Harnessing the power of CTSA-CDRN data networks: Using social determinants of health, frailty and functional status to identify at-risk patients and improve risk adjustment

利用 CTSA-CDRN 数据网络的力量：利用健康、虚弱和功能状态的社会决定因素来识别高危患者并改善风险调整

• 批准号: 10199784
• 负责人: PAULA K SHIREMAN
• 金额: $ 75万
• 依托单位: UNIVERSITY OF TEXAS HLTH SCIENCE CENTER
• 依托单位国家: 美国
• 财政年份: 2018
• 资助国家: 美国
• 起止时间: 2018-07-25 至 2024-06-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10199784
• 关键词: Accounting; Agreement; Ambulatory Care Facilities; American; American College of Surgeons; Authorization documentation; Caring; Clinical Data; Clinical and Translational Science Awards; Community Surveys; Costs and Benefits; Data; Data Set; Data Sources; Discipline of Nursing; Education; Educational process of instructing; Electronic Health Record; Ethnic Origin; Fibrinogen; Future; Goals; Health; Hispanics; Hospitalization; Hospitals; Individual; Inpatients; Institution; Institutional Review Boards; Intervention; Link; Measures; Methods; Minority; Modeling; Natural Language Processing; Occupational Therapy; Operative Surgical Procedures; Outcome; Pathway interactions; Patient-Focused Outcomes; Patients; Physical therapy; Postoperative Complications; Postoperative Period; Poverty; Predictive Analytics; Procedures; Process; Protocols documentation; Quality of Care; Race; Recovery of Function; Registries; Reporting; Research; Research Institute; Resources; Risk; Risk Adjustment; Risk Factors; Site; Socioeconomic Status; Testing; Text; United States Centers for Medicare and Medicaid Services; Vulnerable Populations; base; clinical care; cost effective; data access; design; electronic data; financial incentive; frailty; functional status; health care disparity; high risk; hospital readmission; improved; incentive program; innovation; low socioeconomic status; machine learning algorithm; member; mortality; outcome prediction; payment; postoperative recovery; predictive modeling; programs; readmission rates; safety net; simulation; social; social health determinants; tool; usability

Postoperative complications and readmissions rates are higher in minority and low socioeconomic status (SES) patients. Low SES is associated with frailty, one of the best predictors of 30-day postoperative complications and early hospital readmission. Despite their influence on health outcomes, frailty and social risk factors are not considered in risk adjustment for reimbursement and quality measures. CMS developed financial incentive- based programs to improve quality of care. Yet this strategy disproportionately penalizes minority-serving, major teaching and safety net hospitals (SNH), further constraining resources for the care of vulnerable populations. Our long-term goal is to use frailty and social risk factors to identify at-risk patients to design more effective clinical care pathways. Frailty can be derived retrospectively using the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) dataset. Data networks are powerful research tools that can be used to answer important questions. However, extracting data from EHR is challenging. The Patient-Centered Outcomes Research Institute (PCORI) developed 13 Clinical Data Research Networks (CDRN) that have considerable overlapping membership with Clinical Translational Science Award (CTSA) institutions. While steady progress has been made, multiple barriers exist to efficiently access and use data. We will engage 3 CTSA hubs, each members of a different CDRN, to locally merge identified datasets developing data accessing and linking strategies at diverse institutions for dissemination across sites within CDRNs and to ultimately perform similar studies across CDRNs. We will use the SMART IRB reliance platform to harmonize the regulatory approval process as much as possible for each step of this project to identify barriers to use in data networks. We propose the following Aims: 1) Determine the predictive power of ethnicity, race, SES, and frailty for postoperative complications, mortality and readmissions to improve risk adjustment at 3 CTSA/CDRNs 2) Estimate postoperative functional status using natural language processing (NLP) and machine learning algorithms on inpatient physical therapy (PT), occupational therapy (OT) and nursing notes for ACS NSQIP patients to predict long-term functional status 3) Develop methods to predict long-term loss of independence after major surgery 4) Determine hospital resource utilization stratified by SES, frailty and minority status The significance of our study is the incorporation of social risk factors, frailty and functional status in risk adjustment forming the basis for future interventions by targeting patients at the highest risk for postoperative complications and reducing health care disparities. Our innovative approach harnesses data sources at diverse institutions with the goal of disseminating these methods across 3 CDRNs and the CTSA network.

