ABSTRACT
OBJECTIVE: A recent study leveraging machine learning methods found that postpartum hemorrhage (PPH) can be predicted accurately at the time of labor admission in the U.S. Consortium for Safe Labor (CSL) dataset, with a C-statistic as high as 0.93. These CSL models were developed in older data (2002-2008) and used an estimated blood loss (EBL) of ≥1,000 mL to define PPH. We sought to externally validate these models using a more recent cohort of births where blood loss was measured using quantitative blood loss (QBL) methods. STUDY DESIGN: Using data from 5,261 deliveries between February 1, 2019 and May 11, 2020 at a single tertiary hospital, we mapped our electronic health record (EHR) data to the 55 predictors described in previously published CSL models. PPH was defined as QBL ≥1,000 mL within 24 hours after delivery. Model discrimination and calibration of the four CSL models were measured using our cohort. In a secondary analysis, we fit new models in our study cohort using the same predictors and algorithms as the original CSL models. RESULTS: The original study cohort had a substantially lower rate of PPH, 4.8% (7,279/228,438) versus 25% (1,321/5,261), possibly due to differences in measurement. The CSL models had lower discrimination in our study cohort, with a C-statistic as high as 0.57 (logistic regression). Models refit in our study cohort achieved better discrimination, with a C-statistic as high as 0.64 (random forest). Calibration improved in the refit models as compared with the original models. CONCLUSION: The CSL models' accuracy was lower in a contemporary EHR where PPH is assessed using QBL. As institutions continue to adopt QBL methods, further data are needed to understand the differences between EBL and QBL to enable accurate prediction of PPH. KEY POINTS: · Machine learning methods may help predict PPH.. · EBL models do not generalize when QBL is used.. · Blood loss estimation alters model accuracy..
ABSTRACT
BACKGROUND: A leading cause of preventable maternal death is related to delayed response to clinical warning signs. Electronic surveillance systems may improve detection of maternal morbidity with automated notifications. This retrospective observational study evaluates the ability of an automated surveillance system and the Maternal Early Warning Criteria (MEWC) to detect severely morbid postpartum hemorrhage (sPPH) after delivery. METHODS: The electronic health records of adult obstetric patients of any gestational age delivering between April 1, 2017 and December 1, 2018 were queried to identify scheduled or unscheduled vaginal or cesarean deliveries. Deliveries complicated by sPPH were identified and defined by operative management of postpartum hemorrhage, transfusion of ≥4 units of packed red blood cells (pRBCs), ≥2 units of pRBCs and ≥2 units of fresh-frozen plasma, transfusion with >1 dose of furosemide, or transfer to the intensive care unit. The test characteristics of automated pages and the MEWC for identification of sPPH 24 hours after delivery were determined and compared using sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) and their 95% confidence intervals (CIs). McNemar test was used to compare these estimates for both early warning systems. RESULTS: The average age at admission was 30.7 years (standard deviation [SD] = 5.1 years), mean gestational age 38 weeks 4 days, and cesarean delivery accounted for 30.0% of deliveries. Of 7853 deliveries, 120 (1.5%) were complicated by sPPH. The sensitivity of automated pages for sPPH within 24 hours of delivery was 60.8% (95% CI, 52.1-69.6), specificity 82.5% (95% CI, 81.7-83.4), PPV 5.1% (95% CI, 4.0-6.3), and NPV 99.3% (95% CI, 99.1-99.5). The test characteristics of the MEWC for sPPH were sensitivity 75.0% (95% CI, 67.3-82.7), specificity 66.3% (95% CI, 65.2-67.3), PPV 3.3% (95% CI, 2.7-4.0), and NPV 99.4% (95% CI, 99.2-99.6). There were 10 sPPH cases identified by automated pages, but not by the MEWC. Six of these cases were identified by a page for anemia, and 4 cases were the result of vital signs detected by the bedside monitor, but not recorded in the patient's medical record by the bedside nurse. Therefore, the combined sensitivity of the 2 systems was 83.3% (95% CI, 75.4-89.5). CONCLUSIONS: The automated system identified 10 of 120 deliveries complicated by sPPH not identified by the MEWC. Using an automated alerting system in combination with a labor and delivery unit's existing nursing-driven early warning system may improve detection of sPPH.
Subject(s)
Early Warning Score , Postpartum Hemorrhage/diagnosis , Vital Signs , Adult , Early Diagnosis , Electronic Health Records , Female , Humans , Postpartum Hemorrhage/etiology , Postpartum Hemorrhage/physiopathology , Postpartum Hemorrhage/therapy , Postpartum Period , Predictive Value of Tests , Pregnancy , Reproducibility of Results , Retrospective Studies , Risk Assessment , Risk Factors , Severity of Illness Index , Time FactorsABSTRACT
Consumers have greater access to data, information, and tools to support the management of their health than ever before. While the sheer quantity of these resources has increased exponentially over the past decade, the accuracy of consumer-facing resources is variable, and the value to the individual consumer remains uncertain. In general, the quality of these resources has improved, mostly because of improvements in web and mobile technologies and efforts to restructure health care delivery to be more patient centered. We describe the major initiatives that have led to consumers' increased access to both their own health data and performance data for health care providers and hospitals. We explore how search engines and crowdsourced review websites help and hinder the dissemination of medically accurate information. We highlight emerging examples of websites and apps that enable consumers to make medical decisions more in concert with their preferences. We conclude by describing key limitations of consumer-facing resources and making recommendations for how they may best be curated and regulated.