Dots indicate point estimates; bars indicate 95% CIs. Number of visits in each analysis: referral, 294; pain medication, 271; antibiotics, 107; other medication, 271; laboratory test, 448; radiology test, 153; other test, 147. Analytic reference (vertical line) for all request categories indicates fulfillment of request in the respective category. All analyses adjusted for the following patient characteristics: age, sex, race/ethnicity, education, marital status, body mass index, overall self-rated health, mental health status, medical skepticism, Five Factor Model personality factors, life satisfaction, worry about symptoms, and prior visit with the clinician. All analyses also adjusted for the 6 other request types (3-level categorical variables: no request in category, request in category denied, or request in category fulfilled).
eTable 1. Full results of the models.
eTable 2. Adjusted mean difference in patient satisfaction percentile (by visit, N = 1319) associated with patient characteristics and request disposition
Customize your JAMA Network experience by selecting one or more topics from the list below.
Jerant A, Fenton JJ, Kravitz RL, et al. Association of Clinician Denial of Patient Requests With Patient Satisfaction. JAMA Intern Med. 2018;178(1):85–91. doi:10.1001/jamainternmed.2017.6611
How is denial of specific types of patient requests during office visits associated with patient satisfaction with the clinician?
This was a cross-sectional study of 1141 adults making 1319 office visits to 56 family physicians. Compared with fulfillment of the respective request type, denials of requests for referral, pain medication, other new medication, and laboratory tests were associated with worse patient satisfaction with the clinician.
Denial of some types of requests was associated with worse patient satisfaction with the clinician, but not for others, when compared with fulfillment of the requests, suggesting the need for clinician training in request handling.
Prior studies suggesting clinician fulfillment or denial of requests affects patient satisfaction included limited adjustment for patient confounders. The studies also did not examine distinct request types, yet patient expectations and clinician fulfillment or denial might vary among request types.
To examine how patient satisfaction with the clinician is associated with clinician denial of distinct types of patient requests, adjusting for patient characteristics.
Design, Setting, and Participants
Cross-sectional observational study of 1319 outpatient visits to family physicians (n = 56) by 1141 adults at one Northern California academic health center.
Main Outcomes and Measures
We used 6 Consumer Assessment of Healthcare Providers and Systems Clinician and Group Adult Visit Survey items to measure patient satisfaction with the visit physician. Standardized items were averaged to form the satisfaction score (Cronbach α = 0.80), which was then percentile-transformed. Seven separate linear mixed-effects models examined the adjusted mean differences in patient satisfaction percentile associated with denial of each of the following requests (if present)—referral, pain medication, antibiotic, other new medication, laboratory test, radiology test, or other test—compared with fulfillment of the respective requests. The models adjusted for patient sociodemographics, weight, health status, personality, worry over health, prior visit with clinician, and the other 6 request categories and their dispositions.
The mean (SD) age of the 1141 patients was 45.6 (16.1) years, and 902 (68.4%) were female. Among 1319 visits, 897 (68.0%) included at least 1 request; 1441 (85.2%) were fulfilled. Requests by category were referral, 294 (21.1%); pain medication, 271 (20.5%); antibiotic, 107 (8.1%); other new medication, 271 (20.5%); laboratory test, 448 (34.0%); radiology test, 153 (11.6%); and other tests, 147 (11.1%). Compared with fulfillment of the respective request type, clinician denials of requests for referral, pain medication, other new medication, and laboratory test were associated with worse satisfaction (adjusted mean percentile differences, −19.75 [95% CI, −30.75 to −8.74], −10.72 [95% CI, −19.66 to −1.78], −20.36 [95% CI, −29.54 to −11.18], and −9.19 [95% CI, −17.50 to −0.87]), respectively.
Conclusions and Relevance
Clinician denial of some types of requests was associated with worse patient satisfaction with the clinician, but not for others, when compared with fulfillment of the requests. In an era of patient satisfaction-driven compensation, the findings suggest the need to train clinicians to deal effectively with requests, potentially enhancing patient and clinician experiences.
