US Healthcare Expenditure, Physician Distribution, and Quality of Care

Radhika Rastogi


The US spends an extraordinary amount on healthcare, both in proportion to the national GDP and relative to other economically comparable nations. Healthcare spending currently amounts to one-sixth of the national GDP,1 threatening resource allocation to other government-initiated public programs. In contrast to comparable European nations, the US spends almost 50% more of its GDP on healthcare,2 without achieving equivalent quality.3 Reinhardt et al. posit several explanations for this difference, ranging from differences in GDP to willingness to rationing care and variation in administrative overhead.2 This expenditure-quality discrepancy prompts questions to the relationship between spending and outcomes; Baicker et al. suggests more spending is associated with lower quality care, and Hussey et al. contend that the relationship is unclear to the directionality.4,5 Regardless, the United States has much to improve in both sectors.

Interestingly, these disparities emerge on a regional level within the US. There are known regional variations in spending and, in some cases, these differences have been attributed to discrepancies in the amount of care provided.6 Factors such as price or medication usage are not considered to be prominent causes of the observed spending discrepancies.7,8 Numbers of primary care physicians (PCPs) and specialists also vary by region.9 Generally, a higher number of PCPs per population is associated with more efficient care and better outcomes for patients with chronic diseases.10 Additionally, increasing the number of specialists per resident is associated with poorer quality, greater healthcare disparities between local populations, and greater expenditure.4,11 Thus, physician distribution plays a role in regional variation in spending and quality.

An additional component is variation in racial diversity within different patient populations. Latinos and African Americans have decreased access to healthcare and receive poorer quality services.12,13 While some of these issues have been attributed to the nature of the physician-patient relationship,14 Baicker et al. argue that there is a broader implication of the patients’ and providers’ locations.12 Though a pronounced relationship between geography, healthcare quality, and cost exists, the causes underlying regional variations remain unclear.

Insurance further complicates the interaction between healthcare spending and quality. Differentials in quality of and access to care depend on insurance type, particularly private insurance vs. public options or lack of insurance.15,16 The financial implications of lack of insurance are enormous for individual families, with medical costs being a leading cause of bankruptcy, as well as society in general, given the large costs that hospitals must cover.17 As hospitals typically cover uncompensated care, or care for the uninsured, greater government funding in the form of insurance would decrease hospitals’ burden through cost.18

Given the factors implicated in healthcare cost and quality, I aimed to examine the relationship between physician distribution and specialization, medical expenditure, and quality of care and determine whether insurance status and race play a role in this relationship.


State-based data was collected from several sources. The US Census Bureau was used to determine on state populations.19 The Kaiser Family Foundation website provided data on PCP, cardiologist, and oncologist distribution, percent of minority populations, healthcare expenditure, health outcomes, and insurance status.20 Healthcare expenditure was defined as the amount spent by state of patients’ residence. Health outcomes included deaths due to heart attacks per 100,000 people and deaths due to cancer per 100,000 people. Insurance information included the percentage of state residents with employer-provided insurance, individual insurance, Medicare, Medicare, or without insurance.

The Dartmouth Atlas of Health Care was used to identify measures of healthcare quality.21 Using the Institute of Medicine’s definition of quality, “the degree to which health services for individuals and populations increase the likelihood of desired health outcomes and are consistent with current professional knowledge,”22 the percentage of patients filling prescriptions within a give time period was used to measure quality. In particular, for cardiovascular care quality, I looked at the percentage of patients filling at least one prescription of beta blockers within 6 months of a heart attack and the percentage of patients filling out at least one prescription for a statin within 6 months of a heart attack. To assess psychiatric care quality, I looked at percentage of patients filling out at least one prescription of a selective serotonin reuptake inhibitor (SSRI) or a serotonin and norepinephrine reuptake inhibitors (SRNI).

Statistical analysis

In preliminary analysis, Massachusetts regularly presented as an outlier due to an abnormally high number of both PCPs and specialists per 100,000 residents and was thus discarded from the data set. The following results summarize analysis on the remaining 49 states. All physician variables (number of PCPs, cardiologists, psychiatrists, and oncologists) and state healthcare expenditure were adjusted for population.

First, regression analysis was conducted between state spending and physician distribution variables to evaluate the relationship between number of providers and spending. To evaluate whether the relationships would be preserved if the physician variables were regressed together rather than independently, a model was created using all four physician variables. The four physician variables were then correlated. The quality-provider distribution relationship was analyzed through a regression between the number of providers and the percentage of medication prescriptions filled out. A third regression was performed to evaluate whether a relationship exists between disease burden and number of specialists. Another regression was performed between the percentage of residents with employer-provided, individual, Medicare, Medicaid, or those who were uninsured with healthcare spending.

