Abstract
Introduction
With the rising global prevalence of diabetes, traditional hospital-centered healthcare models are becoming increasingly inadequate to meet patients’ long-term management needs. Community pharmacy services have the potential to address these gaps. However, systematic research on patient preferences for such services remains limited.
Aim
This study aimed to determine the factors influencing diabetic patients’ preferences for community pharmacy service models using a discrete choice experiment (DCE), and to explore how these preferences vary according to glycemic control status.
Method
Attributes and levels of DCE were identified through literature review, expert consultation, and patient pretesting. Six attributes were included: service content, service provider, follow-up program, reduction in cardiovascular events, reduction in hypoglycemic events, and service costs. A mixed orthogonal design generated 18 choice sets, with one scenario serving as reference. The survey was conducted face-to-face with patients with type 2 diabetes attending three community health service centers in Nanjing, China. Conditional logit models were applied to estimate attribute importance and subgroup analyses were performed based on HbA1c levels.
Results
Of the 237 respondents, 190 provided valid responses (response rate: 80.1%). Across the overall sample, the most influential attributes were pharmacy service content (Wald = 389.188, relative attribute importance [RAI] = 4.59), followed by the frequency of hypoglycemic reactions (Wald = 284.155, RAI = 4.19), service cost (Wald = 168.079, RAI = 4.07), reduction in cardiovascular events (Wald = 144.608, RAI = 3.38), service providers (Wald = 142.913, RAI = 3.29), and follow-up programs (Wald = 50.644, RAI = 1.15). Patients preferred drug effectiveness assessments over counselling or adverse reaction monitoring, valued collaborative care involving pharmacists and physicians, and demonstrated willingness to pay approximately ¥20 per session. Subgroup analyses revealed that patients with HbA1c ≤ 8% were more cost-sensitive and favored fixed-interval follow-up, whereas those with HbA1c > 8% preferred individualized follow-up programs and placed less emphasis on service costs.
Conclusion
Patients with diabetes prioritize service quality, outcome-driven care, and multidisciplinary collaboration when selecting community pharmacy services. While hypoglycemia reduction and cost are important across groups, preferences vary according to glycemic control, highlighting the need for tailored patient-centered service models. These findings provide evidence to guide the development of sustainable and responsive community pharmacy services.
Supplementary Information
The online version contains supplementary material available at 10.1007/s11096-025-02080-1.
Keywords: Choice behavior, Community pharmacy services, Diabetes mellitus, Health service accessibility, Interprofessional relations, Patient preference
Impacts statements
Community pharmacy services should prioritize drug effectiveness assessment and hypoglycemia prevention, shifting beyond the traditional dispensing and counselling roles.
Physician–pharmacist collaborative care models are strongly preferred by patients and can enhance trust, adherence, and treatment outcomes.
Differentiated pricing and follow-up strategies tailored to glycemic control levels are needed to align services with patient expectations and improve sustainability.
Introduction
The increasing prevalence of chronic diseases, particularly diabetes, is placing increasing pressure on healthcare systems globally. Traditional hospital-centered models of care are often insufficient to meet patients’ growing and long-term management needs. Community pharmacies have emerged as key components of primary healthcare delivery, offering accessible services that extend beyond dispensing to include medication counselling, chronic disease management, and patient education [1]. International experience demonstrates that well-developed community pharmacy services can improve medication adherence, reduce treatment complications, and enhance health outcomes [2].
Despite these advances, the scope and quality of community pharmacy services remain heterogeneous across different countries. In many settings, pharmacists are still primarily engaged in dispensing, whereas specialized services such as structured counselling, follow-up, and clinical monitoring are less frequently implemented [3, 4]. As the demand for rational medication use increases, challenges persist in service standardization, professional training, and sustainable funding mechanisms [5]. Compared to mature systems abroad, community pharmacy services in some regions remain at an exploratory stage and lack a systematic and standardized service model [6]. In the absence of unified operational standards, the exploration of service models that align with patient needs has become critical [7]. Understanding patient perspectives is particularly important because preferences strongly influence the acceptance, adherence, and success of service models [8, 9].
The discrete choice experiment (DCE) offers a rigorous methodology for quantifying patient preferences. Grounded in Lancasterian and random utility theories, DCE enables the identification of attributes that patients value the most and the trade-offs they are willing to make [10]. This method has been increasingly applied in health service research, including investigations into preferences for medications and healthcare delivery models [11, 12]. Previous studies on diabetes care have highlighted the importance of factors, such as hypoglycemia risk, cardiovascular benefits, and service costs, in shaping patient choices [13–16]. However, most existing research has focused on treatment preferences or institutional care settings, with relatively limited attention paid to the design of community pharmacy service models [17, 18], and few studies have simultaneously addressed all the essential elements of service implementation, such as who provides the service and how follow-up arrangements are organized in a comprehensive manner.
Aim
This study aimed to use a discrete choice experiment to investigate diabetic patients’ preferences for community pharmacy service models and to evaluate how these preferences vary according to the glycemic control level.
Method
Study design
This study employed DCE to investigate community pharmacy service preferences among patients [6]. The methodology and procedures followed the ISPOR Task Force Report guidelines for conducting well-designed DCEs [19]. The overall trial design roadmap is shown in Fig. 1.
