Group D7 Ashim Sarkar (11065) Nikunj Chhaochharia (11212) Priyanshu Dubey (11244) Shah Rishabh Rajendrakumar (11308) Shruti Marwaha (11321) Vijaya Sundara Pandiyan G. (11363)
Problem Statement The U.S. Public Health Service wants to reduce the number of tests required to determine whether a person called for induction has syphilis. It is required to determine whether dividing the people into groups will result in lesser number of tests. If yes then what should be the optimum group size. Methods Case 1 Every person is checked Case 2 Blood samples are pooled in groups Groups are tested for syphilis If a group tests positive then every member of the group is checked individually for syphilis. Variables We define the following variables n The total number of people to be tested p Probability that a person has syphilis. x number of people in a group Thus the number of groups will be n/x Binomial Distribution The three conditions required for binomial distribution are: Trials must be independent Each trial results in success or failure Probability of success(p) must be same
We see that in the problem the first 2 conditions are satisfied. Binomial (cont.) The probability p of getting syphilis varies based on age and social circles, it is assumed to be constant for calculation purposes. Thus all 3 conditions are satisfied and we can apply binomial distribution. Solution using Binomial For case 2 P(a group is investigated further) = P(at least one person has syphilis) = 1 P(no person has syphilis) Using binomial =1 C 0 p 0 (1-p) x =P i
P i is the probability that i th group is investigated further. Solution (cont.) Expected number of groups investigated further:
(n/x)*P i
(since expected value of a binomial distribution is given by N*P where N is the number of trials and P is the probability) Solution (cont.) No. of investigations to be conducted: (n/x)*x*P i + (n/x) = n*P i + (n/x) (1)
If case 1 is followed(no grouping) then number of tests will be : n (2)
Thus, reduction in tests is :(2) (1) =n - n*P i - (n/x) Solution (cont.) Thus percentage decrease in tests is:
1 P i (1/x)
Based on this formula, the results are tabulated in a data table for different values of p and x. ps.xls
Actual Data from Reference 1,070,000 men were examined 83,000 were found to have syphilis
Thus probability of a person having syphilis is (83,000/1,070,000) = 0.077
Reference: www.cdc.gov/stdconference/2004/Slides/B-sessions/B2/McGough.pps Decrease in tests with group size 0.00 10.00 20.00 30.00 40.00 50.00 60.00 0 5 10 15 20 25 30 X axis: group size Y axis: % decrease in no. of tests 4 Result We observe from the tabulation that as p increases, the group method will need more tests as compared to individual tests. If p>0.3 then we should follow case 1. From the graph and from the tabulation, we observe that we get maximum reduction when x=4, for p=0.077(from reference). Thus the group size should be 4. Validation From the tabulation we see that a group of 4 members gives maximum decrease for a probability of 0.077
Let the No. of people (n) be 200,000 Let there be 4 people per group (x) Number of groups is 50,000 Probability (p) of syphilis is 0.077 Validation If everyone is tested (case 1), No. of tests will be 200,000 With groups of 4 (case 2), the number of tests will be: = n*P i + (n/x) Where, 1 C 0 p 0 (1-p) x = P i =104,843 Thus we have almost 47.5% reduction
Conclusion The method of segregation into groups is better for low probability of a person having syphilis. With increase in probability (p>0.3), the method of testing individually is better. For the calculated probability of 0.077, the most efficient group size is of 4 members. Thank You