|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 UTILITY OF CURRENT S URVEILLANCE SYSTEMS TO DETECT RESPIRATOR Y SYNCYTIAL VIRUS SEAS ONS AND IMPLICATIONS FOR IMMUNOPROPHYLAXI S By CHRISTIAN HAMPP A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009
2 2009 by Christian Hampp
3 To my family: my sisters Brbel and Ute, my mother Else, and to the memory of my father Gottfr ied
4 ACKNOWLEDG E MENTS I want to thank my adviser, Almut Winterstein, for her continuous support and guidance. She has been an outstanding mentor and has never stopped challenging me. I would also like to thank my supervisory committ ee members Nabih Asal, Teresa Kauf, Earlene Lipowski and Eileen Schneider for their expertise, advice, and encouragement. My gratitude goes to all faculty members and staff in the department of Pharmaceutical Outcomes and Policy for their support and for e verything they taught me during my time at the University of Florida. Abraham Hartzema, Carole Kimberlin and Richard Segal were great sources of inspiration and always willing to provide their advice. Many thanks also to Jon Shuster for advice on statistic al analyses. support and advice with data programming. For the provision of data, I thank the staff at Public Health Statistics, Office of Vital Statistics and at the Bur eau of Epidemiology, both within the Florida Department of Health. I appreciate access to Medicaid data from the Centers for Medicare and Medicaid Ser vices and the help of Gerrie Ba ros s o at the Research Data Assistance Center to f acilitate data access. Cathy Panozzo at the Centers for Disease Control and Prevention deserves my gratitude for technical assistance and the provision of their surveillance dataset. rida Agency for Healthcare Administration, AHCA. It was conducted in collaboration with the University of Florida Center for Medicaid and the Uninsured. I want to thank my family for their support, guidance and encouragement throughout my life and my girl friend Hee Jung for her love and emotional support Finally, I thank my fellow graduate students and my friends for their friendship and the right amount of distraction.
5 TABLE OF CONTENTS page ACKNOWLEDGEMENTS ................................ ................................ ................................ ............. 4 LIST OF TABLES ................................ ................................ ................................ ........................... 7 LIST OF FIGURES ................................ ................................ ................................ ......................... 8 ABSTRACT ................................ ................................ ................................ ................................ ... 10 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .................. 12 Background ................................ ................................ ................................ ............................. 12 Need for Study ................................ ................................ ................................ ................. 13 Purpose of Study ................................ ................................ ................................ .............. 14 Research Questions and Hypotheses ................................ ................................ ...................... 15 2 LITERATURE REVIEW ................................ ................................ ................................ ....... 20 Respiratory Syncytial Virus ................................ ................................ ................................ .... 20 RSV Disease Epidemiology ................................ ................................ ............................ 21 RSV Infections ................................ ................................ ................................ ................ 22 Diagnostic Tests for RSV ................................ ................................ ................................ 22 The National Respiratory and Enteric Virus Surveillance System ................................ ......... 24 RSV Seasonality ................................ ................................ ................................ ..................... 25 RSV Prevention ................................ ................................ ................................ ...................... 26 RSV Risk Factors and Indications for Immunoprophylaxis ................................ ............ 28 Prior Authorization Requirements ................................ ................................ ................... 29 3 METHODS ................................ ................................ ................................ ............................. 33 Datasets ................................ ................................ ................................ ................................ ... 33 NREVSS ................................ ................................ ................................ .......................... 33 Florida Department of Health RSV Surveillance Data ................................ ................... 33 Medicai d Analytic eXtract Claims Dataset ................................ ................................ ..... 34 State Birth Certificates ................................ ................................ ................................ .... 35 Study Population ................................ ................................ ................................ ..................... 36 ................................ ............... 39 Part II: RSV Epidemiology between Four US States and Five Regions in Florida .............. 43 Part III: Latitude as a Factor in RSV Epidemiology in Florida ................................ ............. 44 Part IV: Patient Factors and Seasonality ................................ ................................ ............... 45 Part V: Timing of Prophylaxis with Palivizumab vs. RSV Seasonality ................................ 46 Part VI: Optimizing Timing of Prophylaxis ................................ ................................ .......... 47
6 4 RESULTS ................................ ................................ ................................ ............................... 51 Sample Characteristics ................................ ................................ ................................ ............ 51 ................................ ............... 53 Part II: RSV Epidemiology between Four US States and Five Regions in Florida .............. 56 Part III: Latitude as a Factor in RSV Epidemiology in Florida ................................ ............. 57 Part IV: Patient Factors and Seasonality ................................ ................................ ............... 58 Part V: Timing of Prophylaxis with Palivizumab vs. RSV Seasonality ................................ 59 Part VI: Optimizing Timing of Prophylaxis ................................ ................................ .......... 60 5 DISCUSSION ................................ ................................ ................................ ......................... 91 Part I: Vali ................................ ............... 91 Part II: RSV Epidemiology between Four US States and Five Regions in Florida .............. 93 Part III: Latitude as a Factor in RSV Epidemiology in Florida ................................ ............. 94 Part IV: Patient Factors and Seasonality ................................ ................................ ............... 94 Part V: Timing of Prophylaxis with Palivizumab vs. RSV Seasonality ................................ 95 Part VI: Optimizing Timing of Prophylaxis ................................ ................................ .......... 96 External Validity ................................ ................................ ................................ ..................... 98 Study Limitations ................................ ................................ ................................ .................. 100 Future Research ................................ ................................ ................................ .................... 102 Summary and Conclusions ................................ ................................ ................................ ... 103 APPENDIX ................................ ................................ ................................ ................................ .. 106 A. Operational Definitions ................................ ................................ .............................. 106 Palivizumab Exposur e ................................ ................................ ................................ ... 106 Risk Factors for RSV ................................ ................................ ................................ ..... 106 Chronic lung disease ................................ ................................ .............................. 106 Prematur ity ................................ ................................ ................................ ............. 107 Congenital heart disease ................................ ................................ ......................... 107 Cystic fibrosis ................................ ................................ ................................ ......... 108 Severe combined or acquired immunodeficiency ................................ .................. 108 Down syndrome ................................ ................................ ................................ ..... 108 Asthma ................................ ................................ ................................ ................... 108 Transplant ................................ ................................ ................................ ............... 109 Malignancy ................................ ................................ ................................ ............. 109 Immunosuppression or antineoplastic agents ................................ ......................... 109 Hospitalizations ................................ ................................ ................................ ............. 109 RSV hospitalization ................................ ................................ ................................ 109 Specific non RSV bronchiolitis or pneumonia ................................ ...................... 110 Unspecific bronchiolitis or pneumonia ................................ ................................ .. 110 B. Supplemental Tables ................................ ................................ ................................ .. 111 LIST OF RE FERENCES ................................ ................................ ................................ ............. 113 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 121
7 LIST OF TABLES Table page 2 1 Historical landmarks in the NREVSS ................................ ................................ .................... 31 4 1 Cohort characteristics ................................ ................................ ................................ ............. 62 4 2 Risk factors for RSV hospitalizations in Florida ................................ ................................ .... 63 4 3 Palivizumab exposure and RSV hospitalizations by state and risk category ......................... 64 4 4 Areas under the curve by state and region ................................ ................................ .............. 65 4 5 Test characteristics at the threshold of 10% median proportion positive laboratory tests ..... 66 4 6 Test characteristics at optimal thresholds of median proportion positive laboratory tests ..... 66 4 7 Mean of absolute differences and direction of difference between season onset according to clinical dataset and surveillance dataset under different definitions for season onset ................................ ................................ ................................ ....................... 67 4 8 Mean of absolute differences and direction of difference between season offset according to clinical dataset and surveillance dataset under different definitions for seas on offset ................................ ................................ ................................ ....................... 68 4 9 Extent of seasonality and seasonality index in each state and regions in Florida .................. 69 4 10 Variation in seasons withi n each state and regions in Florida ................................ .............. 70 4 11 Comparison of season characteristics between regions in Florida ................................ ....... 71 4 12 Linear regressio n analysis of the effects of latitude on season characteristics in Florida .... 71 4 13 Mean of absolute differences and direction of difference between onset of palivizumab utilization and onset of RSV season according to different determinants of RSV season ................................ ................................ ................................ ................................ 72 4 14 Mean of absolute differences and direction of difference between offset of palivizumab utilization and offset of RSV season a ccording to different determinants of RSV season ................................ ................................ ................................ .................... 73 5 1 Cost of prophylaxis per avoided RSV hospitalization ................................ .......................... 105 B 1 List of counties in Florida ................................ ................................ ................................ .... 111 B 2 Coordinates of Florida regions ................................ ................................ ............................. 112 B 3 Week numbers and corresponding calendar months, shown for the year 2000. .................. 112
8 LIST OF FIGURES Figure page 2 1 Map of RSV regions in Florida ................................ ................................ .............................. 32 3 1 Season detection based on clinical dataset ................................ ................................ ............. 49 3 3 Cut off values for areas under the ROC curve ................................ ................................ ....... 50 4 1 Flowchart of sample selection and resulting sample size ................................ ....................... 74 4 2 RSV hospitalization rates and resulting seasons in A) California, B) Florida, C) Illinois and D) Texas ................................ ................................ ................................ ...................... 75 4 3 RSV hospitalization rates and resulting seasons in the regions of Florida. A) Northwest, B) North, C) Central, D) Southwest and E) Southeast ................................ ...................... 76 4 4 NREVSS laboratory tests and resulting RSV season in A) California, B) Fl orida, C) Illinois and D) Texas. ................................ ................................ ................................ ... 78 4 5 NREVSS laboratory tests and resulting RSV season in the regions of Florida. A) Northwest, B) North, C) Central, D) Southwest and E) Southeast. .............................. 79 4 6 Receiver operating characteristics curves for each state ................................ ........................ 81 4 7 Receiver operating characteristics curves for each region in Florida ................................ ..... 82 4 8 Linear effect of latitude on season characteristics in Florida. A) Week of season onset, B ) Week of season offset, C) Season duration and D) Peak week ................................ .... 83 4 9 Heterogeneous effects of RSV risk factors on risk for RSV hospitalizations off season vs. on season ................................ ................................ ................................ ...................... 84 4 10 Palivizumab utilization and RSV seasonality in A) Calif ornia, B) Florid a, C) Illinois and D) Texas ................................ ................................ ................................ ...................... 85 4 11 Palivizumab utilization and RSV seasonality in the region s of Florida. A) Northwest, B) North, C) Central, D) Southwest and E) Southeast ................................ ...................... 86 4 12 A) Monthly RSV hospitalization incidence rates B) N umbers needed to treat with palivizumab by age in the high risk cohort in California ................................ ................... 88 4 13 A) Monthly RSV hospitalization incidence rates B) N umbers needed to treat with palivizumab by age in the high risk cohort in Florida ................................ ....................... 88 4 14 A) Monthly RSV hospitalization incidence rates B) N umbers needed to treat with palivizumab by age in the high risk cohort in Illinois ................................ ....................... 89
9 4 15 A) Monthly RSV hospitalization incidence rates B) N umber s needed to treat with palivizumab by age in the high risk cohort in Texas ................................ ......................... 89 4 16 A) Monthly RSV hospitalization incide nce rates B) N umbers needed to treat with palivizumab in the high risk cohort for each region in Florida ................................ ......... 90 5 1 Distribution of diagnostic codes for bronchiolitis and pneumonia related hospitalizations ................................ ................................ ................................ ................ 105
10 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy UTILITY OF CURRENT S URVEILLANCE SYSTEMS TO DETECT RESPIRATOR Y SYNCYTIAL VIRUS SEAS ONS AND IMPLICATIONS FO R IMMUNOPROPHYLAXIS By Christian Hampp August 2009 Chair: Almut Winterstein Major: Pharmaceutical Sciences To inform timing of immunoprophylaxis for respiratory syncytial virus (RSV) the Centers for Disease Control and Prevention (CDC) monitors the w eekly number s and results of RSV tests from a sample of lab oratories Our objective was to validate use of a 10% threshold of median proportion of positive tests (MPP) to identify RSV seasons. Additionally to help policy makers optimize resource s, we provide monthly RSV incidence s and numbers needed to treat ( NNT ) with palivizumab. Medicaid fee for servi ce recipients under 2 years of age from California, Florida, Illinois and Texas (1999 2004) were categorized for each week as high risk or low ris k for RSV infection based on ICD 9 codes, pharmacy c laims and birth certificates (F lorida only). Subjects were continuously eligible from birth and in ambulatory care for 4 weeks before the current week. The statewide weekly incidence rate s of RSV hospital izations w ere measured for each risk category and adjusted for RSV prophylaxis. Weeks were categorized as on season if the RSV incidence rate in high risk children exceeded the season peak of the incidence rate in low risk children. Receiver operating char acteristics (ROC) curves were used to measure the ability of MPP to discriminate between on season and off season weeks.
11 In California, Florida, Illinois and Texas the area s unde r the ROC curve s were 0.98 (95% confidence interval, 0.96 0.99 ), 0.92 ( 0.88 0.95 ), 0.88 ( 0.83 0.92 ) and 0.92 ( 0.88 0.95 ), respectively. Requiring at least 5 positive tests in addition to the 10% MPP threshold optimized accuracy. S eason onset was detected on average 3.4 (0.0 7.2) weeks apart from hospitalization onset and offset was 2.2 (0.0 4.3) weeks apart from hospitalization offset. P erformance of surveillance was limited in southwest and southeas t regions NNTs differed widely between states, calendar months and age groups and confirmed regional difference s in burden of RSV within Florida. The 10% MPP with at least 5 positive tests is a valid threshold even for geographically diverse states. Higher NNTs for older children highlight the reduced benefit of immunoprophylaxis in the second year of life. NNTs can provide valuable detail about the local burden of RSV and their use should be considered by third party payers.
12 CHAPTER 1 INTRODUCTION Background Respiratory syncytial virus (RSV) is the most frequent cause of lower respiratory tract infections among infants and children. In the United States, RSV causes annually up to 125,000 hospitalizations for bronchiolitis among infants under 1 year 1 While no vaccination is available, palivizumab (Synagis, MedImmune, Inc. Gaithersburg, M D ), a humanized m onoclonal antibody, is able to reduce RSV related hospitalizations 2, 3 According to t he label, monthly injections are necessary to provide protection throughout an RSV season 4 The major limiting factor to the widespread use of RSV prophylaxis is the high drug cost, which often result s in expenses of more than $10,000 to immunize one infant through a six month season 5 These costs limit prophylaxis to patients at increased risk for infection such as children with chronic lung disease (CLD) congenital heart disease (CHD) and certain preterm infants 6 Another option for cost containment is the restriction of prophylaxis to a clearly defined RSV season of high viral activity. The Centers for Disease Control and Prevention (CDC) monitor s RSV activity through its National Respiratory and En teric Virus Surveillance System (NREVSS) 7 A nationwide sample of laboratories report the number of specimens tested for RSV and the number of positive tests. These data are updated every week and published on the NREVSS website. While the RSV season peaks in November/December i n most countries of the northern hemisphere, regions closer to the equator show less seasonal variability and observe cases year round. Furthermore, even within countries, RSV outbreaks differ based on latitude and proximity to a coast 8 In the southern United States, the RSV season starts earlier and lasts longer compared to the rest of the nation 9, 10 Differences in seasonality have been identified even within a single
13 es earlier and longer seasons compared to the rest of the state 11 Need for S tudy Critical for the initiation of RSV prophylaxis with palivizumab is not only the knowledge immunoprophylaxis The CDC developed an RSV season definition that is based on the proportion of positive RSV tests among all tests in its nationwide sample of laboratories. When this proportion exceeds 10% in two consecutive weeks, season onset is assumed in the first week While this definition may seem reasonable, the relation ship between this 10% threshold and the burden of disease as measured, for instance, in RSV related hospital admissions, has never been formally established. The NREVSS measures RSV activity through the use of population based tests, thus providing season estimates for the population as whole, albeit not necessarily representative. Subjects tested for RSV are, naturally, patients with a suspected infection. The majority of infections, however, occur among children at low risk 12 simply because they constitute th e largest part of the infant population. Yet, the definition is used to provide immunization recommendations for high risk infants. Susceptibility at periods of lower viral activity may account for a different temporal pattern in RSV hospitalizations among high risk children Such a scenario could occur if infants experience longer RSV seasons than older children Finally state specific surveillance suggests that the RSV season in Florida differs from the rest of the nation with regard to onset and duratio n, which can be attributed to differences in latitude and climate. Of note, when the CDC introduced its season definition, none of the 74 originally c ontributing laboratories were based in Florida 13 Seasonality in Florida has been described as almost year round in the southeast using the 10% definition, though with a large
14 variation in the absolute number in positive tests 11 This variation is likely correlated with pronounced temporal differences in the burden of disease, but the actual RSV incidence rate for each of the months above this threshold is unclear 14 Purpose of S tudy By establishing a validated RSV season definition this study can maximize the acceptance plans and providers select the optimal timing for RSV prophylaxis. The application of a validated definition to different states provide s knowledge about regional differences in the onset and duration of the RSV season. Physicians can use this state specific in formation to prevent RSV infections most efficiently with immunoprophylaxis at the appropriate time 15 Since April 2008, the Florida Medicaid p rogram has required prior authorization (PA) for palivizumab. Florida Medicaid accounts for regional differences in seasonality by allowing different periods of prophylaxis in different regions. 16, 17 However the regional validity of the NREVSS and the 10% threshold may differ from the statewide validity due to a smaller number of labs and therefore, RSV tests. This study provides information about the regional validity of the surveillance system and the appropriateness of the different regional immunization recommendations in Florida. The investigation of patient factors as determinants of RSV seasonality was another objective of this study. I f the temporal occurrence of RSV infections differs between constituents of the high risk population, immunization recommendations may have to be adjusted. Another objective of t his study was to investigate past patterns of palivizumab utilization and cont rast these patterns with disease occurrence as measured with the current season definition based on surveillance data, clinical information from claims data a nd a fixed immunization schedule ignoring surveillance data Therefore, we provide insight about w hether
15 providers based timing of immunoprophylaxis on NREVSS data or on their clinical observa tion s of disease burden, or neither of these. Since application of the season definition provides only dichotomous information and the burden of disease may diffe r between months categorized as on season, this study also provides monthly incidences of RSV hospitalizations for children at high risk together with a detailed picture of the number s needed to treat (NNT) with palivizumab to avoid one RSV hospitalization In addition, we provide NNTs for different age groups to further inform about differences in the need for prophylaxis even among children with indications. Overall, a validated season definition can provide the necessary evidence to guide ision making about the optimal timing and duration of RSV immunization. From a can aid in optimizing reimbursement policies as region month and age specific NNTs can be used optimize RSV prophylaxis Research Questions an d H ypotheses This dissertation consists of six parts. Where formal statistical hypothesis testing applie d statistical significance was 0 refers to null hypothesis and H A to alternative hypothesis. Part I : Validation of CD c urrent RSV s eason d efinition Research Question 1: Is the current RSV season definition used by the CDC able to detect season onset and offset as measured by RSV related hospital admissions in a pediatric population in California, Florida, Illinois an d Texas and in the 5 regions of Florida ? Research Question 1a : Is the use of percent positive laboratory tests at varying thresholds able to discriminate between explicitly defined weeks of high and low disease incidence?
16 Hypothesis 1a : H A : T he area un der the receiver operating characteristics curve of percent positive lab tests exceeds 0.7. H 0 : The AUC does not exceed 0.7. Research Question 1 b : Does the currently used threshold of 10% positive tests optimize sensitivity and specificity? Hypothesis 1 b1 : H A : Sensitivity at the optimal threshold is significantly higher than sensitivity at the 10% threshold H 0 : There is no significant difference Hypothesis 1 b2 : H A : Specificity at the optimal threshold is significantly higher than specificity at the 10% threshold H 0 : There is no significant difference Research Question 1 c : C ompared to using only the 10% or optimal threshold, can additional criteria improve season definitions with regard to accuracy of detecting onset and offset? Hypothesis 1 c1 : H A : A definition with additional criteria significantly reduces the mean absolute difference to season onset according to RSV hospitalizations compared to a definition purely based on the threshold. H 0 : There is no significant mean difference Hypothesis 1 c2 : H A : A definition with additional criteria significantly reduces the mean absolute difference to season offset according to RSV hospitalizations compared to a definition purely based on the threshold. H 0 : There is no significant mean difference Part II : R SV e pidemiology between f our US s tates and f ive r egions in Florida Research Question 2 a : Across 4 US States, does the degree of seasonality differ ; i.e. how pronounced are differences in hospitalization rates during and outside seasons? Hypothesis 2 a : H A : A 0 : there is no difference in the seasonality index between the four states.
