To examine associations between active patient–physician communication and measures of morbidity in patients with systemic lupus erythematosus (SLE).
To examine associations between active patient–physician communication and measures of morbidity in patients with systemic lupus erythematosus (SLE).
Audiotapes of routine visits between 79 women with SLE and their rheumatologists were coded for active patient participation and the degree of patient-centered communication of the physician, using a validated coding scheme. Measures of SLE activity, functional disability, and permanent organ damage were recorded at the same visit. Permanent organ damage was reassessed in 68 patients after a median of 4.7 years.
Patients who participated more actively in their visits had less permanent organ damage, as measured by the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index, and tended to accrue less organ damage over time. There were no associations between either active patient participation or physicians' patient-centered communication scores and measures of SLE activity or functional disability.
Patients with SLE who participated more actively in their visits had less permanent organ damage, suggesting that involving patients more in their care may decrease morbidity.
Communication between the patient and physician is central to medical care (1). This communication includes the exchange of information about symptoms, diagnoses, prognoses, and treatment options. Effective communication can not only improve patients' knowledge about their illness, but also can enlist them to be partners in their care, improve adherence to treatment, and improve satisfaction with care (1–5). The style of patient–physician communication may also affect health outcomes. In some studies, patients who participated more actively in their visits by asking questions, expressing opinions, and participating in decision making had better outcomes as measured by resolution of symptoms, better self rated health, fewer functional limitations, and improved control of hypertension and diabetes mellitus (3, 6–8). Educational interventions to increase patients' participation in their visits with physicians have also been efficacious (9–11). However, not all studies have found an association between patient–physician interactions that were more participatory, and health outcomes (12–15).
Participatory patient–physician communication is influenced by characteristics of the patient and physician, the length of their relationship, and the reason for the visit, among other factors (16–19). Patients of lower socioeconomic status (SES) are less likely to engage in active communication than those of higher SES (16–18, 20–23). It has been proposed that differences in participatory patient–physician communication could, in part, explain socioeconomic differences in health (21), but this hypothesis has not been tested. Previous studies of health disparities have focused on psychological, social, and environmental factors that might influence health, on differences in access to medical care, or on differences in the technical application of medical care (24–27). Investigation of the potential links between interpersonal aspects of medical care and health disparities is just beginning (28).
We sought to determine if interpersonal aspects of medical care were associated with morbidity in patients with systemic lupus erythematosis (SLE). Specifically, we examined whether patients who participated more actively in their visits with physicians had less active SLE, less functional disability, and less permanent organ damage. Because SES has an important impact on health outcomes in SLE (29–35), we also examined the association between SES and participatory communication styles to determine if differences in communication style explain the association between SES and morbidity in patients with SLE.
Participants were recruited from 4 rheumatology clinics in the San Francisco area: Santa Clara Valley Medical Center (a public hospital in San Jose, CA), Alameda County Medical Center (a public hospital in Oakland, CA), a private rheumatology practice in Alameda county, CA, and Stanford University Medical Center. Participants were selected from the patients currently receiving care at these clinics. Inclusion criteria were fulfillment of 4 or more criteria for the classification of SLE (36), female sex, age 18 years or older, English literacy, and absence of cognitive impairment that would preclude completion of a questionnaire. In addition, participants were required to be ongoing patients in the clinics and to have a scheduled appointment during the study period (August 1995 to August 1996). The study was limited to women because communication styles differ between women and men (1, 18, 21) and because SLE most commonly affects women. Of 139 eligible patients contacted, 100 women (72%) participated (35, 37). The study protocol was approved by the institutional review boards of each study site, and all participants provided written informed consent.
All patients were met at a scheduled followup appointment with their rheumatologist. Patients completed a questionnaire that asked for information on demographic characteristics, functional disability, and depression. At this visit, the patient's rheumatologist scored measures of SLE activity (the Systemic Lupus Activity Measure [SLAM] and the SLE Disease Activity Index [SLEDAI]) and permanent organ damage (the Systemic Lupus International Collaborating Clinics/American College of Rheumatology [SLICC/ACR] Damage Index) based on the medical history, physical examination, and laboratory studies obtained at the visit (38–40).