The importance of striving to improve patients’ subjective experiences of care, commonly referred to as patient satisfaction, is acknowledged by clinicians, administrators, policymakers, and the general public.1 Nonetheless, controversy exists regarding the interpretation and use of patient satisfaction measures seeking to assess the quality of patient-clinician interactions.2-4 In some studies, patient responses to such measures seem to partly reflect the clinician’s interpersonal communication abilities.5-7 The current practice of incentivizing (or penalizing) clinicians based on patient satisfaction is predicated on the notion that the scores reflect clinician interpersonal performance.8,9
While a patient-centered style has been associated with higher satisfaction with clinicians in some studies, such findings offer limited guidance to clinicians. Clinicians face the challenge of attaining favorable satisfaction ratings in time-limited office visits, with multiple competing demands.10,11 There is a need to explore how clinician handling of discrete issues or tasks that commonly arise in office visits influences patient satisfaction with clinicians, to help uncover priority targets for clinician communication efforts and, potentially, skills enhancement training.
Clinician fulfillment or denial of patient requests might be expected to influence patient satisfaction, yet the issue has received little study.12 Patient requests are ubiquitous; prior studies suggest over three-fourths of primary care visits include 1 or more requests.13 Clinicians have long surmised that denial of requests can lead to less favorable patient satisfaction,14,15 and preliminary studies provide some support for this hypothesis.13 However, the studies did not adjust extensively for patient characteristics, many of which confound ratings of satisfaction with clinicians.16 Thus, the independent influences of request denial on satisfaction ratings remain unclear. Furthermore, the studies did not separately examine how patient satisfaction was associated with clinician fulfillment or denial of distinct categories of patient requests (eg, for diagnostic tests vs treatment). This is an important gap, since research suggests that the strength of patient expectations varies across different aspects of care, and that clinicians perceive some types of requests as more difficult to address than others.17,18
We examined the associations of clinician denial of 7 categories of patient requests on patient satisfaction with clinicians.
We conducted the study from July 2015 to May 2016. We obtained ethics approval from the University of California Davis (UCD) institutional review board. Research assistants enrolled a convenience sample of patient participants from the UCD Family Medicine Clinic waiting room. They met the following eligibility criteria and provided written informed consent: (1) age of least 18 years; and (2) able to read, speak, and understand a survey written in English. The clinic conducts approximately 35 000 patient visits per year. We collected most study data via questionnaires that patients completed immediately after office visits with a resident or faculty family physician, before departure from the office. We collected data on the identity of the visit physician and patient’s body mass index (BMI) (calculated as weight in kilograms divided by height in meters squared) from electronic health records. Patients completed the postvisit questionnaire using a study touch screen tablet device. We provided participants with a $10 gift card following questionnaire completion, to compensate them for their time.
Patient satisfaction with the clinician was measured using a scale comprising 6 items from the Consumer Assessment of Healthcare Providers and Systems Clinician and Group (CG-CAHPS) Adult Visit Survey.19,20 Four of the items were from the CG-CAHPS Doctor Communication Composite. Of these, the first asked whether the physician gave them easy-to-understand information about their health questions or concerns (yes, definitely; yes, somewhat; uncertain/mixed; or did not have any health questions or concerns), while the other 3, respectively, asked whether the physician knew important information about their medical history, showed respect for what they had to say, and spent enough time with them (yes definitely, yes somewhat, or no). The fifth CG-CAHPS item was the Overall Doctor Rating item, which asked the patient to rate the physician on a scale of 0 (worst physician possible) to 10 (best physician possible). The sixth CG-CAHPS item was the Recommend Doctor Rating, which asked the patient whether they would recommend the physician to their family and friends (yes, definitely; yes, somewhat; or no).
