Cumulative risk adjusted mortality chart for detecting changes in death rate: observational study of heart surgery

Abstract

Objective: To detect changes in mortality after surgery, with allowance being made for variations in case mix.

Design: Observational study of postoperative mortality from January 1992 to August 1995.

Setting: Regional cardiothoracic unit.

Subjects: 3983 patients aged 16 and over who had open heart operations.

Main outcome measures: Preoperative risk factors and postoperative mortality in hospital within 30 days were recorded for all surgical heart operations. Mortality was adjusted for case mix using a preoperative estimate of risk based on additive Parsonnet factors. The number of operations required for statistical power to detect a doubling of mortality was examined, and control limits at a nominal significance level of P=0.01 for detection of an adverse trend were determined.

Results: Total mortality of 7.0% was 26% below the Parsonnet predictor (P<0.0001). There was a highly significant variation in annual case mix (Parsonnet scores 8.7-10.6, P<0.0001). There was no significant variation in mortality after adjustment for case mix (odds ratio 1-1.5, P=0.18) with monitoring by calendar year. With continuous monitoring, however, nominal 99% control limits based on 16 expected deaths were crossed on two occasions.

Conclusions: Hospital league tables for mortality from heart surgery will be of limited value because year to year differences in death rate can be large (odds ratio 1.5) even when the underlying risk or case mix does not change. Statistical quality control of a single series with adjustment for case mix is the only way to take into account recent performance when informing a patient of the risk of surgery at a particular hospital. If there is an increase in the number of deaths the chances of the next patient surviving surgery can be calculated from the last 16 deaths.

Figures

Figure 1

Probability of detecting an increase in death rate when the rate increases by a factor of 1.1, 1.5, 2, and 4. For example, for 90% probability of detecting a doubling of death rate from a standard rate of 10% to 20%, 16 expected deaths are required at standard rate—that is, 160 operations

Figure 2

Cumulative risk adjusted mortality chart with 99% control limits for change in mortality in last 16 expected deaths