|Home | About | Journals | Submit | Contact Us | Français|
Many patients with diabetes have multiple other chronic conditions, but little is known about whether these patients and their primary care providers agree on the relative importance that they assign these comorbidities.
To understand patterns of patient-provider concordance in the prioritization of health conditions in patients with multimorbidity.
Prospective cohort study of 92 primary care providers and 1,169 of their diabetic patients with elevated clinic triage blood pressure (≥140/90) at nine Midwest VA facilities.
We constructed a patient-provider concordance score based on responses to surveys in which patients were asked to rank their most important health concerns and their providers were asked to rank the most important conditions likely to affect that patient’s health outcomes. We then calculated the change in predicted probability of concordance when the patient reported having poor health status, pain or depression, or competing demands (issues that were more pressing than his health), controlling for both patient and provider characteristics.
For 714 pairs (72%), providers ranked the patient's most important concern in their list of three conditions. Both patients and providers ranked diabetes and hypertension most frequently; however, providers were more likely to rank hypertension as most important (38% vs. 18%). Patients were more likely than providers to prioritize symptomatic conditions such as pain, depression, and breathing problems. The predicted probability of patient-provider concordance decreased when a patient reported having poor health status (55% vs. 64%, p<0.01) or non-health competing demands (46% vs. 62%, p<0.01).
Patients and their primary care providers often agreed on the most important health conditions affecting patients with multimorbidity, but this concordance was lower for patients with poor health status or non-health competing demands. Interventions that increase provider awareness about symptomatic concerns and competing demands may improve chronic disease management in these vulnerable patients.
Patients with diabetes often face the challenge of managing multiple chronic conditions. On average, adults with diabetes have 3.5 other chronic conditions,1 and these comorbidities are associated with increased health care cost and utilization.2,3 Comorbid conditions may interfere with self-management of diabetes,4,5 and acute symptomatic comorbidities such as chronic pain and depression can be especially disruptive.4,6–9 Providers, meanwhile, face the challenge of addressing multiple complex conditions in a brief office visit.10–14 Given our knowledge of the benefits of effective patient-provider communication and trust on a range of health outcomes,15–18 a shared understanding of which of the patient’s health conditions are most likely to affect his future health and function is likely to be important.19–22
Patient-provider concordance has been evaluated previously in the setting of acute problems, with early studies finding that agreement about the nature of an acute problem and the need for follow-up is associated with earlier symptom resolution and retention in outpatient care.19–22 More recently, an emphasis on patient-centered care16,23–25 has renewed interest in patient-provider concordance as a reflection of effective communication and shared decision-making.26,27 Studies attempting to determine correlates of concordance have focused on whether there is agreement regarding the reason for the patient’s visit, the etiology of the patient’s symptoms, and the goals of treatment.28–31 Fewer studies have examined concordance regarding issues specific to chronic disease, and these studies tend to focus on a single health condition.30,32,33 In patients with multiple chronic conditions, little is known about the role of patient-provider concordance and the extent to which patients and providers agree on which of the patient’s conditions are most likely to affect the patient’s health and well-being.
In this study of diabetic patients presenting to clinic with an elevated triage blood pressure, we examined how often the health concerns prioritized by a patient match the conditions that his provider judges as most likely to affect that patient’s health outcomes. We hypothesized that patient-provider concordance would be lower when patients reported that their health was poor, that they were struggling with acute symptomatic conditions, or that they were facing more pressing issues in their lives than their health. On the other hand, we hypothesized that concordance would be higher when the patient rated his relationship with his provider as more favorable on domains such as communication and trust.
We conducted a prospective cohort study of patients with a scheduled primary care visit at nine Veterans Affairs facilities located in three Midwestern states. Details about the study design have been published previously.34 Institutional review boards of all participating facilities approved the study protocol, and all patients and providers gave written informed consent before participating.
One hundred twenty-six primary care providers were initially approached about participating in the study, and 104 consented. Of these, 12 providers changed their patient care responsibilities or left the facility prior to the start of data collection, so 92 providers were ultimately enrolled (median providers per facility was 8 with a range of 2 to 28). Diabetic patients of these providers were referred to study staff before a scheduled primary care visit if their lowest triage systolic blood pressure was 140 mmHg or greater, or if their lowest triage diastolic blood pressure was 90 mmHg or greater. Eligibility criteria required that the health care provider enrolled in the study was the patient’s primary diabetes care provider. Patients with impaired decision-making ability or a terminal disease, patients who did not speak English, and patients who were residents of nursing homes were excluded. Of 1,556 patients approached by study staff, 213 were ineligible, and 1,169 provided written informed consent. The mean number of enrolled patients per enrolled provider was 13 (range 1–16).
