Abstract
PURPOSE Despite evidence suggesting that high-quality primary care can prevent unnecessary hospitalizations, many primary care practices face challenges in achieving this goal, and there is little guidance identifying effective strategies for reducing hospitalization rates. We aimed to understand how practices in the Comprehensive Primary Care Plus (CPC+) program substantially reduced their acute hospitalization rate (AHR) over 2 years.
METHODS We used Bayesian analyses to identify the CPC+ practice sites having the highest probability of achieving a substantial reduction in the adjusted Medicare AHR between 2016 and 2018 (referred to here as AHR high performers). We then conducted telephone interviews with 64 respondents at 14 AHR high-performer sites and undertook within- and cross-case comparative analysis.
RESULTS The 14 AHR high performers experienced a 6% average decrease (range, 4% to 11%) in their Medicare AHR over the 2-year period. They credited various care delivery activities aligned with 3 strategies for reducing AHR: (1) improving and promoting prompt access to primary care, (2) identifying patients at high risk for hospitalization and addressing their needs with enhanced care management, and (3) expanding the breadth and depth of services offered at the practice site. They also identified facilitators of these strategies: enhanced payments through CPC+, prior primary care practice transformation experience, use of data to identify high-value activities for patient subgroups, teamwork, and organizational support for innovation.
CONCLUSIONS The AHR high performers observed that strengthening the local primary care infrastructure through practice-driven, targeted changes in access, care management, and comprehensiveness of care can meaningfully reduce acute hospitalizations. Other primary care practices taking on the challenging work of reducing hospitalizations can learn from CPC+ practices and may consider similar strategies, selecting activities that fit their context, personnel, patient population, and available resources.
- hospitalization
- emergency department
- health care use
- ambulatory care–sensitive conditions
- chronic disease
- primary care
- Medicare
- older adults
- health services for the aged
- vulnerable populations
- access to health care
- patient care management
- comprehensive health care
- change, organizational
- quality improvement
- practice-based research
INTRODUCTION
Substantial evidence demonstrates that many acute hospitalizations in the United States could be avoided by providing patients with timely access to high-quality primary care.1-6 Yet many primary care practices and primary care practitioners (PCPs) face challenges achieving this goal. Beset by increasing patient complexity and administrative burdens, PCPs also face fee-for-service payments insufficient to support their efforts to deliver high-quality care that is accessible, continuous, coordinated, and comprehensive.7-10
To help address such challenges, in January 2017, the Centers for Medicare & Medicaid Services launched Comprehensive Primary Care Plus (CPC+).11 This program gave primary care practice sites financial resources and technical assistance, and promoted regionally based payment reform and primary care transformation to improve quality of care and achieve better health outcomes at lower cost.12,13 More than 3,000 sites participated in CPC+.
Because hospital spending accounts for 41% of annual Medicare Part A and B costs,14 even a modest reduction in the acute hospitalization rate (AHR) could yield savings. There is little evidence, however, identifying effective strategies for reducing avoidable hospitalizations across diverse primary care practice settings and patient populations. As part of the CPC+ independent evaluation, we therefore examined how practices that succeeded in reducing AHR did so.
METHODS
We used Bayesian analyses to identify the CPC+ practice sites—single physical locations where patients are served—with the highest probability of achieving substantial reductions in adjusted Medicare AHR over time. We then conducted telephone interviews and a within- and cross-case comparative analysis of 14 of these primary care practice sites (hereafter referred to as AHR high performers).
AHR High Performer Identification
We defined AHR as the number of hospitalizations at short-stay acute hospitals and critical access hospitals per 1,000 Medicare beneficiaries per year. This AHR measure included emergency department (ED) visits and observation stays if they resulted in an inpatient admission; we excluded hospitalizations for elective surgery and planned procedures.
We used Medicare claims and enrollment data and Bayesian modeling to estimate the probability that each CPC+ practice site would achieve a true reduction in its adjusted AHR that was substantial (at least 5% larger than the average of all CPC+ practice sites). To guard against compositional changes in a practice’s case mix creating the appearance of improvement that did not reflect actual practice transformation, the model adjusted for a range of patient, practice, and market characteristics; to guard against random chance creating the appearance of improvement, especially in small practices, the model shrinks the AHR estimates for all practices toward the mean, with greater shrinkage for small practices (Supplemental Appendix 1).
A total of 2,888 primary care practice sites joined CPC+ in 2017. Using the above methodology, we identified the 25 sites having the highest probability of achieving a substantial reduction in their adjusted AHR between 2016 (the year preceding CPC+) and 2018 (the second year of CPC+ participation).
