eTable 1. Characteristics of patients who completed full baseline surveys and were included in the analysis, versus patients excluded who completed only brief surveys or whose surveys were completed by surrogates.
eTable 2. Odds of reporting excellent quality of care for treatments received (complete regression results for main effects model with respect to preferred and actual decision roles).
eTable 3. Odds of reporting excellent physician communication ratings (complete regression results for main effects model with respect to preferred and actual roles).
Kehl KL, Landrum MB, Arora NK, Ganz PA, van Ryn M, Mack JW, Keating NL. Association of Actual and Preferred Decision Roles With Patient-Reported Quality of CareShared Decision Making in Cancer Care. JAMA Oncol. 2015;1(1):50-58. doi:10.1001/jamaoncol.2014.112
Copyright 2015 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
Shared decision making is associated with improved patient-reported outcomes of cancer treatment, but not all patients prefer to participate in medical decisions. Results from studies of the effect of matching between actual and preferred medical decision roles on patients’ perceptions of care quality have been conflicting.
To determine whether shared decision making was associated with patient ratings of care quality and physician communication and whether patients’ preferred decision roles modified those associations.
Design, Setting, and Participants
We performed a population- and health system–based survey of participants in the Cancer Care Outcomes Research and Surveillance Consortium (CanCORS) study diagnosed with lung and/or colorectal cancer between 2003 and 2005 (56% with colorectal cancer, 40% with non–small-cell lung cancer, and 5% with small-cell lung cancer). The CanCORS study included 9737 patients (cooperation rate among patients contacted, 59.9%) treated in integrated care delivery systems, academic institutions, private offices, and Veterans Affairs hospitals. The medical records were abstracted between October 11, 2005, and April 30, 2009; all analyses were conducted between 2013 and 2014.
We surveyed patients specifically about their preferred roles in cancer treatment decisions and their actual roles in decisions about surgery, chemotherapy, and radiation therapy. We analyzed the responses of 5315 patients who completed baseline surveys and reported decision roles for a total of 10 817 treatment decisions and assessed associations of patients’ decision roles with patient-reported quality of care and physician communication.
Main Outcomes and Measures
The outcomes (identified before data collection) included patient-reported excellent quality of care and top ratings (highest score) on a physician communication scale.
After adjustment, patients describing physician-controlled (vs shared) decisions were less likely to report excellent quality of care (odds ratio [OR], 0.64; 95% CI, 0.54-0.75; P < .001). Patients’ preferred decision roles did not modify this effect (P = .29 for the interaction). Patients describing either actual or preferred physician-controlled (vs shared) roles were less likely to provide a top rating of physician communication (OR, 0.55; 95% CI, 0.45-0.66; P < .001, and OR, 0.67; 95% CI, 0.51-0.87; P = .002, respectively). The preferred role did not modify the effect of the actual role (P = .76 for interaction).
Conclusions and Relevance
Physician-controlled decisions regarding lung or colorectal cancer treatment were associated with lower ratings of care quality and physician communication. These effects were independent of patients’ preferred decision roles, underscoring the importance of seeking to involve all patients in decision making about their treatment.
The Institute of Medicine1 has called for shared decision making and accommodation of patient preferences to improve overall health care quality, and in particular the quality of cancer care.2 Prior studies of shared decision making in cancer treatment have found that most patients prefer to play a role in treatment decisions, but the degree to which their desired role matches their actual role in decision making varies.3- 5 Much of this work has focused on surgical decisions in breast cancer.4- 6 Evidence suggests that patients who are younger, less educated, and who see higher-volume surgeons are less likely to have actual roles that match their preferred roles5 and that patients whose preferred decision-making roles match their actual roles are more satisfied with their treatment choices.4,6 Nevertheless, 1 small study of patients with a variety of cancer types found that patients’ actual roles, but not matching between actual and preferred roles, were associated with greater satisfaction.7
Although their utility as metrics of quality is controversial, patients’ reports of their experiences with care are increasingly important health care performance measures.8,9 Indeed, the Affordable Care Act calls for the use of the patient experience Clinician and Group Consumer Assessment of Healthcare Providers and Systems survey as a comparative measure of physician performance.9 It is possible that patients who are more actively engaged in their decisions, or whose roles match their preferred roles, may have better care experiences.
