A proposed national research and development agenda for population health informatics: summary recommendations from a national expert workshop

Abstract

Objective: The Johns Hopkins Center for Population Health IT hosted a 1-day symposium sponsored by the National Library of Medicine to help develop a national research and development (R&D) agenda for the emerging field of population health informatics (PopHI).

Material and Methods: The symposium provided a venue for national experts to brainstorm, identify, discuss, and prioritize the top challenges and opportunities in the PopHI field, as well as R&D areas to address these.

Results: This manuscript summarizes the findings of the PopHI symposium. The symposium participants’ recommendations have been categorized into 13 overarching themes, including policy alignment, data governance, sustainability and incentives, and standards/interoperability.

Discussion: The proposed consensus-based national agenda for PopHI consisted of 18 priority recommendations grouped into 4 broad goals: (1) Developing a standardized collaborative framework and infrastructure, (2) Advancing technical tools and methods, (3) Developing a scientific evidence and knowledge base, and (4) Developing an appropriate framework for policy, privacy, and sustainability. There was a substantial amount of agreement between all the participants on the challenges and opportunities for PopHI as well as on the actions that needed to be taken to address these.

Conclusion: PopHI is a rapidly growing field that has emerged to address the population dimension of the Triple Aim. The proposed PopHI R&D agenda is comprehensive and timely, but should be considered only a starting-point, given that ongoing developments in health policy, population health management, and informatics are very dynamic, suggesting that the agenda will require constant monitoring and updating.

Background and Significance

Over the last decade, there have been unprecedented efforts to improve health and reduce cost using health information technology (HIT). The wide-scale adoption of HIT, including a quadrupling of electronic health record (EHR) use among physicians, has enabled diverse stakeholders, such as providers, payers, and government agencies, to collaborate using new digital tools to improve the health of defined populations.^1–4 There are an increasing number of population health interventions that use HIT to improve the health of persons enrolled in a specific health plan or cared for by a single provider organization.^5–8 In addition, community-wide government-led or government-coordinated traditional public health IT efforts have experienced a resurgence of interest, if not resources, during the last decade.

In 2014, the Johns Hopkins Center for Population Health IT⁹ hosted a 1-day symposium sponsored by the National Library of Medicine to develop a national research and development (R&D) agenda for the emerging field of population health informatics¹⁰ (PopHI). PopHI integrates aspects of public health, clinical informatics, and healthcare delivery with the target of improving healthcare system effectiveness and the well-being of communities and populations.^5,11,12 Given the transformation of the healthcare and HIT fields currently taking place due to the funding of the Affordable Care Act,¹³ the Health Information Technology for Economic and Clinical Health Act,¹⁴ and other private and public sector reforms,¹⁵ there is a pressing need to set priorities for R&D in the PopHI field.

Symposium attendees included national experts and leading researchers in academia, the public sector, provider organizations, and the information technology (IT) industry. The symposium’s aim was to organize an interactive session for national leaders from various stakeholder groups to discuss challenges and opportunities (C&Os) within the PopHI field. To accomplish this aim, the key objectives of the symposium were to:

• Develop a consensus definition of PopHI;

• Identify key priorities in terms of opportunities for using HIT/informatics tools to improve population health;

• Identify key challenges and technical hurdles in the PopHI field;

• Suggest a national agenda/roadmap for research and technical development in the PopHI field for the next 5–10 years;

• Identify recommendations for how R&D stakeholders (including researchers, policymakers, HIT users and vendors, and funders) could facilitate this agenda; and

• Disseminate findings to those with involvement and interest in the PopHI field.

Working Definition of PopHI

To facilitate discussions, the following working definition of population health was proposed and adopted:

Population health comprises organized activities for assessing and improving the health and well-being of a defined population. Population health is practiced by both private and public organizations. The target “population” can be a specific geographic community or region, or it may represent some other “denominator,” such as enrollees of a health plan, persons residing in a provider’s catchment area, or an aggregation of individuals with special needs. The difference between population health and public health is subtle, and there is not always a full consensus on these definitions. That said, public health services are typically provided by government agencies and include the “core” public health functions of health assessment, assurance, and policy-setting. In the United States, most actions of public health agencies represent population health, but a considerable proportion, if not the majority, of population health services are provided by private organizations.

In this symposium, the terms “population health informatics” and “population health IT” were used interchangeably. “PopHI” has been used throughout this manuscript in place of both terms. The population health implications of all interrelated HIT tools and data sources (eg, data collected by patients, providers, payers, governmental agencies, and other population health stakeholders) were also considered in scope.

