Balancing the Elicitation Burden and the Richness of Expert Input When Quantifying Discrete Bayesian Networks.

Structured expert judgment (SEJ) is a method for obtaining estimates of uncertain quantities from groups of experts in a structured way designed to minimize the pervasive cognitive frailties of unstructured approaches. When the number of quantities required is large, the burden on the groups of experts is heavy, and resource constraints may mean that eliciting all the quantities of interest is impossible. Partial elicitations can be complemented with imputation methods for the remaining, unelicited quantities. In the case where the quantities of interest are conditional probability distributions, the natural relationship between the quantities can be exploited to impute missing probabilities. Here we test the Bayesian intelligence interpolation method and its variations for Bayesian network conditional probability tables, called "InterBeta." We compare the various outputs of InterBeta on two cases where conditional probability tables were elicited from groups of experts. We show that interpolated values are in good agreement with experts' values and give guidance on how InterBeta could be used to good effect to reduce expert burden in SEJ exercises.

A structured expert judgement elicitation approach: how can it inform sound intervention decision-making to support household food security?

OBJECTIVE: To examine structured expert judgement (SEJ) elicitation as a method to provide robust, defensible data for three determinants of household food security (food cost, household disposable income and physical access) for quantifying a proof-of-concept integrating decision support system for food security. DESIGN: SEJ elicitation is a validated method for obtaining unavailable data, but its use in household food security in high-income countries is novel. Investigate Discuss Estimate Aggregate (IDEA) elicitation protocol was implemented, including quantitative and qualitative elements. Using specific questions related to three determinants, food security experts were encouraged to Investigate - estimate individual first-round responses to these questions, Discuss - with each other evidence on the reasoning and logic of their estimates, Estimate - second-round responses, following which these judgements were combined using mathematical Aggregation. SETTING: Victoria, Australia. PARTICIPANTS: Five experts with a range of expertise in the area of household food insecurity participated in the SEJ elicitation process. RESULTS: The experts' ability to provide reliable estimates was tested and informed the aggregation of the collection of individual estimates into a single quantity of interest for use in decision support. The results of the quantitative elicitation show the impact of combinations of varying household income, food cost and physical access on household food security status and severity and is supported by the experts reasoning during elicitation. CONCLUSION: This research provides insight to the application of SEJ where elicited data can inform and support intervention decision-making specific to household food security, especially where evidence is absent or of poor quality.

Fitness predicts long-term survival after a cardiovascular event: a prospective cohort study.

OBJECTIVES: To identify the role of fitness, fitness change, body mass index and other factors in predicting long-term (>5 years) survival in patients with coronary heart disease. DESIGN: Cohort study of patients with coronary heart disease recruited from 1 January 1993 to 31 December 2002, followed up to March 2011 (1 day to 18 years 3 months, mean 10.7 years). SETTING: A community-based National Health Service (NHS) cardiac rehabilitation programme serving the Basingstoke and Alton area in Hampshire, UK. PARTICIPANTS: An unselected cohort of NHS patients, 2167 men and 547 women aged 28-88 years, who attended the rehabilitation programme following acute myocardial infarction, an episode of angina or revascularisation, and had a baseline fitness test. MAIN OUTCOME MEASURES: Cardiovascular mortality and all-cause mortality. RESULTS: A high level of fitness (VO2≥22 mL/kg/min for men, VO2≥19 mL/kg/min for women) at completion of the programme was associated with decreased all-cause death, as was a prescription for statins or aspirin, and female gender. Increase in all-cause mortality was associated with higher age and ACE inhibitors prescription. Higher risk of cardiovascular mortality was associated with increasing age, prescriptions for ACE inhibitor, and diagnosis of myocardial infarction or angina as compared with the other diagnoses. CONCLUSIONS: Prior fitness and fitness improvement are strong predictors of long-term survival in patients who have experienced a cardiac event or procedure. Some secondary prevention medications make a significant contribution to reducing all-cause mortality and cardiovascular mortality in these patients. This study supports public health messages promoting fitness for life.

Matching patients to an intervention for back pain: classifying patients using a latent class approach.

RATIONALE, AIMS AND OBJECTIVES: Classification of patients with back pain in order to inform treatments is a long-standing aim in medicine. We used latent class analysis (LCA) to classify patients with low back pain and investigate whether different classes responded differently to a cognitive behavioural intervention. The objective was to provide additional guidance on the use of cognitive behavioural therapy to both patients and clinicians. METHOD: We used data from 407 participants from the full study population of 701 with complete data at baseline for the variables the intervention was designed to affect and complete data at 12 months for important outcomes. Patients were classified using LCA, and a link between class membership and outcome was investigated. For comparison, the latent class partition was compared with a commonly used classification system called Subgroups for Targeted Treatment (STarT). RESULTS: Of the relatively parsimonious models tested for association between class membership and outcome, an association was only found with one model which had three classes. For the trial participants who received the intervention, there was an association between class membership and outcome, but not for those who did not receive the intervention. However, we were unable to detect an effect on outcome from interaction between class membership and the intervention. The results from the comparative classification system were similar. CONCLUSION: We were able to classify the trial participants based on psychosocial baseline scores relevant to the intervention. An association between class membership and outcome was identified for those people receiving the intervention, but not those in the control group. However, we were not able to identify outcome associations for individual classes and so predict outcome in order to aid clinical decision making. For this cohort of patients, the STarT system was as successful, but not superior.