17 Research Question 2b: Across 5 regions in Florida, does the degree of seasonality diffe r ; i.e. how pronounced are differences in hospitalization rates during and outside seasons? Hypothesis 2b: H A : A indices. H 0 : there is no difference in the seasonality index be tween the 5 regions. Research Question 2c: For each of the states and for the regions in Florida do seasons have a consistent pattern over time or do onset, peak and offset week s fluctuate over time ? Hypothesis 2c 1 : H A : The range of season onset in one state/region over 5 seasons is significantly shorter than 8 weeks H 0 : The range can reach 8 weeks Hypothesis 2c2: H A : The range of peak weeks in one state/region over 5 seasons is significantly shorter than 8 weeks. H 0 : The range can reach 8 weeks. Hypo thesis 2c3: H A : The range of season offset in one state/region over 5 seasons is significantly shorter than 8 weeks. H 0 : The range can reach 8 weeks. Part III: L atitude as a factor in RSV epidemiology in Florida Research Question 3a: Do season onset, of fset duration, peak weeks or peak RSV incidence rate differ between regions in Florida? Hypothesis 3a1: H A : A t least one region differs in week of onset from any other region H 0 : T here is no difference in week of onset between regions in Florida Hypot hesis 3a2: H A : A t least one region differs in week of offset from any other region. H 0 : T here is no difference in week of offset between regions in Florida Hypothesis 3a3: H A : A t least one region differs in season duration from any other region. H 0 : T h ere is no difference in season duration between regions in Florida.
18 Hypothesis 3a 4 : H A : A t least one region differs in timing of peak RSV activity from any other region. H 0 : T here is no difference in timing of peak RSV activity between regions in Florida. Hypothesis 3a 5 : H A : A t least one region differs in the peak RSV incidence rate from any other region. H 0 : T here is no difference in the peak RSV incidence rate between regions in Florida. Research Question 3b: Are differences in season onset, offset, du ration, peak week or peak RSV incidence rate if identified in research question 3a, a factor of latitude? Hypothesis 3b1: H A : Latitude of the regions is associated with season onset. H 0 : Latitude is not significantly associated with season onset. Hypothe sis 3b2: H A : Latitude of the regions is associated with season offset. H 0 : Latitude is not significantly associated with season offset. Hypothesis 3b3: H A : Latitude of the regions is associated with season duration H 0 : Latitude is not significantly asso ciated with season duration Hypothesis 3b 4 : H A : Latitude of the regions is associated with timing of peak week. H 0 : Latitude is not significantly associated with timing of peak week. Hypothesis 3b 5 : H A : Latitude of the regions is associated with the pea k RSV incidence rate H 0 : Latitude is not significantly associated with the peak RSV incidence rate Part IV : Patient f actors and s easonality Research Question 4a: Do risk factors a ffect the risk for RSV hospitalization s different ly during season compare d to outside of the season?
19 Hypothesis 4a: H A : There is a significant interaction term between any of the risk factors and an on season/off season indicator in the prediction of risk for RSV hospitalizations. H 0 : There is no significant interaction. Part V : Timing of p rophylaxis with p alivizumab vs. RSV s easonality Research Question 5: For each state and for each region in Florida : is onset (offset) of palivizumab utilization close st to season onset (offset) as defined by: A. Observed hospitalization rates B. The NREVSS season definition or C. A fixed immunization schedule Hypothesis 5a: H A : There is a significant difference between the mean absolute difference of A,B or C and onset of palivizumab utilization H 0 : There is no significant difference between th e mean absolute difference of A, B or C and onset of palivizumab utilization Hypothesis 5 b : H A : There is a significant difference between the mean absolute difference of A,B or C and offset of palivizumab utilization H 0 : There is no significant differen ce between the mean absolute difference of A, B or C and offset of palivizumab utilization Part VI : Optimizing t iming of p rophylaxis Research Question 6 : For each state and for each region in Florida, w hat are the incidence rates for RSV hospitalization and the NNT s with palivizumab for high risk children in each calendar month, taking patient age into account?
20 CHAPTER 2 LITERATURE REVIEW Respiratory Syncytial Virus The respiratory syncytial virus was first described in 1956 and named chimpanzee coryza agent (CCA) by Morris and colleagues who examined 20 chimpanzees with respiratory symptoms at the Walter Reed Army Institute of Research 18 There, a laboratory worker who had close contact with infected chimpanzees developed upper respiratory sympto ms and was tested positive for CCA antibodies. The same researchers identified CCA antibodies in a sample of adolescents and young adults who were barrack mates of the worker, which suggests that they may have experienced an infection with CCA or a closely related agent earlier. Accordingly, a nother group of researchers hypothesized that the virus was of human origin and introduced to the chimpanzees, causing the outbreak 19 One year lat er, in 1957, Chanock and colleagues identified two respiratory viruses, named Long and Snyder after the hosting patients, and found them indistinguishable from CCA. Due to their shared characteristic of producing syncytial areas in tissue culture, the rese archers grouped these agents together and coined the term respiratory syncytial virus 19, 20 The RS virus is an enveloped, single stranded RNA virus classified in the genus Pneumovirus within the family Paramyxoviridae 21, 22 Two major strains of RSV, A and B, have been identified and differ mainly in their attachment glycoprotein G. Both strains circulate parallel, with A being the dominant strain 22 It has been suggested that strain A causes more severe infections and fluctuation in circulating strains could explain variation in severity between different seasons 23
21 RSV D isease E pidemiology RSV i s the most frequent cause of lower respiratory tract infections among infants and children. By one estimate, RSV causes annually up to 125,000 hospitalizations for bronchiolitis or pneumonia among children younger than 1 year i n the United States 1 However a more recent study reports an annual number of 57,275 RSV hospitalization s for children under the age of 5 years 24 The same study further estimates that 2.1 million children under the age of 5 experience RSV infections each year ; 3% of these infections lead to hospitalization s, 25% are treated in emergency departments and 73% in pediatric practices. Annual RS V related mortality for underlying pneumonia and influenza deaths has been estimated as 3.1 per 100,000 infants younger than 1 year in the US amounting to approximately 124 RSV related deaths yearly 25 RSV is associated with all cause death in 5.4 per 100,000 infant years or 214 deaths in this age group. Another study found that the majority of RSV related deaths is not associated with typical RSV risk factors. Specifically, among infants who died from RSV related causes before reaching the age of 5 years, only 9.9% h ad underlying congenital heart disease, 5.5% chronic lung disease and 4.2% were born prematurely. Consequently, the authors conclude d that immunoprophylaxis of high risk infants w ould not prevent the majority of RSV related deaths 26 T he incidence of RSV related deaths is lower for older children, adolesc ents and adults, but increases with older age 25 Pneumonia and influenza deaths associated with RSV occur at a rate of 7.2 per 100,000 person years above the age of 65 which amounts to 2,388 annual deaths. RSV associated all cause mortality was estimated at 29.6 per 100,000 person years or 9,812 annual deaths between the seasons 1990/91 and 1998/99. Eig hty eight percent of RSV related pneumonia or influenza deaths occurred in the elderly; however this may be due to their overall higher mortality rate.
22 RSV I nfections The clinical picture of RSV infections appears differently in neonates and infants comp ared to older children and adults 21 Infections in the first 4 6 weeks of life are rare, which may be related to the presence of maternal antibodies 21, 23 Infected children between 6 w eeks and 2 years of age usually develop symptomatic lower respiratory tract infections, including bronchiolitis and pneumonia, but also acute otitis media. Symptoms typically appear after an incubation period of 5 days and resolve after a few days to one w eek 21 Older children and adults often experience mild to moderate upper respiratory tract infections which are consequences of recurrent RSV infection s suggesting an incomplete acquired i mmunity after a primary infection 22 Following RSV bronchiolitis, infant patients often exhibit recurr ent wheezing up to ten years after infection, and an increased incidence of asthma in children subsequent to RSV hospitalization has been reported 21, 27, 28 Conversely, a recent study found that RSV infection s increase d the incidence of asthma up to 8 fold only during the first 2 months after RSV hospitalization and there was no increased risk one year after the infection. 29 Diagnostic Tests f or RSV RSV infections cannot be distin guished from non RSV related lower respiratory tract infections based on clinical signs and symptoms, which necessitates the use of laboratory tests. Three approaches to testing are available: cell culture, serology and examination of respiratory secretion s 30 Cell culture allows examination of viral co infections and genetic and a ntigenic change of the RS virus, but its long assay time of 2 5 days decreases utility in clinical practice. The use of serology is limited by a large number of infected patients who remai n serologically negative. Most commonly, laboratories employ antigen based assays because they are i nexpensive and easy to perform. In fact, all tests reported to the NREVSS are antigen based. 31 Different methodologies are available, including immunofluorescent antibody test, enzyme
23 linked immunoassay, direct im munoassay and optical immunoassay. Antigen detection is rapid with a turn around time of 15 30 minutes. The different methods of antigen detection vary in sensitivity and specificity with a lack in these attributes if the viral concentration is low: in eld erly or immunocompromised patients and outside the main RSV season. A newer method for analysis of secretions includes molecular essays using polymerase chain reaction. Due to high sensitivity and very high specificity they are considered gold standard in diagnosing RSV, but widespread utilization is limited by assay cost and labor intensiveness 30, 32 Sen sitivity and s pecificity of a ntigen based t ests : Sensitivity of antigen based tests can vary between 59 and 98% with enzyme linked immunoassay performing the poorest 30 Specificity can range between 75 and 100%; again 75% is associated with enzyme linked immunoassay. A study has found that s ensitivity decrease s with age: 72% sensitivity in a sample of less than one year olds compared to 19% among children older than 1 year 33 This study did not identify differences in sensitivity and specificity with regard to RSV season for the younger age group but a furthe r drop in sensitivity outside the season among the older age group. Since the test results reported to the NREVSS are not limited to infants under 1 year of age, it can be assumed that sensitivity, especially at periods of low viral activity may be subopti mal. This would result in an underestimation of RSV burden outside high risk periods. On an aggregate level however, respiratory lab tests from two single center studies including tests for RSV correlated well with clinical case counts in the same popula tions 34, 35 Furthermore, Light et al. demonstrated that virus circulation as measured by the Florida Department of Health (DoH) RSV survei llance system through laboratory tests parallels the incidence of RSV hospitalization claims in the Florida Medicaid database 36 A correlation
24 between the percentage of positive lab tests and clinical burden of disease is essential for the use of the NREVSS data to detect RSV season onset and offset. The National Respiratory and Enteric Virus Surveillance System The NREVSS mon itors temporal and geographic activity of respiratory and enteric viruses. These viruses include RSV human parainfluenza viruses, respiratory and enteric adenoviruses, rotavirus, and since 2007, rhinovirus, enterovirus and human metapneumovirus. Influenza specimen information, also reported to NREVSS, is integrated with CDC i nfluenza s urveillance data. Collaborating university, community hospital, commercial, and state and county public health laboratories report virus detections, isolations, and electron microscopy report results on a weekly basis. Annual summaries from NREVSS are published in Morbidity and Mortality Weekly Reports (MMWR) 37 The first reference to RSV surveillance by the CDC appeared in an MMWR from 1984 with qualitative rather than quantitat ive information and unclear temporal and geographic coverage 38 A more systematic approach co incided with the first mentioning of the NREVSS in 1990 including data from 95 laboratories in 49 states 39 The same p ublication announced for the season 1990 91 a switch from a monthly postcard reporting system to a weekly telephone based reporting system through a computer polling service with automatic tabulation allowing the publication of available results in the fol lowing week 39 The information supplied by each participating laboratory consists of the weekly number of specimens te sted for RSV and the number of positive tests In 1993, the CDC provided a first definition for virus activity: weeks 40 This definition was refined in the following season as: 41 An update to this definition from 1998 is still in use least half of labs report any RSV detections for at least 2 consecutive weeks and when greater
25 42 Further changes to the system include treating Florida separately from 2005 due to its unique RSV pattern 43 and using inclusion criteria for participating labs for the 2006 repor of specimens positive annually Lastly, from 2007 08, the system has experienced a substantial increase in the number of participating laboratories by using commerc ially available data from Surveillance Data, Inc., which is supported by MedImmune, Inc., the manufacturer of palivizumab (see table 2 1 ). RSV Seasonality The RSV season spans from late fall to spring with peaks in November/December in most countries of th e northern hemisphere, but regions closer to the equator show less of a seasonal variability with year round case occurrences. Even within countries, RSV outbreaks differ based on latitude and proximity to the coast 8 In the southern United States, the RSV season starts ea rlier and lasts longer than in the rest of the nation 9, 10, 44 M ore specifically, the impact of latitude and proximity to the coast may even be detectable within single states such as Florida, with a considerable north south extension and large coast line. In fact, longer seasons have been reported for s outheast Florid a compared to other parts of the state based on surveillance data 11 Immuni zation recommendations are further complicated by annual variation in season onset and duration in the United States necessitating current information on viral activity 9, 44 A biennial pattern of RSV seasonality has been indentified in Sweden, with earlier season s alternating with later seasons and higher hospital admission rates in earlier seasons 45, 46 A similar pattern was found in Germany 47 and Finland 48 however this bie nnial pattern has not been shown for the US. 9 11, 13, 35, 36, 44, 49, 50 It has been acknowledged that the 10% threshold is arbitrary 49 and that its suitability for defining months for
26 RSV prophylaxis is unclear 51 Two studies supported by MedImmune applied a slightly modified season month ) to the Florida DoH RSV surveillance system 11 and test data from three hospitals 49 These studies identified RSV activity almost year round in Florida, reaching epidemic levels i n most months. A n editorial published with the former study found that even across time periods above this threshold, the absolute number of cases is highly variable 52 More specifically, we found and expressed that, in selected months (May through August 2001) the absolute statewide number of positive tests reported in that study was less than 10 as compared to 143 in January indicating a large difference in the burden of dis ease 14 A study from 1998 using laboratory data from two hospitals in Jacksonville, Florida f 2 months instead of the first of 2 weeks as the CDC do es 10 This study reaches similar conclusions with almost year round RSV activity in Florida ; yet again, many summer months show less than one tenth of positive tests of some winter months and still exceed the 10% epidemic threshold. All these observations highlight the problem that a dichotomous categorization into on season and off season periods can mask large differences in the burden of disease even during season. Whether the potentially much lower incidence of RSV in the summer months identified in th e three quoted studies warrants prophylaxis, despite exceeding the 10% threshold, is not clear. RSV Prevention While no vaccination is available for RSV 23, 53 two preventive agents, respira tory s ync ytial virus immune globulin i ntravenous (RSV IGIV, RespiGam, MedImmune, Inc. Gaithersburg, MD) and palivizumab (Synagis, MedImmune, Inc.) are indicated to reduce the risk of RSV related hospitalization. RSV IGIV was approved by the US Food and Dr ug Administration in January 1996 54 The second generation product, palivizumab, was approved in 1998 55 Palivizumab is a humanized monoclonal antibody and can thus avoid the risk for infections
27 potentially associated with the older, pooled human blood pro duct RSV IGIV. Also, the smaller injection volume and less burdensome intramuscular application of palivizumab contributed to its replacement of intravenous RSV IGIV from the market 55 57 As a consequence, the manufacturer discontinued the production of RSV IGIV at the end of 2003 56 A new agent for the prevention of RSV infections, motavizumab (MedImmune, Inc.), has reached phase III of drug development and its sponsor submitted a Biologics License Application to the FDA in January 2008 58 At this time, no published clini cal trial results have been identified as evidence for its efficacy. Currently, palivizumab is the only available pharmaceutical product for the prevention of RSV related hospitalizations. Palivizumab has proven its efficacy in clinical trials, with varyin g estimates for different indications. IMPACT RSV reports a relative risk reduction for RSV hospitalizations (RRR) of 39% for children with CLD, 78% for premature infants and 55% for the combined group 2 In another trial, children with CHD experienced a 45% reduct ion in RSV related hospitalizations 59 To date a reduction in RSV related mortality has not been demonstrated for palivizumab 6, 60 The cost of RSV prevention w ith palivizumab is significant. The average wholesale price of one 50mg vial is $926.48 61 sufficient for one monthly dose for a 3 kg infant at 15mg/kg. With increasing age and body weight, higher monthly doses at higher total cost are required One study found that the average dose in a Medicaid population of 0 2 year old children cost s about $1,700 5 If 6 doses are administered as is common in Florida, total cost of immunoprophylaxis for one child averages more than $10,000 for a single season and significantly more if palivizumab is adm inistered year round Costs per avoided hospitalization are often found to exceed expenses of the actual hospitalization by far suggesting unfavorable cost benefit 5, 62 In
28 the light of the se considerations recommendations limit immuno prophylaxis to patients at highest risk for infection during seasons of high viral activity 6 RSV Risk Fact ors and Indications for I mmunoprophylaxis The American Academy of Pediatrics has defined indications that describe children at high risk for RSV infections 6 and recommends immunoprophylaxi s for: Children less than two years of age with chronic lung disease ; Children with a gestational age of less than 28 weeks if they are not more than 12 months old at season onset (for this study: Prematurity I) ; Children with gestational age of 29 32 wee ks if they are not more than 6 months old at season onset (Prematurity II) ; Children with a gestational age of 32 35 weeks if additional risk factors such as day care attendance or smoking parents are present (Prematurity III) ; Children with hemodynamicall y significant cyanotic and acyanotic congenital heart disease ; Children with cystic fibrosis (CF) ; and Children with severe immunodeficiencies which include severe combined immunodeficiency (SCID) and acquired immunodeficiency syndrome (AIDS). T he last tw o indications are not based on strong evidence for effectiveness (expert opinion only) and more carefully phrased in the guideline The se recommendations are supported by several studies that report increased RSV incidences in some of the above risk group s, yet to a varying extent. For instance, Boyce et al. ( 2000 ) report a risk for RSV hospitalization in infants younger than 6 months with CHD of 120.8 per 1,000 infant years of RSV season compared to 44.1/1000 for low risk infants in the Tennessee Medicaid population. In contrast, Duppenthaler et al. ( 2004 ) estimate a rate of only 25 RSV hospitalizations per 1,000 infant years in the same age group with CHD compared to
29 18/1000 without CHD in Switzerland 12, 63 Not included in the AAP guideline, the presence of Down syndrome has been suggested to be associated with increased risk for RSV infections 64 Furthermore, a few small sample studies suggest a more severe course of RSV infections in im munocompromised children with liver transplant 65, 66 or after chemotherapy for cancer 67, 68 W hether malignancy actually increases the infection risk for RSV is disputed 69 and conclusive evidence is lacking. Internationally, RSV prevention guidelines differ in scope ; for example the Swedish guideline 70 is more restrictive than th e rather inclusive AAP guideline in the US 6 The former gestational age only up to 6 months of age or up t o 24 months of age with CLD and prematurity Prior Authorization R equirements To limit expenditure for RSV immunoprophylaxis, third party payers typically restrict reimbursement for prophylaxis to children who meet certain c riteria and to a limited time period of high risk. A common instrument is a requirement for PA where providers have to confirm the presence of risk factors to the third party payer to request reimbursement. Reimbursement policies in the 4 study states duri ng the study period are detailed here: California : California Medicaid has always required PA for palivizumab since its market introduction in 1998. Requirements changed with updates in AAP guidelines (personal communication, California Medicaid Pharmacy Benefit Division, 02/23/2009) The current PA guideline is in agreement with the AAP guideline, with the exception that it allows for utilization up to the age of 48 months in immunodeficient children 71 It allows for the administration of up to six doses between October and May.