PopHI overlaps with other informatics fields, such as public health informatics, clinical informatics, and consumer health informatics. This overlap is partly due to the similarity of data sources, stakeholders, or operational goals. To help compare and contrast the PopHI field from the commonly understood informatics subdisciplines of public health informatics and clinical informatics, we developed a summary framework, presented in Figure 1.^16–18

Figure 1A.

The varying contexts of population health informatics, public health informatics, and clinical informatics. EHR, electronic health record; HIT, health information technology. ^aThere are no clear boundaries for the intervention targets. Interventions are ordered by the most to the least common target constituents. See Figure 1B for a visual representation of the overlapping intervention targets. Figure 1B Schematic diagram, adapted from the Institute of Medicine’s participating constituents of primary care,¹⁹ depicting the overlapping intervention targets listed in Figure 1A. ^bTotal population of a geographical boundary (eg, a community surrounding a provider), regardless of their attribution to an individual entity (eg, a healthcare provider or an employer). ^cTarget population attributed to an individual entity (eg, a healthcare provider or an employer). In the case of a healthcare provider, the target population will be the entire denominator of their patients or parts of it.

Materials and Methods

The PopHI symposium provided a venue for national experts to brainstorm, identify, discuss, and prioritize top C&Os in the PopHI field. The experts also suggested areas of R&D that could play a critical role in the PopHI field in the future. A program committee was formed to identify potential participants (ie, PopHI experts) and generate a list of topical domains for the brainstorming/breakout sessions.

Participant Characteristics

The program committee compiled a preliminary list of 240 PopHI experts, based on publication records, web resources, affiliation with a PopHI-related center or business, and prior individual encounters. A comprehensive effort was made to reach out to all potential candidates on this list, to ensure a balanced representation from academia, industry, and governmental/nonprofit organizations. A total of 81 experts accepted our invitation and participated in the symposium. The distribution of the attendees’ affiliation was: 34% academia, 32% government or non-profit organizations, 17% IT industry, and 17% healthcare provider or payer institutions. A list of experts who participated in the symposium is presented in the Supplementary Appendix.

Topical Domains

The program committee used the Delphi method to generate a list of six topical domains for the breakout sessions (one domain was discussed per session).^20,21 These domains were generated to guide and facilitate the breakout sessions. The domains provided a wide range of topics that capture various aspects of PopHI. These six domains were:

1. Creating a robust and interoperable information infrastructure;

2. Collaborating across health plans, providers, and vendors;

3. Using quality indicators and other metrics for population health;

4. Developing the next generation of HIT-based tools for integrated delivery systems/networks (IDSs) and Accountable Care Organizations (ACOs);

5. Utilizing computer science (CS) as well as informatics tools and methodologies; and

6. Linking non-clinical health data with clinical data (eg, linking social services data with EHR data).

The committee members drafted 1-page explanations for each domain, which also offered a working definition of PopHI as well as preliminary lists of C&Os and potential R&D priority areas. These documents were distributed to participants beforehand and used in the breakout sessions to stimulate discussions.

Moderation of the Breakout Sessions

The symposium consisted of a welcome address, three keynote speeches (available online at: http://www.jhsph.edu/research/centers-and-institutes/johns-hopkins-center-for-population-health-information- technology/center-for-population-health-it-video-library), six breakout sessions, and a closing session. The symposium participants chose two breakout sessions to attend, and each session was moderated by two experts. Moderators were mainly selected from the program committee. Each session started with the moderators and participants reviewing the 1-page document provided for the session. Following Osborn’s brainstorming methodology,²² moderators focused on collecting as many C&O and R&D issues raised by the participants, avoided criticizing the ideas suggested, welcomed unusual ideas, and used the collective feedback of the group to break up or merge ideas. Breakout sessions started with a discussion of C&O issues and then followed with a discussion of R&D issues. At the end of each session, the participants voted for their top three highest-priority C&O and R&D ideas (separately). Votes were captured by assistant moderators, aggregated, and ranked to identify the top recommendations of each session’s participants. Ideas with no votes were removed from the final recommendations list. At the end of the symposium, the top C&O and R&D ideas were shared with all the participants in the closing session and discussed.

