Choosing a Primary Care Provider: A Formative Qualitative Study to Identify Patient Preferences

Abstract

Purpose

(1) Identify what characteristics patients evaluate when choosing their primary care provider (PCP), (2) characterize how provider characteristics factor into patients’ choice of PCP, and (3) develop a list of features for a quantitative preference study in the form of a discrete choice experiment (DCE).

Methods

We used a cross-sectional, qualitative study to accomplish the objectives. We recruited participants from an internal medicine clinic and the general public in North Carolina. We used convenience and purposive sampling to recruit participants as a single sample. All participants completed a semi-structured interview. We then used a reflexive thematic analysis to analyze the data. We conducted a focus group to confirm and refine the interview findings. Finally, we created and prioritized a list of attributes and attribute levels for use in a future DCE study.

Results

We conducted 12 interviews to reach data saturation. Participants mostly identified as female (75%, n=9), white (67%, n=8), and non-Hispanic/Latino (83%, n=10), with a mean age of 39 years (SD=15). Participants reported examining factors across several domains when deciding on their PCP: (1) knowledge/experience/age, (2) gender alignment, (3) convenience, and (4) cost. Based on the results of the interviews, we selected five preference-sensitive attributes and confirmed these with the focus group. Attributes that will be included in a future study include provider profession, provider race and gender, an attribute describing what the provider is known for, appointment availability, and driving time.

Conclusions

We found that participants reported evaluating several important factors when deciding on selecting a PCP. Elucidating patients’ willingness to accept different types of PCPs has policy implications for provider integration, cost-savings, and workload optimization.

Patient-Friendly Recap.

• More patients are having to wait longer to be seen by their primary care provider (PCP) – a doctor, PA, or nurse practitioner – because of a PCP shortage.

• We wanted to know what people prefer in their PCPs so that we can improve patients’ ability to get seen by them.

• People most cared about how their PCPs treated them and talked with them, not just what college degrees their PCPs had.

• If we have more types of PCPs, it could help more people get seen by a PCP quicker.

A primary care provider (PCP) serves as the central point of contact for patients within the healthcare system and helps provide or coordinate access to a range of healthcare services.¹ PCPs provide comprehensive care, preventive services, and coordination of specialized care. This critical role makes the process of choosing a PCP a significant decision for individuals and their families, one that often involves a careful evaluation of personal preferences, medical needs, and logistical considerations. The importance of PCPs is underscored by the fact that they comprise the only medical specialty whose supply is positively associated with population health.²

The US has faced a long-standing primary care physician shortage; the situation was so dire in the 1960s that two new health professions, the Physician Associate (PA) – previously, Physician Assistant – and Nurse Practitioner (NP), were created to specifically address the shortage of primary care physicians.^3,4 PAs and NPs play an increasingly important role in addressing this shortage, yet there is limited literature on patients’ perceptions of PA and NP contributions to the healthcare system.⁵ Previous work has identified that few patients understand the distinctions between physicians, PAs, and NPs in clinical work, even though such patients probably have already seen a PA or NP for care.⁶ One prior study found that patients’ preferences for a provider may vary depending on the provider’s credentials.⁶ A study of the Association of American Medical Colleges’ annual patient survey found that patients valued physicians’ years of training when choosing a physician but chose a PA or NP because of their attitude/demeanor and greater availability.⁷ Provider selection is highly preference sensitive but has been understudied in the US and would benefit from a stated preferences approach to better understand patient preferences.

A discrete choice experiment (DCE) is a widely used, theoretically-grounded quantitative method to elicit consumer preferences for products or services. DCEs quantify the tradeoffs that individuals are willing to make when choosing an option among a range of alternatives. DCEs involve a survey, for which participants are asked to complete a series of choice tasks, in which the characteristics (“attributes”) of the product or service (in this case, the PCP) are systematically varied by means of an experimental design.⁸ The series of choices is analyzed to derive preference weights that describe the relative importance of each attribute. DCEs have been used to study a variety of preference-sensitive decisions but have yet to be used to study patients’ PCP preferences in the US context. Formative research is needed to support the relevance and realism of attributes.⁹ Conducting qualitative research to assist with attribute generation and level selection for a future DCE is considered a good practice by the Professional Society for Health Economics and Outcomes Research (ISPOR).^10,11 Thus, the objectives of this formative qualitative study were to (1) identify what attributes patients evaluate when choosing their PCP and (2) prioritize a list of attributes and attribute levels for inclusion in a DCE.

