Effective clinical communication, characterized by providers' use of specific communication skills that engage the patient in active discussion, is fundamental to achieving positive patient health outcomes. However, studies have shown that providers tend to use these skills less often with racial/ethnic minority patients, constituting a potential root cause of observed disparities in VA healthcare. In our prior study of patient feedback from over 200 VA- and community-based primary care patients who reviewed a multimedia (audio-video) recording of a visit with their provider, we identified 4 provider clinical communication skills most often considered by patients as important in a conversation: "Agenda-Setting," "Paraphrasing," "Eliciting the Patient's Perspective," and "Strategic Pausing." Our prior work included samples of African American, Hispanic/Latino, and white patients, and these 4 communication skills appealed to patients across race/ethnicity groups. For this project, we developed a training intervention for providers focused on the 4 communication skills. We evaluated our intervention with providers and patients in 2 distinct VA primary care settings.
Specific Aim 1: To demonstrate the efficacy of a training intervention involving 4 clinical communication skills to improve providers' use of these skills and reduce variability in use of these skills with white versus racial/ethnic minority patients.
Specific Aim 2: To demonstrate the impact of providers' use of these clinical communication skills on patient outcomes, overall and across race/ethnicity.
We created a script to demonstrate each of the 4 provider clinical communication skills in action, and we produced a training video for each skill. We created an educational intervention for providers composed of 4 sequential training sessions, each consisting of showing the training video for a skill, engaging the providers in group discussion, and formulating "take-away" notes to enhance providers' recall of the skill and to encourage its use. After each training session, the video and notes for that skill were posted as links on the desktop of each provider's office computer and were made available on-demand to the provider throughout the study.
Provider Inclusion Criterion: Full-time primary care provider who followed patients (n = 16).
We conducted a longitudinal observational study of our intervention at 2 VA primary care clinics, Site A (Houston, TX) and Site B (New Orleans, LA). We completed the intervention at Site A prior to commencing at Site B. At both sites, providers attended each of the 4 training sessions, presented at 3-month intervals at Site A and 1-2 month intervals at Site B.
Patient Inclusion Criterion: Any patient in the regular panel of a participating provider who was a returning patient for that provider.
Patient Exclusion Criteria: Legal blindness, dementia, severe post-traumatic stress disorder, schizophrenic disorder, Alzheimer's disease, bipolar disorder.
Conversations of patients with their provider were audio recorded. Each patient completed scales measuring the patient's perceptions of their provider and the VA as well as a demographic questionnaire. There were 5 measurement periods for the study: baseline, post-session 1, post-session 2, post-session 3, and final. The target enrollment was 28 patients per provider, or 8 at baseline, 4 in each post-session period, and 8 in the final period. The audio recordings were analyzed, and frequency and duration of provider use of each skill during the conversation were tabulated. We examined these data, patients' scale scores, and behavioral outcomes over the 6-month period following each patient's enrollment, across all patients and by patient race/ethnicity, gender, and patient-provider race/ethnicity and gender concordance. We also administered a brief questionnaire to providers 1-2 weeks after each training session, and we recorded their comments on the study during the training sessions.
Mean per-patient frequency of use of each skill from all conversations per measurement period.
Mean per-patient duration of use of each skill from all conversations per measurement period, measured as 1 unit = 1.2 sec, or 1/1,000th of a typical 20-min VA primary care visit.
Trust in Provider Scale: 5 items, "0" = "strongly disagree" to "10" = "strongly agree."
Trust in VA Health System Scale: 4 items, "0" = "strongly disagree" to "10" = "strongly agree."
Provider's Propensity to Share Decision-Making Scale: 5 items, "0" = "strongly disagree" to "10" = "strongly agree."
Provider's Supportiveness Scale: 5 items, "0" = "strongly disagree" to "10" = "strongly agree."
Provider's Sensitivity Scale: 3 items, "0" = "strongly disagree" to "10" = "strongly agree."
Provider's Similarity Scale: 4 items, "1" = "very similar" to "5" = "very different."
Patient Demographics: age, gender, years of education completed, race/ethnicity.
Provider Demographics: age, gender, years of residency, years in practice, race/ethnicity.
