the concordance of the Morisky and RAM measures in the non-depressed and depressed groups, it 82

PAGE 83

was noted that num erous individual s who were discordant were pr esent both in the category of those endorsing maladherence by the Morisky measure but not by the RAM and in the category of those endorsing maladherence by the RAM and not the Morisky measure, so if an effect based on memory does occur, it does not appear to be a biasing effect. It is also possible that the difference in performance of the instrument s across groups was due not to difference in utilization of a cognitive ability, such as memory, but to an attitudinal or motivational difference, such as a bias towards self-negativity in individuals who are de pressed, although again, reconciling this with the lack of a directionality of the effect of depression seems difficult. Certainly, given the extremely small sample sizes considered in the separate depressed and nondepressed populations as well as the number of biva riate tests of association conducted, the possibility that this effect is a form of Type I error, cannot be ruled out. While this study cannot draw a strong conclusi on about the relative efficacy of the two different types of measures, one strength of th e study is that it demonstrates how they behave differently in this sample of people with epilep sy. More studies that compare different adherence measures head to head in the same sample w ould be beneficial in be tter characterizing the strengths and weaknesses of adherence measures. Characteristics of the Study Sample Epilepsy studies are often faced with the challenge of balancing the desire for representativeness of clinical seizure disorder patients and with adequate control for the confounding effects of the heterogeneity of popula tions of seizure disord er patients. At the extreme of generalizability, ve ry heterogenous samples of pa tients can be problematic. For instance, adherence behaviors may be very differe nt in patients who have been seizure free for many years on stable medication management as compared to patients with newly diagnosed seizure disorders. Studies outside of epilepsy have suggested that in dividuals behaviors and 83

PAGE 84

cognitions related to adherence differ when diseases are seen as acute or episodic and when they are seen as chronic (Halm, Mora, & Leventhal, 2006). Patients who have been seizure free for m any years may even be taking their medications for a minimal benefit of prophylaxis in some cases that is, they may not be likely to have seizures if they are maladherent or discontinue their medications. At the other extreme, while very specific populations such as intractable temporal lobe epilepsy patients are fairly well characterized and homogenous, their medicationresistant disorder variants may be different than that seen in the 80% of epilepsy patients whose seizures can be controlled via medications. Th e present study attempted a compromise between these extremes by considering adult patients with a variety of seizure disorder characteristics, but limiting the sample to individuals who have a stab le diagnosis as well as recent seizures. The former limitation was chosen to exclude individual s who have isolated se izures that might not evolve into epilepsy and whose prognosis mi ght depend primarily on diagnosis and / or management of another primary disorder (e.g. diabe tic seizures), as well as to remove the added complexity that newly diagnosed patients may be on complex titration schedules or otherwise experience a relatively large number of medication changes in a s hort period of time. The latter limitation was chosen to exclude seizure-free pa tients who may have relatively fewer risks associated with any deviations from medication adherence. This group may include at least some individuals whose seizures are, or will be, consider ed intractable. It also likely includes at least some people who will achieve sa tisfactory seizure management through medications. While this still certainly represents a subpopulation of people with seizures it is likely to represent a subpopulation in which adherence problems, could they be identified an d intervened upon, might hold potential for clinical benefit. 84

PAGE 85

Beyond the above restriction, no ex plicit restrictions were pla ced on recruitm ent, such as cognitive exclusionary criteria. At the same tim e, the studys relativel y intensive dependence on multiple choice and Likert scale questionnaires di d require a fair amount of reading ability, and patients were not enrolled who would not be able to complete these questionnaires with minimal assistance. This issue was surprisingly common du ring study recruitment. In part as a result of the reading requirements of the study, while the participants had a range of cognitive functioning, the lowest estimated FSIQ in the sample was 76, and previous studies have indicated that a number of epilepsy patients who have had seizures for an extended period of time have cognitive compromise beyond this level (Hermann et al., 2006; Dodrill, 2004). Limiting recruitment to individuals able to indi vidually complete questionnaires is the likely reason for this. On the one hand, this was a necessary limitation for pragmatic reasons. On the other, this study cannot speak to adherence challenges with more cognitively impaired subgroups. With respect to presence of depression, individuals with depr ession were not specifically recruited. Although, as previously noted, the po int prevalence of depression in the study population was well within expectations, it was also observed anecdotally that individuals who, per their clinical records, were more severely depressed, frequently declined to participate in the study. While no formal record of this could be ke pt, it is possible that th e depressed individuals in the study represent a selective su bset of all depressed individuals with seizure disorders. it is possible that severely depressed individuals were under-recruited. To the extent that difficulty in engaging in motivated behaviors is a component of the pathology of depression, it is also notable that individuals who are de pressed and particularly l acking in motivation may be underrepresented in voluntary studies of this kind in general. Th is issue might even raise the 85

