1. Seventy patients had only a 48- to 72-h reading of their skin test. Of these, 18 were positive and 52 were negative, yielding a prevalence of true-positive tuberculin reactors of 25.7 percent. If this percentage is extrapolated to the whole study population, then 25.7 percent of 1,412 or 362 patients would have had positive TBN-ST if the whole population had been read at 48- to 72-h. Since our method detected 259 positive reactors, the estimated sensitivity of our screening method is 259 of 363 or 71 percent.
2. We can also arrive at an estimate of the sensitivity of our methods by focusing on a subgroup that was read as negative prior to 48 h and then had readings at 48- to 72-h. Note that the following results are derived from only 1,232 of the 1,412 patients who had a TBN-ST because for 180 patients, the exact time of the test or the reading of the TBN-ST, or both, were not available. There were 131 patients who had an initial negative reading and a subsequent 48- to 72-h reading. Eleven of these 131 patients had negative readings prior to 48 h but had 10 mm or more of induration at 48- to 72-h. Therefore, this group had a false-negative rate of 11 of 131 or 0.084. If this rate is extrapolated to the 840 patients who had negative readings prior to 48 h and did not have a 48- to 72-h reading, we obtain a prediction of 0.084 X 840, or 71 false-negatives. We detected 214 positives in this group; therefore, our estimated sensitivity is 214/ (214 + 71) or 75 percent. The estimated true prevalence of reactors in our patient population is then (214 + 71)/1,232 or 23.1 percent.