Lastly, to understand the relationship of these variables with state spending, a regression was conducted with state spending as the dependent variable and number of PCPs, psychiatrists, cardiologists, oncologists, percentage of minority residents, median age, deaths due to cancer, deaths due to heart attacks, percent with employer-provided insurance, individual insurance, Medicaid, Medicare, or no insurance, beta-blocker use, statin use, and SSRI use. All variables were adjusted for population.

All analyses were conducted using Stata/MP 13.1.


The primary result of interest was the relationship between state healthcare spending and the number of primary care or specialty physicians. The number of PCPs per 100,000 people was significantly associated with state spending (p<0.001), data-preserve-html-node="true" with every additional PCP adding $30.9 per capita. Similarly, for cardiologists, there was a significant association (p < 0.001), with state spending as each additional cardiologist increased spending by $214.1 per capita. The same relationship was preserved among psychiatrists and oncologists, with each psychiatrist adding $108.4 per capita (p < 0.001) and each oncologist, $315.5 per capita (p < 0.001).

Given that each physician variable was significantly associated with state spending, a model was built to determine whether all four variables were equally as important in explaining state spending. However, it was found that when controlling for the number of psychiatrists, PCPs, and oncologists, the number of cardiologists adjusted for population was no longer significantly associated with state spending (p=0.08). Similarly, the number of psychiatrists was no longer significantly associated with spending (p=0.31), when the model controlled for number of cardiologists, PCPs, and oncologists all adjusted for population. However, the number of PCPs still remained a significant explanatory value (p=0.001), with each additional PCP adding $32.3 per capita, when controlling for the number of specialists. The most surprising result however, was that though the number of oncologists remained significantly associated with state spending (p=0.022), the direction of the relationship was now reversed, such as that each additional oncologist per 100,000 would result in a decrease of $31.1 per capita. To explore why the number of cardiologists and psychiatrists were no longer significantly associated with spending, these two variables were correlated and the high resulting correlation coefficient (r=0.78) between the two might explain the lack of significance due to multicollinearity.

The relationship between number of specialists and quality as measured by percentage of patients picking up a drug prescription was also investigated. Beta-blocker use was significantly associated with the number of cardiologists per resident (p=0.018). For each additional cardiologist per 100,000 people, there was a 0.4% increase in the number of people who picked up at least one prescription in the six months following a heart attack. However, for the other measures, there was no significant relationship. Overall, non-significant inverse relationships were observed; increasing the number of cardiologists and psychiatrists was associated with less statin (p=0.80) and SSRI (p=0.10) use respectively.

In considering whether greater incidence of disease was associated with greater number of specialists, only number of oncologists was significantly associated with number of deaths due to cancer (p=0.001). Increasing number of cardiologists was not associated with increased number of deaths due to heart attacks (p=0.53). Similarly, more psychiatrists per population were not associated with increased incidence of residents reporting poor mental health (p=0.58). Additionally, there were no significant relationships between the percentage of people with employer-provided insurance (p=0.34), individual insurance (p=0.58), Medicaid (p=0.32) or Medicare (p=0.37) or those who were uninsured (p=0.79) and state spending.

In the final regression model, the minority status of the state (p=0.014) and the number of psychiatrists (p=0.024) emerged as significant explanatory variables of state spending, when the other variables of interest were controlled for. For each additional percent minority population, state expenditure decreased by $3299.3 per resident. On the other hand, each additional psychiatrist per 100,000 residents cost the state $76.1 per capita. The disease burden due to cancer (p=0.06), percent of residents who were uninsured (p=0.07), and the percent of residents with individual insurance (p=0.059) were borderline significant.


Significant positive relationships were found between each kind of physician and state spending. Moreover, the only quality measure significantly associated with number of specialists was the percentage of patients picking up beta-blockers within 6 months of a heart attack. In addition, only the number of oncologists in a state was significantly associated with an increased disease burden. Lastly, when all variables accounted for, the percentage of minority residents in a state as well as the number of psychiatrists were the only significant predictors of state spending.