30 Florida : Prio r to 200 8 Florida Medicaid did not restrict access to palivizumab for children under the age of 2 years, neither with regard to risk factors n or timing of immunization. Since April 2008, Florida has used a PA system that allows palivizumab utilization for high risk children according to the AAP guideline for different time periods in 5 regions and up to year round in the southeast of Florida (p ersonal communication with Anne C. Wells, Bureau Chief, Florida Medicaid Pharmacy Services, 02/02/2009) 17 The 5 regions are illustrated in figure 2 1 ; an overview of the county composition of each region is provided in the appendix (table B 1). Illinois : Until 2005, no PA requirement was in place for children under 4 years of age and immunoprophylaxis was not restricted to season months During this time, PA was required fo r children older than 4 years of age Of note, palivizumab has even been used for ventilated children up to the age of 1 8 years Since 2005, the PA requirement mirrors the AAP guideline ( p ersonal communication with Brad Berberet, Assistant Director, Prior Authorization, Department of Pharmacy Practice, University of Illinois at Chicago, 02/23/2009) Texas : Texas Medicaid has had a PA requirement in place during the study period and beyond. The PA was closely modeled after AAP guidelines, allowing prophylaxis for children with CLD, prematurity or CHD during season months according to NREVSS ( p ersonal communication with Judy Devore, Special Assistant to t he Medicaid/CHIP Medical Director, Texas Health & Human Services Commission, 02/2 4 /2009)
31 Table 2 1. Historical landmarks in the NREVSS Season Update to the surveillance system MMWR Reference 1983 84 RSV surveillance first mentioned in available MMWR no explicit season definition 01/1984 38 1989 90 NREVSS with 94 participating labs Change from monthly postcard to weekly telephone based system announced for 1990/91 11/1990 39 1992 93 01/1993 40 1993 94 participating labs reported any RSV detections or Offset: no definition 12/1993 41 1994 95 mean percentage of specimens positive by antigen 12/1994 72 1997 98 detections for at least 2 consecutive weeks and when 12/1998 42 1998 99 Same as above, in addition: Community outbreaks utive weeks 12/1999 73 2005 06 2006 07 2007 08 Florida is added as a separate region. National activity : defined as above (1997 98) Regional activity: consecutive weeks a participating lab reports >1 0% tests preceding 2 consecutive weeks of <10% imens National and Regional season median percentage of specimens positive for RSV median percentage of positiv Expansion of available lab data by inclusion of data from Surveillance Data, Inc. for the season 2007 2008, with support from MedImmune, Inc. reported > 30 weeks and averaged > 10 antigen detection 12/2006 12/2007 12/2008 43 74 31
32 Figure 2 1. Map of RSV regions in Florida 757575 74 75 reprinted with permission of the Florida Department of Health
33 CHAPT ER 3 METHODS Datasets This study was based on two datasets, a surveillance dataset and a clinical, patient level dataset based on medical and pharmacy claims. The surveillance dataset was comprised of NREVSS RSV surveillance data for the states of Californ ia, Illinois and Texas, and DoH surveillance data for the state of Florida. The clinical dataset consist ed of Medicaid data for the same four states commonly referred to as Medicaid Analytic eXtract (MAX) data provided by the Centers for Medicare and Med icaid Services (CMS). The claims dataset for Florida was obtain gestational age estimates Each dataset cover ed the years 1999 2004, thus including 5 RSV seasons. The choice of the 4 states provide d a large sample size as well as geographic diversity with the intent to increase external validity. Characteristics of the source datasets are described below. NREVSS The RSV component of the NREVSS collects weekly information on RSV tests from a sample of laboratories as described in greater detail above. The resulting dataset consist s of a l aboratory identification number and the city and county where the lab is located Laboratories report the test type, the number of total tests and the number of positi ve tests recorded for a given week. Fl orida Department of Health RSV Surveillance D ata As mentioned above, Florida had not been a separate part of the NREVSS before 2005 and only a small number of Florida based laboratories reported to the CDC before that To investigate seasonality on a regional level within the state these data had to be supplemented. The Florida DoH has its own RSV surveillance system in place, using the same laboratory
34 survey approach as the CDC, but with a larger sample in Florida. G eographic detail is provided at the city level and for 5 regions within the state (figure 2 1, see appendix for the regional classification of counties) 75 Me dicaid Analytic eXtract Claims D ataset For each of the four states, eligibility, inpatient, outpatient and pharmacy datasets for Medicaid recipients 0 2 years of age were requested. Eligibility and demographic information w as updated for each month. The MAX dataset has already been reconciled by CMS to display ich eliminate s the need to remove duplicate claims that are generated as results of reimbursement negotiations Since MAX data do not include claims for managed care enrollees, we restricted the study cohort to children who were only in the fee for service (FFS) or primary care case management (PCCM ) program Furthermore, if potential study subjects were enrolled in a behavioral or dental plan, we included them if they were also FFS or PCCM enrollees. We refer to this entire group as the FFS sample. Interna l V alidity : The validity of Medicaid claims data with regard to measuring the incidence of RSV related hospitalizations has not been investigated. Several articles comment on the general utility and limitations of Medicaid data in epidemiologic research. T he greatest strength lies in the size of the r epresented population, allowing for the investigation of effects of rare exposure or rare outcomes with the ability to adjust for multiple confounders 76, 77 With regard to diagnostic validity, it has been recogni zed that hospital based information is superior to other Medicaid datasets 76 Kiyota et al. (2004) calculated a positive predictive value of 94.1% for Medicare administrative claims of discha rge diagnoses for acute myocardial infarction validated against hospital records. 78 Another study looked at gross diagnostic errors in Medicaid data such as childbirth or pregnancy related codes among recipients older than 60 years and concluded that these errors were not widespread in Medicaid claims datasets 79 Members of that study group
35 also looked at Medicaid data obtained from CMS, including our study states California and Florida and investigated longitudinal patterns of inpatient and prescription cla ims to identify breaks. T hey found a linear trend in these claim s suggesting no paucity of data. 80 Furthermore, they found that the proportion of prescriptions with a valid national drug code ( NDC ) ranged from 97 99% be tween the states indicating a high degree of coding validity. RSV hospitalizations were chosen over alternative endpoints such as emergency department visits, physician office visits or death for consistency with the existing literature, the label of paliv izumab which lists the prevention of RSV related hospitalizations as a primary indication and the public health importance of RSV hospitalizations as outlined above. State Birth Certificates Our approach to detecting RSV seasonality in the study sample re quire d the establishment of high and low risk cohort s of Medicaid recipients with regard to their propensity for RSV infection. Part of the high risk classification use d gestational age to define premature birth. Medical claims data are not a valid source of gestational age ; therefore we supplement ed this information with birth certificates. Birth certificate data are collected within in 24 48 hrs after birth by the hospital, which is where 99% of births in the US are delivered 81 Gestational age is calculated based on the when missing, imputed from the clinical estimate (CE) at birth. LMP and CE based estimates were found to agree within 2 weeks in 89.1% of infants. A study based in California validated LMP estimates against a CE estimate using ultrasound at week 15 20 and found that LMP has a false positive rate of 15% for identifying preterm birth and a missed true cases at a rate of 20.5% 82 Although not perfect, th is level of accuracy is acceptable for our study and vital statistics birth data are the only sources for
36 gestational age estimates on a population level. Birth certificates were matched to Medicaid eligibility data based on social security number and date of birth or name and date of birth Study Population To be included in the claims dataset, children had to be younger than 2 years of age with continuous eligibility to Medicaid between birth and the current week. For Florida the availability of a birth certificate with a gestational age between 15 and 50 weeks was further required to exclude invalid estimates Subjects had to be in ambulatory care for at least 4 weeks before the start of each week examined This r equirement ensure d the ability to detect immunoprophylaxis in ambulatory care since prophylaxis administered in the hospi tal is not identifiable from claims data d ue to the aggregate nature of inpatient charges in the Medicaid dataset. As a consequence, th is requirement essentially remove d the first month of life from the analysis. As mentioned before, the first 4 6 weeks of life have a lower RSV incidence. A range of risk factors for RSV infection have been discussed in the literature and are described abo ve. To avoid misclassification, high risk was defined according to the consensus that led to the indications for immunoprophylaxis based on grade I evidence in the AAP guideline 6 To crea te a low risk category, we excluded all children with any potential clinical risk factors for RSV described in the literature To address the paucity of evidence for some of these risk factors, we constructed a logistic regression model based on the Florid a dataset with RSV hospitalization as the dependent variable (equation 3 1 for operational definitions of each variable see appendix ) Adjusting for demographic characteristics, calendar year and month as well as the presence of immunoprophylaxis, we det ermined the contribution of each potential risk factor to th e risk for RSV hospitalizations The model used subject week s as the unit of analysis.
37 1 X 1 + 2 X 2 3 X 3 4 X 4 5 X 5 6 X 6 7 X 7 8 X 8 9 X 9 10 X 10 11 X 11 12 X 12 13 X 13 14 X 14 15 X 15 16 X 16 17 X 17 18 X 18 19 X 19 (3 1) Where: Y: RSV hospitalization X 1 : Calendar year (1999 2004) X 2 : Calenda r month ( 1 12) X 3 : Age ( 0 6, 7 12, 13 24 months) X 4 : Sex X 5 : Race (White, Black, Native American, Asian, Hispanic, Unknown) X 6 : Birth month (1 12) X 7 : Prematurity I X 8 : Prematurity II X 9 : Prematurity III X 10 : CLD X 11 : CHD X 12 : CF X 13 : Immunodeficiency X 14 : Down syndrome X 15 : Asthma X 16 : Transplant X 17 : Malignancy X 18 : Immunosuppression X 19 : Palivizumab prophylaxis The model create d (t 1) dummy variables (where t = the number of levels) for each categorical variable Based on the results of the logistic re gression analysis, a l ow risk category was created excluding all clinical factors that exhibited a trend towards an association with increased risk for RSV infections, even if statistical significance was not met. Subjects were excluded based on clini cal factors and not demographic characteristics that may also have been associated with RSV infections. An exclusion based on demographic risk factors, potentially excluding boys who are at higher risk than girls, would have led to a cohort that lacks comparab ility to the high risk We refer to the excluded children as the elevated risk cohort and present their immunization and RSV hospitalization rates for comparison.
38 The high risk cohort include d ( o perational definitions in the appendix): Children younger than 2 years with CLD who received medication for CLD within 6 months of the current week ; and Children younger than 2 years with hemodynamically significant cyanotic or acyanotic congenital heart disease. In addition in Florida: Children with a gestational age of less than 28 weeks if they were not older than 12 months at the beginning of the current week ; and Children with gestational age of 29 32 weeks if they were not older than 6 months at the beginning of the current week The low risk cohort include d all children who were not part of the high risk cohort at a given week and who did not have any of the following between birth and the current week : Cystic fibrosis Severe combined or acquired immunodeficiency Down syndrome Asthma T ransplant Immunosuppression and Malignancy In addition in Florida: Gestational age of 32 35 weeks and less than 6 months of age at the beginning of the current week We included asthma only if, in addition to a diagnostic code, a prescription for an asthma medication was billed between 120 days and 14 days before the current week. This ensured that the condition was current and prevented the exclusion of children who may have had asthma codes only as a consequenc e of early symptoms of an RSV infection.
39 The remainder of the methods section address es each of the parts I VI outlined in the introduction. Unless otherwise specified, SAS 9.1.3 (SAS Institute, Cary, NC) was used for data analyses and graphs were created in Microsoft Excel 2007 (Microsoft Corp. Redmond, WA) This study has received approval from the institutional review boards of the University of Florida, CMS and the Florida DoH. Part I: efinition Calc ulation of RS V i ncidence r ates : To detect RSV seasons from claims data, weekly incidence rates among low risk and high risk infants w ere calculated. This calculation was complicated by the fact that immunoprophylaxis differentially reduces the incidence rate of RSV ho spitalizations for high risk versus low risk children due to different rates of palivizumab u tilization Simply excluding palivizumab recipients would have distort ed the cohorts since the antibody is not given at random but targeted to higher risk patients even within the high risk cohort. Therefore, two cohorts of pseudo non recipients were created as if no ne had received prophylaxis. The incidence rate in each cohort was based on the assumption that prophylaxis reduced the relative risk of an RSV hospita lization by 50%, the mean of the 45% and 55% RRR shown in the clinical trials 2, 3 As a result, the observed incidence rate of the exposed was assumed to be half the incidence rate had they not been exposed. The corrected weekly incidence rate (I c ) for each cohort was calculated from the number of cases among the truly unexposed (A U ) and a corrected (1/RRR) case number among the exposed (A E ) divided by the sum of person time for the unexposed (N U *T U ) and the exposed (N E *T E ) ( equation 3 2 ). (3 2 )
40 The denominator for the incidence rate was the number of children eligible in a given week ; the numerator was the number of weeks with RSV related hospital admissions identified by inpatient claims with ICD 9 CM codes (International Classification of Diseases, 9th R evision, Clinical Modification) for RSV infections (see appendix). We identified palivizumab exposure as the presence of a claim for palivizumab (based on NDC or procedure codes, see appendix) a subject was exposed for at least half of the current week. Season definitions : A key component in the validation of a season defined by laboratory tests is the defin ition of a gold standard for comparison Since the intended use of the surveillance system is to detect a clinically significant outbreak in th e population, a population based standard derived from the Medicaid claims dataset was needed This standard had to provide estimates for season onset and offset which require d their own definition s ous Diseases and Committee on Fetus and Newborn recommends immunoprophylaxis only for high risk children during the RSV season 6 Consequently, children at low risk for RSV do not have an i ndication for prophylaxis at any time of the year. Thus, it can be argued that the peak RSV incidence rate among low risk children does not reach a level that is considered high risk. Therefore, the peak incidence of low risk children shall be considered a threshold below which RSV risk is low The season for high risk children was defined for this study as every week where the RSV incidence rate among high risk children exceed ed this threshold. To attenuate the impact of outliers, we used a 2 RSV Season onset was defined at the first of two consecutive weeks
41 wher e the high risk incidence exceeded the peak of the low risk incidence. Accordingly, season offset occurred at the last week before two consecutive weeks where the high risk incidence was below the peak incidence of the low risk cohort (figure 3 1 ) We used this approach for statewide data in the 4 states; however the smaller size of the high risk cohort in the individual regions of Florida would have introduced significant misclassification. To illustrate, with a small denominator even a single RSV case c ould have led to a large RSV incidence causing us to consider the corresponding week as a n on season week. To overcome this challenge, we used a different approach to identify clinical RSV seasons in the regions of Florida. For each year (July June) we ex tracted the overall RSV incidence rate at which the statewide model detected season onset in Florida. We also extracted the incidence rate at season offset and considered the average of the onset and offset incidence rate as the threshold incidence for tha t season. This threshold was applied to each region in Florida S eason onset was defined as the first of two consecutive weeks above this threshold. Accordingly, season offset was defined as the last week before two consecutive weeks below this threshold. Validation of the NREVSS RSV season definition : For this analysis, we restricted the observation period to include 5 full seasons. A year started in July (week 27) and ended in June (week 26), thus avoiding partial RSV seasons at the beginning and end of the observation period. In Florida, we were only able to observe 4 seasons (starting July, 2000) because DoH surveillance data w ere not available before the week 1999 42. Each week of the observation period was categorized as on season/off season based on the claims dataset as gold standard and compared to the median proportion positive (MPP) RSV tests from the surveillance dataset The MPP was calculated as the value that divided the laboratories into 2 halves with regard to proportion positive in a given state/region for a given week.
42 When testing the validity of the season definition, two concepts have to be delineated from each other and were tested separately: Is the test (e.g. MPP lab tests) able to differentiate weeks of clinical high burden from wee ks of low burden of disease? Is the threshold (e.g. 10%) the right choice to optimize sensitivity and specificity for correctly identifying RSV seasons ? To answer these questions, receiver operating characteristic (ROC) curve s were plotted with test sensit ivity on the y axis and 1 specificty on the x axis. In addition, s eparate ly for each state and for regions in Florida the following 4 test characteristics were calculated (figure 3 2 ) : Sensitivity: among true on season weeks, % of weeks classified as on season Specificity: among true off season weeks, % of weeks classified as off season Positive Predictive Value (PPV): among weeks classified as on season, % true on season weeks Negative Predictive Value (NPV): among weeks classified as off season, % tru e off season weeks For each observed MPP in each week the resulting (specificity, 1 sensitivity) point was plotted and connecting all points yield ed the ROC curve. An important attribute of ROC curves is the area under the c urve (AUC). The AUC is a measur e of accuracy of a test and varies between 0.5 for a non discriminating test (a random guess) and 1.0 for a test with perfect accuracy 83 A classification of test accuracy has been commonly applied (figure 3 3 ) 84 86 and defines an AUC above 70% as fair. Separat ely for each state and region, we calculated the AUC and concluded that the MPP was adequate if its discriminatory ability exceeded AUC we identified the optimal threshold of MPP For the detection of RSV seasonality and optimization of resource usage, it would be equally important to detect on season weeks as such (sensitivity) as it is to detect off season weeks as such (specificity). Therefore, we identified the optimal threshold as the point
43 where sensitivity and specificity are optimized simultaneo usly. The optimum is part of the output of a SAS macro that we used for our analysis. 87 Confidence intervals for sensitivity, specificity, PPV and NPV were calculated using the Wilson score method. 88 We tested wheth er the optimal threshold provid ed significantly higher sensitivity or specificity than the currently used 10% MPP threshold by inspecting the 95% confidence intervals for overlap. The analysis up to this point can only give information about the accuracy of the observed MPP as a predictor of an RSV season for a randomly chosen week. However, the season definition as used by the NREVSS has additional requirements such as onset is the first of two consecutive weeks where the 10% threshold is exceeded. With this re quirement, the MPP threshold is not the only predictor. Therefore we tested the accuracy of the actual season definition with the added requirements. We also tested alternative definitions with more requirements, namely that in a given state/region and wee k, at least 3 or at least 5 tests have to be positive. With this approach we aimed to eliminate the impact of outliers especially at times of infrequent testing. For all these potential definitions we tested two thresholds: 10% MPP and the optimal MPP for each state/region as derived from the ROC analysis To test accuracy, we calculated the mean of absolute differences between season onset and offset according to each definition and according to RSV hospitalizations. We used one way analysis of variance to test whether any of the definitions differ ed from the other s with regard to accuracy. To inform about the temporal direction of the difference, we also calculate d the mean of actual instead of absolute differences. Part II : RSV Epidemiology between Four US States and Five Regions in Florida For each state and for each region in Florida, we calculated a seasonality index based on seasons according to RSV hospitalizations This seasonality index is the incidence rate ratio of
44 the incidence rate of RSV hospi talization s during season divided by the incidence rate outside of the season. Again, we used the Wilson score method to calculate 95% confidence intervals. 88 To analyze annual variabi lity of seasons within a state / region, we calculated the mean weeks of onset, peak and offset and their standard deviations. We tested the hypothesis that the range of season onset (peak, offset) in one geographical area does not exceed 8 weeks, a range th at we consider ed low variation. The model we used tests whether the week of onset (peak, offset) is uniformly distributed over a k week period, and we used the actual standard deviation to construct a 95% confidence interval for k as follows. A simulation calculated standard deviations of 100,000 samples of 5 values each from a range of 1 k. This simulation provided standard deviations under the null hypothesis that onset (offset, peak) occurs randomly in a range of k weeks. The proportion of occurrences wh ere the actual standard deviation of season onset (offset, peak) is smaller than the simulated standard deviation under the null hypotheses provides evidence against the null hypothesis. For each geographic area, we varied the value of k until the proporti on of simulated standard deviations that were smaller than actual standard deviations was less than 5%. The resulting k was then interpreted as the upper bound of a one sided 95% confidence interval. If the upper bound was larger than 8, we rejected the al ternative hypothesis that season offset (peak, offset) has a range of not more than 8 weeks. Part III: L atitude as a Factor in RSV Epidemiology in Florida For each region in Florida, we used a one way analysis of variance to examine separate ly whether at least one region differs from any other region in the parameters week of onset, offset and peak season duration and peak RSV incidence Next, ArcGIS 9.1 (ESRI, Redlands, C A ) was used to determine coordinates of centroids of the five surveillance regions i n Florida to allow assessment of the influence of latitude on the season parameters above. We only included season parameters where the analysis of variance showed significant variation between regions. Latitude
45 of the centroids (table B 2, appendix) serve d as an independent variable (X ) in a linear regression model with week of onset (offset, peak) or duration (in weeks) or peak RSV incidence rate as dependent variable (Y) : X (3 3) was i nterpret ed as change in timing of onset (offset, peak) relative to change in latitude. One degree increase of latitude corresponds to moving 69 miles (111km) away from the equator 89 JMP 7.0.2 ( SAS Institute, Car y, NC ) was used to conduct the analysis and create graphs. Part IV : Patient F actors and S easonality This analysis determine d whether high risk indications increase the odds for RSV hospitalization differently in off season versus on season periods. Subjec t weeks for each state and on a regional level in Florida were categorized as on season or off season according to RSV hospitalizations as described in part I To gain statistical power, we concatenated all states into a single dataset. We created a logist ic regression model (equation 3 4) including all subject weeks where children were in ambulatory care in the 4 weeks preceding the first day of the current week Apart from the RSV risk factors as described above (equation 3 1) the model further included i nteraction terms betwe en the season indicator and the high risk indications CLD and CHD to determine whether these factors predict the probability of RSV infection differentially between on season and off season periods. We further included interaction te rms between the season indicator and age (0 6, 7 12 and 13 24 months) since age is another strong predictor for RSV hospitalization. 1 X 1 2 X 2 3 X 3 4 X 4 5 X 5 6 X 6 7 X 7 8 X 8 9 X 9 10 X 10 11 X 11 12 X 12 13 X 13 14 X 14 15 X 15 16 X 16 17 X 17 18 X 18 + 19 X 1 X 4 + 2 0 X 1 X 5 + 2 1 X 1 X 9 + 2 2 X 1 X 10 (3 4 )
46 Where: Y: RSV hospitalization X 1 : RSV sea son (yes/no) X 2 : Calendar year (1999 2004) X 3 : Calendar month ( 1 12) X 4 : Age 0 6 months X 5 : Age 7 12 months X 6 : Sex X 7 : Race (White, Black, Native American, Asian, Hispanic, Unknown) X 8 : Birth month (1 12) X 9 : CLD X 10 : CHD X 1 1 : CF X 1 2 : Immunodeficiency X 1 3 : Down syndrome X 1 4 : Asthma X 1 5 : Transplant X 1 6 : Malignancy X 1 7 : Immunosuppression X 1 8 : Palivizumab prophylaxis X 19 : Interaction RSV season Age 0 6 months X 20 : Interaction RSV season Age 7 12 months X 21 : Interaction RSV season CLD X 22 : Interaction RSV season CHD The model analysis create d (t 1) dummy variables (where t = the number of levels) for each categorical variable For each of the variables included in interaction terms, we calculated odds ratios and 95% confidence intervals stratified b y off season and on season and plotted the results in a forest plot to facilitate comparison (SigmaPlot 10.0, Systat Software, Inc., San Jose, CA) Part V: Timing of Prophylaxis with Palivizumab vs. RSV Seasonality From the Medicaid pharmacy claims datas et separate by state region and year, the weekly number of palivizumab prescriptions was identified (NDC and procedure codes: see appendix) and divided by the number of children eligible to the study in that week. We detected onset of utilization as the f irst of two consecutive weeks where the utilization rate exceeded 50% of the peak utilization rate in the respective geographical area for that season (July June) We chose
47 50% over any lower threshold to detect strong population coverage. Offset was defin ed as the week preceding two consecutive weeks with a utilization rate below 50% of the peak utilization rate. Annual patterns in utilization were contrasted with seasons identified from : Observed RSV hospitalization rates The current 10% threshold of the NREVSS season definition and A fixed immunization schedule For the latter, we used the first week of October and the first week of November as comparators for utilization onset and the last week of March and April as comparators for utilization offset. To determine which of these comparators is closest to the onset (offset) of utilization, we calculated the mean absolute difference between utilization onset (offset) and RSV season onset (offset) according to the comparators. Confidence intervals for mean s were calculated using the t distribution with n 1 degrees of freedom instead of the z distribution to account for the small number of observations. 90 One way analysis of variance was used to test whether any of the comparators was closer to utilization onset (offset) than any of the others and we used the lowest point estimate of the mean of absolute differences to determine which of the comparators was most likely to have trigger ed immunoprophylaxis. To inform about the tempor al direction of the difference, we also calculate d the mean of actual instead of absolute differences. Part V I : Optimizing T iming of P rophylaxis The goal of this analysis was to provide information on the actual burden of disease in high risk children for each calendar month. We calculated monthly RSV hospitalization incidence rates for high risk children in the Medicaid claims dataset as follows: the denominator was defined as the number of subject months; the numerator as the number of subject months whe re at least one RSV related hospital admission identified by inpatient claims occurred. As
48 described above in part I, a corrected incidence rate (here monthly) was calculat ed to account for a reduced RSV incidence as a consequence of immunoprophylaxis. We calculated the NNT according to equation 3 5 : (3 5 ) The absolute risk reduction (ARR) was calculated based on the incidence rate multiplied with the RRR, for which we assumed the effectiveness estimate of 50% as described in part I. For eac h stat e, we created a matrix of monthly RSV incidence rates and of monthly NNTs for three age categories: 0 6 months 7 12 months and 13 24 months We also calculated RSV incidence rates and NNT for the regions in Florida, but we collapsed the age categories into a single category since the case numbers were too small for a more detailed analysis.