A total of 1,056 patients completed a survey with information about their sociodemographic characteristics, their health and competing demands, and their relationship with their primary care provider. All providers completed a baseline survey with questions about personal characteristics and then completed a second brief survey after a visit with an enrolled patient (completion rate, 99%). Both the patient survey and the provider brief survey included questions asking the participants to rank the three most important health conditions affecting the patient. Additional information such as patient age, prescribed medications, and comorbidities was obtained from Veterans Health Administration automated data sources.
The primary outcome was whether a patient and their provider agreed in their prioritization of the patient’s health conditions. Both the patient survey and the provider survey contained a list of nine health conditions that included both symptomatic chronic conditions (pain, depression, and breathing difficulties) and chronic conditions less likely to be acutely symptomatic (hypertension, diabetes, hyperlipidemia, heart disease or heart failure, obesity, and tobacco use) (Text Box 1). These conditions were chosen based on their prevalence among diabetic VA patients and were presented in a fixed order. The patient survey stated: “Please think about your current health and health concerns and indicate…which of the listed health concerns are your three most important concerns.” The provider survey stated: “From the following list of conditions and medical concerns, please indicate the top three most important medical concerns that are likely to affect health outcomes for this patient, whether or not these concerns were addressed in today’s visit.” Both surveys included space to rank two “other” health concerns. A concordance score was constructed using the number of matches between the patient’s ranked list and the provider’s ranked list, with an additional point given to the pair if the provider included the patient’s top concern in their list of three health concerns. The score was constructed in this way in order to capture the magnitude of agreement between patient and provider while also acknowledging the clinical importance of having a provider prioritize the patient’s most important concern.
Table 1 summarizes the patient and provider characteristics that we examined. Patients’ non-health competing demands were assessed using their response on a 5-point scale to the question, “I have more pressing issues in my life than my health.”5 Health status was assessed using a 5-point scale, which was later collapsed to excellent-to-good and fair-to-poor due to skewed data. Single-item measures of health status provide more global but less descriptive information than multi-item measures; however, studies have shown these measures to be reliable and valid, as well as predictive of mortality.35–38 Patient pain was assessed using a yes/no question in which patients were asked whether they have had pain present most of the time for 6 months or more during the past year.7 Depression was assessed using two 4-point questions drawn from Corson’s PHQ-2 measure 39 that queried about “little interest or pleasure in doing things” and “feeling down, depressed, or hopeless” over the last 2 weeks. Because of the recognized clinical association between pain and depression,40 these conditions were combined into a single variable in the main model and were then examined separately in a sensitivity analysis. Patient assessment of the patient-provider relationship was assessed through eight questions derived from the Primary Care Assessment Survey in the following domains: Trust in the provider, communication quality, interpersonal treatment, and knowledge of the patient.41 Factor analysis demonstrated that each pair of questions loaded onto a single factor; thus, the eight items were combined by summing them into a single scale that had a Cronbach’s alpha of 0.94. Primary analysis looked at this overall measure of the patient-provider relationship, and the four separate domains were evaluated in a sensitivity analysis. Other patient characteristics included age, education, income, partner status, and race. Patient chronic comorbidities were assessed using International Classification of Diseases, Ninth Revision (ICD-9) codes, as described by the Department of Veterans Affairs Health Economics Resource Center.42 Provider characteristics were obtained through the baseline provider survey and included gender and provider type (physician, nurse practitioner, or physician assistant). Time since the last primary care visit was assessed through review of electronic medical records.
We first examined patterns of concordance to determine how often patients and providers prioritized the same health condition, and how often a provider’s list included the patient’s top priority and vice versa. We then examined associations between the independent variables and patient-provider concordance using ordinal logistic regression. The proportional odds assumption for the dependent variable was tested using a likelihood ratio test and the Brant test.43 Standard errors were adjusted to account for within-provider clustering effects using the Huber-White sandwich estimator.44 Using the ordinal logistic model, we calculated cumulative probabilities of concordance (for a concordance score <3 and ≥3), holding other covariates at their mean value in order to present the effect of independent variables on concordance. Standard errors for the predicted probability of a concordance score were calculated using the delta method.45 Multiple imputation by chained equations was used to impute missing values in the dataset.46 All analyses were performed using Stata 10. (StataCorp 2007, Stata Statistical Software: Release 10. College Station, TX; StataCorp LP).
Of the 1,056 patients who completed a survey, 1,028 (97%) ranked their three most important health concerns. Of these, 24 patients (2%) were excluded from the final regression because they wrote in concerns but did not rank any of the nine health conditions listed in the survey. These patients’ responses are included in a summary of all write-in responses in the Appendix. Table 1 lists the characteristics of the remaining 1,004 patients and the 92 providers included in the final analysis. The mean patient age was 65 (SD 11), and the population was predominantly Caucasian (77%). The mean number of health conditions extracted from ICD-9 codes was 8 (SD 3).