These findings are particularly interesting in context of regional variation in healthcare cost and quality and physician distribution. Like Baicker et al, I found that increasing numbers of specialists were associated with increasing costs.4 However, in contrast to Baicker et al.’s findings about the negative relationship between numbers of PCPs and cost,4 I observed that increasing numbers of PCPs were associated with increased state spending. This may be explained by differences in measurement of cost. Baicker et al. use Medicare spending as gathered from the Dartmouth Atlas, which restricts populations to specific groups, such as the elderly or the disabled. Using state spending by state of patients’ residence as the measure of cost encompasses a greater consumer base, but does not take into account patients who travel to receive healthcare. Though Medicare is a state-sponsored program, evidence suggests that there are significant differences in state spending and Medicare spending, perhaps due to variations in the number of Medicare beneficiaries in different states.23 Results that indicate lack of association between state spending and the proportion of residents in various healthcare insurance plans further corroborate Kronick & Gilmer’s results.23 These results may allay fears that changes in the number of people enrolled in certain insurance plans will increase government expenditure on a state-level. The distinction between Baicker et al.’s and my findings has implications for the policy proposals, especially if the aim is to reduce healthcare-related spending, since the use of PCPs to address other healthcare concerns may be more costly than Baicker et. al predict.

An additional difference between our findings emerges from the analysis on quality. While a lack of association between number of specialists and quality was observed in this study, with the exception of a positive relationship between number of cardiologists and beta-blocker usage, Baicker et al. found a negative relationship between number of specialists and quality measures.4 This distinction may be explained by the use of different quality measures. While Baicker et al. analyzed records specifically used as quality metrics, this study extrapolated from the IOM definition of quality and investigated prescription patterns instead.

Unlike the conclusions in the independent regressions as well as Baicker et al.’s report,4 in which a positive relationship between spending and number of specialists was observed, a negative relationship between number of oncologists and healthcare spending was determined. This reversal in the coefficient may indicate an interaction between number of PCPs and oncologists. This variable could explain referral patterns and ways in which patients accessing multiple providers impacts the quality and cost of their care. While reports of economic and quality benefits from care coordination have been varied, these differences may depend on type and severity of disease.24 Investigating this interaction further may yield relevant findings for oncology, but also other fields, given the range of prognoses faced by patients diagnosed with cancer.

Moreover, it was found that increasing the number of minority residents was associated with decreased cost per resident. These results corroborate findings of decreased access to and utilization of healthcare based on race.11,25 While most studies distinguish between African Americans and Latinos as the two separate groups of analysis, my calculation of minority as a sum of the percentages of African American, Latino, and Asian residents did not alter the implications of the findings. It remains evident that racial disparities need to be addressed.


The most significant limitation to this study is the inability to account for severity of illness in each state. Though it is possible that there is an even distribution of severity of illness among the states, if there are specialized centers in particular areas that draw particular patient populations or environmental exposures due to industries in a particular area, variables may remain unaccounted for. In addition, the data was limited to three specialties within medicine, but expenditure from surgery as well as pediatric subspecialties would also be relevant as there are distinct forms of cost-incurring care at different stages of health. Furthermore, the data sets used were from different years, spanning 2010 to 2013, depending on the source and variable. Since there has been no significant change in population demographics in the recent years nor has the healthcare reform been adequately enacted to have impacted spending or quality measures, it is unlikely that the data would changed substantially in the time between the sets.

Policy Implications

While our results suggest that addition of physicians would increase cost, expanding primary care has empirically remedied issues of healthcare access and quality.10 Integrating PCPs into a hospital system may facilitate coordinated care through technological, personal, and institutional resources.

Moreover, involving health extenders such as insurance representatives and pharmacists, may help target the quality measures of prescription drug pick-up.26 In the case of cardiovascular medication, Bitton et al. show that better adherence is associated with better outcomes as well as lower cost as a result of fewer complications.27 Investment in physician assistant training or insurance company coordination may provide support and encourage better adherence among patients, addressing both quality and cost concerns. Furthermore, it is imperative that the roots of racial disparities be elucidated and addressed. Policy changes include heightened research efforts, enforced training on cultural sensitivity among healthcare providers, and greater involvement of social workers.

About the Author

Radhika Rastogi is a currently a senior at Harvard College, concentrating in Organismic and Evolutionary Biology with a secondary in History of Science. She hopes to address concerns about quality and access to health care in the US by pursuing medicine in the future.