49 Figure 3 1 Season detection based on clinical d ataset For each week: Season according to RSV hospitalizations Yes No Season according to surveillance Yes A True positive B False positive PPV: A/(A+B) No C False negative D True negative NPV: D/(C+D) Sensitivity: A/(A+C) Specificity: D/(B+D) Figure 3 2 Calculation of test c haracteristics
50 Area under the curve 1 Perfect 0.9 <1 Excellent 0. 8 <0.9 Good 0.7 <0.8 Fair 0.6 <0.7 Poor <0.6 Fail Figure 3 3 Cut off values for areas under the ROC c urve
51 CHAPTER 4 RESULTS Sample C haracteristics After applying all inclusion and exclusion criteria, the final sample consisted of a total of 109,665, 551 subject weeks from 2,654,647 children (figure 4 1). In California, the final FFS sample with 4 weeks of ambulatory care preceding the current week represent ed 40.4% of the original population of all Medicaid recipients born between 1999 and 2004 In Fl orida, birth certificates were match ed to 79.1% of subjects and the final sample represent ed 55.6% of the original population In Illinois and Texas, the final sample retain ed 77.4% and 62.0% of the original population s respectively The number of contin uously eligible weeks in California was smaller than in the other states suggesting that with increasing age, more children were moved to managed care and therefore los t study eligibility which is also evidenced by the younger average age in California c ompared to other states (table 4 1 ) Results from the logistic regression model based on the Florida dataset confirm that the AAP indications CLD, prematurity I and I I, and CHD are associated with a significant increase in risk for RSV (table 4 2 ). C ystic fibrosis Down syndrome asthma and i mmunosuppression were also associated with an increased risk for RSV. T ransplant s, m alignancy and i mmunodeficiency d id not reach statistical significance since their 95% confidence intervals included unity ; however cas e numbers were small. Since all the above factors ha d point estimates above 1 and may therefore be associated with increased risk, we categorized subjects with any of the se risk factors into the cohort of elevated risk if they did not fall into the high ri sk cohort C ohort characteri stics are presented in table 4 1 Briefly, the proportion of males was similar across states at about 51% but the racial compositions differ ed widely. While 61.9% and
52 65.4% of the California and Texas sample were Hispanic, this applies to less than a third of the sample s in Florida and Illinois, where the proportions of Whites and Blacks were higher than in California or Texas. The average age was comparable in Florida, Illinois and Texas; but age was lower in California. Betwee n the four states, differences in the incidence of RSV risk factors were present. A ll risk factors except Down syndrome were less prevalent in California, an observation that may be related to the younger average age if some of the risk factors bec o me appa rent later in life Results of the incidence s of RSV hospitalizations and palivizumab exposure by risk category can be found in table 4 3. Consistent with the description of sample demographics, we f ou nd fewer children at high risk in California. Florida ha d the largest proportion of high risk children (2.51%) which was a consequence of including prematurity as a risk factor only in Florida. The highest incidence of palivizumab exposure among high risk children was also observed in Florida, potentially as a result of longer seasons. The overall incidence rate s of RSV hospitalizations were fairly consistent between the states ranging from 0.09% of high risk weeks with an RSV hospitalization in Illinois to 0.12% in Florida and Texas. A similar pattern was obs erved for children at elevated risk, their proportion was smallest in California and largest in Florida, which also ha d the highest rate of palivizumab injections in this cohort T he number of RSV hospitalizations was again similar between the states, rang ing from 0.04% of weeks at elevated risk in California and Illinois to 0.05% in Florida and Texas. Finally, the largest proportion of children at low risk was found in California, palivizumab exposure among low risk children was highest in Texas and the in cidence of RSV hospitalizations was again similar across states at 0.03% in Illinois and 0.04% in the other states.
53 Part I: efinition Weekly incidence rates of RSV hospitalizations for the high risk and low risk coh ort s together with the resulting RSV seasons are plotted in figure 4 2 for each state. We observe d a pattern of very distinct seasons in California with almost no activity outside of the season. Florida experienced longer and less regular seasons with resi dual activity outside the seasons. RSV seasons in Illinois and Texas seem ed more consistent over time compared to Florida Texas show ed some activity outside the season although much less pronounced than in Florida. Figure 4 3 present s the weekly RSV inci dence rates and seasons based on hospitalizations for all children regardless of risk category in the regions of Florida after applying the season al incidence thresholds derived from the high risk/low risk comparison on the state level The northwest, nort h and central regions show ed a distinct seasonal pattern with few cases off season. The southwest show ed a less regular pattern with periods of lower activity that were nevertheless considered on season. Finally, the southeast region experienced t he leas t pronounced seasonality: p eriods of high er and lower activity were detectable, yet again some periods with lower activity were still considered on season. For each of the Florida regions, the incidence rate of RSV hospitalizations reached the highest peak in the fi r st study season (99/01) with smaller peaks thereafter Figure 4 4 summarizes the surveillance dataset for the 4 states. In the figure, the 10% MPP threshold is h ighlighted in red and the optimal threshold as explained below in green Only 4 seasons are displayed for Florida since state specific surveillance was initiated later. Observations correspond to results of the RSV seasonality according to hospitalizations. California experienced very concise seasons with almost no activity outside of the season despite year round testing. Florida experienced longer seasons with some activity outside of the season. Illinois and Texas also show ed a distinct seasonal pattern; however some outliers with a high
54 MPP outside of the season could be traced ba ck to single positive tests during periods w h ere testing was infrequent Figure 4 5 shows results of surveillance in the regions in Florida. Only 3 seasons were available in the northwest and 4 in each of the remaining regions. This figure illustrates the difficulty of establishing a clear cut season when the frequency of tests was low, for example in the north and northeast regions. Here, off season periods coincide d with a number of outliers with a high MPP that were again a consequence of few positive t ests at times of rare testing. Conversely, in the southwest and southeast regions, tests were performed year round which reduced the impact of outliers outside the season. Nevertheless, the southeastern region did experience some RSV activity year round ev en though the 10% MPP threshold was not always reached. R esults of the ROC analysis are presented in figure 4 6 for the 4 states and in figure 4 7 for the regions in Florida. Highlighted are specificity/1 sensitivity pairs for the currently used 10% thres hold and the point that optimizes sensitivity and specificity. The corresponding p oint estimates and 95 % confidence intervals of AUCs are listed in table 4 4. With the exception of Illinois, t he statewide AUCs reach ed the threshold of excellence (>0.9). I n Florida, none of the regional AUCs reach ed the statewide AUC. In the north, southwest and southeast region s AUCs were statistically significantly lower than the statewide estimate. All regional estimates f e ll into of the southeast, which perform ed only fair with the Table 4 5 shows further test characteristics of the approach of using 10% MPP to identify RSV seasons. Here again, statewid e tests perform ed well with sensitivity and specificity exceeding 0.70, however the north, central southeast and southwest regions of Florida
55 experienced poor test specificity indicating that approximately half of the off season weeks were categorized as on season by using the 10% MPP threshold as the sole criterion Optimal MPP thresholds as a result of the ROC analysis are presented in table 4 6. With the exception of Florida they lie below the 10% threshold. Both s ensitivity and specificity were improv ed for the regions in Florida, yet sensitivity in the southeast and specificity in the north region s were still less than excellent The small NP V value for the southeast and southwest regions did not improve with the use of the optimal threshold. The smal l NP V indicates that many weeks categorized as off season were in fact on season weeks but since the number of off season weeks was small in these regions the absolute amount of misclassification by the surveillance system is limited To evaluate the va lidity of the NREVSS for indicating RSV seasons, we investigated how the definition has been applied, specifically with the requirement that two consecutive weeks ha d to exceed the threshold to start a season. Table 4 7 shows the accuracy of this approach and reports the mean difference and direction of difference between season onset according to NREVSS compared to onset based on RSV hospitalizations. The currently used 10% threshold differ ed on average by 4.6 weeks (95% CI, 0.2 9.1 ) from the actual seas on onset in the states T his difference was not systematic as evidenced by the direction of 0.6 weeks which means that on average NREVSS onset precede d the RSV hospitalization onset by 0.6 weeks. In fact, none of the definitions differ ed systematically i n a certain direction when applied statewide as indicated by direction estimates for the mean statewide difference that d id not exceed 1 week in either direction. In the regions of Florida, season onset according to the current season definition differ ed b y 6.1 weeks (95% CI, 0.3 11.8 ) from the actual onset and preceded i t on average by 4.6 weeks. Using the optimal threshold for each state and region did not lead to major
56 improvement, however adding further requirements did. Using the 10% MPP threshold in addition to requiring at least 5 positive tests in a given week for a surveillance area improved accuracy to a mean difference of 3.4 weeks (95% CI, 0.0 7.2 ) in the states and 4.5 weeks (95% CI 0.0 9.8 ) in the regions of Florida. Using the optimal MPP threshold with the requirement of 5 positive tests led to a further but marginal improvement to 3.2 weeks (95% CI, 0.0 7.0) in states and 3.9 weeks (95% CI, 0.0 9.3 ) Nevertheless, t hese differences d id not reach statistical signi ficance. Table 4 8 shows a similar pattern for the prediction of season offset with the added requirement of 5 positive tests performing the best but differences between the definitions were less pronounced. Of not e with a mean difference exceeding 6 week s the definitions were performing much poorer in the regions of Flor ida compared to the state level. A ll definitions except the current CDC definition predict ed an earlier offset compared to the actual o ff set in the regions of Florida Part II: RSV Epide miology between Four US States and Five Regions in Florida Table 4 9 provides evidence for a pronounced difference in the seasonality indices between states California and Illinois ha d a large seasonality index, indicating that the risk of an RSV infectio n during a season was more than 12 times the risk of an infection outside of a season. The risk for an on season infection was 9.7 times (95% CI, 9.0 10.1) higher in Texas and only 3.6 times (95% CI, 3.3 3.8 ) higher in Florida. The latter was a result of a low on season activity in Fl orida (0.060 RSV hospitalizations per 100 subject weeks ; 95% CI, 0.059 0.062) combined with a high off season activity (0.017; 95% CI, 0.016 0.018). The on season activity in Florida only reache d half of the activity in California and Texas, a difference that was statistically significant as evidenced by non overlapping confidence intervals. Similarly, the off season activity in Florida was significantly higher than in any other state. The regions in Florida differed wid ely with regard to their seasonality indices; however this difference was largely driven by
57 differences in the off season activity. The northwest and north regions ha d an off season activity comparable with California and Texas resulting in a seasonality i ndex in the northeast that was only slightly below the Texas estimate Moving south in Florida increase d the off season RSV activity to 0.023 RSV hospitalizations per 100 subject weeks (95% CI, 0.020 0.026 ) in the southeast leading to a seasonality index of only 2.5 (95% CI, 2.2 2.8 ) which suggest s that the risk for an RSV infection during a season was only increased by 2.5 fold compared to off season periods. Table 4 1 0 shows how seasons var ied within geographic areas between different years. With the exception of Florida, variation in season onset was limited to about 2 months or nine weeks, but the upper limit of the 95% confidence interval for the range of season onset well exceed ed the null hypothesis threshold of 8 weeks and we can conclude that t here was significant variation in season onset in each state and each region in Florida. With the exception of California, peak weeks show ed a similar picture. In California peak weeks varied over a range of only 4 weeks ( upper limit, one sided 95 % confid ence interval: 8 ) indicating a fairly stable peak activity. Also, RSV season offset in California was fairly stable: it varied only over 4 weeks however with a larger confidence interval. Between onset, offset and peaks, we observe d the largest statewide v ariation in Florida. A large contributor to this effect were t he central southwest and southeast regions of Florida which show ed a pronounced variation with regard to season offset, reaching a range of 21 weeks in the southwest. Part III: L atitude as a F actor in RSV Epidemiology in Florida Differe nces in the RSV seasons between the regions of Florida were pronounced. Table 4 11 indicates that all season parameters ex cept peak incidence rate differed significantly between the regions. The earliest season o nset was observed in the southeast region for the 28 th week of the year (95% CI, 26.4 30.8) the latest onset occurred in the northwest on average during the
58 45 th week (95% CI, 43.0 47.4). T he earliest offset was observed in the north (8.2; 95% CI, 3.0 13.4) and th e latest in the southwest (20.4; 95% CI, 15.2 25.6 ). Seasons were shortest in the north with an average duration of 20.6 weeks (95% CI, 14.5 26.7 ) and longest in the southeast with an average duration of 42.8 weeks (95% CI, 36.7 48.9 ) The southeast also experience d an early peak week (39.6 ; 95% CI, 37.3 41.9 ) while the other regions peak ed closer to the end of the year. Results from the linear regression analysis show that latitude was a strong linear predictor for these differences. With each degree increase in latitude, season onset occurred 3.23 weeks later (95% CI, 2.49 3.97 ) (table 4 1 2 ). Moving north one degree in Florida was associated with a 2.10 weeks earlier offset (95% CI, 3.55 ( 0.65 )) These effects add ed up to a 5.03 weeks shorter season (95% CI, 6.61, ( 3.45 )) for each degree north. The peak week occur red 2.31 weeks later (95% CI, 1.50 3.12 ) for each degree north but the peak incidences did not change (model R 2 =0.00). These associations ar e illustrated in figure 4 8 Part IV : Patient F actors and S easonality Of the 4 interaction terms tested, 2 showed statistical significance : age 7 12 months and CLD. While age 7 12 months was associated with a 25% higher risk for RSV hospitalization s on season vs. off season (OR= 1.25 ; 95% CI, 1.13 1.39), CLD was associated with a 35% lower risk during seasons compared to off season (OR=0.65 ; 95% CI, 0.53 0.80). Of note, this is not an indicator that CLD lower ed the risk for RSV during seasons ; it merely means that the effect of CLD on RSV risk was stronger outside of the season. The status of the tested variables as risk factors for RSV was confirmed and figure 4 9 which shows for all risk factors, independent of the RSV season that the ir odds ratios exceed ed one The figure also shows that being in the youngest age category was associated with a similar risk on season and off season, which also applies to CHD. The significant interaction terms between age 7 12 months and RSV season and between CLD and RSV season are reflected in the difference in point estimates with
59 non overlapping confidence intervals. During season, the odds ratio for age 7 12 months was 2.88 (95% CI, 2.71 3.05 ) and 2.30 (95% CI, 2.09 2.53 ) off season. CLD had an on season odds ratio of 2.46 (95% CI, 2. 24 2.67 ) which increased to 3.79 (95% CI, 3.14 4.59 ) outside of the RSV season. Finally, an important observation in figure 4 9 relates to the absolute extent of RSV risk associated with each risk factor. Belonging to the youngest age category increase d the risk for RSV by more than 7 times compared to age 13 24 months while the other risk factors had a much lower influence on the RSV risk. Part V: Timing of Prophylaxis with Palivizumab vs. RSV Seasonality Figures 4 10 and 4 11 illustrate patterns of p alivizumab utilization contrasted to the incidence rate of RSV hospitalization s for each of the states and for the regions of Florida. A consistent utilization pattern could be observed on a state level with little variation between the years. An exception was the first season in Texas which we excluded from further analysis. Its distinct pattern seems to illustrate slow acceptance after market introduction of palivizumab and was therefore not considered to be a valid comparator for the following seasons. T he northwest and north regions of Florida and to some extent the central region exhibit ed a similar utilization pattern compared to the states The southwest show ed a less distinct pattern with some utilization year round, however season s were still recogn izable. In the southeast, utilization occurred y ear round in the later seasons and although the curves show ed some degree of a seasonal pattern they allow the estimation of an onset or offset in utilization. Therefore the southeast of Florida was omitted from further utilization analysis. Tables 4 1 3 and 4 1 4 quantify the relationship between utilization and RSV activity. The onset of utilization in the 4 states was closes t to a fixed date, namely the first week of October in Florida and Illinois a nd the first week of November in California and Texas
60 In all regions of Florida, onset of utilization was closes t to early October. The direction of the difference was very close to 0.0 suggesting a small random distribution around the first week of Octo ber. Actual season onset according to RSV hospitalizations and onset estimates according to surveillance data were further remote from onset of utilization compared to the fixed dates. With the exception of the central and southwest regions of Florida uti lization started before the onset of RSV hospitalizations. The results for offset were similar with the exception of Florida and Illinois w h ere utilization offset was closes t to the offset of RSV hospitalizations Nevertheless both estimates were very clo se to the fixed dates of the last week of March and the last week of April. Utilization offset was observed in California and Illinois in the last week of April. Offset in the regions of Florida coincide d with the fixed dates of late March or April with th e exception of the northwest where offset occurred closest to the offset of RSV hospitalizations Yet again, the fixed date of last week of March was almost equally close. Overall, RSV season offset precede d the offset of palivizumab utilization in all sta tes with the exception of the state of Florida and in Florida only in the south we st. Part V I : Optimizing T iming of P rophylaxis Incidence rates of RSV hospitalizations and corresponding NNTs by calendar month and age category are displayed in figures 4 12 through 4 15 for each of the 4 states. Figure 4 1 6 illustrate s regional RSV incidence rates and NNTs in Florida with age categories collapsed into a single category C olor coding of the figures was based on threshold s that facilitate orientation but no NN T thresholds have been established in the literature Coding thresholds were consistent between t he figures to allow cross state comparisons of burden of disease. Consistent with our findings in part I, we observe d a very distinct season in California (fig ure 4 12 ), and we f ou nd the high seasonality index from part II confirmed in the high RSV hospitalization rates on season while almost no activity was measured outside of the season. We c ould further observe a clear
61 age pattern with the oldest age category being at much lower risk for infections compared to the youngest age category. In Florida, the seasonal pattern was less distinct (figure 4 13 ), however months with high activity were still distinguishable from months of lower activity, but unlike in Cali fornia, we dot not identify months with virtually no activity. The age difference in risk for RSV infections was also pronounced in Florida with the oldest age category being at comparably low risk. Illinois and Texas (figures 4 1 4 and 4 1 5 ) exhibit ed a si milar pattern a s California, with distinct periods of high and low RSV activity. Comparing NNTs across states, we f ou nd that immunoprophylaxis with palivizumab was less beneficial in Florida where NNTs were never below 100 regardless of age or calendar mon th. NNTs below 200 were found even for the oldest age category from 13 2 4 months in all states except Florida where the lowest NNT in this age group was 252, followed by 320 as the second lowest. A n examination of the regional RSV incidence rates and NNTs (figure 4 1 6) reveals that the statewide picture in Florida was merely a blend of a very different burden of disease in the regions. Although the northwest and north region s show ed a pattern more similar to the other states, even here we find months with virtually zero activity as we f ou nd in the other states, most pronounced in California. The southwest and southeast regions exhibit ed a pattern of prolonged viral activity, most obvious in the southeast. However even in the southeast, we were able t o distinguish between months of relatively high and relatively low activity. April through July in the southwest region and May and June in the southeast show NNTs exceeding 650 while the winter months ha d a peak activity that was comparable to other regio ns at their peak months.
