Ultimately questions about why there is variation in health spending come down to differences in individual physicians practice patterns, or in the health of treated populations or to regulatory factors. Teasing those out proves difficult. A new study in the Annals of Family Medicine examines the effect of spending and utilization patterns at the institutions where physicians are trained, largely teaching hospitals. (Annals Article) Rather than using hospital referral regions, in this study the researchers examined effects at the hospital service area level, of which there are a whopping 3,436. Over 80% of these have just one hospital. In addition to looking at utilization and spending patterns, quality was evaluated, but in a very minor way by looking at diabetes care. The core data related to care for Medicare beneficiaries delivered by primary care physicians in 2011. Only physicians who saw at least 40 Medicare beneficiaries were included. Spending for these physicians was compared with spending in the HSA where they received their training. Spending patterns were assigned to low, average or high-cost groups, based on average cost per beneficiary.
Patient data was adjusted for a variety of demographic and health characteristics that might account for differences in spending. After application of inclusion criteria, over 3000 primary care physicians treating over 500,000 patients were included in the analysis. The per beneficiary annual spending categories were not excessively wide, $6,102 to $8,683 for low-cost; $8,691 to $9,879 for average; and $9,880 to $16,542 for high-cost. Physicians who practiced in low-cost HSAs were more likely to train in low-cost HSAs and those in high-cost HSAs more likely to have trained in high-cost HSAs. Physicians who trained in high-cost HSAs had more per beneficiary spending regardless of the category of the HSA they practice in, at an annual average of $1,644. General internists were more likely than family practice doctors to have trained in a high-cost HSA and therefore appeared to have higher spending patterns. Physicians who trained at institutions that turned out fewer total doctors, more rural ones and more primary care doctors were more likely to have a low-cost spending pattern. Perhaps the most interesting finding is that spending differences based on training institution appear to evaporate after 15 years in practice. No explanation was given for this phenomenon, but perhaps local practice patterns overwhelm training ones at some point. There were no differences in diabetes care quality associated with training facility spending category. Another fascinating piece of research on spending variation.