The study sample included 83 431 nonelderly adults and 49 197 households, and this analysis was performed at the household level. The pooled sample years 2012 and 2013 indicate the pre-ACA period; 2014 and 2015, the post-ACA period. Figures for out-of-pocket and premium spending are the sum of expenditures for all household members. The percentage exceeding the threshold represents the percentage of nonelderly adults who are members of families for whom total family spending divided by total family income exceeded affordability (19.5%). FPL indicates federal poverty level. For all comparisons, P = .17 for the test of a difference between the income disparities in prevalence of high-burden spending in the pre-ACA cohort vs post-ACA cohort.
eMethods. Detailed Descriptions of Covariates Used in all Adjusted Multivariable Models
eTable 1. Secondary Analysis: Post-ACA Changes in Mean Out-of-Pocket and Premium Contribution Spending, Adjusted for Medical Utilization
eTable 2. Secondary Analysis: Post-ACA Odds of Exceeding Affordability Thresholds for Out-of-Pocket Expenses and Premiums, Adjusted for Medical Utilization
eTable 3. Placebo Test for Change in Mean Spending in Pre-ACA period, 2012 to 2013
eTable 4. Placebo Test for Change in High-Burden Spending in Pre-ACA period, 2012 to 2013
eTable 5. Part 1 of 2-Part Model: of Post-ACA Changes in Likelihood of Out-of-Pocket and Premium Contribution Spending
eTable 6. Part 2 of 2-Part Model: Post-ACA Changes in Mean Out-of-Pocket and Premium Contribution Spending Among Individuals With Any Amount of Spending (Nonzero Spending)
eTable 7. Post-ACA Changes in Mean Out-of-Pocket and Premium Contribution Spending, Unadjusted
eTable 8. Post-ACA Odds of Exceeding Affordability Thresholds for Household Spending on Out-of-Pocket Expenses and Premiums, Unadjusted
eFigure. Income-Related Disparities in the Prevalence of High-Burden Spending Before and After the ACA
Customize your JAMA Network experience by selecting one or more topics from the list below.
Goldman AL, Woolhandler S, Himmelstein DU, Bor DH, McCormick D. Out-of-Pocket Spending and Premium Contributions After Implementation of the Affordable Care Act. JAMA Intern Med. 2018;178(3):347–355. doi:10.1001/jamainternmed.2017.8060
Was implementation of the Affordable Care Act associated with reduced spending on out-of-pocket medical expenses and household premium contributions among nonelderly adults?
In this nationally representative survey of validated spending data from 83 431 US adults, mean out-of-pocket spending decreased by 11.9% in the first 2 years after the insurance expansions, driven by reductions among persons eligible for the Medicaid expansion and those eligible for cost-sharing and premium subsidies on health insurance exchanges. Premium contributions increased by 12.1%, mainly owing to large increases in the higher-income group, whereas total health spending by households decreased in the Medicaid-eligible (lowest-income) group by 16.0%.
Implementation of the Affordable Care Act was associated with reduced out-of-pocket spending for US medical care, particularly among those with lower incomes, but not with reduced premiums.
The Affordable Care Act (ACA) was associated with a reduced number of Americans who reported being unable to afford medical care, but changes in actual health spending by households are not known.
To estimate changes in household spending on health care nationwide after implementation of the ACA.
Design, Setting, and Participants
Population-based data from the Medical Expenditure Panel Survey from January 1, 2012, through December 31, 2015, and multivariable regression were used to examine changes in out-of-pocket spending, premium contributions, and total health spending (out-of-pocket plus premiums) after the ACA’s coverage expansions on January 1, 2014. The study population included a nationally representative sample of US adults aged 18 to 64 years (n = 83 431). In addition, changes were assessed in the likelihood of exceeding affordability thresholds for each outcome and spending changes for income subgroups defined under the ACA to determine program eligibility at 138% or less, 139% to 250%, 251% to 400%, and greater than 400% of the federal poverty level (FPL).
Implementation of the ACA’s major insurance programs on January 1, 2014.
Main Outcomes and Measures
Mean individual-level out-of-pocket spending and premium payments and the percentage of persons experiencing high-burden spending, defined as more than 10% of family income for out-of-pocket expenses, more than 9.5% for premium payments, and more than 19.5% for out-of-pocket plus premium payments.
