The contingency of medicare physician spending on population densities and sizes

The exploration of the spatial association between Medicare physician spending and population densities and sizes could possibly facilitate the investigation of the causal mechanisms beneath the variation in medical care. We acquired the U.S. Medicare physician expenditures and regional demographic and geographic data in 2006 from the Dartmouth Atlas of Health Care. Six geographic units—states, counties, Hospital Referral Regions, Hospital Service Areas, Metropolitan Statistical Areas, and state non-Metropolitan Statistical Areas—were used as units of study. Pearson correlation analysis, multivariable regression, and partial correlation analysis were employed. Among six geographic units, Pearson correlation coefficients between Medicare physician expenditures and logarithmic population densities ranged from 0.42 to 0.63 (p < 0.05 for all), and between the expenditures and logarithmic population sizes from 0.31 to 0.65 (p < 0.05 for all). When population health, differential demand, market structure, and data reporting bias were controlled, population densities and sizes were positively associated with Medicare physician expenditures in most models. Population densities and sizes could explain considerable amounts of regional variation in Medicare physician spending. We concluded that Medicare physician spending was contingent on population densities and sizes. Because population densities and sizes are produced by more fundamental qualities such as natural environments and resources and thus are not easily manipulated, they are suggestive in health policy studies. Further research might investigate population distribution associated properties such as geographic distribution of health care resources, spatial dynamics of medical technology distribution, and cultural and psychological factors.