The visits were also audiotaped using a portable tape recorder placed on a desk or table in the examining room. Patients and physicians were aware the visit was being audiotaped. We attempted to audiotape every patient, but the visits of 12 patients were not taped because of logistic problems. The audiotapes of 9 additional patients could not be coded or analyzed (3 tapes were lost, 4 were poorly audible, 1 interview included both the patient and her husband, and 1 interview was conducted in Spanish). Therefore, 79 patients were included in this study. There were no differences in demographic or clinical characteristics between the 79 patients included in this study and the 21 patients not included.
In 2000–2001, study investigators contacted participants to learn if they had developed any new permanent organ damage since the initial evaluation. SLICC/ACR Damage Indexes were scored based on medical record review by investigators unaware of the results of the audiotape analysis. For participants who were lost to followup or who had died, the last available records were used. Of the 79 patients, 68 (86%) had SLICC/ACR Damage Indexes measured at least 1 year after the initial evaluation. We did not measure SLE activity or functional disability at the followup assessment.
Audiotapes were coded to obtain measures of patient participation and physicians' facilitation of patient involvement, using a method developed by Street et al (23, 41–43). Verbal acts of patient participation and physician patient-centered communication occur infrequently (less than 10% of patient and physician utterances) compared with other speech acts (e.g., information-giving by the patient) (1). Thus, to code behaviors of interest, it was not necessary to transcribe and code all talk. Trained coders first listened to an audiotape of the visit. Each time the coder perceived that a particular behavior occurred (e.g., the patient asked a question or the doctor engaged in partnership building), that portion of the dialogue was transcribed, along with several conversational turns before and after the identified event. Next, the coders divided the discourse into “utterances.” Utterances are the oral analogs of a simple sentence and may be in the form of a complete sentence, independent clause, nonrestrictive dependent clause, or evaluation (44). The coder then listened to that part of the tape again, following the transcribed segment, and placed the utterances into the relevant categories of patient participation and physicians' patient-centered communication. To establish the reliability of coding, transcripts of some audiotapes were created so that assessment could be made of coder agreement on identifying a speech act for coding and on categorizing utterances into particular categories (43).
Three types of patient communication behaviors were coded: question asking, assertive responses, and expressions of concern. Question asking included utterances in interrogative form intended to seek information. Assertive responses represented utterances in which the patient made recommendations, stated a preference for treatment, made a request, disagreed with the doctor, or offered an opinion. Expressions of concern included statements in which the patient expressed worry, anxiety, anger, and other negative emotions (e.g., frustration). Expressions of concern may be signaled by word choice (e.g., “worried,” “concerned,” “upset”) or by an emotional tone of voice. Patients' communicative acts were quantified by counting frequencies of question asking, assertiveness, and expressions of concern, and adding these to create a composite measure of active patient participation.
Two types of speech acts were coded as representing physicians' efforts to facilitate patient involvement: partnership building and supportive talk. Partnership building refers to responses that encourage patients to express their opinions, talk about their feelings, ask questions, and participate in decision making (1, 41, 42). Partnership building also includes talk that affirms and accommodates the patients' preferences. Supportive talk includes statements of reassurance, support, empathy, and other verbal displays of sensitivity (43). Physicians' partnership building and supportive talk were tallied as counts, and added to create an index of patient-centered communication.
To establish reliability, each of the 2 coders initially coded the same subset of 25 interviews that had been completely transcribed. Reliability (Cohen's kappa) was acceptable for both the patients' (questions = 0.92, assertive utterances = 0.78, expressions of concern = 0.82) and physicians' (partnership-building = 0.83, supportive talk = 0.79) communication. The remaining interactions were coded individually and independently. The coders were unaware of the study hypotheses and of patients' sociodemographic and clinical characteristics.
Four measures of morbidity were studied, including 2 alternative measures of current SLE activity, the SLAM and SLEDAI. The SLAM scores the presence and severity of 24 clinical manifestations and 8 laboratory measures of SLE activity (38). The SLEDAI assesses 16 clinical manifestations and 8 laboratory measures (39). Higher scores on each index indicate more active SLE. Both the SLAM and SLEDAI have been shown to be valid measures of SLE activity (45–49). The SLICC/ACR Damage Index is a measure of cumulative organ damage (40). Scores for damage in each organ system are added, so that possible scores of the Damage Index range from 0 (no damage) to 46 (severe damage). Higher scores predict mortality (50). The questionnaire completed by participants included the Health Assessment Questionnaire (HAQ) disability index, a 20-item self-rated measure of functional disability that asks respondents to report the degree of difficulty they have in performing daily activities in 8 functional areas (51, 52). Possible scores range from 0 (no difficulty in all areas) to 3.0 (inability to perform tasks in each area).