Fig. 1.
Design roadmap
Identification of attributes and levels
Attributes and levels were identified through literature review, field interviews, and pretesting. Relevant literature on DCEs in diabetic patients and studies on treatment and care preferences were searched using the China National Knowledge Infrastructure (CNKI), Web of Science, and Medline. From this process, a broad list of potential attributes was extracted, including the level of glycemic control, treatment cost, pharmacy service provider, duration of pharmacy service, follow-up schedule, frequency of hypoglycemia, weight change, gastrointestinal response, cardiovascular events, service content, and access to pharmacists.
To ensure representativeness, validity, and actionability [19], face-to-face interviews were conducted with two patients with diabetes, two clinicians with more than 10 years of experience in endocrinology, one pharmacist, and one methodologist.
Subsequently, to refine this list and ensure its representativeness and practicality, we conducted structured face-to-face interviews with an expert panel of six key stakeholders. This panel included two individuals with type 2 diabetes to incorporate the service user perspective. Two clinicians, each with more than ten years of experience in endocrinology, provided a clinical care perspective. One practising community pharmacist contributed insights into the service feasibility and delivery mechanisms. A research methodologist with expertise in health preference studies provided guidance on the design of DCE and the development of the attribute framework.
Based on these discussions, six attributes were prioritized: pharmacy service content, service provider, follow-up program, cardiovascular events, frequency of hypoglycemic reactions (per year), and the cost of services. These attributes were included in the DCE survey draft.
A pre-test involving cognitive interviews was conducted with 20 patients with diabetes receiving medication. The process was designed to assess the clarity, relevance, and comprehensibility of the draft survey. Participants were asked to ‘think aloud’ as they completed the choice tasks and were probed for feedback on specific attributes and descriptions.
Patient feedback directly informed us of several refinements of the final survey instrument. For instance, participants indicated that the term “drug effectiveness assessment” was more concrete and appealing than the more general “medication guidance,” reinforcing its prominence in the design. Additionally, some found the original descriptions for follow-up programs ambiguous; this led to a simplification and clarification of the levels to “individualized follow-up” and “fixed-interval follow-up”. Suggestions regarding the questionnaire layout, such as improving the visual separation between choice sets, were also implemented to reduce the cognitive burden and enhance usability.
Based on this feedback, the final version of the DCE survey was refined to ensure that it was patient centered and easily understood. All six initial attributes were retained as they were deemed highly relevant. Ultimately, the questionnaire retained six attributes categorized into three domains: pharmacy service programs, clinical outcomes, and service costs. Seventeen attribute levels were included, as detailed in Table 1.
Table 1.
Attributes and levels
| Attributes | Levels |
|---|---|
| Pharmacy Services Programme | |
| Pharmaceutical services | Drug effectiveness assessment |
| Pharmacy counselling and medication guidance | |
| Adverse reaction monitoring | |
| Pharmacy service providers | Offline pharmacist |
| Offline pharmacist and online physician | |
| Offline pharmacist and offline physician | |
| Follow-up programme | Individualized follow-up visits based on condition |
| Follow-up at fixed intervals | |
| Clinical complications | |
| Probability of cardiovascular events (over 10 years) | 5% Reduction |
| 10% Reduction | |
| 15% Reduction | |
| Frequency of hypoglycemic reactions (per year) | 0 Times per year |
| 1 Times per year | |
| 2 Times per year | |
| Cost of pharmacy services | |
| Cost of pharmacy services (per visit) | ¥0 |
| ¥20 | |
| ¥40 | |
Experimental design and choice sets
An orthogonal experimental design (OED) was applied to generate DCE choice sets. Each attribute contained two to three levels and a full factorial design resulted in 486 possible combinations (3 × 3 × 3 × 3 × 3 × 3 × 2). This was judged to impose an excessive cognitive burden on the respondents. To reduce the complexity, a mixed orthogonal design was used, yielding 18 virtual pharmacy scenarios (Supplemental Table 1).
Model 9, which contained moderate levels of each attribute, was designated as the reference model (Scheme B) and paired with other scenarios (Scheme A) to form the final choice sets, Supplemental Table 2 is an example selection set. To assess validity, a deliberately designed “dominant choice set” was embedded in the questionnaire, wherein Option A was demonstrably superior to Option B across all key attributes. Questionnaires that failed to select this superior option were deemed invalid, indicating that participants might not have comprehended or responded conscientiously. A clear completeness criterion was established, excluding questionnaires with completion rates of below 70%. In subsequent conditional logit model analyses, the software automatically handled sporadic missing data within individual choice sets, performing parameter estimation solely based on fully responded choice sets.
Study population
Patients were eligible if they were aged 18 years or older, had a confirmed diagnosis of diabetes mellitus, were conscious and able to comprehend the questionnaire, had used glucose-lowering medication within the past two weeks, had long-term attendance at community health centers, and provided written informed consent. Patients were excluded if they were unable to complete the questionnaire independently or had cognitive impairment, agitation, delirium, or severe mental disorders.
