TimeCodeWAS is a tool for exploring time-dependent associations between an endpoint and health codes (diagnoses, lab tests, procedures, etc.).

The TimeCodeWAS analysis is done in three steps:

1. Build a matched case-control set for this endpoint. Each case is matched to 5–10 controls on sex and year of birth, and controls must be at risk at the same age as the case's age of onset. TimeCodeWAS allows for controls to be reused.
2. Count cases and controls. For each health code, and for each time window: tally the counts into a 2×2 table. To be eligible in a time window, a case or control must have data coverage for at least 50% of that time window.
3. Perform association tests. If all cells of the 2×2 table have N≥10, then a χ² test is performed, otherwise a Fisher's exact test is done. The corresponding p-value and confidence interval are reported below, along with the standard odds ratio when a χ² test is done, or Fisher's MLE odds ratio otherwise.

TimeCodeWAS currently uses the same default set of time windows for all the endpoints. Please reach out to us if you have suggestions for better time windows for some endpoints. The default time windows are: