Lorenzo M. Waguespack 1 Socioeconomic Determinants of Healthcare Quality and Outcomes in U.S. States Lorenzo M. Waguespack Advisor: Dr. Michelle Phillips University of Florida March 3 2018 Abstract: As healthcare becomes a more fundamental and increasingly discussed political issue healthcare quality lies at the foundation of creating an ideal healthcare system. What socioeconomic factors influence the quality of healthcare and efficacy of treatment s in U.S. states? This paper uses a linear regression in an attempt to uncover determinants of a lower colorectal cancer mortality rate and improved healthcare outcomes in U.S. states from 2004 and 2014. Our regression results indicate that contrary to th the uninsured rate and real GDP per capita were not statistically significant indicators of colorectal c ancer mortality and healthcare outcomes. Despi te this, the results support the hypothesis that both higher educational attainment and a higher urban share of counties would be associated with a lower colorectal cancer mortality rate, ultimately concurring with the findings of the studies by Ross et al. (1995) and Aboagye et al. (2014) cited within this paper. Overall, in addition to educational attainment and urbanization, the linear regressions indicate that the colorectal cancer mortality rate is heavily influenced by initial conditions and factors that affect the incidence rate such as diet, racial breakdown, and med ical advancements of the time period.
Lorenzo M. Waguespack 2 I. Introduction In the U.S., one of the most polarizing, yet fundamental, political topics is Healthcare, with nearly 20% of respondents in an August 2017 Gallup poll 1 naming Healthcare the most import ant problem facing the country. Whether the discussion is over access to care, quality, cost, or any other factor, many Americans a re concerned about the h ealthcare they receive and the factors that influence it and for good reason. For decades, one of the hallmarks of a d eveloped nation has been an effective healthcare system that tackles existing medical challenges and allows its residents to enjoy a quality of life that was not previously attainable. For example, in 1975, the U.S. 5 year survival rate for colorectal canc er, the second deadliest form of cancer in the U.S., was 48.6%. Just over 30 years later, in 2009, the 5 year survival rate of colorectal cancer had increased to 66.4% 2 Improvements in screening technology, preventative care, and overall treatment have pr opelled this success in the U.S., however, all U.S. state healthcare systems are not equal, and they do not report identical results in their treatment of the disease. This leaves us with a suitable indicator of healthcare quality and efficacy of treatment s that we can use to compare and contrast all 50 U.S. states: Colorectal Cancer Mortality Rate. This study will further analyze perceived determinants of healthcare quality within U.S. states, examining the correlation between colorectal cancer mortality r ate and socioeconomic factors such as real GDP per capita educational attainment, urban share of counties, racial breakdown and ch anges in the uninsured rate. 1 Ga llup, Inc. August 2017 2 National Cancer Institute, SEER Program, Cancer Stat istics.
Lorenzo M. Waguespack 3 II. Sample The sample for this study includes all 50 U.S. states for the years 2004 and 2014, however, it excludes Washington, D.C. based on its small size, lack of a state government, and the exclusive jurisdiction granted to Congress by the U.S. Constitution over t 3 Despite the exclusion of D.C. based on its political structure, the 50 U.S. states were determined to be suitable for this sample Due to federalism and the U.S. system of government, U.S. state healthcare systems a re similar, but not equal, and make differing decisions on numerous healthcare factors allowing us to compare and contrast the effects of those decisions In addition, U.S. states are relatively equal in available infrastructure to treat the disease, as a ll have at least one cancer hospital or treatment center, with 24 states having a cancer hospital currently ranked in the top 50 by U.S. News & World Report in 2017 and 40 states having one in the top 150 4 This provides us with 100 observations and allow s us to use the colorectal cancer mortality rate in U.S. states as an indicator for healthcare quality and efficacy of treatment. III. Dependent Variable As stated previously, the dependent variable being used for this study is the Colorectal Cancer Mortality Rate per 100,000 population in 2014, the most recent year of data available from the Department of Health and Human Services, and 2004, a decade earli er. I recognize that there may be some objections to using this variable as an indicator of healthcare quality in each 3 U.S. Const. art. I, Â§ 8. 4 U.S. News and World Report, 2017 Best Hospitals for Cancer.
Lorenzo M. Waguespack 4 U.S. state, however, it is also used by the U.S. Department of Health and Human Services as a measure of effective treatment 5 In additio n, the re are links between improvements in the colorectal cancer mortality rate and effective preventative care that are impossible to ignore. These links are supported in a study by Zauber  on the impact of screening on colorectal cancer mortality, which states that observational studies of screening colonoscopies suggest a reduction in mortality of over fifty percent 6 The expectation is that states with better healthcare quality will also have superior preventative care and screening infrastructure and therefore, a lower colorectal cancer mortality rate, all else being equal. IV. Ind ependent Variable s In this study, to examine and correct for the effect of the level of health insurance coverage in each state, the uninsured rate will be used as an independent variable. To measure the uninsured rate this paper uses data from the U.S. Census Bureau to determine the percentage of the population with no heal th insurance coverage throughout the entire year for each state in the years 2004 and 2014. The hypothesis is that states with lower uninsured rates will also have lower colorectal cancer mortality rates due to an expected increase in screenings and access to treatments that stems from a greater percentage o f the population being insured. This variable will undoubtedly capture decreases in the uninsured rate that are, at least in part, due to Affordable Care Act provisions. Unfortunately, because of the rel atively recent implementation 5 Department of Health and Human Services, Agency for Healthcare Research and Quality. 6 Zauber A. G. (2015). The Impact of Screening on Colorectal Cancer Mortality and Incidence Has It Really Made a Difference? Digestive Diseases and Sciences 60(3), 681 691.
Lorenzo M. Waguespack 5 of the ACA, this study will not incorporate enough data from after the enactment of the healthcare law to comment on its effects on healthcare quality and outcomes. From 2010 to 2014, numerous states enacted laws that limited local ACA implementation, opted out of reforms such as the individual and employer mandates, or chose not to expand Medicaid under the new health law. Due to this, gains made from the ACA also vary temporally and quantitatively at the state level, and this only emphasizes the need for more data and observations from years after the laws implementation to make accurate conclusions about its effects. I n Figure 4.1 below, which was drawn from a journal article authored by former President Barack Obama 7 we c an see evidence that the implementation of the Affordable Care Act was a significant catalyst for lower uninsured rates in the United States. Therefore, although this paper will not be able to make connections between the Affordable Care Act and healthcare quality, any future discussion on changes to the U.S. uninsured rate during this time period should note the relevance of the ACA to those changes: Figu re 4.1 7 Obama, B. (2016). United States Health Care Reform Progress to Date and Next Steps. J AMA 316(5), 525 532.
Lorenzo M. Waguespack 6 This study includes Real GDP per Capita of each state in chained 2009 dollars as an independent variable to reflect the impact of income on healthcare quality. Intuitively, the expectation beforehand is that higher GDP per capita will lead to improved healthcare quality and outcomes as famil ies and individuals have more money to spend on healthcare. African American % of Population (%AFR_AMERICAN) Numerous studies have shown evidence of higher colorectal cancer mortality rates for African Americans, which can be attributed to a few different factors. Most notably, regression analysis in a study by Dimou, Syrigos, and Saif  confirmed that there American predominance in right measurably deadlier than left sided colon tumors 8 To control for the effect this might have on a state s overall colorectal cancer mortality rate, this study will include the percenta ge of the population that is African American as an independent variable. Colorectal Cancer Incidence Rate (COLO_INCIDENCE) In order to account for discrepancies in factors that are linked to the incidence of colorectal cancer mortality such as sex and di et, this study will use the colorectal cancer incidence rate for each state in 2004 and 2014 as an independent variable. We would also expect of that state as the average age of diagnosis for colon cancer is 68 for men and 72 for women. In 8 Dimou, A., Syrigos, K. N., & Saif, M. W. (2009). Disparities in colorectal cancer in African Americans vs Whites: Before and after diagnosis. World Journal of Gastroenterology: WJG, 15(30), 3734 3743.
Lorenzo M. Waguespack 7 addition, screenings for colorectal cancer are not recommended until age 50 and older 9 We include this variable because we are specifically interested in the efficacy o f treatment in each state, and by correcting for the incidence of colorectal cancer we are able to analyze the ability of each state to limit colorectal cancer mortality rate regardless of the incidence rate. For example, we would expect a state with a hig her colorectal cancer incidence rate to also have a higher colorectal cancer mortality rate than other states, all else being equal. Educational Attainment ( %BACHELOR ) There is well established evidence that the level of educational attainment in a region or state can impact healthcare outcomes. For instance, a 1995 study on links between education and health published in the American Sociological Review 10 exper ience better health than the poorly educated, as indicated by high levels of self reported This paper will include a variable measuring the population of a state that is 25 and older who have earned at least a degree to account for educational attainment in a state. Based on prior research, the hypothesis is that higher levels of educational attainment in a state will be correlated with a lower colorectal can cer mortality rate. Year (YEAR) As we are using data from both 2004 and 2014, a dummy variable representing the year will be used to control for nationwide medical advancements and healthcare improvements over 9 American Cancer Society, Colo rectal Cancer Facts & Figures 2017 2019. 10 Ross, C.E., & Wu, C. (1995). The Links Between Education and Health. American Sociological Review, 60(5), 719 745.
Lorenzo M. Waguespack 8 the decade between 2004 and 2014. A value of 1 will be assigned to data from 2014, while a value of 0 will be assigned to data from 2004. Urban Share of Counties ( % URBAN COUNTY) A 2014 study on rural urban differences in access to specialist providers of colorectal cancer care in the United States 11 fo und that there is statistically significant evidence that a urban disparity exists in the density of gastroenterologists, general surgeons, and radiation affect access to these services and may negatively influence Due to this, we will include a variable measuring the share of counties in a state that are deemed u rban. This study will use the 2003 and 2013 Rural Urban Cont inuum Codes to determine whether a county is rural or urban 12 For this paper, similar to the study referenced above, c odes between 1 and 5 were determined to be urban, while codes 6 through 9 were categorized as rural. Based on the prevailing research, we hypothesize that an increase in the urban share of counties in a state will be associated with a decrease in the colorectal cancer mortality rate. 11 Aboagye, J.K., Kaiser, H.E., & Hayanga A.J. (2014). Rural Urban Differences in Access to Specialist Providers of Colorectal Cancer Care in the United States: A Physician Workforce Issue. JAMA Surgery, 149(6), 537 543. 12 U.S. Department of Agriculture, Economic Research Service, Rural Urban Co ntinuum Codes
Lorenzo M. Waguespack 9 V. Summary Statistics and Correlation Matrix The summary statistics for each variable are provided in Table 5.1 below: Table 5.1 Mean St. Dev Minimum Maximum Count COLO_MORTALITY 16.371 2.639 10.9 22.4 100 %UNINSURED 11.99 9 3.604 3.3 23.6 100 REAL_GDP 46563.3 8617.3 30509 70986 100 %AFR_AMERICAN 11.201 9.545 0.5 37.8 100 COLO_INCIDENCE 49.061 8.588 25.0 72.5 100 %BACHELOR 28.019 4.984 15.3 41.2 100 YEAR 0.5 0.503 0 1 100 %URBAN_COUNTY 51.823 24.175 9.4 100 100 In addition, due to concerns about collinearity between the independent variables in this study, a correlation matrix is provided in Table 5.2 below to show the correlation between each of the independent variables. By analyzing the table, it can be seen that no two independent variables shared a correlation coefficient with a magnitude greater than 0.58 4 :
Lorenzo M. Waguespack 10 Table 5.2 %UNINSURED REAL_GDP %AFR_AMERICAN COLO_INCIDENCE %BACHELOR YEAR %URBAN_COUNTY %UNINSURED 1.00 REAL_GDP 0.299 1.00 %AFR_AMERICAN 0.197 0.077 1.00 COLO_INCIDENCE 0.134 0.232 0.0 60 1.00 %BACHELOR 0.448 0.584 0.094 0.385 1.00 YEAR 0.338 0.138 0.053 0.472 0.245 1.00 %URBAN_COUNTY 0.222 0.338 0.326 0.054 0.420 0.045 1.00 VI. Regression Results Model: Y = 0 1 X 1 2 X 2 3 X 3 4 X 4 5 X 5 6 X 6 7 Z 7 + X 1 = % Uninsured X 2 = Real GDP per Capita X 3 = % African American X 4 = Colorectal Cancer Incidence Rate X 5 = X 6 = Urban Share of Counties
Lorenzo M. Waguespack 11 Z 7 = Year (Dummy Variable) = Error Term COLO_MORTALITY 0 1 %UNINSURED 2 REAL_GDP + 3 %AFR_AMERICAN + 4 COLO_INCIDENCE 5 %BACHELOR 6 %URBAN_COUNTY 7 YEAR + ERROR In addition to the model described above, this study will also run a second regression that does not include the Colorectal Cancer Incidence Rate in order to observe the impact this has on the regression statis tics, results, and interpretation. Regression Statistics : Table 6.1 REGRESSION (1) REGRESSION (2) R SQUARED 0.7 62 0.678 ADJUSTED R SQUARED 0.7 44 0.657 OBSERVATIONS 100 100 Regression Results: Table 6.2 : Regression 1 COEFFICIENT ST. ERROR P VALUE %UNINSURED 8. 059 8 5.414 0.140 REAL_GDP 0.0000299 0.000019 5 0.128 %AFR_AMERICAN 7.9529 1.621 0.00000398 COLO_INCIDENCE 0.1335 0.023 0.000000146 %BACHELOR 9.0114 4.299 0.0388 YEAR 2.1623 0.361 0.0000000394 %URBAN_COUNTY 1.6504 0.704 0.0212 INTERCEPT 11.0307 2.527 0.0000331