In this nationally representative survey of 83 431 adults (weighted frequency, 49.1% men and 50.9% women; median age, 40.3 years; interquartile range, 28.6-52.4 years), ACA implementation was associated with an 11.9% decrease (95% CI, −17.1% to −6.4%; P < .001) in mean out-of-pocket spending in the full sample, a 21.4% decrease (95% CI, −30.1% to −11.5%; P < .001) in the lowest-income group (≤138% of the FPL), an 18.5% decrease (95% CI, −27.0% to −9.0%; P < .001) in the low-income group (139%-250% of the FPL), and a 12.8% decrease (95% CI, −22.1% to −2.4%; P = .02) in the middle-income group (251%-400% of the FPL). Mean premium spending increased in the full sample (12.1%; 95% CI, 1.9%-23.3%) and the higher-income group (22.9%; 95% CI, 5.5%-43.1%). Combined out-of-pocket plus premium spending decreased in the lowest-income group only (−16.0%; 95% CI, −27.6% to −2.6%). The odds of household out-of-pocket spending exceeding 10% of family income decreased in the full sample (odds ratio [OR], 0.80; 95% CI, 0.70-0.90) and in the lowest-income group (OR, 0.80; 95% CI, 0.67-0.97). The odds of high-burden premium spending increased in the middle-income group (OR, 1.28; 95% CI, 1.03-1.59).
Conclusions and Relevance
Implementation of the ACA was associated with reduced out-of-pocket spending, particularly for low-income persons. However, many of these individuals continue to experience high-burden out-of-pocket and premium spending. Repeal or substantial reversal of the ACA would especially harm poor and low-income Americans.
As implied by its name, the Affordable Care Act (ACA) aimed to ameliorate the burden of health care costs on US families. Before its implementation in 2014, households faced increasing costs for their share of health insurance premiums1,2 and out-of-pocket costs, such as deductibles and co-payments.3,4 In 2012, 41% of nonelderly US adults reported struggling to pay medical bills or going into medical debt.5
The ACA expanded health insurance coverage, mostly by offering free or subsidized coverage to low- and middle-income families. Although these measures halved the uninsured rate,6 the acquisition of insurance among the previously uninsured might have influenced health care spending in different ways. The ACA’s mandate that individuals purchase insurance meeting minimum coverage standards7 might have increased households’ spending on premiums, and out-of-pocket spending may have increased owing to greater use of medical services for previously unaddressed medical needs. Conversely, the ACA may have led to decreased out-of-pocket spending because of the financial protections afforded by insurance coverage, and premiums may have decreased because of the ACA’s regulatory reforms of the insurance market.8
Premiums for insurance policies sold on the ACA exchanges have increased since the ACA’s debut,9 although subsidies have blunted the effects of these increases for many individuals. Meanwhile, private insurance deductibles increased by 12% from 2015 to 2016,10 outstripping income growth.11
Data from the first year after ACA implementation suggest that self-reported household out-of-pocket spending decreased,12 but the full consequences of the ACA may not be apparent until more time has elapsed. To date, no studies have quantified the association of the ACA nationally with households’ premium contributions and out-of-pocket expenditures beyond the first year of implementation. We analyzed nationally representative data to assess the association of the ACA with the overall burden of health care spending borne by families.
We analyzed data from the Medical Expenditure Panel Survey (MEPS) from January 1, 2012, through December 31, 2015. The MEPS is a nationwide survey of the US civilian noninstitutionalized population that includes approximately 15 000 households annually. We used data from the MEPS Person Round Plan file to analyze families’ contributions to private health insurance premiums13 and data from the MEPS Household Component to assess family income and out-of-pocket spending on health care. This study was approved by the institutional review board of the Cambridge Health Alliance, Boston, Massachusetts, which waived the need for informed consent for this analysis of deidentified data.
Our study included adults aged 18 to 64 years. The ACA’s individual mandate and main insurance expansion programs became effective on January 1, 2014. Therefore, we considered the pre-ACA period to be 2012 and 2013. We considered 2014 and 2015 (the most recent available data) to be the post-ACA period.
