Julie P.W. Bynum, MD, MPH

 Chiropractic care is a service that operates outside of the conventional medical system and is reimbursed by Medicare. Our objective was to examine the extent to which accessibility of chiropractic care affects spending on medical spine care among Medicare beneficiaries.

 Retrospective cohort study that used beneficiary relocation as a quasi-experiment.

 We used a combination of national data on provider location and Medicare claims to perform a quasi-experimental study to examine the effect of chiropractic care accessibility on healthcare spending. We identified 84,679 older adults enrolled in Medicare with a spine condition who relocated once between 2010 and 2014. For each year, we measured accessibility using the variable-distance enhanced 2-step floating catchment area method. Using data for the years before and after relocation, we estimated the effect of moving to an area of lower or higher chiropractic accessibility on spine-related spending adjusted for access to medical physicians.

 There are approximately 45,000 active chiropractors in the United States, and local accessibility varies considerably. A negative dose—response relationship was observed for spine-related spending on medical evaluation and management as well as diagnostic imaging and testing (mean differences, $20 and $40, respectively, among those exposed to increasingly higher chiropractic accessibility; 

 <.05 for both). Associations with other types of spine-related spending were not significant.

 Among older adults, access to chiropractic care may reduce medical spending on services for spine conditions.

Chiropractic care is a Medicare-reimbursed service that operates outside of the conventional medical system and provides a sizable amount of the nation’s spine care.

Medicare beneficiaries can use chiropractic care without a medical referral.

Previous studies have not considered how use of chiropractic care affects spending on costly medical services.

We found some evidence of a relationship between lower accessibility of chiropractic care and higher spending on diagnostic imaging and testing for spine conditions.

Back pain and neck pain are associated with an estimated direct cost of $86 billion to the United States economy.

 Among the top contributors to disability, back pain and neck pain outrank chronic diseases such as chronic pulmonary disease, ischemic heart disease, and diabetes.

 Although back pain can be attributed to a serious underlying medical problem, up to 85% of cases are classified as nonspecific (ie, cases without an underlying pathology).

 Approximately one-fourth of all adults will experience nonspecific back pain during a 3-month period, and at any one point in time, approximately 30% of older adults suffer from back pain.

 Older adults are particularly vulnerable to back pain and most who experience nonspecific back pain have future recurrences

; thus, it is among the most common reasons older Americans visit physicians.

With more than 100 million visits for spine conditions each year,

 the chiropractic profession is the largest health service that operates outside of the conventional medical system

 and the only such service reimbursed by Medicare.

 Medicare beneficiaries can use chiropractic care without a medical referral; however, the only reimbursable modality is manual therapy, which costs approximately $30 to $50 per visit. Chiropractors cannot be reimbursed directly for diagnostic imaging nor prescribe medications. In light of the ongoing debate regarding coverage of presumably nonessential services, there is a specific need to more rigorously examine how chiropractic care may affect national healthcare spending.

Previous observational research that has examined spending among chiropractic patients compared with medical care patients suggests that chiropractic care may reduce utilization of and expenditures on medical services for back pain.

 Traditional observational studies use assignment of patients to a specific treatment type (eg, chiropractic vs usual care) to study the cost of chiropractic care. These designs, however, are unable to account for the potential impacts from those who choose to use chiropractic care in place of medical care. To date, previous studies have neglected to consider such indirect effects of the service on the system at large.

Therefore, we performed a quasi-experimental study to examine how accessibility of chiropractic care among Medicare beneficiaries affects healthcare spending for spine conditions. To do so, we used a cohort of older adults in Medicare with back and/or neck pain who relocated and experienced a change in geographic accessibility of chiropractic care. In this way, the change in accessibility of chiropractic care served as a proxy for access to the service among Medicare beneficiaries, providing us with the unique opportunity to observe how either reducing or increasing access affects spine-related spending among older adults.

We examined the effect of accessibility of chiropractic care on spine-related spending using a quasi-experimental design. For our study, we geocoded Medicare beneficiaries and identified a cohort with back and/or neck pain who relocated once from 2010 to 2014. Then, using the variable-distance enhanced 2-step floating catchment area method

 to estimate provider to population ratios, we determined the effect of a change in accessibility of chiropractic care on spine-related spending. This method for estimating provider accessibility is the gold standard because it has a higher resolution that incorporates travel time, takes into account the location of healthcare in adjacent areas, and is less affected by choice of geographic scale.

As our study used administrative claims and publicly available data on healthcare providers, it received an expedited review by the University of Michigan Health Sciences and Behavioral Sciences Institutional Review Board.

