Incorporation of quantitative imaging data using artificial intelligence improves risk prediction in veterans with liver disease.

BACKGROUND AND AIMS: Utilization of electronic health records data to derive predictive indexes such as the electronic Child-Turcotte-Pugh (eCTP) Score can have significant utility in health care delivery. Within the records, CT scans contain phenotypic data which have significant prognostic value. However, data extractions have not traditionally been applied to imaging data. In this study, we used artificial intelligence to automate biomarker extraction from CT scans and examined the value of these features in improving risk prediction in patients with liver disease. APPROACH AND RESULTS: Using a regional liver disease cohort from the Veterans Health System, we retrieved administrative, laboratory, and clinical data for Veterans who had CT scans performed for any clinical indication between 2008 and 2014. Imaging biomarkers were automatically derived using the analytic morphomics platform. In all, 4614 patients were included. We found that the eCTP Score had a Concordance index of 0.64 for the prediction of overall mortality while the imaging-based model alone or with eCTP Score performed significantly better [Concordance index of 0.72 and 0.73 ( p <0.001)]. For the subset of patients without hepatic decompensation at baseline (n=4452), the Concordance index for predicting future decompensation was 0.67, 0.79, and 0.80 for eCTP Score, imaging alone, or combined, respectively. CONCLUSIONS: This proof of concept demonstrates that the potential of utilizing automated extraction of imaging features within CT scans either alone or in conjunction with classic health data can improve risk prediction in patients with chronic liver disease.

Engaging Operational Partners Is Critical for Successful Implementation of Research Products: a Coincidence Analysis of Access-Related Projects in the Veterans Affairs Healthcare System.

BACKGROUND/OBJECTIVE: The Veterans Health Administration (VHA) has prioritized timely access to care and has invested substantially in research aimed at optimizing veteran access. However, implementing research into practice remains challenging. Here, we assessed the implementation status of recent VHA access-related research projects and explored factors associated with successful implementation. DESIGN: We conducted a portfolio review of recent VHA-funded or supported projects (1/2015-7/2020) focused on healthcare access ("Access Portfolio"). We then identified projects with implementable research deliverables by excluding those that (1) were non-research/operational projects; (2) were only recently completed (i.e., completed on or after 1/1/2020, meaning that they were unlikely to have had time to be implemented); and (3) did not propose an implementable deliverable. An electronic survey assessed each project's implementation status and elicited barriers/facilitators to implementing deliverables. Results were analyzed using novel Coincidence Analysis (CNA) methods. PARTICIPANTS/KEY RESULTS: Among 286 Access Portfolio projects, 36 projects led by 32 investigators across 20 VHA facilities were included. Twenty-nine respondents completed the survey for 32 projects (response rate = 88.9%). Twenty-eight percent of projects reported fully implementing project deliverables, 34% reported partially implementing deliverables, and 37% reported not implementing any deliverables (i.e., resulting tool/intervention not implemented into practice). Of 14 possible barriers/facilitators assessed in the survey, two were identified through CNA as "difference-makers" to partial or full implementation of project deliverables: (1) engagement with national VHA operational leadership; (2) support and commitment from local site operational leadership. CONCLUSIONS: These findings empirically highlight the importance of operational leadership engagement for successful implementation of research deliverables. Efforts to strengthen communication and engagement between the research community and VHA local/national operational leaders should be expanded to ensure VHA's investment in research leads to meaningful improvements in veterans' care. The Veterans Health Administration (VHA) has prioritized timely access to care and has invested substantially in research aimed at optimizing veteran access. However, implementing research findings into clinical practice remains challenging, both within and outside VHA. Here, we assessed the implementation status of recent VHA access-related research projects and explored factors associated with successful implementation. Only two factors were identified as "difference-makers" to adoption of project findings into practice: (1) engagement with national VHA leadership or (2) support and commitment from local site leadership. These findings highlight the importance of leadership engagement for successful implementation of research findings. Efforts to strengthen communication and engagement between the research community and VHA local/national leaders should be expanded to ensure VHA's investment in research leads to meaningful improvements in veterans' care.

Engaging Stakeholders in Identifying Access Research Priorities for the Department of Veterans Affairs.