62 Table 4 1 Cohort c haracteristics Variable California Florida Illinois Texas Total subject weeks Demographics 16,617,845 19,903,113 29,252,601 43,891,992 Sex, male Race /Ethnicity 8,466,264 (50.9) 10,168,281 (51.1) 14,942,4 50 (51.1) 22,375,862 (51.0) White Black Native American Asian Hispanic Other Age [months ] 3,667,695 (22.1) 1,106,694 (6.66) 127,637 (0.77) 516,133 (3.17) 10,286,924 (61.9) 912,762 (5.50) 7.97 (6.56) 6,851,466 (34.4) 5,711,267 (28.7) 10,856 (0.05) 148,503 (0.75) 5,683,232 (28.6) 1,497,789 (7.53) 10.47 (6.66) 10,244,069 (35.0) 8,428,941 (28.8) 57,983 (0.20) 708,637 (2.42) 9,346,077 (32.0) 466,894 (1.59) 11 33 (6.66) 9,596,615 (21.9) 5,019,089 (11.4) 130,669 (0.30) 313,256 (0.71) 28,688,524 (65.4) 143,839 (0. 33) 10.34 (6.57) RSV risk factors Prematurity I n/a 117,241 (0.59) n/a n/a Prematurity II n/a 99,024 (0.50) n/a n/a Prematurity III Chronic lung disease Congenital heart disease Cystic fibrosis Immunodeficiency Down syndrome Asthma Transplant Malig nancy Immunosuppression n/a 41,894 (0.25) 72,759 (0.44) 5,396 (0.03) 9,179 (0.06) 28,814 (0.17) 530,733 (3.19) 2,126 (0.01) 46,988 (0.28) 1,135,201 (6.83) 297,633 (1.50) 127,706 (0.64) 234,441 (1.18) 15,782 (0.08) 45,534 (0.23) 38,633 (0.19) 1,334,749 (6.7 1) 3,243 (0.02) 179,291 (0.90) 2,878,774 (14.5) n/a 118,490 (0.41) 303,589 (1.04) 17,571 (0.06) 17,891 (0.06) 43,114 (0.15) 1,328,029 (4.54) 4,606 (0.02) 282,293 (0.97) 3,186,507 (10. 9 ) n/a 234,136 (0.53) 409,400 (0.93) 12,273 (0.03) 22,445 (0.05) 78,312 ( 0.18) 2,983,022 (6.80) 3,669 (0.01) 61,948 (0.14) 6,844,538 (15.6) *weeks are preceded by a 4 weeks ambulatory care period, table shows number of subject weeks (percentages) for categorical variables and mean (standard deviation) for age.
63 Table 4 2 Ri sk factors for RSV hospitalization in Florida Factor Comparison OR [95% CI] Factor Comparison OR [95% CI] Year Calendar month Age [months] Sex Race 1999 vs 2004 2000 vs 2004 2001 vs 2004 2002 vs 2004 2003 vs 2004 1 vs 12 2 vs 12 3 vs 1 2 4 vs 12 5 vs 12 6 vs 12 7 vs 12 8 vs 12 9 vs 12 10 vs 12 11 vs 12 0 6 vs 13 24 7 12 vs 13 24 F vs M Black vs White Native American vs White Asian vs White Hispanic vs White Unknown vs White 1.17 [1.07 1.30] 1.08 [1.01 1.17] 1.11 [1.03 1.20] 1.14 [1.06 1.22] 1.06 [0.99 1.14] 0.68 [0.63 0.74] 0.56 [0.51 0.61] 0.40 [0.36 0.44] 0.22 [0.19 0.25] 0.16 [0.13 0.18] 0.13 [0.11 0.15] 0.17 [0.15 0.20] 0.29 [0.26 0.33] 0.51 [0.47 0.57] 0.84 [0.77 0.91] 1.05 [0.97 1.13] 10.10 [9.32 10.94] 2.61 [2.39 2.84] 0.82 [0.78 0.85] 1.10 [1.04 1.16] 0.76 [0.25 2.37] 0.63 [0.45 0.90] 1.28 [1.21 1.35] 1.21 [1.10 1.32] Birth month Prematurity I Prematurity II Prematurity III Chronic lung disease Congenital heart disease C ystic fibrosis Immunodeficiency Down syndrome Asthma Transplant Malignancy Immunosuppression Current p alivizumab exposure 1 vs 12 2 vs 12 3 vs 12 4 vs 12 5 vs 12 6 vs 12 7 vs 12 8 vs 12 9 vs 12 10 vs 12 11 vs 12 0.92 [0.82 1.03] 0.81 [0.71 0.91] 0.7 0 [0.62 0.80] 0.62 [0.54 0.70] 0.65 [0.57 0.73] 0.61 [0.54 0.69] 0.71 [0.63 0.79] 0.79 [0.71 0.87] 0.87 [0.79 0.97] 0.99 [0.89 1.10] 1.06 [0.95 1.17] 1.44 [1.20 1.74] 1.79 [1.51 2.11] 1.73 [1.56 1.91] 2.70 [2.27 3.22] 2.16 [1.90 2.48] 2.07 [1.24 3.45] 1.17 [0.72 1.88] 2.76 [2.08 3.66] 1.94 [1.78 2.12] 1.97 [.063 6.15] 1.10 [0.86 1.14] 2.08 [1.93 2.24] 0.95 [0.81 1.11] Abbreviations: OR: odds ratio, CI: confidence interval
64 Table 4 3. Palivizumab exposure and R SV hospitalizations by s tate and risk c ategory Risk categories California Florida Illinois Texas Total subject weeks High risk Palivizumab exposure ** Palivizumab doses RSV hospitalization s Elevated risk Palivizumab exposure Palivizumab doses RSV h ospitalization s Low risk Palivizumab exposure Palivizumab doses RSV hospitalization s 16,617,845 105,113 (0.63) 9,889 (9.41) 2,298 (2.19) 108 (0.10) 1,347,922 (8.11) 5,848 (0.43) 1,288 (0.10) 550 (0.04) 15,164,810 (91.3) 28,944 (0.19) 7,556 (0.05) 5,5 06 (0.04) 19,903,113 500,060 (2.51) 89,043 (17.8) 21,149 (4.23) 607 (0.12) 3,576,688 (18.0) 42,469 (1.19) 10,006 (0.28) 1,762 (0.05) 15,826,365 (79.5) 36,381 (0.23) 8,334 (0.05) 5,727 (0.04) 29,252,601 388,337 (1.33) 25,797 (6.64) 5,826 (1.50) 367 (0.0 9) 3,698,822 (12.6) 9,846 (0.27) 2,111 (0.06) 1,498 (0.04) 25,165,442 (86.0) 42,790 (0.17) 9,724 (0.04) 6,802 (0.03) 43,891,992 592,460 (1.35) 50,995 (8.61) 11,599 (1.96) 695 (0.12) 7,607,583 (17.3) 29, 076 (0.38) 6,418 (0.08) 4,042 (0.05) 35,691,949 ( 81.3) 142,725 (0.40) 33,711 (0.09) 15,493 (0.04) *weeks are preceded by a 4 weeks ambulatory care period; table shows numbers (percentages) **palivizumab exposure occurs when a dose was given between 28 days before and t date
65 Table 4 4 Areas under the curve by state and r egion State/region Area under the curve [95% CI] States California Florida Illinois Texas Florida regions Northwest North Ce ntral Southwest Southeast 0.98 [0.9 6 0.99] 0.9 2 [0. 88 0.9 5 ] 0. 88 [0. 83 0. 92 ] 0.9 2 [0.88 0.9 5 ] 0. 88 [0. 83 0.9 3 ] 0.8 0 [0. 75 0. 86 ] 0. 88 [0. 84 0.9 2 ] 0.8 3 [0. 78 0. 88 ] 0. 76 [0. 68 0. 85 ] Abbreviation: CI: confidence interval
66 Table 4 5. Test characteristics at the threshold of 10% median proportion positive laboratory t ests State/region Threshold Sensitivity [95% CI] Specificity [95% CI] PPV [95% CI] NPV [95% CI] States California Florida Illinois Texas Florida regions Northwest North Centra l Southwest Southeast 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 0.81 [0.70 0. 88 ] 0. 88 [0.83 0.92 ] 0. 83 [0. 74 0. 89 ] 0.9 1 [0.8 4 0.9 6 ] 0. 88 [0. 79 0.9 3 ] 0. 88 [0. 81 0.9 3 ] 0.9 6 [0.9 2 0.98 ] 0.7 5 [0. 69 0. 81 ] 0.8 1 [0. 75 0. 86 ] 0.95 [0.9 1 0.97] 0.7 1 [0.6 0 0.8 0 ] 0. 82 [0. 76 0.8 7 ] 0. 80 [0.7 4 0.8 6 ] 0. 79 [0.7 0 0.8 6 ] 0. 56 [0. 47 0. 63 ] 0. 5 8 [0. 4 9 0. 67 ] 0. 6 7 [0. 55 0. 77 ] 0.5 1 [0. 37 0.6 5 ] 0.8 7 [0.77 0.93 ] 0. 88 [0.8 2 0.9 2 ] 0. 72 [0. 62 0.7 9 ] 0.7 2 [0.6 3 0.7 9 ] 0. 77 [0. 68 0. 85 ] 0.5 8 [0. 51 0.6 6 ] 0.7 5 [0.6 9 0. 81 ] 0. 86 [0.8 0 0.9 1 ] 0. 89 [0.8 3 0.93] 0.9 3 [0. 88 0.96 ] 0.7 2 [0.6 1 0.8 1 ] 0.9 0 [0.8 4 0.9 4 ] 0.9 4 [0. 89 0.9 7 ] 0. 89 [0. 81 0. 94 ] 0. 87 [0. 79 0.9 2 ] 0.9 3 [0. 84 0.97 ] 0.5 1 [0.4 0 0.6 1 ] 0. 37 [0. 26 0. 49 ] *High lighted are point estimates <0.7. Abbreviations: PPV: positive predictive value, NPV: negative predictive value CI: confidence interval Table 4 6. Test c haracteristics at optimal thresholds of median proportion positive laboratory t ests State/region Th reshold Sensitivity [95% CI] Specificity [95% CI] PPV [95% CI] NPV [95% CI] States California Florida Illinois Texas Florida regions Northwest North Central Southwest Southeast 5 .5 13.2 8.3 9.5 13.6 16.7 14.8 12.0 1 3.6 0.9 5 [0.8 7 0.9 8 ] 0.8 1 [0.7 4 0 86 ] 0.8 9 [0. 81 0.9 4 ] 0.92 [0.85 0.96] 0. 84 [0. 75 0.9 0 ] 0.7 6 [0. 67 0.8 4 ] 0.8 6 [0.7 9 0. 91 ] 0.7 1 [0. 64 0. 77 ] 0. 68 [0. 62 0. 74 ] 0. 88 [0.8 2 0.9 2 ] 0.9 0 [0. 82 0.9 5 ] 0.8 0 [0. 74 0. 86 ] 0.8 0 [0. 74 0. 86 ] 0. 86 [0. 78 0.9 1 ] 0. 64 [0.7 9 0.85] 0 75 [0. 66 0. 83 ] 0. 85 [0.7 5 0.9 2 ] 0.79 [0.65 0.89] 0.7 5 [0. 65 0.8 3 ] 0.9 5 [0.9 1 0 .98 ] 0. 71 [0. 62 0. 79 ] 0.7 2 [0. 64 0.8 0 ] 0.8 3 [0. 74 0.9 0 ] 0 .66 [0. 57 0. 74 ] 0. 82 [0. 75 0. 88 ] 0.9 3 [0. 87 0.9 6 ] 0. 94 [0. 89 0.97] 0.98 [0.9 4 0.99] 0. 67 [0 57 0. 75 ] 0.9 3 [0.8 8 0.9 6 ] 0.95 [0.9 0 0.98] 0. 87 [0. 79 0.9 2 ] 0.8 1 [0.7 4 0. 87 ] 0.80 [0.7 1 0.8 7 ] 0. 52 [0. 43 0. 61 ] 0. 35 [0. 26 0. 45 ] *Highlighted are point estimates <0.7. Abbreviations: PPV: positive predictive value, NPV: negative predictiv e value CI: confidence interval
67 Table 4 7. Mean of absolute differences and direction of difference between season onset according to clinical dataset and surveillance dataset under different definitions for season onset State/region 10% MPP Optimal MPP 10% MPP, Optimal MPP, 10% MPP, Optimal MPP, p value States California Florida Illinois Texas Florida regions Northwest North Central Southwest Southeast Mean states Mean regions 3.2 [0.8 5 .6] + 0.4 4.8 [0 .0 12.1] +1.8 4.2 [2.0 6.4] 0.6 6.4 [0.8 12.0] 4.0 3.7 [2.9 4.4] +1.0 n/a* 9.3 [5.5 13.0] 9.3 5.0 [2.7 7.3] 0.0 7.0 [0 .0 18.0] 7.0 4.6 [0.2 9.1] 0.6 6.1 [0.3 11.8] 4.6 3.2 [0 .0 6.4] 2.0 4.3 [0 .0 11.5] +3.8 4.4 [2.1 6.6] 1.6 6.4 [0.8 12.0] 4.0 3.7 [2.9 4.4] +1.0 n/a* 5.3 [0 .0 12.1] 4.8 4.5 [2.3 6.6] +4.5 5.8 [0 .0 17.6] 4.8 4.6 [0.1 9.1] 1.0 4.8 [0 .0 10.7] 2.2 3.2 [0.8 5.6] + 0.4 4.8 [0 .0 12.1] +1.8 3.4 [0.2 6.6] +0.2 5.0 [0 .0 10.2] 2.6 3.7 [2.9 4.4] +1.0 1.8 [0 .0 4.5] 0.8 7.3 [3.3 11.2] 7,3 4.8 [2.2 7.3] +0.3 7.0 [0 .0 18.0] 7.0 4.1 [0 .0 8.4] 0.1 4.9 [0 .0 10.5] 3.8 3.2 [0 .0 6.4] 2.0 4.3 [0 .0 11.5] +3.8 2.8 [0.4 5.2] 0.0 5.0 [0 .0 10.2] 2.6 3.7 [ 2.9 4.4] +1.0 1.5 [0 .0 3.1] +1.5 5.3 [0 .0 12.1] 4.8 4.5 [2.3 6.6] +4.5 5.8 [0 .0 18.6] 4.8 3.8 [0 .0 8.1] 0.2 4.2 [0 .0 10.0] 1.5 3.0 [0.8 5.2] + 0.2 4.8 [0 .0 12.1] +1.8 3.0 [0 .0 6.2] +0.6 3.0 [0 .0 6.2] 0.6 4.7 [3.2 6.1] +2.0 2. 0 [0 .0 6.2] +2.0 4.5 [3.8 5.2] 4.5 4.5 [1.5 7.5] +0.5 7.0 [0 .0 18.0] 7.0 3.4 [0 .0 7.2] +0.5 4.5 [0 .0 9.8] 2.6 3.2 [0 .0 6.4] 2.0 4.3 [0 .0 11.5] +3.8 2.6 [0.3 4.9] +0.2 3.0 [0 .0 6.2] 0.6 4.7 [3.2 6.1] +2.0 3.0 [0 .0 6.7] +3.0 1.8 [0 .0 4.5] 0.8 4.5 [2.3 6.6] +4.5 5.8 [0 .0 17.6] 5.3 3.2 [0 .0 7.0] +0.3 3.9 [0 .0 9.3] 0.5 0.999 0.999 0.681 0.567 0.366 0.853 0.158 0.998 0.999 0.730 0.797 Table 4 7 shows the absolute mean difference in weeks [95% confidence interval] and direction of difference, direction is positive if onset according to definition occurred after onset of RSV hospitalizations; *season definition not applicable due to indistinctive pattern ; Abbreviation: MPP: median proportion of positive te sts
68 Table 4 8 Mean of absolute differences and direction of difference between season offset accord ing to clinical dataset and surveillance dataset under different definitions for season offset State/region 10% MPP Optimal MPP 10% MPP, Optimal MPP, 10% MPP, Optimal MPP, p value States California Florida Illinois Texas Florida regions Northwest North Central Southwest Southeast Mean states Mean regions 1.8 [0.8 2.8] +0.2 3 .0 [0 .0 6.9] +1.0 2.8 [0.4 5.2] +2.8 4.2 [1.1 7.3] +4.2 2.3 [0.4 4.2] +1.0 n/a* 2.0 [0 .0 4.7] +1.5 8.5 [1.0 16.0] 4.0 8.8 [1.5 16.0] 8.8 2.9 [0.2 5.7] +2.1 6.2 [0.0 12.6] +0.1 2.6 [0 .0 6.0] +2.2 2.0 [0.2 3.8] 1.0 3.2 [0.8 5.6 ] +3.2 4.2 [1.1 7.3] +4.2 2.7 [0.1 5.3] 2.0 n/a* 5.8 [0.5 11.0] 5.8 7.8 [0 .0 17.8] 7.3 13.5 [2.8 24.1] 13.5 3.1 [ 0 3 5.8] +2.2 7.2 [0.0 15.7] 6.2 1.8 [0.8 2.8] +0.2 3.0 [0 .0 6.9] +1.0 2.0 [0.8 3.2] +0.4 2.8 [0.8 4.8] +2.8 2. 3 [0.4 4.2] +1.0 8.0 [3.6 12.4] +1.0 2.0 [0 .0 4.7] +1.5 8.5 [1.0 16.0] 4.0 8.75 [1.5 16.0] 8.8 2.4 [0.2 4.5] +1.1 6.1 [0.0 12.3] 1.9 2.6 [0 .0 6.0] +2.2 2.0 [0.2 3.8] 1.0 1.6 [0.5 2.7] +0.4 2.8 [0.8 4.8] +2.8 2.7 [0.1 5.3] 2. 0 5.3 [0 .0 10.7] 4.8 5.8 [0.5 11.0] 5.8 7.8 [0 .0 17.8] 7.3 13.5 [2.8 24.1] 13.5 2.3 [0.1 4.4] +1.1 7.2 [0.0 15.4] 6.6 1.8 [0.8 2.8] +0.2 3.0 [0 .0 6.9] +1.0 1.8 [0.2 3.4] 1.4 2.2 [0.4 4.0] +2.2 2.7 [0.1 5.3] 2.0 9.0 [0 .0 23 .2] 5.5 1.5 [0 .0 4.5] +1.0 8.5 [1.0 16.0] 4.0 8.75 [1.5 16.0] 8.8 2.2 [0.0 4.3] +0.5 6.3 [0.0 14.7] 3.9 2.6 [0 .0 6.0] +2.2 2.0 [0.2 3.8] 1.0 1.8 [0.2 3.4] 1.4 2.2 [0.4 4.0] +2.2 4.0 [1.8 6.2] 4.0 9.0 [0 .0 23.3] 8.5 5.8 [0.5 11.0] 5.8 7.8 [0 .0 17.8] 7.3 13.5 [2.8 24.1] 13.5 2.2 [0.0 4.3] +0.5 8.2 [0.0 17.8] 7.8 0.941 0.888 0.444 0.367 0.882 0.914 0.158 0.999 0.771 0.507 0.904 Table 4 8 shows the absolute mean difference in weeks [95% confidence i nterval] and direction of difference, direction is positive if offset according to definition occurred after offset of RSV hospitalizations; *season definition not applicable due to indistinctive pattern ; Abbreviation: MPP: median proportion of positive te sts
69 Table 4 9 Exten t of seasonality and seasonality index in each state and regions in Florida State/region On season RSV IR [95% CI] Off season RSV IR [95% CI] Seasonality index [95% CI ] States California Florida Illinois Texas Florida regions Northwe st North Central Southwest Southeast 0.118 [0.115 0.121] 0.060 [0.059 0.062] 0.089 [0.087 0.090] 0.131 [0.129 0.132] 0.072 [0.067 0.078] 0.060 [0.055 0.065] 0.076 [0.073 0.079] 0.080 [0.076 0.084] 0.055 [0.054 0.057] 0.009 [0.009 0.0 10] 0.017 [0.016 0.018] 0.007 [0.007 0.008] 0.013 [0.013 0.014] 0.008 [0.007 0.010] 0.009 [0.008 0.011] 0.013 [0.011 0.014] 0.016 [0.013 0.019] 0.023 [0.020 0.026] 12.6 [11.8 13.4] 3.6 [3.3 3.8] 12.1 [11.4 12.8] 9.7 [9.0 10.1] 9. 0 [7.2 11.1] 6.4 [5.4 7.9] 6.0 [5.4 6.8] 5.1 [4.2 6.2] 2.5 [2.2 2.8] Abbreviation s : IR: Incidence rate [RSV hosp./100 subject weeks] CI: confidence interval
70 Table 4 1 0 Variation in seasons within each state and regions in Florida State/r egion Mean week (SD) R ange ( min max ) U pper limit, one sided 95 % CI of range Season onset States California Florida Illinois Texas Florida regions Northwest North Central Southwest Southeast Peak we e k States California Florida Illinois Texas Florida regi ons Northwest North Central Southwest Southeast Season offset States California Florida Illinois Texas Florida regions Northwest North Central Southwest Southeast 51.0 (3.45) 32.2 (4.60) 48.6 (3.13) 45.6 (2.51) 45.2 (1.48) 40.8 (1.79) 37.4 (2.88) 36.6 (2.97) 28.6 (2.07) 6.2 (1.10) 49.0 (3.39) 5.0 (2.45) 4.2 (1.92) 52 .0 (2.92) 50.2 (2.17) 48.0 (2.35) 48.4 (2.30) 39.6 (2.70) 12.4 (1.34) 15.2 (3.96) 13.8 (2.05) 11.0 (2.00) 14.2 (3.03) 8.2 (3.63) 13.6 (5.73) 20.4 (7.99) 18.2 (6.10) 9 ( 48 04 ) 12 ( 2 4 35 ) 7 (46 52) 7 (44 50) 5 (34 47) 5 (39 43) 7 (35 41) 9 (32 40) 6 (26 31) 4 (5 8) 9 (46 54) 8 (1 7) 7 (1 6) 8 (47 54) 6 (47 52) 6 (46 51) 7 (46 52) 8 (36 43) 4 (11 14) 11 (11 21) 5 (12 16) 6 (9 14) 9 (10 18) 9 (3 12) 17 (5 21) 21 (13 33) 15 (8 2 2) 25 33 23 18 11 13 21 22 15 8 25 18 14 21 16 17 17 20 10 29 15 15 22 26 41 57 44
71 Table 4 11 Comparison of season characteristics between regions in Florida Dependent variable Model R 2 p value Region Mean [95% CI] Week of onset Week of offset Season duration [weeks] Peak week Peak RSV incidence [RSV hosp /100 subject weeks] 0.87 0.41 0.66 0.78 0.27 <0.001 0.025 <0.001 <0.001 0.156 Northwest North Central Southwest Southeas t Northwest North Central Southwest Southeast Northwest North Central Southwest Southeast Northwest North Central Southwest Southeast Northwest North Central Southwest Southeast 45.2 [43.0 47.4] 40.8 [38.6 43.0] 37.4 [35.2 39.6] 36.6 [34.4 38.8 ] 28.6 [26.4 30.8] 14.2 [9.0 19.4] 8.2 [3.0 13.4] 13.6 [8.4 18.8] 20.4 [15.2 25.6] 18.2 [13.0 23.4] 24.0 [17.9 30.1] 20.6 [14.5 26.7] 29.4 [23.3 35.5] 36.8 [30.7 42.9] 42.8 [36.7 48.9] 52.0 [49.7 2.3] 50.2 [47.9 52.5] 48.0 [45. 7 50.3] 48.4 [46.1 50.7] 39.6 [37.3 41.9] 0.166 [0.115 0.217] 0.132 [0.080 0.183] 0.158 [0.107 0.209] 0.210 [0.159 0.261] 0.126 [0.075 0.177] Abbreviation: CI: confidence interval Table 4 1 2 L inear regression analysis of the effects of latitude on season characteristics in Florida Dependent variable Model R 2 p value Estimated effect of latitude [ 95% CI] Week of onset Week of offset Season duration Peak week Peak RSV incidence 0.76 0.23 0.63 0.58 0.00 <0.001 0.010 <0.001 <0.001 0.824 3.2 3 [2.49 3.97] 2.10 [ 3.55 ( 0.65 ) ] 5.03 [ 6.61 ( 3.45 ) ] 2.31 [1.50 3.12] 0.00 [ 0.02 0.01] Abbreviation: CI: confidence interval
72 Table 4 1 3 Mean of absolute differences and direction of difference between onset of palivizumab utilization and onset of RSV season according to different determinants of RSV season State/region Utilization mean [95% CI] RSV hosp. [95% CI] direction 10% MPP [95% CI] direction 1st week, October [95% CI] direction 1st week November [95% CI] direction p va lue States California Florida Illinois Texas Florida regions Northwest North Central Southwest 46.2 [44.2 48.2] 38.0 [36.0 40.0] 41.6 [40.2 43.0] 42.8 [40.2 45.3] 39.4 [36.3 42.5] 40.4 [38.5 42.3] 40.4 [37.4 43.4] 39.8 [36. 5 43.1] 4.8 [0.0 9.8] +4.8 5.8 [1.1 10.5] 5.8 7.0 [2.8 11.2] +7.0 3.8 [1.2 6.3] +3.2 5.8 [1.5 10.0] +5.8 2.0 [0.0 4.2] +0.4 4.2 [0.5 7.9] 3.0 4.0 [2.2 5.8] 3.2 5.2 [1.1 9.3] +5.2 4.8 [1.5 8.0] 4.3 6.4 [1.0 11.8] +6 .4 5.8 [3.0 8.5] 0.8 7.7 [0.0 15.7] +7.7 7.0 [5.2 8.8] 1.0 12.0 [6.9 17.0] 12.0 7.0 [2.6 11.4] 3.0 6.2 [4.2 8.2] 6.2 2.0 [0.0 4.0] +2.0 1.6 [0.2 3.0] 1.6 2.8 [0.2 5.3] 2.4 1.8 [0.0 3.8] +0.6 1.2 [0.2 2.2] 0.4 2.0 [0.8 3.2] 0.4 2.2 [0.8 3.6] +0.2 1.6 [0.0 3.9] 1.6 6.6 [5.2 8.0] +6.6 3.0 [1.2 4.8] +3.0 2.3 [0. 4 4.2 ] +2.2 5.2 [2.4 8.0] +5.2 4.2 [2.6 5.8] +4.2 4.2 [1.2 7.2] +4.2 4.8 [1.6 8.0] +4.8 0.105 0.051 0.025 0.114 0.135 <0 .001 <0.001 0.045 Table shows the absolute mean difference in weeks the smallest difference is highlighted direction is positive if onset occurred after onset of utilization ; *Florida southeast region: no onset/offset of utilization detectable ; Abbrev iations: MPP: median proportion of positive tests CI: confidence interval
73 Table 4 1 4 Mean of absolute differences and direction of difference between offset of palivizumab utilization and offset of RSV season according to different determinants of RSV season State/region Utilization mean [95% CI] RSV hosp. [95% CI] direction 10% MPP [95% CI] direction Last week, March [95% CI] direction Last week, April [95% CI] direction p value States California Florida Illinois Texas Florida regions Northwest North Central Southwest 16.0 [14.2 17.8] 14.6 [12.0 17.2] 14.8 [12.8 16.8] 15.3 [13.1 17.4] 14.2 [12.2 16.2] 14.0 [11.2 16.8] 14.4 [11.5 17.3] 16.2 [12.0 20.5] 3.6 [1.3 5.8] 3.6 2.2 [0.0 4.9] +0.6 1.4 [0.0 3 .3] 1.0 4.0 [0.3 7.7] 4.0 1.2 [0.0 2.8] 0.0 5.8 [1.0 10.6] 5.8 3.2 [0.4 6.0] 0.8 5.4 [0.0 11.0] +4.2 3.4 [0.0 6.9] 3.4 3.3 [1.7 4.8] +1.8 2.6 [0.0 5.8] +1.8 3.3 [ 0.5 6.0] +0.3 3.0 [1.8 4.2] +1.6 6.0 [6.0 6.0] +6.0 4.0 [1.0 7.0] +2.0 5.3 [0.0 11.5] +0.8 3.2 [1.2 5.2] 3.2 2.6 [1.5 3.7] 1.8 2.4 [1.0 3.8] 2.0 2. 5 [0.1 4.9] 2.5 1.8 [0.2 3.4] 1.4 2.0 [0.2 3.8] 1.2 2.4 [0.5 4.3] 1.6 4.2 [1.8 6.6] 3.4 1.2 [0.0 2.8] +1. 2 2.6 [0.2 5.0] +2.6 2.4 [0.3 4.5] +2.4 2.0 [0.2 3.8] +2.0 3.0 [1.0 5.0] +3.0 3.2 [.08 5.6] +3.2 2.8 [0.1 5.5] +2.8 2.2 [0.0 5.0] +1.0 0.230 0.828 0.725 0.606 0.162 0.104 0.714 0.500 Table shows the absolute mean difference in weeks t he smalles t difference is highlighted direction is positive if offset occurred after offset of utilization ; *Florida southeast region: no onset/offset of utilization detectable; Abbreviations: MPP: median proportion of positive tests, CI: confidence inter val
74 Figure 4 1 Flowchart of sample selection and resulting sample size
75 A B C D Figure 4 2. RSV hospitalization r ates and resulting s easons in A) California, B) Florida, C) Illinois and D ) Texas
76 A B C Figure 4 3. RSV h ospitalization r ates a nd r esulting s easons in the regions of Florida. A) N orthwest, B) N orth, C) C entral, D) S outhwest and E) S outheast
77 E F Figure 4 3. Continued
