Purpose: Why clinicians are reluctant to treat smear negative pulmonary tuberculosis? In discussions during training in clinical reasoning doctors say they fear that, when the threshold is lowered by half, they will treat twice as much patients, of whom a lot without the disease. This seems to indicate that clinicians intuitively imagine a linear, flat distribution of post test probabilities in a given patient population. But, in reality, the higher the accuracy of the test or the combination of tests, the more patients would be distributed at two opposite sides of post test probability, leaving a middle field empty. (figure 1)
Method: We asked 38 participants and 3 tutors in a 6-week workshop on clinical research and evidence based medicine their idea about the distribution of post test probabilities. Participants came from four continents. Forty were medical doctors or specialists, 1 pharmacist; 23 were involved in academic teaching in their countries. All had recently revised the theory of clinical epidemiology, including distribution patterns.
Result: Eleven supposed a normal distribution, one a flat one, 12 a distribution skewed to the right and 15 a skewed to the left. Only 2 participants designed a U-shaped distribution.
Conclusion: We conclude that misconceptions about post test probability distributions might have a negative influence on the application of threshold and expected utility theory. Training in medical decision making should include the impact of test accuracy on the actual distribution of post test probabilities.
Figure 1: histogram of post test probabilities of children suspect of tuberculosis, after diagnostic workup. (based on a cohort used in “Effect of applying a treatment threshold in a population. An example of pulmonary tuberculosis in Rwanda. Van den Ende J, Mugabekazi J, Moreira J, Seryange E, Basinga P, Bisoffi Z, Menten J, Boelaert M. J Eval Clin Pract. 2010 Jan 14.” ) 
See more of: The 32nd Annual Meeting of the Society for Medical Decision Making