The minimum sample size was determined using the formula N > 500c/(a × t), where c is the maximum number of attribute levels, a denotes the number of choice sets that each respondent must complete, and t denotes the number of alternatives within each selection set [20, 21]. In this study, all attribute levels were either two or three; hence, c = 3, a = 17, and t = 2 (Scheme A versus Scheme B). Therefore, the minimum sample size required to meet the basic estimation requirements of the model was approximately 45 participants [22]. The ISPOR guidelines emphasize that when subgroup analyses are planned, the sample size must be calculated separately for each subgroup. To allow subgroup analysis, patients were categorized by glycemic control into three groups (HbA1c < 7.0%, HbA1c between 7.0 and 8.0%, and HbA1c > 8.0%) based on HbA1c targets from the 2020 Chinese Expert Consensus [23]. Glycemic control status was determined using self-reported HbA1c values, which is a pragmatic approach in this community-based survey, although it may be subject to recall bias. A minimum of 45 patients was required in each subgroup, resulting in a minimum sample size of 135. The final effective sample size was 190 cases, comprising 64, 63, and 63 cases across the three subgroups, respectively. The sample size for each subgroup substantially exceeded the statistical requirement of 45 cases, fully meeting the statistical criteria.
Data collection
Data were collected through face-to-face interviews conducted by trained investigators. Before completing the questionnaire, the investigators explained the objectives of the study and provided informed consent. The complete questionnaires were collected immediately after drug administration.
Statistical analysis
The data were entered into Stata version 15.0. A conditional logit (CL) model was constructed to estimate patient preferences for pharmacy service attributes. The dependent variable was the patient’s choice, coded as 1 for the selected option and 0 for the non-selected option. The independent variables were attribute levels coded as 0, 1, or 2, as specified in Supplemental Table 3.
Stepwise regression analyses were performed using SPSS version 23.0, and non-significant predictors were removed to improve model fit. The regression coefficients indicate the relative magnitude of the influence for each attribute level, with higher coefficients reflecting stronger preferences. Wald chi-square values and odds ratios (ORs) were calculated to quantify the attribute importance and level-specific preferences. The greater the Wald chi-square value, the stronger the patient’s preference for that attribute; the larger the OR, the greater the preference for that attribute.
Relative attribute importance (RAI) was calculated by standardizing across subgroups. Specifically, each attribute’s RAI was divided by the largest RAI within the subgroup (the frequency of hypoglycemia in patients with HbA1c > 8%) and multiplied by 10 [24]. This enabled a comparison of the attribute importance across patient groups with different levels of glycemic control.
Ethics approval
The research protocol was reviewed and approved by the Ethics Committee of the Nanjing Drum Tower Hospital (approval number [2020-233-02]). All participants were informed of the study objectives and written informed consent was obtained prior to participation.
Results
Patient characteristics
A total of 237 patients met the inclusion criteria and participated in this survey. After excluding 47 participants who selected the inferior option in the validity check or who did not complete the questionnaire, 190 valid responses were obtained, yielding a response rate of 80.1%.
The sociodemographic and clinical characteristics of the participants are summarized in Table 2. The mean age was 45.6 years (standard deviation [SD] 12.2), with the majority aged 65 years or younger (93.6%). Most participants were married (75.7%) and had a moderate level of education. Regarding occupation, 41.0% were freelancers, 37.8% were company employees, 10.5% were retired, 7.3% were laborers, and 3.1% were civil servants. The annual income was less than RMB 30,000 for 14.7% of the respondents, between RMB 30,000 and 50,000 for 20.0%, between RMB 50,000 and 80,000 for 27.3%, between RMB 80,000 and 100,000 for 24.5%, and greater than RMB 100,000 for 20.0%. Most respondents were covered by medical insurance (92.6%) and 6.9% were self-funded. The duration of diabetes exceeded 10 years in 26.3% of participants. Notably, 63.6% of respondents reported experiencing adverse drug reactions, highlighting the importance of optimizing pharmacy service models to improve satisfaction and outcomes in this population.
Table 2.
Patient characteristics
| Characteristic | Classification | Amount (n, %) |
|---|---|---|
| Gender | Male | 102 (53.6) |
| Female | 88 (46.3) | |
| Age | Average age (SD) | 45.6 (12.2) |
| ≤ 40 years | 90 (47.3) | |
| 41–65 years | 88 (46.3) | |
| ≥ 66 years | 12 (6.3) | |
| Education background | Junior high school and below | 42 (22.1) |
| High school | 60 (31.5) | |
| Junior college | 44 (23.1) | |
| Undergraduate and above | 44 (23.1) | |
| Occupation | Employee | 72 (37.8) |
| Worker | 14 (7.3) | |
| Retired | 20 (10.5) | |
| Freelancer | 78 (41.0) | |
| Civil servant | 6 (3.1) | |
| Annual income (¥10,000/year) | < 3 | 28 (14.7) |
| 3–5 | 38 (20.0) | |
| 5–8 | 52 (27.3) | |
| 8–10 | 39 (24.5) | |
| > 10 | 38 (20.0) | |
| Payment methods | Medical insurance | 176 (92.6) |
| Self-pay | 14 (7.3) | |
| Diabetes duration (year) | ≤ 1 | 64 (33.6) |
| 2–5 | 46 (24.2) | |
| 6–10 | 30 (15.7) | |
| > 10 | 50 (26.3) | |
| Ever had an adverse drug reaction | Yes | 121 (63.7) |
| Not | 69 (36.3) | |
| Marital status | Married | 144 (75.7) |
| Other | 46 (24.2) |
SD, standard deviation; ¥, Chinese Yuan. Data are presented as number (percentage) unless otherwise specified
Overall preferences
Using drug effectiveness assessment, one hypoglycemic event reduction per month, no service cost, a 5% reduction in cardiovascular events, offline pharmacist and online physician service delivery, and fixed-interval follow-up as reference levels, the conditional logit regression model indicated that patients were most inclined to select drug effectiveness assessment, minimization of hypoglycemic events and cardiovascular events, affordable service costs, and an individualized follow-up program (Fig. 2).
