Estimated effect of an increase in the provision of one-to-one nursing on the mortality rate
| One-to-one nursing measured as a proportion of intensive care days* | |
| OLS | |
| Estimated effect of a 10 percentage point increase in one-to-one nursing on the monthly mortality rate (deaths per 100 infants receiving intensive care per month) | 0.6 |
| 95% CI | 0.1 to 1.2 |
| p Value | 0.05 |
| Number of unique neonatal units | 43 |
| Total observations | 2228 |
| Instrumental variables estimator | |
| Estimated effect of a 10 percentage point increase in one-to-one nursing on the monthly mortality rate (deaths per 100 infants receiving intensive care per month) | −0.6 |
| 95% CI | −1.2 to −0.0 |
| p Value | 0.04 |
| Number of unique neonatal units | 43 |
| Total number of observations | 2140 |
| One-to-one nursing measured as a proportion of admissions† | |
| OLS | |
| Estimated effect of a 10 percentage point increase in one-to-one nursing on the monthly mortality rate (deaths per 100 infants receiving intensive care per month) | 0.3 |
| 95% CI | 0.1 to 0.6 |
| p Value | 0.002 |
| Number of unique neonatal units | 43 |
| Total observations | 2228 |
| Instrumental variables estimator | |
| Estimated effect of a 10 percentage point increase in one-to-one nursing on the monthly mortality rate (deaths per 100 infants receiving intensive care per month) | −0.4 |
| 95% CI | −0.7 to −0.0 |
| p Value | 0.03 |
| Number of unique neonatal units | 43 |
| Total number of observations | 2140 |
Results are presented from two different estimators: (1) ordinary least squares, which does not control for the correlation between the one-to-one nursing and unobserved case-mix differences; (2) instrumental variables estimator, which does control for the correlation between the one-to-one nursing and unobserved case-mix differences using historical nursing levels. Results are interpreted as change in the number of deaths per 100 infants receiving neonatal intensive care per month resulting from a 10 percentage point increase in one-to-one nursing.
Regressions control for the mean values of gestational age, birth weight z-score, antenatal steroid receipt and gender, as well as year, calendar month and neonatal unit effects (see online supplementary appendix B). The number of observations are fewer for the instrumental variables estimation as lagged variables are used as instruments. Sensitivity analyses are presented in online supplementary appendix C.
*The proportion of intensive care days on which a one-to-one nurse to patient ratio was provided.
†The proportion of intensive care admissions who received at least one day of care on which one-to-one nursing was provided.