Abstract Sat405: Do Neuroprognostic Studies in Comatose Cardiac Arrest Survivors Report Factors Contributing to Self-Fulfilling Prophecy Bias in Their Methodology? A Systematic Review

Abstract

Background: Self-fulfilling prophecy bias (SFPB) occurs when a perceived prognosis guides treatment decisions, thereby modulating outcomes and potentially inflating the prediction performance of prognostic tools. We hypothesize that factors influencing the degree to which studies are impacted by SFPB are underreported in cardiac arrest (CA) neuroprognostic studies. Goal: Demonstrate the degree to which the CA neuroprognostic literature objectively accounts for factors related to SFPB. Methods: We searched MEDLINE, Embase, and Cochrane databases. Two independent reviewers screened and extracted data using Distiller SR® from observational studies and systematic reviews/meta-analyses (published from inception to 12/31/2022) evaluating the performance of guideline-recommended neuroprognostic tools in comatose CA survivors (Glasgow Coma Scale < 8 or motor subscore < 6) aged > 17 years. Data were synthesized descriptively; heterogeneity of studies precluded meta-analysis. Results: Of 6161 articles, 221 were included. SFPB was reported in 49% (109/221) of studies with 19% (43/221) also reporting strategies to mitigate this bias. Mortality was reported in 63% (139/221) of studies, of which 44% (61/139) reported timing and 31% (43/139) reported mode of death. Furthermore, WLS practices—study-specific, institutional, or cultural—were reported in 46% (64/139) of studies, with 11% (7/64) reporting that cultural or protocol restrictions precluded WLS. The remaining 89% (57/64) of studies mentioned WLS, though WLS rates were provided in only 47% (27/57) with only one study excluding the tool investigated from neuroprognostic impressions informing WLS decisions. Additionally, 56% (15/27) of studies reported perceived poor neurologic prognosis as the most common indication, and in 48% (13/27) WLS occurred within 1 week of CA. Blinding was reported in 55% (122/221) of studies, including blinding of data analysts (35%; 43/122), investigators (45%; 55/122), outcome assessors (25%; 30/122), and treatment team (20%; 25/122). The remaining 45% (99/221) of studies were either unblinded or omitted reporting on blinding. Conclusions: Factors contributing to SFPB are drastically underreported and undermine the reliability of studies informing neuroprognostication tools and guidelines in CA survivors. Our findings can inform standardized reporting guidelines for neuroprognostic studies aiming at increasing transparency and rigor in future studies.