Optimising the prediction of depression remission: A longitudinal machine learning approach

Forthcoming in Neuropsychiatric Genetics

Ewan Carr¹, Marcella Rietschel², Ole Mors³, Neven Henigsberg⁴, Katherine J. Aitchison^5,6,7,8, Wolfgang Maier^9,10, Rudolf Uher¹¹, Anne Farmer¹², Peter McGuffin¹², Raquel Iniesta^1,13*

¹Department of Biostatistics & Health Informatics, Institute of Psychiatry, Psychology and Neuroscience (IoPPN), London, United Kingdom.

²Department of Genetic Epidemiology in Psychiatry, Central Institute of Mental Health, Medical Faculty, Mannheim, Germany

³Psychosis Research Unit, Aarhus University Hospital - Psychiatry, Palle Juul Jensens Boulevard 175, 8200 Aarhus N, Denmark

⁴Croatian Institute for Brain Research, Medical School, University of Zagreb, Salata 3, 10 000, Zagreb, Croatia

⁵College of Health Sciences, Departments of Psychiatry and Medical Genetics, University of Alberta, 116 St and 85 Ave, Edmonton, AB, T6G 2R3, Canada

⁶Neuroscience and Mental Health Institute, University of Alberta, 116 St and 85 Ave, Edmonton, AB, T6G 2R3, Canada
⁷Women and Children's Health Research Institute, University of Alberta, 116 St and 85 Ave, Edmonton, AB, T6G 2R3, Canada
⁸Northern Ontario School of Medicine, 955 Oliver Rd, Thunder Bay, ON, P7B 5E1, Canada

⁹Department of Neurodegenerative Diseases and Geriatric Psychiatry, University Hospital Bonn, Bonn, Germany

¹⁰German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany

¹¹Dalhousie University Department of Psychiatry, 5909 Veterans' Memorial Drive, Halifax, B3H 2E2, Nova Scotia, Canada

¹²Social, Genetic and Developmental Psychiatry Centre, Institute of Psychiatry, Psychology and Neuroscience, King's College London, 16 De Crespigny Park, Denmark Hill, London, SE5 8AF, UK

¹³King’s Institute for Artificial Intelligence, King’s College London, UK

^*Corresponding author.

Abstract

Decisions about when to change antidepressant treatment are complex and benefit from accurate prediction of treatment outcome. Prognostic accuracy can be enhanced by incorporating repeated assessments of symptom severity collected during treatment. Participants (n=714) from the Genome-Based Therapeutic Drugs for Depression study received escitalopram or nortriptyline over 12 weeks. Remission was defined as scoring ≤7 on the Hamilton Rating Scale. Predictors included demographic, clinical, and genetic variables (at 0 weeks) and measures of symptom severity (at 0, 2, 4 and 6 weeks). Longitudinal descriptors extracted with growth curves and topological data analysis were used to inform prediction of remission. Repeated assessments produced gradual and drug-specific improvements in predictive performance. By week four, models’ discrimination in all samples reached levels that might usefully inform treatment decisions (AUC=0.777 for nortriptyline; AUC=0.807 for escitalopram; AUC=0.794 for combined sample).   Decisions around switching or modifying treatments for depression can be informed by repeated symptom assessments collected during treatment, but not until four weeks after the start of treatment.  

Keywords

• Depression remission
• Repeated measures
• Machine learning
• Topological data analysis