For 714 (72%) pairs, the patient’s most important concern was ranked somewhere in the provider’s list of three conditions. Table 2 shows the frequency with which each health condition was ranked by patients as an important concern and by providers as a condition likely to affect the patient’s health outcomes. Overall, 45 (4%) patient-provider pairs did not have any matching health concerns, 251 (25%) pairs had 1 match, 551 (55%) pairs had 2 matches, and 157 (16%) pairs had 3 matches.
Both patients and providers ranked diabetes and hypertension most frequently in their list of the three most important health concerns for the patient (Table 2). However, although providers ranked hypertension as the most important health condition for 384 (38%) patients, only 184 (18%) patients listed hypertension as their most important health concern. Patients were more likely than providers to list “losing weight or being more active” in their top three concerns (35% vs. 21%, respectively). “Other health concern” was ranked among the top three concerns 51 times by patients and 305 times by providers. There were 26 and 326 write-in responses contributed by patients and providers, respectively (Appendix).
Three of the listed health conditions (pain, depression/anxiety, and breathing/pulmonary problems) were considered “symptomatic.” Patients were more likely than providers to rank these conditions as their most important concern and as one of their three most important concerns (Table 2). Even when a patient listed one of these conditions as their most important concern, these conditions were frequently not ranked in the top three by the patient’s provider. For example, of the 22 patients who listed “feeling less blue, down, or anxious” as their most important health concern, only two had a provider who ranked the corresponding “depression, anxiety, mood disorders” as one of the three conditions most likely to affect health outcomes for the patient. Similarly, of the 62 patients who listed “relieving pain or discomfort” as their most important health concern, only 20 had a provider who ranked “pain or physical discomfort” in their list of three conditions likely to affect the patient’s health outcomes.
Six hundred five (60%) patient-provider pairs had high concordance (concordance score ≥3), meaning that the patient’s provider ranked the same three health concerns as the patient, or that the provider ranked two of the same health concerns as the patient and that these two included the patient’s most important concern (Table (Table33).
Ordinal logistic regression revealed that the predicted probability of patient-provider concordance decreased significantly when a patient reported having non-health competing demands (46% vs. 62%, p<0.01) and when a patient reported having poor health status (55% vs. 64%, p<0.01). A nonsignificant trend toward decreased concordance was also seen in patients with both pain and depression, compared with those who had only mild depressive symptoms or neither pain nor depression (56% vs. 64%, p=0.12). In a sensitivity analysis, we examined a model with pain alone and a model with depression alone. Both models showed nonsignificant trends toward decreased concordance when either condition was present (58% vs. 63%, p=0.12 for pain; 57% vs. 62%, p=0.46 for depression), with the results for pain remaining unchanged when we adjusted the model for opioid use (results not shown).
The predicted probability of concordance was higher for patients with fewer health conditions (p<0.01). The only other patient characteristic that was associated with greater concordance was being married or in a relationship (p<0.01). Patients also had an increased likelihood of concordance with their provider if their provider was male (p<0.05). There was no significant association between our measure of the patient-provider relationship and patient-provider concordance. The four domains of this measure (trust, communication quality, interpersonal treatment, and knowledge of the patient) were examined separately in a sensitivity analysis, and no association between any of these domains and concordance was seen.
In this study of multimorbid diabetic patients with elevated blood pressure in primary care clinic, we found that patients’ most important health concerns frequently matched the conditions that their providers ranked as most likely to affect these patients’ health outcomes. The probability of concordance, however, was significantly lower for patients with poorer health status and those with non-health competing demands, even after controlling for the patient’s comorbidity count.
There are several possible explanations why patient-provider concordance is lower when patients have poor health or other non-health competing demands. One scenario is that this discordance reflects a breakdown in communication. For example, the provider may not fully recognize how the patient’s health conditions and non-health concerns are affecting his health and well-being, or the provider might not effectively communicate to the patient the morbidity and mortality risks of his conditions. This explanation, if true, would be troubling because several studies have demonstrated an association between effective patient-provider communication and a multitude of positive patient-centered and clinical outcomes.4,15–17 However, when we adjusted our model for the patient’s assessment of the quality of the patient-provider relationship, the association between our main independent variables and concordance remained unchanged.
A second scenario is that patients with poor health or non-health competing demands may be more likely to face functional limitations, financial stress, and other barriers to care, and are therefore more likely to prioritize a symptomatic condition that is exacerbating these existing challenges. Providers of these patients, meanwhile, may focus on the long-term health consequences of asymptomatic hypertension or uncontrolled diabetes. While both sets of priorities are valid, previous studies suggest that poor health status and non-health competing demands may interfere with self-management of diabetes as well as productive clinic-based management decisions and processes.5,6,30,47 If patients and providers disagree on the importance they ascribe to a patient’s different health conditions, this discordance may intensify barriers to self-management and clinical decision-making for that patient.