Demographic information and information on psychosocial factors that might be associated with active participation by patients in their visit was obtained by questionnaire. Socioeconomic status was measured using either the number of years of formal education or the Hollingshead Two-Factor Index of social position, which combines scores for education level and occupational prestige (53). The Hollingshead Index can range from 11 to 77, with lower scores indicating higher socioeconomic status. Depression was measured using the Center for Epidemiologic Studies Depression Scale (CES-D), a 20-item scale that asks about the frequency of depressive thoughts and feelings (54).
Two analyses were performed: a correlational analysis between the communication measures and the 4 measures of morbidity using the cross-sectional data obtained at the initial study visit, and a longitudinal analysis of whether the communication measures obtained at the initial study visit predicted future changes in the SLICC/ACR Damage Index. Univariate associations between patient characteristics and the 2 communication measures, active patient participation and physician patient-centered communication, were tested using Spearman rank correlations (for continuous measures) and Wilcoxon's rank sum tests (for dichotomous measures). Spearman rank correlations were also used to test the associations between each communication measure and each measure of morbidity. Multivariate associations between active patient participation and each morbidity measure were tested using regression models that included patient age, ethnicity, and education level as covariates. Linear models were specified when the SLAM, SLEDAI, and HAQ disability index (log transformed to normalize the distribution of the data) were the dependent variables. Poisson regression models were specified when the SLICC/ACR Damage Index and the change in SLICC/ACR Damage Index over time were the dependent variables. Adjustments for the duration of SLE (for the cross-sectional analyses) or the interval between SLICC/ACR Damage Index assessments (for the predictive analysis) were incorporated in these models. These analyses were then repeated using the physician's patient-centered communication score as the independent variable of interest. We used generalized estimating equations for all models to adjust the standard errors of the regression coefficients for the between-patient correlations that arise due to the clustering of patients interviewed by the same physician (55, 56). All analyses were performed using SAS programs (Statistical Analysis Systems, Cary, NC).
The 79 women in this study were diverse in age, ethnicity, socioeconomic status, and duration of SLE (Tables 1 and 2). Scores on the SLAM and SLEDAI indicated moderately active SLE at the study visit, and scores on the HAQ disability index indicated mild functional disability. Fifty patients (63%) had a SLICC/ACR Damage Index of 1 or higher, indicating some permanent organ damage. Patients had visits with 1 of 7 attending rheumatologists (2 women, 5 men; who interviewed 1, 5, 5, 7, 7, 11, and 37 patients, respectively) or 1 of 4 rheumatology fellows (2 women, 2 men; who interviewed 1, 1, 1, and 3 patients, respectively).
|Mean ± SD||Patient active participation score||Physician patient-centered communication score|
|Age, years||43.7 ± 11.0||0.18||0.11||0.00||0.99|
|Duration of SLE, years||9.3 ± 8.0||0.24||0.03||0.07||0.51|
|Education level, years||14.2 ± 2.4||−0.08||0.49||−0.10||0.37|
|Hollingshead Index†||40.3 ± 17.0||0.00||0.99||0.05||0.69|
|Depression (CES-D)‡||19.6 ± 11.2||−0.07||0.52||−0.06||0.63|
|n (%)||Median patient active participation score||P||Median physician patient-centered communication score||P|
|African American||14 (18)||11||3.5|
|Not married||47 (59)||11||3|
|Not employed||53 (67)||10||3|
The median patient active participation score was 10 (range 1–60; mean 12.3). The median number of questions asked was 2 (range 0–34), assertive statements was 5 (range 0–19), and expressions of concern was 1 (range 0–12). The median physician's patient-centered communication score was 3 (range 0–17; mean 4.2). The median number of partnership-building statements was 3 (range 0–14), and supportive statements was 0 (range 0–10). Patients' and physicians' scores were comparable to scores in other studies using this coding scheme (17, 23, 41, 43). There were differences among rheumatologists in their patients' active participation scores (P = 0.03) and in their own patient-centered communication scores (P = 0.05).