First, we identified all older adults (65 years and older) enrolled in Medicare Part B throughout 2010 to 2014—this resulted in the identification of 16,842,729 beneficiaries (

]). For each calendar year, we merged beneficiary residential zip codes with hospital referral regions (HRRs) that represent the 306 regional US healthcare markets. We then identified 986,076 beneficiaries who relocated at least once by changing HRR. We restricted our sample to the 887,917 beneficiaries who relocated only once during the 5-year time period—634,935 had at least 1 year of data prior to and after the relocation (

Of those 634,935, we identified 84,679 beneficiaries who had a back and/or neck pain visit using an established list of 

International Classification of Disease, Ninth Revision, Clinical Modification

 Our aim was to identify a cohort of beneficiaries who had back and/or neck pain prior to relocation (ie, those most likely to be affected by accessibility of chiropractic care). Therefore, we restricted our sample to beneficiaries who had 2 separate claims for back or neck pain at least 4 months apart in the year prior to relocation.

For each year, we gathered data on provider location from historic versions of the National Plan and Provider Enumeration System (NPPES). We used provider specialty code 35 in NPPES data to identify the practice locations of chiropractors.

 We removed chiropractors who were clinically inactive in Medicare by linking NPPES provider data to the 20% Carrier file and removing any providers who did not submit a claim in the relevant year. Medicare beneficiaries were aggregated to Zip Code Tabulation Areas (ZCTAs) according to their zip code of residence for each year. ZCTAs are generalized areal representations of US Postal Service zip code service areas. Once assigned a ZCTA, beneficiaries in our study cohort were assigned measures of chiropractic care accessibility. For each ZCTA, we measured chiropractic care accessibility using an enhanced version of the variable-distance enhanced 2-step floating catchment area method first developed by Luo and Wang

Our measure was constructed based on the 2010 US Census block—level population aggregated up to ZCTA level in order to assign each patient estimates of chiropractic care accessibility. First, we calculated a drive time–based service area for each practice location. US Census block centroids within each service area provided population estimates. These estimates were summed to generate provider to population ratios for each practice location. Next, we estimated chiropractic care accessibility based on the weighted sum of all practice locations within a given drive time from each Census block centroid. For locations within a threshold distance (eg, 5, 10, or 15 miles), scores were summed and scaled according to drive time from the block centroid. These estimates of chiropractic care accessibility were then aggregated to the ZCTA in order to assign a measure of provider accessibility around each ZCTA population-weighted centroid.

Our primary dependent variable was annual spine-related spending. We used the combination of ResDAC Carrier, Medical Provider Analysis and Review (referred to as MedPAR), and Outpatient files to calculate total spending on back and neck pain diagnoses by summing across all types of inpatient and outpatient care for each calendar year. We calculated spine-related spending on inpatient and ambulatory care (office-based and outpatient claims combined) separately. To examine more subtle differences in spending, we used the Berenson-Eggers Type of Service categories for “evaluation and management” and “procedures”; furthermore, we combined the categories “imaging” and “testing.” Healthcare spending was adjusted for inflation to 2014 dollars using the Consumer Price Index for medical services.

To adjust for baseline differences, we extracted several socio&shy;demographic characteristics, including age, sex, and race/ethnicity, from the Master Beneficiary Summary File. To account for differences in health status (and changes in health status over time) for each calendar year, we used all administrative data to calculate a comorbidity score using the Charlson-Deyo Comorbidity Index.

 We adjusted for differences in accessibility of primary care physicians.

 Accessibility of primary care physicians (internal medicine, family practice, and general practitioners) was calculated using an approach identical to that described for chiropractic care.

We displayed our national estimate of chiropractic care accessibility for each ZCTA and calculated the coefficient of variation. In order to illustrate high versus low area accessibility, we converted data to standard normal deviations for 2014 ZCTAs (the most recent year of data). We also plotted the unadjusted chiropractor to population ratios collapsed by quintile (

We used data from the year before and the year after relocation (referred to herein as baseline year and postrelocation year). The estimated effect of a change in chiropractic care accessibility was evaluated in 2 ways. First, we simply identified whether the beneficiary relocated to a higher or a lower chiropractic care accessibility quintile relative to their baseline location. Second, we examined the magnitude of the increase or decrease in chiropractic care accessibility by reporting results according to the number of quintiles up or down a beneficiary moved by relocating, regardless of starting point, to identify the equivalent of a dose response. The comparison group in these analyses was all beneficiaries who moved but did not experience a change in accessibility (ie, remained in the same quintile). The 2 approaches highlight different dimensions of relocation change: The first focuses on the initial level of accessibility but aggregates change into broad categories of any increase or decrease, and the second captures how much change in accessibility occurred, with less focus on the starting level.

 The end point for all analyses was the change in spine-related spending (eg, spending on inpatient care, ambulatory care), which was normally distributed. Across baseline accessibility of chiropractic care, age and sex differed very little; however, race and health status varied (

). Therefore, we adjusted for differences in our analyses using linear regression. Specifically, our models adjusted for baseline chiropractic care accessibility, beneficiary characteristics (age, sex, race, baseline comorbidities, and change in comorbidities), and access to primary care physicians (at baseline and a change in access). Analyses were based on complete case analysis and we assumed any missing values to be missing completely at random. A 2-sided 

 value of less than .05 was considered statistically significant. Geospatial analyses were conducted using ArcGIS version 10.5 (Esri; Redlands, California) and analyses of claims data were conducted using SAS version 9.4 (SAS Institute; Cary, North Carolina).

