Follow the Money: Factors Associated with the Cost of Treating High-Cost Medicare Beneficiaries
Originally published by the Center for Studying Health System Change
Published: February 2011
Updated: April 6, 2026
Journal article by James D. Reschovsky, Jack Hadley, Cynthia B. Saiontz-Martinez, and Ellyn R. Boukus. Published in Health Services Research, online ahead of print, February 2011.
A small share of Medicare beneficiaries accounts for a strikingly large portion of program spending. This concentration of costs has made the highest-cost patients a central focus for policymakers searching for ways to slow the growth of federal health expenditures. Yet the factors that actually drive spending among these patients remain surprisingly poorly understood. Are the costs primarily a reflection of how sick these individuals are, or do the physicians they see, the practices where they receive care, and the health care markets where they live play equally significant roles?
A study published in Health Services Research in February 2011 set out to untangle these questions. Researchers James D. Reschovsky, Jack Hadley, Cynthia B. Saiontz-Martinez, and Ellyn R. Boukus analyzed data from approximately 1.6 million elderly Medicare fee-for-service beneficiaries, drawing on 12 months of claims from the 2005-2006 period. The claims data were linked to physician survey responses and detailed information about the health care markets in which patients lived, creating an unusually rich dataset for examining what pushes costs higher among the sickest and most expensive Medicare enrollees.
Health Status as the Dominant Cost Driver
The study's most consistent finding was that health status overwhelmingly predicted which beneficiaries would end up in the top quartile of Medicare spending. After applying risk adjustment methods to account for the number and severity of diagnoses, the researchers found that clinical complexity was far and away the strongest determinant of high costs. Patients with multiple chronic conditions, serious acute episodes, or combinations of both were the ones generating the largest share of expenditures.
This may sound obvious on its surface, but the implications are substantial. If the main reason certain Medicare beneficiaries are expensive is that they are genuinely very ill, then strategies aimed at reforming how physicians practice or restructuring health care markets may have only modest effects on the total spending attributable to these patients. The disease burden itself is the engine behind the costs, and no amount of payment redesign or market restructuring can make a patient with advanced heart failure and diabetes inexpensive to treat.
Physician and Practice Factors: Mostly Insignificant
One of the more striking results involved the limited role that physician and practice characteristics played in explaining high costs. The researchers examined a range of factors: whether the patient's usual physician was a primary care doctor or a specialist, the size and organizational structure of the practice, the adoption of health information technology, and the degree to which physicians reported having enough time during office visits. Most of these variables turned out to be statistically insignificant or only weakly associated with cost differences among high-cost beneficiaries.
There were a few exceptions worth noting. Beneficiaries whose usual source of care was a medical specialist rather than a generalist tended to have higher costs within the top quartile. Physicians who reported that they did not have adequate time for office visits were also associated with somewhat higher spending among their high-cost patients, a finding that raises questions about whether rushed encounters lead to less efficient diagnostic workups or poorer coordination of complex care. Practices operating on a for-profit basis showed a positive association with higher costs as well.
But the overarching message from the physician and practice analysis was one of limited explanatory power. The characteristics of where and how patients received their outpatient care simply did not explain much of the variation in spending once health status was accounted for.
Health Care Market Conditions
The study also investigated whether features of the local health care market influenced costs for high-spending beneficiaries. Variables such as the overall supply of physicians, the number of hospital beds per capita, and the degree of managed care penetration were tested for their relationship to expenditures.
Again, the results pointed toward limited influence. Most market-level factors showed no meaningful connection to the costs incurred by high-cost patients. Two exceptions emerged: a greater supply of medical specialists in a given market was associated with higher costs among top-quartile beneficiaries, and higher Medicare fee levels in a region were also linked to greater expenditures. These findings are consistent with longstanding observations in health economics — areas with more specialists and higher reimbursement rates tend to see more intensive utilization patterns — but their practical significance for policy was modest.
Care fragmentation also surfaced as a factor. Beneficiaries who received care from a larger number of different providers, without a clear coordinating physician, tended to accumulate higher costs. This finding resonated with a growing body of evidence suggesting that poorly coordinated care leads to duplicative testing, conflicting treatment plans, and avoidable hospitalizations — all of which drive up spending without necessarily improving outcomes.