We observed several notable patterns in the patient- and provider-ranked lists. While diabetes and hypertension were ranked highly by the majority of patients and providers, providers were far more likely to rank hypertension as the most important health concern for the patient (38% vs. 18%). This is consistent with previous findings that many diabetic patients are unaware of the importance of blood pressure control despite evidence and guidelines that emphasize the critical importance of this issue.48–52 Patients, in contrast, were modestly but consistently more likely than providers to rank weight loss and symptomatic concerns such as pain, depression, and breathing problems among their top three concerns. Strikingly, very few of the patients who listed pain or depression as their top health concern had a provider who ranked these conditions as likely to affect the patient’s health outcomes (9% and 32%, respectively). This discordance is concerning, not only because it raises the possibility that providers are unaware of the extent to which these conditions affect their patients, but also because pain and depression can be barriers to effective diabetes self-management,7,8 and (in the case of depression) may worsen glycemic control and increase the risk of mortality.53,54 Therefore, it is plausible that by deemphasizing symptomatic conditions, providers are actually neglecting some of the most important medical concerns that are likely to affect health outcomes in these patients.
There are several limitations to this study that should be noted. First, we set out to measure the degree to which the patient and provider had a shared set of priorities about the most important problems facing the patient. Thus, our concordance score does not explicitly measure the somewhat different concept of which conditions providers thought the patient would have prioritized. This latter concept is also of interest and merits additional research. Second, all enrolled patients and providers were aware that this was a study of diabetic patients, and patients and providers were filling out their surveys after the patient had an elevated blood pressure in triage. Not surprisingly, a majority of both patients and providers ranked diabetes and hypertension among the top three health conditions affecting the patient. This limited our ability to fully evaluate concordance patterns among other health conditions. Finally, patients and providers ranked “other health concern” in 5% and 30% of instances, respectively. Write-in responses were assessed qualitatively to better understand the health conditions that most frequently take a patient and provider’s time and attention away from diabetes and hypertension. These responses, summarized in the Appendix, suggest that the most common other concerns listed by providers were unlisted chronic conditions such as renal disease and cancer, and issues related to medication adherence. Only 26 write-in responses were contributed by patients (vs. 326 by providers), so we were unable to assess concordance using these data. Of the 26 write-in responses from patients, only two (cancer and memory loss) matched write-in responses of the corresponding provider. Therefore, for the purposes of our multivariate analysis, we only used the listed nine conditions to determine patient-provider matches. While this may have inflated the frequency of concordance because participants were constrained by a limited number of choices, it likely did not influence our other results because we adjusted our analyses for total number of conditions.
While previous work has focused extensively on patient-provider concordance regarding a patient’s presenting complaint,28,29,31 to our knowledge this is the first study to evaluate patient-provider concordance regarding priority given to the chronic health conditions of multimorbid patients. Although we did not assess the influence of this concordance on patient-centered and clinical outcomes, previous studies in the setting of acute conditions have found an association between higher patient-provider concordance and symptom resolution, improvement in mental health and function, and retention in outpatient care.19,20,55 Our finding that patients with poor health status are less likely to share priorities with their provider is thus concerning, as poor health status has been associated with increased risk of mortality,35,38 and this population is therefore one in which effective communication and shared understanding of priorities are likely to be critical in developing goals of care and treatment strategies.56
In conclusion, diabetes patients with multimorbidity and their primary care providers often agree on the most important health conditions affecting these patients’ health. Our findings, however, reinforce the need for heightened provider recognition of patients’ symptomatic conditions as well as their non-health competing demands. Fortunately, there is growing evidence that interventions can increase provider awareness about patient concerns and priorities,57–59 and that patient-centered approaches can improve diabetes self-management in the face of multimorbidity and other competing demands.60,61 Future research should focus on how best to encourage and implement these practices in primary care in order to optimize chronic disease management in this vulnerable population.
The authors thank Rob Holleman for assisting with data management, Mandi Klamerus for project management, and Shirley Chen for helping with manuscript preparation.
Financial Disclosures None reported.
Funding/Support This work was supported by the Robert Wood Johnson Clinical Scholars Program and an associated VA Advanced Fellowship, as well as research grants from the US Department of Veterans Affairs Health Services Research and Development Service (IIR02-225) and the Michigan Diabetes Research and Training Center Grant (P60DK-20572). Dr. Zikmund-Fisher is supported by a career development award from the American Cancer Society (MRSG-06-130-01-CPPB).
Role of the Sponsor The funding sources had no role in the design and conduct of the study, in the collection, analysis, and interpretation of the data, or in the preparation, review, or approval of the manuscript.
Conflict of Interest None disclosed.
Affiliation Changes Following Manuscript Submission: Dr. Zikmund-Fisher is now at the Department of Health Behavior and Health Education, University of Michigan.