Search

Background Postoperative hemodynamic deterioration among cardiac surgical patients can indicate or lead to adverse outcomes. Whereas prediction models for such events using electronic health records or physiologic waveform data are previously described, their combined value remains incompletely defined. The authors hypothesized that models incorporating electronic health record and processed waveform signal data (electrocardiogram lead II, pulse plethysmography, arterial catheter tracing) would yield improved performance versus either modality alone. Methods Intensive care unit data were reviewed after elective adult cardiac surgical procedures at an academic center between 2013 and 2020. Model features included electronic health record features and physiologic waveforms. Tensor decomposition was used for waveform feature reduction. Machine learning–based prediction models included a 2013 to 2017 training set and a 2017 to 2020 temporal holdout test set. The primary outcome was a postoperative deterioration event, defined as a composite of low cardiac index of less than 2.0 ml min˗1 m˗2, mean arterial pressure of less than 55 mmHg sustained for 120 min or longer, new or escalated inotrope/vasopressor infusion, epinephrine bolus of 1 mg or more, or intensive care unit mortality. Prediction models analyzed data 8 h before events. Results Among 1,555 cases, 185 (12%) experienced 276 deterioration events, most commonly including low cardiac index (7.0% of patients), new inotrope (1.9%), and sustained hypotension (1.4%). The best performing model on the 2013 to 2017 training set yielded a C-statistic of 0.803 (95% CI, 0.799 to 0.807), although performance was substantially lower in the 2017 to 2020 test set (0.709, 0.705 to 0.712). Test set performance of the combined model was greater than corresponding models limited to solely electronic health record features (0.641; 95% CI, 0.637 to 0.646) or waveform features (0.697; 95% CI, 0.693 to 0.701). Conclusions Clinical deterioration prediction models combining electronic health record data and waveform data were superior to either modality alone, and performance of combined models was primarily driven by waveform data. Decreased performance of prediction models during temporal validation may be explained by data set shift, a core challenge of healthcare prediction modeling. Editor’s Perspective What We Already Know about This Topic What This Article Tells Us That Is New

Additional file 1: Supplemental Table 1. Interaction of intraoperative management and reduced left ventricular ejection fraction on the occurrence of the primary outcome. Supplemental Table 2. Multivariable logistic regressions showing effect of remifentanil on each outcome. Supplemental Table 3. Multivariable logistic regressions showing effect of fluid balance on each outcome.

Additional file 2: Supplemental Fig. 1. Assignment of surgical risk category based on the incidence of the primary outcome for each surgical type. The surgical risk categories are displayed using distinct colors as defined in the figure legend. High-risk surgery consisted of acute care and trauma surgery. Moderate-risk surgery consisted of thoracic, vascular, otolaryngology, and neurosurgery. Low-risk surgery consisted of urology and gynecology, general surgery, orthopedics, and plastic surgery.

Variability persists in intraoperative red blood cell (RBC) transfusion rates, despite evidence supporting associated adverse sequelae. We evaluated whether beliefs concerning transfusion risk and safety are independently associated with the inclination to transfuse. We surveyed intraoperative transfusion decision-makers from 33 cardiac surgery programs in Michigan. The primary outcome was a provider’s reported inclination to transfuse (via a six-point Likert Scale) averaged across 10 clinical vignettes based on Class IIA or IIB blood management guideline recommendations. Survey questions assessed hematocrit threshold for transfusion (“hematocrit trigger”), demographic and practice characteristics, years and case-volume of practice, knowledge of transfusion guidelines, and provider attitude regarding perceived risk and safety of blood transfusions. Linear regression models were used to estimate the effect of these variables on transfusion inclination. Mixed effect models were used to quantify the variation attributed to provider specialties and hematocrit triggers. The mean inclination to transfuse was 3.2 (might NOT transfuse) on the survey Likert scale (SD: .86) across vignettes among 202/413 (48.9%) returned surveys. Hematocrit triggers ranged from 15% to 30% (average: 20.4%; SE: .18%). The inclination to transfuse in situations with weak-to-moderate evidence for supporting transfusion was associated with a provider’s hematocrit trigger (p < .01) and specialty. Providers believing in the safety of transfusions were significantly more likely to transfuse. Provider specialty and belief in transfusion safety were significantly associated with a provider’s hematocrit trigger and likelihood for transfusion. Our findings suggest that blood management interventions should target these previously unaccounted for blood transfusion determinants.

Supplemental Material, sj-pdf-1-jic-10.1177_0885066620988675 for Use of the Change in Weaning Parameters as a Predictor of Successful Re-Extubation by Suraj Trivedi, Ryan Davis, Milo C. Engoren, Javier Lorenzo, Graciela Mentz, Elizabeth S. Jewell and Michael D. Maile in Journal of Intensive Care Medicine