Data Collection and Analysis

After the symposium, all the ideas that were discussed and captured were aggregated and ranked based on the votes collected during the breakout sessions. C&O and R&D ideas were kept separate during this process so that unique trends and themes in each area could be identified. Using grounded theory,^23,24 the ideas were analyzed and grouped to form common themes in PopHI. In this process, each idea was broken down by two independent reviewers (H.K. and E.L.) into a maximum of four keywords. Keywords were merged and conflicts were resolved by further discussion. Keywords were collated and grouped into 13 distinct categories or themes, representing different aspects of PopHI. Lastly, each of the ideas was assigned to one of the themes, and themes were ranked based on the number of votes each received.

Results

Summary of Findings

A total of 60 unique recommendations were generated for C&O issues in PopHI. Matching the same domains, a total of 55 recommendations were generated for R&D issues. Tables 1 and 2 present a detailed list of C&Os and the R&D recommendations and their ranking by number of votes within each domain, respectively.

Table 1.

Key PopHI C&Os Identified by the Breakout Session Participants

Breakout Session	Domain
A	Interoperability and Information Infrastructure Challenges
1	Unclear vision and lack of a shared model for a PopHI infrastructure (eg, functions, requirements, and use cases)
2	Mixed incentives for clinical organizations to focus on true population health over the long term
3	Need for effective patient identification methods to link medical, public health, and social service data systems
4	Inconsistent adoption of existing interoperability standards
5	Siloed and misaligned cross-disciplinary systems with different functions and context
6	Absence of a convener to develop needs-based infrastructure strategy
7	Public concern over data ownership/access and need to balance public good and individual consent
8	Limited sustainability and lack of funding, especially for non-medical population health infrastructure
9	Need to balance both patient centricity and populations
10	Hard to measure impact of population health policy
B	Cross-Organization Collaboration Challenges
1	Disparate and not fully functional data exchange agreements and models
2	Concept of population value vs patient/fee-for-service model in transition
3	Lack of financial incentives to use and exchange data
4	Different time horizons of population health impact for different parties
5	Challenges in disseminating/translating evidence-based knowledge regarding HIT-based population health interventions
6	Instability of cross-organizational collaborations and private/public partnerships
7	Challenges of blending medical, public health, and consumer-targeted interventions
8	Need for robust standardization of claims to track cross-organizational events
9	Uncertain impact of Affordable Care Act on population health initiatives
10	Low scalability and flexibility of HIT infrastructure
C	Challenges Related to PopHI-Based Indicators and Metrics
1	Current metrics have not embraced new electronic data sources (eg, consumer m-health/biometrics, unstructured EHR data)
2	Lack of adequate data validity, reliability, accessibility, and comparability (including a lack of data challenges)
3	Complexity of standardized population denominators for measurement
4	Lack of a standardized/prioritized population health metric HIT-derived report card
5	Challenges in aligning population health metrics with clinical metrics
6	Limited financial aid and support for shared development of population health e-Metrics
7	Difficulty operationalizing population health metrics in practice
8	Lack of clarity of the role of payers, providers, government, and vendors in developing population health e-Metrics
9	Need to develop metrics that are linked to scientific evidence, both in terms of population health and other outcomes
10	Need for a coordinating body and coordinated infrastructure for population health e-Metrics
D	Challenges in Applying PopHI Within Integrated Provider Systems
1	Lack of appropriate population health electronic quality measures for IDS/ACO populations (eg, measuring value)
2	Absence of proof-of-concept and evidence using enterprise-level HIT population health interventions
3	Complexity of developing actionable team-based PopHI systems and their absence in MU Stage 3
4	Historic imbalance of population vs individual provider focus within ACO EHR/HIT systems
5	Lack of governance structures for effective and efficient data-sharing in an ACO
6	Need for national benchmarks and normative information about ACO PopHI systems
7	Issues with human factors of HIT in ACO context and changing provider behavior
8	Lack of population health decision-support tools for value-based care delivery models
9	Misalignment of individual and population-based measures for ACOs and PCMHs
10	Ambiguous definition of population and patient attribution in ACOs
E	Challenges in CS and Informatics Methods for Population Health
1	Handling the inherent heterogeneity and complexity of population health data
2	Developing new tools to address data privacy across a population
3	Low levels of collaboration between HIT experts and population health experts
4	Lack of standardized data architectures for population health
5	Barriers associated with accessing population-level health data for R&D
6	Challenges associated with visualizing health data trajectories across a population
7	Addressing the potential hype of Big Data in the larger context of population health
8	Shortage of HIT talent in dealing with new sources of population data
9	Increasing variability of data sources and data patterns in population health
10	High rates of imperfect information and low-quality data in population health data sources
F	Challenges Facing the Integration of Social and non-medical Factors into PopHI Systems
1	Confusion regarding what information should be protected when using non-clinical data
2	Challenges of using social/non-health data to generate signals for population health interventions
3	Technical and political barriers to integrating non- clinical and person-specific data
4	Lack of a conceptual framework and infrastructure for considering and incorporating social/non-clinical data in population health analysis
5	Lack of measurement standards for integrating social/non-clinical data with population health data
6	Current low data quality and low availability of important social/non-clinical data
7	Challenges in funding and sustaining the integration of non-clinical data
8	Fragmentation of non-clinical data systems across a defined population
9	Need to expand evidence base on relationship between non-clinical data and population and clinical health outcomes
10	Challenges in the prioritization of non-clinical data sources and specifications