In a prior analysis, our research group examined the roles in decisions reported by patients in the Cancer Care Outcomes Research and Surveillance (CanCORS) study,10 a large, population- and health-system based study of patients with lung and colorectal cancer. Among 10 939 treatment decisions made by 5383 patients, 39% were categorized as “patient controlled,” 44% as “shared,” and 17% as “physician controlled.” In the present study, we examined patients’ preferred roles in decisions to better understand the relative influence of preferred vs actual roles in decisions regarding surgery, chemotherapy, and radiation therapy. Specifically, we assessed associations between patients’ actual roles in decisions and (1) patient-reported quality of care for each treatment modality received (surgery, chemotherapy, and/or radiation therapy) and (2) patient ratings of physician communication. In addition, because evidence suggests that there may be benefits to matching actual to preferred roles,4,6 we assessed whether associations between actual role and patient-reported quality or physician communication ratings were modified by patients’ preferred roles in decision making.
Patients reporting physician-controlled decisions provided lower ratings of care quality and physician communication than patients reporting shared decisions.
These lower ratings were evident regardless of patients’ preferred role in decision making.
Physicians should work to involve all patients in decisions about cancer care.
This study was approved by the human research ethics committees at all participating institutions. Written informed consent was obtained at 2 of the study sites and waived by the other review boards. Verbal informed consent was obtained from each participant in lieu of written consent.
The CanCORS study investigated care processes, patient experiences with care, and outcomes among patients newly diagnosed with lung and/or colorectal cancer between 2003 and 2005 and living within 1 of 5 geographic regions (Northern California, Los Angeles County, North Carolina, Iowa, or Alabama) or receiving care in one of 5 health maintenance organizations or 15 Veterans Affairs sites.11 Cases were identified using rapid case ascertainment based on registry data.12,13 Patients (or surrogates if patients had died or were too ill) were surveyed 3 to 6 months after diagnosis. The American Association for Public Opinion Research14 survey response rate was 51.0%; the cooperation rate was 59.9%.15 Additional details about patient eligibility have been described previously15; the CanCORS cohort is representative of US patients with lung and colorectal cancer.15 Information on cancer type, histology, and stage were obtained from registry data and medical records, which were abstracted between October 11, 2005, and April 30, 2009. Between 2013 and 2014, we analyzed the subset of patients who were alive at the time of the baseline survey and completed a full baseline interview themselves (n=5518, Figure) and who answered questions about their preferred role in medical decisions in general and their actual role in decisions about at least 1 of the following: surgery, chemotherapy, and radiation therapy (n=5315 patients). Characteristics of these patients compared with respondents who completed surrogate survey versions or brief surveys or did not answer questions about decision roles are listed in eTable 1 in the Supplement.
Patients reported their perception of the overall quality of care for each treatment modality they received. Response options included “excellent,” “very good,” “good,” “fair,” or “poor.” To facilitate presentation of results, responses were grouped into excellent vs all other responses because most patients (67.8%) responded excellent.
As previously described,13,16 we derived 5 questions related to physician communication from the Consumer Assessment of Healthcare Providers and Systems survey.17 The questions were as follows:
How often did your doctors
listen carefully to you?
explain things in a way you could understand?
give you as much information as you wanted about your cancer treatments, including potential benefits and side effects?
encourage you to ask all the cancer-related questions you had? and
treat you with courtesy and respect?
Response options were “always” (3 points), “usually” (2 points), “sometimes” (1 point), and “never” (0 points). We averaged these scores and grouped the results into 2 categories: top-rated physicians (score of 3, awarded by 55.8% of patients) and others (score of <3). Patients rated their physicians as a group rather than rating each physician separately.
We assessed patients’ overall preferred roles in decision making for their cancer treatment, and for each treatment modality considered (surgery, chemotherapy, and/or radiation therapy), we also assessed the actual role patients played in decision making for that modality. Preferred and actual roles for decisions were ascertained using the 5-item Control Preferences Scale.3,18,19 Response options for preferred roles were as follows:
You prefer to make decisions about treatment with little or no input from your doctors.
You prefer to make the decisions after considering your doctor’s opinion.
You prefer that you and your doctors make the decisions together.
You prefer that your doctors make the decisions after considering your opinion.
You prefer your doctors make the decision with little or no input from you.
You made the decision with little or no input from your doctors.
You made the decision after considering your doctors’ opinions.
You and your doctors made the decision together.
Your doctors made the decision after considering your opinion.
