High-Risk Patients and Utilization of Primary Care in the US Veterans Affairs Health System | Health Care Delivery Models | JAMA Network Open | JAMA Network
[Skip to Navigation]
Sign In
Invited Commentary
Health Policy
June 29, 2020

High-Risk Patients and Utilization of Primary Care in the US Veterans Affairs Health System

Author Affiliations
  • 1Division of Healthcare Delivery Science & Innovation, Department of Population Health Sciences, Weill Cornell Medicine, New York
JAMA Netw Open. 2020;3(6):e209518. doi:10.1001/jamanetworkopen.2020.9518

Data are the new gold. This aphorism often describes the role of data-centric, platform-based industries, such as Google, Amazon, and Facebook, in the US $3 trillion health care industry. Big data analytic technologies such as machine learning, data science, and artificial intelligence (AI) are inspiring algorithm-based decision-making that promise to transform health care from a service industry to an information industry. Big data analytics derive power and potential value from the near real-time aggregation and analysis of vast volumes of data mined from multiple sources, creating predictive models, and providing actionable information to inform health care decisions. Big data are both a product and a function of technology and ever-growing analytic and computational power.

In health care, an area seeing growing utilization of data and predictive modeling is risk stratification for care management such as the Medicare Comprehensive Primary Care Plus Initiative.1,2 For several reasons, the US Veterans Affairs (VA) health care system is uniquely prepared to lead innovative applications of data for care management. The VA is one of the largest integrated health care systems in the country with an electronic medical record that spans the entire system and allows for large-scale analysis and modeling to predict care needs and allocate resources. The VA has a relatively high-complexity patient population, owing in part by social determinants of health. The VA is also the closest the US has to an equal access health care system that is not burdened with the fee-for-service payment model that incentivizes volume over quality of care. Lastly, the implementation of patient-centered medical home models throughout the VA health care system gives the VA the ability to rigorously examine the use of big data and analytics in patient risk stratification and care management in primary care.

Accordingly, Chang et al3 provide an elegant analysis of VA big data to examine care management of high-risk patients in primary care. The authors define high-risk based on a validated predictive risk score, the Care Assessment Need (CAN) score, which is calculated on a weekly basis and used to predict risk of hospitalization within 90 days. Based on this score, the top 5% of patients (high-risk) have about a 20% risk of hospitalization within 90 days. For this analysis, the study sample included all VA patients assigned to primary care as of September 2015 (N = 4 309 192). The authors hypothesized that, in comparison with low-risk patients, high-risk patients would use more face-to-face care and less secure messaging; that at least 50% of them would be assigned to specialized primary care programs rather than general primary care; and that those assigned to general primary care would have a higher utilization rate of primary care and medical subspecialty care. In this study, specialized primary care refers to programs such as geriatric, HIV, homeless, and women’s health clinics.

The major findings of the study are the following.3 First, the authors found that male sex, marital status (not married), older age, and race (African American) were associated with higher likelihood of being identified as a high-risk patient. They also found that high-risk patients, compared with low-risk patients had higher likelihood of face-to-face encounters, telephone encounters, and use of secure messaging during the year prior to being identified as high risk. The authors also found that high-risk patients, compared with low-risk patients, had more hospitalizations and emergency department utilization. Most high-risk patients (88%) were assigned to regular primary care instead of specialized primary care clinics. High-risk patients who were assigned to specialized primary care clinics had more outpatient visits. On the other hand, high-risk patients assigned to general primary care had more face-to-face encounters and medical specialist use, but fewer emergency department, mental health, and add-on service (eg, telemedicine) use.3

There are important factors to consider in interpreting these findings. First, as Chang et al3 stated, specialized primary care models, which are intended to benefit high-risk patients, are generally associated with high resource utilization4,5 and are geographically concentrated in urban areas.6 It is therefore not surprising that this analysis found that even within the VA, most of the specialized primary care sites are often located in academic VA medical centers in urban areas. Most general primary care clinics are in community-based sites known as community-based outpatient clinics. The geographic differences in the distribution of care sites confound the comparison of general primary care clinics to specialized primary care clinics.

Another issue that often challenges analyses such as this is the definition of high-risk. The definition depends on whose risk we are concerned with—the patient, health care system, insurance company, or society. The authors addressed this challenge by focusing on the health care system as the key stakeholder of interest. If the definition were based on the perspective of a different stakeholder, such as the insurance company (which is less of an issue for VA insured patients), some of the assumptions that went into the risk score calculation model could have raised concerns about possible overrepresentation of African American patients in the high-risk category. This concern arises in part because of recent reports of how health care data sets used for predictive modeling might be biased against minority patients.7

These minor issues notwithstanding, this is a well-designed and effectively articulated application of big data and analytics in defining high-risk patients, identifying factors that determine the risk score, and in evaluating how high-risk patients are distributed across primary care programs in one of the largest integrated health care systems in the US. The study by Chang and colleagues3 points out that there is room for improvement in assigning high-risk patients to specialized primary care programs, and shows why consideration be given to appropriately resourcing general primary care clinics to optimize care for high-risk patients.

Back to top
Article Information

Published: June 29, 2020. doi:10.1001/jamanetworkopen.2020.9518

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 Ibrahim SA. JAMA Network Open.

Corresponding Author: Said A. Ibrahim, MD, MPH, MBA, Department of Population Health Sciences, Weill Cornell Medicine, 402 E 67th Street, Floor 2 Room LA 215, New York, NY 10065 (sai2009@med.cornell.edu).

Conflict of Interest Disclosures: None reported.

Funding/Support: Dr Ibrahim is supported in part by a K24 Mid-Career Development Award from the National Institute of Arthritis and Musculoskeletal and Skin Diseases (grant K24AR05 5259).

Role of the Funder/Sponsor: The funder had no role in the analysis and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Rothman  AA, Wagner  EH.  Chronic illness management: what is the role of primary care?   Ann Intern Med. 2003;138(3):256-261. doi:10.7326/0003-4819-138-3-200302040-00034PubMedGoogle ScholarCrossref
Rich  E, Lipson  D, Libersky  J, Parchman  M. Coordinating care for adults with complex needs in the patient-centered medical home: challenges and solutions. White paper. Agency for Healthcare Research and Quality. 2012. Accessed April 27, 2020. https://pcmh.ahrq.gov/sites/default/files/attachments/Coordinating%20Care%20for%20Adults%20with%20Complex%20Care%20Needs.pdf
Chang  ET, Zulman  DM, Nelson  KM,  et al.  Use of general primary care, specialized primary care, and other Veterans Affairs services among high-risk veterans.   JAMA Netw Open. 2020;3(6):e208120. doi:10.1001/jamanetworkopen.2020.8120Google Scholar
Porter  ME, Pabo  EA, Lee  TH.  Redesigning primary care: a strategic vision to improve value by organizing around patients’ needs.   Health Aff (Millwood). 2013;32(3):516-525.PubMedGoogle ScholarCrossref
Hochman  M, Asch  SM.  Disruptive models in primary care: caring for high-needs, high-cost populations.   J Gen Intern Med. 2017;32(4):392-397.PubMedGoogle ScholarCrossref
Hong  CS, Siegel  AL, Ferris  TG.  Caring for high-need, high-cost patients: what makes for a successful care management program?   Issue Brief (Commonw Fund). 2014;19:1-19.PubMedGoogle Scholar
Obermeyer  Z, Powers  B, Vogeli  C, Mullainathan  S.  Dissecting racial bias in an algorithm used to manage the health of populations.   Science. 2019;366(6464):447-453.PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words