Determining Patient Panel Size in Primary Care: A Meta-Narrative Review

Abstract

The optimal patient panel size (PPS) in primary care and the factors determining it remain unclear. We conducted a meta-narrative review of the literature to evaluate factors influencing PPS and assess its association with patient outcomes. A comprehensive search of electronic databases was performed from inception through December 2023, focusing on original studies reporting factors used to determine PPS and related outcomes (eg, clinical outcomes, process measures, and resource utilization). A total of 48 studies were included, identifying 7 key factors influencing PPS. Smaller panels were associated with improved patient satisfaction, continuity of care, and health promotion, while clinical outcomes, utilization, and costs showed minimal impact by PPS. Panel size was primarily associated with patient age, sex, comorbidities, and practice type and structure. Community-based centers typically managed smaller panels, often staffed by female clinicians and serving socioeconomically disadvantaged populations with greater health needs than hospital-based practices. Female clinicians were also independently associated with managing smaller panels, higher quality care indicators, fewer emergency department visits, and improved patient satisfaction. Determining the ideal PPS is a multifaceted process influenced by practice setting, patient demographics, and clinician characteristics. While practice-related factors showed limited association with PPS, patient-reported outcomes were more closely linked to it. Primary care practices should tailor panel sizes to their patient populations, emphasizing a patient-centered approach and ensuring adequate infrastructure support to optimize care delivery.

Introduction

Patient panel size (PPS) refers to the number of patients under the care of a single clinician. A primary care clinician’s panel serves as a formalized link, a long-term bond of care between a clinician and their patients. Determining appropriate PPS is challenging for primary care practices due to ongoing ramifications of excessive PPS. ¹ Challenges in determining appropriate PPS are compounded by a combination of factors, including clinician shortages, increasing demand for primary care services from an aging population, rising numbers of individuals seeking care, and variability in practice infrastructure and patient needs.^1,2 Ultimately, the quality of health care delivery is threatened by a delay in patient access to care, which is often due to clinicians handling larger patient panels than they can manage.^3,4 Larger PPS may also contribute to physician burnout, which further augments the problem.^5,6 Understanding what constitutes an ideal PPS for individual practices may help improve patient satisfaction, continuity of care, clinical outcomes, quality of care, and access, while creating a fair and equitable distribution of workload for clinicians.

Consideration has been given to various strategies to alleviate the current primary care clinician shortage. One proposal has been to train more physicians, but Bodenheimer et al ⁷ suggest this may not be the only solution. Empowering other health care personnel such as nurses, pharmacists, and medical assistants to be more involved in patient care could reduce the burden on physicians and allow them more time to care directly for patients. Indeed, up to a quarter of a physician’s time may be saved by delegating some aspects of patient care to licensed and non-licensed personnel.^{7 -9} The idea of transitioning to a team-based care model, in which subpanels of patients are assigned to other clinical and nonclinical staff, would allow for the opportunity to increase physician PPS, while using other resources to achieve maximum efficiency and patient and care teams satisfaction.^10,11 Furthermore, exercising PPS control may help primary care practices provide preventative, acute, and chronic care to more patients, while improving patient and physician satisfaction.^12,13

“What is the optimal PPS a primary care clinician should care for?” is a complex question thought to be influenced by various factors, including patient demographics, physician characteristics, practice infrastructure, regulatory organizations, payers requirements.^4,7,14 Attempts to answer this question have resulted in a body of research around methods and formulas to determine PPS (eg, visits per patient per year, provider capacity per year), but these methods face discrepancies that include miscounting, duplicating, and overestimating.^15,16 This leads to the question, “What factors determine the ideal primary care PPS?” To address these questions, we performed a systematic review with the aim of synthesizing up-to-date evidence about PPS in primary care practices to determine whether the PPS is associated with differences in patient-centered and health care delivery–related outcomes. We further aimed to report which predictive factors should be considered when determining an appropriate PPS.

Methods

Literature Search

A medical librarian conducted a comprehensive literature search with input from senior study investigators (A.M.A.D., M.H.M.) with experience in systematic reviews. The search was performed in 6 electronic databases: MEDLINE, EMBASE, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, CINAHL, and Scopus from database inception through December 2023. We included original prospective and retrospective studies that reported various primary care PPSs and the estimates used to create them. Studies using predictive models to ascertain PPS were also included, as were studies reporting relevant outcomes and their relationship to PPS (Table 1). To ensure a comprehensive search, we included international studies and expanded the search to include additional search terms (eg, caseloads, managed panel). Study team members also searched references for additional studies and updated the original literature search.

Table 1.

PICO Criteria and Outline of Variables of Interest.

Population	Studies reporting panel size (per practice or individual) in primary care
Comparison	Studies reporting a different panel size in a comparable primary care practice
Outcomes	• Indicators of quality: Patient satisfaction, clinician satisfaction, accessibility, continuity/disruption of care (eg, wait time, scheduling) • Clinical outcomes
Subgroup analyses (variables of interest)	• Health care system/practice size • Specialty (eg. internal medicine, pediatrics, family medicine) • Clinician characteristics (eg, type of practice, age, knowledge of their panel size) • Methodology used to determine panel size • Practice type/style/structure (eg, fixed size, fixed salary, team support/composition) • Practice infrastructure (eg, EHR, data systems) • Health coverage • Demographic and population type influence (eg, age, sex, geographic location, SES, geographic location) • Complexity of conditions (eg, type of patients) • Other subspecialties

Abbreviations: EHR, electronic health record; PICO, patient, problem or population, intervention, comparison, control or comparator, outcome(s); SES, socioeconomic status.

Study Eligibility

The search criteria included studies that explored the impact of any factors on PPS. We also included articles that associated patient care outcomes with PPS. After excluding duplicates, we identified original studies eligible for further review by screening abstracts and titles. If a study was deemed relevant, the manuscript was obtained and reviewed in full-text version for further assessment by the reviewers. All steps were done in duplicate, independently and blindly by 2 reviewers.

Data Extraction

We extracted data in duplicate and blindly on various study characteristics, including author, year of publication, aim, study design, setting, practice and PPS characteristics, clinical outcomes, and estimated effects of PPS. We also extracted factors associated with predicting PPS, including clinician and patient characteristics and comorbidities.

Risk of Bias Assessment

We used the Newcastle-Ottawa Scale (NOS) to appraise the risk of bias (methodologic quality) in the included observational longitudinal studies and a modified 3-item NOS for cross-sectional and case series studies.

Data Synthesis

Across the included studies, we observed considerable heterogeneity in study methodology, outcome reporting, and methods for determining PPS and predicting factors. Therefore, we chose to use a metanarrative approach rather than pursue conventional systematic review and meta-analysis. The study team members (A.M.A.D., W.H.F., H.M., P.B.) synthesized the narrative evidence according to the RAMESES guidelines, ¹⁷ a proposed guiding analytic framework, by mapping out benchmarks or predictors in studies that reported various PPSs and the factors that influenced these PPSs in their respective primary care practices as shown in Figure 1.

Figure 1.

Proposed analytic framework to explore patient panel sizes in primary care.

Results

The initial search yielded 1329 studies. After abstract screening, full-text reviewing, and additional reference searches, a total of 48 studies were identified meeting the inclusion criteria (Figure 2). The studies were conducted in or presented data from various countries, including the US (30), UK (11), Canada (7), and Spain (1). Thirty-one studies examined PPS and care delivery–related outcomes (Table 2)^{12,18 -47} and 20 investigated predictive factors of PPS (Table 3).^{10,26,27,41,42,48-62} Three studies^26,27,41 analyzed both (included in Tables 2 and 3). Across all data, 32 panels were analyzed with a median size of 1824 patients (IQR: 1493-2260), while panel sizes ranged from 265 to 13 147 patients. In US-based studies, which included 14 panels, the median panel size was 2263 patients (IQR: 1888-2504), with a similar range of 393 to 13 147. Reporting on solo and group practices, including their sizes, was limited and inconsistent across studies. Family medicine practices were the primary focus in both datasets (~65%), while community internal medicine practices had moderate representation in US studies. Pediatric-focused practices were least reported.

