Determinants of and Trends in Total and Condition Specific Health Care Spending Per Privately Insured Adult
- 1. Department of Health Policy and Management, Emory University, USA
This paper examines the key factors associated with the level and changes in total spending per privately insured person 18-64 over the past decade. We also examine changes in out of pocket spending over time. This will include estimating the difference in out of pocket spending by race, gender, ethnicity as well as the mix of medical conditions under treatment. Nearly 160 million Americans are covered through employer sponsored insurance and just under 35 million through individual/nongroup plans . Understanding trends in private insurance spending and the underlying factors associated with these trends provides important insights into the health care system overall.
The analysis starts by looking at trends in total per capita spending, total private insurance spending and out of pocket payments from 2010 through 2020. Then the analysis focuses on spending among patients with the most expensive medical conditions. This includes:
- Mental health disorders
- Heart Failure and Heart Disease
In addition, we examine the impact of multiple chronic conditions on the level and growth in private insurance spending over time.
MATERIALS AND METHODS
The analysis examines how changes in out of pocket spending, rising chronic disease prevalence, and other demographic factors affect total spending per capita. A major focus will be on trends and differences over time between Non-Hispanic White, NonHispanic Black, Non-Hispanic other adults and Hispanic Adults on the level and growth in health care spending.
Data for this analysis came from the 2010 - 2020 Medical Expenditure Panel Survey (MEPS-HC) medical conditions, events, and consolidated data files . MEPS-HC is a nationally representative survey of the civilian noninstitutionalized population conducted by the Agency for Healthcare Research and Quality (AHRQ). The survey collects self-reported medical condition information, insurance coverage, patient demographics, health services utilization and health care spending.
We used the medical condition and event files to define eleven treated conditions used in our analysis: diabetes, hypertension, hyperlipidemia, mental health disorders, cancer, trauma, heart disease, rheumatoid arthritis, asthma, chronic kidney disease (CKD), and chronic obstructive pulmonary disease (COPD). The regression models focused on the top five conditions, heart disease, cancer, mental disorders, trauma and COPD/asthma. MEPS respondents self-reported medical conditions that were then professionally coded into ICD-9-CM diagnosis codes for years 2010 to 2015 and ICD-10-CM codes for 2016 to 2020. Clinical classification software was then used to collapse the ICD-9-CM codes into mutually exclusive clinical classification categories (CCC) for 2010 to 2015 and refined clinical classification software was used to collapse the ICD-10-CM into mutually exclusive refined clinical classification categories (CCR) for 2018-2020. We defined each of the eleven conditions based on CCC (2010-2015), ICD-10-CM (2016-2017), and CCR (2018-2020) codes with one or more associated inpatient, outpatient, office-based, emergency department (ED), home health, or prescription medication health care event. (Appendix A).
Condition specific total spending included all spending on health care events that occurred during a given calendar year and were directly related to treating the condition. More specifically, we summed inpatient, outpatient, ED, office-based, home health, and prescription drug expenditures. When the health care event was associated with multiple conditions, the expenditures for that event were split evenly across the conditions.
Our analyses were limited to adults with 12 months of private insurance, ages 18 to 64 years old. Any respondent with a missing survey weight or missing values in any of the model covariates were excluded resulting in an analytic sample of 92,792.
We used generalized linear model (GLM) with gamma distribution and log-link function to predict three types of annual expenditures (total amount for all health care utilization, total amount paid by private insurance, and total out-of-pocket) in three time periods: 2010 – 2013, 2014 – 2017, and 2018 – 2020. We then estimated counterfactuals for the latter two time periods by calculating predicted spending using the characteristics of the 2010 – 2013 patients with the regression coefficients for each of the respective latter two time periods.
In each of our models, we controlled for patient characteristics, including age, sex, race/ethnicity, education, region, health status, income level, smoking, and total number of treated conditions.
We used Stata, version 17.0, for data analysis . Sample weights and survey estimation commands were used to adjust for the complex survey design of MEPS. All spending amounts are presented in 2020 dollars, using the GDP implicit price deflator .