Fig. 2.
CL regression results
The RAI analysis (Fig. 3) demonstrated that, in descending order, the most influential factors were pharmacy service content (Wald = 389.188, RAI = 4.59), frequency of hypoglycemic events (Wald = 284.155, RAI = 4.19), service cost (Wald = 168.079, RAI = 4.07), reduction in cardiovascular events (Wald = 144.608, RAI = 3.38), service providers (Wald = 142.913, RAI = 3.29), and follow-up programs (Wald = 50.644, RAI = 1.15).
Fig. 3.
Relative attribute importance. Note: RAI, relative attribute importance. For each attribute, RAI was calculated as the difference between its maximum and minimum preference coefficients. These values were then standardized within each subgroup by dividing by the maximum RAI in that subgroup (here, the RAI for hypoglycaemia frequency in the HbA1c > 8.0% subgroup) and multiplying by 10. The standardised values represent the relative importance of each attribute within its subgroup; a higher value indicates greater importance [25]
Subgroup analyses
Subgroup analyses stratified by HbA1c levels are presented in Supplemental Table 4. Across all groups, patients consistently prioritized drug effectiveness assessment, which was the dominant determinant of service preference. Similarly, all groups demonstrated a strong preference for larger reductions in cardiovascular events and hypoglycemic episodes, and all favored the combination of offline pharmacists and physicians as the preferred service delivery model.
Notable differences emerged between the subgroups (Table 3), particularly when comparing patients with HbA1c ≤ 8% (combining the < 7.0% and 7.0–8.0% subgroups) and those with HbA1c > 8%. Patients with HbA1c levels ≤ 8% were more sensitive to service costs, although they were still willing to pay for pharmacy services. Overall, while all groups recognized the value of pharmacy services, the degree of emphasis on cost and follow-up design varied according to glycemic control status.
Table 3.
Subgroup analysis
| Attributes (reference level) | HbA1c < 7.0% (N = 64) | HbA1c 7.0–8.0% (N = 63) | HbA1c > 8.0% (N = 63) | ||||||
|---|---|---|---|---|---|---|---|---|---|
| β (SE) | 95% Cl | P | β (SE) | 95% Cl | P | β (SE) | 95% Cl | P | |
| Pharmacy Services (drug effectiveness assessment) | |||||||||
| Pharmacy counselling and medication guidance | − 2.035 (0.148) | [− 2.438, − 1.614] | < 0.0001 | − 2.045 (0.208) | [− 2.453, − 1.637] | < 0.0001 | − 4.784 (0.356) | [− 5.482, − 4.086] | < 0.0001 |
| Adverse reaction monitoring | − 1.526 (0.192) | [− 1.902, − 1.150] | < 0.0001 | − 1.053 (0.180) | [− 1.406, − 0.700] | < 0.0001 | − 2.510 (0.301) | [− 3.100, − 1.920] | < 0.0001 |
| Follow-up program (Individualized follow-up visits based on condition) | |||||||||
| Follow-up at fixed intervals | 1.200 (0.175) | [0.857, 1.543] | < 0.0001 | 0.540 (0.154) | [0.238, 0.842] | < 0.0001 | − 0.494 (0.187) | [− 0.860, − 0.128] | 0.009 |
| Pharmacy service providers (Offline pharmacist and online physician) | |||||||||
| Offline pharmacist and offline physician | 1.232 (0.181) | [0.877, 1.587] | < 0.0001 | 0.763 (0.166) | [0.438, 1.088] | < 0.0001 | 1.119 (0.232) | [0.664, 1.574] | < 0.0001 |
| Offline pharmacist | − 1.068 (0.188) | [− 1.436, − 0.700] | < 0.0001 | − 0.875 (0.175) | [− 1.218, − 0.532] | < 0.0001 | − 0.371 (0.276) | [− 0.912, 0.170] | 0.182 |
| Probability of cardiovascular events (5% reduction) | |||||||||
| 10% reduction | 2.156 (0.212) | [1.741, 2.571] | < 0.0001 | 1.735 (0.198) | [1.347, 2.123] | < 0.0001 | 1.300 (0.262) | [0.786, 1.814] | < 0.0001 |
| 15% reduction | 2.950 (0.245) | [2.470, 3.430] | < 0.0001 | 2.150 (0.218) | [1.723, 2.577] | < 0.0001 | 1.512 (0.292) | [0.940, 2.084] | 0.001 |
| Frequency of hypoglycemic reactions (2 times per year) | |||||||||
| 1 times per year | 1.609 (0.205) | [1.207, 2.011] | < 0.0001 | 1.501 (0.201) | [1.107, 1.895] | 0.001 | 3.843 (0.338) | [3.180, 4.506] | < 0.0001 |
| 0 times per year | 2.216 (0.230) | [1.765, 2.667] | < 0.0001 | 1.765 (0.212) | [1.349, 2.181] | < 0.0001 | 5.059 (0.424) | [4.228, 5.890] | < 0.0001 |
| Cost of pharmacy services (¥0) | |||||||||
| ¥20 | 2.429 (0.225) | [1.988, 2.870] | < 0.0001 | 1.574 (0.200) | [1.182, 1.966] | < 0.0001 | 0.886 (0.284) | [0.329, 1.443] | 0.002 |
| ¥40 | − 0.618 (0.188) | [− 0.986, − 0.250] | 0.001 | − 0.663 (0.177) | [− 1.010, − 0.316] | < 0.0001 | 1.137 (0.385) | [0.382, 1.892] | 0.003 |
β, regression coefficient, representing preference strength (positive values indicate preference for that level, negative values indicate lack of preference). SE, standard error. CI, confidence interval, calculated as β ± 1.96 × SE. P-value based on Wald test. N denotes the effective sample size for each subgroup