PAGE 86

need for stratified sam pling of depressed individu als in studies of adherence and depression, for instance by placing additional emphasis on duration and severity of depression, and whether treatment is being received for depression. Additi onal research is also needed to determine whether participants who are experiencing a depre ssive episode that may be in the context of underlying bipolar disorder s hould be considered or ex cluded from analysis. Study Limitations This study has a number of limitations. Most prominently, the smaller-than-anticipated sample size limits the interpretability of negative re sults, such as the lack of relationship between adherence and depression, as well as limiting th e ability to assess more complex mediational models such as the hypothesis th at changes in healthcare knowledge utilization, beliefs, and attitudes may mediate the relationship between de pression and adherence. In spite of designing a relatively brief study protocol, providing financial compensation, attempting to make the study convenient for patients by conducting recruitment at outpatient epilepsy clinics, and recruiting at multiple epilepsy clinics in different locations the rate of participant recruitment was significantly lower than expected. Several fact ors may contribute to this. As mentioned previously, a number of individu als who were identified as potential candidates during prescreening were found to have s ubstantial cognitive impairments such as extreme psychomotor retardation or extremely poor reading ability, and these participants were frequently unable to participate in the study because they were una ble to complete the questionnaire packet or complete the interview independently. Considering those patients who met eligibility criteria and were able to participate in the study at a basic level, successful recruitment rates were still low. Most patients who elected to enroll in the study expressed their pleasure in doing so, frequently noting that they enjoyed opportuni ties to support scien tific progress in un derstanding epilepsy and that they enjoyed giving back to the epile psy community. Informally, participants who did 86

PAGE 87

not enroll indicated concerns over tim e pressure or lack of transportation as barriers to enrolling in the study. While attempts were made to make alternative arrangements to enroll participants outside of their clinic visit, this proved time consumi ng and also resulted frequently in prospective participants failing to attend their scheduled study session. Prospective participants also informally cited the length of the study as a concern. Finally, the total volume of patients in outpatient epilepsy clinics at the recruitment sites was affected adversely during the course of the study by the departure of two epileptologists fr om the University, re ducing the number of individuals who could be recruited for the study. In addition to logistical barriers, the recruitment technique may also entail systematic biases rela ted to individuals who did or did not agree to participate, such as individuals who see themselv es as epilepsy advocates being more likely to participate and individuals who experience stigma being less likely to participate. Future studies of this kind may be more likely to succeed in recr uitment if a fixed protocol is maintained for a relatively long period of time (e.g., se veral years) as part of a broa der research init iative within an epilepsy clinic. Combining epidemiological tec hniques such as brief mail-in or internet-based surveys with targeted recruitment of a subset of participants for more detailed in-person assessments may also allow for both large sample sizes for basic analyses and the ability to more thoroughly assess possible aspects of healthcare knowledge and attitudes that impact adherence. While the study recruited both urban and rural participants, it did so only in a single geographic region of the United States, possibly limiting interpre tation to other regional or international groups of patients. Likewise, the choice to include a broad sample of epilepsy patients has the potential to obscu re any subgroup-specific mechanisms of adherence. Similarly, as all participants were recrui ted from neurology clinics in academic medical centers that primarily fulfill a tertiary medical care role, participants may possibly not be representative of 87

PAGE 88

patients who receive care from community neurol ogists, for instance. This larger population may also enable recruitment of larger study samples. As discussed previously, this study, like all adherence studies, is also limited by the lack of a fully reliable measure of adherence. While the use of multiple measures of adherence is a strength of this study, the use of only self-report measures a nd no measures of objective adherence such as pill counts (t hrough manual medication audits or electronic monitoring) or blood levels is a potential wea kness, as using multiple classes of adherence measurements may provide the ability to provide a more comprehens ive assessment of adhe rence. Future studies may therefore benefit from combining self-report and objective adherence measurements in a single study protocol. Conclusion The objective of this study was, first, to adap t the finding in a fairly large body of literature that co-morbid depression reduces the rate of adherence to medi cation regimens for chronic or serious illnesses substantially. As an extension of this, the stu dy also sought to demonstrate possible mechanisms whereby depression might a ffect the relatively complex behavior of medication-taking, and how this might lead to observable differences in epilepsy management. While some of the proposed mechanisms were in deed found to be associated with self-reported adherence, particularly the presence of a chance lo cus of control with resp ect to medical health and the presence of specific concerns about ep ilepsy medications, others, such as epilepsy knowledge, had effects contrary to expectations, and others, such as illness severity perception and apathy, had non-significant rela tionships with adherence. However, surprisingly, depression itself was not found to play a role in this process. Indeed, many of the effects seemed to be blunted in their effect in the depressed individuals in the stud y sample. It was hypothesized that people with epilepsy and co-morbid depressi on might make up a high-risk group for which 88

PAGE 89

89 future interventions might be planned targeti ng the areas of healthcar e beliefs and attitudes represented in the study measures in order to he lp them achieve better adherence and thereby possibly even better seizure management. However, there is no evidence to support this approach in the present study. Nonetheless, while the relationship with depression was not found in this study, interventions to address epilepsy patients locus of control as we ll as psychoeducation to address concerns they have over their epilepsy medicatio ns appear, based on the present findings, might be promising areas of intervention for epilepsy pa tients. If further research verifies that the findings related to cognition a nd adherence are generally true for people with epilepsy, these findings also argue for an increased role of assessing cognitive func tions, and memory in particular, for people with epilepsy. These asse ssments, in addition to the traditional role of neuropsychological assessment in assessing epileps ys impact on the brain and also in assisting in localizing seizure foci, might be important even for patients who are not necessarily current surgical candidates for the purpose of identifyi ng individuals who need increased healthcare support, particularly in th e area of medication management and psychoeducation. Future studies should continue to assess the relationship between adherence and depression in people with epilepsy, ideally using a larger sample and measures of adherence from multiple classes, such as the combination of a self -report measure with el ectronic monitoring of medication usage. Such studies may not only have greater ability to shed light on adherence behaviors but would also provide guidance on the clinical and research use of these different measurements of adherence. More research should also be done to determine how difficulties in health locus of control and medication concerns arise in people with epilepsy, in order to better understand how to work with patients to overcome these problems in the clinic.