Our analyses were restricted to beneficiaries who had evidence of a back and/or neck condition in the baseline year. In a subanalysis we repeated all analyses on a subset of 57,807 older adults who also had spine claims in the postrelocation year.

Among the 84,679 older adults enrolled in Medicare with a spine condition prior to relocating, 9.5%, 18.0%, 22.9%, 26.7%, and 22.9% resided in quintiles 1 through 5 of chiropractic care accessibility, respectively. The mean (SD) age of older adults with a spine condition who moved once was 77.0 (7.2) years; 68.4% were female, and 3.9% were black.

The total number of active chiropractors varied little from 2010 to 2014—from a high of 45,264 in 2012 to a low of 44,040 in 2014. Across the country, local accessibility varied considerably, with a higher concentration of chiropractors observed in the upper Midwest and a relatively lower concentration in the South (coefficient of variation = 93%) (

; eAppendix Figure 3). Across US ZCTAs in 2014, the mean (SD) ratio of chiropractors per 100,000 population was 15.6 (14.4). Ratios for quintile 1 were 0.0 to 4.6; quintile 2, 4.7 to 9.4; quintile 3, 9.5 to 15.0; quintile 4, 15.1 to 23.6; and quintile 5, 23.7 to 445.5. The several ZCTAs with very high chiropractor to population ratios were in the upper Midwest, where the profession originated and a large chiropractic school is located (in Davenport, Iowa).

Chiropractic Care Accessibility and Spine-Related Spending

When examining whether the beneficiaries relocated to an area with higher or lower chiropractic care accessibility relative to where they started, we found few differences in mean spending on total, inpatient, and ambulatory spine care (

). The only marginally statistically significant differences in spending were mean increases of $513 in total spine-related spending (

 = .05) and $434 in inpatient spine-related spending (

 = .06) among older adults who resided in quintile 3 and who moved to an area of lower chiropractic care accessibility. Although attenuated, some small differences were observed by category of ambulatory care spine spending (

Among residents of quintile 5 who moved to an area of lower chiropractic care accessibility, mean spending on medical procedures decreased by $53 (

 = .04). Residents of quintile 2 who moved to an area of higher chiropractic care accessibility had a decrease of $30 (

 <.001) in mean spending on spine diagnostic imaging and tests.

We found more consistent changes when examining the association of spending with the magnitude of the change in chiropractic care accessibility (

). In general, increases in spending were observed among older adults who moved to areas with lower accessibility of chiropractic care—however, the only statistically significant increase was for inpatient spending among those who moved to areas with chiropractic accessibility 2 quintiles higher than where they started (mean difference in spending of $291; 

 = .05). In addition, a small dose—response relationship was observed for ambulatory care spine spending on “evaluation and management” and “imaging and testing” (Figure 3). For both, decreases of 3 and 4 quintiles of chiropractic accessibility were associated with approximate mean increases in spending of $20 (

The inverse association between accessibility of chiropractic care and spending on spine imaging and testing persisted when restricted to the subset of older adults who had a spine condition in the postrelocation year.

Despite a sizable literature devoted to describing health services that function outside of the conventional medical system in the United States, our study is among the first to attempt to quantify the effect of access to such a service on spending using a quasi-experimental approach. We applied state-of-the-art geospatial methods both for developing the geographic accessibility measures and for using relocation as a tool for approximating causal mechanisms in observational studies. Originating from the field of economics, natural experiments are being used by a growing number of health researchers who seek to identify causal mechanisms from nonexperimental data.

 Taking advantage of unique opportunities to balance confounding factors and designs that allow for temporal observations to be made offers stronger evidence of cause and effect.

 For instance, in an influential report, Song et al used Medicare patients who relocated to examine variation in practice intensity.

 Although effects of chiropractic care accessibility on spending were inconsistent and overall small, we did find some evidence of chiropractic care accessibility affecting spending on certain types of spine-related care. Specifically, we observed small, consistent differences in spending on diagnostic imaging and tests for spine conditions. This is among the first evidence to suggest a potential reduction in medical service use due to a health service that operates primarily outside of traditional pathways of care.

To our knowledge this is also the first application of the variable-distance enhanced 2-step floating catchment area method to national provider data. This approach has several advantages over other measures, including that it (1) provides a more realistic estimation of impact of moving that incorporates travel time in both the urban and rural settings, (2) allows for movement/interaction across areas to account for accessibility in adjacent locales, and (3) is less sensitive to changes in scale because it uses a continuous measure of impedance to calculate provider to population ratios. Use of this method led us to identify relatively high accessibility of chiropractic care in the upper Midwest and lower accessibility in the South, which aligns with previous reports.

