In the United States, a relatively small proportion of complex patients—defined as having multiple comorbidities, high risk for poor outcomes, and high cost—incur most of the nation's health care costs. For these patients especially, fragmentation and poor coordination of care across settings and providers can lead to unnecessary spending on redundant laboratory testing, repeated imaging, and avoidable emergency department (ED) visits and hospitalizations. Improved care coordination and management of complex patients could reduce costs while increasing quality of care. However, care coordination efforts face multiple challenges, such as segmenting populations of complex patients to better match their needs with the design of specific interventions, understanding how to reduce spending, and integrating care coordination programs into providers' care delivery processes.
Innovative uses of analytics and health information technology (HIT) may address these challenges. Analytics are predictive algorithms that use various types of data and may help create better risk stratification approaches that more effectively target patients for interventions. HIT includes tools that may facilitate communication and improve timely decisionmaking, particularly because patients with complex needs tend to have large care teams and generate substantial volumes of data during their care. As new payment models spread, there is increasing interest in predicting and managing care and its costs, for complex patients in particular.
This project reviewed the literature and held discussions with subject matter experts (SMEs) to understand how analytics and HIT are being used to identify and support the coordination of care for complex patients. Our goal was to summarize emerging evidence and best practices that can inform the development and dissemination of more-effective analytics, HIT functionalities, and care models to meet the needs of complex patients.
To better understand the state of knowledge and to inform the SME discussions, we conducted a targeted literature review with the goal of identifying analytics projects that use data and algorithms to find complex patients as part of an intervention and HIT functionalities designed to facilitate care coordination and communication among providers caring for the same patient. We developed a conceptual framework to guide this search: We searched multiple research databases, including PubMed, Web of Science, and SCOPUS, and limited the search to articles published from 2008 to the present. In total, 122 articles were selected for abstraction. One reviewer captured information from the articles using a literature abstraction form based on our conceptual framework, and the research team reviewed that information.
We identified SMEs through the literature review, recommendations from colleagues, and snowball sampling. From this initial list, we selected 35 SMEs to represent a wide range of stakeholder perspectives from the private sector, academic institutions, and federal agencies. We conducted in-depth discussions ranging from 30 to 60 minutes using a discussion guide that identified key topics to be covered. Midway through conducting these discussions (in January 2015), we convened a technical expert panel to discuss the results of the SME discussions to date and of the literature review. Based on feedback from the panel, we further refined our plans for the next phase of SME discussions.
Based on the SME discussions, we identified key themes, including technology goals and barriers and opportunities for progress, and formulated recommendations for how to advance analytic and care coordination functionalities further to better meet the needs of complex patients and their care teams.
Because our review of the peer-reviewed literature used restrictive criteria, we found few papers that addressed analytics and HIT functionalities for complex patients, a result consistent with a previous review on complex patients.
Based on discussions with SMEs, we characterized the purpose of the analytics focused on complex patients into three distinct goals: (1) identify complex patients; (2) identify the subset of complex patients who could be helped by an intervention; and (3) match the subgroups of complex patients to specific interventions. We also identified an additional crosscutting goal to improve the ability to predict the onset of complexity earlier in time so the health care system can intervene preemptively in disease progression.
We found that most efforts sought to address the first goal and (to some extent) the crosscutting goal, while few attempted to address the second and third goals. While SMEs suggested that a growing number of organizations are prioritizing and investing in analytics to identify complex patients, models have limited effectiveness and lack evidence of impact.
SMEs described numerous barriers to progress in analytics, with the foremost among them being those related to data—particularly poor data quality and lack of data related to social determinants of health. SMEs also cited lack of experience using analytics as a challenge for both providers and care coordinators. A third barrier related to financial incentives: While reimbursements are shifting slowly and beginning to reward providers and organizations that identify and treat complex patients, even emerging payment models might not create the right incentives to prioritize devoting resources to the sickest patients, because the return on investment is unclear.
We identified five key HIT functionalities in current use or being piloted for care coordination: dashboards, patient relationship managers, event alerts, referral tracking, and care plans. Dashboards support ad hoc searches and prompt discussions among care team members. Patient relationship managers allow care coordinators to manually track interactions with the patient and manage the patient's to-do list. Event alerts are triggered by the ED visit, hospitalization, or other events and sent to members of a care team. Referral tracking helps to ensure referrals happen and that the summary reports are returned to the referring provider. SMEs suggested that the development and use of referral-tracking functionality was increasing and that the referral-tracking requirement for patient-centered medical homes could be driving the development of this functionality.
Functionalities for care plans—which are designed to communicate instructions for a patient's care—varied widely. Some care plans offered task-tracking capabilities and various kinds of communication, such as one that allowed ad hoc communication similar to social media. We did not identify consistent definitions of care plan contents, which ranged from static text describing physician instructions to highly structured content accessed by multiple users based on a set of discretely coded problems. Most care plan functionality targeted care coordinators as the primary users; few efforts engaged physicians as users. Such responsibilities as updating the care plan varied from giving the care coordinator exclusive control to allowing for broader permission that included others on the care team, including patients.
Barriers to further developing care coordination functionality, especially care plans, were substantial and included: unclear definitions of what it means to be a member of a care team; lack of concepts, frameworks, or understanding of what activities are involved in care coordination and should be best facilitated using HIT; and lack of interoperability between care coordination products and electronic health records. SMEs also discussed the challenge of establishing a sustainable business model for developing and using these functionalities, because the move toward accountable care is proceeding slowly.
We summarize challenges that must be addressed for the success of future work in both analytics to identify complex patients and HIT functionalities to coordinate care. For analytic models to be useful, issues of poor data quality and lack of use of novel data types must be addressed. Making better use of model outputs by integrating predictive model output into clinical workflows is also needed. For work on HIT functionalities for care coordination, existing functionality that supports care plans and communication among care teams has important limitations, and interoperability between care coordination systems and other HIT software is lacking. Also, the lack of evaluative studies suggests that best practices for using any of these functionalities are unknown.
Based on the findings from this work, we propose research options to consider when addressing these challenges as part of a five-part framework: (1) understand the problem and barriers to progress; (2) develop technology and related process and workflow changes; (3) evaluate and generate evidence of impact; (4) implement and disseminate technology and related process and workflow changes; and (5) create incentives that promote the use of technology and related process and workflow changes. In the near term, there is a need for more knowledge about best practices and the need for a convener to bring together and align key stakeholders to accelerate innovation. Longer-term efforts will need to focus on training providers and integrating these technical advances into clinical practice.