Participants also responded to a series of previously validated items asking whether they had made requests in any of the following service categories (yes/no) and, if so, whether they perceived that the clinician had fulfilled or denied the request: referral to another clinician; pain medication; antibiotic; other new medication; laboratory testing; radiology testing; or other testing (eg, sleep study).21,22
We also measured a number of factors previously shown to be associated with patient satisfaction with clinicians.6,16,23-30 Sociodemographics measured included age, sex, race/ethnicity, education level, and marital status. Body mass index was gathered from visit vital signs documentation in the electronic health record. Overall self-rated health was assessed with a single validated item: “In general, would you say your health is (excellent, very good, good, fair, or poor)?”31 Mental health status was assessed using the validated Mental Health Inventory, a 5-item measure assessing both depressive and anxiety disorders (range of scores, 4-100, higher scores indicate fewer symptoms).32 Patients’ degree of skepticism regarding medical care was assessed using a validated 4-item scale (range of scores, 1-5; higher scores indicate more skepticism).33,34 Five Factor Model (FFM) personality factors—agreeableness, conscientiousness, extraversion, neuroticism, and openness—were measured with the well-validated Big Five Inventory.35 The Big Five Inventory consists of 44 statements; for each statement, patients are asked to indicate their level of agreement or disagreement on a 5-point scale (strongly disagree to strongly agree) (range of scores for each factor, 1-5; higher scores indicate higher standing on the factor). Global life satisfaction was assessed using the 5-item Satisfaction with Life Scale, a validated measure of subjective well-being with high temporal reliability (range of scores, 5-35, higher scores indicate greater life satisfaction).36 Three questions with 5-point Likert response scales, assessing bothersomeness of and worry and concerns about any symptoms the patient was experiencing, respectively, were combined into a “worry about symptoms” measure (range of scores, 3-15; higher scores indicate greater worry).21,22 Finally, patients were asked to indicate whether they had previously seen the clinician.16
Analyses were conducted using Stata statistical software (version 15.0; StataCorp). We standardized (ie, Z score transformed) each of the 6 CG-CAHPS items and averaged them to form each patient’s satisfaction score. Cronbach α for this scale in our sample was 0.80. The satisfaction score distribution was highly skewed, with most scores clustered around an upper limit. To reduce the skewness of the distribution, we percentile-transformed the satisfaction scores, replacing ties with the mean percentile.37,38 Higher percentiles indicate higher satisfaction. Percentiles are more generalizable across a sample of visits than absolute rankings, since absolute rankings depend on the number of visits being ranked, while percentiles do not.
We used 7 linear models (1 for each type of request under study), each with patient satisfaction percentile as the dependent variable, and each excluding visits without the type of request under consideration. Cross-nested mixed models were used to allow adjustment for the possible nesting of patient visits within clinicians and clinicians within patients. The key independent variable in each analysis was whether the respective request type was either denied or fulfilled (the latter used as reference). Negative parameter estimates in the model indicate lower (worse) adjusted mean patient satisfaction percentiles, while positive estimates indicate higher (better) adjusted mean satisfaction percentiles. All models adjusted for the following patient characteristics, previously found to be associated with patient satisfaction ratings6,16,23-30: age, sex, race/ethnicity, education, marital status, BMI, overall self-rated health, mental health status, medical skepticism, FFM personality factors, life satisfaction, worry about symptoms, and prior visit with the clinician (vs none). Finally, all models also adjusted for the other 6 types of requests (using 3-level variables for each: no request in category, request in category denied, or request in category fulfilled [reference category]).
We also conducted a single-model sensitivity analysis of all study visits. The model included all 7 request categories (each represented by 3-level variables: no request in category, denied request in category, or fulfilled request in category [reference category]) and adjusted for all of the patient characteristics listed herein.
The questionnaire was completed immediately following 1319 visits, made to a total of 56 physicians by 1141 patients. The mean age of the 1141 patients was 45.6 (16.1) years, and 902 (68.4%) were female. Characteristics of patients, by visit, are summarized in Table 1. Regarding the distribution in total requests (all categories combined), 422 visits (32.0%) had 0, 432 visits (32.8%) had 1, 265 visits (20.1%) had 2, 115 visits (8.7%) had 3, 55 visits (4.2%) had 4, 17 visits (1.3%) had 5, 12 (0.9%) visits had 6, and 1 visit (0.1%) had 7. The sample mean (SD) unstandardized patient satisfaction score was 23.2 (2.5) (range of scores, 9.0-25.0).
Patients reported that 1441 (85.2%) of their requests were fulfilled by the clinician. Table 2 shows patient requests and rates of clinician fulfillment by service category. Laboratory tests were most frequently requested, followed by pain medication, other medication, and referrals, while antibiotics were least frequently requested. Patient-reported request fulfillment exceeded 80% for all service categories except radiology testing (66% fulfilled) (Table 2).
Table 3 summarizes the adjusted results of the 7 cross-nested mixed models of patients’ adjusted mean satisfaction with the clinician conducting the visit, as a percentile by visit. The full results of the models are available in eTable 1 in the Supplement). As depicted in the Figure, compared with fulfillment of a request in the respective category, denials of requests for pain medication, referral, other new medication, and laboratory tests were associated with significantly lower (worse) adjusted mean patient satisfaction percentile. Denials of requests for radiology tests and other tests (each compared with fulfillment of the respective request type) were associated with non–statistically significantly lower adjusted mean satisfaction percentiles. Denial of requests for antibiotics was not associated with lower satisfaction.
