Computer assisted clinical decision support tool for management of statins

描述(申请人提供)：高胆固醇血症(特别是低密度脂蛋白胆固醇(LDL-C))是动脉粥样硬化性心血管疾病(ASCVD)的一个主要的、可改变的危险因素，ASCVD是美国的主要死亡原因。今天，据估计，美国有4100万人患有高胆固醇血症，这4100万人中有75%的人服用七种他汀类药物中的一种，这七种药物在降低升高的低密度脂蛋白和心血管发病率方面非常有效。然而，近55%接受他汀类药物治疗的患者在第一年的治疗中没有达到目标低密度脂蛋白-C水平，导致了可预防的死亡和不必要的医疗费用。实现目标低密度脂蛋白-C水平的最重要障碍是无法为从大型循证数据集合成的优化他汀类药物治疗提供实时建议。在缺乏这种决策支持的情况下，临床医生必须任意选择他汀类药物，并在较长时间内滴定剂量，从而产生可预防的成本。在退伍军人医院的初步研究表明，他汀类药物管理器(SM)是一种正在申请专利的计算机化、基于电子健康记录(EHR)的算法，可以高精度地预测达到目标低密度脂蛋白水平的可能性。他汀类药物管理器使用基于个体患者特征(包括伴随的临床条件和药物)的多变量Logistic回归模型来预测特定剂量的特定他汀类药物将达到目标低密度脂蛋白水平的可能性。SM确保在治疗方案开始时以正确的剂量为每个患者开出正确的他汀类药物。他汀类药物管理算法的进一步开发、扩展和商业化被设想为减少达到目标低密度脂蛋白胆固醇水平的高昂成本、延长的时间和频繁的挫折，潜在地减少副作用，改善治疗依从性，并最终降低与高低密度脂蛋白胆固醇相关的ASCVD的风险。仅在美国，与改善ASCVD结果的医疗保健相关的经济节省估计就有数千万美元或数亿美元。这项第一阶段研究的第一个目的是使用 对退伍军人管理局地区医疗网络中约201,000名服用他汀类药物的患者进行抽样调查，以确认SM通过选择最有效的他汀类药物和剂量来达到目标低密度脂蛋白水平，预测LCL-C目标实现的准确性(可靠性)。我们还将探索算法的扩展，以包括与他汀类药物相关的和可能影响最佳他汀类药物和剂量选择的紧急不良事件。第二个目标是使用退伍军人管理局国家公司中所有接受他汀类药物治疗的患者(约500万人)的数据来确定SM的内部(预测)有效性 数据仓库。我们将比较在大范围的他汀类药物处方和剂量下获得的低密度脂蛋白胆固醇水平与SM预测的水平。在第一阶段完成后，SM将在两个大型回溯性EHR研究中得到进一步验证，从而将SM定位为第二阶段的前瞻性外部验证研究。最终，SM将设计为满足主要集成医疗系统的要求，作为其EHR系统范围内的嵌入式应用程序。

DESCRIPTION (provided by applicant): Hypercholesterolemia (particularly low-density lipoprotein-cholesterol (LDL-C)) is a major, modifiable risk factor for atherosclerotic cardiovascular disease (ASCVD), the primary cause of death in the US. Today, an estimated 41 million people in the US are hypercholesterolemia and 75% of these 41 million people take one of seven statin drugs that are remarkably effective in reducing elevated LDL-C and cardiovascular morbidity. However, nearly 55% of statin-treated patients do not achieve target LDL-C levels during the first year of treatment, resulting in preventable mortality and unnecessary health care costs. The most important barrier to achieving target LDL-C levels is the inability to deliver real-time recommendations for optimized statin treatment synthesized from large, evidence-based datasets. In the absence of such decision support, clinicians must choose statins arbitrarily and titrate doses over a prolonged period, generating preventable costs. Preliminary research in a VA hospital setting indicates that Statin Manager (SM), a patent-pending computerized, electronic health care record (EHR)-based algorithm can predict with high accuracy the probability of achieving target LDL-C levels. Using multivariate logistic regression models based on individual patient characteristics, including concomitant clinical conditions and medications, Statin Manager predicts the probability that target LDL-C levels will be achieved by specific statins at specific doses. SM ensures that the right statin, in the right dosage, is prescribed for each patient at the beginning of the treatment regimen. Further development, extension, and commercialization of the statin management algorithm is envisioned to reduce the high cost, extended time and frequent frustration of experimentation to achieve target LDL-C levels, potentially reduce side effects, improve treatment adherence and ultimately reduce the resultant risk of ASCVD associated with elevated LDL-C. The economic savings associated with improved healthcare for ASCVD outcomes is estimated in the tens or hundreds of millions of dollars annually in the US alone. The first aim of this Phase I study uses a sample of ~201,000 statin-treated patients in a regional VA healthcare network to confirm the precision (reliability) of SM in predicting achievement of LCL-C goal by selecting the most efficacious statin and dose to achieve targeted LDL-C levels. We will also explore extension of the algorithm to include statin-related and emergent adverse events potentially impacting optimal statin and dose selection. The second aim is to determine the internal (predictive) validity of SM using data from all statin- treated patients (~5,000,000) in VA's national Corporate Data Warehouse. We will compare LDL-C levels achieved over a broad range of prescribed statins and doses with those predicted by SM. Upon completion of Phase I, SM will have been further validated in two large retrospective EHR studies, thus positioning SM for a prospective, external validation study in Phase II. Ultimately, SM will be designed to meet the requirements of major integrated healthcare systems for inclusion as an embedded application in their EHR system-wide.