Active participation scores were not associated with patient age, but patients with longer durations of SLE participated more actively in their visits than patients with shorter durations (Table 1). Active participation scores were not associated with SES, as measured by either education level or the Hollingshead Index, or with depression, as measured by the CES-D. Median scores were also similar between those with 12 years of education or less, and those with more than 12 years of education (10 and 10, respectively). Active participation scores were lower and reflected fewer questions and assertive statements by Asians than other ethnic groups (P = 0.003) (Table 2). Physicians' patient-centered communication scores did not vary with patients' demographic characteristics (Tables 1 and 2), but were highly correlated with patients' active participation scores (r = 0.57, P < 0.0001).
In univariate analyses, patients who had higher SLAM scores tended to participate more actively in their visits, and physicians exhibited more patient-centered communication with patients who had more active SLE (Table 3). However, these associations were confounded by ethnicity, as Asian patients had less active SLE than patients of other ethnicities (median SLAM 3 versus 7, P = .0004; median SLEDAI 0 versus 8, P = 0.02) and Asian patients also participated less actively in their visits. Among non-Asians, there were no associations between active patient participation scores and either the SLAM (r = 0.06, P = 0.60) or SLEDAI (r = 0.05, P = 0.70), or between physician patient-centered communication scores and the SLAM (r = 0.13, P = 0.29). There was a weak positive correlation between physician patient-centered communication scores and the SLEDAI among non-Asians (r = 0.22, P = 0.07). Scores on the SLICC/ACR Damage Index and the HAQ disability index did not differ between Asians and non-Asians. There were no associations between communication scores and the HAQ disability index.
|Median (range)||Patient active participation score||Physician patient-centered communication score|
|HAQ disability index||0.75 (0–2.625)||0.19||0.09||0.08||0.47|
|SLICC/ACR Damage Index||1 (0–6)|
|Duration <3.0 years (n = 21)||−0.18||0.45||−0.07||0.77|
|3.0–7.9 years (n = 20)||0.11||0.66||−0.07||0.78|
|8.0–14.9 years (n = 21)||−0.36||0.11||−0.14||0.54|
|≥15.0 years (n = 17)||−0.44||0.08||−0.12||0.66|
Because organ damage accumulates over time, associations with measures of damage must be adjusted for the duration of SLE. Therefore, we analyzed associations with the SLICC/ACR Damage Index among patients stratified by quartile of duration of SLE. There were no associations between active patient participation and the Damage Index among those with less than 3 years of SLE or with 3–7.9 years of SLE. However, moderately strong associations were present among those with 8–14.9 years of SLE (r = –0.36) and in those with 15 years or more of SLE (r = –0.44; Table 3). In these subgroups, patients who participated more actively in their visits had lower SLICC/ACR Damage Indexes than those who participated less actively, but the small size of these subgroups limited the statistical power of this analysis. To determine if this association reflected only the patients of 1 practice, we further stratified those with 8 years of SLE or more into 2 groups: those of the physician who contributed 37 patients, and those of all other physicians. The correlation between patient active participation scores and the SLICC/ACR Damage Index was slightly higher in the multiphysician subgroup (r = –0.47, P = 0.04) than in the single-physician subgroup (r = –0.39; P = 0.11), indicating that this association was not specifically associated with 1 physician's practice. There were no associations between physicians' patient-centered scores and the Damage Index.
In multivariate analyses that adjusted for patient age, ethnicity, education level, and duration of SLE, there were no associations between active patient participation scores or physicians' patient-centered communication scores and the SLAM, SLEDAI, or HAQ disability index (Table 4). For example, SLAM scores increased on average 0.002 points with each additional point on the active patient participation score. In contrast, active patient participation was associated with lower SLICC/ACR Damage Indexes. For each 1-point increase in the active patient participation score, the Damage Index decreased on average by 7% (odds ratio [OR] for each 1-point increase in patient active participation score 0.93; 95% confidence interval [95% CI] 0.91, 0.94; P < 0.0001). This association remained after additional adjustment for depression by including the CES-D in the model (OR for each 1-point increase in patient active participation score 0.96; 95% CI 0.94, 0.98; P < 0.0001).
