Patient experience in primary care : association with patient , physician and practice characteristics in a fee-for-service system

METHODS: The data came from the Swiss part of the Quality and Costs of Primary Care (QUALICOPC) study, an international cross-sectional survey. A random sample of 194 Swiss family physicians and 1540 of their patients were included in this analysis. We assessed patient experience using three scores characterising access, communication and continuity-coordination. Multilevel statistical methods were used to analyse these scores based on patient-level, physician-level and practice-level factors.


Introduction
Quality of care is a concept classically used to evaluate and compare primary healthcare systems.Besides clinical and process outcomes or global patient satisfaction tools, patient-reported experience measures appear nowadays to be an essential tool for assessing quality of care [1][2][3][4][5][6][7][8][9][10].Numerous studies have already investigated factors that may be the most predictive for a patient's experience of care.Factors most often reported relate to the patients themselves (ethnic minorities, younger age, multiple chronic conditions) [11][12][13][14][15][16].For physician or practice characteristics, small practices, a good team climate and long weekly working hours are associated with a better patient experience [12,[17][18][19][20][21].However, the findings are somewhat inconsistent or even conflicting and the authors underlined the need for further studies to better investigate the role of practices and contexts [8,17].Moreover, previous studies usually focused on one dimension of patient experience [12,13,22], and studies combining several dimensions in the same context are sparse.They also often focused on patient characteristics [11,14,15,23] or on physicians or practices [17,18,20]; few were able to explore both simultaneously [8,12,13].Finally, most of the studies were conducted in countries with pay-for-performance systems, such as the United Kingdom, with limited choice of provider and where patient experience surveys are used to work out part of the practice's income [8].In a fee-forservice system like Switzerland, without quality incentives and with a free choice of the practitioner, the impact of the element of satisfaction might be different [24].The aim of this study was to investigate to what extent patient, physician and practice characteristics are associated with patient experience for major dimensions in family medicine including access, interpersonal communication, continuity and coordination, in the context of a liberal payfor-service system.

Study population and setting
Data came from the Swiss participation in the Quality and Costs of Primary Care (QUALICOPC) study, a cross-sectional European survey coordinated by the Nivel Institute from The Netherlands [25].This project aimed to analyse and compare primary healthcare systems across Europe.Surveys were carried out in 31 European countries (European Union 27 -except France, FYR Macedonia, Iceland, Norway, Switzerland and Turkey) and 3 non-European countries (Australia, Canada and New Zealand).In each country, a random nationally representative sample of around 200 family physicians was drawn.Only one family physician per practice was eligible to participate.In Switzerland, the participating physicians stemmed from a random sample of family physicians stratified by canton, via their participation in a practice-based research network.The representativeness in terms of gender, location and age was cross-checked against national statistics and considered satisfactory [26].In each practice, nine patients (randomly drawn) filled out a patient experience questionnaire about the consultation that had just taken place.The resulting sample of patients consisted of 1800 persons.

Ethics approval and consent to participate
Ethical approval (Reference CER-VD 410/11) for the QUALICOPC study was acquired in accordance with the legal requirements of each country.Details about the study protocol and questionnaire development have been published elsewhere [25,27].The Swiss data collection took place between January and June 2012 in Switzerland and was coordinated by the University of Lausanne.Our study was approved by the Swiss ethical review board, "Commission cantonale Vaud d'éthique et de recherche sur l'être humain" (Reference CER-VD 410/11).In accordance with the Ethics Committee and to the extent that no biomedical data were collected, the physicians and patients provided their informed oral consent for their participation.In accordance with the contract (INT040-NC24) between the Netherlands Institute for Health Services Research (Nivel) and the national coordinators, the national coordinators were entitled to use their own national data.

Data collection and questionnaires
Family physicians completed a self-administrated questionnaire sent by post.Questions related to personal, organisational and practice characteristics.Sociodemographic features of family physicians were described in terms of sex, age and rural/urban practice areas.Questions regarding organisational and practice characteristics concerned general features including solo/group practice, family physician in practice as a unique activity, primary care access including consultation length, weekly workload, care collaboration, including workforce in the practice [28].Exposure to work-related stress among physicians was explored by a proxy of Siegrist's effort/reward imbalance model [29].The family physician model claims that to the need for a high level of effort at work (both quantitative and qualitative demands) and inadequate rewards in return (in terms of money, esteem from colleagues or society, job security) can generate stress at work.Such repetitive exposures could have an impact both on a physician's well-being and performance [30].The patient questionnaire was administrated by field workers and explored different aspects of their visit relating to access, interpersonal communication, continuity and coordination, comprehensiveness, trust and patient activation.
For the present study we focused on the first four issues, which were the most detailed in the questionnaire and considered as major dimensions in primary care evaluation.We investigated experience of access with 10 items, communication with 15 items, continuity and coordination with 10 items (see appendix 1).Sociodemographic characteristics of the patients were sex, age, place of birth, language area (Switzerland includes three main linguistic areas: German, French, Italian), income, level of education and employment status.Moreover, we measured global health with two items: perceived health (four levels from bad to very good) and presence of longstanding illness (yes/no).We also collected information about the number of the visits to a family physician in the last 6 months.Translators translated the initial English master version questionnaire into the three national Swiss languages -German, French and Italian (no cross-validation).