ACO, Accountable Care Organization; C&Os, challenges and opportunities; CS, computer science; e-Metric, electronically based metric; EHR, electronic health record; HIE, health information exchange; HIT, health information technology; IDS, integrated delivery system/network; IT, information technology; m-health, mobile health; MU, Meaningful Use; PCMH, patient-centered medical home; PopHI, population health informatics; R&D, research and development. Each C&O is ordered according to its prioritization by each breakout session (based on a vote taken by the participants).

Table 2.

PopHI R&D Priorities Identified by the Breakout Session Participants

Breakout Session	Domain
A	Creating a Robust and Interoperable PopHI Infrastructure
1	Create a widely accepted vision of how to address top PopHI challenges, as well as the funding to support those efforts
2	Develop methods for securely linking records between different stakeholder organizations at the population level
3	Promote a shared knowledge base on geographically linked health and social data for use in PopHI
4	Design end-user and population health decision-support tools that fully consider usability
5	Align population health priorities with Learning Health System goals
6	Develop approaches for integrating PopHI into clinical decision support at the point of care
7	Establish a baseline to monitor progress towards future population health priorities
8	Develop a road map for a unified person-centric community health record
9	Explore whether various clinical data are a suitable fit for population health uses
10	Cultivate a community of stakeholders to adopt best practices for PopHI
B	Improving Collaboration on PopHI Across Organizations and Stakeholders
1	Develop a consensus on what data to collect, integrate, and share across organizations and stakeholders
2	Develop and share new analytic solutions and models targeted at PopHI across providers and agencies
3	Build on established standards (eg, HL7) to develop new PopHI tools, measures, and methods
4	Develop a set of metrics and decision support guidelines that stakeholders will agree to
5	Create methods and tools to measure data completeness and quality across various stakeholders
6	Empower the adoption of current interoperability frameworks for data exchange and further enhance these frameworks
7	Construct mechanisms to automatically link administrative data (eg, insurance claims) and EHR data for a defined population
8	Develop and maintain a standardized repository of useful geographic and social data sources for population health across organizations
9	Leverage existing efforts in the HIE/clinical data exchange community for population health
10	Find ways to increase patient and community engagement in sharing data across healthcare providers
C	Creating Electronic Population Health Indicators and Metrics
1	Develop predictive models that are actionable when population health metrics/targets are not achieved
2	Build an HIT framework to capture and integrate patient/consumer reports on the population scale
3	Create a methodology to align existing population health metrics across multiple programs/stakeholders
4	Establish a framework to support the roll-out of new standardized population health metrics
5	Augment strategies to identify and prioritize the development of new critical data elements for metrics
6	Find ways to use HIE data to report population health metrics
7	Standardize elements that are needed in key quality measures of population health
8	Develop and support a national centralized, but federated, reporting mechanism for population health
9	Collaboratively develop an attribution methodology for population health metrics
10	Maintain a compendium of interoperable open-source tools for population health
D	Creating the Next Generation of PopHI Tools for IDSs and ACOs
1	Foster pragmatic approaches to data-sharing and integration between all parties
2	Integrate practice needs for population health data management and governance
3	Improve human factors and workflow issues (eg, population health data visualization)
4	Define a unifying framework for population health and stakeholder motives
5	Define boundaries for population management that HIT cannot solve
6	Simulate Triple Aim’s ROI model for ACOs in the context of PopHI
7	Establish, support, and evaluate networks as well as a learning collaborative of PopHI stakeholders
8	Develop policies to manage, incorporate, and share consumer-/patient-generated data to be applied to PopHI
9	Address community health issues by sharing population health data and interventions across IDSs in an area
10	Conduct ACO-based needs assessments to identify PopHI priorities
E	Advancing CS, Informatics Tools, and Methodologies in PopHI
1	Develop a set of “grand computing challenges” in PopHI, to offer incentives
2	Develop synthetic datasets that can be shared freely with computer scientists for research and development
3	Develop guidelines for large-scale data-sharing and analysis by non-health researchers
4	Advance interdisciplinary research in the area of actionable PopHI (eg, change management, workflow reasoning)
5	Assess and expand the data science and informatics workforce needed to support and improve PopHI
F	Advancing the Integration of Social and non- medical Data into PopHI Applications
1	Reliably link nonpersonal data (eg, environmental) to person-specific data (eg, health data)
2	Develop population-based decision-support tools for public health and social service agencies
3	Develop analytical tools that work across disparate population-based data types
4	Derive innovative approaches to scale the use of non-health data for population health
5	Identify non-health data sources that address high-priority population health issues
6	Measure the value of various non-health data sources for population health analytics
7	Standardize information provenance to facilitate user understanding of data sources and reliability
8	Assemble guidelines/regulations to govern the use of non-health data for population health
9	Develop tools to help clinical providers use non-health data for community health assessments
10	Furnish policies and frameworks to address privacy and confidentiality in using non-health data