|Model||Dependent variable||Independent variables of interest||Point estimate†||95% CI||P|
|1||SLAM||Active patient participation||0.002||−0.04, 0.05||0.93|
|Education level||−0.21||−0.45, 0.02||0.08|
|2||SLAM||Patient-centered communication||−0.05||−0.29, 0.18||0.65|
|Education level||−0.22||−0.48, 0.04||0.09|
|3||SLEDAI||Active patient participation||0.046||−0.03, 0.13||0.24|
|Education level||−0.36||−0.58, −0.13||0.002|
|4||SLEDAI||Patient-centered communication||0.039||−0.76, 0.85||0.93|
|Education level||−0.34||−0.46, −0.22||<0.0001|
|5||HAQ disability index||Active patient participation||0.005||−0.002, 0.02||0.20|
|Education level||−0.016||−0.04, 0.007||0.17|
|6||HAQ disability index||Patient-centered communication||0.003||−0.03, 0.04||0.83|
|Education level||−0.015||−0.04, 0.007||0.18|
|7||SLICC/ACR Damage Index||Active patient participation||0.93||0.91, 0.94||<0.0001|
|Education level||0.86||0.77, 0.96||0.004|
|8||SLICC/ACR Damage Index||Patient-centered communication||0.98||0.94, 1.03||0.46|
|Education level||0.90||0.85, 0.96||0.0006|
Followup SLICC/ACR Damage Indexes were obtained for 68 patients at a median of 4.7 years after their initial evaluation (range 1.3–5.5 years). Thirty-eight patients (56%) had an increase in the Damage Index over this time (24 with a 1-point increase, 4 with a 2-point increase, 8 with a 3-point increase, and 2 with a 4-point increase). In multivariate analyses that adjusted for age, ethnicity, education level, and the length of time between assessments, more active patient participation at baseline and more patient-centered communication by physicians at baseline were associated with a lower likelihood that the SLICC/ACR Damage Index would increase over time, but these associations were not statistically significant (Table 5).
|Model||Dependent variable||Independent variables of interest||Odds ratio||95% CI||P|
|1||Change in Damage Index||Active patient participation||0.98||0.95, 1.01||0.15|
|Education level||0.91||0.88, 0.94||<0.0001|
|2||Change in Damage Index||Patient-centered communication||0.95||0.88, 1.02||0.13|
|Education level||0.91||0.88, 0.94||<0.0001|
Patients with higher levels of education had lower scores on the SLAM, SLEDAI, and SLICC/ACR Damage Index at the initial study visit (Table 4), and were less likely to have an increase in the SLICC/ACR Damage Index over time (Table 5). There were no differences in the associations between the education level and the morbidity measures in models that included or excluded the patients' active participation scores or physicians' patient-centered communication scores, indicating that the communication measures did not mediate the association between SES and morbidity. Results were similar when the Hollingshead Index was used as the measure of SES.
In this study, women with SLE who participated more actively in their visits with physicians had less cumulative organ damage due to SLE. With each additional 1-point increase in the patient active communication score, the average SLICC/ACR Damage Index decreased by 7%. Based on these results, a woman who expressed 9 questions, concerns, or assertions during her visit would be expected to have, on average, a SLICC/ACR Damage Index that was half that of a women who asked no questions, voiced no concerns, and made no assertions. These differences might pertain if patients' typical participatory styles were captured in this study.
Because these results were based on a cross-sectional analysis, it is not possible to know the direction of any causal association between patients' active participation and cumulative organ damage. Patients who participate more actively in their visits may develop fewer complications as a result of their communication style. They may obtain more information, express their concerns more fully, engage more in problem solving with their physicians, make their preferences better known, adhere more to treatment, assume more responsibility for and control of their health, develop greater confidence in their ability to influence their health, receive more psychosocial support from their physicians, be more satisfied with their care, and have better continuity of care (3, 4, 57, 58). Alternatively, the presence of more complications of SLE may cause patients who have more organ damage to interact less in their visits. Two aspects of our analysis suggest that active participation may lead to less organ damage, rather than the converse. First, the longitudinal component of the study indicated that the degree of active participation by patients might predict the likelihood of future organ damage. Patients who participated more actively in their visits tended to develop organ damage less frequently over the next 4.7 years than patients who participated less actively. Although this association was not statistically significant, the direction of the association supports the possibility that active communication affects the development of future complications. This finding is consistent with intervention studies that demonstrate that changes in participatory behavior can lead to better health outcomes (9–11). Second, the association between patients' active participation scores and the SLICC/ACR Damage Index persisted after adjustment for depression, suggesting that this association was not confounded by the patient's mood.
