Citation: Shao, Y.; Cheng, Y.;
Gottipati, S.; Zeng-Treitler, Q.
Outcome Prediction for Patients with
Bipolar Disorder Using Prodromal
and Onset Data. Appl. Sci. 2023, 13,
1552. https://doi.org/10.3390/
app13031552
Academic Editors: Krzysztof
Ejsmont, Aamer Bilal Asghar,
Yong Wang and Rodolfo Haber
Received: 6 December 2022
Revised: 13 January 2023
Accepted: 18 January 2023
Published: 25 January 2023
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Article
Outcome Prediction for Patients with Bipolar Disorder Using
Prodromal and Onset Data
†
Yijun Shao
1,2
, Yan Cheng
1,2
, Srikanth Gottipati
1
and Qing Zeng-Treitler
1,2,
*
1
Biomedical Informatics Center, George Washington University, Washington, DC 20037, USA
2
Washington DC VA Medical Center, Washington, DC 20422, USA
* Correspondence: zengq@gwu.edu; Tel.: +1-202-994-0477
† This paper is an extended version of our abstract published in the Conference Proceedings of American
Society for Clinical Pathology (ASCP) 2019 Annual Meeting: Innovations in Personalized Medicine from
Biomarkers to Patient-Centered Care, Phoenix, AZ, USA, 11–13 September 2019.
Abstract:
Background: Predicting the outcomes of serious mental illnesses including bipolar disorder
(BD) is clinically beneficial, yet difficult. Objectives: This study aimed to predict hospitalization
and mortality for patients with incident BD using a deep neural network approach. Methods: We
randomly sampled 20,000 US Veterans with BD. Data on patients’ prior hospitalizations, diagnoses,
procedures, medications, note types, vital signs, lab results, and BD symptoms that occurred within
1 year before and at the onset of the incident BD were extracted as features. We then created novel
temporal images of patient clinical features both during the prodromal period and at the time of the
disease onset. Using each temporal image as a feature, we trained and tested deep neural network
learning models to predict the 1-year combined outcome of hospitalization and mortality. Results:
The models achieved accuracies of 0.766–0.949 and AUCs of 0.745–0.806 for the combined outcomes.
The AUC for predicting mortality was 0.814, while its highest and lowest values for predicting
different types of hospitalization were 90.4% and 70.1%, suggesting that some outcomes were more
difficult to predict than others. Conclusion: Deep learning using temporal graphics of clinical history
is a new and promising analytical approach for mental health outcome prediction.
Keywords: prediction; bipolar disorder; deep neural network; support vector machine
1. Introduction
Outcome prediction can facilitate physician and patient decision making as well as
disease management. In the management of serious mental illnesses, the accurate and early
identification of patients at high risk of mortality and other adverse event will enable a
targeted monitoring and interventions to improve prognosis. While many risk indices and
calculators have been developed for a wide range of diseases and conditions, the outcome
prediction of mental illness, especially, serious mental illnesses such as bipolar disorder,
remains very challenging [1,2].
A wide array of features has been used in predictive modeling in mental health,
ranging from early treatment response to genetics. The AUC of these predictive models,
however, rarely reached 80%, which some argued is the threshold of clinical utility [
1
].
For instance, with data from mood disorder patients with first lifetime episodes of major
depression, Tondo et al. used multivariate, logistic modeling and Bayesian methods to
predict a later diagnosis of unipolar versus bipolar depressive disorder. It reported correct
classification rates of 64–67% of cases and an AUC of 0.72 [3].
Challenges faced by outcome prediction in patients with mental illnesses include
inadequate sample size, heterogeneity of the patients sharing the same diagnosis, and
individual choices and disparities [
1
,
4
]. However, we can potentially overcome these
challenges with larger samples, more nuanced feature representations, and state-of-the-art
machine learning methods.
Appl. Sci. 2023, 13, 1552. https://doi.org/10.3390/app13031552 https://www.mdpi.com/journal/applsci