We stratified the sample into 4 income groups according to statutory thresholds set by the ACA to define eligibility for Medicaid or subsidized insurance. Lowest income included those with family incomes of 138% or less of the federal poverty level (FPL), the group eligible for Medicaid in states that expanded the program under the ACA. Low income included those with family incomes of 139% to 250% of the FPL, most of whom were eligible for subsidized premiums and reduced cost sharing on the ACA’s exchanges. Middle income included those with family incomes of 251% to 400% of the FPL, who generally qualified for premium subsidies but not for reduced cost sharing. Higher income included those with family incomes above 400% of FPL, who are not eligible for subsidies. The FPL for a family of 3 was $20 090 in 2015.14
The MEPS Household Component provides estimates of annual total out-of-pocket health care spending by individuals, which includes payments for inpatient stays, outpatient encounters, physician fees, and prescription drugs. Respondent-reported out-of-pocket expenditures are supplemented with information from providers of medical services and pharmacies and reflect the amounts actually paid.15 Expenditure data were collected in household interviews in the MEPS Household Component and in interviews with clinicians and pharmacies in the MEPS Medical Provider Component. Data from the MEPS Household Component on out-of-pocket spending were verified with data from the MEPS Medical Provider Component when available, and in the event of a discrepancy, data from the MEPS Medical Provider Component were considered to be more accurate. If no data from either source were available, the amount was imputed. Premium payments tabulated in the MEPS Person Round Plan were self-reported and reflected the total amount that a respondent’s family paid out of pocket after ACA-related subsidies were applied.16
To capture families’ spending burden for out-of-pocket expenses, we summed out-of-pocket expenditures for all family members and divided this figure by the family’s income. We similarly calculated family premium expenditures and combined health expenditures (family-level out-of-pocket plus family premium) as a proportion of family income. We defined family using the MEPS definition as cohabitating individuals linked by blood, marriage, adoption, foster care, or self-identification as a family unit (rather than by the health insurance unit, which only includes spouses and dependents), as have other studies of high-burden spending.17,18 Although family eligibility for ACA programs is determined according to the health insurance unit, we chose the MEPS definition because resources are typically shared by all family members living in a household. We calculated the proportion of nonelderly individuals belonging to families whose health care spending exceeded affordability thresholds (defined below) in the periods before and after the ACA’s major programs began.
We used spending exceeding 10% of family income as the high-burden threshold for out-of-pocket spending, a widely used marker of underinsurance.19,20 We also analyzed the bottom 2 income groups using 5% of family income as the out-of-pocket threshold, an accepted measure of affordability for low-income populations.19,21 For premium payments, we used 9.5% of family income as the high-burden threshold based on the ACA’s provision that allows persons whose employer-offered insurance premium exceeds 9.5% of family income to forgo the employer-based option in favor of an exchange plan.22 Finally, we summed the out-of-pocket and premium thresholds to create a combined definition of high-burden total health spending of 19.5% (10% out-of-pocket plus 9.5% premium thresholds).
We compared annual out-of-pocket and premium payments before (2012-2013) and after (2014-2015) implementation of the ACA using several approaches. We fit separate linear regression models to the logarithms of the following 3 outcomes: out-of-pocket spending, premium payments, and out-of-pocket spending plus premium payments. To retain zero spenders, we added US $1.00 to all figures before log transformation. Regression results are reported as percentage of change in mean annual US dollars. We chose this approach because it yields a summary measure of the consequences of the ACA for the whole population. We also conducted sensitivity analyses of these outcomes using 2-part models23,24: a logit model for the probability of any spending, followed by a linear model of logged expenditure, applied only to individuals with any spending.
All multivariable models included terms for key sociodemographic characteristics, self-reported health status, income, and employment status to control for secular changes in the composition of the US population and potential economic changes occurring during the study period (eMethods in the Supplement provides further description of covariates). To determine whether potential reductions in spending in the post-ACA period were owing to decreased use of heath care services, we performed a secondary analysis that also adjusted for the following 4 measures of resource use: outpatient visits, inpatient admissions, emergency department visits, and filled prescriptions. Detailed descriptions of utilization variables are presented in eTable 1 in the Supplement.
We used multivariable logistic regression to estimate the likelihood of a nonelderly adult’s family exceeding high-burden spending thresholds in the pre-ACA vs the post-ACA periods, controlling for previously described demographic variables. We also conducted a secondary analysis of high-burden spending that adjusted for use of health care resources (eTable 2 in the Supplement). All linear and logistic regression analyses were conducted for the overall study population and for each income group.
To examine the association of the ACA with disparities in medical spending burdens for each income group, we performed logistic regression analyses in which the outcome was whether the respondent exceeded the spending thresholds. The prognostic variables included time (pre-ACA vs post-ACA periods), income group, and an interaction term between time and income group, in addition to adjustments for sociodemographic characteristics, health status, employment, and income. We used type 3 Wald tests on the interaction term to assess whether the income-related pattern of high-burden spending changed in the post-ACA period. P < .05 indicated significance.
