4J
ORAL ABSTRACTS: SHARED DECISION MAKING AND PATIENT CENTERED DECISION SUPPORT
Angela Fagerlin, PhD
University of Utah / Salt Lake City VA
Professor and Chair Population Health Sciences/Research Scientist
Purpose: Despite substantial evidence that patients value engagement with SDM tools, sustained use of such tools in real life clinical practice has proven challenging. We studied change in usage patterns of a shared decision making (SDM) tool integrated within the electronic health record (EHR).
Method: In 2010 a dynamic SDM tool, designed to support statin decisions by clinicians and patients, was integrated within a university medical center's EHR. Unique patient and clinician use of the tool was recorded for 6 years. Multiple uses by the same patient within 180 days counted as one use. No advertising or education accompanied the launch.
We calculated percent likelihood that clinicians would use the tool again as usage increased. We also measured time to subsequent use for a subset of experienced users.
Result: Over 6 years 1,569 different clinicians accessed the tool with 47,981 unique patients. After first use, 58% of clinicians returned. Following uses 2-5: 80%, 88%, 91% and 95% respectively returned. Clinicians using the tool 6-20 times, averaged a return rate of 97.2%
Of 244 primary care clinicians (PCCs), 73% (178) used the tool once, 96% used it twice and 98% returned for use #3. After use 3-20, average return was 99.3%.
PCCs who used the tool with >100 patients, (N=86) accounted for 86% of all 26,668 uses in primary care. Time to second use averaged 104 days; after five uses average return was <30 days; after 100 uses return was <7 days and when over 400 uses, return was under 48 hours.
Conclusion: True SDM occurs between a patient and clinician in the office and with EHR workflow integration, our tool addresses a known barrier to tool utility and uptake. Over 6 years, through word of mouth and occasional clinical presentations, a substantial majority of PCCs who tried the tool, integrated regular use into their practice.
Use was particularly sticky as >99% of PCCs with >=3 uses returned with more patients. Assuming a clinician performs a stable rate of statin conversations, accelerated tool use by over 50-fold illustrates an increasing percent of patients with whom clinicians choose to use the tool.
As next steps, the tools are integrated in two new institutions and uptake will be followed.
Methods: We conducted semi-structured interviews with 16 purposefully sampled surrogates of previously hospitalized ciTBI patients (up to two years from injury) recruited from two level-1 trauma-centers in the Northeast and Mid-Atlantic region, and 20 subspecialty physicians (neurocritical care,neurosurgery,trauma,palliative care) from seven academic hospitals in the Northeast,Mid-Atlantic,South,West, and Midwest. We explored the factors that physicians consider when they make predictions about long-term outcomes, and preferences for communicating these predictions for both physicians and surrogates. Two independent reviewers analyzed transcribed interviews using deductive and inductive approaches (NVIVO-software), with adjudication by a third reviewer when necessary. The sample size was determined by theme saturation.
Results: Overall, 82% of surrogates stated a preference for numeric estimates from the physicians to predict outcomes, as compared to purely qualitative predictions(“highly likely”or “unlikely”). All surrogates explained that numerical values were “more direct”,“clearer, more concise,and less confusing” than descriptive predictions. In contrast, 75% of physicians stated that they do not typically give numeric predictions when discussing outcomes in ciTBI due to distrust in the accuracy of existing models: “Better have good data to do that with, and most often, we do not.”While physicians were all aware of the surrogates’ preferences for numeric risk prediction, they also voiced concern that surrogates might misinterpret numeric risk estimates: “Families will often hook onto those numbers and make an absolute judgment…they become simplified and [these numbers are] used against you later.”
Conclusion: We found lack of concordance between physicians and surrogate decision-makers of ciTBI patients regarding the communication of predicted long-term outcomes. Shared-decision making tools in ciTBI could potentially help set more realistic expectations by explaining to surrogates the limitations of risk prediction models and helping them to better interpret probabilistic information.