Your doctors made the decision with little or no input from you.
Our analyses adjusted for age at diagnosis, sex, race and ethnicity, marital status, education, geographic region, income, enrollment in an integrated health care system (patients enrolled through the US Department of Veterans Affairs, Kaiser Permanente of Northern California, Kaiser Permanente of Southern California, or other health maintenance organization sites), number of self-reported comorbid conditions,21- 23 health status before diagnosis (which is associated with patient satisfaction24; our measure included a subset of 5 questions from the 12-Item Short-Form Health Survey,25 and categorized in quartiles),25 and depression (positive response for ≥6 of the 8-item Center for Epidemiological Studies–Depression scale26). We also adjusted for treatment modality, cancer type, and stage at diagnosis. The patient-reported quality outcome was assessed only among patients who reported receiving the treatment corresponding to each decision. For the analyses in which the outcome variable was ratings of physician communication, we also adjusted for whether patients reported receiving the treatment corresponding to each decision. Variables were categorized as listed in Table 1.
The 5315 patients in our cohort reported decision roles for 10 817 treatment decisions (4559 surgery, 3928 chemotherapy, and 2330 radiation therapy decisions; Figure). For analyses of perceived quality of care for each treatment, these decisions constituted the unit of analysis, with 1 observation per decision. Each patient could have up to 3 observations if they participated in decisions about surgery, chemotherapy, and radiation therapy. There were 8201 decisions made by 5176 patients who received a treatment under consideration and rated overall quality of care for that treatment.
For analyses of physician communication, we included only 1 decision per patient, since patients provided overall ratings of communication with their physicians rather than rating each treating physician. We identified the most frequently discussed treatment decision (ie, surgery, chemotherapy, or radiation therapy) for each cancer type and stage to select the treatment for which we would include the actual decision role. Among patients with stage I or II non–small-cell lung cancer or stage I to III colorectal cancer, we included surgery decisions; for patients with stage III to IV non–small-cell lung cancer, small-cell lung cancer, or stage IV colorectal cancer, we included chemotherapy decisions. There were 4848 patients who made such decisions. We further restricted analysis to the 4830 patients who answered at least 3 of the 5 questions about communication with their physicians. For patients who answered only 3 or 4 of the 5 items, we averaged their responses; we also performed sensitivity analyses in which multiple imputation was used to impute missing responses for the questions. Results were similar.
In unadjusted analyses, we used bivariable logistic regression to assess the associations of decision roles and other clinical characteristics with excellent patient-reported quality and top ratings of physician communication. For the patient-reported quality outcome, we used a robust covariance estimator to adjust standard errors for repeated measures within patients. We report P values for tests of combined significance of the categorical independent variables.
We used multivariable logistic regression to assess the association of actual and preferred roles in decisions with (1) patient-reported quality of care for the treatment considered during each decision (with a robust covariance estimator to account for repeated measures within patients) and (2) top ratings of physician communication, adjusting for all noted patient characteristics. For each dependent variable, we examined models that included actual decision role and preferred decision role. We also examined models that included the interaction of actual and preferred role. Finally, we examined the effect of actual role in decision making, stratified by preferred decision roles. We calculated adjusted probabilities of each outcome for particular role variables by taking the mean of predicted probabilities generated by the model for each observation, allowing other covariates to retain values from the original data. In sensitivity analyses, we also used proportional odds models in which the patient-reported quality outcome ranged from 0 to 4, per the original survey scale, and in which the physician communication rating was categorized into tertiles. We also conducted sensitivity analyses stratified by patient sex. Results of all sensitivity analyses were similar.
No data were missing for our dependent variables, per our cohort definitions. Missing data were infrequent for demographic and clinical factors (<10% nonresponse for all items; Table 1). For adjusted analyses, we used multiple imputation to impute missing data for our independent variables; we did not use imputed data for the decision role variables, the primary independent variables of interest.27 For patient-reported quality, 10 of 8201 decisions were made by patients completing only partial versions of their surveys, for which data were not imputed for some control variables; these decisions were excluded from analysis, and the final analysis cohort for that outcome included 8191 decisions made by 5170 patients. Similarly, for ratings of physician communication, 5 of 4830 decisions were excluded from models owing to nonimputed missing data for partially completed surveys, leaving a final analysis cohort of 4825 patient decisions. Two-sided P < .05 was considered statistically significant. Analyses were conducted using SAS (version 9.2) and Stata (version 13) software.