METHODS

Reflexivity Statement

One author (AAA) conducted the interviews and focus group. The interviewer is an experienced qualitative researcher. The PI is an experienced researcher and principal faculty member in both physician (MD) and PA training programs. The interviewer and PI collaborated intensively throughout the data collection process. The research team included additional experienced health services researchers (CE, JO) and currently practicing primary care clinicians (SG, JV, RG). One PCP (SG) is a PA, one is a physician (MD), and the other an NP (RG). To minimize bias toward any particular profession, the research team collectively created and iteratively revised the interview guide, with input from two PAs, a physician, an NP, and non-clinician researchers.

Study Design

We conducted a qualitative study using a constructivist approach. The constructivist approach to qualitative research posits that individuals can experience multiple and subjective realities.¹² We chose a constructivist approach for this study to better understand the subjective experiences and realities of individuals.¹² We used the Standards for Reporting Qualitative Research (SRQR) to guide the reporting of this study, and they can be found in Appendix A.¹³

Sample and Sampling

Between October 2022 and March 2023, we conducted 12 individual, in-depth semi-structured interviews in North Carolina. We used convenience sampling to recruit general members of the public using a public-facing recruitment website for study sign-up, where people could learn more about studies at an academic medical center. We also used purposive sampling at a general internal medicine primary care office that serves a large number of underserved patients to enhance recruitment of participants who are typically underrepresented in research. Those who were eligible for inclusion were (1) at least 18 years of age, (2) seen by a healthcare clinician in the past, and (3) able to read English. One interview was conducted in Spanish. Prior work has found that sample sizes as low as n=9 can reach thematic saturation.¹⁴ The PI (CG) and interviewer (AAA) met continuously throughout the data collection process to determine when thematic saturation had been met. We determined that thematic saturation was met with 10 interviews and conducted two additional interviews to confirm saturation. That is, we conducted one-on-one interviews with 12 participants, at which point thematic saturation was deemed sufficient. Most participants identified as female (75%, n=9), white (67%, n=8), and non-Hispanic/Latino (83%, n=10), with a mean age of 39 years (SD=15, Range: 21–63). Most participants (67%, n=8) reported their PCP to be a physician (MD/DO). Most participants identified their highest grade of schooling as a college graduate or more (67%, n=8) with a combined household annual income of $75,000 or more (50%, n=6).

Data Collection

Semi-structured interviews were used to explore which provider attributes patients evaluate when choosing their PCP and which PCP attributes are most desired. All interviews were audio-recorded using WebEx by Cisco (San Jose, California). Participants were asked to speak about their experiences in selecting and interacting with their current and past PCPs, as well as their selection criteria for an ideal provider.

The ISPOR guidelines, prior literature, and Andersen’s Behavioral Model of Health Services Utilization informed the study design and development of the interview guide.^11,15 Specifically, we used the Andersen Behavioral Model to elicit why a patient would decide to seek care from their PCP. During the interviews, participants completed an adaptive choice-based conjoint (ACBC) exercise, and participants described their ideal PCP. Participants were also asked whether their current PCP matches their most desired characteristics. Finally, we asked about the provider’s credentials given that patients frequently encounter non-physician providers, such as PAs and NPs. Prior to fielding the interview guide, two authors (AAA and CG) conducted a pilot test with four non-healthcare providers who were unaware of the study’s hypotheses. The research team modified the interview guide and items after the pilot to ensure comprehension and clarity. The final interview guide is presented in Appendix B.

Data Analysis

Two experienced qualitative researchers (AAA and KR) reviewed the transcripts and developed a codebook to capture the coding hierarchy from concepts found in the data. Using the research questions, concepts from the interview guide, and ideas that emerged from the participant discussions, both deductive and inductive coding strategies were utilized to develop the codebook.

