Adjusted ORs indicating the change in the odds of a patient death associated with below-target registered nurse (RN) staffing
| Outcome: patient death | |||
| OR | 95% CI | P value | |
| Binary indicator of RN staffing below target by: | |||
| 4 hours or more | 1.0424 | 0.9257 to 1.1739 | 0.4927 |
| 8 hours or more | 1.0587 | 0.9371 to 1.1961 | 0.3596 |
| 12 hours or more | 1.0933 | 0.9508 to 1.2572 | 0.2108 |
| 16 hours or more | 1.2282 | 1.0034 to 1.5035 | 0.0463 |
| 20 hours or more | 1.2626 | 0.9841 to 1.6199 | 0.0667 |
| 24 hours or more | 1.3701 | 1.0146 to 1.8501 | 0.0399 |
Results listed are from six separate conditional logistic regression models, each with ward fixed effects. Unit of analysis is the hospital ward-day (n=19 287). 95% CI, with SEs clustered at the hospital ward level. All models also control for the mean patient age; the patient age squared; the sex and ethnicity mix of patients; the mean Elixhauser Comorbidity Index; pairwise-interaction terms between mean age, sex, ethnicity and Elixhauser Comorbidity Index; the mean hospital length of stay of patients treated in the ward that day; the number of patients treated in the ward that day; a dummy for each day of the week; a dummy for month of the year and a dummy indicating whether the day was a bank holiday. Target RN staffing is defined as the total number of rostered hours for RNs (those in band 5 or above), which is equal to the sum of filled and unfilled hours.