Analysis
For the present analysis, we dropped patients without appointments, considering that their experience might be significantly different (142 patients).A unique practice identification number linked the family physician responses to the responses of his or her patients, allowing for multilevel analyses of the data in order to take into account the nested nature of the observations (nine patients in each practice).All the items exploring experience were coded 0 for positive experience and 1 for negative experience.For each dimension a score of "bad experience" was calculated by adding the responses for each item of the dimension.For each dimension, the item-test correlations, as well as the average inter-item correlation, were very close, meaning that the contribution of each factor to the global score was equivalent.The internal consistency of each score was poor (Cronbach's alpha >0.35 and >0.40) reflecting more a global score for each domain, including several sub-dimensions, rather than a homogeneous variable.Finally, the correlation between dimensions was very low, meaning good independence of the three scores.Then we carried out analyses in two steps.In the first step, associations between each dimension score of patient experience (dependent variable) and sociodemographic and personal patient characteristics were considered one at a time in multilevel Poisson regression models (this was appropriate considering the low numbers of items reporting bad experience).For reasons of convenience, most of the dependent categorical variables related to practice char-acteristics were dichotomized.The variables associated at a p-value of 0.2 or less were then included in a multiple model.Finally, we performed a manual backward stepwise selection (removal of the least significant variable at each step) to obtain a final model for patient characteristics (named "final model patients" in the tables).In the second step, we studied separately the association between each dimension (dependent variable) and each family physician characteristic, using the final model selected for patient characteristics.The variables associated with the dimension at a p-value of 0.2 or less were then included in the joint multiple multilevel Poisson regression model.Finally, we performed a manual backward stepwise selection with a p-value for selection set at 5% to obtain a final model including both patient and family physician characteristics (named "global final model" in the tables).Note that the results are expressed as incidence rate ratios (IRRs): for example, an IRR of 1.2 for a given independent variable would mean the score is 20% higher for this variable.Had we considered the number of positive items as dependent variables, no such interpretation would have been possible.Analyses were performed using STATA software.

Results
After exclusion of patients with missing data for any of the studied variables (119 patients), the dataset contained 1540 patients and 199 practices.In the patient sample, 57% were female; the median age was 59 years and 74% of the patients were born in Switzerland.The most frequent reason for the family physician visit was a scheduled appointment (medical check-up, renewal of prescription and other reasons) and not for acute symptoms.We interviewed patients in 199 different practices.Physicians were mainly men (78%), with a median age of 56 years.They worked in group practices in 52% of the cases and in urban areas in 48% (table 1).The mean score of poor experience regarding access was 0.92/9.When two-way associations were considered, poor experience of access was associated with every sociodemographic feature (except language skills) and personal patient characteristics.In the final multiple model including only patient characteristics, poor experience of access decreased with patient age and level of education, but increased with poor perceived health.Practices in rural areas had lower scores of poor access compared with urban ones (IRR 0.85, 95% confidence interval [CI] 0.73-1.00).The presence of other paramedical disciplines in the practice was also associated with poorer access (IRR 1.27, 95% CI 1.06-1.52).No other physician or practice attribute was associated with patient reported experience of access (table 2).The mean score of poor experience regarding communication was 1.21/15.In univariate analyses, poor experience was associated with patient sociodemographic features but not with personal medical ones.In multiple analysis, patient gender was associated with poor experience of communication with an IRR of 0.91 (95% CI 0.83-1.01)for women.Poor experience of communication was also lower among French and Italian speaking patients, IRRs of 0.88 (95% CI 0.78-0.99)and 0.54 (95% CI 0.45-0.67),respectively.The association observed with a high level of education persisted in the final model, including both patient and physician features with IRRs of 0.84 (95% CI 0.75--0.94)with upper secondary and 0.84 (95% CI 0.72-0.97)with post-secondary levels.Communication was better among patients with good language skills, and those who perceived their health as fair compared with very good (IRR 0.84, 95% CI 0.72-0.99).Moreover, poor experience of communication increased with the daily number of face-toface consultations (IRR 1.16, 95% CI 1.08-1.25)and decreased with the physician's weekly workload (IRR 0.87, 95% CI 0.81-0.93)(table 3).The mean score of poor experience for continuity and coordination of care was 1.63/10 (table 1).Younger patients

