Clinical trial simulation -- 2
Budget: $250 – $750 USD
I am looking for someone to develop a simulation for a book I am writing about evidence in medicine. Let’s assume a 19th century physician is searching for a cure for yellow fever. He has access to 100 different herbal preparations. One of them (let’s say #57) actually does help, i. e. it increases chances of survival 3x. Let's say 30-day mortality of yellow fever is 30%, and it goes down to 10% when this herb is given. The 99 other herbs have no effect. The question is, how long would it take this physician to determine which herb it is? The additional assumptions are as follows:
1. The physician can prescribe any one of the 100 herbal treatments to any patient, and they will accept and be 100% compliant.
2. 20% of patients will be lost to follow-up and their outcome will be unknown. For the other 80%, the physician will know whether they died or recovered from their illness.
3. The physician is following some kind of an algorithm in which they start by randomly selecting which herb to give to which patient, and over time they get somewhat biased by the outcomes they see. If one lifetime is not enough, the physician writes down their observations, passes on to the next generation and the search continues.
The output would be a plot where on the X axis is the number of patients treated, and on the Y axis is the probability (as calculated by the physician) of treatment #57 being the effective treatment. At the beginning, it will be 1%, but presumably after 1000, 10,000 or a certain large number of patients the confidence in #57 will grow and in the others will fall.
1. The physician can prescribe any one of the 100 herbal treatments to any patient, and they will accept and be 100% compliant.
2. 20% of patients will be lost to follow-up and their outcome will be unknown. For the other 80%, the physician will know whether they died or recovered from their illness.
3. The physician is following some kind of an algorithm in which they start by randomly selecting which herb to give to which patient, and over time they get somewhat biased by the outcomes they see. If one lifetime is not enough, the physician writes down their observations, passes on to the next generation and the search continues.
The output would be a plot where on the X axis is the number of patients treated, and on the Y axis is the probability (as calculated by the physician) of treatment #57 being the effective treatment. At the beginning, it will be 1%, but presumably after 1000, 10,000 or a certain large number of patients the confidence in #57 will grow and in the others will fall.