Techniques to monitor for endophthalmitis and other cataract surgery complications
Ng, J. Q., Morlet, N., Bremner, A. P., Bulsara, M. K., Morton, A. P., & Semmens, J. B. (2008). Techniques to monitor for endophthalmitis and other cataract surgery complications. Ophthalmology, 115(1), 3-10. doi:10.1016/j.ophtha.2007.09.010
Objective: To report complication incidence to identify outliers more reliably, to provide feedback on performance, and to generate more timely alerts.
Design: Data from a retrospective entire-population study was used as an example for the charting methods.
Participants: The Western Australian (WA) Data Linkage System identified all cataract or lens-related procedures undertaken in WA and those operations complicated with endophthalmitis over 20 years from 1980.
Methods: Use of risk-adjusted charts to assess complication incidence between hospitals. We compare these with ones that demonstrate individual hospital performance. The latter also adjust for risk and enable reporting at the time of complication rather than after a data collection period.
Main Outcome Measure: Excessive complication risk (postoperative endophthalmitis).
Results: Confidence limits allow comparison of hospitals performing different numbers of operations; the 95% Poisson prediction interval was exceeded by 4 possible-outlier hospitals. Case-mix risk adjustment better narrowed them to probable outliers (now only 2 hospitals). However, 2 high-volume nonoutlier hospitals had a short duration of significantly higher risk of endophthalmitis with cumulative sum analysis. Their endophthalmitis numbers were not excessive, and they were not identified as outliers by the other methods.
Conclusion: Simple ranking (or league) tables are not useful enough; someone is always first and last. Chance and circumstance will push all towards the middle with time. Risk-adjusted observed versus expected charting better identifies outliers than a funnel plot. Better still, the use of cumulative sum analysis can help surgeons distinguish between failures due to random processes and those that are associated with problems that require investigation to search for potentially correctable causes.