Sensitivity, Specificity, Positive Predictive Value, & Negative Predictive Value
Please review this Excel spreadsheet: Week 2 Epidemiological Examples (please see attached document)
In your post, you are to select a disease that no one else has selected. You can post early in the week to reserve your disease and then just reply to your original post when your material is done. Answer the following questions with your initial post.
1. What disease did you select? (Liver and pancreatic cancer
2. Find two different incidence rates for this disease. This can be either the rates of the disease in two different countries, age groups, ethnicities, occupation, or any other classifier that might be important to your disease. Provide the two different rates here. Cite your source.
3. Select a screening test or a confirmatory test for your disease. Report on what the sensitivity and specificity percentages are and cite your source. Also, provide the cost of the test.
4. Using the method outlined in the excel document determine the PPV and NPV of using both screening tests in the two populations you selected.
5. How much did each true positive cost?
6. Does it make sense to implement mandatory screening in either population? Why or why not?
Sen & Spec
| Example 1 | |||
| Disease (+) | Disease (-) | ||
| Test (+) | a (True Positive) | b (False Positive) | All Test Positive |
| Test (-) | c (False Neg) | d (True Negative) | All Test Negative |
| All Diseased | All Well | Total Pop | |
| Sensitivity | 90% | a/a+c | |
| Specificity | 95% | d/d+b | |
| Fake Data on an UNKNOWN DISEASE AND TEST | |||
| Disease (+) | Disease (-) | ||
| Test (+) | 475 | 4974 | 5449 |
| Test (-) | 53 | 94499 | 94551 |
| 528 | 99472 | 100000 |
What you see below is a 2 x 2 table. We will be using it to explain how to calculate sensitivity and specificity. Once that is explained, we will move on how to use sensitivity and specificity data along with incidence information to estamate how many people will be found using a screening program.
As you can see from our fictitious example, the fake screening test that we are talking about using would give 53 people a negative result when they were sick, 4,974 a positive result when they were not sick. This doesn't sound like a great test, but that is all dependent on the natural history of the disease, mortality associated with it, and the communibility.
Please continue to the next sheet labeled PPV & NPV.
PPV & NPV
| Example 1 | |||
| Disease (+) | Disease (-) | ||
| Test (+) | Sensitivity (a) | 1 – Specificity (b) | All Test Positive |
| Test (-) | 1 – Sensitivity ( c ) | Specificity (d) | All Test Negative |
| Incidence Number | Population -Incidence Number | Total Pop | |
| PPV = A/A+B | |||
| NPV = D/D+C | |||
| Step 1: Insert Sensitivity and Specificity | |||
| Disease (+) | Disease (-) | ||
| Test (+) | 90% | 1 – Specificity (b) | All Test Positive |
| Test (-) | 1 – Sensitivity ( c ) | 95% | All Test Negative |
| Incidence Number | Population -Incidence Number | Total Pop | |
| Step 2: Calculate C & D | |||
| Disease (+) | Disease (-) | ||
| Test (+) | 90% | 5% | All Test Positive |
| Test (-) | 10% | 95% | All Test Negative |
| Incidence Number | Population -Incidence Number | Total Pop | |
| Step 3: Incidence data | |||
| Pop A | |||
| Disease (+) | Disease (-) | ||
| Test (+) | 90% | 5% | All Test Positive |
| Test (-) | 10% | 95% | All Test Negative |
| 250 | 750 | 1000 | |
| Pop B | |||
| Disease (+) | Disease (-) | ||
| Test (+) | 90% | 5% | All Test Positive |
| Test (-) | 10% | 95% | All Test Negative |
| 3 | 997 | 1000 | |
| Step 4: Completing the table | |||
| Pop A | |||
| Disease (+) | Disease (-) | ||
| Test (+) | 225.00 | 38.00 | 263.00 |
| Test (-) | 25.00 | 712.00 | 737.00 |
| 250 | 750 | 1000 | |
| Pop B | |||
| Disease (+) | Disease (-) | ||
| Test (+) | 3.00 | 50.00 | 53.00 |
| Test (-) | 0.00 | 947.00 | 947.00 |
| 3 | 997 | 1000 | |
| Step 5: Calculate the PPV and NPV | |||
| Population A | |||
| PPV | 85.55% | ||
| NPV | 96.61% | ||
| Population B | |||
| PPV | 5.66% | ||
| NPV | 100.00% | ||
| Step 6: Finances | |||
| Cost of Fake Test | 25$ | ||
| Pop A | Cost per Positive | 1000 x $25 = $ 25,000 | $25,000/225 = $111 dollars per positive found |
| Pop B | Cost per Positive | 1001 x $25 = $ 25,000 | $25,000/3 = $8,333 dollars per positive found |
| Summary | Notice how cheap the cost per positive is when you screening in a population with a high incidence |
1. Sensitivity and Specificity calculations are always on tests for disease. You can go into any Pharmacy and look at their pregnancy, drug, or paternity tests and they have those numbers on them. The next topic is going to be how do we use that information in developing screening policies. All diseases occur in different populations at different rates, if a disease is higher in prevalence in a given population then their positive preditive value (finding cases) will increase. I am going to show you how to use incidence data along with sensitivity and specificity data to generate a PPV and NPV.
2. Our previous Sen and Spec Calculations were 90% & 95% respectively. Let's insert those into the appropriate charts based on the information we have.
3. Let's say you wanted to find out what the positive predictive values (PPV) and the negative predictive values (NPV) would be if you screened two different populations. All you have is incidence data on the disease estimates in the population now. Population A = 250 per 1,000 population Popuation B = 3 per 1,000 population Put the incidence data into the 2 x 2 table. 4. Multiply the percentages in A & C x the incidence total in A + C Multiple the percentages in B & D x the incidence total in B+D 5. Calculate the PPV and the NPV 6. Take the number of True Positives for Populations A and then Population B and multiply by the cost of our fake test.