PAGE 90

LIST OF REFERE NCES Adams, S. J., O'Brien, T. J., Lloyd, J., Kilpatric k, C. J., Salzberg, M.R ., & Velakoulis, D. (2008). Neuropsychiatric morbidity in focal epilepsy.. British Journal of Psychiatry 192, 464-469. Adda, C. C., Castro, L. H., Alm-Mar e Silva, L. C., de Manreza, M. L., & Kashiara, R. (2008). Prospective memory and mesial temporal epile psy associated with hippocampal sclerosis. Neuropsychologia 46, 1954-1964. Asadi-Pooya, A. A. (2005). Drug compliance of children and adolescents with epilepsy. Seizure. 14, 393-395. Baiardini, I., Braido, F., Giardini, A., Majani, G., Cacciola, C., Rogaku, A., Scordamaglia, A., & Canonica, G. W. (2006) Adherence to treatmen t: assessment of an unmet need in asthma. Journal of Investigational Alle rgology & Clinical Immunology 16, 218-223. Barton, C., Clarke, D., Sulaiman, N., & Abra mson, M. (2003). Coping as a mediator of psychosocial impediments to optimal management and control of asthma. Respiratory Medicine, 97, 747-761. Beghi, E., De Maria, G., Gobbi, G., & Veneselli, E. (2006). Diagnosis and Treatment of the First Epileptic Seizure: Guidelines of the Italian League Against Epilepsy. Epilepsia, 46, Suppl 5, 2-8. Berg, A. T. (2006). Defining Intractable Epilepsy. Advances in Neurology 97, 5-10. Bissonnette, J. M. (2008). Adherence: a concept analysis. Journal of advanced nursing 63, 634643. Broadbent, E., Petrie, K. J., Main, J., & We inman, J. (2006). The Br ief Illness Perception Questionnaire. Journal of Psychosomatic Research 60, 631-637. Briesacher, B.A., Andrade, S. E., Fouayzi, H., & Chan, K. A. Comparison of drug adherence rates among patients with seven different medical conditions. Pharmacotherapy 28, 437443. Brunbech, L. & Sabers, A. (2002). Effect of Antiepileptic Drugs on Cognitive Function in Individuals with Epilepsy: A Comparative Review of Newer Versus Older Agents. Drugs 62, 593-604. Buck, D., Jacoby, A., Baker, G. A., & Chadwick, D. W. (1997). Factors Influencing Compliance with Antiepileptic Drug Regimes. Seizure 6, 87-93. Carpay, J. A., Aldenkamp, A. P., & van Donselaa r, C. A. (2005). Complaints Associated With The Use Of Antiepileptic Drugs: Re sults From A Community-Based Study. Seizure 14, 198-206. 90

PAGE 91

Cassano, P. & Fava, M. (2002). Depressi on and Public Health: An Overview. Journal of Psychosomatic Research 53, 849-857. Chengappa, K. N. R., Gershon, S., & Levine, J. (2001). The evolving role of topiram ate among other mood stabilizers in the ma nagement of bipolar disorder. Bipolar Disorders 3, 215232. Cooper, J., Hall, L., Penland, A., Krueger, A., & May, J. (2009). Measuring medication adherence. Population Health Management 12, 25-30. Cramer, J. A. Glassman, M., & Rienzi, V. ( 2002). The Relationship Between Poor Medication Compliance and Seizures. Epilepsy & Behavior 3, 338-342. Cramer, J. A., Blum, D., Reed, M., Fanning, K ., & Epilepsy Impact Project Group. (2003). The influence of comorbid depression on quali ty of life for people with epilepsy. Epilepsy & Behavior, 4, 515-521. Cramer, J. A., Blum, D., Fanning, K., Reed, M ., & Epilepsy Impact Project Group. (2004) The impact of comorbid depression on health res ource utilization in a community sample of people with epilepsy. Epilepsy & Behavior 5, 337-342. Cramer, J. A., Brandenburg, N., & Xu, X. (2005). Differentiating an xiety and depression symptoms in patients with partial epilepsy. Epilepsy & Behavior 6, 563-569. Cramer, J. A., Mattson, R. H., Prevey, M. L., Sche yer, R. D., & Ouellette, V. L. How often is medication taken as prescribed? A novel assessment technique. Journal of the American Medical Association 261, 3273-3277. Cramer, J. A., Scheyer, R. D., & Mattson, R. H. (1990). Compliance declines between clinic visits. Archives of Internal Medicine, 150, 1509-1510. Curtis, J. R., Xi, J., Westfall, A. O., Cheng, H ., Lyles, K., Saag, K. G., & Delzell, E. (2009). Improving the prediction of medication comp liance: the example of bisphosphonates for osteoporosis. Medical Care 47, 334-341. Dalessandro, M., Conti, C. M., Gambi, F., Fala sca, K., Doyle, R., C onti, P., Caciagli, F., Fulcheri, M., & Vecchiet, J. (2007) Antidepressant therapy can improve adherence to antiretroviral regimens among HIV-in fected and depressed patients. Journal of Clinical Psychopharmacology, 27, 58-61. DiIorio, C., Osborne Shafer, P., Letz, R., Henry, T., Schomer, D. L., & Yeager, K. (2003). The association of stigma with self-management and perceptions of health care among adults with epilepsy. Epilepsy & Behavior 4, 259-267. DiMatteo, M. R., Lepper, H. S., & Croghan, T. W. (2000). Depression Is a Risk Factor for Noncompliance With Medical Treatment. Archives of Internal Medicine, 160, 2101-2107. Dodrill, C. B. (2004). Neuropsychological effects of seizures. Epilepsy & Behavior 5, S21-24. 91

