Table 5.
Piecewise logistic regression analysis of the comprehensive prevention programme effects* on violence
| Full model† | Simple model‡ | |||
| OR (95% CI) | P value | OR (95% CI) | P value | |
| Trend prior to intervention (per month) | 1.09 (0.81 to 1.49) | 0.5848 | – | – |
| Immediate change in level | ||||
| A | 0.31 (0.03 to 3.20) | 0.3236 | – | – |
| BC added to A | 2.19 (0.70 to 6.82) | 0.1773 | – | – |
| D added to ABC | 1.05 (0.28 to 3.88) | 0.9406 | – | – |
| E added to ABCD | 5.73 (2.08 to 15.77) | 0.0007 | – | – |
| Change in trend (per month) | ||||
| A | 0.95 (0.55 to 1.65) | 0.8657 | 0.87 (0.82 to 0.92) | <0.0001 |
| BC added to A | 0.61 (0.33 to 1.13) | 0.1188 | – | – |
| D added to ABC | 1.85 (0.98 to 3.48) | 0.0572 | 1.45 (1.14 to 1.84) | 0.0022 |
| E added to ABCD | 0.35 (0.17 to 0.70) | 0.0031 | 0.65 (0.45 to 0.93) | 0.0194 |
Components: A corresponds to computerised triage algorithm, BC corresponds to signage and messages broadcast on TV in the waiting rooms, D corresponds to mediator and E corresponds to video surveillance.
*Logistic generalised estimating equation model adjusted for waiting time>2 hours. OR<1 represent a decline and inversely, OR>1 represent an increase in monthly likelihood of violence occurrence during admission at OED per month.
†Full model included time effect and immediate changes after each component’s implementation and changes in slopes.
‡Parsimonious model after backward selection.
OED, ophthalmology emergency department.