Physician-Patient Racial/Ethnic Concordance - Coded as 1 = match, 0 = mismatch.
Physician-Patient Gender Concordance - Coded as 1 = match, 0 = mismatch.
Proportion of patient's appointments kept to total appointments scheduled during the 6-month period following enrollment.
Unscheduled patient visits to VA during the 6-month period following enrollment (Yes/No).
Proportion of patient's VA pharmacy prescriptions filled to total VA prescriptions during the 6-month period following enrollment.
Length of patient-provider relationship in months prior to enrollment.
Provider post-session feedback questions: 2 items, response options from lowest to highest.
Sample descriptive statistics: The sample consisted of 18 providers and 369 patients:
7 providers (39%) were white, 6 (33%) African/African American, 1 (6%) Hispanic/Latino, and 4 (22%) of other race/ethnicity, 7 (39%) women, and 11 (61%) men; and 129 patients (35%) were white, 207 (56%) African American, 30 (8%) Hispanic/Latino, 3 (1%) of other race/ethnicity, 31 (8%) women and 338 (92%) men. There were 58 (16%) patient-provider pairs concordant on white race/ethnicity, 67 (18%) on African/African American race/ethnicity, 1 on Hispanic/Latino race/ethnicity, none on other race/ethnicity, 23 (6%) pairs concordant on female gender, and 202 (55%) on male gender.
Skills use metrics: Mean differences in provider skill use and duration across consecutive measurement periods were not statistically significant; therefore, we aggregated post-baseline measurements and compared the aggregate to baseline. For all 4 skills jointly, 12 of 18 providers (67%) exhibited increases in mean use and duration per patient compared to baseline; 7 of 12 (58%) increases were statistically significant (p < .05). Agenda-Setting was the most used skill across the patient sample, with a count of 174 uses and duration of 447 units (8 min, 55 sec), and per-patient use ranged from 1 to 36 times and duration from 5 to 96 units. Mean use per patient increased 0.3, from 0.3 to 0.6, from baseline to post-baseline, and 6 of 18 providers (33%) showed increases in mean use and duration; these increases were statistically significant. Paraphrasing was used 26 times with duration of 104 units (2 min, 6 sec); Eliciting the Patient's Perspective was used 11 times with duration of 45 units (54 sec); and Strategic Pausing was used 5 times with duration of 21 units (25 sec). Given the scarcity of use of these skills across the patient sample, no further analyses of these were performed.
Agenda-Setting use and duration increased from baseline to post-baseline for Hispanic/Latino patients. The difference in mean use and duration between whites and all other patient race/ethnicity groups decreased from baseline to post-baseline; however, this decrease was not statistically significant. With women, Agenda-Setting was used 6 times with duration of 11 units (14 sec), and no other skills were used. Given the scarcity of use of skills in the women sample, no gender-wise analyses were performed. Both male-concordant and male-discordant (male patient, female provider) pairs showed increases in use and duration of Agenda-Setting and all 4 skills jointly; increases were larger for male-discordant than male-concordant pairs.
Scale scores: Score distributions for all scales exhibited negative skewness ("ceiling effect"). Score mean differences across periods were not statistically significant; therefore, we aggregated post-baseline measurements and compared the aggregate to baseline. Two of 18 providers (11%) showed a score mean increase on all 6 scales, 2 (11%) showed an increase on 5 of 6, 2 (11%) showed an increase on 4 of 6, and 3 (17%) showed an increase on 3 of 6; all were statistically significant. For white patients, there was a statistically significant increase on the Provider's Supportiveness scale. There were no statistically significant increases on any scale for other patient race/ethnicity groups. Correlations of scale scores with behavioral outcomes were not statistically significant.
