Abstract
Objective: The Person-Centered Primary Care Measure (PCPCM) is a patient assessment of their longitudinal experience of care with a clinician and care team, evaluating core functions of primary care in the health system. However, the optimal process of implementation across health systems, including how and when to administer the survey, reporting, and process improvement activities tied to survey data, has not been established.
Methods: We distributed the 11-question PCPCM experience survey to 329,450 patients empaneled across 78 primary care practices between April 2023 and January 2024. We evaluated survey completion parameters, psychometric properties, and mean responses in relationship to patient-level demographic variables.
Results: In this large, heterogeneous system of primary care practices, the PCPCM survey was successfully distributed using the Press Ganey (South Bend, IN) platform. We found a low response rate (6.4%), but demonstrated good internal consistency, with a skew toward higher scores. PCPCM scores varied by age, sex, race, primary care clinician type, and the number of years the patient had been at their current primary care practice. Responses varied significantly by patient race, but differences were small and not uniform in direction. Black or African American patients were significantly less likely to believe that the care provided by the practice was informed by knowledge of their community, compared with all other racial groups.
Conclusions: The PCPCM was implemented successfully in a large network of primary care practices, but more work is needed to improve the response rate. Future work should focus on the use of the PCPCM for practice and clinician feedback and validation of individual PCPCM items.
- Continuity of Patient Care
- Family Medicine
- Implementation Science
- Patient Care Team
- Patient-Centered Care
- Patient Reported Outcome Measures
- Primary Health Care
- Quality Improvement
Introduction
Primary care as a health care service is measured in many ways by health systems, including disease-specific quality measures, completion of preventive care tasks, operational measures of productivity and efficiency, and patient experience measures. Traditional measures have not captured the core function of primary care in a health system, defined as care that is first contact, comprehensive, coordinated, and continuous.1–2 In 2017, attendees of the Starfield Summit III convened to build on the results of a crowdsourced survey of diverse stakeholders to consider how to more closely capture the work of primary care clinicians.3 Findings from the Starfield Summit were combined with findings from the crowdsourced surveys and a review of the literature to culminate in the development of the Person-Centered Primary Care Measure (PCPCM; see Appendix).4 The PCPCM asks patients and families to assess their longitudinal experience of care with a clinician and care team, evaluating access, continuity, comprehensiveness, coordination, advocacy, family and community context, and goal-directed care. It can be used to focus efforts on improving the characteristics of primary care that matter most for individual and population health outcomes. The PCPCM is comprehensive, concise and has demonstrated good reliability and construct validity.4–5 The PCPCM is therefore a measure of the patient experience and perception of the core functions of primary care in the health care system.
The National Quality Forum and the Centers for Medicaid and Medicare Services have endorsed the PCPCM as a clinical quality measure. Although it has been available for use in the Merit-based Incentive Payment System (MIPS) since 2022, few large health systems have implemented it. Best practices for implementation, including how and when to administer the survey, reporting, and process improvement activities tied to survey data, have therefore not been established.6
To date, there are no published studies of the PCPCM performance in a large heterogeneous health care system. Prior studies are small in scale and limited to specific homogenous patient populations. Li et al reported PCPCM results from 6 affiliated academic family medicine practices in Toronto, Canada, showing associations with patient characteristics and association with measures of patient satisfaction. They showed that implementation was possible in this setting and that there was variability in survey responses by age, gender, health status, and number of years in care with a clinician.7 The PCPCM was adapted for use in a small sample of patients in urban Japan (not tied to a single practice or group) and in Hong Kong, and, in both studies, had good internal consistency, reliability, and validity.8–9 Finally, Ronis et al adapted the instrument for use with pediatric patients in a resident continuity clinic with a high proportion of low-income, African American patients and families, where they also showed good survey performance independent of key patient socio-demographic factors.10
Given the promise of the PCPCM - efficient, meaningful measurement of primary care’s core functions in a health system - it is important to prove scalability, incorporate into existing patient-reported measurement systems, and evaluate the performance in the context of a dynamic health system with diverse types of practices that vary by type, specialty, catchment, and payor mix. This study aimed to document the initial implementation process across a large health system that, because of its size and history of recent mergers, represents a diverse patient population engaged in a wide variety of different practice sizes and types. We document the challenges and successes experienced in this process, as well as an exploratory evaluation of the measure performance, including replication of the PCPCM’s psychometric properties and analysis of demographic differences in PCPCM scores.
Methods
Study Design
We conducted a cross-sectional survey study of empaneled primary care patients within Jefferson Primary Care, the primary care system of Jefferson Health located in the greater Philadelphia area, between April 2023 and January 2024. The project was reviewed and approved by the Thomas Jefferson University Office of Human Research (iRISID-2023 to 2528).