PAGE 92

Dunbar, J. (1984). Adherence m easures and thei r utility. Controlled C linical Trials, 5, 515-521. Duncan, J. S., Sander, J. W., Sisodiya Walker, M. C. (2006). Adult epilepsy. The Lancet, 367, 1087-1100. Drevets, W. C. (2001). Neuroimaging and neuropat hological studies of depression: implications for the cognitive-emotional f eatures of mood disorders. Current Opinion in Neurobiology 11, 240-249. Edeh, J. & Toone, B. (1987). Relationship betw een interictal psychopath ology and the type of epilepsy. Results of a survey in general practice. British Journal of Psychiatry 151, 95101. Enriquez-Caceres, M. & Soto-Santillana, M. (1998). Non-compliance with pharmacological treatment in patients with epilepsy. Revista Neurologia ., 42, 647-654. Ettinger, A. B., Manjunath, R., Candrilli, S. D., & Davis, K. L. (2009) Prevalence and cost of nonadherence to antiepileptic drugs in elderly patients with epilepsy. Epilepsy & Behavior 14, 324-329. Ettinger, A., Reed, M., Cramer, J., & Epileps y Impact Project Group. (2004). Depression and comorbidity in community-based patients with epilepsy or asthma. Neurology, 28, 10081014. Evans, D. L., et al (2005). Mood Disorders in the Medi cally Ill: Scientif ic Review and Recommendations. Biological Psychiatry 58, 175-189. Faught, E., Duh, M. S., Weiner, J. R., Gurin, A., & Cunnington, M. C. (2008). Nonadherence to antiepileptic drugs and increased mort ality: findings from the RANSOM Study. Neurology, 71, 1572-1578. Faught, R. E., Weiner, J. R., Gurin, A., Cunni ngton, M. C., & Duh, M. S. (2009). Impact of nonadherence to antiepileptic drugs on health ca re utilization and costs: findings from the RANSOM study. Epilepsia 50, 501-509. Fisher, J. D., Fisher, W. A., Amico, K. R., & Harman, J. J. (2006). An InformationMotivation Behavioral Skills Model of Adhe rence to Antiretroviral Therapy. Health Psychology 25, 462-473. Fleiss, J. L., Tytun, A., & Ury, H. K. (1980). A Simple Approximation For Calculating Sample Sizes For Comparing Independent Proportions. Biometrics 36, 343. French, J. A. (2007) Refractory Epilepsy: Clinical Overview. Epilepsia 48, Suppl. 1, 3-7. Gaitatzis, A., Trimble, M. R., & Sander, J.W. (2004). The psychiatric comorbidity of epilepsy. Acta Neurologica Scandinavica 110, 207-220. 92

PAGE 93

Gao, X. & Nau, D. P. (2000). Congruence of Three Self-Report Measures of Medication Adherence Am ong HIV Patients. Annals of Pharmacotherapy, 34, 1117-1122. Garca-Morales, I., de la Pea Mayor, P., Kanner, A. M. (2008). Psychiatric comorbidities in epilepsy: identification and treatment. The Neurologist, 14, S15-25.. Garg, S. K., Kumar, N., Bhargava, V. K., & Prabha kar, S. K. (1998). Effect of grapefruit juice on carbamazepine bioavailability in patients with epilepsy. Clinical Pharmacology and Therapeutics, 64, 286-288. Glauser, T. A. & Pippenger, C. E. (2000). Cont roversies in blood-level monitoring: reexamining its role in the treatment of epilepsy. Epilepsia 41, Supp. 8, S6-15. Golden, C. J. (1978). Stroop Color and Word Test Chicago: Stoelting. Goldstein, L. H., Holland, L., Soteriou, H., & Melle rs, J. D. C. (2005). Illness representations, coping styles and mood in adults with epilepsy. Epilepsy Research 67, 1-11. Goldstein, D. B., Need, A. C., Singh, R., Sisodiya, S. M. (2007). Potential genetic causes of heterogeneity of treatment effects. American Journal of Medicine 120, S21-25. Gomes Mda, M., Maia Filho Hde, S., & No, R. A. (1998). Anti-epileptic drug intake adherence. The value of the blood drug level measurement and the clinical approach. Arquivos de Neuro-Psiquiatria 56, 708-713. Gould, O. N., McDonald-Miszczak, L., & Ki ng, B. (1997) Metacognition and medication adherence: how do older adults remember? Experimental Aging Research 23, 315-342. Gramstad, A., Iversen, E., & Engelsen, B. A. (2001) The impact of affectivity dispositions, selfefficacy and locus of control on psychosocial adjustment in patients with epilepsy. Epilepsy Research 46, 53-61. Grosset, K. A., Bone, I., & Grosset, D.G. (2005). Suboptimal medication adherence in Parkinson's disease. Movement Disorders 20, 1502-1507. Gurege, O. (1991). Interictal ps ychopathology in epilepsy. Prevalence and pattern in a Nigerian clinic. British Journal of Psychiatry 158, 700-705. Halm, E. A., Mora, P., & Leventhal, H. (2006). No Symptoms, No Asth ma: The Acute Episodic Disease Belief Is Associated With Poor Se lf-Management Among Inne r-City Adults With Persistent Asthma. Chest 129, 573-580. Hama, S., Yamashita, H., Shigenobu, M., Watanabe A., Kurisu, K., Yamawaki, S., & Kitaoka, T. (2006). Post-Stroke Affective Or Apathetic Depression And Lesion Location: Left Frontal Lobe And Bilateral Basal Ganglia. European Archives of Clinical Neuroscience, Nov 25, 2006 (Epub ahead of print). 93

