(PS10-52) Variables Associated with Nonresponse in Anxiety and Depression Among Patients Receiving Treatment in a Partial Hospitalization Program

Abstract Text: Partial hospital programs (PHPs) represent a key point along the psychiatric treatment continuum. As in any clinical setting, a portion of patients receiving psychological treatment do not meaningfully respond to treatment in PHPs. Identifying variables associated with nonresponse may allow clinicians to better predict who is less likely to respond, and alter treatment procedures accordingly. This project aimed identify baseline variables that are most predictive of nonresponse among patients who did not respond to treatment in a PHP. Analysis was conducted with two samples of patients who presented to treatment with clinically elevated anxiety (n=1227; Mage = 36.68, 67.2% female, 74% white) and clinically elevated depression (n=1272, Mage = 36.59, 68.9% female, 74.2% white). Patients were classified as nonresponders if they had a reliable change index score lower than the 1.96 threshold proposed by Jacobson and Truax (1991). 24.8% (n=315) of the anxiety sample and 36.3% (n=445) of the depression sample were classified as nonresponders. To identify variables that may be associated with treatment nonresponse, two logistic least absolute shrinkage and selection operator (LASSO) regression models were conducted with anxiety and depression samples, including demographic, diagnostic, symptom severity, functioning, coping skills, and quality of life variables. LASSO regression is a machine learning approach in which variable selection and regularization is completed automatically based upon a tuning parameter (lambda; λ) and k = 10 cross-validation, allowing selection of favorable covariates from a large list of possible covariates. Variables selected as covariates by the LASSO model were included in two binary logistic regression models predicting nonresponse in anxiety and depression samples, respectively. In the depression sample, days absent from the program (OR=1.18, 95% CI=1.08-1.28, padjusted <.001) and higher baseline anxiety symptoms (OR=1.02, 95% CI=1.01-1.03) increased likelihood of depression nonresponse, whereas completing treatment (OR=0.52, 95% CI=.31-86), higher discharge coping ability (OR=0.87, 95% CI =0.79-0.95, padjusted <.001), better functioning at discharge (OR=0.92, 95% CI=0.87-0.96, padjusted <.001), and higher positive mental health at discharge (OR=0.93, 95% CI=0.89-0.98, padjusted=.01) decreased likelihood of depression nonresponse. In the anxiety sample, days absent from the program (OR=1.18, 95% CI=1.09-1.28, padjusted <.001) increased likelihood of anxiety nonresponse, whereas completing treatment (OR=0.56, 95% CI=0.34-0.89), having a comorbid anxiety disorder (OR=0.66, 95% CI=0.45-0.94), higher discharge coping ability (OR=0.83, 95% CI=0.76-0.90, padjusted <.001), and higher positive mental health at discharge (OR=0.94, 95% CI=0.90-0.98, padjusted=.018) decreased likelihood of anxiety nonresponse. These results suggest that patients with greater levels of clinical severity, more functional impairment, and more absences were less likely to respond to PHP treatment. Addressing barriers to treatment adherence and encouraging the uptake and use of coping skills to increase functioning may help to reduce the rates of treatment nonresponse in PHPs.