PCORI funded the Pilot Projects to explore how to conduct and use patient-centered outcomes research in ways that can better serve patients and the healthcare community. Learn more.
Few studies of patient recall have included information about specific provider behaviors believed to be associated with better patient recall after the provider-patient dialogue, such as “teach back,” open questions, agenda setting, patient participation, and elements of shared decision making. To address this gap, researchers wanted to test the feasibility of using observational and interview methods to assess characteristics of provider-patient communication in outpatient care. They developed a detailed characterization of decision processes (“resolutions”), tested the reliability and validity of these resolutions, and linked them to patient understanding and recall.
The objective of Methods for Evaluating Decisions and Information in Clinical Communication (Project MEDICC) was to assess the frequency of provider communication behaviors hypothesized to be associated with better patient recall and components of shared decision making. Researchers examined the relationship between features of clinical interactions and patient recall, satisfaction, and activation. They hypothesized that routine decision-making dialogue (e.g., decisions about medical interventions such a stopping, starting, or changing medications, and behavioral recommendations such as salt restriction or exercise) can be reliably identified and classified; that decision making is rarely shared and often poorly recalled; and that when it is shared, patients are more engaged, more adherent, and more satisfied.
Participants, Interventions, Settings, and Outcomes
Participants were 11 cardiologists, eight nephrologists, and 101 patients. Patients’ mean age was 57; 54% were women; 72% were white, 20% were black, and 8% were “other.” Providers were a convenience sample from participating clinics. Patients were consecutively sampled from those newly referred with a diagnosis of chronic kidney disease or heart disease, or those who were experiencing an exacerbation or imminent significant clinical decision. Exclusion criteria were limited English or cognitive impairment.
The setting was two outpatient cardiology clinics and one nephrology clinic affiliated with academic medical centers.
Outcome measures included patient recall of treatment decisions and information received (operationalized as three levels of quality), satisfaction with interaction and decision, intention to adhere to recommendations, patient scores on the Patient Activation Measure, and patient scores on subscales of the Primary Care Assessment Survey.
Data sources included audio-recorded and transcribed routine clinical encounters and follow-up interviews with patients.
Clinical encounters were audio recorded, transcribed by an outside firm, and corrected by project staff. Staff conducted patient follow-up surveys by telephone approximately one week later. This process was repeated at subsequent visits occurring during the study period. Trained research assistants coded the transcripts by using versions of the Generalized Medical Interaction Analysis System, Comprehensive Analysis of the Structure of Encounters System, and OPTIONS system for coding elements of shared decision making.
Researchers generated descriptive data on the frequency of items coded for recall, the quality of patient recall, the prevalence of communicative behaviors hypothesized to be associated with better recall, and OPTIONS codes. The research team treated medical and behavioral resolutions separately, as understanding may require different kinds of background knowledge and place different cognitive demands on patients. As the provider behaviors were uncommon, researchers could not test the association with recall. They tested the bivariate association of recall quality with various independent variables. Researchers constructed multivariate ordered logit models to predict patient recall using patient education, number of items to be recalled in a visit, and the ratio of provider to total (patient plus provider) utterances in resolutions within the visit (“verbal dominance”). They tested patient age, gender, race/ethnicity, and provider specialty as covariates and constructed separate models for behavioral and medical resolutions.
Quality of Data and Analysis: Missing data were assumed to be random. Researchers tested intuitively plausible covariates in the models and intuitively plausible interactions, and they retained parsimonious models.
Recall was strongly associated with patients’ level of formal education (p<.0001). About 38% of resolutions were recalled freely and accurately by people with less than a high school diploma whereas 65% were recalled freely and accurately by people with a college degree. This difference remained significant when selecting only medical or only behavioral resolutions. Provider verbal dominance was significantly associated with poorer recall in medical processes, and in all processes, although not in behavioral processes.
Patient age, gender, race/ethnicity, and the specialty (nephrology vs. cardiology) were not associated with recall quality. In a multivariate model including the interaction between the number of resolutions in a visit and patient education (< some college), researchers found that the interaction term, verbal dominance, and the resolution count per visit were all significant predictors of poorer recall quality, whereas the effect of patient education was no longer significant. More items to be recalled was more strongly associated with worse recall for medical resolutions for patients with less formal education, but less strongly associated for behavioral resolutions. Verbal dominance remained a significant predictor of poorer recall. Scores on the Patient Activation Measure were positively and significantly associated with recall quality for behavioral decisions but not for medical decisions.
This is a fairly small study limited to three sites and a convenience sample of providers.
Patient recall can be enhanced by limiting the number of decisions and recommendations in an encounter and by encouraging patients to participate more actively in decision-making dialogue. Recall of medical resolutions seems to be more sensitive to patient engagement and participation than is recall of behavioral resolutions. However, providers rarely engage in behaviors to encourage patient participation. These findings may support the development of evidence-based interventions to improve interaction quality, shared decision making, and ultimately patients’ health outcomes.
Future work will apply these methods to more specific research questions including the relationship between diagnosis-related interaction and risk of diagnostic error; interactions in primary care management of chronic pain; and primary care of older patients.