BACKGROUND: Postoperative pulmonary complications can have a significant impact on the morbidity and mortality of patients undergoing major surgeries. Intraoperative lung protective strategies using low tidal volume (TV) ventilation and positive end-expiratory pressure (PEEP) have been demonstrated to reduce the incidence of pulmonary injury and infection while improving oxygenation and respiratory mechanics. The purpose of this study was to develop decision support systems designed to optimize behavior of the attending anesthesiologist with regards to adherence with established intraoperative lung-protective ventilation (LPV) strategies. METHODS: Over a 4-year period, data were obtained from 49,386 procedures and 109 attendings. Cases were restricted to patients aged 18 years or older requiring general anesthesia that lasted at least 60 minutes. We defined protective lung ventilation as a TV of 6-8 mL/kg ideal body weight and a PEEP of ≥4 cm H2O. There was a baseline period followed by 4 behavioral interventions: education, near real-time feedback, individualized post hoc feedback, and enhanced multidimensional decision support. Segmented logistic regression using generalized estimating equations was performed in order to assess temporal trends and effects of interventions on adherence to LPV strategies. RESULTS: Consistent with improvement in adherence with LPV strategies during the baseline period, the predicted probability of adherence with LPV at the end of baseline was 0.452 (95% confidence interval [CI], 0.422-0.483). The improvements observed for each phase were relative to the preceding phase. Education alone was associated with an 8.7% improvement (P < .01) in adherence to lung-protective protocols and was associated with a 16% increase in odds of adherence (odds ratio [OR] = 1.16; 95% CI, 1.01-1.33; P = .04). Near real-time, on-screen feedback was associated with an estimated 15.5% improvement in adherence (P < .01) with a 69% increase in odds of adherence (OR = 1.69; 95% CI, 1.46-1.96; P < .01) over education alone. The addition of an individualized dashboard with personal adherence and peer comparison was associated with a significant improvement over near real-time feedback (P < .01). Near real-time feedback and dashboard feedback systems were enhanced based on feedback from the in-room attendings, and this combination was associated with an 18.1% (P < .01) increase in adherence with a 2-fold increase in the odds of adherence (OR = 2.23; 95% CI, 1.85-2.69; P < .0001) between the end of the previous on-screen feedback phase and the start of the individualized post hoc dashboard reporting phase. The adherence with lung-protective strategies using the multidimensional approach has been sustained for over 24 months. The difference between the end of the previous phase and the start of this last enhanced multidimensional decision support phase was not significant (OR = 1.08; 95% CI, 0.86-1.34; P = .48). CONCLUSIONS: Consistent with the literature, near real-time and post hoc reporting are associated with positive and sustained behavioral changes aimed at adopting evidence-based clinical strategies. Many decision support systems have demonstrated impact to behavior, but the effect is often transient. The implementation of near real-time feedback and individualized post hoc decision support tools has resulted in clinically relevant improvements in adherence with LPV strategies that have been sustained for over 24 months, a common limitation of decision support solutions.