Figure 2.

Flowchart of study selection.

Table 2.

Summary of Included Studies on Panel Size and Care Delivery–Related Outcomes.

Study ID (study design)	Study practice setting Practice location	Panel description	Conditions studied/scope
Abdelhamid et al, 2010 18 (cross-sectional)	Eighteen primary/general practices: • Small (501-5000) • Medium (5001-10 000) • Large (>10 000) United Kingdom	Sample of 253 pts from all practices: • Small (37) • Medium (131) • Large (85)	Asthma
Angstman et al, 2016 19 (cross-sectional)	Academic medical center (Mayo Clinic) United States	2959 pts managed by 9 primary care teams, including 36 family physicians	Diabetes
Baker and Streatfield, 1995 20 (cross-sectional)	Survey to 89 general practices (solo and multi-partnered) in the Southwestern Region Health Authority with a total panel size of 7190 pts (1150-16 000). United Kingdom	19 580 pts • Sample of 220 pts/practice	Surgical care and experience
Balasubramanian et al, 2010 21 (cross-sectional)	Academic medical center (Mayo Clinic) United States	Primary care group practice • 1200 pts/clinician • 17 clinicians • 20 400 pts	General care
Broadbent et al, 2008 22 (cross-sectional)	Eighteen primary/general practices: • Small (501-5000) • Medium (5001-10 000) • Large (>10 000) United Kingdom	A sample (320 pts) from all practices stratified by the Indices of Multiple Deprivation: • Small (49) • Medium (150) • Large (121)	Osteoarthritis
Campbell et al, 2001 23 (cross-sectional)	A sample of 10 practices from 6 authorities within 3 regions in the National Health Service that are nationally representative in terms of rurality, socioeconomic deprivation, and geographic dispersion of population United Kingdom	11 831 pts surveyed • Sample response from 4493 pts	Chronic condition care • Angina • Asthma • Diabetes Preventive care
Cheung et al, 2017 24 (population-based cohort study)	9014 PCPs 1 018 647 pts Canada	Panels were categorized based on overall ambulatory volume, as well as diabetes-specific volume and number of PCPs in each category • ≤20 pts/day (3165 PCPs→0.1% had >301 pts with diabetes) • 21-25 pts/day (1821 PCPs → 0.6% had >301 pts with diabetes) • 26-30 pts/day (1339 PCPs→2.5% had >301 pts with diabetes) • 31-40 pts/day (1546 PCPs → 7.7% had >301 pts with diabetes) • > 40 pts/day (1143 PCPs → 22% had >301 pts with diabetes)	Diabetes
Dahrouge et al, 2012 12 (cross-sectional)	137 primary care practices in Ontario • 35 FFS • 35 CHC • 35 FHN • 32 HSO Canada	Survey included: • 3284 pts   ○ FFS (849)   ○ CHC (856)   ○ FHN (827)   ○ HSO (752) • 288 family physicians	Preventive care
Dahrouge et al, 2016 26 (cross-sectional)	137 primary care practices in Ontario • 35 FFS • 35 CHC • 35 FHN • 32 HSO Canada	8 265 930 pts • Mean (range) panel size: 1824 (1493-2260) • 4195 clinicians   ○ Male (2891)   ○ Female (1304)	Provider’s sex impact Preventive care
Dahrouge et al, 2016 25 (cross-sectional)	137 primary care practices in Ontario • 35 FFS • 35 CHC • 35 FHN • 32 HSO Canada	8 265 930 pts • Mean (range) panel size: 1824 (1493-2260) • 4195 clinicians   ○ Male (2891)   ○ Female (1304)	Chronic condition care • Diabetes • Asthma • CHF Preventive care
Dai et al, 2019 27 (retrospective survey)	Various practices, most small, solo, single-specialty group practices United States	27 836 family practitioners • Mean (range) panel size: 2263 (1888-2504)	Multispecialty
Dobscha et al, 2008 28 (cross-sectional)	A convenience sample from an ongoing trial (SEACAP) at a single VHA medical center serving Oregon and Washington United States	46 of 54 eligible staff clinicians Total active panel: 42 000 pts	Pts in clinicians’ panels prescribed opioids
Hernandez et al, 2012 29 (pilot study) a	Home-Based Primary Care Model: The study used a patient-centered care team, including 3 physicians (2 full-time equivalents), a social worker, a nurse practitioner, and an administrative assistant at Mount Sinai Visiting Doctors Program in New York, New York. United States	The model served 393 pts (40% increase in each physician’s panel size)	Extension of primary and palliative care services to 50% more homebound individuals per primary care provider while maintaining high quality care
Hippisley-Cox et al, 2001 30 (cross-sectional)	206 single-handed practices 606 partnerships practices United Kingdom	Mean number of panel size: • Single-handed practices (2299) • Partnerships practices (2058)	Chronic condition care • Asthma • Diabetes • Epilepsy
Hogg et al, 2009 31 (qualitative nested case series)	137 primary care practices in Ontario • 35 FFS • 35 CHC • 35 FHN • 32 HSO Canada	Survey included: • 5361 pts charts   ○ FFS (1375)   ○ CHC (1219)   ○ FHN (1494)   ○ HSO (1273) • Mean panel size per full-time physician   ○ FFS (2500)   ○ CHC (3200)   ○ FHN (4500)   ○ HSO (1700)	Health promotion
Kamnetz et al, 2018 32 (cross-sectional)	374 Primary care practices: family medicine, general internal medicine, pediatrics, adolescent medicine Public academic health system 279 000 medically homed pts United States	Study sampled 112 PCPs, 27 clinics, and 150 000 pts • Varied sociodemographic case mix • Medicare, Medicaid, HMOs, commercial and other insurance	Multispecialty
Katz et al, 2013 33 (cross-sectional) a	VA primary care patient data in Region 23 in Iowa United States	180 808 pts • Mean (IQR) panel size: 1178 (982-1295)/provider	Primary care
Kontopantelis et al, 2010 34 (cross-sectional)	GP Access survey by Department of Health in England: 4 922 080 surveyed pts from 8307 general practices United Kingdom	1 999 523 pts (40.6% response rate) Practice panel lists ranges: • <2000 • 2000-5999 • 6000-9999 • ≥10 000 Practice FTE providers ranges: • 1 • 2-5 • >5	Primary care
Lewis and Holcomb, 2012 35 (cohort study)	Military family practice setting caring for approximately 77 000 beneficiaries piloted a patient-centered model of care that proposed making the provider, RN, and RN case manager equally responsive and responsible for the empaneled patient population (My RN model) United States	My RN model: • 1193 pts/provider • 2 providers Usual care: • 1127.7 pts/provider • 12 providers	Chronic condition care • Asthma • Diabetes Preventive care
Majeed et al, 2003 36 (retrospective review)	69 general practices United States	382 188 pts • Mean (range) practice size: 5762 (265-13 147) • 6888 pts with IHD	IHD
Margolius et al, 2018 37 (retrospective review)	Tertiary academic health care system MetroHealth System in Ohio Diverse payer mix United States	114 clinicians, 87 physicians, and 27 advanced practice providers with independent panels • Family medicine, internal medicine, pediatrics and geriatrics • Mean panel size: 1146	Multispecialty
Mayo-Bruinsma et al, 2013 38 (cross-sectional)	137 primary care practices in Ontario • 35 FFS • 35 CHC • 35 FHN • 32 HSO Canada	137 practices, 363 providers, and 5144 pts Survey included: • 5361 pts • Mean panel size per full-time physician   ○ FFS (2500)   ○ CHC (3200)   ○ FHN (4500)   ○ HSO (1700)	Family-centered care
Millett et al, 2007 39 (cross-sectional)	QOF data for England and Scotland from the new General Medical Services contract for general practices with population of 55 522 778 pts United Kingdom	55 522 778 pts • 8970 general practices • 1 852 762 pts with diabetes • Practice panel ranges:   ○ 0 to 2999   ○ 3000 to 4999   ○ 5000 to 7999   ○ 8000 to 9999   ○ ≥10 000	Chronic condition care • Diabetes
Mittelstaedt et al, 2013 40 (cross-sectional)	Academic center (Oregon Health & Science University) United States	4 family medicine clinics • Mean (SD) pts/provider: 577.4 (315.8) • 63 clinicians	Interpersonal continuity
O’Shea et al, 2022 41 (retrospective data)	2017 to 2021 data from the VHA United States	5 014 445 pts 6202 primary care providers 916 primary care clinics (165 hospital-based and 751 community-based)	Staffing metric for primary care providers using retrospective data
Orueta et al, 2015 42 (cross-sectional)	The public Basque Health Service through primary care clinics in 7 health districts Spain	2 265 058 pts • 130 health centers • 2 207 175 pts (surveyed) • 1479 clinicians	Medication prescription
Panattoni et al, 2014 43 (cross-sectional/ observational)	Large specialty ambulatory care practice Mixed payer United States	13 clinics • 205 physicians (104 family medicine, 101 internal medicine) • 850 000 unique Press Ganey patient surveys	Multispecialty
Schapira et al, 2016 44 (survey)	Non-hospital, hospital, and community-based offices and health centers; 50% patient-centered medical home. United States	668 women’s health primary care clinicians • Physicians, physician assistants, nurse practitioners • Primary care, internal medicine, family medicine, OB/GYN	Breast and cervical cancer screenings
Stefos et al, 2011 45 (cross-sectional)	VA primary care patient data United States	162 000 new pts • 1187/provider • 2624 clinicians	Chronic condition care • Diabetes • IHD • Hypertension Preventive care • Alcohol misuse screen • Hyperlipidemia screen • Pneumococcal immunization • Cancer screen
Thalanany and Derrough, 2005 46 (cross-sectional)	The Public Health Directorate at the Billericay, Brentwood, and Wickford Primary Care Trust practice data United Kingdom	15 practices • 2083 pts/provider	Preventive care • Pneumococcal immunization
Vedavanam et al, 2009 47 (cross-sectional)	Primary practices in the Norfolk area • Small (501-5000) • Medium (5001-10 000) • Large (>10 000) United Kingdom	18 practices • 279 pts   ○ Small (43)   ○ Medium (139)   ○ Large (97)	Depression

