Annual trends in administrative data outcomes (in those admitted for pneumonia)
| Initial change* per year (95% CI) | Year of rate-change† | Subsequent change* per year (95% CI) | Direction of change‡ | Estimate in 2014§ | |
|---|---|---|---|---|---|
| 30-day mortality | 1.8 (0.4 to 3.1) | 2007 | −3.1 (−4.5 to −1.7) | ↑ ↓ | 0.22 |
| 30-day readmission | |||||
| All-cause | 10.0 (4.2 to 16.1) | 2002 | 1.5 (0.4 to 2.6) | ↑ ↑ | 0.16 |
| Pneumonia | 3.4 (1.7 to 5.2) | − | ↑ | 0.04 | |
| Median length of stay (days) | 0.00 (−0.06 to 0.06) | 2006 | −0.25 (−0.30 to −0.20) | ↔ ↓ | 4 |
Length of stay modelled using piecewise median regression, all other trends using piecewise linear Poisson regression.
*Percentage change in rate trend for mortality, readmission; absolute change in median length of stay.
†Allowing change in trend over time if significant improvement over linear trend alone (p<0.05), with year of the change point chosen by profile likelihood8 (see Methods).
‡Single arrows represent direction of change over the whole time period; paired arrows represent direction and relative size of change in the periods before and after the year when data suggested trend in rates changed (as in †).
§Proportion of admissions/median value, as estimated by the best-fitting model as presented in the table.