Robustness considerations in Bayesian analysis.

P Gustafson

doi:10.1177/096228029600500403

Home/Archive/PubMed

PubMed · 9004378

Robustness considerations in Bayesian analysis.

P Gustafson

View original source ↗Export metadata ↓

Abstract

All statistical analyses demand uncertain inputs or assumptions. This is especially true of Bayesian analyses. In addition to the usual concerns about the agreement of the data and model, a Bayesian must contemplate the effect of an uncertain prior specification. The degree to which inferences are robust to changes in the prior is of primary interest. This article discusses some robust techniques that have been suggested in the literature. One goal is to make apparent the relevance of some of these techniques to biostatistical work.

Explore related subjects

Adrenergic beta-Antagonists Bayes Theorem Clinical Trials as Topic Data Interpretation, Statistical Humans Markov Chains Meta-Analysis as Topic Models, Statistical Monte Carlo Method Normal Distribution Probability Journal Article

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

P Gustafson. 1996. Robustness considerations in Bayesian analysis.. https://doi.org/10.1177/096228029600500403

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed1998

National patterns and predictors of beta-blocker use in patients with coronary artery disease.

BACKGROUND: Prior studies suggest underuse of beta-blockers in patients with coronary artery disease, but these studies have been based on selected populations of recently hospitalized patients. OBJECTIVE: To describe national patterns and determinants of beta-blocker use in the ambulatory setting. METHODS: We analyzed 11745 visits by patients with coronary artery disease to randomly selected, office-based physicians in the National Ambulatory Medical Care Surveys for 1980, 1981, 1985, and 1989 through 1996. We used multiple logistic regression to determine the independent effect of sociodemographic and clinical factors on beta-blocker use. OUTCOME MEASURE: Beta-blocker use at patient visits. RESULTS: Beta-blocker use was reported in only 20.9% of office visits by patients with coronary artery disease and no strong contraindications between 1993 and 1996. In multivariate analyses, age younger than 75 years, residence in the Northeast, and visits to cardiologists and internists compared with family and general practitioners predicted greater use of beta-blocker therapy. White race and private insurance also were significant predictors of beta-blocker use between 1980 and 1996. Longitudinal analyses revealed a significant decline in beta-blocker use from 1980 to 1990, followed by a gradual increase in recent years. CONCLUSIONS: Beta-blockers appear to be underused in ambulatory patients with coronary artery disease. Our data suggest that nonclinical factors may influence rates of use, indicating the need for closer scrutiny of variations in physician prescribing practices.

Adrenergic beta-Antagonists↗

PubMed1998

Clinical effects of beta-adrenergic blockade in chronic heart failure: a meta-analysis of double-blind, placebo-controlled, randomized trials.

BACKGROUND: beta-Blockers have improved symptoms and reduced the risk of cardiovascular events in studies of patients with heart failure, but it is unclear which end points are most sensitive to the therapeutic effects of these drugs. METHODS AND RESULTS: We combined the results of all 18 published double-blind, placebo-controlled, parallel-group trials of beta-blockers in heart failure. From this combined database of 3023 patients, we evaluated the strength of evidence supporting an effect of treatment on left ventricular ejection fraction, NYHA functional class, hospitalizations for heart failure, and death. beta-Blockers exerted their most persuasive effects on ejection fraction and on the combined risk of death and hospitalization for heart failure. beta-Blockade increased the ejection fraction by 29% (P<10(-9)) and reduced the combined risk of death or hospitalization for heart failure by 37% (P<0.001). Both effects remained significant even if >90% of the trials were eliminated from the analysis or if a large number of trials with a neutral result were added to the analysis. In contrast, the effect of beta-blockade on NYHA functional class was of borderline significance (P=0.04) and disappeared with the addition or removal of only 1 moderate-size study. Although beta -blockade reduced all-cause mortality by 32% (P=0.003), this effect was only moderately robust and varied according to the type of ss-blocker tested, ie, the reduction of mortality risk was greater for nonselective beta-blockers than for beta1-selective agents (49% versus 18%, P=0.049). However, selective and nonselective beta-blockers did not differ in their effects on other measures of clinical efficacy. CONCLUSIONS: These analyses indicate that there is persuasive evidence supporting a favorable effect of beta-blockade on ejection fraction and the combined risk of death and hospitalization for heart failure. In contrast, the effect of these drugs on other end points requires additional study.

Adrenergic beta-Antagonists↗

PubMed1998

Influences of educational interventions and adverse news about calcium-channel blockers on first-line prescribing of antihypertensive drugs to elderly people in British Columbia.

BACKGROUND: The way in which dissemination of evidence changes medical practice needs to be better understood. Controversy about calcium-channel blockers (CCB) in the past 3 years has provided a natural experiment, enabling assessment of the impact of media stories, a national warning letter, a teleconference, small group workshops, and newsletters on first-line prescribing of antihypertensive drugs. METHODS: We included all physicians (4403) in British Columbia who prescribed a thiazide diuretic, beta-blocker, inhibitor of angiotensin-converting enzyme (ACE), or CCB as the first antihypertensive agent for 36,507 residents aged 66 years and over, with no previous or concurrent sign of underlying cardiovascular disease. We used a database covering all prescriptions to elderly people to measure the change in proportion of newly treated patients who received each class of drug as first-line therapy. We used a matched cohort design for assessment of the teleconference and workshops, a randomised community design for the newsletters, and time-series analysis for the media impacts. FINDINGS: The proportion of patients who received a CCB as first-line therapy declined gradually from 22% in early 1994 to 15% in late 1996. This proportion was not affected by two waves of adverse news about CCBs in 1995, but fell by 5% for 5 months and by 3% for 1 month after two waves in 1996. The proportion of patients who received either a CCB or an ACE inhibitor as first-line therapy, contrary to guidelines, was still 42% overall in 1996. The workshops and newsletters were followed by shifts from first-line CCB to first-line thiazide prescribing. INTERPRETATION: Changes in prescribing practices occur gradually with the accumulation of small impacts from educational interventions and lay media attention.

Adrenergic beta-Antagonists↗