Mortality Trends After a Voluntary Checklist-based Surgical Safety Collaborative

Abstract

Objective: 

To determine whether completion of a voluntary, checklist-based surgical quality improvement program is associated with reduced 30-day postoperative mortality.

Background: 

Despite evidence of efficacy of team-based surgical safety checklists in improving perioperative outcomes in research trials, effective methods of population-based implementation have been lacking. The Safe Surgery 2015 South Carolina program was designed to foster state-wide engagement of hospitals in a voluntary, collaborative implementation of a checklist program.

Methods: 

We compared postoperative mortality rates after inpatient surgery in South Carolina utilizing state-wide all-payer discharge claims from 2008 to 2013, linked with state vital statistics, stratifying hospitals on the basis of completion of the checklist program. Changes in risk-adjusted 30-day mortality were compared between hospitals, using propensity score-adjusted difference-in-differences analysis.

Results: 

Fourteen hospitals completed the program by December 2013. Before program launch, there was no difference in mortality trends between the completion cohort and all others (P = 0.33), but postoperative mortality diverged thereafter (P = 0.021). Risk-adjusted 30-day mortality among completers was 3.38% in 2010 and 2.84% in 2013 (P < 0.00001), whereas mortality among other hospitals (n = 44) was 3.50% in 2010 and 3.71% in 2013 (P = 0.3281), reflecting a 22% difference between the groups on difference-in-differences analysis (P = 0.0021).

Conclusions: 

Despite similar pre-existing rates and trends of postoperative mortality, hospitals in South Carolina completing a voluntary checklist-based surgical quality improvement program had a reduction in deaths after inpatient surgery over the first 3 years of the collaborative compared with other hospitals in the state. This may indicate that effective large-scale implementation of a team-based surgical safety checklist is feasible.