Predicting future major depression and persistent depressive symptoms: Development of a prognostic screener and PHQ-4 cutoffs in breast cancer patients.

Reference
Weihs KL, Wiley JF, Crespi CM, Krull JL, Stanton AL. 2018. Predicting future major depression and persistent depressive symptoms: Development of a prognostic screener and PHQ-4 cutoffs in breast cancer patients. Psychooncology. 27:420–426. doi:10.1002/pon.4472.
Abstract

OBJECTIVE: Create a brief, self-report screener for recently diagnosed breast cancer patients to identify patients at risk of future depression.

METHODS: Breast cancer patients (N = 410) within 2 ± 1 months after diagnosis provided data on depression vulnerability. Depression outcomes were defined as a high depressive symptom trajectory or a major depressive episode during 16 months after diagnosis. Stochastic gradient boosting of regression trees identified 7 items highly predictive for the depression outcomes from a pool of 219 candidate depression vulnerability items. Three of the 7 items were from the Patient Health Questionnaire 4 (PHQ-4), a validated screener for current anxiety/depressive disorder that has not been tested to identify risk for future depression. Thresholds classifying patients as high or low risk on the new Depression Risk Questionnaire 7 (DRQ-7) and the PHQ-4 were obtained. Predictive performance of the DRQ-7 and PHQ-4 was assessed on a holdout validation subsample.

FINDINGS: DRQ-7 items assess loneliness, irritability, persistent sadness, and low acceptance of emotion as well as 3 items from the PHQ-4 (anhedonia, depressed mood, and worry). A DRQ-7 score of >=6/23 identified depression outcomes with 0.73 specificity, 0.83 sensitivity, 0.68 positive predictive value, and 0.86 negative predictive value. A PHQ-4 score of >=3/12 performed moderately well but less accurately than the DRQ-7 (net reclassification improvement = 10%; 95% CI [0.5-16]).

INTERPRETATION: The DRQ-7 and the PHQ-4 with a new cutoff score are clinically accessible screeners for risk of depression in newly diagnosed breast cancer patients. Use of the screener to select patients for preventive interventions awaits validation of the screener in other samples.