Abstract 316: Missing Data and the SOFA Score in Cardiac Arrest Research: A Scoping Review
Introduction: The Sequential Organ Failure Assessment (SOFA) score is often used as an outcome or exposure in cardiac arrest studies. SOFA requires lab values and vital signs at certain time points which often results in missing data. How this missing data is handled is unknown. Methods: We performed a scoping review of PubMed, EMBASE, and Web of Science. English language peer-reviewed manuscripts were included. Titles/abstracts were screened by two independent reviewers to assess if they met inclusion criteria. Studies that met inclusion criteria were retrieved in full; those that did not were excluded. Disagreements between reviewers were resolved by a third reviewer. Results: The initial search provided 408 abstracts, 142 underwent full-text review (kappa: 0.91), and 66 were included (5 randomized controlled trials, 26 prospective and 25 retrospective studies). The studies had a median of 151 (IQR: 55, 278) subjects. SOFA was used as an outcome in 36 (55%) and a primary outcome in 10 (15%). Only 27 (41%) studies reported a method to handle missing SOFA data. The most common method was to exclude subjects with missing data (81%). Other methods were use of maximum SOFA while subjects were alive (11%), modified SOFA after excluding subjects who died prior to the timepoint (11%), and earlier and later SOFA to impute values (4%). When SOFA was the primary outcome, 4 (40%) reported a method; 3 (75%) excluded subjects and one (25%) used modified SOFA. Two studies conducted sensitivity analyses to test assumptions used to handle missing SOFA (one imputed values for death/discharge, one adjusted for mortality difference prior to SOFA measurement). Only 9 studies (14%) mentioned quantity of missing SOFA, ranging from 0-76% (median: 10% [IQR: 6%, 42%]). In the 50 studies using SOFA at time points after baseline, only 11 (22%) mentioned mortality prior to SOFA measurement; when mentioned, it ranged from 3%-76% with a median of 12% (IQR: 6%-35%). Conclusion: Missing data for SOFA scores used in cardiac arrest studies is pervasive yet often not acknowledged and/or handled with described or consistent methods. These findings illustrate that studies using SOFA may exhibit substantial bias and results could be misinterpreted, particularly if patients with missing data are excluded.