This study employed a conditional logit (CL) model to examine patient choice behavior. Multiple metrics were calculated to evaluate the goodness of fit of the model. The final model yielded a log-likelihood of − 624.73. Compared with a null model containing only a constant term, the likelihood ratio test was statistically significant (χ2 = 168.08, p < 0.001), confirming that the six included attributes jointly contributed to explaining patient preferences. The McFadden’s pseudo-R2 value was 0.130. While values between 0.2 and 0.4 are typically considered indicative of a good fit in discrete choice models, values between 0.1 and 0.2 are generally acceptable, especially in behavioral studies where unobserved heterogeneity is common. These metrics indicate that the model provides a statistically significant and substantively meaningful representation of the preferences of diabetic patients for community pharmacy service attributes.
Nonlinear effects test
Quadratic terms were added to the model to examine the potential nonlinearity in the effects of continuous attributes (cardiovascular event reduction and hypoglycemia frequency). The results indicated that the nonlinear effects were non-significant (cardiovascular event reduction: β = − 0.032, P = 0.421; hypoglycemia frequency: β = 0.021, P = 0.587) and did not substantially improve the model fit (likelihood ratio test χ2 = 2.17, P = 0.338), supporting the validity of the linear specification. Notably, the “pharmaceutical service expenditure” attribute exhibited non-monotonic variation: compared to ¥0 per visit, the preference coefficient for ¥20 per visit was positive (β = 1.479, P < 0.001), whereas the coefficient for ¥40 per visit was negative (β = − 0.583, P = 0.005). After recoding this variable as categorical, model fit improved marginally, as indicated by a lower Akaike Information Criterion (AIC), suggesting the presence of a threshold effect or diminishing marginal utility for this attribute.
Discussion
This study employed DCE to quantitatively examine the preferences of patients with diabetes in community pharmacy service models. Six attributes were evaluated, revealing the inherent trade-offs in patient decision making. Consistent with existing evidence that HbA1c levels influence treatment choices [26], patients were stratified into three subgroups based on glycemic control targets outlined in the 2020 Chinese Expert Consensus [27]. These findings provide empirical support for the design of patient-centered pharmacy services.
Key findings and comparison with previous research
The most important determinant of preference was pharmacy service content (RAI = 4.59), with “drug effectiveness assessment” strongly preferred over counselling or adverse reaction monitoring. This indicates a shift from traditional dispensing to outcome-driven professional service. Hypoglycemia frequency and service cost ranked second and third (RAI = 4.19 and 4.07), reflecting patient concern about immediate safety and rational willingness to pay for services perceived as valuable. Notably, while a fee of approximately ¥20 per visit was generally acceptable, the preference for the ¥40 level varied across the glycemic control subgroups. Cardiovascular event reduction was valued less (RAI = 3.38), suggesting a preference for short-term tangible benefits over long-term risk reduction.
These results align with international DCE studies highlighting hypoglycemia risk as a key factor in diabetes care. This focus on hypoglycemia aligns with DCE studies on diabetes medications, which consistently identified hypoglycemia risk as a critical factor for patients [28, 29]. In DCE studies of diabetes services across Indonesia, Thailand, and other nations, reducing hypoglycemic episodes was ranked by patients as one of the most critical attributes. However, while prior research has focused on medication selection [26, 30], this study extends this evidence to service models, confirming hypoglycemia prevention as a central preference driver.
This corroborates the regional universality of our study finding that hypoglycemia prevention constitutes a core service element. This finding is consistent with DCE studies conducted in middle-income countries such as Indonesia and Malaysia, where patients often demonstrate high sensitivity to out-of-pocket costs [13, 31, 32]. In contrast, studies in some high-income settings suggest a relatively lower marginal impact of costs within the ranges studied, possibly due to different insurance structures [12, 33].