78 A B C D Figure 4 4. NREVSS laboratory tests and resulting RSV season in A) California, B) Florida, C) Illinois and D ) Texas The red line marks the 10% threshold and the green line marks the optimal threshold MPP: Median proportion of positive RSV laboratory tests.
79 A B C Figure 4 5 NREVSS laboratory tests and resulting RSV season in the regio ns of Florida. A) N orthwest, B ) North, C) Central, D) Southwest and E) S outheast The red line marks the 10% threshold and the green line marks the optimal threshold MPP: Median proportion of positive RSV laboratory tests.
80 D E Figure 4 5 Co ntinued
81 Figure 4 6 Receiver operating characteristics curves for each state
82 Figure 4 7 Receiver operating characteristics curves for each r egion in Florida
83 A B C D Figure 4 8 Linear effect of latitude on season chara cteristics in Florida. A) Week of season onset, B) Week of season offset, C) Season duration and D) Peak week
84 Figure 4 9 Heterogeneous effect s of RSV risk factors on risk for RSV hospitalizations off season vs. on se ason
85 A B C D Figure 4 10 Palivizumab utilization and RSV seasonality in A) California, B) Florida, C) Illinois and D ) Texas
86 A B C Figure 4 11 Palivizumab utilization and RSV seasonality in the regions of Florida. A) N orthwest B) N orth, C) Central, D) Southwest and E) S outheast
87 D E Figure 4 11. Continued
88 A B Figure 4 1 2 A) Monthly RSV h ospitaliza tion i ncidence r ates [per 100 subject months] B) NNT ( numbers needed to treat ) with palivizumab by age in the hi gh risk c ohort in California A B Figure 4 1 3 A) Monthly RSV hospitalization incidence rates [per 100 subject months] B) NNT (numbers needed to treat) with palivizumab by age in the high risk cohort in Florida
89 A B Figure 4 1 4 A) Monthly RSV hospitalization incidence rates [per 100 subject months] B) NNT (numbers needed to treat) with palivizumab by age in the high risk cohort in Illinois A B Figure 4 1 5 A) Monthly RSV hospitalization incidence rates [per 100 subjec t months], B) NNT (numbers needed to treat) with palivizumab by age in the high risk cohort in Texas
90 A B Figure 4 1 6 A) Monthly RSV hospitalization incidence r ates [per 100 subject months], B) NNT (numbers needed to treat) with palivizumab in t he high risk cohort for each r egion in Florida
91 CHAPTER 5 DISCUSSION This study can be divided into two major sections one section unrelated to the type of immunoprophylaxis (parts I IV) and a second one with content specific to palivizumab (parts V and VI) In the first section we used RSV hospitalization data to defi ne season onset and offset and validate the currently used surveillance system (part I). Next, we compared extent and variability of seasonality between 4 states and 5 regions of Florida (p art II). In part III, we examined difference s in seasonal characteristics as a factor of latitude in Florida P art IV was centered on the question whether patient characteristics play a role in timing of RSV infections relative to the RSV season. This firs t section was not dependent on palivizumab as an agent for immunoprophylaxis and our conclusions about RSV seasonality and the validity of the NREVSS should still be relevant when a new prophylactic agent or vaccine becomes available. The s ubject of t he second section was palivizumab ; specifically, we examine d whether historically palivizumab utilization was triggered by disease occurrence or followed a fixed immunization schedule (part V). Finally, in part VI we provide d monthly NNTs for palivizumab to overcome limitations of a dichotomous season definition as a sole basis for RSV immunoprophylaxis Part I: efinition This analysis show ed that the approach of using MPP of a sample of RSV laboratory tests was able to detect seasons of RSV hospitalizations for large US states However, limitations to this approach become apparent on the regional level in Florida with smaller AUCs and evidence for poor specificity. Using the optimal MPP threshold as derived from the R OC analysis was requirement that two consecutive weeks have to be above the threshold predicted a season onset
92 that was on average 4.6 weeks apart from the season onset according to hospitalizations in the states and 6.1 weeks in the regions (table 4 7) Both measures were improved to 3.4 weeks and 4.5 weeks difference with the added requirement of at least 5 positive tests in a given week. Using the optimal threshold an d requir ing that at least 5 tests be positive in each state or region further reduced the difference to actual onset to 3.2 weeks in the states and 3.9 weeks in the regions. Given that this is only a marginal improvement over the 10% MPP combined with 5 po sitive tests and given the distinct advantage of the 10% threshold that it is universal and does not have to be predetermined for each geographic area as would be the case for the optimal threshold we recommend the continued use of the 10% MPP threshold b ut advocate for the added requirement of 5 positive tests in a given week. This definition also optimizes the prediction of season offset with only 2.2 weeks difference to the actual offset (table 4 8) on a state level. However, on a regional level in Flor ida the average absolute difference was 6.3 weeks and the direction 3.9, indicating that predicted offset would occur before the actual offset. We hypothesize that the suboptimal performance of RSV surveillance on a regional level in Florida was a conse quence of two factors. First, if the number of lab oratories in a geographic area and the number of tests during a certain time period is small, outliers can skew the detection of a season. This became especially apparent in the Florida north region (figure 4 5 B) where single positive tests during periods of infrequent testing caused MPPs of up to 70% even outside the season In the north region, on average only 1.9 laboratories reported in a given week compared to 4.8 in the southeast region. Yet t he numb er of laboratories in Illinois and Texas was only slightly higher with an average o f 2.1 and 3.0, respectively but we saw fewer outliers in these states which may be a consequence of the more distinct RSV season with higher seasonality indices in these sta tes Nevertheless, it is obvious that a small number of reported
93 tests can make the detection of a season difficult by either missing true cases or by providing a small denominator thus increasing the impact of single positive tests. The inclusion of Surv eillance Data Inc. RSV data 74 into NREVSS has the potential to overcome this limitation by substantially increasing the number of participating laboratories. Second the south west and especially southeast regions experienced a smaller degree of seasonality with seasonality indices of only 5.2 and 2.5 respectively ( table 4 9) and we observed extended periods where the MPP was just above or just below the threshold in the southe ast (figure 4 5 F). Hence, small variations in the number of positive tests could have moved the MPP above the threshold causing a different estimate of seasonality. Given the fact that seasonality was not very pronounced in the southern part of Florida, i t was expected and confirmed in our study that any season definition has some degree of inaccuracy on a regional level in Florida. Therefore, it is not surprising that even our recommended season definition has its poorest perform t followed by the southwest In fact, the remaining regions show ed a level of accuracy comparable to a state level. Part II: RSV Epidemiology between Four US States and Five Regions in Florida Implications of small seasonality indices in the southern regi ons of Florida were discussed considerations that payers in these states are facing and will be discussed in part VI. Our findings of s eason al variations within g eographic areas over time deserve attention. Season onset varied within states over 7 weeks in Illinois and Texas and 12 weeks in Florida. Even the smallest variation of 7 weeks highlights the need for current surveillance information on RSV activity. One may argue that using the center week of this range to predict the next season has a similar accuracy as the NREVSS which is on average 3.4 weeks apart from the actual RSV season onset (table 4 7). Nevertheless o ur study period only included 5 seasons; it is
94 very possible that the range of season onset increases with more seasons observed which is emphasized through the large confidence intervals of the range of season onset In this case, a fixed immunization recommendation based on previous seasons alone would not be able to respond to changing seasons. Only current data can indicate current RSV activity in the light of year to year variation. Part III: L atitude as a Factor in RSV Epidemiology in Florida Florida Medicaid divides the state into 5 regions w ith regard to RSV seasonality and the r egional differences observed in our study confirm the need for this approach. Treating Florida as a homogenous geographic area would overestimate season duration in the north and underestimate it in the south, therefo re not optimizing immunoprophylaxis. Our analysis showed that latitude was a strong linear predictor for these regional differences as evidenced by large coefficients of determination (R 2 ) in the linear regression models (table 4 12) M oving 1 degree south ( 111 km ) was associated with 5 weeks longer season duration Prior to this study, the division was based on surveillance data only and we now can recommend the continued use of the 5 regions based on actual hospitalization data. At this point, we cannot p rovide explanation as to why latitude plays such a decisive role. It has been suggested that proximity to the coast can influence RSV seasonality 8 yet coastal proximity characterizes most metropolitan areas in Florida, independent of latitude. Climate differences may play a role since the climate becomes more tropical and less seasonal towards the south but these influen ces deserve further research. Regardless of the explanation, the presence of the regional differences in RSV seasonality needs to be taken into account for decisions about the timing of immunoprophylaxis. Part IV : Patient F actors and S easonality We found that the risk factor CLD was more important outside the season than during the season. This observation would suggest that at periods of low viral activity, children at higher
95 risk for RSV were attacked first. However we could not confirm this association for the risk factors CHD and young age (0 6 months) Conversely, belonging to t he second age group (7 12 months) increased the RSV risk more during season than off season, yet to a small extent. Taken together, our observations do not provide conclusive ev idence to support a hypothesis of a different seasonal pattern relative to the risk status of a child. Part V: Timing of Prophylaxis with Palivizumab vs. RSV Seasonality Palivizumab utilization was consistently closest to a fixed date, not to the onset o f RSV hospitalizations or RSV season according to NREVSS. This observation suggests that practitioners may not have based their timing of immunization on current surveillance, but rather followed local PA requirements or the AAP guideline that recommends u tilization between November and April. The only PA requirement during our study period that restricted immunoprophylaxis to fixed calendar months was used in California and allowed palivizumab use between October and May, but only up to 6 doses. Here, we o bserved utilization between November and April, a period that covered 6 doses. Texas Medicaid adjusted their PA period to current NREVSS data, yet utilization seemed not to have been triggered by surveillance data but rather followed a fixed schedule name ly from November through April The AAP guideline recommends that local surveillance data be taken into account but we cannot confirm that this was common practice. Since 3 out of 4 study states did not connect prior authorization to RSV surveillance dat a from NREVSS and prescribers seem to have followed the AAP guideline rather than current surveillance, research into the acceptance of RSV surveillance data would be helpful. After our study period, Medicaid agencies in Florida and Illinois have updated t heir PA requirements to include that palivizumab not be given outside of the local season This suggests that more states started to consider data from NREVSS in their reimbursement decisions; however our sample of only 4 states cannot give a representati ve
96 picture of PA throughout the nation. A survey of both third party payers and practitioners could help understand the basis of timing of immunoprophylaxis in clinical practice. From our limited sample it seems that practitioners use d NREVSS data only ind irectly since the ir prescribing decisions were restricted by PA requirements. The predominant users of NREVSS are potentially third party payers who have an incentive to minimize palivizumab utilization outside of the RSV season. In all states and in most regions of Florida, utilization was started before the onset of RSV hospitalization s and ceased after the offset of viral activity indicating sufficient temporal coverage. Nevertheless, immunoprophylaxis was initiated early, as much as 7 weeks on average before RSV onset in Illinois. While this practice can be beneficial from a pure clinical perspective, it may need to be reviewed in the light of resource optimization. Once again, the incorporation of current surveillance data can help target immunoprophy laxis to periods of high RSV activity. Part V I : Optimizing T iming of P rophylaxis The analysis of monthly RSV incidences and corresponding NNTs deserves interpretation. An NNT of 200 translates into 200 children who have to receive immunoprophylaxis in a given month to prevent one case of RSV hospitalization; hence a small NNT generally indicates a more beneficial intervention. Since palivizumab is a costly drug, third party payers interpret NNTs with regard to related expenses. Our own research has shown that a single dose costs on average $1,338 in the age category 0 6 months, $1,750 from 7 12 months and $2,087 for children from 13 24 months of age from a Medicaid perspective 5 For a NNT of 200, this means that for the youngest age group at high risk for RSV and therefore with indication for palivizumab $267,600 have to be spen t on immunoprophylaxis to prevent one hospitalization ( table 5 1 ) Older age categories are associated with lower RSV risk and, due to higher body weight, with higher cost
97 for prophylaxis. A NNT of 500 for the category of children between one and two years of age requires $1,043,500 to prevent a single hospitalization. The sam e study found that the actual cost for RSV hospitalizations average s well below $10,000. For subjects with indication for immunoprophylaxis, the average hospitalization cost was $7,198.01 Hospital claims ranged from $927.65 to $54,689.31 and 90% of paid c laims were below $16,714.47. Half of the hospitalizations were associated with claims below $4,551.38. We compared these estimates with cost estimates from the Healthcare Cost and Utilization Project (HCUP ) which uses data based on all payer information. A search for RSV hospitalizations in the year 2006 found 66,266 RSV related hospitalizations nationwide for infants up to one year of age. The average charge was $11,745 with an average length of stay of 3.4 days and a total of 45 in hospital deaths. 91 Of note, HCUP does not provide cost estimates and, because of discrepancies, charges should not be used to approximate actual cost. 92 To overcome this limitation, w e applied a cost to charge ratio (CCR) of 0.53, which yielded an average cost estimate of $6,225 for one RSV hospitalization f or all children, regardless of their risk status This CCR has been used with HCUP data from 2002, 93 but we were unable to obtain a more recent estimate. The consistency between Medicaid and HCUP cost estimates strengthen s the generalizability of our findings namely the distinct difference between immunization and hospit alization cost. S tatewide RSV incidence rates decreased with increasing age which is con sistent with the RSV literature 24, 94 and confirms results of our multivariate model from part I. The model revealed that after adjusting for clinical risk factors, the youngest age category (0 6 months) ha d a 10.1 fold increased RSV risk compared to the oldest age category (1 3 24 months) while the middle category only has a 2.6 fold higher risk (table 4 2 ) Since the NNT is inversely related to RSV incidence, we found a distinct increase in NNT associated with the older age groups. Both
98 factors, higher cost for prophylaxis and lower RSV incidence in older children should be taken into account in the evaluation of the appropriateness of immunoprophylaxis in the second year of life The analysis of their own monthly RSV hospitalization data should be considered by each state or third party payer. It offers several advantages over only using surveillance data to guide RSV immunoprophylaxis. First, while surveillance data only provide information about presence or absence of viral activity, hospitalization data can inform about the extent of the RSV burden. Payer s may find that in some months, the burden is too low to justify prophylaxis even though the surveillance system would indicate RSV season. Second, hospitalization data can provide insight about RSV risk factors such as age. Different third party payers may have different willingness to pay threshold s and can determine which groups of patients have acceptable NNTs using their own criteria Similarly, since the burden of disease differs between the states as evidenced in part II, states may use hospitalization data to determine whether lower local RSV activity justifies prophylaxis at all External V alidity Generalizability of our study findings needs to be evaluated on two different levels: first, whether our results derived from a Medicaid sample are generalizable to the general pediatric population and second, whether our sample of 4 states is nationally representative. A strength of the selection of Medicaid data for our study is that a large proportion of all young childre n are enrolled in Medicaid at a given point in time During the year 2000, one of our study years, 532,610 infants were born in California. 95 Of these, 212,361 ( 39.9 % ) were enrolled in the Medicaid program at some point during the year 2000 After applying our eligibility criteria, our final sample retained 111,622 ( 21.0% ) infants with Medicaid fee for service eligibility at birth In Florida, of 204,305 live births, 50 .4 % were enrolled in Medicaid in 2000 and 32.1% were
