US Health Care Spending by Race and Ethnicity, 2002-2016

Abstract

Importance: Measuring health care spending by race and ethnicity is important for understanding patterns in utilization and treatment.

Objective: To estimate, identify, and account for differences in health care spending by race and ethnicity from 2002 through 2016 in the US.

Design, setting, and participants: This exploratory study included data from 7.3 million health system visits, admissions, or prescriptions captured in the Medical Expenditure Panel Survey (2002-2016) and the Medicare Current Beneficiary Survey (2002-2012), which were combined with the insured population and notified case estimates from the National Health Interview Survey (2002; 2016) and health care spending estimates from the Disease Expenditure project (1996-2016).

Exposure: Six mutually exclusive self-reported race and ethnicity groups.

Main outcomes and measures: Total and age-standardized health care spending per person by race and ethnicity for each year from 2002 through 2016 by type of care. Health care spending per notified case by race and ethnicity for key diseases in 2016. Differences in health care spending across race and ethnicity groups were decomposed into differences in utilization rate vs differences in price and intensity of care.

Results: In 2016, an estimated $2.4 trillion (95% uncertainty interval [UI], $2.4 trillion-$2.4 trillion) was spent on health care across the 6 types of care included in this study. The estimated age-standardized total health care spending per person in 2016 was $7649 (95% UI, $6129-$8814) for American Indian and Alaska Native (non-Hispanic) individuals; $4692 (95% UI, $4068-$5202) for Asian, Native Hawaiian, and Pacific Islander (non-Hispanic) individuals; $7361 (95% UI, $6917-$7797) for Black (non-Hispanic) individuals; $6025 (95% UI, $5703-$6373) for Hispanic individuals; $9276 (95% UI, $8066-$10 601) for individuals categorized as multiple races (non-Hispanic); and $8141 (95% UI, $8038-$8258) for White (non-Hispanic) individuals, who accounted for an estimated 72% (95% UI, 71%-73%) of health care spending. After adjusting for population size and age, White individuals received an estimated 15% (95% UI, 13%-17%; P < .001) more spending on ambulatory care than the all-population mean. Black (non-Hispanic) individuals received an estimated 26% (95% UI, 19%-32%; P < .001) less spending than the all-population mean on ambulatory care but received 19% (95% UI, 3%-32%; P = .02) more on inpatient and 12% (95% UI, 4%-24%; P = .04) more on emergency department care. Hispanic individuals received an estimated 33% (95% UI, 26%-37%; P < .001) less spending per person on ambulatory care than the all-population mean. Asian, Native Hawaiian, and Pacific Islander (non-Hispanic) individuals received less spending than the all-population mean on all types of care except dental (all P < .001), while American Indian and Alaska Native (non-Hispanic) individuals had more spending on emergency department care than the all-population mean (estimated 90% more; 95% UI, 11%-165%; P = .04), and multiple-race (non-Hispanic) individuals had more spending on emergency department care than the all-population mean (estimated 40% more; 95% UI, 19%-63%; P = .006). All 18 of the statistically significant race and ethnicity spending differences by type of care corresponded with differences in utilization. These differences persisted when controlling for underlying disease burden.

Conclusions and relevance: In the US from 2002 through 2016, health care spending varied by race and ethnicity across different types of care even after adjusting for age and health conditions. Further research is needed to determine current health care spending by race and ethnicity, including spending related to the COVID-19 pandemic.

Figure 1.. Estimated Health Care Spending by Race and Ethnicity From 2002 Through 2016

A, Total health care spending for 6 mutually exclusive race and ethnicity groups in the US from 2002 and 2016. B, The proportion of total spending, the total US population, and the total US insured population in each race and ethnicity group. Uncertainty intervals were based on completing the analysis on 1000 independently bootstrapped samples of the underlying data.

Figure 2.. Estimated Age-Standardized Health Care Spending per Person by Race and Ethnicity and Type of Care in 2016

Error bars indicate 95% uncertainty intervals (UIs). UIs were based on completing the analysis on 1000 independently bootstrapped samples of the underlying data. The coefficient of variation indicates dispersion relative to the mean. A larger coefficient of variation means a relatively larger amount of variation. Values in parentheses represent uncertainty intervals for each coefficient of variation. Types of care are mutually exclusive.

Figure 3.. Age-Standardized Spending per Person Attributable to Utilization and Price and Intensity of Services in 2016

Each bar represents the relative difference in spending for each group, relative to all-population spending for each type of care. Types of care are mutually exclusive. Green arrows indicate the relative differences attributed to differences in utilization, while the orange arrows indicate the relative difference attributed to differences in price and intensity of treatment. Decomposition was not performed for nursing facility care due to the lack of utilization data for this type of care. The corresponding dollar amounts by race and type of care can be found in Figure 2.

Figure 4.. Statistically Significant Differences in Estimated Health Care Spending and Utilization in 2016

Each cell reports the relative difference of spending or utilization rates per notified case, comparing that specific race or ethnicity group and the condition-specific all-population mean for the population aged 20 years or older. Cells without values reported indicate where estimates were repressed because the differences from the all-population mean were not statistically significant (α = .05). Dental care was omitted because these conditions are not treated as part of dental care. Nursing facility care was omitted due to a lack of utilization data. The corresponding estimates of relative differences in spending per utilization by type of care can be found in eTable 6B in the Supplement. In addition, eTable 6A presents these estimates as absolute differences in dollars and utilization counts rather than percent differences. COPD indicates chronic obstructive pulmonary disease.