Thread by @DearPandemic

1/ Q: What’s my personal risk for a bad #Covid_19 outcome should I become infected?

A: A variety of new “risk calculators” can help you answer this very question (Links below).
BIG CAVEAT: There are huge margins of error on the results, often making risk scores LESS ACCURATE.

2/
Background: As several parts of the country #reopen, we need to make our own, individualized risk/benefit calculations. It’s a tough puzzle – we all have incredibly unique circumstances, constraints, and supports under which we’re operating.

3/ Both clinical medicine and health care policy rely on statistically sophisticated prediction models to estimate an individual’s risk for an outcome of interest (good or bad).

These models suggest answers to the Q: “What’s my personal risk for X outcome if I contract #COVID?”

4/ Much of the buzz around machine learning and AI in medicine relates to their improving the accuracy of person-specific “clinical risk scores.”
In fact, Nerdy Girl @lindsleininger spent a chapter of her career developing, testing, and translating such models for policy-makers.

5/ Many new #COVIDー19-specific risk calculators can meet individuals’ needs in making risk/benefit calculations during the pandemic (see table one): cebm.net/covid-19/what-…

6/ It’s so, so important note that these are built on new data that haven’t been widely validated. These tools perform best at the POPULATION level, as opposed to the INDIVIDUAL level.

7/ For example, we know that on average 100,000 people with preexisting heart conditions are more likely to have a poor #COVID__19 prognosis than 100,000 people with no conditions.
But for any one person within either of those groups there’s a large margin of error.

8/
If, on balance, you feel like having incomplete information is better than no information about your individual risk, then by all means check out the calculators. And note that it’s best to compare estimates from a variety of tools, versus having complete faith in one tool.

9/ Finally, look for calculators with LARGE underlying data sets that GENERALIZE WELL TO YOUR CONTEXT.

Here's the link again to the risk calculators (see table 1): cebm.net/covid-19/what-…