ACO, Accountable Care Organization; EHR, electronic health record; HIE, health information exchange; HIT, health information technology; IDS, integrated delivery system/network; IT, information technology; HL7, Heath Level 7; PopHI, population health informatics; R&D, research and development; ROI, return-on-investment. Each R&D priority is ordered according to its prioritization by each breakout session (based on a vote taken by the participants).

Key Priority Areas Identified in the Breakout Sessions

An Interoperable Information Infrastructure for PopHI

The breakout session participants identified the need for a clear, shared vision for a common, interoperable, and robust information infrastructure as one of PopHI’s main obstacles, as was the lack of organizational incentives to focus on population health over the long-term. Other top challenges include the need for effective patient identification methods to share data across systems, inconsistent adoption of interoperability standards, and challenges associated with siloed functions and service provisions.

The highest-priority R&D recommendations to help address the obstacles identified by the breakout session participants included: collectively develop a shared national vision for PopHI, develop methods for securely and privately linking patient medical records across stakeholders, and promote a shared knowledge base for community health trends and population decision support. Other notable R&D priorities included designing end-user tools that incorporate usability methods and heuristics as well as fostering the integration of social science data into PopHI’s overarching architecture.

Cross-Organization Collaboration on PopHI

Independent providers, payers, agencies, and IT companies do not have well-established programs or policies for sharing data across boundaries for population health. Collaboration challenges among stakeholders are due to limited PopHI data use and exchange agreements as well as a lack of aligned incentives, such as evolved value-based payment models. Other important impediments to achieving cross-organization collaboration on PopHI include: inadequate evidence-based knowledge regarding HIT-based population health interventions, the challenging long-term time horizon of PopHI projects’ return-on-investment, the instability of many partnerships between private and public organizations, and the difficulty of blending medical, public health, and consumer-targeted interventions.

To help address these challenges, the breakout session participants recommended that the utmost priority be to reach a consensus on what data to collect, integrate, and share as part of PopHI. Other R&D priorities identified by the participants included: discovering and sharing new PopHI analytic solutions and models among stakeholders; defining population health data-sharing methods using established standards, such as Health Level 7 (HL7); and developing a set of metrics and decision support guidelines. The participants also proposed creating methods to assess the quality of shared data and techniques for linking disparate data sources (eg, linking EHRs with claims data) as priorities.

Population Health e-Metrics

In terms of key challenges, the breakout session participants agreed that population health metrics that can be derived from HIT systems are still immature and must be advanced.¹² The PopHI field has a critical role in collecting and collating the data necessary to develop and evaluate these electronically based metrics (or “e-Metrics”). The participants proposed a variety of informatics challenges for the development and adoption of such performance measures. For example, understanding how new data sources, such as patient-generated data and unstructured EHR data, can be utilized will be key to developing more appropriate community or other population-focused measures. In addition, there needs to be appropriate informatics methods for improving the reliability, accessibility, and comparability of such measures. Differentiating population denominators (eg, geographic) from provider-centric ones, aligning population and clinical metrics, and developing a prioritized population/community report card were also identified by the participants as informatics-related challenges to developing e-Metrics.