Interviews
From February to December of 2020—the fourth year of CPC+ implementation and the first year of the COVID-19 pandemic—we conducted telephone interviews at 14 of the 25 identified AHR high-performer practice sites (the other 11 sites declined to participate). We conducted an initial 60-minute interview with 2 or 3 practice or system leaders at each of the 14 sites. We used a grounded theory approach, asking the same open-ended questions in each interview to identify the factors (care delivery activities, practice characteristics, and community context) that respondents perceived as influencing AHR reductions.15 We then conducted 60-minute follow-up interviews with 1 to 6 staff individually at 9 of the 14 sites to gather detail (the other 5 sites declined to participate). We customized follow-up interviews based on findings from the initial interviews and the respondent’s role at the practice. Across the 14 sites, we interviewed 64 respondents: 19 physicians, 14 practice administrators, 10 system-level leaders, 10 care managers, and 11 other practice staff (eg, nurses, pharmacists) (Supplemental Appendix 2).
Health services researchers (D.M.P., L.F., R.M.M., V.P., A.S., R.S., and S.H.) conducted all interviews for assigned AHR high performers, each paired with a physician with primary care research experience (A.S.O., D.R.R., R.E.P., and E.C.R). We acquired verbal consent, and recorded and transcribed the interviews.
Qualitative Analysis
Within- and cross-case analysis proceeded in stages.16,17 After completing interviews, we drafted a case report for each AHR high performer based on interview notes. We then coded interview transcripts using NVivo version 12 (QSR International) using codes aligned with interview questions and open coding to capture factors influencing AHR. We met weekly to resolve coding discrepancies and revise codes. We used coded data to finalize case reports. We then scored the influence of factors present in each report from 0 (not contributing to AHR reduction) to 3 (major contributor to AHR reduction). Scoring took into consideration respondents’ perceptions, when the factor was introduced or modified, and the proportion of patients potentially influenced by the factor. The assigned researcher (D.M.P., L.F., or V.P.) and all physicians reviewed each case report and independently scored factors. We held a series of meetings to reach consensus on final scores.18 We entered scores and substantiating data into a matrix with AHR high performers as columns and factors as rows.
Three authors (D.M.P., L.F., and V.P.) used the matrix to detect similarities and differences across the cases, merge and distinguish concepts, identify factors present for 4 or more cases, and generate findings.19 They met weekly to reach consensus on factors and referred back to transcripts and coded data as needed. To check that variation in the number of interviews conducted across AHR high performers did not bias results, we compared results from the 11 cases having multiple interviews with the results from the 5 cases having a single interview. Factors present at the 5 practices were represented in the sample of 11, although fewer factors emerged overall for the cases with less data.
After the analysis was complete, we conducted 3 virtual panels with 17 staff from 12 AHR high performers to confirm that our findings aligned with their perceptions and to discuss implications.
The New England Institutional Review Board granted the study an exemption from review.
RESULTS
The 14 AHR high performers experienced a 6% average decrease in the Medicare AHR between 2016 and 2018, in contrast to an average increase of 5% in the CPC+ practices that did not meet the criteria for probability of high performance. Table 1 displays selected characteristics for each participating AHR high performer.
Selected Characteristics of AHR High Performers at Baseline, 2016
Consistent with our purposive selecting of practices with highest probability of reduction in AHR, these practices differed from the full set of CPC+ practices (Table 2). The 14 participating AHR high performers were larger, employed more practitioners, and served more fee-for-service beneficiaries than did CPC+ practices overall and thus received larger CPC+ payments. All 14 had primary care transformation experience. AHR high performers also served patients with slightly higher medical complexity. They were more likely to be located in rural areas and in the western United States. AHR high performers’ counties had more acute care hospital beds than those of CPC+ practices overall.
Comparison of AHR High Performers With All CPC+ Practices at Baseline, 2016
Activities Perceived as Reducing AHR
Our analysis of factors revealed 8 care delivery activities that AHR high performers perceived as reducing their AHR between 2016 and 2018. The activities aligned with 3 overarching strategies: improve access to primary care, expand care management, and increase comprehensiveness of care. Respondents perceived each strategy to increase their practice’s capacity to meet patients’ needs in a timely fashion, providing an alternative to ED or hospital care. Each AHR high performer used a combination of activities within and across strategies, and attributed varying levels of influence to each activity on their AHR. Table 3 shows the prevalence of activities within strategies across AHR high performers. We discuss findings for the 3 strategies below and provide illustrative quotes in Table 4.