PAGE 94

Harden, C. L. (2002). The Co-m orbidity Of Depression And Epilepsy: Epidemiology, Etiology, And Treatment. Neurology 59, Suppl. 4, S48-S55. Haut, S. R., Shinnar, S., & Mosh, S. L. (2005). Seizure clustering: risks and outcomes. Epilepsia 46, 146-149. Hazzard, A., Hutchinson, S. J., & Krawiecki, N. (1990). Factors related to adherence to medication regimens in pediatric seizure patients. The Journal of Pediatric Psychology 15, 543-55. Hermann, B. P., Seidenberg, M., Dow, C., Jones, J., Rutecki, P., Bhattacharya, A., & Bell, B. (2006). Cognitive prognosis in chronic temporal lobe epilepsy. Annals of Neurology 60, 80-87. Horne, R. & Weinman, J. (1999). Patients' beliefs about prescribed medicines and their role in adherence to treatment in chronic physical illness. Journal of Psychosomatic Research 47, 555-567. Horne, R. (2006). Compliance, Adherence, and Concordance: Implications for Asthma Treatment. Chest 130, 65S-72S. Hovinga, C. A., Asato, M. R., Manjunath, R., Wheless, J. W., Phelps, S. J., Sheth, R. D., PinaGarza, J. E., Zingaro, W. M., & Haskins, L. S. (2008). Association of non-adherence to antiepileptic drugs and seizures, quality of lif e, and productivity: survey of patients with epilepsy and physicians. Epilepsy & Behavior 13, 316-322. Hughes, D., Cowell, W., Koncz, T., Cramer, J. (2007). Methods for integrating medication compliance and persistence in pharmacoeconomic evaluations. Value Health 10, 498-509. Jerant, A., DiMatteo, R., Arnsten, J., Moore-Hill, M., & Franks, P. (2008). Self-report adherence measures in chronic illness: retest reliability and predictive validity. Medical Care 46, 1134-1139. Johnson, E. K., Jones, J. E., Seidenberg, M., & Hermann, B. P. (2004). The Relative Impact of Anxiety, Depression, and Clinical Seizure Features on Health Related Quality of Life in Epilepsy. Epilepsia, 45, 544-550. Johannessen, S. I. & Landmark, C. J. (2008). Value of therapeutic drug monitoring in epilepsy. Experimental Review of Neurotherapeutics 8, 929-939. Johnson, M. O., Elliott, T. R., Neilands, T. B., Morin, S. F., & Chesney, M. A. (2006). A Social Problem-Solving Model of Adhe rence to HIV Medications. Health Psychology, 25, 355363. Jones, J. E., Hermann, B. P., Woodard, J. L., Barr y, J. J., Gillliam, F., Kanner, A. M., & Meador, K. J. (2005). Screening for Major Depre ssion in Epilepsy with Common Self-report Depression Inventories. Epilepsia 46, 731-735. 94