The sensitivity analysis of all study visits with all 7 request types in 1 model revealed findings consistent with those of the 7 separate analyses (results are available in eTable 2 in the Supplement).
We found that clinician denial of various types of patient requests during office visits, compared with fulfillment of the respective request types, was associated with patient satisfaction with the clinician, adjusting for patient factors previously shown to be associated with such ratings. Specifically, denials of requests for referral, pain medication, other (nonpain, nonantibiotic) new medication, and laboratory tests were associated with significantly lower (worse) adjusted satisfaction percentiles, each relative to fulfillment of requests in the respective category. Our findings may have relevance given the prevailing focus on patient satisfaction in US health systems. Clinician handling of requests for pain medication, other new medications, referrals, and laboratory tests seems the most critical. For visits wherein patients' requests for these services were denied, patient satisfaction ratings were 10 to 20 percentiles lower when compared with visits wherein requests for these services were fulfilled. The sizes of the associations suggest that clinicians who are less likely than their colleagues to fulfill patient requests for these services could face a penalty in satisfaction ratings, potentially affecting clinician career satisfaction, compensation, and handling of subsequent requests for these services. Additional studies exploring these possibilities are warranted.
The mechanisms of our findings are uncertain; our study was not designed to address this issue. However, it is not surprising that denial of requests for pain medication would be associated with less favorable patient satisfaction, given research suggesting that patients bring high expectations for pain control to visits.39 Similarly, prior work indicates that patients often request diagnostic testing to help assuage worries about having a serious health condition, potentially explaining the association of laboratory test fulfillment with satisfaction with clinician ratings in our study.40 The same mechanism may largely explain the association we observed between referral denial and lower satisfaction with clinician.41 The mechanisms of the association of denial of other (nonpain, nonantibiotic) new medication requests with patient satisfaction with clinician are less clear, given the breadth of medications encompassed by this category; additional studies will be required to explore this issue. The absence of a negative adjusted association between antibiotic request denial and satisfaction may reflect the small sample size for that request category (and associated wide 95% CIs; see Table 3 and the Figure). It may also be that clinicians have more experience at denying antibiotic requests than other types of requests, given sustained and publicized efforts to improve antibiotic stewardship.42
Regardless of their mechanisms, in the current era of clinician compensation tied partially to patient satisfaction ratings, we believe our findings provide guidance to clinicians and health care administrators charged with delivering high technical quality, cost-effective care while also optimizing patients’ subjective experiences of care.43 Recommendations to clinicians for optimizing patient satisfaction ratings assert this can best be achieved by using a patient-centered communication approach.9 Yet the patient-centered communication construct is broad in scope and relatively complex to apply, and its use in primary care has been associated with increased visit length.7,44 It is challenging for clinicians to approach the many tasks and issues they face in brief office visits using the patient-centered communication paradigm while still meeting clinical productivity targets (also influencing clinician compensation).45 In the face of these competing imperatives, clinicians may be tempted to adopt a default approach of simply acquiescing to patient requests, including requests for low-value care, in an effort to maintain both favorable patient satisfaction ratings and clinical productivity targets.46 Further reflecting these competing demands, prior work indicates that clinicians perceive visits in which patients request diagnostic testing as being more difficult than visits without such requests.17
Taken together, the foregoing observations suggest the potential utility of targeted clinician training in brief communication techniques to address patient requests. Training might be developed to provide clinicians with communication approaches that foster a positive patient experience without simply acquiescing to requests for low-value care, thereby avoiding the harms of unnecessary evaluation and treatment, maintaining good stewardship of resources, and potentially enhancing clinician career satisfaction.47-50 One potentially fruitful strategy may be to train clinicians to offer a watchful waiting (“wait and see”) option in response to patient requests when appropriate (eg, in handling requests for services that are not clearly indicated based on the clinician’s initial evaluation).51 Such an approach has face appeal in being focused and relatively brief while also involving the patient in decision-making and offering a middle ground between immediate acquiescence and flat denial of requests, consistent in spirit with the broader, patient-centered communication paradigm.7 In an observational analysis, we found that a clinician offer of a watchful waiting option was associated with less low-value test ordering, with no decrement in patient satisfaction.52 However, few randomized clinical trials of watchful waiting have been conducted, and none have examined the effectiveness of the approach, specifically in responding to patient requests in primary care or the resulting impact on satisfaction, underscoring the need for work in this realm.53-55
A key strength of our study was adjustment for a broad array of patient characteristics known to influence patient satisfaction with clinicians.16 Nonetheless, patient characteristics and contextual factors (eg, the details of the requests) not measured in our study might also influence patient satisfaction with clinician, warranting further study. Our study had other limitations. The sample sizes for some request categories were relatively small. Still, the findings of the 7 separate analyses were consistent with those of a single-model sensitivity analysis with a larger sample size (all study visits included) (eTable 2 in the Supplement). The observational cross-sectional nature of the analyses precludes causal inference. The study was conducted within a convenience sample of English-speaking adults in a single family medicine residency training clinic at an academic health center in Northern California, and focused on only 1 aspect of the patient experience. Thus, the generalizability of the findings to other geographic regions, care settings, patients, and satisfaction ratings (eg, with the health system in general) is uncertain.