Prevalence and Perceived Level of Contribution of Activities (Within Strategies) to Reduce Acute Hospitalizations Within and Across AHR High Performers
Hypothesized Pathways and Illustrative Quotes for Activities (Within Strategies) to Reduce Acute Hospitalizations at AHR High Performers
AHR Reduction Strategies and Associated Activities
Improve Access to Primary Care
AHR high performers reported that improving access to primary care, combined with promoting patients’ knowledge of the importance of access and how to access primary care, increased their likelihood of addressing patients’ concerns quickly.
Many AHR high performers said they increased the number of same-day visits, encouraging patients to see a PCP for urgent needs or concerns (thereby avoiding ED visits). AHR high performers that hired staff to provide same-day visits (either nurse practitioners who explicitly focused on same-day visits or PCPs who did not yet have full patient panels) were able to expand access to more patients than those that added same-day slots to existing practitioners’ schedules. A few AHR high performers added staff to increase access by lengthening their hours of operation during the week and on weekends, and one shared that weekend hours were “more impactful [on AHR] than longer days.”
A few AHR high performers increased timely telephone access to the practice by providing high-risk patients with the telephone number of a care manager. They noted that care managers’ familiarity with the patient helped to rapidly address patient needs or connect the patient with the PCP.
Some health system–affiliated AHR high performers perceived their AHR improvements were achieved through system-owned urgent care centers providing patients an alternative to the ED when PCPs were not readily available. They noted that these system-affiliated centers had access to patients’ information and could contact PCPs to schedule primary care follow-up appointments through shared health information technology, unlike independently operated urgent care centers.
AHR high performers proactively promoted the use of primary care (through verbal and written communication, posters, and portal messages) as an alternative to the ED for managing new or worsening concerns.
Expand Care Management
Most AHR high performers credited the expansion of their care management with helping to reduce AHR. By identifying patients at high risk for ED or hospital use and addressing patient needs with focused outreach to supplement traditional PCP visits, AHR high performers perceived they were able to avert hospitalizations by intervening earlier in the course of illness.
Most AHR high performers followed up with patients within 48 hours of a hospital discharge to provide information and linkages to primary care and thereby prevent additional hospitalizations. They called patients to check on their health; review medications; answer questions; provide disease-specific education; connect them to needed supports (eg, medical equipment, social services); and schedule follow-up appointments with the PCP. AHR high performers instituted or expanded these efforts during the first 2 years of CPC+ by hiring or redeploying staff to this role. Many AHR high performers extended these efforts to patients who visited the ED or experienced observation stays. Various AHR high performers perceived that follow-up calls were most effective when made by care managers who had specific skills (eg, nursing or social work background, ability to build rapport, empathy) and who used purposeful processes (eg, reviewing discharge reports to prepare for calls, asking questions, and following through to ensure patients’ needs were met)—in contrast to automated calls or calls by less-skilled staff focused on scheduling follow-up appointments. Follow-up calls were especially effective when care managers making the calls were connected to a care team. Receipt of complete and timely information from discharging facilities, and, in a few cases, the discharging hospital scheduling the patients’ follow-up appointments with their PCP, enabled their work.
Many AHR high performers credited long-term care management as contributing to improvements in AHR. Although strategies varied across AHR high performers, long-term care management consistently involved continuous relationship-based support outside of PCP visits that was matched to patients’ needs, conditions, and abilities. AHR high performers added staff to provide these services to additional patients at the practice site. They noted that care managers were most effective when they knew how to prioritize patients, were skilled problem solvers, and could build trust with patients and PCPs. To enroll the patients at highest risk for hospitalization, several AHR high performers used enhanced risk score algorithms and/or developed capabilities to detect frequent users of the hospital and ED.
Several AHR high performers developed specialized care management programs at their site, targeting subgroups of patients based on condition prevalence in their population, as well as available practice resources, and likely intervention effectiveness. For example, one monitored the frequency of albuterol refills among patients with chronic obstructive pulmonary disease as an early indicator of higher risk of disease exacerbation. Another AHR high performer developed an outpatient program to better manage care of people with sickle cell disease, which reduced admissions for its complications.
Increase Comprehensiveness of Care
Many AHR high performers perceived that increasing the comprehensiveness of care helped reduce AHR by better managing new conditions and preventing exacerbations of patients’ chronic conditions. These AHR high performers expanded the breadth of services offered at the practice site to treat patients’ range of needs. Examples of new or enhanced services included behavioral health, social work, and enhanced medication management. As one PCP described the influence of broadening the practice’s capabilities on AHR, “It’s all together. It’s everybody, truly all-hands-on-deck wrapping ourselves around; we all bring something to the table that’s different. It’s synergistic.”