  1. Martin, A. B., Lassman, D., Washington, B. & Catlin, A. Growth In US Health Spending Remained Slow In 2010; Health Share Of Gross Domestic Product Was Unchanged From 2009. Health Aff. (Millwood) 31, 208–219 (2012).
  2. Reinhardt, U. E., Hussey, P. S. & Anderson, G. F. U.S. Health Care Spending In An International Context. Health Aff. (Millwood) 23, 10–25 (2004).
  3. U.S. Health in International Perspective. (Institution of Medicine of the National Academies, 2013). at
  4. Baicker, K. Medicare Spending, The Physician Workforce, And Beneficiaries’ Quality Of Care. Health Aff. (Millwood) (2004). doi:10.1377/hlthaff.w4.184
  5. Hussey, P. S., Wertheimer, S. & Mehrotra, A. The Association Between Health Care Quality and CostA Systematic Review. Ann. Intern. Med. 158, 27–34 (2013).
  6. Fisher, E. S. et al. The Implications of Regional Variations in Medicare Spending. Part 1: The Content, Quality, and Accessibility of Care. Ann. Intern. Med. 138, 273–287 (2003).
  7. Gottlieb, D. J. et al. Prices Don’t Drive Regional Medicare Spending Variations. Health Aff. (Millwood) 29, 537–543 (2010).
  8. Zhang, Y., Baicker, K. & Newhouse, J. P. Geographic Variation in Medicare Drug Spending. N. Engl. J. Med. 363, 405–409 (2010).
  9. Rosenthal, M. B., Zaslavsky, A. & Newhouse, J. P. The Geographic Distribution of Physicians Revisited. Health Serv. Res. 40, 1931–1952 (2005).
  10. Ferrer, R. L., Hambidge, S. J. & Maly, R. C. The Essential Role of Generalists in Health Care Systems. Ann. Intern. Med. 142, 691–699 (2005).
  11. Starfield, B., Shi, L., Grover, A. & Macinko, J. The effects of specialist supply on populations’ health: assessing the evidence. Heal. Aff. Proj. Hope Suppl Web Exclusives, W5–97–W5–107 (2005).
  12. Baicker, K., Chandra, A. & Skinner, J. Geographic Variation in Health Care and the Problem of Measuring Racial Disparities. Perspect. Biol. Med. 48, 42–S53 (2005).
  13. Waidmann, T. A. & Rajan, S. Race and Ethnic Disparities in Health Care Access and Utilization: an Examination of State Variation. Med. Care Res. Rev. 57, 55–84 (2000).

  14. Saha, S., Arbelaez, J. J. & Cooper, L. A. Patient–Physician Relationships and Racial Disparities in the Quality of Health Care. Am. J. Public Health 93, 1713–1719 (2003).

  15. Asch, S. M. et al. Who Is at Greatest Risk for Receiving Poor-Quality Health Care? N. Engl. J. Med. 354, 1147–1156 (2006).

  16. Bethell, C. D. et al. A National and State Profile of Leading Health Problems and Health Care Quality for US Children: Key Insurance Disparities and Across-State Variations. Acad. Pediatr. 11, S22–S33 (2011).

  17. Himmelstein, D. U., Thorne, D., Warren, E. & Woolhandler, S. Medical bankruptcy in the United States, 2007: results of a national study. Am. J. Med. 122, 741–746 (2009).

  18. Hadley, J., Holahan, J., Coughlin, T. & Miller, D. Covering The Uninsured In 2008: Current Costs, Sources Of Payment, And Incremental Costs. Health Aff. (Millwood) 27, w399–w415 (2008).

  19. US Census Bureau, D. I. D. Population Estimates. at

  20. State Health Facts. Henry J Kais. Fam. Found. at

  21. Prescription Medication use in Medicare Part D. Dartm. Atlas Heal. Care at

  22. Lohr, K. Medicare: A Strategy for Quality Assurance. (National Academies Press, 1990).

  23. Kronick, R. & Gilmer, T. P. Medicare And Medicaid Spending Variations Are Strongly Linked Within Hospital Regions But Not At Overall State Level. Health Aff. (Millwood) 31, 948–955 (2012).

  24. Peikes D, Chen A, Schore J & Brown R. Effects of care coordination on hospitalization, quality of care, and health care expenditures among medicare beneficiaries: 15 randomized trials. JAMA 301, 603–618 (2009).

  25. Darkins, A. et al. Care Coordination/Home Telehealth: the systematic implementation of health informatics, home telehealth, and disease management to support the care of veteran patients with chronic conditions. Telemed. J. E-Heal. Off. J. Am. Telemed. Assoc. 14, 1118–1126 (2008).

  26. Gaskin, D. J., Briesacher, B. A., Limcangco, R. & Brigantti, B. L. Exploring racial and ethnic disparities in prescription drug spending and use among medicare beneficiaries. Am. J. Geriatr. Pharmacother. 4, 96–111 (2006).

  27. Indicators Spotlight: Health Care Extenders. Natl. Cent. Chronic Dis. Prev. Heal. Promot. at

  28. Bitton, A., Choudhry, N. K., Matlin, O. S., Swanton, K. & Shrank, W. H. The impact of medication adherence on coronary artery disease costs and outcomes: a systematic review. Am. J. Med. 126, 357.e7–357.e27 (2013).