Participants
Jefferson Health (JH) is a nonprofit health system with 17 hospitals across Southeastern Pennsylvania and Southern New Jersey. Jefferson Primary Care (JPC) serves approximately 560,000 people in 98 practices across the region. For this study, we targeted our survey to patients from 76 practices in JPC that use the Epic Systems health record (Verona, WI). This included 53 primary care practices staffed by family medicine physicians and 23 staffed by internal medicine physicians. The study population was patients empaneled to these practices, defined as having a visit to any primary care provider (PCP) within 24 months before survey distribution and having a primary care physician or advanced practice clinician in the PCP field of the health record. This definition aligns with the definition of empanelment used in JH primary care practices for operational purposes.
Implementation and Survey Design
In Fall 2022, the research team met with a team from the Larry A. Green Center that had developed the PCPCM survey, to discuss PCPCM survey implementation at JH. The research team then reviewed options for implementation during late fall 2022 with the Thomas Jefferson University patient experience office in collaboration with Press Ganey (PG, South Bend, IN, https://www.pressganey.com). Based on our existing contract with PG, we were able to add the new measure, separate from ongoing visit-related patient experience surveys, at no additional cost. We created a survey that included the 11-item PCPCM4 as well as additional questions (“How long have you been a patient at your current primary care practice?,” and 2 free text prompts, “Is there anything additional you would like to share regarding your experiences with your primary care practice?” and “How can your care team best improve your experience as a patient in the future?”) We met with PG and Jefferson Health patient experience teams to systematically identify our population of interest, develop a process of ongoing monthly deployment of the PCPCM to patients as described below, and regularly collect respondent and nonrespondent data.
Empaneled adult patients at 76 primary care practices received survey links via text message or by e-mail (based on documented patient preference in the EMR) using the PG system starting in April 2023 (for patients with birthdays between January 1 and April 1 batched into the first group) and monthly surveys for subsequent birth months. Each patient received 1 reminder message and then 2 messages by e-mail (if available) or text message to complete the survey 1 month after their birth month during this period. Survey responses were collected by PG and combined with patient level demographic variables, practice site, and PCP name assigned in the Epic EHR and were distributed to the research team using a secure web-interface platform in batches. This data also included patient level demographic variables, practice affiliation, and PCP name for nonrespondents to facilitate comparison.
The PCPCM survey presents patients with 11 items, each of which is then rated by the patient using a scale of “Definitely,” “Mostly,” “Somewhat,” and “Not at all.” For scoring, this scale is converted to a 1 to 4 numeric scale with “Definitely” = 4 and “Not at all” = 1. The PCPCM questions and additional items can be found in the Appendix for reference.
Data Analysis
Descriptive statistics were calculated. The relationship between demographic variables and survey completion was analyzed using logistic regression, and adjusted odds ratios were reported. A single composite clinician-level PCPCM score was created by averaging across the 11 PCPCM items. To understand how the PCPCM composite score and individual item scores varied by patient demographic variables, we evaluated group differences with 2 approaches: univariate one-way ANOVAs and a multivariable full factorial ANOVA (ie, including all two-way interactions). One-way ANOVAs were completed to compare our responses with those in the current literature and then a full factorial ANOVA was conducted to account for all variables of interest within a single model and evaluate any interactions between demographic variables. Finally, validation evidence for the internal structure of the PCPCM survey was obtained using exploratory factor analysis (EFA) with the hypothesis of a single factor extraction. All analyses were conducted using SPSS (IBM, Armonk, NY).
Results
Demographics
Overall, 329,540 patients received the survey using the PG distribution platform (Figure 1). A total of 21,154 patients (6.4%) completed the survey, with 71.1% of respondents using a touch device (smartphone or tablet). Approximately 27.2% of respondents were 65 or older, 57.9% were female, and 65.1% identified as white or white (Table 1). The primary care clinician was a physician (MD or DO) for 73.4% of respondents and 40.2% had received care for over 10 years at their current primary care practice (Table 1).
Study participant flow diagram.
