(268) Data Missingness Reporting and the Use of Methods to Address it in Critical Care Cohort Studies

Introduction: Failure to recognize and address data missingness in cohort studies may result in biased results. We reviewed published ICU cohort studies to characterize missingness reporting and the use of methods to address it.

Methods: We searched the 2022 table of contents of 29 critical care/critical care subspecialty journals having a 2021 impact factor ≥3 to identify published, English-language, prospective clinical cohort or retrospective database cohort studies enrolling ≥100 adult or pediatric patients. In duplicate, two trained researchers conducted a manuscript/supplemental material PDF word search for ‘missing*’and extracted the following: study type, patient age, ICU type, sample size, whether result missingness was reported (and exceeded >5% for ≥1 variable) and the use of methods to address missingness [i.e., imputation, an analytic strategy (e.g., sensitivity analysis for the primary outcome) or both].

Results: Among the 656 identified articles, 334 (50.9%) mentioned result missingness. These articles were published in higher impact journals (p < 0.01) and enrolled larger samples (p =0.03). Degree of missingness was reported for ≥1 variable in 263/334 (78.7%) and exceeded >5% for ≥1 variable in 175/263 (66.5%). Complete case analysis was used in 120/334 (35.9%). A method to address missingness was used in nearly half (161/334, 48.2%): imputation only (101/161, 62.7%), an analytic strategy (22/161, 13.7%), or both (38/161, 23.6%). Only the presence of missingness >5% (101/161, 62.7%) was independently associated with greater missingness method use (aOR 2.27, 95% CI 1.45-3.56). Among the 139 studies using imputation, multiple imputation was used in 84 (60.4%) and single imputation in 50 (36.0%). Among the 61 studies employing an analytic method, a sensitivity analysis was used in 42 (68.9%) but only 13/42 (31.0%) of these studies considered missingness as a separate variable category.

Conclusions: Among published critical care cohort studies, only half mentioned result missingness, only 40% reported actual missingness, and only 25% employed a method to manage missingness. Rigorous missingness methods like multiple imputation or primary outcome sensitivity analysis are not commonly used.