HSR&D Home » Research » EDU 08-430 – HSR&D Study
Intervening to Prevent Contextual Errors in Medical Decision Making
Saul J. Weiner, MD
Jesse Brown VA Medical Center, Chicago, IL
Funding Period: July 2009 - December 2012
Our prior work has focused on identifying and predicting serious and common medical errors that occur when physicians fail to take into account patients' unique circumstances, or context, when planning their care. Such "contextual errors" can have adverse effects that are as significant as those that result from overlooking signs of a pathophysiologic condition. The failure, for instance, to recognize that a patient is not able to take a medication correctly (e.g. because of cognitive disabilities or cost) may have as deleterious consequences as the failure to prescribe the correct medicine. The purposes of this study were to develop and test an intervention to train physicians to avoid such errors by contextualizing medical decisions when planning patient care, and to assess whether contextualized care is associated with improved health care outcomes.
Our proposed randomized medical education intervention study was designed to test three hypotheses:
1.Contextually-trained physicians are more likely to elicit and incorporate contextual information essential to planning appropriate care when tested during encounters with standardized patients than those not so trained.
2.Contextually-trained physicians are more likely to adapt a plan of care to their real patient's needs and circumstances than those not so trained.
3.Patients whose care is contextualized are more likely to have positive outcomes than patients whose care does not address identified contextual factors.
To determine whether a care plan is contextualized we employed a structured approach to reviewing the medical record and coding audio-recorded encounters to answer three questions: Are there clues to underlying essential contextual factors, which we term "contextual red flags"? If so, did either the physician explore them for underlying contextual factors that could be addressed in a care plan or the patient unprompted reveal such information? If so, did the physician address the contextual factors in the recommended care plan? A detailed step-by-step protocol, with examples, is available online in a manual titled "Content Coding for Contextualization of Care," or "4C." The manual may be accessed at http://dvn.iq.harvard.edu/dvn/dv/4C.
To determine whether a care plan was associated with an improved health care outcome, we monitored the patient's medical record for nine months after the visit to determine if the original contextual red flag partially or fully resolved.
In addition, half of physicians in the study were randomly selected to participate in four one-hour seminar series on patient centered decision making based on a curriculum designed for medical students and described in a prior publication. Skills acquisition was assessed using standardized patients, both following the intervention and in the control group.
Among 222 eligible resident physicians with primary care clinics at the VA, 139 consented to participate over a 30-month period. Twenty-six percent were first-year residents, 40% were second-year residents, and 34% were third-year residents. Fifty percent were women, and 50% had participated in the intervention. Among the population of study-eligible categorical residents in these programs during the study period, 48% were women, 29% were first-year residents, 36% were second-year residents, and 35% were third-year residents. The consenting residents did not differ from the eligible population in either sex (P = 0.34) or training year (P = 0.41).
A total of 1,799 patients were informed of the study at the time of check in. Among these patient 160 were not interested in meeting with the RA, 754 declined during the consent process, and 111 were called to their appointments before the consent process could be completed, leaving 774 who consented to participate. These patients were provided audio recorders to conceal during the encounter and return following their visit.
Each physician was repeatedly audio recorded until we had collected three encounters The final data set consisted of 403 encounters containing a total of 548 identified contextual red flags.
As Assessed with Standardized Patients
There was a significant interaction in performance between intervention/control group and baseline/contextual cases (F(1,489=4.5, p=.035). Residents in the intervention performed correctly on the contextual cases 62% of the time (95% CI [50%-72%]), while residents in the control group performed correctly only 44% of the time (95% CI [34%-55%]). The two groups did not differ in performance on baseline cases (intervention 78% [67%-86%], control 80% [72%-87%]).
In Care of Real Patients
In a mixed-effects logistic regression on encounters (clustered by physician) there was no effect from whether the physician had participated in a contextualization of care seminar (P = 0.33).
Among 548 contextual red flags, 208 contextual factors were confirmed, either when physicians probed or patients volunteered information. Outcome data were available for 157 contextual factors, and contextualized planning was documented for 96 of these. When care planning addressed contextual factors, health care outcomes improved in 68 (71%), compared with 28 (46%) when contextual factors were not addressed (P = 0.002).
The coding system developed for this study, 4C, provides a tool for assessing physician performance at patient-centered decision making, a key component of patient centered care.
Applying 4C, this study has demonstrated that when physicians adapt care plans to individual patient's life circumstance and needs - i.e. their context - health care outcomes improve. We also found that a four hour training intervention can improve physician skills at contextualizing care, as measured using standardized patients, but that these skills are not then employed in actual practice as assessed using data collected from real patient encounters. The implications are than more sustained interventions will be required to change physician performance.
In response VISN 12 has initiated a QI project at two facilities employing the methodology developed for this study to assess physician performance at contextualizing care. The data will be shared at regular intervals with participating attending physicians in primary care, and their performance tracked over time to see if a cycle of continuous feedback leads to improved care.
External Links for this Project
Dimensions for VADimensions for VA is a web-based tool available to VA staff that enables detailed searches of published research and research projects.
If you have VA-Intranet access, click here for more information vaww.hsrd.research.va.gov/dimensions/
VA staff not currently on the VA network can access Dimensions by registering for an account using their VA email address. Search Dimensions for this project
DRA: Health Systems
DRE: Diagnosis, Prevention, Treatment - Comparative Effectiveness
Keywords: Behavior (provider), Education Research, Safety
MeSH Terms: none