Center for Urban Population Health, University of Wisconsin School of Medicine and Public Health, Madison, USA.
Am J Crit Care. 2011 Nov;20(6):e122-30. doi: 10.4037/ajcc2011775.
Long-term care hospitals are Medicare providers of postacute care that have a mean length of stay of 25 days or more. Early identification and timely transfer of patients requiring mechanical ventilation to such hospitals may improve the efficiency of inpatient care.
To develop a predictive model and a simplified score for use on day 7 of hospitalization to assess whether a patient receiving mechanical ventilation is likely to require an additional 25 days of hospitalization (ie, would qualify for transfer to a long-term care hospital).
A retrospective, cross-sectional study using hospital discharge and billing data from the 2005 Nationwide Inpatient Sample for 54 686 Medicare beneficiaries admitted to US community hospitals who met the study's eligibility criteria. The outcome was overall length of stay (≥32 vs <32 days). Split-sample validation was used. Multivariable survey-logistic regression analyses were performed to assess predictors and probability of the outcome. A simplified score was derived from the final predictive model.
The discriminatory power of the predictive model was 0.75 and that of the simplified score was 0.72. The model calibrated well. All predictors were significantly (P < .01) associated with a hospitalization of 32 days or longer; having a tracheostomy was the strongest predictor (odds ratio, 4.74). The simplified scores ranged from -5 to 110 points and were categorized into 3 classes of risk.
Efforts to aid discharge decision making and optimize hospital resource planning could take advantage of our predictive model and the simplified scoring tool.
长期护理医院是医疗保险提供的急性后期护理机构，患者平均住院时间为 25 天或以上。早期识别和及时将需要机械通气的患者转移到这类医院，可能会提高住院护理的效率。
开发一种预测模型和简化评分，用于患者住院第 7 天，评估接受机械通气的患者是否需要再住院 25 天（即符合转入长期护理医院的条件）。
采用回顾性、横断面研究，使用 2005 年全国住院患者样本中的住院和计费数据，纳入符合研究条件的 54686 名接受机械通气的 Medicare 受益人的美国社区医院住院患者。结局为总住院时间（≥32 天与<32 天）。采用样本拆分验证。采用多变量调查逻辑回归分析评估预测因素和结局的可能性。从最终预测模型中得出简化评分。
预测模型的判别能力为 0.75，简化评分的判别能力为 0.72。模型校准良好。所有预测因素与住院 32 天或以上均显著相关（P<.01）；气管造口术是最强的预测因素（比值比，4.74）。简化评分范围为-5 至 110 分，分为 3 个风险等级。
为了帮助做出出院决策和优化医院资源规划，可以利用我们的预测模型和简化评分工具。
