The Occurrence of and Risk Factors for Depressive Symptomatology in MyocarditisSurvivors: A Cross-Sectional Survey-Based Study Using Machine Learning

Jean Marrero-Polanco ¹ Laura Suarez Pardo ² Shehzad K Niazi ³ Daniel G Smith ² Cynthia J Stoppel ² Candace Moose ⁴ Arjun P Athreya ¹ Leslie T Cooper ^4,5 William V Bobo ^6*

• ¹ Department of Molecular Pharmacology and Experimental Therapeutics, Mayo Clinic, Rochester, Michigan, United States
• ² Department of Psychiatry and Psychology, Mayo Clinic, Rochester, Minnesota, United States
• ³ Department of Psychiatry and Psychology, Mayo Clinic, Jacksonville, United States
• ⁴ The Myocarditis Foundation, Kingwood, TX, United States
• ⁵ Department of Cardiovascular Medicine, Mayo Clinic, Jacksonville, United States
• ⁶ Department of Behavioral Sciences and Social Medicine, College of Medicine, Florida State University, Tallahassee, Florida, United States

The frequency and impact of depressive symptoms in myocarditis survivors are poorly understood.Objectives: We conducted a cross-sectional study to identify risk factors and the relative importance of each for predicting clinically significant depressive symptomatology in myocarditis survivors.Participants completed an electronic survey assessing sociodemographic, general health, and myocarditis-related variables, as well as self-reported cardiac symptoms and personal and family mental health history. Participants also completed the Center for Epidemiologic Studies Depression Scale (CES-D), Beck Anxiety Inventory (BAI), revised Impact of Events Scale (IES-R), and other validated measures of social support, quality of life, resiliency, childhood adversity, treatment distress, and somatic symptom burden. Clinically significant depressive symptomatology was defined as a CES-D total score >16. We used supervised machine learning to examine which and how well psychosocial and other types of variables predicted clinically significant depressive symptomatology in myocarditis survivors. Finally, we calculated the variable importance for each variable from the trained models and examined the rank ordering of predictors.Results: Ninety-six of 113 respondents (85.0%) with complete survey data were included in the analyses. Forty-three (44.8%) respondents had clinically significant depressive symptomatology.When predicting depressive symptomatology, random forests achieved a mean AUC of 0.91 (95% CI 0.87-0.95) and a significantly higher accuracy than that of the null information rate (0.84 vs 0.55, p < 0.005), with correspondingly high sensitivity (0.84) and specificity (0.85). Emotional wellbeing, quality of life, history of depression, anxiety, and resilience were the top predictors in variable importance analyses, ahead of self-reported cardiovascular symptoms, other myocarditisrelated variables, and family history of depression.Myocarditis survivors are at high risk for clinically significant depressive symptomatology. Psychosocial factors that are measurable in routine practice may be more predictive of significant depressive symptomatology than demographics, family history, or selfreported cardiovascular symptoms.