Table 1 presents 10-year trends in average private insurance and out-of-pocket spending, as well as trends in chronic disease prevalence between 2010 and 2020. Real per insured private insurance spending increased from $3,540 in 2010 to $4,967 by 2020, an average annual increase of 3.4 percent. These results may have been impacted by the COVID-19 pandemic. The Centers for Medicare and Medicaid Services (CMS) reported that private health spending declined by 1.2 percent in 2020 . However, CMS reports that the decline was largely due to a reduction in private insurance enrollment. We report average annual increases per insured and the results are virtually identical. The average increase between 2000 and 2009 was 3.47 percent and between 2000 and 2010 was 3.45 percent. We also did the analysis dropping the year 2020 using 2018-2019 as the third time period. The regression results were virtually identical to the results reported here.
The average annual increase in spending between 2000 and 2009 Out-of-pocket spending among privately insured adult increased from $703 in 2010 to $897 by 2020, an average annual increase of 2.5 percent.
Trends in the prevalence of chronic disease among privately insured adults were relatively stable over time. Fifty two percent adults had no treated chronic conditions in 2010 compared to 57 percent by 2020. The percent of adults with 3 or more chronic conditions treated decreased slightly over time, falling from 9.4 percent in 2010 to 7 percent in 2020.
Trends in chronic disease prevalence were generally stable over time as well. Diabetes prevalence remained around 6.2 percent from 2010 through 2020. Hypertension prevalence showed a slight decline from 18.5 to 15.3 percent. The remaining chronic conditions examined were virtually constant over time.
Table 2 examines changes in real per insured spending by the most expensive and prevalent chronic conditions. Between 2010 and 2020 private insurance spending per adult rose from $3,716 to $4,893 a 31.7 percent increase. The 2020 spending level however was most likely impacted by COVID. The pattern of spending changes varied widely across chronic condition. Per insured spending on hyperlipidemia declined nearly 50 percent per person over the ten-year period, falling from $735 to $371. This decline reflects a dramatic change in the mix of statins with a large rise in the use of generic statins and a 90 percent reduction over a ten-year period for brand name statins . Similarly, increased use of generic drugs to treat hypertension (generic Lipitor) rose over the ten-year period resulting in 16 percent decrease in spending.
In contrast, spending on other chronic conditions increased sharply. Spending to treat diabetes more than doubled, rising from $2,380 in 2010 to $4,875 by 2020. Cancer spending showed the next largest rise, from $6,997 in 2010 to $8,960 by 2020, over a 28 percent increase. Spending on trauma cases and mental disorders increased by roughly the same amount 25 and 23 percent respectfully over the ten-year period.
Table 3 presents the full regression results for private insurance spending focusing on the number of chronic conditions treated. We focus on the results from 2010-2013 and 2018- 2020 through the 2014-2017 results are reported as well. In both periods, women incurred higher expenses (about $2,000) than men. Self-reported health status also impacted spending, private insurance spending for those that reported poor health were over $15,000 higher compared to those reporting excellent health (results similar in both periods).
Private insurance spending was considerably lower for racial and ethnic minorities. In 2010-2013 and 2018-2020 spending was approximately $900 lower for Non-Hispanic other adults compared to Non-Hispanic White adults. Insurance spending was $713 lower in 2010-2013 and $619 lower in 2018-2020 for NonHispanic Black adults compared to Non-Hispanic White adults. Finally, spending on Hispanic adults was over $500 lower in the earlier period, and $213 lower in the latter period compared to non-Hispanic White adults.
In both time periods, higher levels of education were associated with higher private insurance spending. In the 2018- 2020 time period, adults with a college degree spending $2,700 more than adults with no degree.
Somewhat surprisingly smokers incurred lower health care spending in both periods, ranging from $700 to over $840 lower per year.
Finally, even when controlling for self-reported health status, the number of chronic conditions treated has a substantial impact on private insurance spending.