Most of the 5315 patients (56%) had colorectal cancer; 40% had non–small-cell lung cancer, and 5% had small cell lung cancer. Most of the patients (58%) preferred shared roles in decision making about their cancer; 36% preferred patient-controlled decisions, and 6% preferred physician-controlled decisions. Other patient characteristics are listed in Table 1.
Patients in our cohort made 10 817 treatment decisions; 42% regarding surgery, 36% regarding chemotherapy, and 22% regarding radiation therapy. Participants reported that their actual decision-making process was patient controlled in 39% of decisions, shared in 44% of decisions, and physician controlled in 17% of decisions.
For 67.8% of treatments received by patients, patients reported their care by the physician performing the treatment as excellent. In adjusted analyses examining preferred and actual roles in decisions, the interaction of preferred and actual role was not statistically significant (P = .29), and only the main effects model is presented herein. Preferred role was not associated with ratings of quality, but patient reports that treatment decisions were physician controlled (vs shared) were associated with lower odds of excellent patient-reported quality (odds ratio [OR], 0.64; 95% CI, 0.54-0.75; P < .001) (Table 2). In models stratified by preferred role, the negative associations between physician-controlled (vs shared) decisions and patient-reported quality were evident regardless of preferred role (Table 2).
Overall, 55.8% of patients gave their physicians the highest possible rating of communication. In adjusted analyses, the interaction of preferred and actual decision roles was not statistically significant (P = .76), and we present only the main effects model (Table 3). Both preferred and actual decision roles were associated with top ratings. Patients who preferred physician-controlled to shared decisions were less likely to give top ratings to their physicians (OR, 0.67; 95% CI, 0.51-0.87; P = .002), as were patients who reported actually experiencing physician-controlled vs shared decisions (OR, 0.55; 95% CI, 0.45-0.66; P < 001). In models stratified by preferred role, the association of actual physician-controlled decisions with lower ratings remained evident, although this finding did not reach statistical significance for the relatively small number of decisions made by patients preferring physician-controlled decisions (OR, 0.61; 95% CI, 0.33-1.13; P = .12).
In this large, population- and health-system–based cohort of patients with recently diagnosed lung and colorectal cancer, we found that among patients receiving a treatment under consideration, those reporting physician-controlled vs shared decisions were less likely to report excellent quality of care for that treatment. This effect was not modified by preferred role in treatment decision making, implying that shared decision making was associated with higher perceived quality, even for patients preferring less active roles in medical decisions. Of note, patients were only asked to report quality of care when they also reported receiving the treatment in question. Therefore, factors such as whether these patients were treated, or were eligible for treatment, could not have contributed to their perceptions of involvement in treatment decisions or of the quality of care resulting from the decisions.
Similarly, patients who reported physician-controlled decisions gave lower ratings of physician communication compared with those reporting shared decisions, and patients’ preferred role in decisions did not modify this effect. Interestingly, even after adjustment for actual decision role and whether patients received the treatments under consideration, patients who expressed a preference for physician-controlled decisions were independently less likely to rate physician communication highly. The explanation for this counterintuitive finding is not obvious, especially since other recent work has indicated higher levels of trust in physicians among patients preferring physician-controlled decisions.28 These patients may have different overall attitudes toward clinicians, and further work is needed to understand how they approach decision making throughout their treatment courses.
The Institute of Medicine highlighted the importance of engaged and well-informed patients, along with shared decision making, as central to a high-quality cancer care delivery system.2 Patient ratings of their experiences are also playing increasingly important roles as performance metrics.8,9 Although some patients may prefer that physicians take a leading role in decision making,3,29 other evidence suggests that patients also want information about their treatments and prefer to take part in decisions30 and that patient preferences for involvement in decisions have increased over time.31 Our findings suggest that providing information to patients with colorectal and lung and engaging them in shared decisions is valuable, even for patients who express preferences for physicians to control the decision-making process. Notably, the lack of an effect of matching between actual and preferred roles differed from prior findings in breast cancer.4,6 This may reflect heterogeneity of the effect of role matching across disease types, or temporal changes in attitudes or expectations about decision making. In sensitivity analyses, we found no evidence of effect modification by sex.