Patients favored physician-pharmacist collaborative care over pharmacist-only services. This preference is consistent with findings from primary care settings in the UK, where patients valued an expanded pharmacist role within a collaborative framework [15]. This further supports broader evidence that such collaboration can enhance trust and service uptake [25, 34–37]. The follow-up plan was the least influential attribute (RAI = 1.15), possibly reflecting the limited patient awareness of the importance of structured follow-up.
Implications for practice
Research indicates that patients accepted a fee of around ¥20 per visit for valued services, while ¥40 was less preferred in the total sample. However, this pattern differed by glycemic control: patients with HbA1c > 8% did not show the same resistance to ¥40, suggesting varied cost sensitivity. This provides direct evidence for formulating tiered pricing policies: basic services (low-cost/free) will incorporate fundamental services such as medication counselling and routine follow-ups into public health subsidies or basic medical insurance coverage to enhance service accessibility. Advanced services (value-based pricing) for highly valued specialist services such as “drug effectiveness assessment,” “personalised follow-up programmes” and “hypoglycaemia risk management”, a fee structure of approximately ¥20–¥30 per visit may be established. This is supported by partial reimbursement by medical insurance or dedicated chronic disease management funds.
Patients strongly preferred collaborative services involving both in-person pharmacists and in-person physicians, indicating that policy must address responsibility allocation and remuneration distribution within such collaborations: the medical insurance reimbursement catalogue should introduce a dedicated category for “pharmaceutical services for chronic conditions”, explicitly permitting separate billing for services such as assessments, follow-ups, and education provided by pharmacists in collaboration with physicians. Establish a certification system and service standards for pharmacists’ chronic disease management services, ensuring that their professional capabilities align with fee-based services to enhance trust in pharmacy services among patients and the medical insurance system.
The subgroup analysis revealed distinct preferences regarding costs and follow-up. Patients with HbA1c ≤ 8% demonstrated greater cost sensitivity and preferred fixed-interval follow-up. In contrast, patients with HbA1c > 8% placed a higher value on individualized follow-up plans and, as indicated by their positive coefficient for the ¥40 level, were relatively less sensitive to higher service costs, showing a greater willingness to pay for personalized care. Consequently, more intensive, individualized pharmaceutical follow-up and assessment services could be offered to poorly controlled patients, with appropriate reimbursement rates. For well-controlled patients, low-cost or free periodic assessments and education should predominate, with a focus on maintenance and prevention.
Strengths and limitations
This study has several strengths. First, attributes and levels were rigorously identified through literature review and expert interviews to ensure content validity. The use of an orthogonal design reduced the respondent burden while preserving statistical efficiency. Second, this is the first systematic DCE conducted in China to explore patient preferences in community pharmacy service models, thereby filling an important research gap. Third, subgroup analyses by glycemic control revealed heterogeneity in preferences, providing nuanced insights into tailoring services.
This study has several limitations. First, as with all stated preference studies, choices made in hypothetical scenarios may not accurately predict real-world utilization behavior, which may affect the external validity of the findings, particularly regarding willingness-to-pay estimates. Second, the sample was limited to patients from three urban districts in Nanjing, China, potentially introducing a selection bias and restricting the generalizability of the results to rural or other regional populations. The inclusion of participants from diverse geographical and socioeconomic backgrounds in future studies is recommended to assess the transferability of the identified preference patterns. Third, glycemic control status was categorized based on self-reported HbA1c values rather than on clinically verified laboratory records. This method may introduce measurement errors or recall bias, which could affect the accuracy of subgroup comparisons and the robustness of conclusions related to glycemic control. Objective biochemical confirmation in future studies would improve the validity of such stratification. Fourth, although subgroup analyses indicated that cost sensitivity correlates with glycemic control levels, the potential influence of socioeconomic confounding factors such as income, education, and type of health insurance coverage were not formally examined.
Conclusion
Patients with diabetes prefer community pharmacy services that emphasize drug effectiveness assessment, reduction of hypoglycemic risk, affordable costs, and physician–pharmacist collaboration. Preferences varied significantly with glycemic control. Patients with HbA1c ≤ 8% were more cost-sensitive and preferred fixed-interval follow-up. Conversely, patients with HbA1c > 8% placed higher values on individualized follow-up and demonstrated greater willingness to pay for services. These insights highlight the need for tailored patient-centered service models for diabetes management.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
None.
Author contributions
Conceptualization, Zizhen Jia, Yu Sun and Linyan Lan; Data curation, Zizhen Jia and Yu Sun; Funding acquisition, Han Xie and Cheng Ji; Methodology, Linyan Lan and Zizhen Jia; Project administration, Linyan Lan and Cheng Ji; Supervision, Xiangxiang Xie and Han Xie; Validation, Yu Sun; Writing—original draft, Zizhen Jia; Writing—review and editing, Zizhen Jia and Cheng Ji.
Funding
This study was supported by the National Natural Science Foundation of China (grant number: 72104105) and the National Research Center for the Development of Licensed Pharmacists, China Pharmaceutical University (grant number: NRCLPD202505).
Data availability
The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.
Declarations
Competing interests
The authors declare no competing interests.
Ethics approval
The research protocol was reviewed and approved by the Ethics Committee of the Nanjing Drum Tower Hospital (approval number [2020-233-02]). All participants were informed of the study objectives and written informed consent was obtained prior to participation.
