A disease surveillance system should be simple, flexible and acceptable to the population. For example, to detect hunting-related shooting injuries, the requirements for a hunter to report an episode should not be onerous or many shooting injuries will go unrecorded.. The surveillance system should also be representative of the population and provide a timely alarm. Like a smoke detector without a power source, a surveillance system that is not able to recognize a disease outbreak quickly and accurately is not very useful.
Evaluating the effectiveness of a surveillance system is easiest if there is a ‘gold standard’ that allows the researcher to compare surveillance results to the true condition. If it can be known whether or not the condition is present, a 2 × 2 table can be useful as shown below.
The count of true cases that are detected by the surveillance system is given in cell A. Similarly, cell B contains the count of non-cases that the surveillance system mistakenly considers as disease. The row totals are A + B and C + D, the numbers detected and not detected by surveillance. Given these data, sensitivity and positive predictive value can be calculated.
|Detected by Surveillance||Yes||
|A + B|
|C + D|
|A + C||B + D||Total|
- Sensitivity: the proportion of true positives that are detected by the surveillance system A/(A+C)
- Positive Predictive Value: the proportion of true positives among those detected by the surveillance system A/(A+B)
You have now completed reading the course notes for Week 4....show what you have learned in answering this week's discussion question.