To identify potential secular trends in the pre-ACA period, we conducted a placebo analysis in which we compared spending in 2012 with that in 201325 (eTables 3 and 4 in the Supplement). We used weights provided by the MEPS to generate nationally representative estimates and SAS software procedures (version 9.4; SAS Institute, Inc) that account for the survey’s complex sample design. All expenditure data was adjusted to 2015 US dollars using the Consumer Price Index.26
Our sample included data from 83 431 nonelderly adults (weighted frequency, 49.1% men and 50.9% women; median age, 40.3 years; interquartile range, 28.6-52.4 years) and 49 197 households. Sociodemographic and health characteristics changed little from the pre-ACA to post-ACA periods, although employment increased slightly from 76.6% to 78.6% (Table 1). Insurance coverage increased in the full sample and in each income group. The proportion of persons with Medicaid increased by 3.2 percentage points, whereas 3.4% of the study population reported participation in exchange plans at the close of 2015.
For the full sample, adjusted mean annual out-of-pocket spending decreased by 11.9% (95% CI, −17.1% to −6.4%; P < .001), from $619.23 to $548.48, in the post-ACA period compared with the pre-ACA period (Table 2). We observed significant decreases in out-of-pocket spending for the lowest-income (−21.4%; 95% CI, −30.1% to −11.5%; P < .001), low-income (−18.5%; 95% CI, −27.0% to −9.0%; P < .001), and middle-income (−12.8%; 95% CI, −22.1% to −2.4%; P = .02) groups but not for the higher-income group.
Adjusted mean premium spending increased by 12.1% (95% CI, 1.9% to 23.3%; P = .02) in the full sample and by 22.9% (95% CI, 5.5% to 43.1%; P = .01) in the higher-income group. Premium spending did not change in other income groups. Combined health spending (out-of-pocket plus premium contribution) decreased by 16.0% (95% CI, −27.6% to −2.6%; P = .02) in the lowest-income group but did not change in the full sample or the other income groups.
In our analysis using the 2-part model, likelihood of any out-of-pocket spending decreased slightly (odds ratio [OR], 0.94; 95% CI%, 0.89-0.99) and likelihood of making any premium contribution increased (OR, 1.07; 95% CI, 1.01-1.14) in the full sample (eTable 5 in the Supplement). Results of the second part of the model, which analyzed changes in logged expenditure among those with nonzero spending, were substantially similar to the results of our primary analysis (eTable 6 in the Supplement).
In secondary analyses adjusted for use of health care resources, we observed decreases in mean out-of-pocket spending for the full sample (−15.3%; 95% CI, −19.0% to −11.4%; P < .001) and for each income group (range, −20.7% [95% CI, −28.2% to −12.4%] to −12.1% [95% CI, −18.4% to −5.3%]; P ≤ .002) but no changes in premium spending (eTable 1 in the Supplement). Combined health spending decreased significantly in the full sample (−6.5%; 95% CI, −12.5% to 0.2%; P = .045) and in the lowest-income (−13.5%; 95% CI, −24.1% to −1.4%; P = .03) and low-income (−21.4%; 95% CI, −34.1 to −6.2%; P = .01) groups after adjustment for use of health care services (eTable 1 in the Supplement). Results of unadjusted analyses of out-of-pocket and premium spending were similar except that the decrease in total health spending in the lowest-income group (−12.2%; 95% CI, −25.4% to 3.5%; P = .12) did not reach statistical significance (eTable 7 in the Supplement).
In the full sample, the adjusted odds of a nonelderly adult’s family incurring out-of-pocket expenditures exceeding 10% of their income decreased by 20.5% (OR, 0.80; 95% CI, 0.70-0.90) after ACA implementation (Table 3). In the income-stratified analysis, those in the lowest-income group had a 19.6% reduction in the adjusted odds of experiencing high-burden spending after ACA implementation (OR, 0.80; 95% CI, 0.67-0.97); we found no changes for other income groups. In the subgroup analysis using a 5% threshold, the lowest-income group experienced a 21.3% decrease (OR, 0.79; 95% CI, 0.68-0.91) and the low-income group experienced a 26.7% decrease (OR, 0.73; 95% CI, 0.63-0.86) in the adjusted odds of exceeding the threshold for out-of-pocket spending.