BACKGROUND: The Veterans Access Research Consortium (VARC), a Department of Veterans Affairs (VA) Consortium of Research focused on access to healthcare, has been funded by VA's Health Services Research and Development Service (HSR&D) to develop a research roadmap for healthcare access. The goal of the roadmap is to identify operationally aligned research questions that are most likely to lead to meaningful improvements in Veterans' healthcare access. OBJECTIVES: To describe the process of soliciting diverse stakeholder perspectives about key priorities on which VA's HSR&D access agenda should focus and identify the results of that process. METHODS: We used a modified Delphi approach to engage researchers and VA operational partners in a process to develop recommendations regarding the access-related research questions VA should prioritize. We then collaborated with three Veteran Engagement Groups (VEGs) across the country to solicit Veterans' reactions to the Delphi results and their perspectives about access-related issues affecting access to VA health care. RESULTS: The Delphi panel consisted of 22 research and operational experts, both internal and external to VA. The Delphi process resulted in five research questions identified by the panelists as highest priority for VA to pursue, each representing one of the following domains: (1) measurement of access, (2) barriers to access, (3) equity and subpopulations, (4) effective interventions to improve access, and (5) consequences of poor/better access. Veterans' perspectives focused primarily on the barriers to access domain. Veterans indicated several barriers that might be addressed through research or operational initiatives, including poor communication about services, weak connections to and partnerships with local community care facilities, and poor provision of telehealth resources and education. CONCLUSIONS: Engaging multiple methods to solicit stakeholder perspectives enables more nuanced understanding of access-related priorities for VA. Future research should consider utilizing such an approach to identify additional research and/or operational priorities.

Understanding providers' attitudes and key concerns toward incorporating CVD risk prediction into clinical practice: a qualitative study.

BACKGROUND: Although risk prediction has become an integral part of clinical practice guidelines for cardiovascular disease (CVD) prevention, multiple studies have shown that patients' risk still plays almost no role in clinical decision-making. Because little is known about why this is so, we sought to understand providers' views on the opportunities, barriers, and facilitators of incorporating risk prediction to guide their use of cardiovascular preventive medicines. METHODS: We conducted semi-structured interviews with primary care providers (n = 33) at VA facilities in the Midwest. Facilities were chosen using a maximum variation approach according to their geography, size, proportion of MD to non-MD providers, and percentage of full-time providers. Providers included MD/DO physicians, physician assistants, nurse practitioners, and clinical pharmacists. Providers were asked about their reaction to a hypothetical situation in which the VA would introduce a risk prediction-based approach to CVD treatment. We conducted matrix and content analysis to identify providers' reactions to risk prediction, reasons for their reaction, and exemplar quotes. RESULTS: Most providers were classified as Enthusiastic (n = 14) or Cautious Adopters (n = 15), with only a few Non-Adopters (n = 4). Providers described four key concerns toward adopting risk prediction. Their primary concern was that risk prediction is not always compatible with a "whole patient" approach to patient care. Other concerns included questions about the validity of the proposed risk prediction model, potential workflow burdens, and whether risk prediction adds value to existing clinical practice. Enthusiastic, Cautious, and Non-Adopters all expressed both doubts about and support for risk prediction categorizable in the above four key areas of concern. CONCLUSIONS: Providers were generally supportive of adopting risk prediction into CVD prevention, but many had misgivings, which included concerns about impact on workflow, validity of predictive models, the value of making this change, and possible negative effects on providers' ability to address the whole patient. These concerns have likely contributed to the slow introduction of risk prediction into clinical practice. These concerns will need to be addressed for risk prediction, and other approaches relying on "big data" including machine learning and artificial intelligence, to have a meaningful role in clinical practice.

Quality Improvement and Personalization for Statins: the QUIPS Quality Improvement Randomized Trial of Veterans' Primary Care Statin Use.

