Around five percent of the population is affected by a rare disease, most often due to genetic variation. A genetic test is the quickest path to a diagnosis, yet most suffer through years of diagnostic odyssey before getting a test, if they receive one at all. Identifying patients that are likely to have a genetic disease and therefore need genetic testing is paramount to improving diagnosis and treatment. While there are thousands of previously described genetic diseases with specific phenotypic presentations, a common feature among them is the presence of multiple rare phenotypes which often span organ systems. Here, we hypothesize that these patients can be identified from longitudinal clinical data in the electronic health record (EHR). We used diagnostic information from the EHRs of 2,286 patients that received a chromosomal microarray and 9,144 matched controls to train and test a prediction model. We identified high prediction accuracy (AUROC = 0.97, AUPR = 0.92) in a held-out test sample and in 172,265 hospital patients where cases were defined broadly as interacting with a genetics provider (AUROC = 0.9, AUPR = 0.63). High probabilities (median = 0.97) were associated with 46 patients carrying a known pathogenic copy number variant (CNV) among a subset of 6,445 genotyped patients. Our model identified many more patients needing a genetic test while increasing the proportion having a putative genetic disease compared to the current nonsytematic approach. Taken together, we demonstrate that phenotypic patterns representative of a genetic disease can be captured from EHR data and provide an opportunity to systematize decision making on genetic testing to speed up diagnosis, improve care, and reduce costs.

中文翻译：

---

临床数据中的表型特征可对患者进行系统鉴定以进行基因检测

约有5％的人口患有罕见病，最常见的原因是遗传变异。基因测试是诊断的最快途径，但是大多数人在接受测试之前要经历数年的诊断性征兆，如果他们接受的话。识别可能患有遗传疾病并因此需要进行基因检测的患者，对于改善诊断和治疗至关重要。尽管有成千上万的先前描述的具有特定表型表现的遗传疾病，但它们之间的共同特征是存在多种稀有表型，这些表型通常跨越器官系统。在这里，我们假设可以从电子健康记录（EHR）中的纵向临床数据中识别出这些患者。我们使用了来自2286例接受染色体微阵列和9例患者的EHR的诊断信息 144个匹配的控件，用于训练和测试预测模型。我们在保留的测试样本中和172265名住院患者中确定了较高的预测准确性（AUROC = 0.97，AUPR = 0.92），这些患者的病例被广泛定义为与遗传学提供者互动（AUROC = 0.9，AUPR = 0.63）。在6,445个基因分型患者中，有46位患者携带已知病原体拷贝数变异（CNV），相关性较高（中位数= 0.97）。与目前的非定型方法相比，我们的模型确定了更多需要进行基因检测的患者，同时增加了可能患有遗传病的比例。综上所述，我们证明可以从EHR数据中捕获代表遗传疾病的表型，并为系统化基因检测决策提供机会，从而加快诊断速度，