Abbreviations: CHC, community health clinic; CHF, congestive heart failure; FFS, fee for service; FHN, family health network; FTE, full-time equivalent; GP, general practitioner; HMO, health maintenance organization; HSO, health service organization; IHD, ischemic heart disease; OB/GYN, obstetrics and gynecology; PCP, primary care physician; pts, patients; QOF, quality and outcomes framework; RN, registered nurse; VHA, Veterans Health Administration.

^aAbstract only.

Table 3.

Summary of Studies on Predictive Factors of Panel Sizes.

Study author, year (study design)	Study/practice setting country	Panel description	Conditions studied/scope
Altschuler et al, 2012 10 (case studies/predictive model)	Academic center (Duke University’s Department of Community and Family Medicine) United States	2500 pts designated to predicted models of care per delegation assumptions: • Nondelegated (983) • Delegated-1 (1947) • Delegated-2 (1523) • Delegated-3 (1387)	Predictive model • Preventive care • Acute and chronic care
Chung et al, 2012 48 (cross-sectional/predictive model)	Primary care practices in 7 clinics from the Palo Alto Medical Foundation (190 practices) United States	Using billing data for pts from the 7 practices • Adults (281 842) • Children (119 593)	Weight/impact of patient age
Dahrouge et al, 2016 26 (cross-sectional)	137 primary care practices in Ontario • 35 FFS • 35 CHC • 35 FHN • 32 HSO Canada	8 265 930 pts • Mean (range) panel size: 1824 (1493-2260) pts • 4195 Clinicians   ○ Male (2891)   ○ Female (1304)	Impact of provider’s sex Preventive care
Dai et al, 2019 27 (retrospective survey data)	Various practices, the majority small, solo, single-specialty group practices United States	27 836 family practitioners • Mean (range) panel size: 2263 (1888-2504) pts	Multispecialty
Green et al, 2007 50 (cross-sectional/ predictive model)	A probability model was developed to estimate overflow frequency level measure for any practice using NAMCS data United States	NAMCS reports that total number of annual visits to general and family practitioners in the US is a rate of 0.761. The authors use their previous data to estimate panel sizes (capacity utilizations) for different parameter values, primary care type	Predictive model to estimate panel size
Green and Savin, 2008 49 (cross sectional/predictive model)	A probability model was developed to use Queue with State-Dependent No-Shows approach to estimate panel size United States	The authors used data from the Columbia MRI facility and the Johns Hopkins Mental Health Care facility data published in a previous study to validate their model	Reducing delays for medical appointments in primary care
Huang et al, 2023 51 (retrospective data)	1 year of appointment data from Department of Family Medicine at Mayo Clinic United States	82 881 pts 105 clinicians	Retrospective model that determined optimal panel size based on 2019 appointment data
Hugo et al, 2000 52 (cohort)	137 general practices United Kingdom	434 clinicians	Eating disorder
Marx et al, 2009 54 (cohort)	Community- and hospital-based primary care clinics from the San Francisco Department of Public Health network; the authors described the process and outcome on establishing panels and development of a web-based panel application (NetAccess) for use by administrators and providers United States	Community-based clinic (13) • 3 516 447 pts • Active panel size: 2632 Hospital base clinic (4) • 25 649 pts • Active panel size: 997	Preventive care • Medication prescription
Marx et al, 2011 53 (cohort)	Community-based; the authors used 2-year retrospective data to predict utilization in the third year and assess increase/change in panel size United States	7 core primary care clinics • 25 000 active pts • Minimum 1125 pts/paid primary care FTE • Active panel size: 2595 to 5056	Primary care • Clinic capacity
Muldoon et al, 2013 55 (cross-sectional)	63 community health clinics in Ontario Canada	143 831 pts • Mean (SD) panel size/provider: 444.9 (178.3) pts	Chronic care
Murray et al, 2007 56 (cross-sectional/ predictive model)	Utilization data from previous primary care practice experience that guided building a predictive model to estimate panel size United States	The model variables • Provider visits per day • Provider days per year • Visits per patient per year • Pts age • Pts sex	Predictive model to estimate panel size
O’Shea et al, 2022 41 (retrospective data)	2017 to 2021 data from the Veterans Health Administration United States	5 014 445 pts • 6202 primary care providers 916 primary care clinics (165 hospital-based and 751 community-based)	Staffing metric for primary care providers using retrospective data
Orueta et al, 2015 42 (cross-sectional)	The public Basque Health Service through primary care clinics in 7 health districts Spain	2 265 058 pts • 130 health centers • 2 207  175 pts (surveyed) • 1479 clinicians	Medication prescription
Ozen and Balasubramanian 2013 57 (predictive model/case study)	Primary Care Internal Medicine United States	27 000 patients • 39 physicians • Case study 2 to 4 physicians • Mean panel size (range): 1060 (1053-1077) pts • Varied case mix	Multispecialty
Porter et al, 2022 58 (simulation study)	Hypothetical patient panels	Hypothetical panels of 2500 pts representative of the adult U.S. population based on the 2017 to 2018 National Health and Nutrition Examination Survey	Preventative, chronic disease, and acute care
Potts et al, 2011 59 (cross-sectional/predictive model)	POINT data using a predetermined disease-burden scale for 6 chronic conditions and their assigned 4-level risk to build a predictive model to minimize clinician burden and assess person-needed-for-hire at the Ohio Kaiser Permanente Medical Group United States	80 090 clinicians • Mean (range) panel size per provider: 2650 (1061-4627) pts	Chronic condition care • Asthma • Chronic kidney disease • Coronary artery disease • Diabetes • Heart failure • Hypertension
Rajkomar et al, 2016 60 (retrospective data)	UCSF primary care health system • 52 386 patients Cluster patients into phenotypes based on utilization United States	34 748 patients with at least 1 encounter (including office visit, telephone, electronic messaging, or medication refill) before February 2013	Multispecialty
Reckrey et al, 2015 61 (cohort)	Home-based primary care model (team approach) vs usual care The study used a patient-centered care team with 3 physicians (2 FTEs), a social worker, a nurse practitioner, and an administrative assistant compared to usual care at Mount Sinai Visiting Doctors Program in New York, NY United States	Team approach • 130 pts/provider • 347 pts Usual care: • 90 pts/provider • 1074 pts	Primary care • Hospitalizations • 30-day readmission
Rossi and Balasubramanian 2018 62 (retrospective longitudinal survey)	MEPS data sample of panel sizes 1500 and 2000 United States	Estimated panel sizes of 1500 and 2000 • Assumed patient data tied to 1 primary care physician • Varied case-mix	Multispecialty