Interview transcripts were imported and coded in ATLAS.ti version 23 software using thematic analysis.¹⁶ The two qualitative researchers independently coded the textual data and met weekly to compare coding. Discrepancies in coding were discussed and resolved iteratively using consensus building (ie, the researchers explored different interpretations of the data, considered the context of the coded segment, clarified code definitions as needed to align with reconciliation decisions, and recoded segments based on agreed-upon definitions). In some instances, both researchers’ codes were applied to the data when both interpretations were agreed upon (eg, double-coding “PCP selection” and “Credential and Training” when the participant was explaining the level of experience or education contributes to their choice of who to see). The codebook was adjusted as needed based on these discussions. Text segments were reviewed by code and summarized. Code summaries were synthesized into themes and organized using principles of reflexive thematic analysis.¹⁷ After identifying themes, the valence of the attributes that were captured were then categorized as positive, negative, or neutral/indifferent.

Focus Group Discussion

We recruited two additional participants for an internet-based focus group to gather additional data and create the attributes and levels that might be included in the future DCE. We chose not to conduct the focus group with the previous participants to assess whether the previous participants’ experiences in choosing PCPs were similar and to evaluate the initial attributes selected by the research team.¹⁸ The focus group occurred in June 2024. Participants were recruited through the same public-facing website as the original participants; neither of the focus group participants participated in the interviews. During the focus group, participants were asked to complete two hypothetical choice tasks with two potential provider selection scenarios with one fixed attribute (credentials, eg, MD/DO or PA/NP) and varied attributes (eg, distance from house, cost). Participants were asked which provider they preferred, and which attribute(s) were most influential in making that choice. Participants were also asked about information outside of what we presented that also influenced their choice. Participants then discussed and helped define other potential attributes and attribute levels for inclusion into the DCE (eg, provider demeanor). The focus group data were summarized separately using a rapid analysis approach that focused on cross-cutting preferences and experiences.

Ethical Considerations

This study was approved by the institutional review board (IRB) at the first author’s medical center. All participants provided informed consent to participate in the study.

Selection of Attributes for Inclusion in the DCE

The research team met three times in 2024 to generate the initial list of attributes and potential levels. The team continually refined and iterated on the list of attributes and levels via once monthly team meetings and emails until March 25, 2025. The PI also conducted a search of PubMed/MEDLINE to identify other DCEs that studied patient choice in PCP or other types of healthcare providers (eg, dentists). We used the published literature to evaluate important correlates of patient satisfaction as potential attributes in the future DCE. Attributes used in these studies were extracted and compared to those identified using the semi-structured interviews. Initial attributes were then refined over email. We paid particular attention to choice-relevant and modifiable PCP characteristics.

RESULTS

Main Qualitative Findings

Participants reported examining several factors related to the PCP themselves, including their experience, age, and gender concordance. Participants sought knowledgeable and experienced providers, those who were older, and those who shared the same gender as the patient. Participants also reported valuing convenience in terms of the location of the practice and ease of scheduling appointments. Cost considerations, such as insurance coverage and transparent pricing, were also significant factors in selecting a PCP. Table 1 presents the themes, key findings from the interviews, and illustrative quotes.

Table 1.