Footnotes
Yu Sun, Zizhen Jia and Linyan Lan have contributed equally to this work.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Han Xie and Cheng Ji are co-corresponding authors and contributed equally to this work. Correspondence may also be addressed to: Cheng Ji, Email: getcct@sina.com.
References
- 1.Hedima EW, Okoro RN. Primary health care roles of community pharmacists in low- and middle-income countries: a mixed methods systematic review. BMC Health Serv Res. 2025;25(1):1269. 10.1186/s12913-025-13387-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Yang Y, Li C, Qin W, et al. Current status and development recommendations for community pharmacy services in China. Chin J Pharm Res. 2023;42(12):1042–6. 10.13506/j.cnki.jpr.2023.12.016. [Google Scholar]
- 3.Fan J, Shen Y, Cheng J, et al. Survey on the current status of pharmaceutical services in community health service centers in Nanjing’s main urban area. Pharm Guide. 2023;42(11):1697–703. 10.3870/j.issn.1004-0781.2023.11.020. [Google Scholar]
- 4.Jiang B, Chen X, Qi J, et al. Investigation and analysis of the demand of pharmacy service for community residents. Chin Pharm Aff. 2017;31(4):430–4. 10.16153/j.1002-7777.2017.04.014. [Google Scholar]
- 5.Soekhai V, de Bekker-Grob EW, Ellis AR, et al. Discrete choice experiments in health economics: past, present and future. Pharmacoeconomics. 2019;37(2):201–26. 10.1007/s40273-018-0734-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Guangdong Pharmaceutical Association. Community pharmacy care standards (first edition). Todays Pharm. 2022;32(12):881–7. 10.12048/j.issn.1674-229X.2022.12.001. [Google Scholar]
- 7.Fusiak J, Sarpari K, Ma I, et al. Practical applications of methods to incorporate patient preferences into medical decision models: a scoping review. BMC Med Inform Decis Mak. 2025;25(1):109. 10.1186/s12911-025-02945-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.National Health Commission, Ministry of Industry and Information Technology, National Healthcare Security Administration, et al. Opinions on reforming and improving the coordinated management mechanism for medicines at the primary healthcare level and expanding the range of medicines available at the primary healthcare level. Chin Rural Health. 2024;16(12):4. 10.3969/j.issn.1674-361X.2024.12.003. [Google Scholar]
- 9.Yan Q, Yang R, Huang F, et al. Development and application of a home-based pharmaceutical service model for community chronic disease patients based on family doctor contract services. Mod Med Health. 2024;40(2):339–41. 10.3969/j.issn.1009-5519.2024.02.035. [Google Scholar]
- 10.Ryan M, Farrar S. Using conjoint analysis to elicit preferences for health care. BMJ. 2000;320(7248):1530–3. 10.1136/bmj.320.7248.1530. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Bridges JF, Hauber AB, Marshall D, et al. Conjoint analysis applications in health—a checklist: a report of the ISPOR good research practices for conjoint analysis task force. Value Health. 2011;14(4):403–13. 10.1016/j.jval.2010.11.013. [DOI] [PubMed] [Google Scholar]
- 12.de Bekker-Grob EW, Ryan M, Gerard K. Discrete choice experiments in health economics: a review of the literature. Health Econ. 2012;21(2):145–72. 10.1002/hec.1697. [DOI] [PubMed] [Google Scholar]
- 13.Presley B, Groot W, Widjanarko D, et al. Preferences for pharmacist services to enhance medication management among people with diabetes in Indonesia: a discrete choice experiment. Patient Educ Couns. 2021;104(7):1745–55. 10.1016/j.pec.2020.12.005. [DOI] [PubMed] [Google Scholar]
- 14.Raghunandan R, Howard K, Marra CA, et al. Identifying New Zealand public preferences for pharmacist prescribers in primary care: a discrete choice experiment. Patient. 2022;15(1):77–92. 10.1007/s40271-021-00529-9. [DOI] [PubMed] [Google Scholar]
- 15.Tinelli M, Ryan M, Bond C. Patients’ preferences for an increased pharmacist role in the management of drug therapy. Int J Pharm Pract. 2009;17(5):275–82. 10.1211/ijpp.17.05.0003. [PubMed] [Google Scholar]
- 16.Kawaguchi T, Azuma K, Yamaguchi T, et al. Preferences for pharmacist counselling in patients with breast cancer: a discrete choice experiment. Biol Pharm Bull. 2014;37(11):1795–802. 10.1248/bpb.b14-00452. [DOI] [PubMed] [Google Scholar]
- 17.Hertroijs DFL, Elissen AMJ, Brouwers MCGJ, et al. Preferences of people with Type 2 diabetes for diabetes care: a discrete choice experiment. Diabet Med. 2020;37(11):1807–15. 10.1111/dme.13969. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Mc Morrow L, O’Hara MC, Hynes L, et al. The preferences of young adults with Type 1 diabetes at clinics using a discrete choice experiment approach: the D1 Now Study. Diabet Med. 2018;35(12):1686–92. 10.1111/dme.13809. [DOI] [PubMed] [Google Scholar]