The incremental spending for each additional condition was about $2,000 in 2010-2013 and was similar in 2018-2020. The major difference was the incremental spending associated with moving from 3 to 4 or more conditions treated in 2018-2020. In this case, the additional spending was over $4,200 higher (p<.10).
Table 4 presents the same set of results for out-of-pocket spending trends over time. Like total spending, females spent approximately $380 more out of pocket compared to males in both time periods. Reported health status also influenced spending as those reporting in poor health spent $1,000 to $1,200 more out-of-pocket in the two time periods. Racial minorities spent less out-of-pocket compared to non-Hispanic White adults. In the latest period, out-of-pocket spending was approximately $350 to $419 lower per year for racial and ethnic minorities.
Out-of-pocket was over $200 higher for adults aged 50 to 64 compared to those under 50. Out-of-pocket spending also increased with higher levels of education. Adults with a college degree spent $340 to $400 more per year out-of-pocket compared to those without a high school diploma. As before smokers spend less out-of-pocket ($120 to $160) compared to non-smokers.
As before, even accounting for self-reported health status, out-of-pocket spending increased sharply with the number of chronic conditions treated. In both time periods examined, outof-pocket spending was $1,200 to nearly $1,300 more per year for those with 4 or more conditions compared to those with no chronic conditions.
Out-of-pocket spending for those with one chronic condition was higher in 2018-2020 ($212, p<.05) compared to 2010-2013. Out-of-pocket spending for other chronically ill patients were similar in the two time periods.
The results in Table 5 estimates the marginal impact of race and ethnicity and the most expensive and prevalent chronic conditions on private insurance spending in two time periods (2010-2013 and 2018 and 2020) as well as over the two periods. Other demographic results were similar in these models that were reported above and therefore are not shown.
In both time periods, health spending was approximately $2,000 higher for females. Spending on racial and ethnic minorities in the 2010-2013 time period were uniformly lower compared to non-Hispanic White adults. This ranged from $508 per year lower for Hispanic Adults to over $900 per year for nonHispanic other adults.
Spending increased sharply with the number of chronic conditions treated. In 2010-2013 incremental spending was $1,983 higher and those with 4 or more conditions $7,659 higher for those with 4 or more conditions compared to those with no chronic conditions. Spending on chronic disease increased sharply over the ten-year period. Private insurance spending for each category of the number of chronic conditions increased by nearly $1,200 (for those with one condition) to over $4,500 more in 2018-2020 for those with 4 or more conditions. These changes within each category were all statistically significant (p<.05).
The results presented in Table 6 estimate the incremental out-of-pocket expenditures by race and ethnicity and number of chronic conditions treated. Out-of-pocket spending increased sharply as the number of chronic conditions treated increased. Relative to adults with no chronic conditions, out-of-pocket spending was $415 higher for those with 1 condition rising to nearly $1,300 higher for those with 4 or more conditions treated. As before, out-of-pocket spending for racial and ethnic minorities were lower than from Non-Hispanic White adults.
Out-of-pocket spending rose over the two time periods for adults with one chronic condition (p<.05). For these adults, spending was $212 higher in 2018-2020 compared to 2010- 2013.
We next estimate the change in private insurance spending for five of the most expensive chronic conditions over time. Using a Wald Chi-Square test we compare the impact of changes in chronic care spending on total private insurance spending (Table 7), and out-of-pocket spending (Table 8), for three time periods; 2014-2017 and 2010-2013 and 2018-2020 and 2010-2013.
Spending for four of the five chronic conditions examined increased significantly between 2018-2020 and 2010-2013. Cancer spending was $4,282 higher in the latter period compared to the earlier period. Similarly spending to treat heart disease was $3,417 higher, trauma spending $1.300 and treatment of mental disorders $965 higher in 2018-2020 compared to 2010- 2013. Spending to treat heart disease was also higher ($2,592) in 2014-2017 compared to 2010-2013.
