Ren Boyu, Yoon WonJin, Thomas Spencer, Savova Guergana, Miller Timothy, Hall Mei-Hua
Laboratory for Psychiatric Biostatistics, McLean Hospital, Belmont, MA, United States.
Department of Psychiatry, Harvard Medical School, Boston, MA, United States.
JMIR Ment Health. 2025 Sep 12;12:e71630. doi: 10.2196/71630.
Patients with mood or psychotic disorders experience high rates of unplanned hospital readmissions. Predicting the likelihood of readmission can guide discharge decisions and optimize patient care.
The purpose of this study is to evaluate the predictive power of structured variables from electronic health records for all-cause readmission across multiple sites within the Mass General Brigham health system and to assess the transportability of prediction models between sites.
This retrospective, multisite study analyzed structured variables from electronic health records separately for each site to develop in-site prediction models. The transportability of these models was evaluated by applying them across different sites. Predictive performance was measured using the F1-score, and additional adjustments were made to account for differences in predictor distributions.
The study found that the relevant predictors of readmission varied significantly across sites. For instance, length of stay was a strong predictor at only 3 of the 4 sites. In-site prediction models achieved an average F1-score of 0.661, whereas cross-site predictions resulted in a lower average F1-score of 0.616. Efforts to improve transportability by adjusting for differences in predictor distributions did not improve performance.
The findings indicate that individual site-specific models are necessary to achieve reliable prediction accuracy. Furthermore, the results suggest that the current set of predictors may be insufficient for cross-site model transportability, highlighting the need for more advanced predictor variables and predictive algorithms to gain robust insights into the factors influencing early psychiatric readmissions.
患有情绪或精神障碍的患者意外再次入院率很高。预测再次入院的可能性可为出院决策提供指导并优化患者护理。
本研究旨在评估来自电子健康记录的结构化变量对麻省总医院布莱根健康系统内多个地点全因再次入院的预测能力，并评估预测模型在不同地点之间的可移植性。
这项回顾性多地点研究分别针对每个地点分析了电子健康记录中的结构化变量，以建立本地预测模型。通过将这些模型应用于不同地点来评估其可移植性。使用F1分数衡量预测性能，并进行了额外调整以考虑预测变量分布的差异。
研究发现，不同地点再次入院的相关预测因素差异显著。例如，住院时间仅在4个地点中的3个是强有力的预测因素。本地预测模型的平均F1分数为0.661，而跨地点预测的平均F1分数较低，为0.616。通过调整预测变量分布差异来提高可移植性的努力并未改善性能。
研究结果表明，需要建立针对各个地点的特定模型才能实现可靠的预测准确性。此外，结果表明当前的预测变量集可能不足以实现跨地点模型的可移植性，这凸显了需要更先进的预测变量和预测算法，以便深入了解影响早期精神科再次入院的因素。
