Table 3. Logistic Regression Associations Between Covariates of Interest and 90-Day Hospital Admission, by Model Type (N = 4685).
| Covariates | Odds ratio (95% CI)a | ||
|---|---|---|---|
| Model with EHR-based covariates | Model with EHR- and survey-based covariates, forward selection | ||
| All | Forward selection | ||
| EHR-based measure | |||
| Demographic characteristic | |||
| Age, y | |||
| <60 | 1 [Reference] | ||
| 60-80 | 1.05 (0.70-1.59) | ||
| >80 | 0.90 (0.54-1.51) | ||
| Sex | |||
| Male | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| Female | 0.50 (0.24-1.04) | 0.52 (0.25-1.07) | 0.48 (0.23-1.00) |
| Race/ethnicity | |||
| Non-Hispanic White | 1 [Reference] | ||
| Non-Hispanic Black | 1.10 (0.78-1.55) | ||
| Hispanic | 1.63 (0.95-2.77) | ||
| Other | 1.16 (0.52-2.61) | ||
| VA copay status | |||
| Exempt | 1 [Reference] | 1 [Reference] | |
| Nonexempt | 0.74 (0.49-1.12) | 0.71 (0.47-1.08) | |
| Residence | |||
| Nonrural | 1 [Reference] | ||
| Rural | 0.88 (0.67-1.16) | ||
| Chronic health condition | |||
| Gagne comorbidity scoreb | |||
| For 1-unit increase (when initial score in −2 to 0 range) | 0.82 (0.48-1.39) | 0.84 (0.50-1.43) | 0.84 (0.49-1.43) |
| For 1-unit increase (when initial score in 1 to 13 range) | 1.09 (0.63-1.89) | 1.08 (0.63-1.88) | 1.08 (0.62-1.87) |
| Non-Gagne health condition (presence vs absence [reference]) | |||
| Anxiety | 0.95 (0.65-1.38) | ||
| Bipolar disorder | 1.80 (0.99-3.28) | 1.75 (0.98-3.14) | 1.71 (0.94-3.08) |
| Dyslipidemia | 0.86 (0.66-1.11) | ||
| Posttraumatic stress disorder | 0.68 (0.47-0.97) | 0.70 (0.49-0.99) | 0.74 (0.52-1.06) |
| Schizophrenia | 1.99 (1.02-3.86) | 2.09 (1.08-4.03) | 2.00 (1.03-3.90) |
| Traumatic brain injury | 2.03 (0.41-10.01) | ||
| Spinal cord trauma | 2.04 (0.69-6.05) | ||
| Other clinical characteristic | |||
| BMI | |||
| <25 | 1 [Reference] | ||
| 25 to <30 | 1.00 (0.68-1.46) | ||
| 30 to <35 | 0.96 (0.65-1.44) | ||
| ≥35 | 1.09 (0.72-1.64) | ||
| Frailty (JEN Frailty Index, for 1-unit increase) | 1.23 (1.14-1.33) | 1.24 (1.15-1.33) | 1.23 (1.14-1.32) |
| Alcohol use (AUDIT-C score) | |||
| None (0) | 1 [Reference] | 1 [Reference] | |
| Low (1-2 [female], 1-3 [male]) | 0.94 (0.70-1.27) | 0.95 (0.71-1.28) | |
| Moderate (3-7 [female], 4-7 [male]) | 1.19 (0.75-1.91) | 1.22 (0.76-1.94) | |
| High (>7) | 2.14 (1.11-4.11) | 2.23 (1.18-4.21) | |
| Survey-based measure | |||
| Workload domain | |||
| Global health status | |||
| Poor/fair | 1 [Reference] | ||
| Good/very good/excellent | 1.36 (1.02-1.81) | ||
| Smoking status | |||
| Never | 1 [Reference] | ||
| Past | 1.11 (0.79-1.55) | ||
| Current, less than daily | 0.89 (0.45-1.75) | ||
| Current daily | 1.63 (1.10-2.40) | ||
| Medication | |||
| Secure | 1 [Reference] | ||
| Insecure | 0.63 (0.40-1.00) | ||
| Depression symptoms, PHQ-2 | |||
| 0-2 | 1 [Reference] | ||
| ≥3 | 1.33 (0.97-1.82) | ||
| Capacity/resilience domain | |||
| Lack confidence in completing medical forms | 1.32 (0.91-1.91) | ||
| Low resilience, score | |||
| >3 | 1 [Reference] | ||
| ≤3 | 1.16 (1.00-1.34) | ||
| Married | |||
| No | 1 [Reference] | ||
| Yes | 0.69 (0.52-0.90) | ||
| Preference/expectations | |||
| Powerful others health-related locus of control, score | |||
| ≤18 | 1 [Reference] | ||
| >18 to 22 | 1.02 (0.77-1.36) | ||
| >22 to 30 | 0.71 (0.49-1.04) | ||
| Mean AIC (SD) | 1980.2 (1.7) | 1963.9 (1.6) | 1947.7 (2.5) |
Abbreviations: AIC, Akaike Information Criterion; AUDIT-C, Alcohol Use Disorders Identification Test–Consumption; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); EHR, electronic health record; PHQ-2, 2-item Patient Health Questionnaire; VA, Veterans Affairs.
Calculated using PROC MIANALYZE within SAS software, version 9.4, to combine estimates from 10 imputed data sets (SAS PROC MI) per model type. Although the primary analyses were not based on statistical significance, the 95% CIs are provided to inform uncertainty around the magnitude of the estimates from the covariates selected as improving estimates of the outcomes based on AIC.
Gagne score was included in the model as a piecewise continuous variable composed of 2 pieces, one for Gagne scores with initial value of −2 to 0 and the other for Gagne scores with initial values of 1 to 13. The reported Gagne odds ratios and 95% CIs are for a 1-unit increase in the Gagne score.