Demographic Differences in PCPCM Survey Completion
Respondent versus Nonrespondent Characteristics
A logistic regression was performed (n = 329,474) to examine the effects of 3 demographic variables, sex, age, and race, on completion of the PCPCM survey (Table 1). Pseudo R-squared estimates for the model were small: 3.5% (Cox & Snell R2) to 9.3% (Nagelkerke R2). There were demographic differences in respondents and nonrespondents in sex, race, and age (p < 0.001). Women were more likely to complete the survey than men (aOR = 1.11, 95% CI: 1.08–1.15). White or white patients were more likely to complete the survey than Black or African American patients (aOR = 1.27, 95% CI: 1.22–1.32), Asian or Pacific Islander patients (aOR = 1.79, 95%CI: 1.65–1.94), and Hispanic or Latino patients (aOR = 1.15, 95% CI: 1.06–1.24). Lastly, for patient age, patients who were 65+ were more likely to complete the survey than 18 to 35-year-olds (aOR = 8.91, 95% CI: 8.39–9.46), 36 to 49-year-olds (aOR = 4.94, 95% CI: 4.71–5.18), and 50 to 64-year-olds (aOR = 2.07, 95% CI: 2.01 to 2.14). The relative proportions between our sample and the total surveyed population are visualized for racial groups in Figure 2 and for age groups in Figure 3.
Relative proportion by race.
Relative proportion by age group.
A second logistic regression was run including only patients who opened the survey (n = 46,514) to explore whether the device the patient used influenced the relative odds of survey completion while controlling for the previous demographic variables. Pseudo R-squared for this model were larger: 23.3% (Cox & Snell R2) to 31.2% (Nagelkerke R2). Patients using a touch device were much more likely to complete the survey, compared with those using desktop devices (aOR= 4.95, 95% CI: 4.74–5.17).
Person-Centered Primary Care Measure (PCPCM) Psychometrics and Composite Means
Exploratory factor analysis confirmed the 11 items formed a single factor (eg, primary care quality). This single factor solution explained 64% of the variation in the 11 items. The scale demonstrated excellent reliability providing evidence that the 11 scale items are closely related (α = 0.94).
The sample’s mean composite PCPCM score was 3.36 with a standard deviation of 0.71. All but 2 questions had a mean above 3.0 representing ‘Mostly’ agree. Question 6 “My doctor and I have been through a lot together” and question 9 “The care I get in this practice is informed by knowledge of my community” had means of 2.79 and 2.99, respectively. There was a significant difference in mean scores by age, sex, race, primary care clinician type, and the number of years the patient had been at their current primary care practice (Table 2).
Demographics of Respondents and Person-Centered Primary Care Measure (PCPCM) Score Demographic Comparisons
To further understand variability in the composite PCPCM score, a full factorial ANOVA was executed with patient age, race, sex, and years with their primary care practice as predictors, along with all interactions. No interactions were statistically significant. The main effects of patient age (p < .001), race (p = .011), and years at current practice (p < 001) were statistically significant, while the effect of sex was not (p = .187). post hoc comparisons were completed using Tukey’s HSD test. The mean composite PCPCM score of individuals 65 or older was the highest of all 4 age groups (p < 0.001). Patients who were 50 to 64-years old had a higher composite PCPCM score than patients who were 36 to 49-years old and 18 to 35-years old (p < 0.001). The mean PCPCM score did not significantly differ between patients who were 36 to 49-years old and 18 to 35-years old (p = .241).
Patient’s reported length of time with their current primary care practice was stratified into 4 categories. Each category had a statistically distinct PCPCM mean score, with patients who had a longer relationship with their practice having a significantly higher mean PCPCM score (p < 0.001). Patient age and years with their practice were significantly correlated (rs = 0. 216, p < .001), but there was no interaction between these variables (p = .506). For patient race, the mean PCPCM score was higher for White or White patients than Black or African American patients (p = .041) and other/unknown race patients (p = .003). It was not significantly different from the means for patients that identified as Asian or Pacific Islanders or as Hispanic or Latino.
Person-Centered Primary Care Measure (PCPCM) Item Means
Examining the individual PCPCM item means, we found similar ranked order differences to the mean composite PCPCM discussed above. There was a significant difference between age groups for all 11 PCPCM items (p < 0.001) (Figure 4). Similarly, there was a significant difference based on the duration of time at their current primary care practice for all 11 PCPCM items (p < 0.001) (Figure 5). This effect was the strongest for Q6, “My doctor and I have been through a lot together” (ƞ2 = 0.12). Generally, higher patient age and duration of time at their current primary care practice was related to higher PCPCM item means although differences were not statistically significant across all levels and effect sizes were small (Figure 2 and 3).
Person-centered primary care measure items by patient age. Abbreviation: PCPM, Person-Centered Primary Care Measure.
Person-centered primary care measure by years with practice. Abbreviation: PCPM, Person-Centered Primary Care Measure.
For race, significant differences were found for all items except Q8, “The care I get takes into account knowledge of my family.” However, racial differences were small in effect size and not uniform in direction. The greatest effect is found for Q9 “The care I get in this practice is informed by knowledge of my community.” Black or African American patients reported a significantly lower score for this item than all other racial groups (p < 0.001, ƞ2 = 0.006).