Purpose:
Patients facing a high-stakes clinical decision often confront a bewildering array of treatment options that may impact many facets of their lives. A good decision requires consideration of patients' goals and preferences and providers' scientific rationale for treatment. However, preference-assessment instruments typically focus on pre-selected clinical outcomes and preference attributes that may not be relevant to patients. We sought to develop a patient-centered approach to elicit treatment goals and to compare the treatment goals of patients with Multiple Sclerosis (MS) to MS healthcare providers.
Methods:
To generate a comprehensive list of treatment goals, we conducted 6 structured focus groups using Nominal Group Technique among MS patients and MS healthcare providers. Groups were conducted in Georgia, Colorado, Massachusetts, and online. Each group of 5-9 participants responded to one question about their treatment goals. Responses were shared, consolidated, and ranked (top 9). Weights were assigned and scores summed to develop a prioritized list. Goals were consolidated across groups and the combined list was then presented to a larger sample (n=41) to rate and sort into categories using their own criteria (“How do you see these going together?”). A co-occurrence matrix was created based on how frequently items were sorted into the same category. Multidimensional Scaling was used to create an optimal geometric solution for the matrix (“cognitive map”) and Hierarchical Cluster Analysis and clinical judgement via patient and provider consultation was used to identify and label clusters of related attributes.
Results:
34 unique patient-generated treatment goals were included in the combined list. Ten clusters were identified and prioritized (highest to lowest): Brain Health, Disability Concerns, Avoiding Flare-ups/Progression, Caring/Informed Medical Team, Quality-of-Life, Safe Treatments, Symptom Management, Financial Concerns, Lifestyle & Daily Living, and Avoiding Care Facilities. Goals generated by patients were primarily focused on managing symptoms and limiting loss of function, while providers focused on slowing disease progression. The ratings and groupings of items also differed between the groups.
Conclusions:
Unlike other preference assessment approaches that focus on assigning weights to investigator-identified outcomes and attributes, this patient-centered approach prioritizes and maps outcomes and attributes elicited by patients, minimizing investigator bias and maximizing their relevance to patients. Cognitive mapping can be used to identify patient goals and gaps among priorities considered by patients and their providers.
Method: A prospective cohort study enrolled eligible patients with knee or hip osteoarthritis, lumbar herniated disc (LDH) or spinal stenosis (LSS). Participants were surveyed one week after their visit with a specialist to assess knowledge, preferred treatment, and baseline quality of life (QoL) (EQ-5D, Knee injury and Osteoarthritis Outcome Score (KOOS), Harris Hip Score, Oswestry Disability Index (ODI)). Patients with a passing knowledge score (60% or higher) who received their preferred treatment (either surgery or non surgical) were considered to have made an informed, patient-centered (IPC) decision.
A follow-up survey at six months assessed QoL, regret and satisfaction. We tested hypotheses that patients who made IPC decisions would have higher QoL, higher satisfaction and less decision regret at follow-up. Regression analyses accounted for clustering of patients within clinicians and controlled for surgery, age, gender, joint and baseline QoL. With 550 surveys the study had more than 80% power to detect a difference of 0.05 on the EQ-5D between those who made IPC decision or not.
Result: The response rate to the initial survey was (652/926, 70.3%) and to the follow up survey was (551/648, 85%). About half received a decision aid (45.5%) and about half had surgery (49.0%) within the 6 months of the initial visit. Patients who made IPC decisions (36.0%) reported significantly better overall and disease-specific QoL, across all topics. The unadjusted increase was 0.06 for EQ-5D, p=0.006, 4.72 points for KOOS symptoms, p=0.009, 2.93 points for Harris Hip Score p<0.0001, and -7.59 points on ODI p<0.0001. Participants who made IPC decisions were more likely to be extremely satisfied with improvement in symptoms (76.68% vs. 41.86%, p=0.0003), more likely to be very or extremely satisfied with treatment (70.68% vs. 34.66%, p=0.0003), and had less regret (5.2% vs. 15.0% p=0.0006).
Conclusion: Patient engagement in elective surgery decisions is important ethically, and evidence from this study suggests that well-informed patients that receive their preferred treatment have higher satisfaction and small improvements in health outcomes.