Discussion
Experience in family medical care regarding access, communication and continuity coordination of care in Switzerland seems to be very good.Most of the factors we found associated with poor experiences were related to patient characteristics such as age, linguistic area and health status.However, several factors relating to the physicians and the organisation of the practices were also predictive of experience, depending on the dimension.Among patients attributes age was an important factor, with younger age associated with poorer experience for two out of three dimensions, namely access and continuity coordination (in the adjusted final models).These results are in line with previous findings [11][12][13][14] suggesting that younger patients may be more demanding than older ones [31].The results also show that a patient's experience is associated with health status, in particular for access and continuity coordination (less clear association for communication), but in a different ways according to the dimension.First, poor experience of access was higher among patients with poor perceived health.This might reflect the difficulties for patients with poor health to reach the practices because of physical limitations, for instance.However, based on more objective indicators such as the presence of chronic disease and the number of visits in family medicine, continuity coordination was reported as higher among patients with poor health.This is probably because patients with chronic disease may benefit more from coordinated care and better care management.The discrepancy between the indicators illustrates the importance of not considering patient-reported experience of care as a unique concept, but rather to explore its different components.Interestingly, the analysis showed differences in experience of communication and continuity coordination between different linguistic areas.However, Switzerland has a unique global primary care organisation functioning across the country.The physicians and practice features we included in the analyses do not explain these variations.An explanation might be that the level of expectation regarding care depends on the cultural background, which is different according to the particular linguistic area.This is probably a key element to consider systematically in the field of patient-reported outcome measures, as other authors have already suggested.Regarding this issue, qualitative research that is more relevant than quantitative studies should be developed [22].Physician and practice attributes were less predictive of poor experience of care, but were also more difficult to investigate in view of the study design.However, several as-sociations are observed.Patients reported better access in rural areas than in urban ones.In a more detailed analysis (not shown), we observed that for the access score, the items best correlated with the rural or urban areas were difficulty to get an appointment and considering the practice too far from the house.In this latter domain, patients' perceptions of what is acceptable in terms of access might differ according to their rural/urban origin.The presence in the practice of other paramedical professionals (except a medical secretary) was also associated with a poorer experience of access.This result suggests that access is better perceived in small practices -not necessarily in solo practices (no difference observed), but in practices with a small workforce.These results are in line with previous studies, despite differences in the definition of "small practices" [12,18,20].Communication was better when the number of daily face-to-face consultations was lower and when the global time (hours per week) spent with patients was higher.These two results imply that long consultations are associated with a better experience of communication.Some practice characteristics were also found to be associated with a better experience of continuity coordination than others: for example, working with other fam- The concept of imbalance between effort at work and rewards in return, in terms of money, esteem from colleagues or society and job security, is well known.A more detailed analysis showed that the continuity-coordination dimension was preferentially associated with items relating to the dissemination of information between family physicians or between family physicians and specialists.We assumed that family physicians having a feeling of low esteem from colleagues might be less willing to share and exchange information.Few studies, with inconsistent results, have investigated this issue, but Campbell et al. also described better patient satisfaction in practices with a good team climate [20,21,[32][33][34].
The results of this study should be interpreted with its limitations in mind.The representativeness of the samples might be biased, in particular the physician sample.Despite random sampling and good representativeness in terms of age, gender and rural/urban partition [26], the low participation acceptation rate in a practice-based research network (although classically observed) might introduce some level of bias for other unmeasured characteristics.In particular, participation in a research network might select physicians with specific modes of organisation or functioning.Additionally, the German-speaking physicians are a little under-represented in the network (less than 10% variation), probably because of the research team is based in the French-speaking area.Even though we cannot accurately estimate the impact of such potential biases, we assume that their impact on the results is negligible.Moreover, the lack of representativeness is less problematic in analytical analyses than in descriptive ones.Conversely, the participation rate among patients was high (around 84%), but data collection at the practice may have generated a declaration bias.A relatively small sample size, particularly regarding physician data, probably limits the possibility of observing more significant associations.Due to "missing" and sometimes "non-applicable" data related to each item, the number of respondents for each of the three dimensions was not the same.This could be a limitation in the objective to compare the results of the three multiple

Orginal article
models.Lastly, the study was cross-sectional and thus the causal nature of the associations cannot be assessed.

Conclusion
According to our study, in Switzerland, with a liberal feefor-services healthcare system and free choice of the physician, patient experience in family medicine is very good, regardless of the domain (access, family physician-patient communication or continuity coordination of care).Having patient, physician and practice data allowed us to show that variations in patients' reported experience were mainly due to patient characteristics, in particular, age and health status.However, associations between patient experience and some practice characteristics or organisation deserve attention for overall improvement of quality of care.In particular, intermediate-sized practices (between solo practices and large clinics), providing substantial time spent with the patients, represent an interesting configuration for best patient experience.Finally, the issue regarding the role of cultural aspects in patient reported experience, illustrated in this Swiss context, remains relevant, pertaining to quality of care, wherever the healthcare system.

Table 1 :
Patient, physician and practice characteristics.
Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.

Table 2 :
Factors associated with poor experience of access in family medicine.

Table 3 :
Factors associated with poor experience of communication in family medicine.
Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.CI = confidence interval; IRR = incidence rate ratio Multilevel analyses: * variables kept for multiple analyses (p ≤0.20 in multivariate analysis); ¶ interquartile range; ‡ Including final model patient variables; † Final model including patients' characteristics; § Final model including patient and physicians' characteristics