Abbreviations: CHC, community health clinic; FFS, fee for service; FHN, family health network; FTE, full-time equivalent; HSO, health service organization; MEPS, Medical Expenditure Panel Survey; NAMCS, National Ambulatory Medical Care Survey Series; POINT, Permanente Online Interactive Network Tool; pts, patients; UCSF, University of California San Francisco.

Reported outcomes varied between studies and included clinical outcomes, processes, patient-reported outcomes, practice-centered outcomes, and system-related outcomes. Studies that explored the relationship between PPSs and health care delivery–related outcomes (ie, patient- and practice-centered outcomes) are summarized in Table 4.^{12,18 -20,22 -28,30 -47,52,55,57,58,60,62} Findings from all studies that highlighted the relationship between predictive factors and their respective PPSs are summarized in Table 5.^{10,13,15,21,24,27,32,37,41,43,48 -51,53,54,57,59 -64} Figure 1 provides a mapped-out analytic framework to follow these reported findings.

Table 4.

Summary of Study Outcomes and Effect Size.

Study author, year (country)	Assessed outcomes	Estimated effect of panel size
Abdelhamid et al, 2010 18 (UK)	Patient education Clinical process and outcomes Adherence to treatment Medication prescription	No effect No effect No effect No effect
Angstman et al, 2016 19 (USA)	Access (third available appointments) Disease quality management Patient satisfaction Cost Percentage daily fill rate Poor quality rankings (≤25th percentile)	Larger panel (longer; OR 10.91; 95% CI, 1.36-87.26; adjust for panel size above the mean of 2959 pts) Larger panel (worse) No effect No effect No effect Larger panel (OR 7.61; 95% CI, 1.13-51.46; P = .04; adjust for panel size above the mean of 2959 pts)
Baker and Streatfield, 1995 20 (UK)	Overall patient satisfaction Targeted satisfaction and access • Availability (ability to see clinicians) • Continuity of care (same clinicians) • Accessibility (ease to see clinicians) • Medical care (error-free care) • Premises (comfort, up to date)	Larger panel (regression coefficient for total list size per 1000, −0.78; SE, 0.18; P < .001) Larger panel, regression coefficient (SE) • −1.90 (0.33) • −1.47 (0.35) • −0.67 (0.13) • −0.70 (0.12) • −0.99 (0.40)
Broadbent et al, 2008 22 (UK)	Patient education Regular assessment and treatment Medication prescription (osteoarthritis)	No effect No effect No effect
Campbell et al, 2001 23 (UK)	Disease quality management (diabetes) Access to care Interpersonal care Continuity of care Overall patient satisfaction Preventive care	Larger panel better (AOR, 2.16; 95% CI, 0.22-4.10; P = .029) Smaller panel (AOR, 0.87; 95% CI, 0.76-0.99; P = .038) Smaller panel (AOR, 0.82; 95% CI, 0.74-0.90; P < .001) No effect No effect No effect
Cheung et al, 2017 24 (Canada)	DM Care Quality Indicators: • Eye examination • LDL cholesterol testing • Hemoglobin A1c • Prescriptions of ACEIs/ARBs • Prescriptions of statins ED visits for hypo/hyperglycemia	High DM-specific rates was associated with higher quality performance in all quality indicators Higher overall ambulatory volume was associated with lower performance in all quality indicators
Dahrouge et al, 2012 12 (Canada)	Preventive care (cancer screening and screening examinations)	Smaller panels (<1600 vs >1600 [β, 6.8; 95% CI, 3.1-10.6])
Dahrouge et al, 2016 25 (Canada)	Continuity of care Comprehensiveness Preventive care and premises (cancer screening and screening examinations) Disease quality management (diabetes, asthma and CHE) Medication prescription (Metformin, lipid lowering agents, anti-hypertensive agents	Smaller panel (best in medium panel of 1800-2700) Smaller panel Smaller panel (likelihood decreased with increasing panel size from 1200 to 3900; larger panels had more up-to-date screening likelihood No effect No effect
Dahrouge et al, 2016 26 (Canada)	Comprehensiveness Sex (female) Female physicians Access (low-urgency ED visits) Admissions (ambulatory care sensitive conditions)	Smaller panel Smaller panel size Larger panel (worked with) Smaller panels (1200-1799 vs 1800-2399 vs 2400-2999) had fewer low-urgency ED visits Smaller panels
Dai et al, 2019 27 (USA)	Variations in family physicians practices with NPs and PAs	Larger panel size (NPs  =  259; PAs =  410; Both  =  245; P < 0.05)
Dobscha et al, 2008 28 (USA)	Job satisfaction (clinicians) Pts in clinicians’ panels prescribed opioid	Larger panel (r, −0.680; P < .001) Larger panel (r, −0.563; P < .001)
Hippisley-Cox et al, 2001 30 (UK)	Singlehanded general practitioners	Larger panel
Hogg et al, 2009 31 (Canada)	Health promotion discussion (likelihood) Longer booking interval	Smaller panel (AOR, 0.92; 95% CI, 0.85-1.01) Smaller panel (AOR, 0.89; 95% CI, 0.81-0.97)
Hugo et al, 2000 52 (UK)	Referrals to specialists (eating disorders)	Large panel (RR per unit increase in number of partners, 1.11; 95% CI, 1.07 -1.16; P < .001)
Kamnetz 2018 32 (USA)	Development of a utilization-based weighting system that accounts for patient complexity Impact on access	Decreased open panels for family physicians and pediatricians Increased open panels for general internists
Katz et al, 2013 33 (USA)	Continuity of care Excellent communication	No effect Larger panel (AOR, 1.007; 95% CI, 1.005; 1.009 for each 10-pt increase in panel size)
Kontopantelis et al, 2010 34 (UK)	Targeted satisfaction and access • Availability (ability to see a clinician) • Continuity of care (specific clinician) • Accessibility (advanced scheduling) • Surgery opening hours	Smaller panel • Small practices (OR per 1000 increase in pts, 0.616; P ≤ .0001) • Small practices (OR per 1000 increase in pts, 0.556; P ≤ .0001) • Small practices (OR per 1000 increase in pts, 0.407; P ≤ .0001) • Small practices (OR per 1000 increase in pts, 0.839; P ≤ .0001)
