Suppose a computerized database contains all charts of patients at nine hospital

Question

Suppose a computerized database contains all charts of patients at nine hospitals in cleveland, ohio. one concern of the group conducting the study is the possibility that the attending physician underreports or over reports various diagnoses that seem consistent with a patient's chart. an investigator notes that 50 of th 10000 people in the database are reported as having a particular viral infection by their attending physician. a computer using an automated method of diagnosis claims that 68 of the 10000 people have the infection, 48 of them from the attending physician's 50 positives and 20 from the attending physician's 9950 negatives



Test the hypothesis that the probability of detecting this viral infection are the same for the computer and the attending physician.

Explanation / Answer

One might immediately think of testing the 50 versus 68 using the standard chi-squared test. However, we can have a more powerful result by using the fact that there were different diagnoses on 22 people. For these, the computer diagnosed the disease in 20 and the physician 2.
Under b(22,.5), the probability of 0-2 diagnoses by the physician is
P(0) = .5^22
P(1) = C(22,1).5^22 = 22*.5^22
P(2) = C(22,2).5^22 =231*.5^22

P(0) + P(1) + P(2) = (1+22+231).5^22 = 254/4194304 = 127/2097152 0.0000605583190917969

Doubling this to get the p-value for a 2-sided test, you would still have a p-value of 0.000121116638183594, which means you would reject the null hypothesis of equal rates of diagnosis as long as > 0.000121116638183594

As = .05 is standard, you reject the null hypothesis of equal rates of diagnosis.

Related Questions

Navigate

• Browse (All)
• Browse S
• Subjects

---

---

Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.