Lorenzo M. Waguespack 12 Table 6.3: Regression 2 COEFFICIENT ST. ERROR P VALUE %UNINSURED 9 342 5.171 0.074 REAL_GDP 0.0000209 0.0000225 0.355 %AFR_AMERICAN 8 975 1.863 0.00000566 %BACHELOR 21 081 4.328 0.00000454 YEAR 3.374 0.337 1.941 x 10 16 %URBAN_COUNTY 0.834 0.797 0.298 INTERCEPT 23.541 1.445 5.518 x 10 29 Uninsured Rate (%UNINSURED) The regression resul ts for this variable were not statistically significant at the 95% original hypothesis is not supported The p values for the first and second regressions were 0.140 and 0.074 respectfully I n considering any results from this variable i t should be noted that this paper does not distinguish between the type of healthcar e insurance, and therefore, this paper would not be able to make any reasonable conclusions on specific poli cies such as Medicaid expansion. Moreover, due to the recent implementation of Affor dable Care Act provisions, additional years of data will be necessary to properly observe the effect of the legislation on the uninsured rate, the colorectal cancer mortality rate, and healthcare quality overall. Real GDP per Capita ( REAL_GDP ) W ith a p value of 0.1 28 the res ult for Real GDP per Capita is not stat istically signific ant at the 95% confidence level in our main regression Similarly, the variable also did not produce a statistically significant result in the second regression, with a p value of 0.355 African American % of Population (%AFR_AMERICAN)
Lorenzo M. Waguespack 13 A s we hypothesized from prior research into colorectal cancer mortality and race, the main regression results for this variable were statistically significant at the 99% confidence level, with the coe fficient indicating a relatively small increase in the co lorectal cancer mortality rate of .079 per 100,0 0 0 for each percentage point increase in the percentage of the population that is African American. Additionally, in our second regression, the results were statistically significant and indicated a slightly larger increase in the colorectal cancer mortality rate for each per centage point increase in the variable, further supporting our hypothesis. Finally, t his result also supports the findings of Dimou, Syrigos, and Saif (2009) which were referenced earlier. Colorectal Cancer Incidence Rate (COLO_INCIDENCE) While we did not include this in our second regression, the main regression result s for the variable which accounted for discrepa ncies in factors affecting the colorectal cancer incidence r ate such as se x, age, and diet in each state, were significant at the 99% confidence level with a p value of 0.000000 15 The resu lts support the hypothesis made earlier that a higher colorectal cancer incidence rate would lead to a higher colorecta l cancer mortality rat e as well, with the coefficient of 0.138 implying an increase in the colorectal cancer mortality rate of 0.134 per 100,000 for each increase in the colorectal cancer incidence rate of 1 per 100,000. Educational Attainment ( % BACHELOR ) The results for our variable measuring educational attainment were statistically significant in both regressions and support the original hypothesis Intuitively, this is what would be expected, as the links between higher education and improved healthcare are well established, however, it is interesting that this connection was still seen in our study of the U.S., where educational attai nment is already high er than others even among OECD countries possible that among states or countries with much lower educational attainment the coefficient for this
Lorenzo M. Waguespack 14 variable would be much higher, as it is reasonable to expect educational attainment to have a decreasing marginal benefit on healthcare. Year (YEAR) As we expected due to technological change, t he main regression results for our dummy variable controlling for the year support our original hypothesis and allow us to reject the null hypothesis with statistical significance at the 99% confidence level due to a p value of 0.00000 00 4 The coefficient, calcula ted to be negative at 2.16, implies a decrease in the colorectal cancer mortality rate of 2.16 per 100,000 over the ten year period from 2004 to 2014, most likely due to technological and medical advancements. In the second regression, the variable was st atistically significant at an even lower p value, adding additional support for the validity of our findings. Urban Share of Counties (%URBAN_COUNTY ) In our main regression, the results for the variable measuring the urban share of counties support the or iginal hypothesis that a higher urban share of counties will lead to a lower colorectal cancer mortality rate. In the second regression, however, the variable was not statistically significant at any generally accepted level. The coefficient from our main regression also indicates a very limited impact on the colorectal cancer mortality rate of just .0165 per 100,000 for a percentage point increase in the urban share of counties. VII. Conclusion The regression results particula rly with our first model, indicate t hat colorectal cancer mortality, at least in U.S. states, is heavily dependent on initial conditions such as racial