Lewis and Holcomb 2012 35 (USA)	Managing larger panel size Preventive care (cancer screening and screening examinations) • Access to care • Utilization • Continuity • Satisfaction • Productivity	Larger panel (pilot model managed larger panel size) • Larger panel (pilot model improved screening) • Larger panel (pilot model increased access) • Larger panel (pilot model lessened ED visits) • Larger panel (pilot model lessened no-show rate and maintained seeing the same pts) • No effect • No effect
Majeed et al, 2003 36 (UK)	Disease quality management (IHD) Medication prescription (statin, aspirin, β blocker, ACEI) Screening tests	No effect No effect No effect
Margolius et al, 2018 37 (USA)	Waits for appointments Number of clinical days	Lower FTE associated with worse access. No effect; panel size, without adjustment for FTE and number of clinicians per site, had no correlation with access
Mayo-Bruinsma et al, 2013 38 (Canada)	FCC	Larger panel (OR, 0.92; 95% CI, 0.84-1.01; P = .095) of pts reporting FCC (lower reporting of hereditary conditions in pts’ families, household income, and living situations, as well as awareness of the signs of child abuse)
Millett et al, 2007 39 (UK)	Disease quality management (diabetes) • Clinical processes • Surrogate outcomes	Larger panel • Larger panel • Larger panel (modest [<5%] for BMI, smoking cessations, neuropathy testing, retinal screening, hemoglobin A1c, blood pressure, and cholesterol)
Mittelstaedt et al, 2013 40 (USA)	Interpersonal continuity (Usual Provider Continuity Index)	No effect
Muldoon et al, 2013 55 (Canada)	Poverty Comorbidities (measured by SAMI) Primary care staffing Office layout	Smaller panel (β coefficients, −0.072) Smaller panel (β coefficients, −0.291) No effect No effect
O’Shea et al, 2022 41 (USA)	Clinic location • Rural • Urban	Effect on staffing in 3.5 year study period • Understaffed for 20.5% of months • Understaffed for 13.7% of months
Orueta et al, 2015 42 (Spain)	Higher prescription costs Referrals Number of doctor visits Hospitalizations for ambulatory care sensitive conditions	Smaller panel (<1700), rate ratio 0.85 (0.77-0.93) No effect No effect No effect
Ozen and Balasubramanian, 2013 57 (USA)	Impact of case-mix on timely appointments Benchmark overflow value for a group practice Panel redesign options for least disruption to improve access	Effect on panel size varies by model: • Capacity based—switches pts to even panel sizes among physicians (similar panel sizes) • Heuristic 1—switches more pts (with lower utilization); effect on individual physicians varies depending on case-mix, those with more patients with higher comorbidities end up with smaller panel sizes; those with more pts with fewer comorbidities have increased panel sizes. • Heuristic 2—fewer pts switched but across all comorbidity levels, varying effects on individual physicians
Panattoni et al, 2014 43 (USA)	Relationship between a physician’s clinical FTE • Continuity of care • Access to care • Patient satisfaction	FTE associated with: • Better continuity of care • Better access to care • Worse patient satisfaction scores
Porter et al, 2022 58 (USA)	Difference in mean time required to provide preventative, chronic disease, and acute care from 2500 panel size to 1500, 2000, and 3000 • PCP-only care • Team-based care	PCP-only care: −10.7 h (1500 pts) −5.3 h (2000 pts) +5.3 h (3000 pts) Team-based care: −3.7 h (1500 pts) −1.9 h (2000 pts) +1.9 h (3000 pts)
Rajkomar et al, 2016 60 (USA)	Weighted algorithm of predictive panel sizes 7 phenotypes based on utilization • A, >1 visit/year • B, C, and D, <2 visits/year • E and F, >7 visits/year • G, high outlier >30 visits/year	Patient populations decreased in inactive clusters (26 091-17 205) and low clusters (11 830-591) Weighted sizes increased in medium clusters (9404-12 402) and high clusters (5043-22 169) Panel size increased on average by 12.8%
Rossi and Balasubramanian, 2018 62 (USA)	PCP weekly office visit distribution Weekly distribution of non-PCP events that do not require face-to-face communication	Including coordination time for non-PCP specialty visits will increase estimated number of office visits/week • Panel of 2000: 93.54 office visits + 13 non-PCP specialty visits (coordination) = 106.54 office visits • Medicare panel requires 1.68 physicians to cover 157.02 weekly PCP office visits
Schapira et al, 2016 44 (USA)	Breast cancer risk assessments HER decision support Performance reports Panel reports for screenings and follow-up	PCMH designation associated with greater use of comparative performance reports of guideline adherent breast or cervical cancer screening Greater use of automated reports of pts due for breast or cervical cancer screening No panel size effect
Stefos et al, 2011 45 (USA)	Overall patient satisfaction Increased wait time Preventive care (alcohol misuse and hyperlipidemia screening, pneumococcal immunization)	Larger panel (log −0.0003) Larger panel (log −0.00013) Larger panel (log −0.0004 for all indicators)
Thalanany 2005 46 (UK)	Preventive care (pneumococcal immunization)	Larger panel (>4000 pts; AOR, 1.45; P < 0.0001; pts >80 years with clinical risk)
Vedavanam et al, 2009 47 (UK)	Medication prescription Patient education Clinical processes (diagnosis, treatment, follow-up)	No effect No effect No effect

Abbreviations: ACEI, angiotensin converting enzyme inhibitor), ARB, angiotensin receptor blockade; BMI, body mass index; DM, diabetes mellitus; ED, emergency department; FCC, family-centered care; FTE, full-time equivalent; IHD, ischemic heart disease; LDL, low-density lipoprotein; NP, nurse practitioner; OR, odds ratio; PA, physician assistant; PCMH, patient-centered medical home; PCP, primary care physician; pts, patients; RR, relative risk; SE, standard error.

Table 5.

Summary of Predictive Care Models.