Relationship of skills use metrics to scale scores: Across the full patient sample, both time and provider were statistically significant predictors of frequency (adjusted R2 (AR2) = .49) and duration (AR2 = .39) of Agenda-Setting, while provider only was a predictor of frequency (AR2 = .45) and duration (AR2 = .33) of all 4 skills jointly. For white patients, time and provider were predictors of frequency (AR2 = .45) of Agenda-Setting, and provider was a predictor of frequency (AR2 = .43) and duration (AR2 = .38) of all 4 skills jointly. For African American patients, provider was a predictor for frequency (AR2 = .54) and duration (AR2 = .35) of Agenda-Setting, and for frequency (AR2 =.48) and duration (AR2 = .30) of all 4 skills jointly; also, Trust in VA System was a statistically significant covariate for duration of all 4 skills jointly. There were no statistically significant predictors for Hispanic/Latino patients. No other study variables, when included as covariates in the models, were statistically significant.
Across the full patient sample, duration of Agenda-Setting and of all 4 skills jointly were statistically significant predictors of Provider's Supportiveness (AR2 = .07 and .09), and duration of all 4 skills jointly was a predictor of Provider's Sensitivity (AR2 = .07). For African American patients, duration of all 4 skills jointly was a predictor of Provider's Supportiveness and Provider's Sensitivity (AR2 = .15 and .13), provider's years in practice was a predictor of Provider's Supportiveness and Provider's Sensitivity (AR2 = .12 and .10), and provider's years in practice and patient's years of education completed were predictors of Provider's Similarity (AR2 = .10 for both). For Hispanic/Latino patients, duration of Agenda-Setting was a predictor of Provider's Supportiveness (AR2 = .39), and duration of all 4 skills jointly was a predictor of Provider's Sensitivity (AR2 = .41). There were no predictors of scale scores for white or women patient groups, or for behavioral outcomes for any group.
Provider post-session questionnaire and other provider feedback: The first question to providers was "How often have you been able to use this skill with your patients?" For Agenda-Setting, 6 (33%) chose "With every patient," 4 (22%) chose "With several patients a day," 1 (6%) chose "With one patient a day," 5 (27%) chose an option of less than one patient a day, and 2 (11%) did not respond. For Paraphrasing, 9 (50%) chose an option of at least one patient a day, and 9 (50%) chose less. For both Eliciting the Patient's Perspective and Strategic Pausing, 10 (56%) chose an option of at least one patient a day, and 8 (44%) chose less. For all 4 skills, there were 19 (26%) responses of "With every patient," 15 (21%) of "With several patients a day," 2 (3%) of "With one patient a day," 10 (14%) of less than one patient a day, and 26 (36%) non-responses.
The second question to providers was "Have you noticed any changes in your conversations with patients since you started using this skill?" For Agenda-Setting, 2 (11%) chose "Longer," 3 (17%) chose "More detailed," 2 (11%) chose "More personal," 4 (22%) chose "More two-way discussion," and 8 (44%) chose "No change." For Paraphrasing, 2 (11%) chose "Longer," 6 (33%) chose "More detailed," 2 (11%) chose "More personal," 4 (22%) chose "More two-way discussion," and 4 (22%) chose "No change." For Eliciting the Patient's Perspective, 3 (17%) chose "Longer," 4 (22%) chose "More detailed," 4 (22%) chose "More personal," 3 (17%) chose "More two-way discussion," and 4 (22%) chose "No change." For Strategic Pausing, 2 (11%) chose "Longer," 2 (11%) chose "More detailed," 2 (11%) chose "More personal," 4 (22%) chose "More two-way discussion," and 8 (44%) chose "No change." For all 4 skills, there were 9 (13%) responses of "Longer," 15 (21%) of "More detailed," 10 (14%) of "More personal," 18 (25%) of "More two-way discussion," and 20 (27%) of "No change."
Written comments on the questionnaire forms included, for Agenda-Setting, "very good technique;" and for Paraphrasing, "have been using this for years," but also "it worked against me a couple times, patient got angry." There were no written comments for Eliciting the Patient's Perspective or for Strategic Pausing. Verbal comments made by providers during training sessions included, "I can see the value of doing this," "this is like something I already do," and "this has been useful with my patients," but also "when I ask, 'do you have an agenda for discussion today,' some patients say, 'no, you are the one who called me in and therefore you must have an agenda,'" and "I've noticed my attention to doing the skills has increased, but I'm not sure it's affected my discussions."
Our findings addressed our research hypotheses as follows:
H1A:Compared to baseline, provider use of each skill with patients will increase after the introduction of that skill.