PAGE 95

Jones, R. M., Butler, J. A ., Thomas, V. A., Peveler, R. C., & Prevett, M. (2006). Adherence to treatment in patients with epilepsy: Associat ions with seizure control and illness beliefs. Seizure, 15, 504-508. Kanner, A. M. & Balabanov, A. (2002). Depres sion and Epilepsy: How Closely Related Are They? Neurology 58, Suppl. 5, S27-S39. Kanner, A. M. (2003). Depressi on in Epilepsy: Prevalence, Clinical Semiology, Pathogenic Mechanisms, and Treatment. Biological Psychiatry, 54, 388-398. Kanner, A. M. (2008). Depression in epileps y: a complex relation with unexpected consequences. Current Opinions in Neurology 21, 190-194. Karve, S., Cleves, M. A., Helm, M., Hudson, T. J., West, D. S., & Martin, B. C. (2008). An empirical basis for standardizing adherence m easures derived from administrative claims data among diabetic patients. Medical Care 46, 1125-1133. Kemp, S., Feely, M., Hay, A., Wild, H., & Cooper, C. (2007). Psychological factors and use of antiepileptic drugs: pilot work using an objective measure of adherence. Psychology, Health, & Medicine 12, 107-113. Kennedy, S. H., Javanmard, M., & Vaccarino, F. J. (1997). A review of functional neuroimaging in mood disorders: positron emission tomography and depression. Canadian Journal of Psychiatry, 42, 467-475. Krishnamoorthy, E. S. (2003). Treatment of depr ession in patients with epilepsy: problems, pitfalls, and some solutions. Epilepsy & Behavior 4, S46-54. Kyngs, H. (2001). Predictors of good compliance in adolescents with epilepsy. Seizure 10, 549553. Lee, W. C., Arcona, S., Thomas, S. K., Wang, Q., Hoffmann, M. S., & Pashos, C. L. (2005). Effect of comorbidities on medical care use and cost among refractory patients with partial seizure disorder. Epilepsy & Behavior 7, 123-126. Lehane, E. & McCarthy, G. (2009). Medication non-adherence--exploring the conceptual mire. International Journal of Nursing Practice 15, 25-31. Litvan, I., Cummings, J. L., & Mega, M. (1998) Neuropsychiatric featur es of corticobasal degeneration. Journal of Neurology, Neurosurgery, & Psychiatry 65, 717-721. Long, L., Reeves, A. L., Moore, J. L., Roach, J., & Pickering, C. T. (2000). An Assessment of Epilepsy Patients Knowle dge of Their Disorder. Epilepsia 41, 727-731. Lpez-Gngora, M., Martnez-Domeo, A., Garca, C., Escartn, A. (2008). Effect of levetiracetam on cognitive functions and qua lity of life: a one-year follow-up study. Epileptic Disorders, 10, 297-305. 95

PAGE 96

Loring, D. W. & Meador, K. J. (2001). The evolving role of topiram ate among other mood stabilizers in the manageme nt of bipolar disorder. Epilepsia 42 (S8), 24-32. Lugo Gonzalez, M.E., Santos Ortiz, M. C., Davila Torres R. R., Rios Motta, M. (2001). Aspects Related To Drug Adherence In A Group Of Care givers Of Children And Adolescents With Epilepsy. Puerto Rico Health Sciences Journal 20, 257-267. Lusic, I., Titlic, M., Eterovic, D. (2005). Epile ptic Patient Compliance With Prescribed Medical Treatment. Acta Medica Croatia 59, 13-18. Magai, C., Consedine, N., Neugut, A. I., & Hershman, D. L. (2007). Common psychosocial factors underlying breast cancer screening a nd breast cancer treatment adherence: a conceptual review and synthesis. Journal of Womens Health 16, 11-23. Manjunath, R., Davis, K. L., Candrilli, S. D., & Ettinger, A. B. (2009). Association of antiepileptic drug nonadherence with risk of seizures in a dults with epilepsy. Epilepsy & Behavior, 14, 372-378. Marin, R. S., Biedrzycki, R. C. & Firinciogulla ari, S. (1991). Reliabili ty and validity of the Apathy Evaluation Scale. Psychiatry Research 38, 143-162. Marinker, M. (ed.) Royal Pharmaceutical Society of Great Britain. (1997). From compliance to concordance: towards shared goals in medicine taking London: RPS, 1997. Martins, S. P. & Damasceno, B. P. (2008). Pr ospective and retrospec tive memory in mild Alzheimer's disease. Arquivos de Neuro-Psiquiatria 66, 318-322. Mattson, M. E., Friedman, L. M. (1984). Issues in medication adherence a ssessment in clinical trials of the National Hear t, Lung, and Blood Institute. Controlled Clinical Trials 5, 488496. May, T. W. & Pffflin, M. (2002). The Efficacy of an Educational Treatment Program for Patients with Epilepsy (MOSES): Resu lts of a Controlled, Randomized Study. Epilepsia 43, 539-549. McAuley, J. W., McFadden, L. S., Elliott, J. O., & Shneker, B. F. (2008). An evaluation of selfmanagement behaviors and medication a dherence in patients with epilepsy. Epilepsy & Behavior, 13, 637-641. Meador, K. J., Loring, D. W., Hulihan, J. F., Karmin, M. & Karim, R. (2003). Differential Cognitive And Behavioral Effects Of Topiramate And Valproate. Neurology 60, 14831488. Mitchell, W. G., Scheier, L. M., & Baker, S.A. (2000). Adherence To Treatment In Children With Epilepsy: Who Follows "Doctor's Orders"? Epilepsia 41, 1616-1625. Modi, A. C., Morita, D. A., Glauser, T. A. (2008). One-month adherence in children with newonset epilepsy: white-coat compliance does not occur. Pediatrics 121, e961-966. 96