Method: We conducted a qualitative study at two academic medical centers, among women who had experienced periviable deliveries (22-25 weeks) within the past 3 years. Using semi-structured interview guides, we asked participants what advice they would offer healthcare providers and other women/families based on the lessons learned from their own experiences. Interviews were audio-recorded and transcribed verbatim. De-identified transcripts were independently coded by a team of 3 trained coders.
Result: In our preliminary analyses of thirteen interviews, four broad categories emerged: General Advice for Healthcare Providers, Advice to Improve Counseling, Advice for Moms and Families, and ‘Changes’ that would improve care or clinical experiences. Participants advised healthcare providers to be more ‘patient’ and to appreciate the novelty of the event for patients and the uniqueness of each family’s experience and perspective. In counseling, they were encouraged to be ‘realistic’ and ‘factual’ to prepare families for the full range of potential outcomes. Joint OB/Neonatology counseling was favored. Descriptions and images of periviable neonates, resuscitation attempts, and intensive care units were suggested to prepare parents and set expectations--though others stated that no advice or counseling could ever ‘prepare’ a parent for the experience. Guilt and blame were repeated themes. Healthcare providers were encouraged to reassure pregnant and postpartum women that it was not their fault, and moms were encouraged to seek support or counseling to cope with feelings of guilt and self-blame. Women and families were encouraged to ask questions, get engaged, avoid the internet, and try to find joy and meaning in whatever time they had with their child. Finally, systems changes that could better accommodate and facilitate mother/baby contact and/or time spent together were felt to be needed.
Conclusion: A robust literature on patient perspectives in periviable care is lacking, yet sorely needed as we seek to develop decision-support tools to facilitate more shared, informed decision-making for periviable delivery management and resuscitation decisions. Our findings will present patient perspectives to improve interactions with healthcare providers; optimize the experience of pregnant women and their families; and advance efforts toward developing more patient-centered systems of periviable care.
Method: Data were collected over 3 years (2013-2015) of annual OSCEs. Trained standardized patients (SP) portrayed a pregnant woman presenting to clinic at 23 weeks gestation with preterm premature rupture of membranes (PPROM). Residents were instructed to counsel the patient as they would in their typical practice and develop a plan for her care. OSCE sessions were directly observed, video- and audio-recorded. Braddock’s 9-item measure of shared decision-making was adapted to a 10-item scoring rubric, adding one item to rate empathic behaviors. Each item was rated as 0 (absent), 1 (partial), or 2 (complete). Two coders independently rated the encounters. Discrepancies were resolved by consensus.
Result:
Twenty-six fourth-year residents participated. All residents provided ‘complete’ discussions of the clinical issue/nature of the decision, and all received ‘complete’ or ‘partial’ ratings for informing the patient of her prognosis (62% and 38%, respectively) and assessing the patient’s understanding (4% and 96%). The vast majority of residents received ‘complete’ or ‘partial’ ratings for discussions of the patient’s role in decision-making (42% and 50%, respectively); discussions of management alternatives (69% and 23%); and discussions of risks and benefits of those alternatives to the infants (31% and 58%) and to mom (46% and 46%). Discussions of the patient’s goals of care and larger context of the decisions (i.e., related to resuscitation preferences, disability, or quality-of life concerns) were absent in 69% of residents’ discussions. Likewise, the majority of residents (62%) had no discussion of the patient’s preference, typically suggesting that no decision be made until after talking to neonatology. Only 42% of residents discussed uncertainties related to the decision.
Conclusion: We developed an OSCE to evaluate the degree to which OB/GYN residents utilize SDM when counseling patients about periviable birth and delivery management. Residents consistently convey information related to diagnosis, prognosis, decision-making roles, risk, benefits and alternatives; but often fail to address uncertainties and patients’ goals or preferences. This suggests that, while ample medical information is conveyed, critical elements of SDM related to patient’s values, goals and preferences are not explored. Interventions and training are needed to facilitate SDM in periviable care.