 Although our analyses cannot fully explain this pattern, the profession did originate in the upper Midwest, suggesting that cultural factors may play a role. A recent report that compared availability of different provider types found that chiropractors were more likely to locate in areas of higher income and health status.

 campaign’s mantra that “less is more” for clinical management of back pain,

 trends indicate worrisome increases in the use of opioid analgesic medications, overreliance on medical specialists, and unwarranted diagnostic imaging.

 All of these practices lead to higher healthcare costs.

Medicare spends $400 million to $500 million on chiropractic care each year,

 and chiropractic care has been scrutinized several times by the Office of the Inspector General.

 A prior study uncovered an association between higher accessibility of chiropractic care and lower reliance on primary care services, suggesting that chiropractic care may substitute for medical care.

 We sought to determine whether or not chiropractic care is merely additive to the system (ie, patients use chiropractic care who would have otherwise not used health services or patients use chiropractic care in addition to other health services). We find that chiropractic care may be associated with small savings in aspects of ambulatory care. Although these are small per-person dollar differences ($40 reduction in annual spending), the cumulative effect could be quite large given the prevalence of spine conditions among older adults. Thus, it is conceivable that CMS is recapturing a portion of the payout for coverage of chiropractic care.

Our study has several potential limitations that must be acknowledged. The cohort consisted of older adult Medicare beneficiaries who relocated once from 2010 to 2014, which could be perceived to limit the generalizability of our findings. However, considering that each beneficiary included in our cohort served as his or her own control, there is no reason to believe that the change in spending is not more broadly generalizable. A particular strength of using Medicare beneficiary relocation as an exogenous change unlikely to be related to our primary variable of interest is that it can be considered a more rigorous study design compared with traditional observational studies. However, we found differences in race and health status according to the change in chiropractic care accessibility. Despite accounting for these differences in our analyses, we cannot completely rule out residual confounding. Lastly, because our study used administrative data, we focused on spending, which is only one aspect of the important issues related to management of back and neck pain. We cannot say whether reductions in pain, improvements in quality of life, or use of pharmacological agents were any different from the data in this study. These are particularly important avenues for investigation in the context of the current opioid crisis and will be addressed in future work.

This study is among the first to examine whether access to chiropractic care, a health service that provides a significant amount of the nation’s conservative management of nonspecific back pain, has any effect on Medicare spending. We found some evidence of a relationship between lower accessibility of chiropractic care and higher spending on diagnostic imaging and testing. Future work is required to determine if indeed access to chiropractic care for Medicare beneficiaries in any way breaks the pathway to care that is discordant with practice guidelines.

 University of Michigan Institute for Social Research (MAD), Ann Arbor, MI; University of Michigan School of Nursing (MAD, OY, HL, MGT), Ann Arbor, MI; Institute for Healthcare Policy and Innovation, University of Michigan (MAD, OY, MGT, JPWB), Ann Arbor, MI; University of Michigan School of Public Health (OY), Ann Arbor, MI; National Center for Geospatial Medicine, Rice University (JT), Houston, TX; Department of Internal Medicine, Geriatric and Palliative Medicine, University of Michigan Medical School (JPWB), Ann Arbor, MI.

 This work was supported by award number 1R01AT009003 from the National Center for Complementary and Integrative Health at the National Institutes of Health. The views expressed herein do not necessarily represent the official views of the National Center for Complementary and Integrative Health nor the National Institutes of Health.

Articles

Among older adults who have a spine condition, access to chiropractic care may reduce medical spending on diagnostic services.

Access to Chiropractic Care and the Cost of Spine Conditions Among Older Adults

 Alternative payment models, such as accountable care organizations, hold provider groups accountable for an assigned patient population, but little is known about unassigned patients. We compared clinical and utilization profiles of patients attributable to a provider group with those of patients not attributable to any provider group.

 Cross-sectional study of 2012 Medicare fee-for-service beneficiaries 21 years and older.

 We applied the Medicare Shared Savings Program attribution approach to assign beneficiaries to 2 mutually exclusive categories: attributable or unattributable. We compared attributable and unattributable beneficiaries according to demographics, dual eligibility for Medicaid, nursing home residency, clinical comorbidities, annual service utilization, annual spending, and 1- and 2-year mortality. We estimated multivariate regression models describing correlates of attribution status.

 Most beneficiaries (88%) were attributable to a provider group. The remaining 12% were unattributable. Beneficiaries unattributable to any provider group were more likely to be younger, male, and from a minority group; to have disability as the basis for enrollment; and to live in high-poverty areas. Unattributable beneficiaries included 3 distinct subgroups: nonusers of care, decedents, and those with healthcare service use but no qualifying evaluation and management visits. Many unattributable Medicare beneficiaries had minimal use of healthcare services, with the exception of a small subgroup of beneficiaries who died within the attribution year.

 Attribution approaches that more fully capture unattributable patients with low service use and patients near the end of life should be considered to reward population health efforts and improve end-of-life care.

