Appendix Table 3:

Association between Practice Location Rurality and Low Value Care Spending of Attributed Medicare Patients Across Specialty

CharacteristicPanel A: OLS EstimatesPanel B: Logit Odds Ratio Estimates
Physician Characteristics
    Non-US Medical Graduate−0.2160.41580.9850.1138
Patient Size
    50 to 149−2.7790.46750.4160.0560
    150 to 299−1.8110.41710.540.0701
    Internal Medicine
    Family Medicine−1.1820.34990.680.0687
    Other Specialty−0.8290.77420.840.2193
Graduation Year
    1980 to 1989−0.1660.38341.0900.1236
    1990 to 2000−0.7410.40860.9720.1122
Practice Region
    RuralXInternal Medicine
    RuralXFamily Medicine0.7640.58221.4210.4270
    RuralXOther Specialty−0.8311.23000.9220.5493
Patient Characteristics
    Age (years)
        65 to 69
        70 to 74−0.0370.03960.9950.0105
        75 to 79−0.0640.03450.9890.0102
        80 to 84−0.1310.03380.9830.0102
    ElixHauser Comorbidity Index
        1 to 20.050*0.02361.0150.0085
        3 to 50.2140.03071.0580.0074
    Practice Location Characteristics
        <12 years schooling−0.0960.03360.978*0.0105
        < 200% FPL−0.0770.01520.9790.0050
        Number of Observations6,905
  • OR, odds ratio; FLP, federal poverty line; OLS, ordinary least squares. Note: The dependent variable for Panel A was per-patient low value care (LVC) Medicare spending in dollar amounts, while the ORs in Panel B were calculated from logit regression results, where the dependent variable was equal to 1 if the primary care physician's (PCP) per-patient LVC Medicare spending was in the top quintile. RuralXSpecialty indicated a set of interaction terms between the rurality of PCP's practice location and PCP's specialty type. Practice was considered to be in a rural area if it was located in any of the six non-metropolitan county categories in the Rural Urban Continuum Code that divided all US counties into three metropolitan and six nonmetropolitan county categories. PCPs were categorized into three specialty types: those that practice internal medicine, family medicine, and other PCP specialties. Both models were estimated using sample weights that reflect the oversampling of physicians in smaller states.

  • * P < .05.

  • P < .01.

  • P < .001.