Strengths of our analysis include the large, multiregional, and representative15 cohort of patients with recently diagnosed lung and colorectal cancer. Several limitations remain, however. First, data were collected 3 to 6 months after diagnosis, so reports of roles played in treatment decisions may in some cases have been ascertained several months after those decisions took place.10 In addition, we asked patients for preferences regarding decision making and for ranking of their physicians’ communication in general, not with regard to specific therapies, and it is possible that preferred roles may have varied by the treatment under consideration.30 It is also possible that patients’ impressions of their decision-making processes may have been most influenced by the provider who played the most prominent role in their care.
As with any survey, our analysis was subject to nonresponse bias, although prior analysis of the CanCORS data has demonstrated that the demographics of the cohort changed little as data collection proceeded from initial ascertainment to final enrollment,15 and the response rate was relatively high, particularly for a population-based survey. Still, our analysis focused on patients who were alive and healthy enough to complete a full baseline survey approximately 3 to 6 months after diagnosis. Our outcomes were subjective, patient-reported measures; nevertheless, we believe that patients’ perceptions of care quality and communication with their physicians are important, and they are playing increasingly important roles in quality-improvement efforts.
We adjusted for many clinical and demographic characteristics in this analysis; however, as in any observational study, we cannot exclude the possibility that unmeasured confounders may in part explain the associations between patient decision roles and ratings of care quality or physician communication. Finally, some evidence suggests that patients’ descriptions of their preferred decision roles may vary according to the measurement scale used.31 We used the well-accepted Control Preferences Scale,3,18,19 but these results do not exclude the possibility that, for example, even patients who prefer to control treatment decisions might also want clear treatment recommendations from their physicians.
Among patients with newly diagnosed lung and colorectal cancer who made decisions about surgery, chemotherapy, and/or radiation therapy, patients who experienced physician-controlled vs shared decision making were less likely to report excellent quality of care and top ratings of physician communication. These associations were similar regardless of patients’ preferred roles in decisions. Given the increasing emphasis on patient experiences and ratings in health care, these results highlight the benefits of promoting shared decision making among all patients with cancer, even those who express preferences for less active roles.
Accepted for Publication: November 17, 2014.
Corresponding Author: Nancy L. Keating, MD, MPH, Department of Health Care Policy, Harvard Medical School, 180 Longwood Ave, Boston, MA 02115 (email@example.com).
Published Online: February 12, 2015. doi:10.1001/jamaoncol.2014.112.
Author Contributions: Dr Kehl had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kehl, Landrum, Arora, van Ryn, Keating.
Acquisition, analysis, or interpretation of data: Kehl, Landrum, Arora, Ganz, Mack, Keating.
Drafting of the manuscript: Kehl.
Critical revision of the manuscript for important intellectual content: Kehl, Landrum, Arora, Ganz, van Ryn, Mack, Keating.
Statistical analysis: Kehl, Landrum, van Ryn, Keating.
Obtained funding: Keating.
Administrative, technical, or material support: van Ryn.
Study supervision: Keating.
Conflict of Interest Disclosures: None reported.
Funding/Support: This work of the CanCORS Consortium was supported by grants from the National Cancer Institute (NCI) to the Statistical Coordinating Center (U01 CA093344) and to the following NCI-supported Primary Data Collection and Research Centers: Dana Farber Cancer Institute/Cancer Research Network (U01 CA093332), Harvard Medical School/Northern California Cancer Center (U01 CA093324), RAND Corporation/University of California, Los Angeles (U01 CA093348), University of Alabama at Birmingham (U01 CA093329), University of Iowa (U01 CA093339), and University of North Carolina (U01 CA093326); it was also supported by a Department of Veterans Affairs grant to the Durham VA Medical Center (CRS 02-164). Drs Keating and Landrum were supported by NCI grant 1R01CA164021-01A1.
Role of the Funder/Sponsor: The CanCORS initiative was funded as a cooperative agreement and hence involved significant participation from several scientific staff members from the NCI and the Department of Veterans Affairs. Specifically, investigators from the NCI and the Department of Veterans Affairs collaborated in the design and conduct of the study and in data collection and management at some study sites. For this article, Dr Arora, an investigator at NCI, participated in analysis and interpretation of the data and review and approval of the manuscript; the manuscript was also reviewed by 2 other members of NCI scientific staff before submission.
Previous Presentation: This research was presented in part in abstract form at the 2014 annual meeting of the American Society of Clinical Oncology; June 3, 2014; Chicago, Illinois.