PAGE 97

Mohr, D. C., Goodkin, D. E., Likosky, W ., Gatto, N ., Neilley, L. K., Griffin, C., Stiebling, B. (1996). Therapeutic expectations of patie nts with multiple sclerosis upon initiating interferon beta-1b: relationshi p to adherence to treatment. Multiple Sclerosis 2, 222-226. Morisky, D. E., Ang, A., Krousel-Wood, M., & Ward, H. J. (2008). Predictive validity of a medication adherence measure in an outpatient setting. Journal of Clinical Hypertension 10, 348-354. Morisky, D. E., Green, L. W., & Levine, D. M. ( 1986). Concurrent and pred ictive validity of a self-reported measure of medication adherence. Medical Care 24, 67-74. Mula, M., Schmitz, B., Sander, J. W. (2008). The pharmacological treatment of depression in adults with epilepsy. Expert Opinion on Pharmacotherapy 9, 3159-3168. Mller, T., Fuchs, G., Hahne, M., Klein, W., & Schwarz M. (2006). Diagnostic aspects of early Parkinson's disease. Journal of Neurology 253, IV29-31. Naar-King, S., Podolski, C.-L., Ellis, D. A., Frey, M. A., & Templin, T. (2006). Social Ecological Model of Illness Management in High-Risk Youths With Type 1 Diabetes. Journal of Consulting and Clinical Psychology, 74, 785-789. Payakachat, N., Summers, K. H ., & Barbuto, J. P. (2006). A Co mparison of Clinical Practice Guidelines in the Initial Pharmacological Ma nagement of New-Onset Epilepsy in Adults. Journal of Managed Care Pharmacy 12, 55-60. Rabkin, J. G., Ferrando, S. J., van Gorb, W., Ri eppi, R., McElhiney, M., & Sewell, M. (2000). Relationships Among Apathy, Depression, and Cognitive Impairment in HIV / AIDS. The Journal of Neuropsychiatry and Clinical Neurosciences 12, 451-457. Radloff, L. S. The CES-D Scale: A Self-Report Depression Scale for Research in the General Population. Journal of Applied Psychological Measurement 1, 385-401. Ramaratnam, S., Baker, G. A., & Goldstei n, L.H. (2005). Psychological Treatments For Epilepsy. Cochrane Database of Systematic Reviews 19. Reisberg, B., Prichep, L., Mosconi, L., John, E. R., Glodzik-Sobanska, L., Boksay, I., Monteiro, I., Torossian, C., Vedvyas, A., Ashraf, N., Jamil, I. A., de Leon, M. J. (2008). The pre-mild cognitive impairment, subjective cognitive impairment stage of Alzheimer's disease. Alzheimers & Dementia 4, S98-S101. Remien, R. H., Hirky, A. E., Johnson, M. O., Wei nhardt, L. S., Whittier, D., & Le, G. M. (2003). Adherence to Medication Treatment: A Qualita tive Study of Facil itators and Barriers Among a Diverse Sample of HIV+ Men and Women in Four U.S. Cities. AIDS and Behavior, 7, 61-72. Resnick, B., Zimmerman, S. I., Magaziner, J., & Adelman, A. (1998). Use of the Apathy Evaluation Scale as a measure of motivation in elderly people. Rehabilitation Nursing 23, 141-147. 97

PAGE 98

Rey, A. (1964). Lexam en clinique en psychologie Paris: Presses Universitaires de France. Ried, S., Specht, U., Thorbecke, R., Goecke, K., & Wohlfarth, R. (2001). MOSES: An Educational Program for Patients wi th Epilepsy and Their Relatives. Epilepsia 42, Suppl. 3, 76-80. Rodin, E. A. (1978). Various prognostic aspects in patients with epilepsy. Folia Psychiatrica Et Neurologica Japonica 32, 408-418. Sacco, W. P., Wells, K. J., Vaughan, C. A., Friedman, A., Perez, S., & Matthew, R. (2005). Depression in adults with type 2 diabetes: th e role of adherence, body mass index, and selfefficacy. Health Psychology, 24, 630-634. Safren, S. A., Otto, M. W., Worth, J. L., Salo mon, E., Johnson, W., Mayer, K., & Boswell, S. (2001). Two strategies to incr ease adherence to HIV antiret roviral medication: life-steps and medication monitoring. Behaviour Research and Therapy 39, 1151-1162. Sander, J. W. (2004). The Use Of Antiepile ptic Drugs Principles And Practice. Epilepsia, 45, Suppl. 6, 28-34. Savage, R. M. & Gouvier, W. D. (1992). Rey Aud itory-Verbal Learning Te st: the effects of age and gender, and norms for delayed recall and story recognition trials. Archives of Clinical Neuropsychology 7, 407-414. Shakya, G., Malla, S., Shakya, K. N., & Shre stha, R. (2008). Therapeutic drug monitoring of antiepileptic drugs. Journal of the Nepal Medical Association 47, 94-97. Sheehan, D. V., Lecrubier, Y., Sheehan, K. H., Amor im, P., Janavs, J., Weiller, E,, Hergueta, T., Baker, R., & Dunbar, G. C. (1998). The Mi ni-International Neurops ychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. Journal of Clin ical Psychiatry 59, Suppl. 20, 22-33. Shipley, W. C. (1940). A self-administered scal e for measuring intelle ctual impairment and deterioration. Journal of Psychology 9, 371-377. Snodgrass, S. R., Vedanarayanan, V. V., Parker, C. C., & Parks, B. R. (2001). Pediatric Patients With Undetectable Anticonvulsant Blood Leve ls: Comparison With Compliant Patients. Journal of Child Neurology 16, 164-168. Spector, S., Cull, C., & Goldstein, L. H. ( 2001). High and Low Perceived Self-Control of Epileptic Seizures. Epilepsia 42, 556-564. Swinkels, W. A., van Emde Boas, W., Kuyk, J., van Dyck, R., & Spinhoven, P. (2006). Interictal depression, anxiety, personality traits, and psychological dissociation in patients with temporal lobe epilepsy (TLE) and extra-TLE. Epilepsia 47, 2092-2103. Thapar A., Roland M., & Harold G. (2005). Do depression symptoms predict seizure frequency-or vice versa? Journal of Psychosomatic Research 59, 269-74. 98

