From the Viterbi Family Department of Ophthalmology and Shiley Eye Institute, University of California San Diego, (S.L.B., B.R.S.), La Jolla, California; UCSD Health Department of Biomedical Informatics, University of California San Diego, (S.L.B., B.R.S., P.P., J.K., L.B., T.-T.K., L.O.-M.), La Jolla, California.
From the Viterbi Family Department of Ophthalmology and Shiley Eye Institute, University of California San Diego, (S.L.B., B.R.S.), La Jolla, California; UCSD Health Department of Biomedical Informatics, University of California San Diego, (S.L.B., B.R.S., P.P., J.K., L.B., T.-T.K., L.O.-M.), La Jolla, California.
Am J Ophthalmol. 2021 Jul;227:74-86. doi: 10.1016/j.ajo.2021.01.008. Epub 2021 Jan 23.
To (1) use All of Us (AoU) data to validate a previously published single-center model predicting the need for surgery among individuals with glaucoma, (2) train new models using AoU data, and (3) share insights regarding this novel data source for ophthalmic research.
Development and evaluation of machine learning models.
Electronic health record data were extracted from AoU for 1,231 adults diagnosed with primary open-angle glaucoma. The single-center model was applied to AoU data for external validation. AoU data were then used to train new models for predicting the need for glaucoma surgery using multivariable logistic regression, artificial neural networks, and random forests. Five-fold cross-validation was performed. Model performance was evaluated based on area under the receiver operating characteristic curve (AUC), accuracy, precision, and recall.
The mean (standard deviation) age of the AoU cohort was 69.1 (10.5) years, with 57.3% women and 33.5% black, significantly exceeding representation in the single-center cohort (P = .04 and P < .001, respectively). Of 1,231 participants, 286 (23.2%) needed glaucoma surgery. When applying the single-center model to AoU data, accuracy was 0.69 and AUC was only 0.49. Using AoU data to train new models resulted in superior performance: AUCs ranged from 0.80 (logistic regression) to 0.99 (random forests).
Models trained with national AoU data achieved superior performance compared with using single-center data. Although AoU does not currently include ophthalmic imaging, it offers several strengths over similar big-data sources such as claims data. AoU is a promising new data source for ophthalmic research.
（1）利用 All of Us（AoU）数据验证先前发表的预测青光眼患者手术需求的单中心模型，（2）使用 AoU 数据训练新模型，（3）分享有关这种新型眼科研究数据源的见解。
机器学习模型的开发和评估。
从 AoU 中提取了 1231 名被诊断为原发性开角型青光眼的成年人的电子健康记录数据。将单中心模型应用于 AoU 数据进行外部验证。然后，使用 AoU 数据使用多变量逻辑回归、人工神经网络和随机森林来训练新的预测青光眼手术需求的模型。进行了五重交叉验证。基于接收者操作特征曲线下的面积（AUC）、准确性、精度和召回率来评估模型性能。
AoU 队列的平均（标准差）年龄为 69.1（10.5）岁，女性占 57.3%，黑人占 33.5%，显著超过单中心队列的代表性（P 分别为 =.04 和 P <.001）。在 1231 名参与者中，有 286 名（23.2%）需要进行青光眼手术。当将单中心模型应用于 AoU 数据时，准确性为 0.69，AUC 仅为 0.49。使用 AoU 数据训练新模型可获得更好的性能：AUC 范围从 0.80（逻辑回归）到 0.99（随机森林）。
使用全国 AoU 数据训练的模型与使用单中心数据相比表现更好。尽管 AoU 目前不包括眼科成像，但它提供了一些优于类似大数据源（如索赔数据）的优势。AoU 是眼科研究的一种很有前途的新数据源。
