Predictive Utility of In-Hospital Screening for Postdischarge Depression Symptoms Among Survivors of Critical Illness.

BACKGROUND: Although postdischarge symptoms of depression are highly prevalent after critical illness, the best timing and strategies for screening are unclear. RESEARCH QUESTION: What is the value of using the 2-item Patient Health Questionnaire (PHQ-2) depression scale in the hospital to predict elevated postdischarge depression symptoms? STUDY DESIGN AND METHODS: This was a secondary analysis from a randomized factorial trial designed to optimize a postdischarge distress intervention among survivors of cardiorepiratory failure managed in ICUs in 3 states. The primary outcome was the 9-item Patient Health Questionnaire (PHQ-9) score measured immediately after discharge, whereas the key predictor was in-hospital PHQ-2 score. Performance metrics of the PHQ-2 across different score cut points were compared to identify the optimal threshold value for predicting clinically important postdischarge symptoms. Linear and logistic regression models examined the PHQ-2's performance as a continuous and binary variable alone, as well as with the addition of sociodemographic and clinical characteristics. RESULTS: Among 164 participants, clinically important depression symptoms were present after discharge among 72%, as defined by a PHQ-9 score of ≥ 5, and among 39%, as defined by a PHQ-9 score of ≥ 10. Compared with a PHQ-2 score threshold of ≥ 2, a value of ≥ 1 provided higher sensitivity (78% vs 58%), although lower specificity (61% vs 80%) among those with a PHQ-9 score of ≥ 5; metrics were similar with the use of a PHQ-9 score of ≥ 10. Compared with models with binary PHQ-2 scores alone, continuous PHQ-2 scores provided better discrimination (area under the receiver operating characteristic curve [AUC], 0.61-0.69 vs 0.74, respectively), but similar prediction accuracy (Brier score, 0.17-0.19 vs 0.17, respectively). The addition of sociodemographic and clinical characteristics improved the PHQ-2's discrimination (AUC range, 0.61-0.74 vs 0.77-0.83, respectively) and amount of variance explained (R 2 range, 0.18-0.36 vs 0.37-0.48, respectively). INTERPRETATION: Our results show that the in-hospital PHQ-2 demonstrated moderate predictive performance as a screening tool for postdischarge depression symptoms among survivors of critical illness. The addition of sociodemographic and clinical variables improved performance. CLINICAL TRIAL REGISTRATION: ClinicalTrials.gov; No.: NCT04038567; URL: www.clinicaltrials.gov.