Failure of a patient-centered intervention to substantially increase the identification and referral for-treatment of ambulatory emergency department patients with occult psychiatric conditions: a randomized trial [ISRCTN61514736]

Table 3 Psychiatric diagnosis, consultation and referral in patients with a PRIME-MD diagnosis, by study limb (N (%))

	Study Limb
	Tell Neither	Tell Patient	Tell Both
Patients with a PRIME-MD diagnosis and a:	9	20	22
Physician psychiatric diagnosis	0 (0%)	2 (10%)	4 (18%)
Physician psychiatric consultation or referral	1 (11%)	3 (15%)	4 (18%)
Physician psychiatric diagnosis, consultation, or referral	1 (11%)	3 (15%)	4 (18%)
Frequentist 95% CI	0%, 48%	3%, 38%	5%, 40%
Bayesian Analysis of these data*	mean 95% Credible Interval
Prior distribution for limb	7% 0%, 23%	20% 0%, 87%	20% 0%, 87%
Likelihood for limb (from observed data)	11% 0%, 37%	15% 3%, 33%	18% 5%, 36%
Posterior distribution for limb	9% 1%, 22%	15% 4%,33%	18% 6%, 36%

*All probabilities are modelled as beta distributions. The traditional, frequentist 95% confidence intervals for the trial can be compared to the 95% credible intervals of the posterior distributions produced by the Bayesian analysis which combines prior beliefs about the results of each limb with the observed data (see text and [14]).

ISSN: 1471-227X