Varenicline and Nicotine Replacement Use Associated With US Food and Drug Administration Drug Safety Communications
Varenicline and Nicotine Replacement Use Associated With US Food and Drug Administration Drug Safety Communications
复制标题
DOI：
10.1001/jamanetworkopen.2019.18962
复制
发表时间：
2019-09-01
影响因子：
13.8
通讯作者：
Good, Chester B.
中科院分区：
文献类型：
--
作者：
Desai, Ravi J.;Good, Meghan M.;Good, Chester B.
要点问题基于常规收集的理赔数据对心力衰竭患者结果的预测是否可以通过机器学习方法并结合链接的电子病历来改进？结果在这项包含 9502 名患者记录的预后研究中，机器学习方法在根据管理要求预测心力衰竭关键结果方面仅比逻辑回归提供有限的改进。从电子病历中纳入额外的预测因素可以改善对死亡率、心力衰竭住院治疗和家庭损失的预测，但成本不高。意义基于仅索赔预测因子的模型可以在预测心力衰竭关键患者结局方面实现适度的区分和准确性，而机器学习方法和结合电子病历中的其他预测因子可能会在选定结果的风险预测方面提供一定的改进。重要性心力衰竭（HF）患者的准确风险分层对于部署旨在改善患者生活质量和结果的有针对性的干预措施至关重要。目标将机器学习方法与传统逻辑回归在预测心力衰竭患者关键结局方面进行比较，并评估利用电子病历 (EMR) 衍生信息增强基于索赔的预测模型的附加价值。设计、设置和参与者进行了一项为期 1 年随访的预后研究，包括来自马萨诸塞州波士顿 2 个医疗保健提供者网络（“提供者”包括医生、临床医生、其他医疗保健专业人员及其组成网络的机构）的 9502 名参加医疗保险的心力衰竭患者。该研究于2007年1月1日至2014年12月31日期间进行；对2018年1月1日至12月31日的数据进行分析。主要结果和措施使用逻辑回归、最小绝对收缩和选择操作回归、分类和回归树、随机森林和梯度增强建模（GBM）对全因死亡率、心力衰竭住院、最高成本十分位数和居家日损失大于25%进行建模。所有模型均使用网络 1 中的数据进行训练，并在网络 2 中进行测试。根据判别、Brier 评分和校准选择最有效的建模方法后，计算精确回忆曲线下面积 (AUPRC) 和决策曲线的净效益估计值，以重点关注使用仅索赔与索赔 + EMR 预测器时的差异。结果共纳入 9502 名平均 (SD) 年龄为 78 (8) 岁的心力衰竭患者：6113 名来自网络 1（训练集），3389 名来自网络 2（测试集）。梯度增强模型始终提供最高的辨别力、最低的 Brier 分数以及所有 4 个结果的良好校准；然而，逻辑回归具有大致相似的性能（基于仅索赔预测因素的逻辑回归的 C 统计：死亡率，0.724；95% CI，0.705-0.744；心衰住院，0.707；95% CI，0.676-0.737；高成本，0.734；95% CI，0.703-0.764；仅家庭日损失索赔， 0.781；95% CI，0.764-0.798；GBM 统计：死亡率，0.727；95% CI，0.708-0.747；HF 住院，0.745；高费用，0.733； 0.703-0.763；家庭日损失，0.790；95% CI，0.773-0.807）。与仅索赔的 GBM 相比，索赔+EMR 获得了更高的 AUPRC，预测死亡率（0.484 对 0.423）、心力衰竭住院（0.413 对 0.403）和家庭时间损失（0.575 对 0.521），但不预测成本（0.249 对 0.252）。在死亡率和家庭时间损失结果的各种阈值概率下，索赔+EMR 与仅索赔 GBM 的净收益较高，但其他 2 个结果类似。结论和相关性机器学习方法在预测关键心力衰竭结果方面比传统逻辑回归仅提供有限的改进。将 EMR 中的其他预测因子纳入基于索赔的模型似乎可以改善对某些（但不是全部）结果的预测。这项预后研究将几种机器学习方法与传统逻辑回归进行了比较，以开发针对心力衰竭患者的全因死亡率、心力衰竭住院、高费用和在家时间损失的预测模型。重要性 美国食品和药物管理局 (FDA) 发布的药物安全通报通常基于批准后有关安全信号的有限证据。 Varenicline 可以作为相关案例研究，因为它是 2008 年和 2009 年 FDA 多次沟通的目标；最终，2016 年 12 月 16 日全球戒烟研究中的不良事件评估 (EAGLES) 驳回了对自杀念头增加以及攻击性和不稳定行为的安全担忧。 目的 研究 FDA 药物安全沟通与伐尼克兰使用之间的关联。 设计、设置和参与者 对退伍军人健康管理局 (VHA) 10 月份门诊数据进行回顾性、纵向、横断面研究2001 年 1 月 1 日至 2018 年 12 月 31 日，以及 2006 年 7 月 1 日至 2018 年 9 月 30 日期间关于伐尼克兰处方的医疗补助药物州使用数据。 主要结果和措施 提取了 VHA 中伐尼克兰和尼古丁替代疗法 (NRT) 的处方记录，以及 VHA 中独特的伐尼克兰和 NRT 用户数量测量了四分之一。进行了中断时间序列分析来描述 FDA 安全警告与伐尼克兰和 NRT 的使用之间的关联。为了测试研究结果的普遍性，我们使用 2006-2018 年医疗补助每季度报销的伐尼克兰处方数量进行了类似的分析。 结果 2007 年 1 月将伐尼克兰纳入 VHA 国家药物处方集后，伐尼克兰的使用量呈现稳步增长，在 2008 年第一季度达到 32 581 季度独立用户的峰值。在 2 月 1 日之后的 12 个月内， 2008 年，公共卫生咨询，VHA 患者的季度伐尼克兰使用量减少了 68.7%（从 32 581 名患者减少到 10 182 名患者；P
Key PointsQuestionCan prediction of patient outcomes in heart failure based on routinely collected claims data be improved with machine learning methods and incorporating linked electronic medical records? FindingsIn this prognostic study including records on 9502 patients, machine learning methods offered only limited improvement over logistic regression in predicting key outcomes in heart failure based on administrative claims. Inclusion of additional predictors from electronic medical records improved prediction for mortality, heart failure hospitalization, and loss in home days but not for high cost. MeaningModels based on claims-only predictors may achieve modest discrimination and accuracy in prediction of key patient outcomes in heart failure, and machine learning approaches and incorporation of additional predictors from electronic medical records may offer some improvement in risk prediction of select outcomes.ImportanceAccurate risk stratification of patients with heart failure (HF) is critical to deploy targeted interventions aimed at improving patients' quality of life and outcomes. ObjectivesTo compare machine learning approaches with traditional logistic regression in predicting key outcomes in patients with HF and evaluate the added value of augmenting claims-based predictive models with electronic medical record (EMR)-derived information. Design, Setting, and ParticipantsA prognostic study with a 1-year follow-up period was conducted including 9502 Medicare-enrolled patients with HF from 2 health care provider networks in Boston, Massachusetts ("providers" includes physicians, clinicians, other health care professionals, and their institutions that comprise the networks). The study was performed from