Thematic Findings From PCP Selection and Preferences

Theme	Key Findings	Illustrative Quote
Selection
Provider characteristics	Participants examine several factors related to the PCP themselves, including experience, age, and gender concordance	I feel more comfortable if they’ve done it for a few years, not just starting out. I’m just more comfortable that way (P01)
	Participants sought knowledgeable and experienced providers who provided high-quality care	For me personally, I would prefer if it’s a woman. I just feel more comfortable. I had a male pediatrician when I was growing up, and I remember there was an age that I just felt a little uncomfortable having a male and talking openly about what I felt I needed to talk about in terms of having a menstrual period and things of that nature (P02)
Convenience and cost	Participants valued convenience in terms of the location of the practice and ease of scheduling appointments	I think having access to your PCP is, or at least to their office is, important. If you can get an appointment within a week or so, especially for sick visits like the same day, it’s probably the best (P04)
Preferences
Ideal characteristics	Participants desired a PCP who embodies key characteristics, including compassion, empathy, knowledge, thoroughness, strong interpersonal skills, and effective listening. These traits are considered to foster trust, comfort, and shared decision-making	The most important [factor] to me, I think, is compassion. For me, if I feel that if that person is compassionate, then I think they will do everything they can within their means to make sure that we are addressing my healthcare needs (P01) When it’s more of a conversation of just like what do you think is the best decision or it really feels like they’re listening to me and…my concerns and taking my input and then they add the-I still trust their medical context, but when I see that they hear me and then they follow up with the clinical-medical reasoning behind what the issue is, it’s much easier to trust them. (P06) Someone who is knowledgeable in just the primary care aspects of everyday life…So being up to date on all the quality that you need to do, whether it’s a colonoscopy or any number of screenings that you may need as you age (P04)
Negative characteristics	Participants are likely to avoid PCPs that are dismissive or rush through a visit	The biggest one would probably just like writing off a complaint that you have as if there’s nothing, instead of really taking what you say to heart, and understanding what you’re going through. (P04)
	Other perceived negative attributes included, but are not limited to, arrogance, intimidation, and lack of personability and care	…[older doctors] come off as more condescending… (P06)
Neutral/indifferent characteristics	Some participants expressed indifference towards a PCP’s educational background, gender, age, or personality	…it’s nice for somebody to be nice…but I don’t mind a person that is abrupt or doesn’t have that type of personality, as long as they know what they’re doing; they can come up with a proper diagnosis and prognosis (P09)

PCP Selection

Participants offered insights into their decision-making factors when selecting a new PCP. The most frequently described factors included experience, age, gender, location, availability, and cost. When discussing experience and age, some participants indicated a preference for having someone who is “knowledgeable” and experienced in treating someone in their age group. Several participants expressed the desire for a PCP whose gender aligns with their own, noting differing levels of comfort or confidence with male and female providers.

Some participants described a priority of convenience when selecting a new PCP, emphasizing the location of the practice and the availability of appointments. Participants wanted practices that were “in the vicinity” (P04) or “close to [their] house” (P06) to minimize their commute and time required for an appointment. A few participants indicated that cost and insurance coverage dictated where they sought primary care. They selected from locations that were covered by their insurance (P04) or that did not require a co-pay (P03).

Similar to the interviews, the focus group participants preferred PCPs to use a holistic approach and shared decision-making in the first choice task. Participants expressed that they felt more confident about their medical care when providers encompassed more than physical health in their treatment plan. Similarly, participants felt more at ease when providers gave them a sense of autonomy over their health with shared decision-making. Participants did not have a preference for an MD or an NP when given a choice; however, they did state that it was easier to get an appointment with an NP. Furthermore, participants stated that they felt NPs had a more “human-centric” approach to care, as NPs spend more time with patients and are more compassionate. Participants reported receiving better care from NPs than MDs but stated that they would equally value MDs when provided with similar care.

In the other choice task, participants in the focus group favored the PA provider in the second item. Availability and out-of-pocket cost (OOPC) were the attributes with the biggest influence for these participants. Participants expressed that if they had an acute medical concern, they would want to be seen by a provider as soon as possible; however, if it were a more in-depth visit, such as one on which they were being followed up regarding a chronic medical decision, they would be willing to wait more time. Cost was also an important factor for participants. Participants would reconsider seeing their preferred provider if they had to pay $50 over $15 OOPC; however, availability was still the preferred attribute if they needed to be seen for an acute medical condition. Participants reiterated that they did not have a preference between a PA or MD and that they would go with the provider that was more readily available and more affordable.

PCP Preferences – ACBC Exercise

Participants were also asked to describe characteristics that their ideal PCP would demonstrate. The most frequently reported responses included being compassionate, knowledgeable, thorough, having good interpersonal skills, being a good listener, and being accessible/available. Participants were asked to identify which of the characteristics were the most important to them in a PCP. The primary responses included good listening or communication, being knowledgeable, and being compassionate. Figure 1 depicts the prevalence of ideas in the data, with larger circles representing the most commonly discussed content and smaller circles representing less frequently discussed topics.