We took steps to reduce the skewness of our patient satisfaction score distribution. Nonetheless, the tendency for scores to cluster near the top end of the range (a common issue in satisfaction studies) may still have limited our ability to capture the impact of clinician request handling on satisfaction.
Patients self-reported requests and whether they were fulfilled or denied by the clinicians. Both underreporting and overreporting of requests and of clinician fulfillment or denial of the requests may have occurred, with the potential for variation in overreporting and underreporting by request type and patient characteristics, yielding uncertain net impact on the study findings. Nonetheless, other ascertainment methods also have drawbacks. For example, making audio or video recordings of the visits to allow coding of patient requests and clinician responses might well alter patient and clinician behaviors related to request making and handling, respectively.56,57
We also lacked a measure of longitudinal continuity with the clinician, which may influence satisfaction. We did adjust for whether the patient had ever seen the clinician, which was associated with satisfaction in most of our models (eTable 1 in the Supplement).
We found that clinician denial of some types of patient requests was associated with reduced patient satisfaction with the clinician, but not for others, when compared with fulfillment of the requests, accounting for patient factors also associated with patient satisfaction. Specifically, patients were less satisfied with clinicians who denied requests for referral, pain medication, other (nonpain, nonantibiotic) new medication, and laboratory tests than with clinicians who fulfilled such requests. In an era of satisfaction score–driven compensation of clinicians, the findings suggest the need to explore the utility of training clinicians to help better handle patient requests, potentially optimizing the patient experience while also enhancing clinician career satisfaction.
Corresponding Author: Anthony Jerant, MD, Department of Family and Community Medicine, University of California Davis School of Medicine, 4860 Y St, Ste 2300, Sacramento, CA 95817 (email@example.com).
Accepted for Publication: September 24, 2017.
Published Online: November 27, 2017. doi:10.1001/jamainternmed.2017.6611
Author Contributions: Dr Jerant had full access to all data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Jerant, Fenton, Kravitz, Magnan, Bertakis, Franks.
Acquisition, analysis, or interpretation of data: Jerant, Fenton, Kravitz, Tancredi, Magnan, Franks.
Drafting of the manuscript: Jerant, Franks.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Fenton, Tancredi, Franks.
Obtained funding: Fenton.
Administrative, technical, or material support: Jerant, Fenton, Bertakis.
Study supervision: Fenton, Bertakis, Franks.
Conflict of Interest Disclosures: None reported.
Funding/Support: This work was supported by a grant from the University of California Davis Department of Family and Community Medicine. Dr Magnan was supported by the National Center for Advancing Translational Sciences, National Institutes of Health, through grant No. UL1 TR001860 and linked award KL2 TR001859.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Meeting Presentation: This study was presented in part at the Society of Teachers of Family Medicine Annual Spring Conference; May 5-9, 2017; San Diego, California.
Additional Contributions: We are grateful to the following individuals, all of whom were employed and compensated as study support staff at the University of California Davis: Rimaben Cabrera, MSW, who managed the project and contributed to study recruitment and data collection; and Eliot Lee and Leyleh Salem for assistance in data collection.
Create a personal account or sign in to: