Patient Preferences and Spending Variation

On various analytic bases, there appear to be geographical variations in health spending, both for Medicare, and to a lesser extent, for the commercial population.  The cause of this variation is generally assumed to be multi-factorial, including doctor practice patterns, patient characteristics such as burden of illness or demographics, and patient preferences.  The last mentioned factor is particularly hard to study but research published in Health Affairs takes a crack at it.   (HA Article)   The authors attempted to take data on Medicare spending, patient characteristics, provider characteristics and survey data from a 2005 Dartmouth survey of 4000 Medicare beneficiaries on their health status and preferences for care; and use all of this to attempt to estimate the effect of patient preferences on spending.  Patient preferences, however, were inferred at a hospital referral region level, and they were based on six survey questions, and this data is almost ten years old.  Divided by quintiles, the number of hospital beds is positively correlated with higher spending while the number of physicians per 100,000 people is negatively correlated.  Not clear in either case what the cause and effect relationship, if any, may be.  A higher mortality rate in a region was also correlated with more spending.  And five of the six care preference questions had a significant association with higher spending.  For example, if a region had more people who expressed a preference to see a doctor right away for a health problem, it tended to have higher spending.  Overall, the highest spending quintile averaged $9011 per beneficiary per year while the lowest averaged $6177.  About 23% appeared to be accounted for by provider supply factors, differences in patients’ health and income for 12.4% and patient preference accounted for 4.6%.  Note that about 60% is unexplained by the set of factors considered in the study.  Preferences appeared to have a bigger effect on end-of-life care spending than on overall spending.  While care preferences appear to have some effect, it is relatively minor.