Geographic Variation in Medicare and the Military Healthcare System

Document Type


Publication Date



Research Objective: Geographic variation in per capita healthcare spending is well-documented using Medicare claims data but it is unclear whether spending patterns in Medicare are representative of the rest of the health care system. Differences in price and Medicare reimbursement rates account for only a small portion of the variation observed and variation in practice patterns likely explains this phenomena. We explored geographic variation in health care spending and health care utilization within the Military Healthcare System (MHS) compared with Medicare spending and healthcare utilization across Hospital Referral Regions (HRRs).

Study Design: A retrospective cohort approach was used. Data on age, sex and race adjusted Medicare expenditures by HRR were obtained from Dartmouth Atlas website. MHS data were obtained from the MHS Data Repository. Patients in the MHS were assigned to one of 306 HRRs using zip code crosswalk obtained from Dartmouth Atlas. Utilization measures of inpatient days, hip and back surgery were also analyzed. Measures of variation included coefficient of variation (CoV) and interquartile range. Depending on the measure, HRRs with less than 5 surgical events over the study period were excluded from our analysis.

Population Studied: Medicare beneficiaries included those above 65 years from 2007 to 2010. MHS Beneficiaries included military personnel, retirees and dependents above 18 years living within the U.S. and enrolled in TRICARE Prime, an HMO-like option in which all enrollees are assigned a Primary Care Manager. For MHS, we included data from 2007 and 2010 (the only 2 years available at the time of this study). The average number of enrollees per year was 3.4 million and 10.1 million within the MHS and Medicare respectively.

Principal Findings: The CoV for spending was higher in the MHS compared with Medicare (0.27, 0.15, respectively). Interquartile range was also higher in MHS compared to Medicare (1.44 and 1.22, respectively). The top 5 Medicare spending markets differed from top MHS spenders although the bottom 5 Markets exhibited some similarity as they were mostly located in the upper Midwest. The CoV for inpatient days was lower in Medicare compared to MHS (0.19, 1.29). Hip and back surgery rates were variable within Medicare and MHS with COV greater than 0.9 in both systems.

Conclusions: Despite being a managed care system, geographic variation in spending exists within the MHS to a higher degree than in Medicare. This may be partially explained by demographic differences in patient population in the MHS in addition to other unaccounted factors. Although variation in inpatient days was similar, other measures such as rates of back and hip surgery were highly variable within both systems.

Implications for Policy or Practice: Assessing variation in healthcare spending and utilization within the MHS compared with Medicare provides insight into potential sources of variation. Higher variation in spending within MHS may reflect its diverse patient population or a lessened delivery of integrated care given similar reimbursement rates to Medicare. This may inform policy solutions aimed at reducing unwarranted variation in health care.


Academy Health Research Annual Conference (ARM)


Minneapolis, MN