I had decided to stop commenting about it because enough is enough, but a recent paper in the BMJ raises a question about one of the criteria ProPublica used to formulate its ratings.
ProPublica defined complications 1) as any patient readmission within 30 days and 2) "any patient deaths during the initial surgical stay."
The authors of the BMJ paper randomly selected 100 records of patients who died at each of 34 hospitals in the United Kingdom. The 3400 records were reviewed by experts to determine whether a death could have been avoided if the quality of care had been better.
The number of patient records in which a death was at least 50% likely to have been avoidable was 123 or 3.6%.
There was a very weak association between the number of preventable deaths and the overall number of deaths occurring at each hospital. By two measures of overall hospital deaths, the hospital standardized mortality ratio and the summary hospital level mortality indicator, the correlation coefficient between avoidable deaths and all deaths was 0.3, not statistically significant.
From the paper: "The absence of even a moderately strong association is a reflection of the small proportion of deaths (3.6%) judged likely to be avoidable and of the relatively small variation in avoidable death proportions between trusts [hospitals]. This confirms what others have demonstrated theoretically—that is, no matter how large the study the signal (avoidable deaths) to noise (all deaths) ratio means that detection of significant differences between trusts is unlikely."
The Surgeon Scorecard was derived from administrative data. No individual analysis of patient deaths was undertaken. According to a ProPublica article discussing some key questions about their methodology, "As for deaths, we took a conservative approach and only included those that occurred in the hospital within the initial stay."
Maybe that wasn't such a conservative approach after all.
And maybe we need to rethink that 2013 paper claiming that medical error caused up to 440,000 deaths per year.