Jump to start of content

Breadcrumb trail:

Can three practice processes improve provider-patient interactions for patients with risk factors for limited health literacy?: A case-control study.


Arnold GK, Baranowski RA. — American Board of Internal Medicine

Presented: Health Literacy Research Conference, October 2012

Background: Providing high-quality health care is complex and becomes more challenging when patients lack sufficient literacy skills to participate actively in their treatment. Similarly, physicians may lack skills for communicating with patients about the nature of their conditions, available treatment options and actions patients must take as part of their treatment plan. Thus, successful health outcomes are jeopardized, and the risks for medical errors and health disparities increase. This study investigated whether primary care practices that (1) have written policies on availability of language services for patients with limited English proficiency, (2) prominently display patients' language preference in their medical records, and (3) assess other communication barriers including health literacy and physical impairments (e.g., hearing, vision), have better patient-visit experiences than matched-control practices that do not take these steps. Indicators of better patient-visit experience were higher patient ratings on quality of physician-patient communication, higher patient ratings on participation in shared decision-making about treatment, and higher overall ratings on quality of care.

Methods: A total of 1,245 internists completed the ABIM Communication-Primary Care Practice Improvement Module (PIM) as part of their maintenance of certification. The PIM focuses on improving physician-patient interactions. The module required administering CG-CAHPS surveys to a sample of 25 patients; in total, 21,347 patients with one or more NAAL-specified risk factors for low health literacy were included in the survey pool. A total of 436 physicians indicated (1) they had written policies on providing language services for patients, (2) the preferred language of patients appeared in their medical records, and (3) other patient-communication barriers were assessed. The 809 practices without these features served as a control pool. We created three composite scores reflecting patients' ratings for (1) how well physicians communicate (0-35 score, Cronbach's α = .86); (2) level of shared decision-making (0-9 score; Cronbach’s α = 0.77); and (3) overall rating of physicians (0-10 score; Cronbach's α = 0.82). The hypothesis was that practices implementing language services and literacy-assessment policies would have higher patient ratings than practices that did not implement these services. A dual-propensity score-matching procedure formed case-control pairs. One score matched 31 variables on physician and practice characteristics. The other score matched 12 demographic variables on patients, and was a weighted score adjusted for clustering of patients within physicians. Cases and controls were matched by shortest Mahalanobis distances. Wilcoxon signed-ranks tests tested ratings with a two-tailed α = 0.1 significance. Rosenbaum's Bias Parameter was used to assess sensitivity of significant results to hidden biases.

Results: A total of 279 practices had the policies (cases) with complete data for analyses; 9,593 patient responses were compared. The differences in communication scores between cases: mean and (SD) 27.8 (1.8) versus controls: 27.6 (1.8) (p = .172) and in overall ratings cases: 9.4 (0.5) versus controls 9.4 (0.5) (p = .912) were inconclusive. The difference in shared decision-making (SDM) scores cases: 7.1 (1.4) versus controls: 6.9 (1.3) (p = .072) approached significance. Rosenbaum’s Bias Parameter for SDM score comparison was 1.02 showing potential sensitivity to hidden biases.

Conclusions: Practices that offer language services for limited English-speaking patients and evaluate other barriers to communication, including literacy, received significantly higher mean SDM ratings suggesting that practices trying to communicate effectively with low-literacy patients may improve their engagement in their treatment options. The results, however, were sensitive to possible hidden biases including time and instrument effects. The SDM results should be replicated. The effects of these policies on physician-patient communication or overall ratings on quality of care were inconclusive. Additional efforts are needed to improve the quality of provider-patient encounters among patients with low health literacy.

For more information about this presentation, please contact Research@abim.org