Not supported; there was an increase across providers as a whole, but it was not statistically significant.
H1B:Compared to baseline, differences in provider use of each skill with white versus minority patients will decrease after the introduction of that skill.
Not supported; there was a decrease across providers as a whole, but it was not statistically significant.
H1C:Increases in provider use of a skill will be associated in time with the introduction of that skill; that is, less often used before, more often used after.
Not supported; increases varied over time (period-to-period) idiosyncratically by provider, and while there was an increase across providers as a whole (from baseline to post-baseline), it was not statistically significant.
H2A:Increases in frequency of a skill will be associated with increases in perceptual and behavioral outcome measures of patients.
Not supported; there was an increase across providers as a whole in mean score for all 6 perceptual scales from baseline to post-baseline, however, none of these increases was statistically significant. There was no evidence of increase in the behavioral outcomes of patients.
H2B:Increases in perceptual and behavioral outcome measures will be greater for minority patients than for patients overall.
Not supported; there were no statistically significant increases from baseline to post-baseline on any perceptual scale for any race/ethnicity minority groups. There was no evidence of increase in the behavioral outcomes of patients.
We found larger increases from baseline in use and duration of Agenda-Setting and all 4 skills jointly among gender-discordant patient-provider pairs versus gender-concordant pairs. These increases were statistically significant. A similar relationship, also statistically significant, was found for African/African American race-concordant versus race-discordant pairs, but for the Agenda-Setting skill only. These findings suggest that gender and race concordance of patient-provider pairs may play a role in clinical communication development, affirming a concept established in prior research on patient-provider communication.
Our multivariate modeling findings suggest that provider identity, or the idiosyncrasies that comprise an individual provider's communication style and ability, is the strongest predictor of communication skill use and duration, which was an anticipated finding. Unfortunately, individuals' idiosyncrasies in communicating are trait-like and thus not amenable to training interventions. We found little statistical support for time (post-intervention versus baseline) as a predictor, which was an unanticipated finding. We had anticipated that results would be more sensitive to time. Also, the Trust in VA System scale did not prove to be a useful covariate for these statistical models. While duration of Agenda-Setting and all 4 skills jointly were statistically significant predictors for scores on 2 of the perceptual scales, we found no statistically significant predictors for the patient behavioral outcomes. Our explanation for this is the scarcity of missed appointments, missed VA pharmacy refills and unscheduled visits to the VA in our sample of patients, rendering a lack of measurable relationships among these variables.
Lastly, the self-report data and comments we collected from providers appeared largely inconsistent with the actual frequency and duration of providers' use of the 4 skills. Specifically, 60% of providers reported using Agenda-Setting with at least one patient per day, and several providers indicated that it was a skill that seemed familiar or natural to them; these findings were supported by the actual use data, which showed about 0.5 use of Agenda-Setting with every patient in the sample, and showed some providers using it notably more often than this. In contrast, 50% or more of providers reported using the other 3 skills, and all 4 skills jointly, with at least one patient per day, but the actual use data showed far less use of these skills. We performed coding double-checks on a sample of patient-provider audio recordings to ensure our accuracy of coding the skills in question, and found minimal discrepancies with the initial coding record. Therefore, we suggest that providers may have interpreted these skills, and used them in conversations, in ways that did not manifest as the skills taught in the training sessions. Alternatively, providers simply may have overestimated the frequency with which they used these skills.
Given the lack of statistical significance in our findings for our research hypotheses, we have no recommendation for assessing our communication skills intervention with a larger VA audience. However, the 4 clinical communication skills examined in this study have been shown in other research to enhance patient-provider communication, particularly Agenda-Setting, and given the feedback from our sample of providers, we believe these skills should be considered for further study within VA. Also, we believe our findings on the effect of gender- and African/African American race-concordance on clinical communication are important and suggestive of further study on this topic within VA.
- Kelly PA, Haidet P, Street RL. Evaluating a patient-centered clinical communication skills intervention to reduce disparities in U.S. veterans’ health care. Presented at: International Conference on Communication in Healthcare; 2012 Sep 1; St. Andrews, Scotland.