99 included in our final sample. Our study sample further included 42.2% of 181,984 live births in Illinois and 46 .2 % were enrolled in Medicaid at some point in 2000 Finally, out of 368,019 live birth s in Texas in the year 2000, 53.6 % were enrolled in Medicaid and we included 39.4 % of all newborns in our final study sample. These data confirm that Medicaid is the single largest healthcare program for infants. Our final sample included as many as 42.2% of al l newborns in Illinois, however significantly fewer in California due to the large managed care population in the latter state. Although our sample covered a large proportion of the pediatric population, representativeness is not automatically guaranteed, especially if the sample differs from the general population with regard to relevant characteristics. Medicaid enrollees are often reported to be at a worse health state compared to the rest of the population which suggests that the Medicaid dataset may ov er represent children at higher risk for RSV hospitalization This is confirmed by the Palivizumab Outcomes Registry reporting an unadjusted odds ratio of 1.76 (p<0.0001) for RSV hospitalization between Medicaid recipients and non recipients 96 This concern can be addressed. Our study include d a division of children into high risk and low risk for RSV hospitalization based on the current AAP guideline According to our definition for seasonality based on claims data, seasons were periods where the RSV incidence rate among high risk infants was above the maximum incidence rate for low risk children. Not relevant to generalizability is the relative size of these groups, which would differ between the Medicaid sample and the general population due to the over representation of high risk infants. It is irrelevant simply because the incidence rate is not a factor of sample size. Our choice of 4 large geographically diverse states allows for geographic generalizations The fact that the AUCs together with sensitivity and specificity were consistently high for the
100 statewide application of RSV surveillance data make s us confident that the validity of the NREVSS and the RSV season definition are applicable on the level of large states or even larger surveillance regions as used by the CDC, independent of their geographic locations. While the season definition is universal, the burden of RSV and the extent of seasonality differ widely between geographic areas. As a consequence, o ur calculation of monthly incidences and corresponding NNTs are only applicable to the study states. As discussed before, third party payers who are interested in the burden of RSV relevant to their population are served best by consulting their own hospit alization data. Study Limitations Several limitations should be considered when interpreting the results of this study. First, we faced unforeseen difficulties in obtaining social security numbers from state V ital S tatistics departments as a linkage variab le between birth certificates and CMS claims data. Alternative variables for linking records were not available either. At this point, the study only includes birth certificates and therefore gestational age estimates for Florida, not for the other states. As a consequence, the high risk cohorts in the other states only include d CLD and CHD, not prematurity I and II, and the low risk cohorts did not exclude prematurity III. In a sensitivity analysis based on the Florida dataset, we tested how the absence of information on prematurity may have affected our results. Ignoring prematurity in Florida, we estimated a season onset that was on average 1.6 weeks apart from estimates with prematurity. Season offset was on average 2.4 weeks apart from estimates based o n the complete dataset. This small difference supports the accuracy of our study results even in the absence of gestational age information in California Illinois and Texas. Nevertheless to maximize accuracy and ultimately strengthen our study, we are st ill pursuing ways to access birth certificates in all study states and we will update all analyses upon successfully defining prematurity.
101 Second, a limitation to our study is the possibility of misclassified RSV infections ; i.e. bronchiolitis and pneumoni a hospitalizations that were due to RSV but not coded as such. We aimed to quantify the potential for misclassification For this, we divided c ases of bronchiolitis and pneumonia hospitalizations into three categories: RSV related specific to other organi sms and unspecific ( see appendix for diagnostic codes) For Florida, we plotted the incidence rate of hospitalizations falling into each of these categories in each calendar month (figure 5 1). We considered t he number of unspecifi ed cases as the maximum of potentially misclassified RSV cases. It is conceivable that some cases with specific codes for other agents were also misclassified RSV cases, however, potentially with a lower likelihood of misclassification since an alternative causal agent was identi fied. The results show that in case all unspecific bronchiolitis and pneumonia hospitalizations were in fact RSV cases, the RSV incidence rate would be underestimated by approximately 40%. Furthermore, given that the incidenc e rates of specified and un spe cified non RSV cases have a strikingly parallel course; it is very likely that many of the unspecified cases are in fact non RSV related bronchiolitis or pneumonia admissions, thus not affecting our study. This misclassification would not change conclusion s about RSV seasonality under the assumption that it occurred to an equal amount in the high risk and low risk cohort. Misclassification of RSV cases would mainly affect part VI, resulting in an underestimate of the monthly RSV incidence rates and in an ov erestimate of NNTs Nevertheless, given the limited amount of misclassification, NNTs would remain high even after accounting for missed RSV cases and conclusions about the need for careful consideration of immunoprophylaxis would remain unchanged. Lastly the accuracy of palivizumab claims in Medicaid claims data is limited as a study found using data from the North Carolina Medicaid program. 97 Briefly, this study compared
102 dates and frequenc ies of palivizumab claims with abstracted ambulato ry and inpatient medical records. The investigators found 87.0% agreement of infants who received any palivizumab injection in both datasets, but agreement was only reached in 46.1% of infants about the number of injections. Although not mentioned in the s tudy, the second finding may be related to injections during inpatient stays that would have been identified in the medical records but not in the Medicaid claims dataset due to the aggregate nature of a Medicaid hospitalization claim We tried to increase accuracy of exposure information in our study by requiring 4 weeks of ambulatory care preceding the current week to ensure that palivizumab administration could have create d an outpatient /pharmacy claim. Furt hermore, we required that a palivizumab recipie within 10 days of the claim. Although this requirement is not very specific, it can exclude palivizumab claims with questionable validity. Inaccuracy in detecting p alivizumab exposure has the theoretical potential to bias RSV incidence estimates since they were adjusted to account for palivizumab exposure. Since palivizumab was targeted to high risk children (table 4 3), only their RSV incidence estimates would be af fected. However, since less than 20% of subject weeks in the high risk cohort were associated with palivizumab prophylaxis, our incidence estimates were mostly driven by observed hospitalizations and only to a small amount affected by adjustment for immuno prophylaxis. Thus, we feel confident that our results regarding RSV seasonality are robust despite the potential for exposure misclassification. Future Research With the inclusion of more RSV laboratories provided by Surveillance D ata I nc ., NREVSS is exp ected to offer more accurate estimates on a smaller geographic level. Future studies should investigate whether regional limitations of RSV surveillance which we identified in Florida could be overcome with the added number of laboratories. Since the inclu sion of the
103 additional laboratories occurred only from the 2006 07 RSV season, more years of observation are necessary to determine beneficial effects of the larger sample. As we outlined before, third party payers are strongly encouraged to use their own inpatient data to determine the absolute burden of RSV relevant to their population. This would enable them to select a population and timing of immunoprophylaxis that is consistent with their own willingness to pay thresholds. Finally, our utilization an alysis showed that historically, immunoprophylaxis seemed not primarily triggered by season detections based on RSV surveillance. Surveying third party payers and practitioners about the basis of their decisions about timing of prophylaxis can help underst and and potentially increase the acceptance of NREVSS. Moreover this research could provide a synopsis of current PA programs and share the experience of their success and limitations. Summary and Conclusions to det ect seasons of RSV activity based on laboratory surveillance. We recommend the continued use of the 10% MPP threshold, however with the added requirement of 5 positive tests in a given week. The performance of the laboratory based approach was suboptimal i n the southwest and southeast regions in Florida and should be applied carefully in areas with less distinct seasonality Next, our study identified differences in seasonality over time and a different extent of seasonality between the 4 study states and e ven within Florida therefore confi rming the importance of a surveillance system that offers current and local information on RSV activity Based on RSV hospitalization data, we confirmed the need to subdivide the state of Florida into 5 regions to appropr iately account for differences in RSV epidemiology. We found that historically, palivizumab utilization seemed not primarily triggered by seasons detected with laboratory surveillance; therefore further research is
104 necessary to help understand the acceptan ce of the NREVSS. Finally, we provided monthly RSV incidence rates for each state and for each region in Florida. The corresponding NNT estimates can provide further detail on the burden of disease to guide reimbursement practices and thus, overcome limita tions of a dichotomous RSV season definition. Higher NNTs for older children as a result of a lower RSV incidence combined with the need for higher and more costly doses of palivizumab after infancy highlight the reduced benefit of immunoprophylaxis in the second year of life.
105 Table 5 1. Cost of prophylaxis per avoided RSV hospitalization Age [months] 0 6 7 12 13 24 0 24 avg cost per dose $1,338 $1,750 $2,087 $1,688 NNT cost per avoided hospitalization 100 $133,800 $175,000 $208,700 $168, 800 200 $267,600 $350,000 $417,400 $337,600 300 $401,400 $525,000 $626,100 $506,400 400 $535,200 $700,000 $834,800 $675,200 500 $669,000 $875,000 $1,043,500 $844,000 Abbreviation: NNT: number needed to treat Figure 5 1. Distr ibution of diagnostic codes for bronchiolitis and pneumonia related hospitalizations
106 APPENDIX A. Operational Definitions The Medicaid prescription drug dataset did not include generic class codes, only NDC. We ordered a crosswalk file from NDC codes to drug names, generic class codes and American Hospital Formulary Service (AHFS) class codes. We were able to match 99.56 % of all prescriptions from the Medicaid dataset. Palivizumab E xposure Any of (NDC codes for palivizumab: 60574411101, 6057441 1201, 60574411301 60574411401 or procedure code for palivizumab: 90387, (additionally in California: C9003, X7439 ; Texas : 1086X, 1095X)) in conjunction with an outpatient visit to a physician, other practitioner, outpatient hospital, clinic, home health, other services, nurse practitioner or private duty nurse (Medicaid statistical information system (MSIS) type of service codes 08, 10, 11, 12, 13, 19, 37 or 38) within 10 days before or after the palivizumab claim. Risk F actors for RSV Chronic l ung d isease Children younger than 2 years with CLD who received medication for CLD (Steroids, bronchodilators, oxygen, diuretics) within 6 months of the current week. They n eed to have at least one (ICD 9 :770.7, 496x) claim at any time before the current week and eit her A) At least one of the following CLD medications based on AHFS codes: 121208 BETA ADRENERGIC AGONISTS 680400 ADRENALS 861600 RESPIRATORY SMOOTH MUSCLE RELAXANTS 481024 LEUKOTRIENE MODIFIERS 402820 THIAZIDE DIURETICS
107 402824 THIAZIDE LIKE DIURETICS 4028 08 LOOP DIURETICS 402816 POTASSIUM SPARING DIURETIC or B) ICD 9 code: V46.2 or C) Oxygen Procedure Code: 93.96, E1390, E1392, E1400, E1401, E1402, E1403, E1404, E1405, E1406, E0424, E0431, E0434, E0439, E0441, E0442, E0443, E0444, E0450 w ithin the previous 6 months of the current week Prematurity Prematurity I: Infants with a gestational age of less than 28 weeks if they are not more than 12 months at the beginning of the current week Prematurity II: Infants with gestational age of 29 32 (32 weeks and 0 day s) weeks if they are not more than 6 months old in at the beginning of the current week Prematurity III: Infants with gestational age of 32 35 (from 32 weeks and 1 day) weeks if they are not more than 6 months old in at the beginning of the current week Co ngenital h eart d isease Children younger than 2 years with hemodynamically significant cyanotic and acyanotic CHD : Acyanotic CHD is only considered significant if medication (ACE Inhibitor, Digoxin, Diuretics, or Oxygen) is necessary to treat the disease. T hey need to have at least one of the following ICD 9 codes: 745.10, 747.41, 745.0, 745.11, 745.2, 747.42, 745.3, 746.1, 746.7, 745.1, 745.12, 745.19, 746.2, 747.3, 747.4, 747.40, 747.49 (cyanotic heart disease) any time before the current week OR [at leas t one of the ICD 9 codes: 746.86, 747.11, 747.22, 745.4, 745.5, 745.6, 745.60, 745.61, 745.69, 745.7, 745.8, 745.9, 746.0, 746.00, 746.01, 746.02, 746.09,746.3, 746.4,
108 746.5, 746.6, 746.8, 746.81, 746.82, 746.83, 746.84, 746.85, 746.87, 746.89, 747.0, 747 .1, 747.10, 746.9, 747, 747.2, 747.20, 747.21,747.29, 747.5, 747.6, 747.60, 747.61, 747.62, 747.63, 747.64, 747.69, 747.8, 747.81, 747.82, 747.83, 747.89, 747.9 (acyanotic CHD) any time before the current week AND [ at least one of the following CHD medicat ions based on AHFS codes within 6 months before the current week: 243204 ANGIOTENSIN CONVERTING ENZYME INHIBITORS 402808 LOOP DIURETICS 402816 POTASSIUM SPARING DIURETICS 402820 THIAZIDE DIURETICS 402824 THIAZIDE LIKE DIURETICS 240408 CARDIOTONIC AGENTS (D igoxin) OR Oxygen Procedure Codes: 93.96, E1390, E1392, E1400, E1401, E1402, E1403, E1404, E1405, E1406, E0424, E0431, E0434, E0439, E0441, E0442, E0443, E0444, E0450] ] Cystic fibrosis At any time before the current week: ICD 9 code: 277.0x Severe comb ined or acquired immunodeficiency At any time before the current week: I CD 9 code: 042xx, 279.2, 279.11 or 758.32 Down syndrome At any time before the current week: ICD 9 code: 758.0 Asthma At any time before 2 weeks before the current week: ICD 9 code 493 xx AND asthma medication between 120 and 14 days before the beginning of the current week : AHFS code: 121208 BETA ADRENERGIC AGONISTS
109 481024 LEUKOTRIENE MODIFIERS 120808 ANTIMUSCARINICS/ANTISPASMODICS only Ipratropium (not nasal) 680400 ADRENALS 861600 RES PIRATORY SMOOTH MUSCLE RELAXANTS Transplant T ransplant (ICD 9: V42.x) or awaiting transplant (ICD 9: V49.83) at any time before the current week Malignancy At any time before the current week: ICD 9 codes: 140 209, 230 234 Immunosuppression or antineoplast ic agents Any time before the current week: AHFS code: 100000 ANTINEOPLASTIC AGENTS 680400 ADRENALS (oral only) OR Drug names: INFLIXIMAB, GOLD, AURANOFIN, AUROTHIOGLUCOSE, LEFLUNOMIDE, ETANERCEPT, ADALIMUMAB, ANAKINRA, LETROZOLE, AZATHIOPRINE, MYCOPHENOL ATE, CYCLOSPORINE, TACROLIMUS ANHYDROUS (excludes topical form), SIROLIMUS, DACLIZUMAB, INTERFERON, LETROZOLE, ALDESLEUKIN, OMALIZUMAB, THALIDOMIDE Hospitalizations RSV hospitalization ICD 9 codes: 079.6, 466.11 or 480.1 in the inpatient dataset.
110 Specific non RSV bronchiolitis or pneumonia ICD 9 codes: 480.0, 480.2, 480.3, 480 .8, 481, 482.xx, 483.xx, 484.xx or 466.19 in the inpatient dataset. Unspecific bronchiolitis or pneumonia ICD 9 codes: 480.9 or 486 in the inpatient dataset.
1 11 B. Supplemental Tables Table B 1. List of c ounties in Florida County FIPS c ode Region County FIPS c ode Region ALACHUA 1 N LAKE 69 C BAKER 3 N LEE 71 SW BAY 5 NW LEON 73 NW BRADFORD 7 N LEVY 75 N BREVARD 9 C LIBERTY 77 NW BROWARD 11 SE MADISON 79 N CALHOUN 13 NW MANATE E 81 SW CHARLOTTE 15 SW MARION 83 C CITRUS 17 C MARTIN 85 SE CLAY 19 N MONROE 87 SE COLLIER 21 SW NASSAU 89 N COLUMBIA 23 N OKALOOSA 91 NW DADE 25 SE OKEECHOBEE 93 SW DESOTO 27 SW ORANGE 95 C DIXIE 29 N OSCEOLA 97 C DUVAL 31 N PALM BEACH 99 SE ESCAMBIA 33 NW PASCO 101 C FLAGLER 35 C PINELLAS 103 C FRANKLIN 37 NW POLK 105 SW GADSDEN 39 NW PUTNAM 107 N GILCHRIST 41 N SANTA ROSA 113 NW GLADES 43 SW SARASOTA 115 SW GULF 45 NW SEMINOLE 117 C HAMILTON 47 N ST. JOHNS 109 N HARDEE 49 SW ST. LUCIE 111 SE HENDRY 51 SW SUMTER 119 C HERNANDO 53 C SUWANNEE 121 N HIGHLANDS 55 SW TAYLOR 123 N HILLSBOROUGH 57 C UNION 125 N HOLMES 59 NW VOLUSIA 127 C INDIAN RIVER 61 SE WAKULLA 129 NW JACKSON 63 NW WALTON 131 NW JEFFERSON 65 NW WASHINGTON 133 NW LAFAYETTE 67 N Abbreviation: FIPS: Federal information processing standard
112 Table B 2 Coordinates of Florida r egions Florida Region Centroid latitude [degrees North ] Centroid longitude [degrees West ] Northwest 30.293 85.597 North 29.909 82.607 Central 28.622 81.629 Southwest 27.083 81.712 Southeast 26.197 81.087 Table B 3. Week numbers and corresponding calendar months, shown for the year 2000 Week Month Week Month 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 2 0 21 22 23 24 25 26 January January January January February February February February March March March March March April April April April May May May May June June June June June 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 5 1 52 July July July July August August August August August September September September September October October October October November November November November November December December December December *week 1 is the first week that end ed in 20 00.
