Detecting Miscoded Diabetes Diagnosis Codes in Electronic Health Records for Quality Improvement: Temporal Deep Learning Approach.
Detecting Miscoded Diabetes Diagnosis Codes in Electronic Health Records for Quality Improvement: Temporal Deep Learning Approach.
复制标题
检测电子健康记录中错误编码的糖尿病诊断代码以提高质量：时态深度学习方法。
DOI：
10.2196/22649
复制
发表时间：
2020-12-17
影响因子：
3.2
通讯作者：
Saltz M
中科院分区：
文献类型：
--
作者：
Rashidian S;Abell-Hart K;Hajagos J;Moffitt R;Lingam V;Garcia V;Tsai CW;Wang F;Dong X;Sun S;Deng J;Gupta R;Miller J;Saltz J;Saltz M
糖尿病影响着美国3000多万患者。由于疾病负担如此之大，即使是分类中的一个小错误也可能是重大的。目前，在就诊时分配的账单代码是反映个人实际疾病的“黄金标准”，因此总体上反映了人口中的疾病流行率。这些代码由训练有素的编码人员和医疗保健提供者生成，但并不总是准确的。这项工作提供了一种可扩展的深度学习方法，可以在多个医疗保健系统中更准确地对糖尿病患者进行分类。我们利用长短期记忆密集神经网络（LSTM-DNN）模型来识别糖尿病患者，使用来自5家急性护理机构的数据，其中包括187，187名患者和275，407次就诊，包括实验室检查结果，诊断/程序代码，药物，人口统计学数据和入院信息。此外，设盲的医生小组审查了不一致的病例，提供了对人群总影响的估计。在预测糖尿病的诊断记录时，我们的模型在来自5个不同卫生机构的异质数据集上实现了84%的F1评分，96%的曲线下面积-受试者工作特征曲线和91%的平均精度。然而，在81%的情况下，该模型不同意记录的表型，一个盲医生小组同意该模型。总的来说，这表明我们研究的人群中有4.3%的糖尿病诊断缺失或不正确。这项研究表明，即使患者数据是嘈杂的，稀疏的和异质的，深度学习方法也可以改善临床表型。
Diabetes affects more than 30 million patients across the United States. With such a large disease burden, even a small error in classification can be significant. Currently billing codes, assigned at the time of a medical encounter, are the “gold standard” reflecting the actual diseases present in an individual, and thus in aggregate reflect disease prevalence in the population. These codes are generated by highly trained coders and by health care providers but are not always accurate. This work provides a scalable deep learning methodology to more accurately classify individuals with diabetes across multiple health care systems. We leveraged a long short-term memory-dense neural network (LSTM-DNN) model to identify patients with or without diabetes using data from 5 acute care facilities with 187,187 patients and 275,407 encounters, incorporating data elements including laboratory test results, diagnostic/procedure codes, medications, demographic data, and admission information. Furthermore, a blinded physician panel reviewed discordant cases, providing an estimate of the total impact on the population. When predicting the documented diagnosis of diabetes, our model achieved an 84% F1 score, 96% area under the curve–receiver operating characteristic curve, and 91% average precision on a heterogeneous data set from 5 distinct health facilities. However, in 81% of cases where the model disagreed with the documented phenotype, a blinded physician panel agreed with the model. Taken together, this suggests that 4.3% of our studied population have either missing or improper diabetes diagnosis. This study demonstrates that deep learning methods can improve clinical phenotyping even when patient data are noisy, sparse, and heterogeneous.
登录
查看更多内容
影响因子：
4.5
作者：
Chen, You;Ghosh, Joydeep;Malim, Bradley
通讯作者：
Malim, Bradley
DOI：
10.1093/jamia/ocw112
发表时间：
2017-03-01
期刊：
Journal of the American Medical Informatics Association : JAMIA
影响因子：
--
作者：
Choi E;Schuetz A;Stewart WF;Sun J
通讯作者：
Sun J
DOI：
10.1146/annurev-biodatasci-080917-013315
发表时间：
2018-01-01
期刊：
ANNUAL REVIEW OF BIOMEDICAL DATA SCIENCE, VOL 1
影响因子：
--
作者：
Banda, Juan M.;Seneviratne, Martin;Shah, Nigam H.
通讯作者：
Shah, Nigam H.
DOI：
10.1073/pnas.1510502113
发表时间：
2016-07-05
影响因子：
11.1
作者：
Hripcsak, George;Ryan, Patrick B.;Madigan, David
通讯作者：
Madigan, David
DOI：
10.1136/amiajnl-2012-001145
发表时间：
2013-01-01
期刊：
Journal of the American Medical Informatics Association : JAMIA
影响因子：
--
作者：
Hripcsak G;Albers DJ
通讯作者：
Albers DJ