A universal AutoScore framework to develop interpretable scoring systems for predicting common types of clinical outcomes

PARCCS: A Machine Learning Risk-Prediction Model for Acute Peripartum Cardiovascular Complications During Delivery Admissions

Background

Maternal mortality in the United States remains high, with cardiovascular (CV) complications being a leading cause.

Objectives

The purpose of this paper was to develop the PARCCS (Prediction of Acute Risk for Cardiovascular Complications in the Peripartum Period Score) for acute CV complications during delivery.

Methods

Data from the National Inpatient Sample (2016-2020) and International Classification of Diseases, Tenth Revision codes to identify delivery admissions were used. Acute CV/renal complications were defined as a composite of pre-eclampsia/eclampsia, peripartum cardiomyopathy, renal complications, venous thromboembolism, arrhythmias, and pulmonary edema. A risk prediction model, PARCCS, was developed using machine learning consisting of 14 variables and scored out of 100 points.

Results

Of the 2,371,661 pregnant patients analyzed, 7.0% had acute CV complications during delivery hospitalization. Patients with CV complications had a higher prevalence of comorbidities and were more likely to be of Black race and lower income. The PARCCS variables included electrolyte imbalances (13 points [p]), age (3p for age <20 years), cesarean delivery (4p), obesity (5p), pre-existing heart failure (28p), multiple gestations (4p), Black race (2p), gestational hypertension (3p), low income (1p), gestational diabetes (2p), chronic diabetes (6p), prior stroke (22p), coagulopathy (5p), and nonelective admission (2p). Using the validation set, the performance of the model was evaluated with an area under the receiver-operating characteristic curve of 0.68 and a 95% CI of 0.67 to 0.68.

Conclusions

PARCCS has the potential to be an important tool for identifying pregnant individuals at risk of acute peripartum CV complications at the time of delivery. Future studies should further validate this score and determine whether it can improve patient outcomes.

Explainable artificial intelligence in emergency medicine: an overview

Artificial intelligence (AI) and machine learning (ML) have potential to revolutionize emergency medical care by enhancing triage systems, improving diagnostic accuracy, refining prognostication, and optimizing various aspects of clinical care. However, as clinicians often lack AI expertise, they might perceive AI as a "black box," leading to trust issues. To address this, "explainable AI," which teaches AI functionalities to end-users, is important. This review presents the definitions, importance, and role of explainable AI, as well as potential challenges in emergency medicine. First, we introduce the terms explainability, interpretability, and transparency of AI models. These terms sound similar but have different roles in discussion of AI. Second, we indicate that explainable AI is required in clinical settings for reasons of justification, control, improvement, and discovery and provide examples. Third, we describe three major categories of explainability: pre-modeling explainability, interpretable models, and post-modeling explainability and present examples (especially for post-modeling explainability), such as visualization, simplification, text justification, and feature relevance. Last, we show the challenges of implementing AI and ML models in clinical settings and highlight the importance of collaboration between clinicians, developers, and researchers. This paper summarizes the concept of "explainable AI" for emergency medicine clinicians. This review may help clinicians understand explainable AI in emergency contexts.