Study author, y (country)	Model description
Altschuler et al, 2012 10 (USA)	Based on the American Academy of Family Physicians’ estimation of a family physician working 43 h per week for 47.1 weeks annually, this model calculates the average annual working hours as 2025. It then determines the panel size by dividing these 2025 h by the total annual hours required per patient for providing preventive, chronic, and acute care services.
Balasubramanian et al, 2010 21 (USA)	Used a computerized simulation (reallocation) model (stochastic linear programming) to improve clinical access and continuity through physician panel redesign at Mayo Clinic
Cheung et al, 2017 24 (Canada)	This population-based cohort study provided another way of looking at panel size which examines a specific outcome and quality care, showing that physicians with larger diabetes-specific panels may provide higher quality diabetes care than physicians with higher overall panel size not specific to a single condition (diabetes mellitus).
Chung et al, 2012 48 (USA)	The study utilized data from over 280 000 patients at the Palo Alto Division of the Palo Alto Medical Foundation to evaluate whether patient clinical conditions should factor into primary care physician (PCP) workload when standardizing panel sizes. Work resource value units were employed to normalize PCP panel workload, with standardized panels developed using age and sex, as well as clinical condition-based risk indicators. Billing data from all patients, irrespective of their insurance status, were used for PCPs in a group practice (n = 190). The research also examined weighting methods for various subgroups, categorized by PCP specialty (family medicine, internal medicine, pediatrics) and patient age (adults vs children), and at different levels of aggregation (individual patient vs PCP).
Dai et al, 2019 27 (USA)	The study analyzed 4 years of demographic questionnaire data from family physicians registering for the American Board of Family Medicine Family Medicine Certification Examination. It found that variations in the configuration of practice teams correlated with the size of family physicians’ patient panels and their scope of practice. This suggests that optimally structured teams could enhance the capacity at which primary care physicians operate compared to less effectively configured teams.
DeCherrie et al, 2012 63 (USA)	In this study comparing a home-based primary care model with usual care, the researchers employed a patient-centered care team approach. This team, part of the Mount Sinai Visiting Doctors Program in New York, New York, consisted of 3 physicians (equivalent to 2 full-time employees), a social worker, a nurse practitioner, and an administrative assistant. The effectiveness of this model was compared to the standard, usual care provided within the same program.
Ghorob, et al 2013 64 (USA)	This review addressed 3 building blocks to successful implementation of same-day appointments for patients. These building blocks include panel size management as the most important block, supplemented by continuity of care and team care.
Green et al, 2007 50 (USA)	A probability model was developed to estimate overflow frequency level measure for any practice using National Ambulatory Medical Care Survey Series data.
Green and Savin, 2008 49 (USA)	A probability model was developed to use Queue with State-Dependent No-Shows approach to estimate panel size.
Huang et al, 2023 51 (USA)	Retrospective optimization model that determined optimal panel size from 2019 appointment data for physicians, advanced practice clinicians, and residents with respect to panel management time (ie, their percent time spent on clinical duties in relation to total work time). The model aimed to maximize clinician capacity through heterogeneous appointment types spread among providers.
Kamenetz et al, 2018 32 (USA)	A utilization-based weighting system was developed to compare both weighted and unweighted panel sizes. This system created weighted panels for PCPs within the organization, utilizing available electronic health record (EHR) data. The aim was to enhance the accuracy of panel size calculations by adjusting for varying patient characteristics, thereby reflecting the actual workload more precisely.
Margolius et al, 2018 37 (USA)	Multiple regression models to evaluate the association between an access metric (wait times until third next available appointment), panel size, and clinician FTE
Marx et al, 2009 54 (USA)	In a pilot model of care (panel management) for the San Francisco Department of Public Health, encompassing 13 community-based and 4 public hospital-based primary care clinics, the target panel size for each clinic was calculated. This was done by multiplying the number of paid primary care FTEs by 1125 and subtracting the clinic’s current adjusted panel size to determine capacity for new enrollees. A web-based tool, NetAccess, provided weekly updated summary statistics for primary care coordinator and PCP panels, including patient panel lists linked to data on utilization and demographics. This enabled providers to manage panels effectively, measure utilization, capacity, productivity, assess patient characteristics, and generate clinical quality indicators based on an accurate denominator.
Marx et al, 2011 53 (USA)	Building on their previous model and experience, the authors used 2-year retrospective data to predict utilization in the third year and assess increase/change in panel size.
Murray et al, 2007 15 (USA)	Predictive model that predicts panel size based on supply and demand while taking into consideration/adjusting for patient sex, age, weight, and practice style. The authors built an input sheet that can be used to calculate panel size.
O’Shea et al, 2022 41 (USA)	Primary care provider staffing metric to identify Veterans Health Administration clinics that were at risk for over or under-staffing based on current VHA care models, panel size, and provider FTE. This metric reported, throughout a 3.5 year period, staffing gaps for all clinics and as well as for both urban and rural clinics.
Ozen and Balasubramanian, 2013 57 (USA)	The study presents a mathematical model designed to optimize physician panels in a multi-physician practice, aiming to minimize the frequency of maximum overflow. Building upon the Green et al 50 framework, it introduces an Excel tool that serves 3 main functions: (1) it quantifies the impact of case-mix; (2) it calculates a benchmark value for overflow in group practices; and (3) it facilitates the evaluation of various long-term panel redesign options. This tool is intended to assist in effectively restructuring physician panels for improved practice management. A preliminary version of the Excel spreadsheet is available for free at people.umass.edu/hbalasub/PanelDesignSpreadsheet.xlsx.
Panattoni 2014 (USA) 43	This study assessed the relationship between a physician’s clinical full-time equivalent (FTE) and 4 key variables: the continuity of care received, the continuity of care provided, access to care, and patient satisfaction with the physician. A multilevel structural equations model was employed for this analysis, utilizing Stata/MP 13.0 GSEM (StataCorp LLC). The model incorporated department level random effects and standard errors correlated at the clinic level, providing a comprehensive understanding of these interrelated factors in a healthcare setting
Potts et al, 2011 59 (USA)	In this study, the authors devised a model to determine the optimal number of nurse practitioners needed to address staffing gaps and alleviate physician workload and burnout. Utilizing data from the POINT network, the model incorporated a disease-burden scale for 6 chronic conditions, each classified into 4 risk levels. The disease scores for each physician were totaled and averaged. This average disease score was then divided by the panel size, providing an average disease burden for each physician within specific geographic regions of their practice. This approach aimed to strategically match nurse practitioners with physicians based on calculated staffing needs.
Rajkomar et al, 2016 60 (USA)	Predictive, weighted model using EHR data to cluster primary care patients into groups that reflect primary care work effort required to care for diverse patients, based on utilization per year; utilization phenotypes used to create weighted panel sizes
Reckrey et al, 2015 61 (USA)	Home-based primary care model (team approach) vs usual care study; the study used a patient-centered care team consisting of 3 physicians (2 FTEs), a social worker, a nurse practitioner, and an administrative assistant compared to usual care at Mount Sinai Visiting Doctors Program in New York, New York
Rossi and Balasubramanian, 2018 62 (USA)	Used MEPS data to quantify 2 types of workload associated with panel size: (1) PCP weekly office visit distribution and (2) weekly distribution of non-PCP events that do not require face-to-face communication
Rowe, et al 2017 13 (USA)	A survey evaluated a type of care model called “Concierge Medicine” or Direct Primary Care where patients pay a predetermined membership fee to enroll. The projected increased panel size examined in this study suggests the emergence of a not yet popular primary care practice model.

Abbreviations: EHR, electronic health record; FTE, full-time equivalent; MEPS, Medical Expenditure Panel Survey; PCP, primary care physician; POINT, Permanente Online Interactive Network Tool.

PPS and Care Delivery–Related Outcomes

Of the 31 studies examining PPS and care delivery–related outcomes, 23 were cross sectional or survey studies, 3 were retrospective chart review, 2 were cohort studies, 1 had a mixed methods design, and 1 was a quasiexperimental pilot study. Table 2 incudes the baseline characteristics, clinical outcomes, and reported effect sizes of these studies. We identified reported trends between PPS and (1) patient-centered outcomes (clinical and wellness outcomes); (2) patient-reported outcomes (patient satisfaction); (3) quality of care measures; and (4) utilization. A detailed narrative of these findings is included in Supplemental Appendix.

