March 26, 2010 2:25 PM Subscribe

In epidemiology and/or medicine, is it called the "positivity rate" or the "positive rate"? Is one (i.e., "positive rate") a preferred term for laypersons? Also how do you calculate this rate? I've seen it done several different ways. Could you provide references to authoritative sources that define it?

For disease screening, the positivity rate/positive rate tells you what percentage of people have tested positive for a disease, but what is considered the correct denominator? Is it:

1. number of people submitting to a test

2. positive results + negative results

3. number of adequately completed tests (but including indeterminate results)

I would be interested in sources recommending proper use and wording for both expert and lay audiences.

(Would rather not link this question to my posting history. Thx.)
posted by *anonymous* to Science & Nature (3 answers total) 2 users marked this as a favorite

For disease screening, the positivity rate/positive rate tells you what percentage of people have tested positive for a disease, but what is considered the correct denominator? Is it:

1. number of people submitting to a test

2. positive results + negative results

3. number of adequately completed tests (but including indeterminate results)

I would be interested in sources recommending proper use and wording for both expert and lay audiences.

(Would rather not link this question to my posting history. Thx.)

I'm not exactly sure what you're looking for, but I haven't heard of positivity used in this context.

The percentage of people who test positive for a disease or condition is the prevalence of that condition. It becomes a valid population rate if the testing is performed randomly. In that case, the denominator is the size of the sample frame, i.e., how many people were sampled. The number of people in the sample tells us something about the accuracy or precision of the estimates, usually conveyed through confidence intervals.

So if people wanted to test the prevalence of HIV, for example, the rate would differ depending on who was being sampled - if it's a random sample of the population of a geographic area, the sampling would be done along those lines, and the denominator would be the size of the sample. If it was among a certain subgroup (e.g., gay men, or injection drug users), the denominator would be the size of that sample.

If it's based on people submitting for a test by showing up at a clinic, say, as opposed to a random sample, the rates are likely biased - often some kind of stratification is performed to try to adjust the results so that they might more accurately represent the population, if that's desired. OTOH, it could be that the clinic is just interested in the prevalence of the condition within the population of people using that clinic.

These rates are typically standardized to become comparable - commonly rates are given per 100,000 or per 1,000.

posted by jasper411 at 7:01 PM on March 26, 2010

The percentage of people who test positive for a disease or condition is the prevalence of that condition. It becomes a valid population rate if the testing is performed randomly. In that case, the denominator is the size of the sample frame, i.e., how many people were sampled. The number of people in the sample tells us something about the accuracy or precision of the estimates, usually conveyed through confidence intervals.

So if people wanted to test the prevalence of HIV, for example, the rate would differ depending on who was being sampled - if it's a random sample of the population of a geographic area, the sampling would be done along those lines, and the denominator would be the size of the sample. If it was among a certain subgroup (e.g., gay men, or injection drug users), the denominator would be the size of that sample.

If it's based on people submitting for a test by showing up at a clinic, say, as opposed to a random sample, the rates are likely biased - often some kind of stratification is performed to try to adjust the results so that they might more accurately represent the population, if that's desired. OTOH, it could be that the clinic is just interested in the prevalence of the condition within the population of people using that clinic.

These rates are typically standardized to become comparable - commonly rates are given per 100,000 or per 1,000.

posted by jasper411 at 7:01 PM on March 26, 2010

The terms you've pulled are lay terms, not epidemiologic ones, and it does seem likely you mean positive predictive value. I would think it is easier for a lay audience to comprehend positive predictive value with the thirteen words: "the proportion of all people with positive tests who truly have a disease" rather than try to capture the meaning using a different pair of words that are equally incomprehensible. Using those 13 words, the answer to your denominator question becomes much more clear: the denominator is the number of people who tested positive.

Another good resource is the Supercourse. Here is one of several powerpoint presentations that define and illustrate some of the concepts you are trying to understand.

As an aside, none of these are likely to be true rates, but rather proportions. A rate usually would have time in the denominator.

posted by gubenuj at 8:29 PM on March 26, 2010

Another good resource is the Supercourse. Here is one of several powerpoint presentations that define and illustrate some of the concepts you are trying to understand.

As an aside, none of these are likely to be true rates, but rather proportions. A rate usually would have time in the denominator.

posted by gubenuj at 8:29 PM on March 26, 2010

This thread is closed to new comments.

True positive rate: out of the tests you make, how percent of the results that show true that are correctly true. So if you are screening for a disease, it is the percent of the tests that show positive that come from people who actually have the disease. A false positive rate would be the percent of tests that show positive that come from subjects who are disease free.

Positive predictive rate: Given a positive result on a test, the chance that the person who took the test has the disease.

posted by procrastination at 3:23 PM on March 26, 2010