- 19.Reed Johnson F, Lancsar E, Marshall D, et al. Constructing experimental designs for discrete-choice experiments: report of the ISPOR conjoint analysis experimental design good research practices task force. Value Health. 2013;16(1):3–13. 10.1016/j.jval.2012.08.2223. [DOI] [PubMed] [Google Scholar]
- 20.Ijzerman MJ, van Til JA, Snoek GJ. Comparison of two multi-criteria decision techniques for eliciting treatment preferences in people with neurological disorders. Patient. 2008;1(4):265–72. 10.2165/1312067-200801040-00008. [DOI] [PubMed] [Google Scholar]
- 21.Rademaker MM, Essers BAB, Stokroos RJ, et al. What tinnitus therapy outcome measures are important for patients?—a discrete choice experiment. Front Neurol. 2021;12:668880. 10.3389/fneur.2021.668880. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Sukpanich R, Sanglestsawai S, Seib CD, et al. The influence of cosmetic concerns on patient preferences for approaches to thyroid lobectomy: a discrete choice experiment. Thyroid. 2020;30(9):1306–13. 10.1089/thy.2019.0821. [DOI] [PubMed] [Google Scholar]
- 23.Pan F. Glycated haemoglobin has important value in diabetes management—Interview with Professor ZHU Dalong, Chairman of the Diabetes Branch of the Chinese Medical Association and Professor of Nanjing Gulou Hospital. Chin Med J. 2020;17(26):1–3. 10.20047/j.issn1673-7210.2020.26.001. [Google Scholar]
- 24.Belvederi Murri M, Mamberto S, Briatore L, et al. The interplay between diabetes, depression and affective temperaments: a structural equation model. J Affect Disord. 2017;219:64–71. 10.1016/j.jad.2017.05.018. [DOI] [PubMed] [Google Scholar]
- 25.Cook CA, Vakayil V, Pribyl K, et al. A pharmacist-driven glycemic control protocol to reduce the rate of severe hypoglycemia in high-risk patients. Hosp Pharm. 2022;57(1):45–51. 10.1177/0018578720973891. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Mansfield C, Sikirica MV, Pugh A, et al. Patient preferences for attributes of type 2 diabetes mellitus medications in Germany and Spain: an online discrete-choice experiment survey. Diabetes Ther. 2017;8(6):1365–78. 10.1007/s13300-017-0326-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Chinese Diabetes Society, Chinese Society of Endocrinology. Expert consensus on glycated haemoglobin control targets and strategies for achieving targets in Chinese adults with Type 2 diabetes mellitus. Chin J Diabetes. 2020;12(1):1–12. 10.3760/cma.j.issn.1674-5809.2020.01.001. [Google Scholar]
- 28.Hauber AB, Nguyen H, Posner J, et al. A discrete-choice experiment to quantify patient preferences for frequency of glucagon-like peptide-1 receptor agonist injections in the treatment of type 2 diabetes. Curr Med Res Opin. 2016;32(2):251–62. 10.1185/03007995.2015.1117433. [DOI] [PubMed] [Google Scholar]
- 29.Purnell TS, Joy S, Little E, et al. Patient preferences for noninsulin diabetes medications: a systematic review. Diabetes Care. 2014;37(7):2055–62. 10.2337/dc13-2527. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Senanayake S, Barnett A, Brain D, et al. A discrete choice experiment to elicit preferences for a chronic disease screening programme in Queensland, Australia. Public Health. 2024;228:105–11. 10.1016/j.puhe.2024.01.007. [DOI] [PubMed] [Google Scholar]
- 31.Wei L, Zhang L, Tan Y, et al. Preference research on community health services for patients with chronic diseases comorbidity. Chin General Pract. 2024;27(04):454–61. [Google Scholar]
- 32.Liu Y, Kong Q, Yuan S, et al. Factors influencing choice of health system access level in China: a systematic review. PLoS ONE. 2018;13(8):e0201887. 10.1371/journal.pone.0201887. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Guerriere DN, Choinière DJ, Dion D, et al. Wait times at a multidisciplinary pain clinic: a discrete-choice experiment of patient preferences. J Pain Res. 2010;3:253–62. 10.2147/JPR.S14875. [Google Scholar]
- 34.Tsimtsiou Z, Kirana PS, Hatzichristou D. Determinants of patients’ attitudes toward patient-centered care: a cross-sectional study in Greece. Patient Educ Couns. 2014;97(3):391–5. 10.1016/j.pec.2014.08.007. [DOI] [PubMed] [Google Scholar]
- 35.Brock KA, Doucette WR. Collaborative working relationships between pharmacists and physicians: an exploratory study. J Am Pharm Assoc. 2004;44(3):358–65. 10.1331/154434504323063995. [DOI] [PubMed] [Google Scholar]
- 36.Jokanovic N, Tan EC, Sudhakaran S, et al. Pharmacist-led medication review in community settings: an overview of systematic reviews. Res Social Adm Pharm. 2017;13(4):661–85. 10.1016/j.sapharm.2016.08.005. [DOI] [PubMed] [Google Scholar]
- 37.Rowen D, Brazier J, Van Hout B. A comparison of methods for converting DCE values onto the full health-dead QALY scale. Med Decis Making. 2015;35(3):328–40. 10.1177/0272989X14559542. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.