Given the reliance of alternative payment models on assignment of patients to provider groups, we compared clinical and utilization profiles of patients attributable to a provider group with those of patients unattributable to any provider group.

We estimated multivariate regression models describing correlates of attribution status.

Most beneficiaries (88%) were attributable to a provider group, whereas 12% remained unattributable.

Beneficiaries unattributable to any provider group included 3 distinct subgroups: nonusers of care, decedents, and those with healthcare service use but no qualifying evaluation and management visits.

Attribution approaches that more fully capture unattributable patients with low service use and patients near the end of life should be considered to reward population health efforts and to improve end-of-life care.

By holding networks of healthcare providers responsible for the total cost and quality of care for a designated population, the accountable care organization (ACO) model creates incentives to coordinate care across providers, reduce unnecessary spending, and improve the quality of care.

 and by January 2017, Medicare held ACO contracts with 525 organizations serving more than 10 million beneficiaries.

 Medicare ACO contract participants that meet quality benchmarks are eligible to share the savings that they generate.

 An ACO’s performance is evaluated based on Medicare methods of measurement and attribution. Researchers have evaluated outcomes of ACOs among attributable beneficiaries,

 but no study has described unattributable beneficiaries.

In Medicare ACO contracts, beneficiaries are attributed to organizations based on their use of primary care services from eligible providers,

 so attribution can influence an organization’s performance under an ACO contract. Among organizations serving patients with complex clinical and psychosocial needs, for example, quality metrics may be hard to achieve, giving organizations incentives to avoid accepting such patients. Conversely, ACO participants that are effectively managing population health may not achieve shared savings for patients who, appropriately, do not use

the primary care services that would make them

eligible for attribution. ACOs have little financial incentive to deliver preventive care that might decrease the chance that healthy patients are attributed to an ACO.

 As advanced payment models mature to include downside risk, and as models like Comprehensive Primary Care expand, it is important to understand which patients are left out so that policy makers can develop regulations that encourage participation and improved care in new payment models.

To date, we have no information on the composition of patients who are not attributable to any provider group under Medicare Shared Savings Program (MSSP) regulations. This paper examines beneficiary characteristics associated with attribution and compares hospitalization, mortality, and spending across attributable and unattributable beneficiaries. Results from our analyses can guide policy on whether additional actions are necessary to adequately give provider participants incentives to improve population health and to ensure that vulnerable beneficiaries—who may benefit the most from improved care coordination&mdash;are not excluded from new payment models.

Using Medicare claims and CMS attribution rules for the most widely adopted model, the MSSP,

 we categorized beneficiaries into 2 mutually exclusive groups: patients who were attributable to provider groups and patients who were not attributable to any provider group. We included only beneficiaries with full parts A and B coverage, limiting the sample to those for whom traditional fee-for-service (FFS) Medicare was the primary payer. We combined all beneficiaries attributable to provider organizations (ie, MSSP, Pioneer ACO, and non-ACO medical groups) into a single category because the purpose of our research was to investigate the characteristics of those falling through the cracks of the current attribution methodology. In 2012, Medicare ACO contracts included 32 Pioneer program participants and 114 MSSP participants (over the first performance period from April 2012 or July 2012 through December 2013) that were responsible for more than 2 million beneficiaries.

 Although Pioneer program participants, which were responsible for nearly 700,000 beneficiaries in 2012,

 faced downside risk for spending, virtually all MSSP participants (n = 110) were in Track 1 and eligible for upside savings only.

 We conducted cross-sectional analyses examining characteristics of beneficiaries according to their attribution status.

Beneficiary Attribution to Provider Organizations 

Following MSSP’s 2-step attribution process, a beneficiary who has at least 1 face-to-face outpatient evaluation and management (E&M) visit is assigned to the provider group that has the highest allowed charges by primary care clinicians (ie, general practice, family practice, internal medicine, and geriatric medicine practitioners) for those visits.

 Patients not seeing primary care clinicians are attributed based on visits to qualifying non—primary care clinicians (ie, physicians in other specialties, nurse practitioners, physician assistants, and clinical nurse specialists).

 Beneficiaries who received care only from nonqualifying clinical providers (eg, interventional cardiologists or certified registered nurses) or in nonstandard settings (eg, emergency departments [EDs]) or who had no visits at all were not able to be assigned to any provider group using current MSSP attribution rules; they formed our “unattributable” group.

For beneficiaries in parts A and B who were 21 years or older, data were drawn from a 40% random sample of 2012 Medicare FFS claims and 2012-2013 beneficiary summary files. Beneficiary summary files provided patient demographics, enrollment status, and date of death. Attribution status, diagnoses, utilization, and spending were obtained using claims. We chose 2012 because that was the most recent year of claims available without the suppression of claims indicating substance use disorder in research files

 and within the period of Medicare ACO implementation.