BACKGROUND: Implementation of new practice guidelines for statin use was very poor. OBJECTIVE: To test a multi-component quality improvement intervention to encourage use of new guidelines for statin use. DESIGN: Cluster-randomized, usual-care controlled trial. PARTICIPANTS: The study population was primary care visits for patients who were recommended statins by the 2013 guidelines, but were not receiving them. We excluded patients who were over 75 years old, or had an ICD9 or ICD10 code for end-stage renal disease, muscle pain, pregnancy, or in vitro fertilization in the 2 years prior to the study visit. INTERVENTIONS: A novel quality improvement intervention consisting of a personalized decision support tool, an educational program, a performance measure, and an audit and feedback system. Randomization was at the level of the primary care team. MAIN MEASURES: Our primary outcome was prescription of a medium- or high-strength statin. We studied how receiving the intervention changed care during the quality improvement intervention compared to before it and if that change continued after the intervention. KEY RESULTS: Among 3787 visits to 43 primary care providers, being in the intervention arm tripled the odds of patients being prescribed an appropriate statin (OR 3.0, 95% CI 1.8-4.9), though the effect resolved after the personalized decision support ended (OR 1.7, 95% CI 0.99-2.77). CONCLUSIONS: A simple, personalized quality improvement intervention is promising for enabling the adoption of new guidelines. CLINICALTRIALS. GOV IDENTIFIER: NCT02820870.

Results From a Trial of an Online Diabetes Prevention Program Intervention.

INTRODUCTION: Online Diabetes Prevention Programs (DPPs) can be scaled up and delivered broadly. However, little is known about real-world effectiveness and how outcomes compare with in-person DPP. This study examined online DPP weight loss and participation outcomes and secondarily compared outcomes among participating individuals with parallel in-person interventions. STUDY DESIGN: A large non-randomized trial supplemented by a comparative analysis of participating individuals from a concurrent trial of two parallel in-person programs: in-person DPP and the Veterans Administration's standard of care weight loss program (MOVE!). SETTING/PARTICIPANTS: Obese/overweight Veterans with prediabetes enrolled in online DPP (n = 268) between 2013 and 2014. Similar eligibility criteria were used to enroll in-person participants between 2012 and 2014 (n = 273 in-person DPP, n = 114 MOVE!) within a separate trial. INTERVENTION: Online DPP included a virtual group format, live e-coach, weekly modules delivered asynchronously, and wireless home scales. In-person programs included eight to 22 group-based, face-to-face sessions. MAIN OUTCOMES MEASURES: Weight change at 6 and 12 months using wirelessly uploaded home scale data or electronic medical record weights from clinical in-person visits. Outcomes were analyzed between 2015 and 2017. RESULTS: From 1,182 invitations, 268 (23%) participants enrolled in online DPP. Among these, 158 (56%) completed eight or more modules; mean weight change was -4.7kg at 6 months and -4.0kg at 12 months. In a supplemental analysis of participants completing one or more sessions/modules, online DPP participants were most likely to complete eight or more sessions/modules (87% online DPP vs 59% in-person DPP vs 55% MOVE!, p < 0.001). Online and in-person DPP participants lost significantly more weight than MOVE! participants at 6 and 12 months; there was no significant difference in weight change between online and in-person DPP. CONCLUSIONS: An intensive, multifaceted online DPP intervention had higher participation but similar weight loss compared to in-person DPP. An intensive, multifaceted online DPP intervention may be as effective as in-person DPP and help expand reach to those at risk.

Designing for Clinical Change: Creating an Intervention to Implement New Statin Guidelines in a Primary Care Clinic.

BACKGROUND: Recent clinical practice guidelines from major national organizations, including a joint United States Department of Veterans Affairs (VA) and Department of Defense (DoD) committee, have substantially changed recommendations for the use of the cholesterol-lowering statin medications after years of relative stability. Because statin medications are among the most commonly prescribed treatments in the United States, any change in their use may have significant implications for patients and providers alike. Prior research has shown that effective implementation interventions should be both user centered and specifically chosen to address identified barriers. OBJECTIVE: The objectives of this study were to identify potential determinants of provider uptake of the new statin guidelines and to use that information to tailor a coordinated and streamlined local quality improvement intervention focused on prescribing appropriate statins. METHODS: We employed user-centered design principles to guide the development and testing of a multicomponent guideline implementation intervention to improve statin prescribing. This paper describes the intervention development process whereby semistructured qualitative interviews with providers were conducted to (1) illuminate the knowledge, attitudes, and behaviors of providers and (2) elicit feedback on intervention prototypes developed to align with and support the use of the VA/DoD guidelines. Our aim was to use this information to design a local quality improvement intervention focused on statin prescribing that was tailored to the needs of primary care providers at our facility. Cabana's Clinical Practice Guidelines Framework for Improvement and Nielsen's Usability Heuristics were used to guide the analysis of data obtained in the intervention development process. RESULTS: Semistructured qualitative interviews were conducted with 15 primary care Patient Aligned Care Team professionals (13 physicians and 2 clinical pharmacists) at a single VA medical center. Findings highlight that providers were generally comfortable with the paradigm shift to risk-based guidelines but less clear on the need for the VA/DoD guidelines in specific. Providers preferred a clinical decision support tool that helped them calculate patient risk and guide their care without limiting autonomy. They were less comfortable with risk communication and performance measurement systems that do not account for shared decision making. When possible, we incorporated their recommendations into the intervention. CONCLUSIONS: By combining qualitative methods and user-centered design principles, we could inform the design of a multicomponent guideline implementation intervention to better address the needs and preferences of providers, including clear and direct language, logical decision prompts with an option to dismiss a clinical decision support tool, and logical ordering of feedback information. Additionally, this process allowed us to identify future design considerations for quality improvement interventions.