In the population health metric domain, the breakout session participants voted heterogeneously on R&D priorities. However, a number of the ideas received more support than others, including: developing actionable predictive models for population health metrics, building an HIT framework to capture and integrate patient-reported outcomes on the population scale, and creating a methodology to align existing population health metrics across programs. Other suggestions included establishing a framework to support the roll-out of new population-health-focused metrics as well as strategies to support the expansion of our collection of the critical data elements necessary for this framework.

The Next Generation of PopHI-Based Tools for IDSs/ACOs

Although a number of HIT tools have been developed over the last few years to support the operations of value-based care delivery networks, the breakout session participants believed that the current state-of-the-art tools represent an imbalance between focusing on populations vs individual providers within IDSs.²⁵ The participants identified a series of challenges that have affected HIT tool development and adoption by IDSs such as ACOs/patient-centered medical homes (PCMHs), including a lack of appropriate population health electronic quality measures, the absence of proof-of-concept and return-on-investment of enterprise-level tool implementation associated with the focus of the Centers for Medicare and Medicaid Services’ Meaningful Use policy to-date, and a lack of governance structures for effective and efficient data-sharing.

The breakout session participants offered a series of R&D recommendations to address these concerns. The highest-priority challenges identified by the participants were the current barriers to fostering pragmatic approaches for data-sharing and to integration between all IDS parties. Other barriers identified by the participants included: the integration of practice needs for population health data management and governance, improvement of human factors and workflow issues (eg, population health data visualization) for all stakeholders, and the identification of a unifying framework for PopHI infrastructure and stakeholder incentives for adoption.

Advancing CS and Informatics Methodologies for Population Health

The breakout session participants agreed that PopHI is an emerging field, but also that many existing informatics and CS methods can be adapted to meet some of the field’s key needs. The participants identified several high-priority target areas for methods advancement. The heterogeneity and complexity of population health data sources were identified as major factors that complicate the use of established CS/informatics methods that are traditionally applied to more straightforward data types. Additional challenges that data scientists and others face include: privacy issues associated with population-level analytics, limited collaboration between HIT and population health experts, and a lack of standardized data architectures for population health.

This breakout session was focused on actual CS and informatics methods that could advance PopHI solutions, so its participants avoided providing a long list of potential technical solutions. Instead, the group provided a number of higher-level recommendations intended to promote the application and development of methods for PopHI. The group’s highest-priority R&D recommendation was to develop and disseminate a “grand computing challenge” for PopHI, in order to encourage collaboration between computer scientists to meet the challenge and to trigger the development of advanced, innovative methods. Other R&D priorities identified by the participants included: developing synthetic population-level datasets that can be shared freely, developing guidelines for large-scale data-sharing as well as analytic frameworks for non-health researchers, and advancing interdisciplinary research in applied PopHI focus areas, such as change management and workflow reasoning.

Integrating Social and Non-clinical Data into PopHI

Population health methods and models can be enhanced by linking social and other non-medical information with clinical data. The main challenges identified by the breakout session participants were: a lack of clarity as to what information should remain protected or confidential when using non-clinical data, challenges associated with generating “signals” across an entire population, technical and political barriers to integrating non-clinical person-specific data, and a lack of standards or frameworks for considering and integrating multisector data (ie, integrating health data with non-health data) at the population level, rather than the individual level.

The high-priority R&D topics recommended by the breakout session participants included: developing solutions to reliably link non-personal data to person-specific data, developing population-based decision-support tools for both public health/community officials and clinicians, and developing analytical tools that work across disparate population health data types.

Identification of Cross-Workgroup Conference Themes

As part of our group process, we extracted 13 overarching themes from the 100+ challenges and recommendations (Tables 1 and 2) produced by the participants in the breakout sessions. These themes represent a wide range of factors that can affect PopHI. Although these themes focus on distinct perspectives, they should not be considered mutually exclusive. Moreover, some of the 13 themes identified have commonalities with a subset of those identified by prior agenda-setting groups that focused on related areas, such as public health informatics.^26–28

The 13 overarching themes of the PopHI symposium were:

1. Policy Alignment: A misalignment of population health strategies and desired outcomes hinders the adoption and implementation of appropriate PopHI solutions. For example, a variety of population health strategies, such as metrics, policies, motivators/incentives (eg, payment reform), or delivery models (eg, ACOs, PCMHs), can be misaligned and, hence, can slow the adoption of PopHI solutions.