113 LIST OF REFERENCES 1. Shay DK, Holman RC, Newman RD, Liu LL, Stout JW, Anderson LJ. Bronchiolitis associated hospitalizations among US children, 1980 1996. Jama. Oct 20 1999;282(15):1440 1446. 2. The IMpact RSV Study Group. Paliv izumab, a humanized respiratory syncytial virus monoclonal antibody, reduces hospitalization from respiratory syncytial virus infection in high risk infants. Pediatrics. Sep 1998;102(3 Pt 1):531 537. 3. The PREVENT Study Group. Reduction of respiratory syn cytial virus hospitalization among premature infants and infants with bronchopulmonary dysplasia using respiratory syncytial virus immune globulin prophylaxis. Pediatrics. Jan 1997;99(1):93 99. 4. MedImmune Inc. Package insert, Synagis. http://www.fda.gov/cder/foi/label/2002/palimed102302LB.pdf Accessed 06/05/2008. 5. Winterstein AG, Hampp C, Saidi A. Final Report to the Agency for Healthcare Administration, Florida Department of Health : Florida Center for Medicaid and the Uninsured, University of Florida; 2008. 6. Committee on Infectious Diseases and Committee on Fetus and Newborn. Revised indications for the use of palivizumab and respiratory syncytial virus immune globulin intra venous for the prevention of respiratory syncytial virus infections. Pediatrics. Dec 2003;112(6 Pt 1):1442 1446. 7. Centers for Disease Control and Prevention. Respiratory Syncytial Virus (RSV) Surveillance. http://www.cdc.gov/surveillance/nrevss/rsv/default.html Accessed 02/23/2009. 8. Stensballe LG, Devasundaram JK, Simoes EA. Respiratory syncytial virus epidemics: the ups and downs of a seasonal virus. Pediatr Infect Dis J. Feb 200 3;22(2 Suppl):S21 32. 9. Mullins JA, Lamonte AC, Bresee JS, Anderson LJ. Substantial variability in community respiratory syncytial virus season timing. Pediatr Infect Dis J. Oct 2003;22(10):857 862. 10. Halstead DC, Jenkins SG. Continuous non seasonal epi demic of respiratory syncytial virus infection in the southeast United States. South Med J. May 1998;91(5):433 436. 11. Bauman J, Eggleston M, Oquist N, Malinoski F. Respiratory syncytial virus: seasonal data for regions of Florida and implications for pal ivizumab. South Med J. Jul 2007;100(7):669 676. 12. Boyce TG, Mellen BG, Mitchel EF, Jr., Wright PF, Griffin MR. Rates of hospitalization for respiratory syncytial virus infection among children in medicaid. J Pediatr. Dec 2000;137(6):865 870.
114 13. Gilchris t S, Torok TJ, Gary HE, Jr., Alexander JP, Anderson LJ. National surveillance for respiratory syncytial virus, United States, 1985 1990. J Infect Dis. Oct 1994;170(4):986 990. 14. Hampp C, Winterstein AG. Response to respiratory synctial virus. South Med J Feb 2008;101(2):212 213; author reply 213 214. 15. Meissner HC, Anderson LJ, Pickering LK. Annual variation in respiratory syncytial virus season and decisions regarding immunoprophylaxis with palivizumab. Pediatrics. Oct 2004;114(4):1082 1084. 16. Flori da Agency for Healthcare Administration. Palivizumab prior authorization Northwest region. http://www.fdhc.state.fl.us/medicaid/prescribed_drug/pharm _thera/paforms/synagis_nw.pd f Accessed 02/23/2009. 17. Florida Agency for Healthcare Administration. Palivizumab prior authorization Southeast region. http://ahca.myflorida.com/medicaid/Prescribed_Drug/pharm_thera/paforms/synagis_se.pdf Accessed 02/28/2009. 18. Blount RE, Jr., Morris JA, Savage RE. Recovery of cytopathogenic agent from chimpanzees with coryza. Proc Soc Exp Biol Med. Jul 1956;92(3):54 4 549. 19. Chanock R, Finberg L. Recovery from infants with respiratory illness of a virus related to chimpanzee coryza agent (CCA). II. Epidemiologic aspects of infection in infants and young children. Am J Hyg. Nov 1957;66(3):291 300. 20. Chanock R, Roiz man B, Myers R. Recovery from infants with respiratory illness of a virus related to chimpanzee coryza agent (CCA). I. Isolation, properties and characterization. Am J Hyg. Nov 1957;66(3):281 290. 21. Ogra PL. Respiratory syncytial virus: the virus, the di sease and the immune response. Paediatr Respir Rev. 2004;5 Suppl A:S119 126. 22. Hall CB. Respiratory syncytial virus and parainfluenza virus. N Engl J Med. Jun 21 2001;344(25):1917 1928. 23. Domachowske JB, Rosenberg HF. Respiratory syncytial virus infect ion: immune response, immunopathogenesis, and treatment. Clin Microbiol Rev. Apr 1999;12(2):298 309. 24. Hall CB, Weinberg GA, Iwane MK, et al. The burden of respiratory syncytial virus infection in young children. N Engl J Med. Feb 5 2009;360(6):588 598. 25. Thompson WW, Shay DK, Weintraub E, et al. Mortality associated with influenza and respiratory syncytial virus in the United States. Jama. Jan 8 2003;289(2):179 186.
115 26. Shay DK, Holman RC, Roosevelt GE, Clarke MJ, Anderson LJ. Bronchiolitis associated mortality and estimates of respiratory syncytial virus associated deaths among US children, 1979 1997. J Infect Dis. Jan 1 2001;183(1):16 22. 27. Meissner HC, Long SS. Respiratory syncytial virus infection and recurrent wheezing: a complex relationship. J Pediatr. Jul 2007;151(1):6 7. 28. Taussig LM, Wright AL, Holberg CJ, Halonen M, Morgan WJ, Martinez FD. Tucson Children's Respiratory Study: 1980 to present. J Allergy Clin Immunol. Apr 2003;111(4):661 675; quiz 676. 29. Stensballe LG, Simonsen JB, Thomsen SF, et al. The causal direction in the association between respiratory syncytial virus hospitalization and asthma. J Allergy Clin Immunol. Jan 2009;123(1):131 137 e131. 30. Henrickson KJ, Hall CB. Diagnostic assays for respiratory syncytial virus disease. Pediatr Infect Dis J. Nov 2007;26(11 Suppl):S36 40. 31. Respiratory syncytial virus activity -United States, July 2007 December 2008. MMWR Morb Mortal Wkly Rep. Dec 19 2008;57(50):1355 1358. 32. Schauer U, Ihorst G, Rohwedder A, et al. Evaluation of resp iratory syncytial virus detection by rapid antigen tests in childhood. Klin Padiatr. Jul Aug 2007;219(4):212 216. 33. Schutzle H, Weigl J, Puppe W, Forster J, Berner R. Diagnostic performance of a rapid antigen test for RSV in comparison with a 19 valent m ultiplex RT PCR ELISA in children with acute respiratory tract infections. Eur J Pediatr. Jul 2008;167(7):745 749. 34. Bourgeois FT, Olson KL, Brownstein JS, McAdam AJ, Mandl KD. Validation of syndromic surveillance for respiratory infections. Ann Emerg Me d. Mar 2006;47(3):265 e261. 35. Fergie J, Purcell K. Respiratory syncytial virus laboratory surveillance and hospitalization trends in South Texas. Pediatr Infect Dis J. Nov 2007;26(11 Suppl):S51 54. 36. Light M, Bauman J, Mavunda K, Malinoski F, Eggleston M. Correlation Between Respiratory Syncytial Virus (RSV) Test Data and Hospitalization of Children for RSV Lower Respiratory Tract Illness in Florida. Pediatr Infect Dis J. Apr 30 2008. 37. Centers for Disease Control and Prevention. Morbidity and Mortali ty Weekly Report. http://www.cdc.gov/mmWR/ Accessed 02/23/2009. 38. Respiratory virus surveillance -United States, 1983 1984. MMWR Morb Mortal Wkly Rep. Jan 27 1984;33(3):29 30. 39. Respiratory syncytial virus an d parainfluenza virus surveillance -United States, 1989 90. MMWR Morb Mortal Wkly Rep. Nov 23 1990;39(46):832 833, 839.
116 40. Respiratory syncytial virus outbreak activity -United States, 1992. MMWR Morb Mortal Wkly Rep. Jan 15 1993;42(1):5 7. 41. Update: re spiratory syncytial virus activity -United States, 1993. MMWR Morb Mortal Wkly Rep. Dec 24 1993;42(50):971 973. 42. Update: respiratory syncytial virus activity -United States, 1997 98 season. MMWR Morb Mortal Wkly Rep. Dec 11 1998;47(48):1043 1045. 43. Br ief report: respiratory syncytial virus activity -United States, 2005 2006. MMWR Morb Mortal Wkly Rep. Dec 1 2006;55(47):1277 1279. 44. Panozzo CA, Fowlkes AL, Anderson LJ. Variation in timing of respiratory syncytial virus outbreaks: lessons from national surveillance. Pediatr Infect Dis J. Nov 2007;26(11 Suppl):S41 45. 45. Reyes M, Eriksson M, Bennet R, Hedlund KO, Ehrnst A. Regular pattern of respiratory syncytial virus and rotavirus infections and relation to weather in Stockholm, 1984 -1993. Clin Micro biol Infect. Feb 1997;3(6):640 646. 46. Eriksson M, Bennet R, Rotzen Ostlund M, von Sydow M, Wirgart BZ. Population based rates of severe respiratory syncytial virus infection in children with and without risk factors, and outcome in a tertiary care settin g. Acta Paediatr. 2002;91(5):593 598. 47. Terletskaia Ladwig E, Enders G, Schalasta G, Enders M. Defining the timing of respiratory syncytial virus (RSV) outbreaks: an epidemiological study. BMC Infect Dis. 2005;5(1):20. 48. Waris M. Pattern of respiratory syncytial virus epidemics in Finland: two year cycles with alternating prevalence of groups A and B. J Infect Dis. Mar 1991;163(3):464 469. 49. Light M. Respiratory syncytial virus seasonality in southeast Florida: results from three area hospitals caring for children. Pediatr Infect Dis J. Nov 2007;26(11 Suppl):S55 59. 50. Singleton RJ, Bruden D, Bulkow LR, Varney G, Butler JC. Decline in respiratory syncytial virus hospitalizations in a region with high hospitalization rates and prolonged season. Pediatr Infect Dis J. Dec 2006;25(12):1116 1122. 51. Meissner HC. Summary. Pediatr Infect Dis J. Nov 2007;26(11 Suppl):S60. 52. Mitchell I. Respiratory syncytial virus: different criteria for palivizumab use in different areas? South Med J. Jul 2007;100(7):661 66 2. 53. Power UF. Respiratory syncytial virus (RSV) vaccines -two steps back for one leap forward. J Clin Virol. Jan 2008;41(1):38 44.
117 54. Food and Drug Administration. Press Release: FDA licences first product to prevent serious RSV disease. http://www.fda.gov/bbs/topics/NEWS/NEW00523.html Accessed 04/14/2008. 55. Food and Drug Administration. Palivizumab product approval letter. http://www.fda.gov/cder/foi/appletter/1998/palimed061998l.pdf Accessed 04/14/2008. 56. MedImmune Inc. MedImmune Annual Report 2003. http://www.medimmune.com/ar/2003 /financials/notes16.html Accessed 04/14/2008. 57. Food and Drug Administration. Palivizumab Product Review. http://www.fda.gov/cder/biologics/review/palimed061998r5a.pdf Acc essed 04/14/2008. 58. MedImmune Inc. MedImmune Submits Biologics License Application to FDA for Motavizumab. http://www.reuters.com/article/pressRelease/idUS179 756+04 Feb 2008+PRN20080204 Accessed 04/23/2008. 59. Feltes TF, Cabalka AK, Meissner HC, et al. Palivizumab prophylaxis reduces hospitalization due to respiratory syncytial virus in young children with hemodynamically significant congenital heart disease J Pediatr. Oct 2003;143(4):532 540. 60. Vogel AM, Lennon DR, Broadbent R, et al. Palivizumab prophylaxis of respiratory syncytial virus infection in high risk infants. J Paediatr Child Health. Dec 2002;38(6):550 554. 61. Drug topics red book: pharmacy's fundamental reference. Montvale, N.J: Thomson PDR; 2008. 62. Kamal Bahl S, Doshi J, Campbell J. Economic analyses of respiratory syncytial virus immunoprophylaxis in high risk infants: a systematic review. Arch Pediatr Adolesc Med. Oct 2002;156(10):1034 10 41. 63. Duppenthaler A, Ammann RA, Gorgievski Hrisoho M, Pfammatter JP, Aebi C. Low incidence of respiratory syncytial virus hospitalisations in haemodynamically significant congenital heart disease. Arch Dis Child. Oct 2004;89(10):961 965. 64. Bloemers BL van Furth AM, Weijerman ME, et al. Down syndrome: a novel risk factor for respiratory syncytial virus bronchiolitis -a prospective birth cohort study. Pediatrics. Oct 2007;120(4):e1076 1081. 65. Blanchard SS, Gerrek M, Siegel C, Czinn SJ. Significant mor bidity associated with RSV infection in immunosuppressed children following liver transplantation: case report and discussion regarding need of routine prophylaxis. Pediatr Transplant. Nov 2006;10(7):826 829. 66. Pohl C, Green M, Wald ER, Ledesma Medina J. Respiratory syncytial virus infections in pediatric liver transplant recipients. J Infect Dis. Jan 1992;165(1):166 169.
118 67. Hall CB, Powell KR, MacDonald NE, et al. Respiratory syncytial viral infection in children with compromised immune function. N Engl J Med. Jul 10 1986;315(2):77 81. 68. El Saleeby CM, Somes GW, DeVincenzo JP, Gaur AH. Risk factors for severe respiratory syncytial virus disease in children with cancer: the importance of lymphopenia and young age. Pediatrics. Feb 2008;121(2):235 243. 69 Sung L, Alonzo TA, Gerbing RB, et al. Respiratory syncytial virus infections in children with acute myeloid leukemia: a report from the Children's Oncology Group. Pediatr Blood Cancer. Dec 2008;51(6):784 786. 70. Management of infections caused by respir atory syncytial virus. Scand J Infect Dis. 2001;33(5):323 328. 71. California Department of Healthcare Services Medi Cal. Provider manuals. http://files.medi cal.ca.gov/pubsdoco/manu als_menu.asp Accessed 02/23/2009. 72. Update: respiratory syncytial virus activity -United States, 1994 95 season. MMWR Morb Mortal Wkly Rep. Dec 16 1994;43(49):920 922. 73. Update: respiratory syncytial virus activity -United States, 1998 1999 season. M MWR Morb Mortal Wkly Rep. Dec 10 1999;48(48):1104 1106, 1115. 74. Brief report: respiratory syncytial virus activity -United States, July 2006 November 2007. MMWR Morb Mortal Wkly Rep. Dec 7 2007;56(48):1263 1265. 75. Florida Department of Health. RSV Regi onal Breakdown http://www.doh.state.fl.us/disease_ctrl/epi/RSV/regbreakdown.pdf Accessed 03/17/2009. 76. Bright RA, Avorn J, Everitt DE. Medicaid data as a resource for epid emiologic studies: strengths and limitations. J Clin Epidemiol. 1989;42(10):937 945. 77. Crystal S, Akincigil A, Bilder S, Walkup JT. Studying prescription drug use and outcomes with medicaid claims data: strengths, limitations, and strategies. Med Care. O ct 2007;45(10 Supl 2):S58 65. 78. Kiyota Y, Schneeweiss S, Glynn RJ, Cannuscio CC, Avorn J, Solomon DH. Accuracy of Medicare claims based diagnosis of acute myocardial infarction: estimating positive predictive value on the basis of review of hospital reco rds. Am Heart J. Jul 2004;148(1):99 104. 79. Hennessy S, Bilker WB, Weber A, Strom BL. Descriptive analyses of the integrity of a US Medicaid claims database. Pharmacoepidemiol Drug Saf. Mar 2003;12(2):103 111. 80. Hennessy S, Leonard CE, Palumbo CM, Newco mb C, Bilker WB. Quality of Medicaid and Medicare data obtained through Centers for Medicare and Medicaid Services (CMS). Med Care. Dec 2007;45(12):1216 1220.
119 81. US Census Bureau. National and state population estimates, selected Age Groups by States and Puerto Rico: 2005. http://www.census.gov/popest/states/asrh/SC EST2005 01.html Acce ssed 02/23/2009. 82. Pearl M, Wier ML, Kharrazi M. Assessing the quality of last menstrual period date on California birth records. Paediatr Perinat Epidemiol. Sep 2007;21 Suppl 2:50 61. 83. Swets JA. Measuring the accuracy of diagnostic systems. Science. Jun 3 1988;240(4857):1285 1293. 84. Vasan RS, Benjamin EJ, Larson MG, et al. Plasma natriuretic peptides for community screening for left ventricular hypertrophy and systolic dysfunction: the Framingham heart study. Jama. Sep 11 2002;288(10):1252 1259. 85. Perneczky R, Pohl C, Sorg C, et al. Complex activities of daily living in mild cognitive impairment: conceptual and diagnostic issues. Age Ageing. May 2006;35(3):240 245. 86. Ahmad Y, Shelmerdine J, Bodill H, et al. Subclinical atherosclerosis in systemic lupus erythematosus (SLE): the relative contribution of classic risk factors and the lupus phenotype. Rheumatology (Oxford). Jun 2007;46(6):983 988. 87. Gnen M. Receiver Operating Characteristic (ROC) Curves. http://www2.sas.com/proceedings/sugi31/210 31.pdf Accessed 02/26/2009. 88. Newcombe RG. Two sided confidence intervals for the single proportion: comparison of seven methods. Stat Med. Apr 30 1998;17(8):857 872. 89. United States De partment of the Interior. National Atlas of the United States, Latitude and Longitude. http://www.nationalatlas.gov/articles/mapping/a_latlong.html Accessed 03/11/2009. 90. Smi thson M. Confidence intervals, p19 Thousand Oaks, Calif.: Sage Publications; 2003. 91. Agency for Healthcare Research and Quality. HCUPnet: A tool for identifying, tracking, and analyzing national hospital statistics. http://hcupnet.ahrq.gov/HCUPnet.jsp Accessed 04/24/2009. 92. Finkler SA. The distinction between cost and charges. Ann Intern Med. Jan 1982;96(1):102 109. 93. Boscoe A, Paramore C, Verbalis JG. Cost of illness of hyponatremia in the Unite d States. Cost Eff Resour Alloc. 2006;4:10. 94. Iwane MK, Edwards KM, Szilagyi PG, et al. Population based surveillance for hospitalizations associated with respiratory syncytial virus, influenza virus, and parainfluenza viruses among young children. Pedia trics. Jun 2004;113(6):1758 1764.
120 95. National Center for Health Statistics. 1 9. Live births by State of occurrence distributed according to resident status: United States and each State, 2000. http://www.cdc.gov/nchs/data/statab/t001x09.pdf Accessed 03/21/2009. 96. Frogel M, Nerwen C, Cohen A, Vanveldhuisen P, Harrington M, Boron M. Prevention of hospitalization due to respiratory syncytial virus: results from the Palivizumab Outcomes Regi stry. J Perinatol. Mar 27 2008. 97. Jacobson Vann J, Feaganes J, Wegner S. Reliability of medicaid claims versus medical record data: in a cost analysis of palivizumab. Pharmacoeconomics. 2007;25(9):793 800.
121 BIOGRAPHICAL SKETCH Christian Hampp was born in Neunkirchen /Saar and raised in St. Ingbert, Germany. He from Saarland University in Saarbrcken in 2003. In 2004, after he became a registered p harmacist in Germany he joined the department of Ph armaceutical Outcomes & Policy at the University of Florida where he was named the first honorary recipient of the DuBow Family Fellowship for Pharmaceutical Outcomes and Policy Research. Christian has authored and coauthored several peer reviewed publica tions and presented at national and international conferences. His research interests focus on drug safety and effectiveness, cost effectiveness and the evaluation and prevention of inappropriate drug use. He is further interested in infectious disease epi demiology and methods to describe temporal and geographic patterns of disease occurrence.