• Raina M Merchant, Robert A Berg, Lin Yang, Lance B Becker, Peter W Groeneveld, Paul S Chan, and American Heart Association's Get With the Guidelines-Resuscitation Investigators.
• Department of Emergency Medicine, University of Pennsylvania, Philadelphia, PA.
• J Am Heart Assoc. 2014 Jan 1;3(1):e000400.

BackgroundIn-hospital cardiac arrest (IHCA) is common and often fatal. However, the extent to which hospitals vary in survival outcomes and the degree to which this variation is explained by patient and hospital factors is unknown.Methods And ResultsWithin Get with the Guidelines-Resuscitation, we identified 135 896 index IHCA events at 468 hospitals. Using hierarchical models, we adjusted for demographics comorbidities and arrest characteristics (eg, initial rhythm, etiology, arrest location) to generate risk-adjusted rates of in-hospital survival. To quantify the extent of hospital-level variation in risk-adjusted rates, we calculated the median odds ratio (OR). Among study hospitals, there was significant variation in unadjusted survival rates. The median unadjusted rate for the bottom decile was 8.3% (range: 0% to 10.7%) and for the top decile was 31.4% (28.6% to 51.7%). After adjusting for 36 predictors of in-hospital survival, there remained substantial variation in rates of in-hospital survival across sites: bottom decile (median rate, 12.4% [0% to 15.6%]) versus top decile (median rate, 22.7% [21.0% to 36.2%]). The median OR for risk-adjusted survival was 1.42 (95% CI: 1.37 to 1.46), which suggests a substantial 42% difference in the odds of survival for patients with similar case-mix at similar hospitals. Further, significant variation persisted within hospital subgroups (eg, bed size, academic).ConclusionSignificant variability in IHCA survival exists across hospitals, and this variation persists despite adjustment for measured patient factors and within hospital subgroups. These findings suggest that other hospital factors may account for the observed site-level variations in IHCA survival.

36 keywords...

hide…