2. Data Governance: Conflicts between data governance, data access, and privacy introduce additional complexities to developing PopHI solutions. For instance, insufficient data governance policies may produce a conflict between population-wide data access and data ownership, thus diminishing the ability to share the necessary data with stakeholders.

3. Data Quality: A lack of essential data quality specifications for many relevant population health data sources reduces their utility for PopHI. In particular, a suboptimal rate of completeness, accuracy, timeliness, reliability, or validity of a data source can produce uncertainty when the data are incorporated into PopHI solutions.

4. Data Management: The current divergence of population health data types, sources, and architectures complicate data management planning and operations. Differences in granularity and heterogeneity across different data sources produce many data management challenges.

5. Sustainability and Incentives: A lack of funding slows down PopHI initiatives and reduces the long-term sustainability of PopHI programs. Difficulty attracting new funding streams may affect the scalability of a project and limit the flexibility of its core business model.

6. Population Metrics: The impracticality and unevenness of population health measures burden PopHI projects. The lack of accurate and reproducible e-Metrics may reduce the effectiveness and actionability of PopHI solutions.

7. Interoperability and Standards: Limited population health standards and interoperability frameworks affect the development of new PopHI solutions. Specifically, the restricted interoperability of new data sources with existing clinical data can limit PopHI developers’ ability to offer new data-driven capabilities in their tools.

8. Stakeholder Collaboration: The instability of cross-organizational collaborations and a lack of effective leadership often results in multistakeholder PopHI programs being unsuccessful. Hence, PopHI solutions may produce suboptimal results if stakeholders do not collaborate effectively.

9. Tools and Infrastructure: The lack of robust population health data infrastructure and tools impedes the development and expansion of a robust HIT ecosystem for population health. Moreover, technology that has been applied almost exclusively in pursuit of clinical applications cannot always be easily integrated into PopHI systems.

10. Ethics and Security: Concerns about the privacy, confidentiality, and security of data restrict the expansion of population-wide HIT solutions. Integrating multiple data sources (eg, integrating non-clinical data with clinical data) could lead to a greater risk of confidentiality breaches beyond the risk that already exists within the context of each dataset.

11. Best Practices and Dissemination: Insufficient best practices for PopHI and limited dissemination mechanisms to introduce them among stakeholders have restricted the adoption of new PopHI solutions. When considering whether to fund such solutions, a lack of successful PopHI case studies may give decision makers pause.

12. Education and Training: The need for additional educational and training programs in PopHI is essential to advancing the science and operations of PopHI. Properly training PopHI experts in both payer and provider settings can help empower cross-organizational PopHI programs.

13. Evaluation and Methods: The development and evaluation of new PopHI methods are essential to advancing the science of population health. The lack of scientific evidence on the effectiveness of PopHI solutions may eventually reduce their adoption, despite some ad-hoc operational success.

Discussion and Recommendations

PopHI is a rapidly growing field that has emerged to address the population dimension of the Triple Aim.²⁹ The Triple Aim is a unifying framework developed by the Institute for Healthcare Improvement that includes the three dimensions of population health, patient experience, and cost reduction as means of optimizing the performance of health systems. This new focus on communities and other “denominator populations” has been propelled by:

1. The Affordable Care Act and other reform initiatives that have prompted the rapid adoption of value-based, HIT-supported healthcare delivery models, such as ACOs and PCMHs^{13,25,30–32};

2. The Health Information Technology for Economic and Clinical Health Act and state-level EHR adoption incentives that have exponentially increased HIT use among healthcare providers^4,14,33;

3. Data interoperability efforts, such as HL7, the Clinical Data Interchange Standards Consortium (CDISC), and the Standards and Interoperability (S&I) Framework, that have led to the integration, consolidation, and wider adoption of relevant data encoding and information exchange standards^34–37;

4. The creation of large population health data repositories that range from ACO-level clinical data warehouses, to regional networks such as the national Patient-Centered Clinical Research Network’s (PCORnet) Clinical Data Research Networks,³⁸ and state-wide efforts such as health information exchanges (HIEs) and all-payer claims databases³⁹; and

5. Advances in Big Data methods of storing, managing, and analyzing large population health datasets originating from heterogeneous sources such as EHRs, insurance claims, HIEs, epidemiological datasets, and biometric/telemonitoring devices.^40,41

These synergetic and ongoing trends have significantly advanced the depth and breadth of the application of informatics to the population health domain and are contributing to an unprecedented array of opportunities and challenges for this developing field.