PAGE 99

Torta, R. & Keller, R. (1999). Behavioral, psychotic, and anxiety di sorders in epil epsy: etiology, clinical features, and therapeutic im plications. Epilepsia, 40, S2-20. Trost, L. F., Wender, R. C., Suter, C. C., Rosenbe rg, J. H., Brixner, D. I., Von Worley, A. M., & Gunter, M. J. (2005). Management of ep ilepsy in adults: Di agnosis guidelines. Postgraduate Medicine 118, 22-26. Trost, L. F., Wender, R. C., Suter, C. C., Rosenbe rg, J. H., Brixner, D. I., Von Worley, A. M., & Gunter, M. J. (2005). Management of ep ilepsy in adults: Treatment guidelines. Postgraduate Medicine 118, 29-33. Ubbiali, A., Donati, D., Chiorri, C., Bregani, V., Cattaneo, E., Maffe i, C., & Visintini, R. (2008). The usefulness of the Multidimensional Health Locus of Control Form C (MHLC-C) for HIV+ subjects: an Italian study. AIDS Care 20, 495-502. Vaur, L., Vaisse, B., Genes, N., Elkik, F., Le grand, C., Poggi, L. (1999). Use of electronic pill boxes to assess risk of poor treatment comp liance: results of a large-scale trial. American Journal of Hypertension 12, 374-380. Vermeire, E., Hearnshaw, H., Van Royen, P., & Denekens, J. (2001). Patient adherence to treatment: three decades of res earch. A comprehensive review. Journal of Clinical Pharmacy and Therapeutics 26, 331-342. Vitolins, M. Z., Rand, C. S., Ra pp, S. R., Ribisl, P. M., & Se vick, M. A. (2000). Measuring adherence to behavioral a nd medical interventions. Controlled Clinical Trials 21, 188S194S. Wagner, R. F. (2003). Clinical Case Formul ation in the Research Program Subjective Theories: Patients with Chronic Diseases. European Journal of Psychological Assessment 19, 185-194. Wallston, K.A., Wallston B.S., & DeVellis, R. (1978). Development of the Multidimensional Health Locus of Control (MHLC) Scales. Health Education Monographs 6, 160-170. Walters, R. J., Hutchings, A. D., Smith, D. F., & Smith, P. E. (2004). In appropriate requests for serum anti-epileptic drug levels in hospital practice. QJM 97, 337-341. Wang, L. & Li, J. (2003). Role of educational intervention in the management of comorbid depression and hypertension. Blood Pressure 12, 198-202. Wechsler, D. (1997). Wechsler Adult Intelligence Scale III San Antonio: The Psychological Corporation. Winawer, M. R. (2006). Phenotype Definition of Epilepsy. Epilepsy & Behavior 8, 462-476. Wing, R. R., Phelan, S., & Tate, D. (2002). The Role of Adherence in Mediating the Relationship Between Depression and Health Outcomes. Journal of Psychosomatic Research 53, 877-881. 99

PAGE 100

100 Yeager, K.A., DiIorio, C., Shafer, P.O., McCarty, F., Letz, R., Henry, T., & Schomer, D.L. (2005). The Complexity Of Treatme nts For Persons With Epilepsy. Epilepsy & Behavior 7, 679-686. Zeber, J. E., Copeland, L. A., Amuan, M., Cram er, J. A., & Pugh, M. J. (2007). The role of comorbid psychiatric conditions in health status in epilepsy. Epilepsy & Behavior 10, 539546.

PAGE 101

BIOGR APHICAL SKETCH Mohan Krishnan graduated from the University of Michigan with a Bachelor of Science in engineering physics, in 1997, and a Master of Science in nuclear engineering and radiological sciences, in 1999. He then spent approximately 5 years working in various engineering and business roles within the automotive industry. Du ring this time, he pursued coursework in psychology at Wayne State Universi ty, and participated in res earch studying the relationship between cardiovascular disease and depression in the elderly at the Wayne State University Institute of Gerontology. He obtained a Master of Science in clinical psychology from the University of Florida in 2006. His masters thesis, submitted in partial fulfillment of requirements for that degree, was entitled R elationships Between Medication Levels And Depressive Symptoms In Older I ndividuals. Mr. Krishnan completed clinical training at Shands at the University of Florida, Shands Jack sonville, and with the Veterans Administration. Currently, Mr. Krishnan is in the process of completing his internship in clinical neuropsychology at the University of Chicago Medical Center. Mr. Krishnan is working toward a doctorate in clinical and h ealth psychology, with a specializa tion in clinical neuropsychology, at the University of Florida. 101