Scholarlink

Background <p>Psychiatric readmission within 30 days of discharge is an important indicator of healthcare quality. Despite the global proliferation of prediction models, no validated tool exists for Middle Eastern or Gulf Cooperation Council populations, where distinctive demographic characteristics and service configurations may limit the transportability of models developed elsewhere.</p> Methods <p>This retrospective cohort study analysed 14,994 psychiatric inpatient admissions from 9,263 unique patients at a tertiary psychiatric facility in the United Arab Emirates (2018–2025), using electronic health record data. The cohort was randomly split into development (70%) and validation (30%) samples. Missing marital status data (28.2%) were addressed using multiple imputation with chained equations (20 imputed datasets). Nonlinearity of count-based predictors was addressed using log-transformations. LASSO logistic regression with ten-fold cross-validation was used for predictor selection and model fitting, with results pooled across imputed datasets. A simplified integer-based clinical risk score was derived from categorised predictors. Model performance was benchmarked against recalibrated LACE and READMIT-subset indices.</p> Results <p>The 30-day readmission rate was 10.6% (<i>n</i> = 1,583). LASSO regression identified prior psychiatric admissions, schizoaffective disorder, Charlson Comorbidity Index, first-generation antipsychotic prescription, long-acting injectable antipsychotic prescription (protective), emergency admission, emergency department visits, and antidepressant prescription as the principal predictors. Marital status was not consistently retained after multiple imputation with married as the reference category, likely reflecting the large proportion of missing data and the characteristics of the unrecorded group. The model demonstrated modest discrimination (AUC 0.649, 95% CI 0.623–0.678 in validation), significantly outperforming both the LACE index (AUC 0.557, <i>p</i> &lt; 0.001) and READMIT-subset (AUC 0.604, <i>p</i> &lt; 0.001). Temporal validation (2018–2023 training, 2024–2025 testing) yielded stable performance (AUC 0.676). A three-tier risk score stratified patients into low (6.7%), medium (10.0%), and high (18.0%) readmission probability groups.</p> Conclusion <p>This study presents the first psychiatric readmission prediction model from the Middle East, developed using penalised regression and multiple imputation. The locally derived model outperformed imported indices, and the clinical risk score offers a practical tool for identifying high-risk patients who may benefit from intensified discharge planning and transitional care.</p> Clinical trial number <p>Not applicable.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Development and validation of a clinical prediction model and risk score for 30-day psychiatric readmission in the United Arab Emirates

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月12日

查看全文
  • Syed Ali Bokhari,
  • Syed Fahad Javaid

摘要

Background

Psychiatric readmission within 30 days of discharge is an important indicator of healthcare quality. Despite the global proliferation of prediction models, no validated tool exists for Middle Eastern or Gulf Cooperation Council populations, where distinctive demographic characteristics and service configurations may limit the transportability of models developed elsewhere.

Methods

This retrospective cohort study analysed 14,994 psychiatric inpatient admissions from 9,263 unique patients at a tertiary psychiatric facility in the United Arab Emirates (2018–2025), using electronic health record data. The cohort was randomly split into development (70%) and validation (30%) samples. Missing marital status data (28.2%) were addressed using multiple imputation with chained equations (20 imputed datasets). Nonlinearity of count-based predictors was addressed using log-transformations. LASSO logistic regression with ten-fold cross-validation was used for predictor selection and model fitting, with results pooled across imputed datasets. A simplified integer-based clinical risk score was derived from categorised predictors. Model performance was benchmarked against recalibrated LACE and READMIT-subset indices.

Results

The 30-day readmission rate was 10.6% (n = 1,583). LASSO regression identified prior psychiatric admissions, schizoaffective disorder, Charlson Comorbidity Index, first-generation antipsychotic prescription, long-acting injectable antipsychotic prescription (protective), emergency admission, emergency department visits, and antidepressant prescription as the principal predictors. Marital status was not consistently retained after multiple imputation with married as the reference category, likely reflecting the large proportion of missing data and the characteristics of the unrecorded group. The model demonstrated modest discrimination (AUC 0.649, 95% CI 0.623–0.678 in validation), significantly outperforming both the LACE index (AUC 0.557, p < 0.001) and READMIT-subset (AUC 0.604, p < 0.001). Temporal validation (2018–2023 training, 2024–2025 testing) yielded stable performance (AUC 0.676). A three-tier risk score stratified patients into low (6.7%), medium (10.0%), and high (18.0%) readmission probability groups.

Conclusion

This study presents the first psychiatric readmission prediction model from the Middle East, developed using penalised regression and multiple imputation. The locally derived model outperformed imported indices, and the clinical risk score offers a practical tool for identifying high-risk patients who may benefit from intensified discharge planning and transitional care.

Clinical trial number

Not applicable.