ForgIng New paths in DIabetes PrevenTion (FINDIT): Study Protocol for a Randomized Controlled Trial.

BACKGROUND: Prediabetes is an asymptomatic condition in which patients' blood glucose levels are higher than normal but do not meet diagnostic criteria for type 2 diabetes mellitus (T2DM). A key window of opportunity to increase engagement of patients with prediabetes in strategies to prevent T2DM is when they are screened for T2DM and found to have prediabetes, yet the effects of this screening and brief counseling are unknown. METHODS: In this parallel-design randomized controlled trial we will recruit 315 non-diabetic patients from the Ann Arbor VA Medical Center (AAVA) who have one or major risk factors for T2DM and an upcoming primary care appointment at the AAVA, but have not had a hemoglobin A1c (HbA1c) test to screen for T2DM in the previous 12 months. After informed consent, participants will complete a baseline survey and be randomly assigned to, at the time of their next primary care appointment, one of two arms: (1) to have a hemoglobin A1c (HbA1c) test to screen for T2DM and receive brief, standardized counseling about these results or (2) to review a brochure about clinical preventive services. Participants will complete surveys 2 weeks, 3 months, and 12 months after their primary care appointment, and a weight measurement 12 months after their primary care appointment. The primary outcome is weight change after 12 months. The secondary outcomes are changes in perception of risk for T2DM; knowledge of T2DM prevention; self-efficacy and motivation to prevent T2DM; use of pharmacotherapy for T2DM prevention; physical activity; participation in weight management programs; and mental health. Quantitative analyses will compare outcomes among participants in the HbA1c test arm found to have prediabetes with participants in the brochure arm. Among participants in the HbA1c test arm found to have prediabetes we will conduct semi-structured interviews about their understanding of and reactions to receiving a prediabetes diagnosis. DISCUSSION: This trial will generate foundational data on the effects of a prediabetes diagnosis and brief counseling on patients' preventive behaviors and mediators of these behaviors that will enable the development of novel strategies to improve patient engagement in T2DM prevention. TRIAL REGISTRATION: ClinicalTrials.gov, NCT02747108 . Registered on 18 April 2016.

Implementation of a prediabetes identification algorithm for overweight and obese Veterans.

Type 2 diabetes prevention is an important national goal for the Veteran Health Administration (VHA): one in four Veterans has diabetes. We implemented a prediabetes identification algorithm to estimate prediabetes prevalence among overweight and obese Veterans at Department of Veterans Affairs (VA) medical centers (VAMCs) in preparation for the launch of a pragmatic study of Diabetes Prevention Program (DPP) delivery to Veterans with prediabetes. This project was embedded within the VA DPP Clinical Demonstration Project conducted in 2012 to 2015. Veterans who attended orientation sessions for an established VHA weight-loss program (MOVE!) were recruited from VAMCs with geographically and racially diverse populations using existing referral processes. Each site implemented and adapted the prediabetes identification algorithm to best fit their local clinical context. Sites relied on an existing referral process in which a prediabetes identification algorithm was implemented in parallel with existing clinical flow; this approach limited the number of overweight and obese Veterans who were assessed and screened. We evaluated 1,830 patients through chart reviews, interviews, and/or laboratory tests. In this cohort, our estimated prevalence rates for normal glycemic status, prediabetes, and diabetes were 29% (n = 530), 28% (n = 504), and 43% (n = 796), respectively. Implementation of targeted prediabetes identification programs requires careful consideration of how prediabetes assessment and screening will occur.