# Predicting Psychiatric Readmission with Machine Learning in Children and Adolescents

> **NIH NIH F31** · CINCINNATI CHILDRENS HOSP MED CTR · 2022 · $39,993

## Abstract

Project Summary/Abstract
Mental health disorders, including anxiety and depression, are common in pediatric patients and significantly
impair behavioral function and quality of life. For those with severe illness, patients may be hospitalized for more
targeted treatment. Despite medication and/or therapy treatment, children and adolescents are frequently
readmitted into psychiatric care as a result of numerous reasons, including treatment ineffectiveness, medication
side effects, and issues with adhering to the treatment plan for the disorder. In fact, 25% of youth are readmitted
within one year of discharge. Additionally, treatment for these disorders can be long and costly to patients and
their families, especially if patients are hospitalized or re-hospitalized, with patients enduring multiple medication
trials before finding the best medication. In order to address these issues with pediatric psychiatric readmission,
this research is focused on the development of a machine learning algorithm to predict psychiatric readmission
in children and adolescents.
The first aim of the proposed research is to develop and establish machine learning algorithms to predict
psychiatric readmission within 30-, 90-, and 180-days of discharge in pediatric patients with anxiety and
depressive disorders using demographic, clinical, and pharmacogenetic data in the electronic health record.
Multiple algorithms will be evaluated to determine the best predictive model for each outcome. Important factors
influencing readmission and model performance for each outcome will be assessed and compared. Additionally,
this will be the first machine learning evaluation of psychiatric readmission in pediatric patients. The second aim
will assess the generalizability of our models using external pediatric psychiatric admission data from a
comparable institution. This validation is significant to ensure our model is applicable to new patients if this were
to be implemented clinically to improve patient care.
The exploratory third aim of this proposal will assess the ability of a model to select commonly prescribed
antidepressant medications that reduce readmission risk. The model will predict the risk of readmission if a
patient had been prescribed each antidepressant, which will be compared to current prescribing practices. This
will evaluate the impact of antidepressants on future psychiatric readmission, which could aid in medication
selection.
This project will be the first to evaluate psychiatric readmission in children and adolescents through a machine
learning approach, with the goal to reduce psychiatric readmission, thereby improving patient care and quality
of life. Further, this research will lay the foundation for future studies evaluating additional data modalities and
outcomes as we move towards more personalized treatments and recommendations for pediatric patients with
mental health disorders.

## Key facts

- **NIH application ID:** 10604849
- **Project number:** 1F31MH132265-01
- **Recipient organization:** CINCINNATI CHILDRENS HOSP MED CTR
- **Principal Investigator:** Ethan Andrew Poweleit
- **Activity code:** F31 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2022
- **Award amount:** $39,993
- **Award type:** 1
- **Project period:** 2022-09-19 → 2024-09-18

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10604849

## Citation

> US National Institutes of Health, RePORTER application 10604849, Predicting Psychiatric Readmission with Machine Learning in Children and Adolescents (1F31MH132265-01). Retrieved via AI Analytics 2026-05-24 from https://api.ai-analytics.org/grant/nih/10604849. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*