Lorenzo M. Waguespack 15 breakdown and factors that affect colorectal cancer incidence. In addition, this study finds that educat ional attainment, technological advancement over time and urbanization are all statistically significant in indicating the colorectal cancer mortality rate of a state. Contrary to our expectations, which were based on intuition and a study by Zauber (2015 ) 13 the uninsured rate was not associated with a change in the colorectal cancer mortality rate at a statistically significant level. This is possibly due to the relatively recent implementation of the Affordable Care Act and its associated provisions, which will most likely take several years to see the full impact on healthcare quality and outcomes. Therefore, it is the recommendation of this paper that further research is done when additional years of data are available. 13 Zauber, A. G. (2015). The Impact of Screening on Colorectal Cancer Mortality and Incidence Has It Really Made a Difference? Digestive Diseases and Sciences 60(3), 681 691.
Lorenzo M. Waguespack 16 Works Cited Aboagye, J.K., Kai ser, H.E., & Hayanga, A.J. (2014). Rural Urban Differences in Access to Specialist Providers of Colorectal Cancer Care in the United States: A Physician Workforce Issue. JAMA Surgery, 149(6), 537 543. American Cancer Society. (2017). Colorectum at a Glance Retrieved from https://cancerstatisticscenter.cancer.org/#!/cancer site/Colorectum Dimou, A., Syrigos, K. N., & Saif, M. W. (2009). Disparities in colorectal cancer in African Americans vs Whites: Before and after diagnosis. World Journal of Gastroenterology: WJG, 15(30), 3734 3743. Gallup, Inc. (2017). Most Important Problems Facing the U.S. Retrieved from http://news.gallup.com/poll/219020/government top problem race immigration.aspx?g_source=CATEGORY_HEALTHCARE&g_medium=topic&g_campaign=til es National Cancer Institute. (2014). Cancer Stat Facts: Colorectal Cancer. Retrieved from https://seer.cancer.gov/statfacts/html/colorect.html Obama, B. (2016). United States Health Care Re form Progress to Date and Next Steps. JAMA, 316(5), 525 532. U.S. Const. art. I, Â§ 8. U.S. News & World Report (2017 ). Best Hospitals for Cancer. Retrieved from https://health.usnews.com/best hospitals/rankings/cancer United States Bureau of Economic Analysis. (2004, 2014). Per capita real GDP by state (chained 2009 dollars) Retrieved from https://www. bea.gov/iTable/index_regional.cfm United States Census Bureau. (2004, 2014). ACS Educational Attainment by Degree Level and Age Group. Retrieved from
Lorenzo M. Waguespack 17 http://www.higheredinfo.org/dbrowser/?year=2014&level=nation&mode=data&state=0&submea sure=232 United States Census Bureau. (2004, 2014) Current P opulation Survey. Retrieved from https://www.census.gov/cps/data/cpstablecreator.html United States Department of Agriculture. (2003 2013 ). Rural Urban Continuum Codes. Retrieved from https://www.ers.usda.gov/data products/rural urban continuum codes/ United States Department of Health and Human Services. (2004, 2014). National Healthcare Quality and Disp arities Reports. Retrieved from https://nhqrnet.ahrq.gov/inhqrdr/data/submit United States Department of Health and Human Services. (2014 ). United States Cancer Statistics: 1999 2014 Incidence an d Mortality Web based Report. Retrieved from https://www.cdc.gov/cancer/colorectal/statistics/index.htm Zauber A. G. (2015). The Impact of Screening on Colorectal Cancer Mortality and Incidence Has It Really Made a Difference? Digestive Diseases and Sciences 60(3), 681 691.