Ability of clinical data to predict readmission in Child and Adolescent Mental Health Services DOI Creative Commons
Kaban Koochakpour,

Dipendra Pant,

Odd Sverre Westbye

et al.

PeerJ Computer Science, Journal Year: 2024, Volume and Issue: 10, P. e2367 - e2367

Published: Oct. 18, 2024

This study addresses the challenge of predicting readmissions in Child and Adolescent Mental Health Services (CAMHS) by analyzing predictability over short, medium, long term periods. Using health records spanning 35 years, which included 22,643 patients 30,938 episodes care, we focused on episode care as a central unit, defined referral-discharge cycle that incorporates assessments interventions. Data pre-processing involved handling missing values, normalizing, transforming data, while resolving issues related to overlapping correcting registration errors where possible. Readmission prediction was inferred from electronic (EHR), this variable not directly recorded. A binary classifier distinguished between readmitted non-readmitted patients, followed multi-class categorize based timeframes: short (within 6 months), medium (6 months - 2 years), (more than years). Several predictive models were evaluated metrics like AUC, F1-score, precision, recall, K-prototype algorithm employed explore similarities through clustering. The optimal (Oversampled Gradient Boosting) achieved an AUC 0.7005, Random Forest) reached 0.6368. resulted three clusters (SI: 0.256, CI: 4473.64). Despite identifying relationships intensity, case complexity, readmission risk, generalizing these findings proved difficult, partly because clinicians often avoid discharging likely be readmitted. Overall, dataset offers insights into patient service patterns, remains challenging, suggesting need for improved analytical consider development, disease progression, intervention effects.

Language: Английский

Ability of clinical data to predict readmission in Child and Adolescent Mental Health Services DOI Creative Commons
Kaban Koochakpour,

Dipendra Pant,

Odd Sverre Westbye

et al.

PeerJ Computer Science, Journal Year: 2024, Volume and Issue: 10, P. e2367 - e2367

Published: Oct. 18, 2024

This study addresses the challenge of predicting readmissions in Child and Adolescent Mental Health Services (CAMHS) by analyzing predictability over short, medium, long term periods. Using health records spanning 35 years, which included 22,643 patients 30,938 episodes care, we focused on episode care as a central unit, defined referral-discharge cycle that incorporates assessments interventions. Data pre-processing involved handling missing values, normalizing, transforming data, while resolving issues related to overlapping correcting registration errors where possible. Readmission prediction was inferred from electronic (EHR), this variable not directly recorded. A binary classifier distinguished between readmitted non-readmitted patients, followed multi-class categorize based timeframes: short (within 6 months), medium (6 months - 2 years), (more than years). Several predictive models were evaluated metrics like AUC, F1-score, precision, recall, K-prototype algorithm employed explore similarities through clustering. The optimal (Oversampled Gradient Boosting) achieved an AUC 0.7005, Random Forest) reached 0.6368. resulted three clusters (SI: 0.256, CI: 4473.64). Despite identifying relationships intensity, case complexity, readmission risk, generalizing these findings proved difficult, partly because clinicians often avoid discharging likely be readmitted. Overall, dataset offers insights into patient service patterns, remains challenging, suggesting need for improved analytical consider development, disease progression, intervention effects.

Language: Английский

Citations

0