January 1, 2007, to December 31, 2014; data were analyzed from January 1 to December 31, 2018. Main Outcomes and MeasuresAll-cause mortality, HF hospitalization, top cost decile, and home days loss greater than 25% were modeled using logistic regression, least absolute shrinkage and selection operation regression, classification and regression trees, random forests, and gradient-boosted modeling (GBM). All models were trained using data from network 1 and tested in network 2. After selecting the most efficient modeling approach based on discrimination, Brier score, and calibration, area under precision-recall curves (AUPRCs) and net benefit estimates from decision curves were calculated to focus on the differences when using claims-only vs claims+EMR predictors. ResultsA total of 9502 patients with HF with a mean (SD) age of 78 (8) years were included: 6113 from network 1 (training set) and 3389 from network 2 (testing set). Gradient-boosted modeling consistently provided the highest discrimination, lowest Brier scores, and good calibration across all 4 outcomes; however, logistic regression had generally similar performance (C statistics for logistic regression based on claims-only predictors: mortality, 0.724; 95% CI, 0.705-0.744; HF hospitalization, 0.707; 95% CI, 0.676-0.737; high cost, 0.734; 95% CI, 0.703-0.764; and home days loss claims only, 0.781; 95% CI, 0.764-0.798; C statistics for GBM: mortality, 0.727; 95% CI, 0.708-0.747; HF hospitalization, 0.745; 95% CI, 0.718-0.772; high cost, 0.733; 95% CI, 0.703-0.763; and home days loss, 0.790; 95% CI, 0.773-0.807). Higher AUPRCs were obtained for claims+EMR vs claims-only GBMs predicting mortality (0.484 vs 0.423), HF hospitalization (0.413 vs 0.403), and home time loss (0.575 vs 0.521) but not cost (0.249 vs 0.252). The net benefit for claims+EMR vs claims-only GBMs was higher at various threshold probabilities for mortality and home time loss outcomes but similar for the other 2 outcomes. Conclusions and RelevanceMachine learning methods offered only limited improvement over traditional logistic regression in predicting key HF outcomes. Inclusion of additional predictors from EMRs to claims-based models appeared to improve prediction for some, but not all, outcomes.This prognostic study compares several machine learning approaches with traditional logistic regression for development of predictive models for all-cause mortality, heart failure hospitalization, high cost, and loss in home time, among patients with heart failure.IMPORTANCE Drug safety communications released by the US Food and Drug Administration (FDA) are often based on limited evidence on safety signals after approval. Varenicline may serve as a relevant case study because it was the target of several FDA communications in 2008 and 2009; ultimately, the Evaluating Adverse Events in a Global Smoking Cessation Study (EAGLES) dismissed safety concerns on increased suicidal thoughts and aggressive and erratic behavior on December 16, 2016.OBJECTIVE To examine the association between FDA drug safety communications and the use of varenicline.DESIGN, SETTING, AND PARTICIPANTS Retrospective, longitudinal, cross-sectional study of Veterans Health Administration (VHA) outpatient data from October 1, 2001, through December 31, 2018, and Medicaid drug state use data from July 1, 2006, through September 30, 2018, on varenicline prescribing.MAIN OUTCOMES AND MEASURES Prescribing records for varenicline and nicotine replacement therapy (NRT) in the VHA were extracted, and the number of unique varenicline and NRT users in the VHA per quarter was measured. An interrupted time series analysis was performed to describe the association between FDA safety warnings and the use of varenicline and NRT. To test the generalizability of the findings, similar analyses were conducted using the number of prescriptions reimbursed for varenicline by Medicaid every quarter in 2006-2018.RESULTS After its addition to the VHA national drug formulary in January 2007, varenicline use presented a steady increase, reaching a peak of 32 581 quarterly unique users in the first quarter of 2008. Within 12 months of the February 1, 2008, public health advisory, quarterly varenicline use in VHA patients decreased by 68.7%(from 32 581 to 10 182 patients; P