The experts who attended the PopHI symposium were in agreement as to the overarching C&Os in the field. More importantly, there was also a high degree of consensus among the experts regarding the national R&D actions needed to move the field forward in the near term. Figure 2 summarizes the key recommended action items that should constitute a national R&D agenda in the PopHI field, collated from all of the symposium workgroups. The 19 high-priority recommendations identified by the symposium participants are grouped into four broad areas: framework/infrastructure, tools, knowledge, and policy.

Figure 2.

Overview of key symposium recommendations for a national PopHI R&D agenda. ^aThe first three categories include key recommendations with policy, privacy and sustainability components; however, these key items were categorized in different groups, due to their focus on those topics. ACO, Accountable Care Organization; CS, computer science; GIS, geographic information system; HIE, health information exchange; IDS, integrated delivery system/network; PopHI, population health informatics; R&D, research and development.

The first series of recommendations in this R&D action list address the need for a standardized national framework and infrastructure to help unify the vision for the PopHI field across stakeholders. Such collaboration would help support standard approaches to integrating and linking multiple types of data. Without a unified framework, the boundaries between population health, public health, and clinical informatics would remain fuzzy, thus making who is accountable for population health efforts unclear. The second series of recommendations are focused on expanding the national PopHI methods “tool chest” in areas intended to address perceived technical challenges for the field. The third, and largest, series of recommendations are meant to contribute to the advancement of the national (and global) scientific knowledge base in areas relevant to PopHI. Finally, the last group of recommendations contains priority government policy actions relevant to PopHI: improving regulations in the privacy/data access domain and improving incentives and sustainability. If substantial progress were to be made within these four areas, researchers, policymakers, and providers alike would have new informatics tools, support, and knowledge to help improve their effectiveness in managing population health at a local, community, or regional level.

The recommendations and proposed R&D topics identified in our PopHI symposium are inherently confined to the results generated by the participating experts and should be considered preliminary. A larger group of experts representing a wider range of stakeholders and a broader geographical area is required to update the proposed recommendations with a more detailed and comprehensive list of suggestions in the near future.

Conclusion

The PopHI R&D agenda proposed herein is based on the discussions of a large and diverse panel of national experts. This agenda is comprehensive and timely, but it should be considered only a starting-point, given that ongoing developments in health policy, population health management, and informatics are very dynamic, suggesting that this R&D agenda will evolve and will require frequent updating. Moreover, although we believe the symposium’s unique focus on population health is critical, we acknowledge that this domain is closely linked to the field of public health informatics. Thus, collaboration and integration with the closely allied activities of groups such as the American Medical Informatics Association’s public health informatics workgroup should be fostered.⁴²

Like any major scientific or technical endeavor, the R&D agenda we have outlined out will require a well-trained workforce. Incorporating PopHI training into the curricula of the informatics, public health, and care management fields will engender a skilled workforce capable of pushing forward with these R&D priorities. Recognizing the immediate need for enhanced training in this domain, the Office of the National Coordinator has recently announced plans to significantly update and expand their sponsored HIT educational curriculum in PopHI.⁴³

Finally, it is our hope that as the PopHI R&D agenda moves forward, it can adapt based on the increasing number of population health interventions that have successfully leveraged PopHI solutions. This approach will be similar to the efforts of the Beacon Communities, which have developed an exemplary community-based HIT system.^6–8 Lessons learned from these and other large-scale HIT initiatives in support of population health improvement should be shared with the health informatics, public health, and health services research communities to contribute to a “Learning Population Health System” that will be capable of continuously updating and refining the proposed national PopHI R&D agenda we present herein.

Contributors

Drs Kharrazi, Weiner, Chin, Lehmann, Advani, Loonsk, and Yasnoff actively participated in the presymposium topic development, chaired the focus groups, and reviewed/merged the comments collected from the participants. Drs Kharrazi, Weiner, and Ms Lasser applied grounded theory to the findings and finalized the data analysis. They also authored the first draft of the manuscript. All authors have read, commented, discussed, and agreed on the integrity of all parts of the manuscript.

Funding

The national agenda-setting PopHI symposium was funded and supported by National Library of Medicine grant number #RLM011955A. DST Health Solutions, PricewaterhouseCoopers LLP, and the Johns Hopkins School of Public Health provided additional financial support for the event.