The middle-income group experienced a 28.3% increase (OR, 1.28; 95% CI, 1.03-1.59) in the odds of premium spending above 9.5% of family income, but we found no change for the full sample or for other income groups. Results of unadjusted analyses of high-burden spending were similar and are presented in eTable 8 in the Supplement. High-burden total spending did not change significantly in the full sample or in any income group. After adjustment for use of health care services in the analysis of high-burden out-of-pocket spending, the lowest-income group no longer showed a significant change, whereas the low-income group had a 36.7% decrease (OR, 0.63; 95% CI, 0.45-0.90) in the odds of spending more than 10% of family income after implementation of the ACA (eTable 2 in the Supplement).
The Figure displays the prevalence of high-burden combined spending (out-of-pocket plus premium spending) for pre-ACA and post-ACA cohorts within each income group. Before the ACA, a steep stepwise increase in high-burden total spending occurred as family income decreased. The ACA did not alter this pattern; the odds of experiencing high-burden spending were approximately 16 times greater among the poor than among the higher-income group before the ACA and 14 times higher after the ACA. Results of analyses for high-burden out-of-pocket and premium spending are shown in the eFigure in the Supplement.
Our placebo test identified pre-ACA decreases for the lowest-income group in adjusted mean out-of-pocket spending (−15.8%; 95% CI, −24.5% to −6.1%; P = .002) and total health spending (−18.2%; 95% CI, −28.8% to −5.9%; P = .01). No other year-to-year changes in mean or high-burden spending were noted in the pre-ACA period (eTables 3 and 4 in the Supplement).
In the first 2 years of implementation, the ACA was associated with a decrease in mean out-of-pocket spending for the overall population, driven by decreases among the lowest- and low-income groups and a reduction in high-burden out-of-pocket spending overall and among the lowest-income group. Mean premium payments increased moderately, whereas the prevalence of high-burden combined health spending and income-based inequalities in high-burden spending did not change.
The ACA was designed to reduce the costs borne directly by households by expanding Medicaid coverage for the lowest-income Americans, offering subsidized coverage for low-income individuals, and mandating more comprehensive benefits and cost-sharing limits for many others with private coverage. The legislation required private insurers to cover 10 essential health benefits,27 prohibited co-payments and deductibles for several preventive services, and banned exclusions for preexisting conditions.28
The ACA’s Medicaid expansion, which generally required enrollees to pay neither premiums nor co-payments, likely accounts for our finding that out-of-pocket spending decreased among the lowest-income group after the law’s implementation. This consequence might have been greater if all states had accepted the ACA’s Medicaid expansion. In the analysis of high-burden spending in this group, controlling for use of health care services slightly attenuated the decrease but did not substantially alter our findings. The differences noted for the lowest-income group in our placebo test in the pre-ACA period suggests that our results for this group may also have been influenced by secular trends or early expansions of Medicaid in some states. Because only 17.8% of adults in the lowest-income group paid insurance premiums before the ACA (eTable 5 in the Supplement), we are not surprised that the legislation had little effect on premiums in this group.
The reduction in out-of-pocket spending for low-income individuals (139%-250% of FPL) that we observed suggests that the ACA’s exchange plans and cost-sharing subsidies were associated with a decreased burden of health care costs for this population. Our finding from models controlling for use of health care services suggests that decreased use of medical services and drugs did not drive this decrease.
The decrease in mean out-of-pocket expenditures by the middle-income group may reflect the modest 5.1% increase in coverage gained in this group under the ACA. Although individuals in this income group were not eligible for subsidized cost-sharing on the exchange, the ACA’s provision that eliminated cost-sharing for preventive services may have decreased out-of-pocket spending. Although many individuals in this group were eligible for premium assistance through the ACA exchanges, the subsidies were apparently insufficient to prevent growth of premium contributions for this group or to reduce their total health spending.