Out-of-pocket spending to treat two chronic conditions also increased over time (Table 8). Spending to treat trauma patients was $201 higher in 2018-2020 compared to 2010-2013. Out-ofpocket spending for patients treated for mental disorders was $280 in 2018-2020 compared to 2010-2013.
CONCLUSIONS, LIMITATIONS AND RECOMMENDATIONS
Real per capita private insurance spending increased an average of 3.4 percent per year between 2010 and 2020. Out-ofpocket spending also increased over this period but by a lower amount of 2.5 percent per year. The analysis highlights several important demographic differences in the level of both total insurance spending and out-of-pocket spending over this period. First, health care spending, even controlling for reported health status and chronic health conditions, were lower for racial and ethnic minorities compared to non-Hispanic White adults. Out-ofpocket spending was also lower for racial and ethnic minorities. These differences however have not increased over the tenyear period examined. The result also show that spending rises with educational attainment and age. Total and out-of-pocket spending was also higher for women.
As may be expected the most important determinants of total and out-of-pocket were self-reported health status and the number of chronic health care conditions treated. Total spending on those with self-reported poor health was over $15,000 higher compared to those reporting excellent health. The significantly higher spending among those with fair or poor health compared to patients with excellent health was observed even when controlling for the number of chronic health care conditions treated.
The analysis also highlighted the substantial increase in total and out-of-pocket spending among patients a greater number of chronic conditions treated. Private insurance spending for adults under treatment for 4 or more chronic conditions were over $12,000 higher compared to adults with no chronic conditions treated. Even among chronically ill adults, spending increased substantially for each additional chronic condition treated. A similar result was found for out-of-pocket spending for chronically ill adults. These high levels of out-of-pocket spending for chronically ill patients are a concern if they deter patients from adhering to medications and seeking timely treatment. One approach would be to lower or eliminate cost sharing on clinically important medications used to treat patients with highly prevalent chronic conditions.
The analysis highlights several areas of interest. The lower spending overall on racial minorities even after accounting for health status raises some issues that require additional study. The analysis also highlighted that higher spending on several chronic conditions were an important factor accounting for the growth in total private insurance spending. Additional treatment costs of cancer and heart disease were the two leading conditions accounting for the increased spending. The increased spending was not associated with higher prevalence of the conditions, they were relatively stable over the ten- year period. Higher treatment costs per case then accounted for the rise.
The analysis showed that the treatment costs of some conditions like hyperlipidemia and hypertension actually declined over time. The lower level of spending is linked to the increased use of generic medications to treat both conditions.
Finally, the results provide important information for capitated health insurance plans in general (Medicare, Medicaid, and private insurance). These plans must predict forward looking treatment costs in setting premiums or negotiation per capita payments. The fact that both reported health status and the number and mix of chronic conditions (which are used in risk adjusting per capita rates for Medicare Advantage plans) are highly predictive of spending highlights the important role that health risk assessments in addition to clinical data in predicting levels of spending. Relying solely on risk adjusting simply using claims data on clinical data could result in underpredicting next year’s spending levels. Both risk assessment information and the clinical data are two of the most important determinants of the level and change in health care spending.
The author thanks the National Pharmaceutical Council for support of the analysis.
- American Community Survey. Health insurance coverage of the nonelderly 0–64. San Francisco (CA): KFF; 2019. 2022.
- Lin SY, Baumann K, Zhou C. Trends in Use and Expenditures for Brand-name Statins After Introduction of Generic Statins in the US, 2002-2018. JAMA Netw Open. 2021; 4: e2135371
- Agency for Healthcare Research and Quality. Medical Expenditure Panel Survey: Household Component full-year files, 2010–2020. Rockville (MD): AHRQ.
- StataCorp. 2021. Stata Statistical Software: Release 17. College Station, TX: StataCorp LLC.
- Centers for Medicare and Medicaid Services, 2023.
- Organization for Economic Co-operation and Development, GDP Implicit Price Deflator in United States [USAGDPDEFAISMEI]. 2022.