术后并发症和再入院率在少数民族和低社会经济地位（SES）中较高 患者低SES与虚弱相关，这是术后30天并发症的最佳预测因素之一 和早期再入院尽管虚弱和社会风险因素对健康结果有影响， 在报销和质量措施的风险调整中考虑。CMS制定了财政激励措施- 以提高护理质量为基础的计划。然而，这一战略不成比例地惩罚少数民族服务，主要 教育和安全网医院（SNH），进一步限制了用于照顾弱势群体的资源。 我们的长期目标是利用虚弱和社会风险因素来识别高危患者，以设计更多 有效的临床护理路径。虚弱可以通过美国外科医生学会的回顾性研究得出 国家外科质量改进计划（ACS NSQIP）数据集。 数据网络是强大的研究工具，可用于回答重要问题。然而，提取 EHR的数据具有挑战性。以患者为中心的结果研究所（PCORI）开发了13个 临床数据研究网络（CDRN），与临床数据研究网络（CDRN）有相当多的重叠成员 翻译科学奖（CTSA）机构。虽然取得了稳步进展，但存在多重障碍 有效地访问和使用数据。我们将聘请3个CTSA中心，每个中心都是不同CDRN的成员， 合并已确定的数据集，在不同机构制定数据访问和链接战略， 在CDRN内跨研究中心传播，并最终在CDRN间进行类似研究。我们将使用 SMART IRB依赖平台，以尽可能协调每个机构的监管审批流程 该项目的步骤，以确定在数据网络中使用的障碍。我们提出以下目标： 1)确定种族、人种、SES和虚弱对术后并发症、死亡率的预测能力 和再入院，以改善3个CTSA/CDRN的风险调整 2)使用自然语言处理（NLP）和机器学习评估术后功能状态 ACS NSQIP的住院物理治疗（PT）、作业治疗（OT）和护理记录算法 患者预测长期功能状态 3)开发预测大手术后长期独立性丧失的方法 4)确定按社会经济地位、脆弱性和少数民族地位分层的医院资源利用率 我们研究的意义在于将社会风险因素、虚弱和功能状态纳入风险中 通过针对术后风险最高的患者进行调整，为未来的干预措施奠定基础。 并发症和减少保健差距。我们的创新方法利用各种数据源 这些方法的目标是在3个CDRN和CTSA网络中传播。

项目成果

Predicting health-related social needs in Medicaid and Medicare populations using machine learning.

• DOI: 10.1038/s41598-022-08344-4
• 发表时间: 2022-03-16
• 期刊: Scientific reports
• 影响因子: 4.6
• 作者: Holcomb J;Oliveira LC;Highfield L;Hwang KO;Giancardo L;Bernstam EV
• 通讯作者: Bernstam EV

Epidemiology of age-, sex-, and race-specific hospitalizations for abdominal aortic aneurysms highlights gaps in current screening recommendations.

• DOI: 10.1016/j.jvs.2022.02.058
• 发表时间: 2022-11
• 期刊: JOURNAL OF VASCULAR SURGERY
• 影响因子: 4.3
• 作者: Li, Shimena R.;Reitz, Katherine M.;Kennedy, Jason;Gabriel, Lucine;Phillips, Amanda R.;Shireman, Paula K.;Eslami, Mohammad H.;Tzeng, Edith
• 通讯作者: Tzeng, Edith

Chasm Between Cancer Quality Measures and Electronic Health Record Data Quality.

• DOI: 10.1200/cci.21.00128
• 发表时间: 2022-01
• 期刊: JCO CLINICAL CANCER INFORMATICS
• 影响因子: 4.2
• 作者: Schorer, Anna E.;Moldwin, Richard;Koskimaki, Jacob;Bernstam, Elmer, V;Venepalli, Neeta K.;Miller, Robert S.;Chen, James L.
• 通讯作者: Chen, James L.

Improving Pharmacovigilance Signal Detection from Clinical Notes with Locality Sensitive Neural Concept Embeddings.

利用局部敏感神经概念嵌入改进临床记录中的药物警戒信号检测。

• 发表时间: 2022
• 期刊: AMIA Joint Summits on Translational Science proceedings. AMIA Joint Summits on Translational Science
• 作者: Mower,Justin;Bernstam,Elmer;Xu,Hua;Myneni,Sahiti;Subramanian,Devika;Cohen,Trevor
• 通讯作者: Cohen,Trevor

Automatic classification of scanned electronic health record documents.

• DOI: 10.1016/j.ijmedinf.2020.104302
• 发表时间: 2020-12
• 期刊: International journal of medical informatics
• 影响因子: 4.9
• 作者: Goodrum H;Roberts K;Bernstam EV
• 通讯作者: Bernstam EV