Discussion
We found it was possible to deploy the PCPCM in a large health system made up of heterogeneous primary care practices using the existing PG patient engagement platform and that showed, on average, consistently high scores on the measure across a diverse group of practices and clinicians. As shown previously, the PCPCM had strong internal structure, and responses skewed high with a mean score of 3.36.9–10 The response rate was much lower than expected (6.4%), with a majority not opening the survey, and less than half completing the survey after opening it. Women, White patients, and older patients were more likely to complete the survey when compared with other demographic groups. For comparison, other studies have reported higher response rates (12.8% in Li et al.) and, in general, our PG response rate for postencounter surveys in our system is consistently around 15%. Anecdotally, author RE, Co-Director of the Larry A. Green Center, and developer of the PCPCM notes that ongoing implementation in a Virginia-based system has response rates between 2 to 12%, and at least one other vendor has found response rates varying from 6 to 30%, dependent on whether the practice type and implementation strategy. While our low response rate limits generalizability, we obtained a large data set including responses from all demographic groups. CMS established standards for using the PCPCM based on the number of responses per clinician assessed (requiring a minimum of 30) and not response rate. In our implementation, most clinicians received more than the required 30 responses, and most practices received well over 150 responses.11 We are actively engaging with patients and patient experience leaders to improve the PCPCM response rate at Jefferson Health. However, considering the low response rate, the results of the evaluation of the PCPCM should be considered in need of confirmation in a sample with a greater response rate.
Exploratory analysis of patient age and years under a clinician’s care showed association with higher PCPCM scores. Younger patients had lower response rates, suggesting that more work is needed to improve engagement with this group in primary care. Older individuals are more likely to have longer-term care relationships with their primary care clinician and may be best suited to opine about their experience with primary care. Based on this observation, it may be necessary to consider practice-level or clinician-level data in the context of differing panel characteristics like patient age or time under care.
The “knowledge of my community” question (Q9) stands out as an area for inquiry in future studies. Average responses were lower on this question compared with the others (except for Q6, “My doctor and I have been through a lot together”), and it varied by years under care. Notably, Black and African American patients responded with lower scores on this question when compared with non-Black patients. We believe that this difference is important and speculate that more work needs to be done to explore how practices and clinicians can better understand patients’ community and its impact on the delivery and function of primary care in health systems. More work is needed to address how experiences of racism affect the characteristics of care measured by the PCPCM, and how antiracism efforts can improve the patient experience of primary care.
In conclusion, the PCPCM instrument can be implemented across a large primary care system through a standard patient experience platform. Further study is needed to develop best practice approaches to improve patient engagement by varying distribution methods, preamble language, in-office marketing and communication, and follow-up reporting about practice improvements that stem from survey responses. In addition, work is needed to validate PCPCM responses with other patient experience measures, operational and utilization data, and HSRN screening responses. An understanding of variability among different demographic groups will help inform how PCPCM responses can be used in clinician- or practice-level comparisons to drive quality improvement efforts. Understanding which factors impact PCPCM responses will guide practice- and system-level quality improvement interventions. These will be aimed at optimizing the core functions of primary care in the health care system and focus attention on areas that may impact both clinical outcomes and patient experience where narrowly focused quality measures fail. Based on our observations, more work is also needed to focus on improving the quality of primary care, measured by the PCPCM, particularly for younger patients and Black patients. Investment of time and resources in developing best practices for implementation and interpretation of the PCPCM is worthwhile given the importance of measuring meaningful aspects of primary care patient experience.
Acknowledgments
The authors acknowledge the essential contributions of clinicians and patients who participated in this study.
Appendix
Person-Centered Primary Care Measure (PCPCM)1
My practice makes it easy for me to get care.
My practice is able to provide most of my care.
In caring for me, my doctor considers all factors that affect my health.
My practice coordinates the care I get from multiple places.
My doctor or practice knows me as a person.
My doctor and I have been through a lot together.
My doctor or practice stands up for me.
The care I get takes into account knowledge of my family.
The care I get in this practice is informed by knowledge of my community.
Over time, my practice helps me to stay healthy.
Over time, my practice helps me to meet my goals.
1The response format is as follows: Definitely (4), Mostly (3), Somewhat (2), Not at all (1)
Notes
This article was externally peer reviewed.
Funding: The authors received no financial support for the research, authorship, or publication of this article.
Conflict of interest: The authors declare that there are no conflicting or competing interests.
- Received for publication December 3, 2024.
- Revision received February 7, 2025.
- Revision received February 11, 2025.
- Accepted for publication February 24, 2025.
