Table 4.
GEE logistic regression models for overall quality of care over time
| Variable | OR | 95% CI | p value | Model QICa |
|---|---|---|---|---|
| Model 1: biannual groupings, with months 1–6 as reference group | n/a | |||
| Months 7–12 | 2.33 | (1.33, 4.11) | p = 0.003 | |
| Months 13–18 | 1.84 | (1.05, 3.20) | p = 0.032 | |
| Months 19–24 | 3.05 | (1.76, 5.29) | p < 0.001 | |
| Model 2: time as a continuous variable | 614.62 | |||
| Months since iCCM services initiation | 1.06 | (1.02, 1.09) | p < 0.001 | |
| Model 3: time as a continuous variable, with a spline knot at month 6 | 612.16 | |||
| Months since iCCM services initiation—months 1–6 | 1.24 | (1.08, 1.43) | p = 0.003 | |
| Months since iCCM services initiation—months 7–24 | 1.01 | (0.98, 1.04) | p = 0.47 | |
aQuasi-likelihood under the independence model criterion (QIC). This is a modification of the Akaike information criterion (AIC) so that it can be applied to GEE regression models to assess goodness of fit of different models. A lower QIC term reflects a better-fitting regression model. It is not applicable when using factor variables, so it is not calculated for Model 1, which uses a categorical time variable