The Field Guide › Paper
A landmark patient-safety paper using NSQIP data on 84,730 general and vascular surgery patients (2005 to 2007) to test what actually distinguishes high-mortality from low-mortality hospitals. Risk-adjusted complication rates were almost identical across the mortality spectrum, roughly 25 to 27 percent: bad things happen at close to the same rate everywhere. What differed dramatically was what happened next. Patients who developed a major complication were nearly twice as likely to die at the highest-mortality hospitals as at the lowest (21.4% versus 12.5%), and this gap in 'failure to rescue' (death following a complication, rather than the complication itself) was the primary driver of the overall mortality variation. The finding reframes the safety question: avoiding every deviation isn't the achievable or even the decisive goal; what separates systems that cope from systems that don't is the capacity to notice a deviation early and respond to it effectively before it compounds. The discussion, careful to flag this as inference rather than something the study itself measured, points to collaborative team communication, nursing staffing ratios and ICU organisation as the likely mechanisms, citing Pronovost's finding that dedicated intensivist rounding cut mortality by a factor of three. A rigorous, large-scale empirical anchor for the argument that weak-signal detection and response capacity, not error elimination, is where resilience actually lives.