Table 4.
Odds ratios for the variable recognition (primary outcome), prescribing antidepressants, referral to specialist mental health services and number of consultations after recognition (secondary outcomes), using mixed effects (logistic or Poisson) regression models on multiple imputed data
| Outcome variable | Odds ratio | 95% confidence interval | Test statistics and p value | Estimated proportion in control group | Estimated proportion in intervention group |
|---|---|---|---|---|---|
| Recognition (primary outcome) | 1.60* | (1.01, 2.53) | z = 1.99, p = 0.047 | 0.31 | 0.42 |
| Number of consultations | 1.78* | (1.14, 2.78) | z = 2.52, p = 0.012 | 0.57a | 1.01a |
| Prescribing antidepressants | 1.07 | (0.52, 2.19) | z = 0.19, p = 0.849 | 0.12 | 0.13 |
| Referral to specialist mental health services | 1.62 | (0.72, 3.64) | z = 1.16, p = 0.247 | 0.05 | 0.08 |
All mixed effects models were estimated on multiple imputed data (20 datasets) having a multilevel data structure of 444 patients (lowest level) of 40 general practitioners (second level) of 23 practices (highest level).
*p value <0.05.
aFor ‘number of consultations’, the incidence rate ratio (IRR) and incidence rates from the Poisson regression are displayed instead of an odds ratio (OR) and probabilities.