Permanent organ damage in SLE most often results from persistently active inflammation or as consequences of medications used to treat active inflammation, yet patients' active participation scores were not associated with either of 2 measures of SLE activity. Active participation scores were also not associated with the degree of functional disability among these patients. Absence of these associations may be due to the variable nature of these measures in a disease marked by exacerbations and remissions, by the stability or novelty of the level of SLE activity at the study visit, or by different reactions to symptoms by patients and physicians. Health benefits from active participation may be more apparent for cumulative outcomes of chronic illnesses, such as permanent organ damage, if patients' communication styles remain generally stable over time.
Although patients of higher SES had less active SLE and less permanent organ damage, there were no associations between measures of SES and measures of participatory communication. Therefore, differences in participatory communication style could not explain the socioeconomic differences in morbidity in this sample. It is possible that the relatively high education level of these patients limited our ability to detect an association between SES and communication style. We used the Hollingshead Index to better differentiate SES among patients of similar education levels, but we also found no association between the Hollingshead Index and patients' active participation scores. There may be a threshold level of education above which there is little or no association with participatory communication style. Several studies that demonstrated associations between education level or social class and patient-physician communication included patients with a wider range of SES (18, 20, 22). Even though we found no association between SES and active communication, the SES differences among our patients were associated with differences in morbidity. It is not clear if the association between SES and participatory communication style differs between women and men. Women tend to be more active participants than men (18, 21, 59–61), and 1 study suggests that women provide more information to their physicians than do men of the same social class (20). Characteristics such as SES may also be less important an influence on patient–physician communication among patients who have longstanding relationships with a given physician (62). Of note, Asian patients were less participatory, possibly due to cultural beliefs (63).
We did not find associations between the physicians' degree of patient-centered communication and measures of morbidity. This may be due to the low levels of patient-centered communication expressed by these physicians, particularly of supportive talk. Fifty-eight percent of visits included statements of reassurance or support. Partnership building statements were somewhat more common, but active communication in this study was most often initiated by the patients. Because all patients in this study had ongoing relationships with these physicians, they may have learned and adapted to the style of interaction preferred by the physician, or selected a physician who matched their communication style preference. These patients may have needed less prompting or may have felt freer to ask questions or express concerns. This possibility is supported by the high correlation between patients' and physicians' communication scores, a finding that also suggests that physicians' communication styles may influence patients' participation in their visits (15, 17).
The strengths of this study include a diverse patient sample, use of several measures of morbidity, use of an established and reliable communication coding process, blinding of the investigators and coders to relevant patient information, and longitudinal followup for assessment of the development of new organ damage. We also studied only patients who had an established relationship with their physician, to reduce variation in the reason for the visit and the tenure of the patient–physician relationship. However, the study also has some limitations. Patients were from one geographic area, only a limited number of patients and physicians were included, and all subjects were women. Study of more physicians may have provided a wider range of communication styles. Almost one-half of patients were of 1 physician, but the findings of these patients were similar to those of patients of the other physicians. We examined only 1 visit of each patient, and therefore may not have captured the patient's typical communication style. Variability in this measure and in the scoring of the activity and damage measures by different physicians may have limited our ability to detect associations. We also studied patients' interactions with their rheumatologists, because these were thought to be most relevant for the outcomes of interest. Patients' interactions with their primary care physicians or other specialists may differ from those with their rheumatologists, and these interactions may also affect morbidity.
We did not find evidence to suggest that differences in communication styles mediated socioeconomic differences in morbidity. However, our findings indicate that patients who participated more actively in their visits with physicians had less permanent organ damage. This association suggests that encouraging active participation and involving patients more in their care may decrease morbidity. Educational interventions that specifically target communication skills may be an effective way to decrease organ damage in patients with SLE (9–11, 64).
We thank Cheryl Kallmann for expert technical assistance.