Table 2.
Determinants of Total Inpatient Days per Year, 1991–1995*
| Full Model | No-Predictors Model | ||
|---|---|---|---|
| Independent Variables | Reference Category | % Total Days per Year Relative to Reference (95% CI)† | |
| HMO Effect | |||
| Proportion of year in group/staff HMO | Continuous FFS | 82 (79, 86) | 80 (77, 83) |
| Proportion of year in IPA HMO | 89 (82, 97) | 88 (82, 94) | |
| Enrollment History Groups | Continuous FFS | ||
| Group/Staff HMO | 73 (71, 76) | 63 (61, 66) | |
| IPA HMO | 64 (60, 67) | 52 (49, 55) | |
| Switch to HMO and Disenroll to FFS | 125 (118, 133) | 128 (117, 140) | |
| Year | 1991 | ||
| 1992 | 97 (95, 98) | 109 (107, 111) | |
| 1993 | 95 (91, 98) | 137 (131, 144) | |
| 1994 | 90 (86, 95) | 153 (147, 159) | |
| 1995 | 80 (76, 84) | 138 (130, 146) | |
| Male | Female | 121 (118, 124) | — |
| African American | Non-African American | 120 (112, 129) | — |
| Medicaid-eligible in 1991 | Not eligible | 151 (145, 157) | — |
| Disabled >64 years | Other>64years | 168 (163, 172) | — |
| Age, years | Age 65–69 years | ||
| 70–74 | 119 (116, 122) | — | |
| 75–79 | 141 (138, 144) | — | |
| 80–84 | 162 (158, 167) | — | |
| 85+ | 161 (150, 173) | — | |
| Death from 1993 through 1996 | Alive | ||
| Death in first half of current year | 429 (388, 475) | — | |
| Death in second half of current year | 743 (698, 791) | — | |
| Death in first half subsequent year | 447 (420, 476) | — | |
| Death in second half subsequent year | 265 (251, 280) | — | |
| Goodness of fit (Efron's R2) | 6.7 | 0.8 | |
| Number of Observations | 1,227,105 | 1,227,105 | |
*
HMO indicates health maintenance organization and FFS indicates fee-for-service.
†
Results are based on a two-part model where the first part is a logistic regression of whether the beneficiary had one or more days in the hospital in a year, and the second part is an ordinary least squares regression of the natural log of total hospital days per year given at least one day. Log days are retransformed to days using the smearing estimate. CI indicates confidence intervals, which are based on standard errors adjusted for clustering by HMO. Confidence intervals assume log normality.