Viterbi Family Department of Ophthalmology, Hamilton Glaucoma Center and Shiley Eye Institute, University of California, San Diego, La Jolla, California, USA; UCSD Health Department of Biomedical Informatics, University of California, San Diego, La Jolla, California, USA.
UCSD Health Department of Biomedical Informatics, University of California, San Diego, La Jolla, California, USA; Interdisciplinary Research on Substance Use Joint Doctoral Program, University of California, San Diego and San Diego State University, San Diego, California, USA.
Am J Ophthalmol. 2019 Dec;208:30-40. doi: 10.1016/j.ajo.2019.07.005. Epub 2019 Jul 16.
To predict the need for surgical intervention in patients with primary open-angle glaucoma (POAG) using systemic data in electronic health records (EHRs).
Development and evaluation of machine learning models.
Structured EHR data of 385 POAG patients from a single academic institution were incorporated into models using multivariable logistic regression, random forests, and artificial neural networks. Leave-one-out cross-validation was performed. Mean area under the receiver operating characteristic curve (AUC), sensitivity, specificity, accuracy, and Youden index were calculated for each model to evaluate performance. Systemic variables driving predictions were identified and interpreted.
Multivariable logistic regression was most effective at discriminating patients with progressive disease requiring surgery, with an AUC of 0.67. Higher mean systolic blood pressure was associated with significantly increased odds of needing glaucoma surgery (odds ratio [OR] = 1.09, P < .001). Ophthalmic medications (OR = 0.28, P < .001), non-opioid analgesic medications (OR = 0.21, P = .002), anti-hyperlipidemic medications (OR = 0.39, P = .004), macrolide antibiotics (OR = 0.40, P = .03), and calcium blockers (OR = 0.43, P = .03) were associated with decreased odds of needing glaucoma surgery.
Existing systemic data in the EHR has some predictive value in identifying POAG patients at risk of progression to surgical intervention, even in the absence of eye-specific data. Blood pressure-related metrics and certain medication classes emerged as predictors of glaucoma progression. This approach provides an opportunity for future development of automated risk prediction within the EHR based on systemic data to assist with clinical decision-making.
利用电子健康记录（EHR）中的系统数据预测原发性开角型青光眼（POAG）患者是否需要手术干预。
机器学习模型的开发和评估。
将来自单一学术机构的 385 名 POAG 患者的结构化 EHR 数据纳入模型中，使用多变量逻辑回归、随机森林和人工神经网络。进行了留一法交叉验证。计算每个模型的接收者操作特征曲线下的平均面积（AUC）、灵敏度、特异性、准确性和 Youden 指数，以评估性能。确定并解释了驱动预测的系统变量。
多变量逻辑回归在区分需要手术治疗的进展性疾病患者方面最为有效，AUC 为 0.67。较高的平均收缩压与需要青光眼手术的几率显著增加相关（比值比 [OR] = 1.09，P <.001）。眼科药物（OR = 0.28，P <.001）、非阿片类镇痛药（OR = 0.21，P =.002）、抗高血脂药物（OR = 0.39，P =.004）、大环内酯类抗生素（OR = 0.40，P =.03）和钙通道阻滞剂（OR = 0.43，P =.03）与需要青光眼手术的几率降低相关。
即使没有眼部特定数据，EHR 中的现有系统数据在识别有进展为手术干预风险的 POAG 患者方面具有一定的预测价值。血压相关指标和某些药物类别是青光眼进展的预测因素。这种方法为基于系统数据在 EHR 中开发自动风险预测提供了机会，以帮助临床决策。