Implications for Bending the Cost Curve
The findings carried important implications for the health reform debates that were prominent at the time of publication. Many of the policy approaches being discussed — accountable care organizations, patient-centered medical homes, bundled payment models, and efforts to shift patients from specialist-dominated care to primary care — were predicated on the assumption that how care was organized and delivered was a major driver of cost variation. The results of this study suggested those reforms, while potentially beneficial for other reasons, might have relatively small effects on spending for the patients who actually consume the most Medicare resources.
This does not mean that delivery system reforms are pointless. Even modest per-patient savings, when multiplied across the large population of high-cost beneficiaries, can translate into meaningful aggregate reductions. And reforms aimed at improving care quality and patient experience have value independent of their effect on costs. But the study cautioned against expecting sweeping cost reductions from broad-based organizational changes alone.
Instead, the researchers suggested that more targeted approaches might hold greater potential. Rather than applying uniform interventions across all high-cost patients, programs that identify specific subgroups — patients with particular combinations of chronic conditions, those experiencing transitions between care settings, or individuals with both medical and behavioral health needs — and tailor interventions to their circumstances could be more effective at reducing unnecessary spending while preserving or improving outcomes.
Study Design and Methodology
The research drew on a unique combination of data sources. Medicare claims provided detailed information about each beneficiary's diagnoses, procedures, and total costs over a 12-month period during 2005-2006. These claims were linked to responses from the Community Tracking Study Physician Survey, which collected data on physician characteristics, practice organization, and reported clinical behaviors. Additional information about local health care markets was incorporated from the Area Resource File and other administrative datasets.
The analytical approach used two-part models to examine both the probability of being a high-cost beneficiary and the level of costs conditional on being in the top quartile. Risk adjustment was performed using hierarchical condition categories (HCCs), a methodology commonly used by the Centers for Medicare and Medicaid Services to predict expected spending based on diagnosis patterns. The linkage to physician survey data allowed the researchers to examine provider-level variables that are typically unavailable in claims-only analyses, giving the study an unusual degree of depth in assessing the physician's contribution to cost variation.
The sample of 1.6 million beneficiaries was drawn from the fee-for-service Medicare population aged 65 and older. By focusing on elderly beneficiaries in traditional Medicare, the study avoided confounding effects from Medicare Advantage plan structure or the distinct clinical profiles of disabled beneficiaries under age 65. The large sample size gave the analysis substantial statistical power to detect even relatively small effects, which made the finding that most physician and market factors were insignificant all the more telling.
Limitations and Context
As with any observational study, there were important limitations. The cross-sectional design meant the researchers could identify associations but could not establish definitive causal relationships. It is possible, for example, that sicker patients systematically sought out specialists as their usual physician precisely because of their complex conditions, making it difficult to separate the effect of specialist care from the underlying severity of illness. Similarly, the physician survey data, while rich, captured self-reported practice characteristics that may not perfectly reflect actual clinical behavior.
The study was also based on data from a period that predated the Affordable Care Act and the subsequent wave of delivery system experiments, including accountable care organizations and value-based payment models. Whether the same patterns would hold in a substantially different policy environment remains an open question. Still, the fundamental finding — that health status dominates the cost picture for the most expensive Medicare patients — is likely robust to changes in payment structures, since the underlying clinical needs of these patients do not change simply because the reimbursement framework does.
The Broader Conversation About Medicare Spending
This research contributed to an evolving understanding of where Medicare dollars actually go and why. The concentration of spending among a relatively small number of high-cost beneficiaries has been documented repeatedly, but the question of what to do about it remains contentious. Some analysts have emphasized the role of waste and inefficiency, arguing that a substantial fraction of high-cost care is unnecessary. Others have pointed to the findings of studies like this one to argue that much of the spending reflects genuine medical need and that the savings achievable through efficiency improvements may be more limited than commonly assumed.
The truth almost certainly lies somewhere in between. There are undoubtedly high-cost patients receiving care that is poorly coordinated, duplicative, or inconsistent with their preferences — and targeted interventions for those individuals can generate real savings and better experiences. But there are also patients whose costs are high simply because they are seriously ill and receiving appropriate, evidence-based treatment. Conflating the two groups and applying blanket cost-reduction strategies risks harming the latter in pursuit of savings from the former.
The researchers ultimately argued for a more granular approach to high-cost Medicare patients — one that recognizes the heterogeneity within this population and develops interventions matched to specific clinical and social circumstances. Rather than broad policy levers applied uniformly, they envisioned a strategy built on careful identification of patient subgroups where targeted care management, better coordination, or alternative treatment pathways could make a meaningful difference without compromising the quality of care for those who genuinely need intensive services.