Long-term residence in nursing homes was determined from the 2012 Minimum Data Set (MDS). Patients’ zip codes were used to identify hospital referral regions (HRRs) of residence; HRRs represent regional healthcare markets in the United States defined by the Dartmouth Atlas. We further used beneficiaries’ zip codes to identify rural and urban areas

 and whether they lived in high-poverty Census tracts (≥20% of residents below the federal poverty line).

The main outcome of interest was whether a beneficiary was attributable to a provider group or not. We estimated correlations between attribution and beneficiary characteristics, including age, sex, race/ethnicity (ie, non-Hispanic white, non-Hispanic black, Hispanic, Asian or Pacific Islander, or other race), residence in a high-poverty Census tract or a rural area, dual eligibility for Medicaid, residence in long-term nursing home (≥100 days in a nursing home according to MDS), disability as a basis of original Medicare entitlement, and date of death.

International Classification of Diseases, Ninth Revision

 codes on claims to identify beneficiaries’ comorbid conditions in 2011 using Hierarchical Condition Categories developed by CMS

]). Chronic conditions included dichotomous variables for cancer, cerebrovascular disease and stroke, chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), connective tissue disease, coronary artery disease, dementia, diabetes, hematologic/thrombotic disease, liver disease, mental health conditions (severe mental illness, substance use disorder, and depression), paralysis (not stroke), Parkinson’s/Huntington disease, peripheral vascular disease, and renal disease. We identified these conditions in 2011 to account for beneficiary diagnosis history and for their chronic nature and association with mortality and costs. We categorized patients with 0 observed chronic conditions, with multiple observed chronic conditions, and with unknown diagnoses in 2011 due to lack of medical claims. We reported utilization (ambulatory care—sensitive condition admissions, admissions to acute care or critical access hospitals, E&M visits, ED visits not leading to admission) and 2012 spending (Medicare payments to all providers, total and by category: inpatient, physician services, home health, and hospice).

To distinguish those who were unattributable because attribution rules may exclude end-of-life care, we examined mortality in 2012 and 2013 (among those alive January 1, 2013). This approach disentangles the mechanical relationship between attribution and death from clinical complexity.

We compared characteristics of attributable and unattributable Medicare enrollees—based on a 40% random sample to minimize computing time&mdash;using multivariate logistic models. We modeled attribution status as a function of demographics, dual eligibility for Medicaid, nursing home residency, disability, measures of chronic condition history, and HRR fixed effects.

We considered the attribution status of those with dual eligibility for Medicaid, who totaled more than 10 million people as of 2016, due to their low socioeconomic status, complex clinical needs, and elevated medical spending.

This study included about 13,000,000 FFS Medicare beneficiaries in 2012. Most of these FFS Medicare beneficiaries (88%) were able to be attributed to a provider group, leaving 12% unattributable (

Compared with attributable beneficiaries, unattributable beneficiaries tended to be younger (65 vs 71 years), male (60% vs 43%), from a minority race or ethnic group (28% vs 18%), living in a high-poverty area (25% vs 22%; 

), and less likely to be dually eligible for Medicaid but no less likely to reside in a rural area (

). Unattributable beneficiaries had fewer observed chronic conditions and were less likely to have 2 or more chronic conditions in 2011 (5.3% vs 22.3%; Table 1) compared with attributable beneficiaries; however, people with no service use or very few visits have no or fewer opportunities for diagnoses to be observed. Mortality was higher in the unattributable group in 2012 (6.6% vs 4.3%; Table 1) but lower in 2013 (2.8% vs 4.9%). More beneficiaries younger than 65 years were unattributable than those older than 65 years (20% vs 10%; Figure 1). We do not report 

 values of differences in variables’ means and proportions across attribution groups, which were all statistically significant (

In terms of geographic distribution, the proportion of unattributable beneficiaries ranged from 6% to 22% across HRRs (

); HRRs with the fewest unattributable beneficiaries (less than 10%) were located primarily in parts of the Midwest, South, and Northeast regions.

not shown, we examined 2011 physician supply in the market—a proxy for access&mdash;and found no evidence for differences across attribution categories.

In the unattributable group, we identified a subpopulation with no healthcare service use or spending in 2012 (63% of the unattributable) and a group that died in 2012 (6%). The remaining beneficiaries (31%) used healthcare services, including primary care services, that did not qualify for attribution (

). Among the unattributable nonusers of care in 2012, few (3.6%) had 1 or more reported diagnoses in 2011. However, spending averaged $9000 in 2011 in this group, primarily from ED visits and hospitalization (not shown).

Unattributable patients with healthcare service use who died (user-decedents) spent $19,000, on average (including $12,000 of inpatient spending and $5000 of hospice spending; Table 2). More than half (60%) of the unattributable decedents died within the first 3 months of 2012, compared with 20% of decedents attributable to provider organizations (

). Of the more than 80,000 unattributable user-decedents, 67% had 1 or more chronic condition diagnoses in claims in 2011; dementia, CHF, renal disease, diabetes, and COPD were common.