Several AHR high performers described using team-based care to allow PCPs to spend more time with complex patients to better understand their needs and assess their health concerns, increasing the breadth and depth of care provided. To accomplish this, one AHR high performer used advanced practice clinicians to manage patients with straightforward issues so that physicians could reserve time for those with more complex health conditions. Other AHR high performers shifted staff roles to help PCPs be more comprehensive; for example, medical assistants took on advanced activities such as reviewing medications, identifying gaps in care, and working as scribes.
Facilitators of AHR Reduction Strategies
Our analysis also identified practice characteristics that facilitated AHR high performers’ ability to implement the 3 AHR reduction strategies. Table 5 describes 4 facilitators that were present across all or most AHR high performers: experience with transformation efforts, use of data, a team-based approach, and interest in innovation.
Facilitators of AHR High Performers’ Efforts to Prevent Acute Hospitalizations
DISCUSSION
The AHR high performers achieving a substantial 2-year reduction in Medicare AHR described a variety of activities they perceived as preventing unnecessary hospitalizations. The activities they perceived as most helpful align with 3 strategies: (1) promoting timely access to primary care, (2) identifying patients at high risk for hospitalization and addressing their needs with enhanced care management, and (3) expanding the breadth and depth of services offered at the primary care practice site. These activities also align with 3 of the defining elements of advanced primary care—accessibility, care coordination (including coordinating transitions of care and managing chronic conditions), and comprehensiveness2,10,20—that all have been shown to be associated with reduced hospitalizations.1,4,21-32
Although many AHR high performers perceived that similar activities reduced AHR, no 2 used the same combination of activities. All AHR high performers leveraged available human and financial resources, chose strategies based on local circumstances and priorities, and dedicated additional staff resources to the selected activities. They used staff with relevant training and commitment, supported staff with a robust care team, and used data to identify the highest-value activities (including identification of patient subgroups). Our analysis also points to the importance of taking advantage of opportunities to innovate and building on prior experience. Our findings may help practices choose a starting point for reducing AHR that matches their patient population, practice capabilities, and resources, and may encourage these practices to try out new activities, learn from them, and continue to transform. AHR high performers’ perceptions of activities most beneficial to AHR reduction are especially relevant for practices participating in Primary Care First, which rewards reduced hospital use while giving practices flexibility in the care delivery innovations used to achieve this outcome.33
Our study has limitations. First, it was designed to gather rich insights and detailed examples, not to provide generalizable findings. Second, the data might be subject to recall bias because we asked respondents to consider activities that occurred between 2016 and 2018, that is, 2 to 4 years before our interviews. Also, the findings are based on respondents’ perceptions of activities that reduced AHR. Finally, reducing health care expenditures (to which hospitalizations contribute the largest share) is a national policy priority, and a key desired outcome for CPC+; thus, we focused on AHR to identify AHR high performers for this study. Although it is important to reduce potentially preventable hospitalizations through better primary care, some people may have unmet need for hospitalizations. Reducing the rate of acute hospitalizations is therefore at best only one aspect of successful performance by primary care practices. Ideally, future research could focus on other dimensions of high-quality primary care as important outcomes of interest. As noted in a recent report by the National Academies of Sciences, Engineering, and Medicine, “primary care is the only health care component where an increased supply is associated with better population health and more equitable outcomes. For this reason, primary care is a common good, making the strength and quality of the country’s primary care services a public concern.”10
Our findings suggest that the AHR can be meaningfully reduced by strengthening the local primary care infrastructure through practice-driven, targeted changes in access, care management, and comprehensiveness of care. Other primary care practices taking on the challenging work of reducing hospitalizations can learn from AHR high-performer practices in the CPC+ program and may consider similar strategies, selecting activities that fit their context, personnel, patient population, and available resources.
Acknowledgments
Nancy T. McCall, a senior fellow at Mathematica and Project Director of the evaluation of CPC+, critically reviewed the manuscript. Debbie Peikes, former senior fellow at Mathematica and former Project Director of the evaluation of CPC+, provided substantial support to the design of this study. Arnold Chen and Jan Genevro, senior researchers at Mathematica, supported data collection activities, and Mario Gruszczynski, James Burnham, and Ify Obi, analysts at Mathematica, supported recruitment.
Footnotes
Conflicts of interest: authors report none.
Funding support: This study was funded by the US Department of Health and Human Services, Centers for Medicare & Medicaid Services under contract HHSM-500-2014-00034I/HHSM-500-T0010.
Disclaimer: The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the US Department of Health and Human Services or any of its agencies.
- Received for publication June 21, 2022.
- Revision received February 7, 2023.
- Accepted for publication February 24, 2023.
- © 2023 Annals of Family Medicine, Inc.