Figure 1.

Flow chart of the study population based on scheduling coordination types, n = 1626.

Compassion was discussed in a number of ways, with several participants directly stating compassion or using similar words like empathy or understanding. Some participants alluded to interpersonal skills using concepts like engagement, connection, and bedside manner. They described a desire for a PCP who makes them feel comfortable and someone who connects with them. More than half the sample mentioned their providers’ ability to listen well. Listening was also brought up in discussions of shared decision-making. A handful of participants identified shared decision-making as an ideal characteristic for a PCP.

Some participants wanted a PCP who is knowledgeable about their health and concerns, as well as what procedures or tests that they should get, especially as they age. Other participants described thoroughness as an ideal treat, emphasizing the value of a thorough exam in not missing an underlying issue. Other factors that emerged included a provider’s reputation, dependability, and approach to holistic care.

Negative Attributes

The most frequently discussed negative attributes that participants wanted to avoid from their PCP were dismissiveness and rushing. Participants described dismissiveness as being, “…patronizing and judgmental and not listening to me”.

Several participants also indicated that rushing, interrupting patients, and appearing distracted all contributed to a more negative experience with a provider, although not all instances were specific to PCPs. One participant speculated that some providers seem rushed because they are short-staffed, and one provider was unprofessional during this participant’s visit (eg, aggressive). Participants also noted that some PCPs were arrogant, intimidating, or not personable/caring. One participant associated age with arrogance, explaining that “[older doctors] come off as more condescending” and younger doctors tended to have “more compassion” (P06).

Neutral/Indifferent Characteristics

Participants described PCP characteristics that were nice to have but not necessary, as well as traits to which they were indifferent. Several participants stated that, although it is nice to have a PCP who is personable, it does not dictate whether they will keep seeing that provider. Other factors listed as nice but not necessary included a PCP who speaks the same language as they do, simplified their health information for them, was nearby, offered weekend hours, and had an established history with them.

Attributes and Attribute Levels

Appendix C presents the final attributes and potential attribute levels selected for inclusion in a future DCE. The research team deliberated on which attributes and levels to include via virtual meetings. The first objective was to identify which attributes should be prioritized for inclusion based on our previous work and the current findings. The research team started deliberation with 21 potential attributes for inclusion. Based on the qualitative findings of both the interviews and focus group, we prioritized shared decision-making, demeanor, and providing holistic care for inclusion. We operationalized these concepts into one attribute to describe the type of care that the provider was best known for providing. Throughout the process, the research team took care not to include attribute levels that would be interpreted as overtly negative by participants and would thus never be preferred by patients (eg, a physical confrontation with a PCP). We did not include cost of care because participants reported that cost, while important, was not as important a factor as availability when given a choice due to insurance cost-sharing. We prioritized convenience by operationalizing the attribute as driving distance measured in minutes (as reported by participants). We also operationalized time to be seen in three levels. Because female participants reported that gender concordance was important to them, whereas male participants did not voice much of a preference for gender concordance, we included a gender attribute. We chose to include provider type, operationalized as physicians (MD/DO), PA, and NP, because prior research has shown that patients see the type of care delivered by these different types of providers variably and a preference for the type of care might also drive preference to a type of provider.⁷

DISCUSSION

This study is among the first that we are aware of to identify PCP preferences among the general public. Through a combination of qualitative interviews and a focus group, we found that participants place higher value on a PCP’s communication style, competence and accuracy of diagnosis, and demeanor/attitude over other characteristics (eg, level of training). Our work further demonstrates that choosing a PCP is a preference-sensitive decision, where individuals place significant importance on their choice. Improving the match between patient preferences and PCP selection has the potential to improve patient satisfaction with their provider and ensure optimal outcomes.