The third group of unattributable beneficiaries (with healthcare use but no qualifying primary care visits) averaged more than $3000 in annual spending, which was one-third of the spending of attributable patients. Very few of these beneficiaries had nonqualifying E&M visits (n = 57), whereas others had other healthcare use in nonattribution settings. Of the unattributable nondecedent healthcare users, 31% had reported clinical conditions in the prior year, with severe mental illness, diabetes, renal diseases, dementia, and COPD being most common.

In multivariate regression analyses, older and female beneficiaries were less likely to be unattributable (0.1 percentage-point decrease per year of age and 4.5 percentage points less likely among women; 

). Minorities and those living in high-poverty Census tracts were more likely to be unattributable. Nursing home residents and dual-eligible beneficiaries were less likely than others to be unattributable to provider groups (4.3 and 4.2 percentage points lower probability, respectively; 

 <.001). Disability as original reason for entitlement was associated with a 4.5 percentage-point increase in the likelihood of being unattributable. Beneficiaries with 0 observed chronic condition diagnoses and those with unknown chronic condition history (nonusers of care and those not in both parts A and B FFS Medicare in 2011) in 2011 were more likely to be unattributable in 2012 relative to those with observed chronic conditions (2.4 and 20.4-22.3 percentage points higher probability, respectively; 

Because Medicaid coverage complements Medicare benefits for certain low-income adults, we explored how attribution status differed in the 22% of our sample who were dually eligible for Medicaid benefits. Although they were less likely than other beneficiaries to be unattributable (Figure 1), results from multivariate analyses of dual-eligible enrollees were consistent with those found in the overall sample (

), with 2 exceptions: Asian/Pacific Islander ethnicity and disability status were both associated with a decrease in the likelihood of being unattributable.

Because existing evidence on ACOs focuses exclusively on attributable Medicare beneficiaries,

 we studied the excluded group of unattributable patients to understand how these omitted groups might influence performance in Medicare ACO programs.

 Using MSSP attribution methods, nearly 12% of beneficiaries were unattributable in 2012. This group was heterogeneous, composed of decedents, beneficiaries using no services, and those using services ineligible for attribution under common methods. The majority of unattributable beneficiaries had no encounter with the healthcare system at all, which may be due to financial and nonfinancial barriers that limit access to care.

 A small but costly group, 6% of unattributable beneficiaries, died in 2012; among them, inpatient and hospice service use was common. The remaining 31% of unattributable beneficiaries had minimal healthcare use, largely for urgent or hospital-based services.

The distribution of unattributable beneficiaries aligned with regional spending patterns in Medicare.

 Beneficiaries living in low-cost areas, such as Minnesota, had higher proportions of unattributable beneficiaries. This pattern indicates that alternative payment models in these regions, ironically, may not reward organizations for their activities targeting healthier groups who use few services. For example, ACOs in low-cost regions were evaluated against lower financial benchmarks than were ACOs in high-cost regions. Beginning in 2017, cost benchmark calculations were improved to integrate regional factors, making an ACO’s benchmark less reflective of its historical spending and more dependent on FFS expenditures in its region.

 Two changes to attribution might capture more low-utilization beneficiaries. First, Track 3 of the MSSP, with 16 participating organizations in 2016, and the Next Generation ACO model—another demonstration project in the CMS Innovation Center&mdash;are currently piloting attestation, allowing beneficiaries to voluntarily choose an ACO,

 which would allow for accountability for preventive services in healthier populations with little or no observed utilization. Second, “sticky” attribution, or allowing utilization in a prior year to count in the current year for those with no use in the current year, could capture more healthy patients; it could also capture more individuals who are very sick and use non—primary care settings due to death early in the year. Although attestation has the benefit of permitting patients to choose, sticky attribution may be more feasible in existing programs because it can be done using just claims.

Facilitating the attribution of healthy patients and those with minimal use through any method may improve the consistency and continuity of care for patients. Because ACOs are responsible for both cost and quality of care for assigned patients, attribution of healthy patients with no healthcare use, as well as of those with some healthcare use but nonqualifying visits for current attribution purposes, would reward providers for care delivered to these patients, without rewarding unnecessary care.

Attribution of nonusers may more appropriately contribute to an ACO’s shared savings, which are conditional on meeting predetermined quality measure benchmarks, and these benchmarks reflect beneficiaries classified as at-risk populations—for example, in benchmark year(s)&mdash;who currently do not use services.

 Moreover, financial benchmarks are calculated using attributed beneficiaries’ historical healthcare use. The current attribution methodology provides incentives for ACOs to bring in healthy patients for primary care visits, but attributing nonusers through alternative methods may be warranted to appropriately reward providers without inducing potentially unnecessary primary care visits.

 A practical solution that would address those nonusers of care who used healthcare in previous years would be to assign them based on sticky attribution.

