Table 2.

Logit Regression Coefficients for Glycemic Control (HbA1c < 8%) by Depression and Patient Location, Adjusted for Covariates at Patient, Practice, and Neighborhood Levels

Without Depression-Rural Interactions in ModelWith Depression-Rural Interactions in Model
CoefficientP ValueCoefficientP Value
Residential location (ref. group Urban)
 Large rural town0.018.0310.012.289
 Small rural town−0.002.734−0.015.139
Interactions terms (ref. group Depression* Urban area)
 (Depression)* (Large rural town)0.019.135
 (Depression)* (Small rural town)0.051<.001
Patient-level control variables
  Annual trend−0.036<.001−0.036<.001
 Patient age, years (ref. group aged 65 to 75)
  18 to 44−0.840<.001−0.840<.001
  45 to 54−0.609<.001−0.609<.001
  55 to 64−0.273<.001−0.273<.001
 Ischemic vascular disease*−0.210<.001−0.210<.001
 Type 1 diabetes−0.844<.001−0.844<.001
 Insurance (ref. group Commercial)
  Dual Medicare/Medicaid−0.181<.001−0.181<.001
  No insurance−0.401<.001−0.401<.001
Practice-level control variables
 Certified as a patient-centered medical home0.049<.0010.049<.001
 Practice ownership (ref. group Large medical group)
  Single-site medical group−0.104.006−0.104.006
  Small medical group−0.087<.001−0.087<.001
Neighborhood-level (Patient ZIP Code) ACS control variables§
 Percent of population White, non-Hispanic0.0027<.0010.0027<.001
 Percent of adults with high school degree only−0.0039<.001−0.0039<.001
 Percent of adults with 4-year college degree0.0008.3060.0008.319
 Percent of households under the federal poverty level−0.0031<.001−0.0030<.001
  • ACS, American Community Survey.

  • * Depression and ischemic vascular disease were reported by the primary care practice to MN Community Measurement (MNCM). MNCM suggested the use of the Major Depression or Dysthymia (DEP-01) Value Set and Ischemic Vascular Disease Value Set, but stated that “Any documentation of a new or existing diagnosis of depression [IV.D] during the measurement period [for IV.D: or year prior] is accepted.”

  • Rurality was determined by mapping patient ZIP code to Rural-Urban Commuting Areas, and summarizing to the WWAMI Rural Health Research Center’s Categorization B. (

  • Small medical groups were defined as 2 to 11 primary care sites; large medical groups were defined as 12 or more primary care sites.

  • § Coefficients represent the impact of a one percentage-point increase in the ACS-based statistic (e.g., 50% to 51%). Educational distribution was represented by percentage of adults with a high school degree but no 4-year college degree, and percentage with a 4-year college degree. We omitted the percentage without a high school degree to avoid collinearity in the estimation. A U-shaped impact of educational distribution in the patient’s neighborhood was evident, with neighborhoods more heavily weighted toward the high-school-only category showing poorer glycemic control.