Our results also confirmed previous research showing that patients want to share responsibility with their PCP when deciding on a course of treatment, consistent with recommendations that healthcare providers should use shared decision-making and include patients’ perspectives and preferences in their treatment plans.^19,20 Our work supports prior research that patients value communication style, empathy, and competence.^21–23 Women preferred gender concordance with their PCP, also consistent with prior studies.²⁴ Our work is unique in that we used a qualitative study design compared to studies by Mercado and Razzouk.^22–23 Mercado and colleagues compared patients’ preferences for their PCP in family medicine versus internal medicine using a cross-sectional rating and ranking questionnaire.²² Razzouk looked at choosing a primary care physician in the context of a Health Maintenance Organization policy and used HEDIS measures.²³ A major drawback of these two studies is the lack of a trade-off component. Since real-world decision-making involves trade-offs, eg, less driving time but longer wait times, more research is needed to elicit those tradeoffs that patients are willing to make. The qualitative, formative work of our study provided a first step towards designing a future study in which policy- and preference-relevant tradeoffs can be identified in a future research project.

Prior work investigating preferences for PCPs has all been conducted in countries other than the US with primary healthcare systems that are remarkably different. The closest in geography to the US and another country with an advanced economy, Canada, included only two similar attributes to our future DCE – time to get an appointment and the type of provider.²⁵ In that study, only physicians, nurses, and resident physicians were included in the qualitative work, which might explain the differences to our study. Other attributes included in this study were how patients booked their appointment, how long it would take to see the provider, and how well the provider knew the patient. Among adults aged 18 and older already using a family practice clinic, the most valued attribute across scenarios was time to get an appointment, while clinic wait time was least important.²⁵

Prior DCEs in provider choice in the US have focused primarily on primary or urgent care provided in retail clinics, such as retail pharmacies.^26,27 These reports found that participants preferred a pharmacist providing primary care services with physician oversight, over PAs or NPs, in a retail pharmacy setting.²⁶ One other study conducted among urban adults in the state of Georgia investigated patients’ preferences regarding providing primary care services in retail clinics from NPs compared to physicians in private clinics.²⁷ In that study, participants preferred a physician in a private office compared to an NP in a retail clinic setting (eg, grocery store or chain pharmacy). The authors did not report whether they used qualitative research to elicit attributes and attribute levels. Again, that study focused more on the setting (eg, retail clinic) and in relation to urgent care level needs or situation, not necessarily preferences for their PCP. Our study attempts to advance what is known about preferences for PCP and PCP services.

It is impossible to predict how our final attribute list might differ if a more diverse or broader sample were used. Prior literature suggests that racial concordance between providers and patients may be associated with greater patient satisfaction and might therefore be an important attribute.²⁸ Recent research using multivariable regression models suggests that racially concordant provider-patient relationships may not be as important as initially thought and that other aspects of the clinic visit or provider-/patient-level factors may be more important in predicting visit satisfaction.^29–30 Future research is needed to better understand other factors that are important regarding patient satisfaction with care.

Like all qualitative studies, our study has limitations. First, we conducted the study in one US state, so the results may differ compared to other US locations. Second, our study sample was not representative of the area in which recruitment occurred; however, we used purposive sampling to complement convenience sampling to increase the representation of participants who are typically underrepresented in research. Third, our list of physician/PA/NP characteristics that we included in the interviews, and that we will include in the future survey, was defined a priori from prior literature and expert opinion; therefore, other attributes could also be influential when choosing a primary care provider. We did not include any measure of cognitive function in our inclusion criteria; however, none of our participants exhibited signs of cognitive impairment, so the risk of this limitation impacting our findings is low. While qualitative research is not designed for generalizability, our findings show participants making meaning of preferences in their choice of PCP.

CONCLUSIONS

Our research builds upon previous work examining factors that most patients want in their PCP. The goal of this study was to provide formative evidence to help the research team identify relevant attributes and attribute levels for a future DCE. We found that choosing a PCP is a preference-sensitive decision, and patients evaluate multiple characteristics simultaneously to help them choose a provider. Based on our findings, patients seem to place the greatest weight on interpersonal skills, the provider’s gender (for women), and clinic location. As such, we created a list of attributes for a future DCE to identify trade-offs that patients are willing to make to get their preferred PCP. The findings of this future DCE will enable policymakers, administrators, and medical educators to improve the match between patients’ preferences for a PCP and actual care.

Data Availability

The study’s data are available upon reasonable request to the corresponding author.