1. Patient-centered outcomes (clinical outcomes)

   • (a) Preventive services: Smaller practices were often associated with a higher likelihood of being up-to-date with preventitive services,^12,25 though this was not consistently reported.^30,44,60
   • (b) Diagnostic tests: Generally, practice size was not associated with the number of diagnostic tests ordered per patient.^18,22,47
   • (c) Health promotion: Better health promotion was associated with smaller PPSs in 1 study. ³¹
   • (d) Clinical services: Health education was not associated with PPS.^18,22,47
   • (e) Diabetes care: While several studies found no association between PPS and diabetic care,^22,25,36 others reported favorable^23,39 or unfavorable^19,61 outcomes for diabetic care with larger panels (≥2500).
   • (f) Medication prescription rate: The consensus among the studies was that PPS did not impact medication prescription rates for various conditions, including asthma, and depression.^{18,22,25,35,36,47} Two studies found that larger PPS was strongly and negatively correlated with opioid prescriptions.^28,42
2. Patient-reported outcomes

   • (a) Patient satisfaction and experience: Larger PPSs were either negatively correlated with^20,28,34,45 or had no significant effect on patient satisfaction.^19,23,35 A 1000 PPS increment was reportedly associated with the following reduction in experience and satisfaction: 2.4% reduction relating to getting through on the phone; 1.0% reduction relating to scheduling an urgent appointment; 1.9% reduction relating to scheduling an advanced appointment; and 1.2% reduction relating to scheduling an appointment with a particular doctor ³⁴ Reduced full-time equivalent (FTE) work hours were also associated with higher patient satisfaction. ⁴³ Mohr et al ⁶⁵ surveyed 222 primary care clinics in the Veterans Health Administration (VHA) system and noted that when workload increased, patient quality of life ratings were negatively impacted. However, higher levels of relational climate (ie, clinicians communicating effectively and providing mutual support to accomplish patient care goals) were associated with better care. ⁶⁵
   • (b) Family and Personal History Reporting/Communication: Larger panels in a practice were associated with lower odds of patients reporting family-centered care (ie, lower focus on individualized, family-centered care aspects due to higher workload or insufficient time per patient) ³⁸ ; for every 1000 additional patients in a practice, the odds of patient-reported family-centered care dropped by 8%. Family-centered care was defined as consideration of hereditary conditions in the patient’s family, household income, and living situations, as well as awareness of the signs of child abuse.
3. Quality of care measures

   • (a) Continuity of care: Continuity was found to be associated with PPS in some studies, with continuity of care improving with smaller PPSs,^12,18,44,62 and to have no effect in others.^23,33,40 Dahrouge et al ²⁶ found that continuity of care was optimized with PPSs of 1800 to 2700 patients per clinician. Lewis and Holcomb ³⁵ found that implementing a coordinated care model that included the same physician and nurse per patient significantly improved continuity of care and the clinician’s ability to see their own patients, even with larger PPSs. Another VHA study of 180 808 patients with the aim of implementing a comparable model found no effect on continuity of care when PPSs were increased. ³³ Increment of clinician FTEs was associated with better continuity of care. ⁴³
   • (b) Access to care: Access was measured by various methods. Generally, the association between access to care and PPS was positive with smaller PPSs^23,34 and negative with larger PPSs, resulting in delays.^19,20,38 One study found that implementing a patient-centered model of continuous care increased access in larger panels. ³⁵ Lower clinician FTEs was associated with less access, ³⁷ and increased FTEs were directly associated with better access to care. ⁴³ The creation of a weighting system accounting for complexity resulted in changes to practice closure, increased total patients, and improved access. After this weighting was implemented, patients’ perceived access improved in family medicine and general internal medicine clinics. ³² Egede et al ⁶⁶ conducted a study within a primary care clinic at an academic medical center in the Midwest over the course of 24 months, reporting improved access to care (as well as provider productivity and patient satisfaction) of approximately 22% by integrating various practice changes, including: consistent visit lengths; addition of early morning and evening hours; handoff of completion of pain management agreements to registered nurses; creation of a team-based care model including 2 advanced practice providers to each team to increase patient volume; standardization of visit length; and incorporation of biweekly meetings with scheduling staff and administration to improve scheduling. Additionally, O’Shea et al ⁴¹ introduced a “gap staffing” metric to assess staffing adequacy in VHA clinics, finding significant staffing gaps, especially in rural areas. They found that lower clinician FTEs were associated with larger panel sizes, which in turn led to less access to care, highlighting the importance of balanced staffing for optimal patient access. ⁴¹
4. Utilization and cost

   • (a) Adherence to treatment: Smaller PPS was significantly associated with improved adherence to certain preventive care (eg, metabolic panels, mammography and colorectal cancer screening), ³⁵ though the significance was not consistent with other quality indicators. ¹⁸
   • (b) Cost: There was a paucity of data about patient empanelment in relation to cost.

Predictive Factors of PPS

There were 20 studies included in this category. Table 3 incudes the baseline characteristics of the included studies, clinical outcomes, and any reported effect sizes. Factors contributing to the calculation of appropriate PPSs included (1) patient, (2) clinician, and (3) practice-related characteristics, as well as (4) utilization and cost. Despite their crucial relevance to PPS, there remains insufficient reporting regarding practice structure resources (eg, team support and composition), practice infrastructure resources (eg, electronic health record, data systems, information technology support), health type coverage, and other integrated subspecialty services within primary care. A detailed narrative of these findings is included in Supplemental Appendix.

Patient-related characteristics

• (a) Age and sex: Chung et al ⁴⁸ summarized the evidence using data from over 280 000 patients and employing work relative value units to standardize primary care physician panel workload. Standardized panels were created using age and sex and clinical condition–based (comorbidities) risk indicators. According to the study, for pediatric panels, age and sex–based adjustment is probably sufficient, but for adult panels, further adjustment reflecting patient clinical conditions is warranted. The authors concluded that PPS standardization should account for unusual but predictable coding patterns. ⁴⁸

• (b) Comorbidities: In addition to Chung et al, ⁴⁸ Ozen and Balasubramanian ⁵⁷ found that when patients have more comorbidities, physicians should consider decreasing PPS.

Clinician-related characteristics

Dahrouge et al ²⁶ found that female clinicians had significantly lower PPSs and their patients were more likely to have received recommended cancer screening and diabetes management, fewer emergency department visits and hospitalizations, and higher referral rates.

Practice-related characteristics

Community-based settings predict smaller PPSs and poorer populations with higher medical comorbidities. ⁵⁵ One study suggested the difference in PPSs between community and hospital settings may be due to the larger number of FTE staff at hospital-based clinics; PPS per clinical FTE, a more accurate measure of clinical burden, was slightly higher for community settings vs hospital-based clinics. ⁵⁴

The addition of nurse practitioners and physician assistants within family physician practices was associated with increased PPSs. ²⁷ Beyond the composition of clinical teams, the physical office layout and support staffing in clinics did not predict changes in PPS. ⁵⁵ Reckrey et al ⁶¹ found that with the support of a nurse practitioner, an administrative assistant, and a social worker, physicians were able to increase their PPS. Potts et al ⁵⁹ found that hiring nurse practitioners to work with physicians who have patient populations with high comorbidities helped maintain manageable PPSs while improving patient and physician satisfaction. Porter et al ⁵⁸ used a hypothetical panel of 2500 patients in a simulation study to quantify time needed to provide care in a single-physician versus team-based care model. They noted that a primary care physician alone would need 26.7 h a day (14.1 h/day for preventive care, 7.2 h/day for chronic disease care, 2.2 h/day for acute care, and 3.2 h/day for documentation and inbox management). The physician’s time would decrease to 9.3 h a day in a team-based structure (2 h/day for preventive care, 3.6 h/day for chronic disease care, 1.1 h/day for acute care, and 2.6 h/day for documentation and inbox management). They concluded that primary care physicians did not have enough time to provide guideline-recommended care alone or in a team-based care model, even though team-based care reduced time requirements by more than half. ⁵⁸ Huang et al ⁵¹ optimized clinician workloads in primary care by modeling patient panel sizes using data from 82 881 patients and 105 clinicians based on their panel management time, improving workload distribution and reducing burnout. The study highlighted significant differences in annual appointment volumes per clinician with an optimization-based approach. ⁵¹ O’Shea et al ⁴¹ examined a novel “gap staffing” modeling metric to assess the relationship between FTE-based staffing size and panel size to ensure adequate access to care. This metric showed that lower clinician FTEs were associated with larger panel sizes, resulting in less access to care, particularly in rural clinics; thus, emphasizing the need for balanced staffing to optimize patient care and access. ⁴¹

