Leveraging HIT For Actionable Knowledge: Synthesized List of Themes by Category AcademyHealth’s site visit interviews and other discussions with the six early HITadopting health systems examined in this project revealed a plethora of themes and issues with potential relevance to other health systems interested in using HIT for research and other analytic purposes. Many of the same themes arose in more than one health system. This document synthesizes (with no discussion) those themes, grouped into four categories, two of which contain sub-categories. The purpose of this list is to serve as a planning tool for comparative issue briefs and papers for the benefit of other health systems and policymakers. October 21, 2010 Funded with the support of the California Health Care Foundation, Oakland, California (Grant #15304) Category 1: Design of HIT Systems The pros and cons of involving researchers in the design of HIT systems. Roles they can productively play in system design. Potential sustainability of a program like PCIP to create a distributed data network. Willingness of physicians to share data for research and analytic purposes beyond initial required period. Lessons for other attempts to create distributed data networks at the level of physician practices. Tradeoffs between promoting bottom-up innovation in the uses of HIT among clinicians versus making more disciplined, top-down decisions about HIT use in order to assure consistency, data quality, and more controlled use of resources. The pros and cons of customized EHR interface and functionality in terms of meeting local needs of preferences and data quality. Pros and cons of requiring a common set of standardized elements while otherwise allowing local variation. Key players and decisions in the process of developing a data warehouse. Relative roles of internal decisions and decisions made in collaboration with outside organizations such as the HMO Research Network or the VA’s VistA systems. Category 2: Data Management Data Storage, Access and IT Support Approaches to data stewardship (category 4 below) are linked to decisions about data storage and access. The need for both national and local databases for national systems like the VA and KP. Parallels for other types of multi-site health systems. Other pros and cons of distributed data networks for research (including physician compliance with data entry, data quality, software patch difficulties, granularity, designing research protocols from hub). Pros and cons of centralized data gate-keeping versus more decentralized approaches. Pros and cons of a virtual research environment (like VINCI) for data access (and security). Tensions within the health systems over having research-specific IT support instead of having a single IT support unit for research, business, and clinical purposes. Data Quality Logistic and data quality challenges in national or other multi-site databases—the need for a standard vocabulary, standard conventions for data entry, handling of missing data, and creation and maintenance of metadata. 1 Challenges of getting physician cooperation in entering data. Divergence in data management needs of researchers and clinicians – e.g. researchers’ need for reproducible results versus clinicians’ updates to patients’ EHRs that rewrite data in particular fields. Limitations to the complexity and usefulness of research that can be undertaken due to the difficulties in reconciling data inconsistencies across linked or pooled databases. Challenges of harmonizing data from different sources within each health system, and the costs of harmonization. The need for natural language processing tools to extract information from text notes. Category 3: Data Stewardship Alternative approaches to data stewardship and their implications: 1. Maintaining all identified data and analysis behind the clinical care firewall through the use of a virtual research environment. Access is limited to authorized users and no information is downloaded (the VA’s VINCI system). Review by IRB for research projects. 2. Researchers work outside the clinical firewall with de-identified analytic data files drawn from a clone of identified EHR database. IT staff responsible for creating analytic files reporting through the clinical care chain of command. Review by IRB of research projects (PAMF). 3. Variations of Approach #2 with different organizational reporting structures for IT support staff responsible for creating de-identified analytic data files (KP, DH, GHS). 4. Data remains with patient care providers in a distributed data network. Researchers only receive aggregated count data (PCIP). Perceived ambiguities of the IRB process for secondary data analysis, and their implications for research involving electronic data. Perceived ambiguities concerning data ownership and patient consent for data from EHRs. Perceived uncertainty concerning the application of IRB/privacy rules to public health research using EHR data. Category 4: Research Research Culture The centrality of an innovation mindset, a commitment to innovation in health care delivery, to the successful use of HIT data for research purposes. The pervasiveness of this orientation throughout the clinical, operational, and research components of the health system. 2 The relationship between researchers and IT professionals – making IT people feel more valuable by showing them the information value of data; research as an avenue for IT to serve a new client and provide new services to old clients; joint gains. The relative merits of delivery organizations versus academia for health services researchers. The centrality of the “business case” for research capabilities to provide actionable knowledge, but also to increase organizational prestige and brand. Role of serendipity in the successful use of HIT data for research: e.g. a specific confluence of events at PAMFRI (new PAMFRI director, merger with Sutter, further rollout of Epic) that provided a new focus on innovative use of data, joint gains in systems and data standardization for all parts of PAMF, and less legacy data. Distinctiveness of health systems operating a health plan for the role of research in the organization; the implications of needing actuarial informatics for creating a mindset open to using research for prediction, a form of actionable knowledge. Clinicians’ incentives and rewards from the health system for conducting research. Other professionals’ incentives to conduct research. Differences in approaches to HIT-based research in safety-net systems versus those that serve broader patient populations. Differences between public and private health systems. Research Benefits and Challenges The increasing breadth of research and analysis in health care delivery systems using HIT and the value and challenges (methodological and other) in each: o Health services research o Clinical and behavioral research o Genomic research o Public health surveillance and population health research o Operational analysis of capacity, work flows, etc. (e.g. Lean Processes) o Physician-engaged QI including disease management and coordination, health promotion and disease prevention, o Pay-for-performance Increasing difficulties in distinguishing analysis for quality improvement (QI) from research. Difficulty in categorizing analysis involving registries as a QI activity versus research. Implications of labeling analysis as QI versus research. Challenges in determining what level of methodological rigor is appropriate for what purpose. Spillover benefits of having in-house research capacity in terms of improved data quality and greater methodological rigor in analyses throughout the health system. Challenges and risks posed by having “data by the pound” capabilities. Benefits and challenges of research collaborations with academic institutions and other health systems: o Benefits in extending staff capabilities o Potential for larger patient pools 3 o Challenges in data interchange and cross-institutional variation in vocabularies/interpretations of similar data elements, data quality, and data stewardship o Differences in incentives to publish and other cultural differences between health systems and academia, or among different collaborating health systems Challenges in moving promising clinical and health services research-based innovations from research to pilots and from pilots to broader patient pools. Funding Sources and Research Agendas How research agendas are determined within organizations. The mix of internal requests, investigator initiated projects, and externally defined questions. Implications of each approach. For national health systems, what leads to particular research and analytic activities being undertaken at the national versus regional or local level? At a particular site versus system-wide? What are the relative merits of research being separate versus integrated with patient care (both organizationally and technologically)? To what extent do research and clinical/operational cultures clash versus support one another? Scientific and Technical Talent Implications of an increased commitment to research and other analytic uses of EHRs for the number of IT professionals needed and their skill sets. Challenges in recruiting scientific talent to conduct research at a health care delivery organization. Criteria for evaluating researchers’ job performance: the role of external scientific activities (publications, conference presentations, service to the research field, externally-funded investigator-initiated research projects, etc.) versus internal contributions to the organization. 4