Previous studies6,29 have highlighted the ACA’s successes. About half of the previously uninsured population gained coverage.6 Several other indicators also improved, such as the likelihood of having access to affordable care and self-reported health status.29
Few prior studies have provided information on the ACA’s effects on US households’ health care expenditures. Earlier analyses found decreases in out-of-pocket spending or in the likelihood of any out-of-pocket expenditure in subgroups of the population, such as the uninsured,30 low-income adults in California,31 and adults with incomes less than 400% of FPL who purchased nongroup market coverage.32 A non–peer-reviewed report12 using data from the Current Population Survey found small decreases in total health spending (out-of-pocket plus premiums) in states with higher marketplace enrollment in 2014, the first year of implementation; however, that analysis did not assess nationwide pre-ACA to post-ACA changes. Our findings are generally consistent with these prior studies’ findings of decreases in financial burden after the ACA but builds on them by using the most reliable data on spending in a nationally representative sample, examining changes overall and for all relevant income groups, and by analyzing data beyond the first year after implementation.
The ACA did not have a greater effect on out-of-pocket spending for several reasons. First, only a small proportion of Americans—6.5% according to our data—became newly insured after the ACA. Second, about 28 million Americans remain uninsured.6 Third, many of those with coverage continued to incur high costs; in 2016, individual deductibles were a mean of $3064 in exchange Silver plans (the metal tier chosen by most exchange enrollees)33,34 compared with $1478 in employer-sponsored plans.35 We cannot exclude the possibility that the ACA had other positive effects on financial well-being, such as a reduction in unpaid bills or medical debt, but these data were not available in the MEPS.
Countervailing factors may have contributed to the failure of the ACA to help reduce premium payments. Although those who switched to subsidized ACA plans may have experienced reductions in premium payments, others who were previously uninsured may have newly begun paying premiums. Depending on income, such premiums could be high. For instance, 85% of persons who remained uninsured after shopping on the exchanges failed to purchase coverage because it was too expensive.36 Insurance reforms, such as protections for preexisting conditions, may have lowered premiums for medically complicated individuals but increased them for others.37
Several limitations of our study should be noted. First, only 2 years of post-ACA MEPS data were available, although marketplace enrollment increased little in 2016.38 The MEPS data do not provide information on Medicare part B premiums; thus, the financial burden on households that included an elderly family member may be underestimated. Data on premiums are self-reported and therefore subject to error. In addition, no data were available on whether uninsured respondents paid the ACA-mandated penalty for lack of health insurance or whether families had outstanding unpaid medical bills. A few of the ACA’s provisions (eg, a ban on lifetime caps in insurance coverage) came into effect in 2010 and would not be captured in our analysis. Last, information on MEPS respondents’ state of residence is not available in the public-use MEPS data, precluding analysis of whether the consequences of the ACA differed between Medicaid expansion and nonexpansion states.
Repeal of the ACA was under consideration in Congress several times in the past year, and the future of the ACA remains uncertain. Our findings carry several implications for the health reform debate. First, the ACA was associated with moderate reductions in the cost burden for lowest-, low-, and middle-income households, which represents incremental but important progress. Repealing or otherwise dismantling the legislation without a suitable replacement could cause financial harm to many lower-income families. As of this writing, the Senate tax bill includes a repeal of the individual mandate. If enacted, the numbers of uninsured persons will increase, along with their out-of-pocket costs. Premiums will likely increase because healthier people will exit the insurance pool.
Second, medical expenses currently consume a large share of many families’ incomes and compound income inequalities.39 Reforms to the ACA that could improve household spending burdens include expanding Medicaid in all states, increasing the generosity of cost-sharing and premium subsidies, and increasing the actuarial values of standard exchange plans.40 International experience suggests that a universal, comprehensive national health insurance program would most effectively reduce household spending and ameliorate disparities.41
Accepted for Publication: November 24, 2017.
Corresponding Author: Anna L. Goldman, MD, MPA, Cambridge Health Alliance, 1493 Cambridge St, Cambridge, MA 02139 (firstname.lastname@example.org).
Published Online: January 22, 2018. doi:10.1001/jamainternmed.2017.8060
Author Contributions: Dr Goldman had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: All authors.
Acquisition, analysis, or interpretation of data: Goldman, Woolhandler, Himmelstein, McCormick.
Drafting of the manuscript: Goldman.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Goldman, Woolhandler, McCormick.
Obtained funding: Bor.
Administrative, technical, or material support: Woolhandler, Bor, McCormick.
Study supervision: Himmelstein, Bor, McCormick.
Conflict of Interest Disclosures: None reported.
Funding/Support: The study had no direct funding. Dr Goldman’s salary is supported by NRSA (National Research Service Award) for Primary Care grant T32HP12706 from the National Institutes of Health. Dr McCormick received salary support for his contributions from internal funds of the Department of Medicine at Cambridge Health Alliance.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.