Tsai Wan-Ju, Qian Tian-Yang, Lu Chun-Mei, Liu Qing, Wang Lai-Shuan
Department of Neonatology, National Children's Medical Center/Children's Hospital of Fudan University, Shanghai, China.
Transl Pediatr. 2021 Feb;10(2):256-264. doi: 10.21037/tp-20-184.
To construct and externally validate a prediction model for neonate unplanned rehospitalization within 31 days of discharge.
A retrospective study was performed in the Department of Neonatology of the Children's Hospital of Fudan University. A binominal regression method was applied to construct and validate the prediction model. Analysis was performed on a total of 11,116 neonates with an index admission between 11/1/2016 and 12/31/2018. Neonates admitted from 11/1/2016 to 1/31/2018 were used for the selection of prognostic variables and construction of the model. Model validation was then performed with neonates admitted from 2/1/2018 to 12/31/2018.
The rehospitalization rate for neonates was 3.27% (373/11,116). A total of 512 neonates were enrolled for the construction of the prediction model. Gestational age (GA), NICU length of stay (LOS), nonmedical order discharge and younger maternal age were strongly correlated with rehospitalization. By incorporating these 4 strong risk factors, we constructed a model to predict neonate unplanned rehospitalization within 31 days of discharge. The formula was turned into a nomogram for use in clinical practice. The nomogram has a total score of 180, with a predicted risk from 0 to 100%. Neonates are at high risk for rehospitalization if they have a total score greater than 39 points, according to the cutoff point established by the Youden index. The model was shown to have good discriminatory ability, with area under the receiver operating characteristic curves of 0.68 and 0.65 in the model construction and validation datasets, respectively. A total of 39 points is the cutoff for follow-up.
The model is able to predict neonate unplanned rehospitalization well. A total score greater than 39 indicates that follow-up is necessary.
构建并外部验证一个预测模型，用于预测新生儿出院后31天内的非计划再入院情况。
在复旦大学附属儿科医院新生儿科进行一项回顾性研究。应用二项回归方法构建并验证该预测模型。对2016年11月1日至2018年12月31日期间首次入院的11116例新生儿进行分析。将2016年11月1日至2018年1月31日入院的新生儿用于选择预后变量并构建模型。然后用2018年2月1日至2018年12月31日入院的新生儿进行模型验证。
新生儿再入院率为3.27%（373/11116）。共纳入512例新生儿用于构建预测模型。胎龄（GA）、新生儿重症监护病房住院时间（LOS）、非医嘱出院和母亲年龄较小与再入院密切相关。通过纳入这4个强风险因素，构建了一个预测模型，用于预测新生儿出院后31天内的非计划再入院情况。该公式转化为列线图用于临床实践。列线图总分180分，预测风险为0至100%。根据约登指数确定的截断点，总分大于39分的新生儿再入院风险较高。该模型显示出良好的鉴别能力，在模型构建和验证数据集中，受试者操作特征曲线下面积分别为0.68和0.65。随访的截断值为39分。
该模型能够较好地预测新生儿非计划再入院情况。总分大于39分表明有必要进行随访。