Furthermore, healthcare providers and policy makers should give greater attention to the consequential subpopulation of decedents (ie, end-of-life patients) currently falling outside of the responsibility of organizations participating in ACOs. One-fourth of Medicare spending is devoted to care in the last year of life,

 so relative savings on end-of-life care contribute outsized savings to the Medicare program. There are also opportunities for improvements in care quality and preference-based care. Organizations that deliver thoughtful end-of-life care, by using providers or settings not recognized by current attribution methods, are not rewarded for this high-quality care. On the flip side, organizations that fail to coordinate end-of-life care or that deliver resource-intensive care not aligned with patient preferences face no penalty. Additionally, higher hospice spending registered in unattributable decedents signals that involvement of hospice and palliative care in new payment initiatives may improve quality of care and potentially reduce costs

 Alternative attribution methods have different strengths and weaknesses. Retrospective attribution, currently used for most MSSPs and done at the end of the performance year, reflects actual care delivery. It removes patients who no longer receive care from an ACO and captures new patients, such as those aging into Medicare during the year (

 However, with retrospective attribution, organizations cannot plan care for patients assigned to them only after care delivery. In contrast, prospective attribution removes this uncertainty, which might facilitate active engagement with patients and can improve management of patient population health. However, prospective attribution is a forecast based on past healthcare use, and it works best when care-seeking patterns are stable and therefore predictable over time. This attribution, although preferred by ACOs,

 may have important cost implications for ACOs if patient care patterns drastically change from one year to another because patients may see any Medicare provider. This type of attribution would also fail to capture patients who initiated care during the year (eAppendix Table 5).

Most of the gaps in attribution rules that we identified may be addressed with a hybrid attribution method. First, for most populations, the current assignment based on E&M visits is appropriate. Second, for those with limited healthcare use and for the healthy, assignment would be based on attestation over a longer time period, therefore “rewarding” providers for early prevention and early detection care patterns. For those with serious illnesses, assignment could be broadened to include hospice and palliative care to encourage better end-of-life care (eAppendix Table 5). Before patients with serious illness reach the end of life, however, a first priority is to engage them in care. Providers could tailor care management and coordination to the needs of patients with chronic conditions who have accessed very little healthcare

 in order to improve quality and health outcomes.

Gender disparities were noted between attribution categories, with disproportionately larger proportions of younger and male populations among the unattributable beneficiaries. This may signal gender differences in healthcare-seeking behavior.

 The slightly higher proportion of those with dual eligibility for Medicaid attributable to a provider group (90%) may reflect greater coverage for out-of-pocket spending in this group, lowering a barrier to access. New initiatives, such as the CMS Financial Alignment Initiative, which integrates care across primary, acute, behavioral health, and long-term care for dual-eligible beneficiaries, offer another avenue to engage these patients.

This cross-sectional exploratory study in the first year of the ACO payment model has limitations. First, methods other than the MSSP, which was used in this study, may result in slightly different distributions of beneficiaries across categories. However, MSSP is the largest Medicare ACO program. Second, a longitudinal study with additional years of data could examine trends in attribution of Medicare beneficiaries as provider groups gain knowledge and develop better strategies to manage attribution. Third, claims-based measures of patient illness cannot capture the illness burden for those without utilization. Thus, our results may reflect care-seeking behavior, as we cannot attribute or observe clinical comorbidities for those with no or little healthcare utilization and for new enrollees in the preceding year.

The promise of ACO contracts to lower spending and enhance quality of care, and, ultimately, to improve population health, can be enhanced by improving the attribution of individuals to providers. By rewarding organizations for keeping patients healthy and out of resource-intensive care settings and by appropriately capturing and rewarding compassionate care, such as hospice services, for patients approaching the end of life, Medicare’s continued development of advanced payment models has the potential to move us closer to the goal of improving population health.

The authors thank Kristen Bronner for her help in map designing.

The Dartmouth Institute for Health Policy & Clinical Practice, Geisel School of Medicine at Dartmouth (MHO, EM, C-HC, SRR, JPWB, VAL, CHC), Lebanon, NH; The National Bureau of Economic Research (EM), Cambridge, MA; Department of Internal Medicine, University of Michigan Medical School (JPWB), Ann Arbor, MI; Institute for Health Policy and Innovation, University of Michigan (JPWB), Ann Arbor, MI.

 Supported by the National Institute on Aging (R03AG049360, R33AG044251, P01-AG019783, and 1K01AG049914), a grant from the Commonwealth Fund (20160616), and a grant from the Agency for Healthcare Research and Quality (5U19HS024075-02). The content of the article is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute on Aging, the Commonwealth Fund, or the Agency for Healthcare Research and Quality.

This article compares clinical and utilization profiles of Medicare patients who are attributed to provider groups with those of patients unattributed to any provider group in accountable care organization models.

Julie P.W. Bynum, MD, MPH

Articles

Access to Chiropractic Care and the Cost of Spine Conditions Among Older Adults

Forgotten Patients: ACO Attribution Omits Those With Low Service Use and the Dying