Utilization and cost

• (a) Annual visits: Two studies concluded that utilization was directly or indirectly predictive of optimal PPS.^60,62 Factors considered included estimating the workload required to deal with non–primary care physician specialty visits (non–face-to-face visits) that primary care physicians coordinate. A third study was able to predict PPS based on visit type (ie, preventative, acute, or chronic). ⁵⁰

• (b) Referrals: Orueta et al, ⁴² analyzing health system data of panels between 1600 and 1890 patients in Spain, found that PPSs larger than 1700 patients did not significantly affect the number of doctor visits or referrals to specialists. On the other hand, Hugo et al ⁵² found that practice size, proximity to the clinic, female general practitioners, general practitioners with UK-based degree and qualification, and full contraceptive services were all associated with higher referral rates.

• (c) Cost: Angstman et al ¹⁹ reported that PPS did not predict changes in costs measured as per member per month charges. Hernandez et al ²⁹ found that a 40% increment in PPS for home-based primary care programs resulted in a 15% cost reduction per patient.

Discussion

Our review identified 48 original studies exploring 7 key factors associated with PPS as dynamic and multifactorial outcome, influenced by practice type, patient population characteristics, and organizational structure. Generally, smaller panel sizes were associated with improved patient-reported outcomes, such as satisfaction, continuity of care, and health promotion, while clinical outcomes, utilization, and costs showed minimal association with panel size. Larger panels were linked to increased clinician burnout, reduced patient access, and lower patient satisfaction. A lack of standardized definitions and reporting on PPS characteristics created significant heterogeneity across studies, limiting the ability to draw consistent conclusions. Community-based centers generally managed smaller panels, often staffed by female clinicians, and served populations with greater socioeconomic disadvantages and comorbidities compared to hospital-based settings. Despite these findings, there was no single clear predictor for optimal PPS.

It is estimated that by 2034, there will be an estimated shortage of up to 48 000 primary care physicians. ⁶⁷ Despite these predictions, studies have demonstrated the risk of insufficient feasibility or sustainability of primary care, highlighting that it would take a primary care physician between 17 to 21 h a day of clinical work to keep up with their patient panel.^14,68 In his essay, Bodenheimer ⁶⁹ hypothesized that primary care’s root problems are due to 2 primary factors: a low percentage of national health expenditure directed to primary care and unmanageably large PPSs. The author, among others, noted that these factors lead to widespread primary care physician exhaustion, cynicism, and burnout, as well as poor patient access, suboptimal medical care, fewer preventative services, and lower patient satisfaction.^{25,69 -72}

We designed our metanarrative review to determine whether there is an association between PPS and outcome measures, as well as to further determine which factors could be used when calculating ideal PPS. In a systematic review, Paige et al ⁷³ concluded that higher PPS was negatively associated with patient health outcomes, clinical quality, patient experience, and health care professional burnout. Similarly, the included evidence demonstrated significant associations between PPS and outcome measures, as well as great variation in how groups and health systems evaluate PPS and the factors and outcomes associated with modifying PPSs for distinct primary care practices. Furthermore, we agree with a study that identified significant variation in how different organizations define primary care panels, the factors for including or removing patients from a panel, and the impact that these variable definitions have on primary care PPS. ⁷⁴ According to Mayo-Smith et al, ⁷⁴ given the considerable variation in defining and measuring primary care panels, caution and judgment should be used when comparing reported PPS findings across studies until a standardized definition is created. Understanding and defining PPS in primary care is an essential building block of population-based care. ⁷ Effective use of this building block may be a key component to the future of effective primary care delivery.

Practice leaders and policy makers are frequently tasked with the establishment of PPS parameters or targets for their practice. “What are your target PPSs?” and “How did you arrive at that number?” are commonly asked questions at primary care leadership conferences. It should be noted that many factors influence the calculation of ideal PPS, including physician and practitioner preferences, organizational logistics, ancillary support structure, and patient population. ⁶⁸ Furthermore, according to a systematic review by the VHA, the optimal PPS requires a complex balance of health care system demands (ie, access, quality of care, cost, and patient experience) and the needs of the health care team (ie, preferences, satisfaction, and minimization of burnout).^10,72 Shekelle et al ⁷² noted that although determining optimal PPS is complex, there was a negatively significant relationship of modest size between increasing PPS and various measures of health care quality and experience.

Variability in PPS is influenced not only by practice-specific factors but also by external factors such as regulatory and payer requirements. Organizations such as the National Committee for Quality Assurance (NCQA) and the Joint Commission as well as medical societies and bodies (eg, American Academy of Family Physicians [AAFP]; Society of General Internal Medicine [SGIM]) set standards for care delivery, yet they do not provide concrete guidance on PPS, leading to inconsistencies in its implementation.^75,76 Additionally, payer-driven metrics and incentives focusing on cost, quality, and patient experience significantly shape panel size decisions. Addressing these influences is essential for developing a comprehensive and standardized framework for PPS optimization.^7,77

There are novel factors that might be considered in future PPS research and frameworks. For example, theories involving patient capacity frameworks especially for chronic care management, such as those proposed in minimally disruptive medicine (MDM), may provide valuable insights. ⁷⁸ These approaches emphasize aligning care demands with patients’ ability to manage their health, which could inform PPS optimization, particularly for high-need or vulnerable populations.^11,79,80

Our review further illustrates the important challenges facing stakeholders and decision makers as they grapple with the complexities of their own practice characteristics and the variability found in the literature. In the current state, it is not feasible to confidently establish optimal PPS for a given practice based on well-established norms or via an extensive review of the literature. Therefore, our metanarrative review demonstrates the need for a standardized and systematic approach to the work of understanding primary care PPS. We propose a framework of 7 categories for conceptualizing primary care panels and practice characteristics and their impact on practice-meaningful outcomes to explore and identify potential commonalities, as included in Figure 1. We have summarized the existing literature into this framework with the hope that it will facilitate future studies in PPS in primary care research.

Strengths and Limitations

Some limitations of our study are inherent to the nature of the study design, observational and nonvalidated predictive model, which may limit generalizability. Additionally, the heterogeneity in reporting outcomes may limit its application. The strengths of this study include its comprehensive nature and selection process following a robust methodologic framework of reporting heterogeneous literature. This novel classification approach could inform many types of future studies and research as noted above. The narrative nature of the summaries may highlight new areas and opportunities for collaboration to direct future practice guidelines and policies.

Conclusions

Our metanarrative review found that the optimal primary care PPS is a multifaceted decision that varies according to practice type and setting. Generally, patient-reported outcomes are more highly associated with PPSs than practice-related factors. For individual primary care practices, patient population characteristics were a major factor for determining appropriate PPS. There was substantial heterogeneity in methodology and approaches to determine PPS and the predictive factors for its determination. Future research should adopt a standardized framework, focusing on the 7 identified factors to better understand and evaluate PPS, ensuring a balance between patient-centered outcomes, system demands, and clinician well-being.