All frameworks and cheat sheets

Bar chart

Rate ratio for death under a biased and a correct analysis

Bar chart of rate ratios for death: 0.73 and 0.78 under a biased cohort design, 0.98 and 0.94 once person-time is allocated correctly.
The same patients, two designs. Only the biased one makes drugs with no cardiac benefit look protective. Source: Samy Suissa, Pharmacoepidemiology and Drug Safety, 2007

Samy Suissa's 2007 study of 3,315 COPD patients in the Saskatchewan Health databases tested two drug classes with no known cardiac benefit. Under a cohort design with immortal time, both appeared to cut deaths, with rate ratios of 0.73 and 0.78. Allocating person-time correctly gave 0.98 and 0.94. Product teams can use it to show why a feature adoption chart can invent a retention effect.

Where this appears

What's in this bar chart

Biased cohort designCorrect person-time analysis
Inhaled beta-agonists0.730.98
Gastrointestinal drugs0.780.94

Neither drug class has a known cardiac benefit. 3,315 COPD patients hospitalised for cardiovascular disease.

Use this image

Free to use in your docs, decks and posts under CC BY 4.0, as long as you credit Koji with a link to this page. To embed it, paste this into your page:

html
<a href="https://www.koji.so/visuals/immortal-time-bias-rate-ratios-biased-vs-correct"><img src="https://www.koji.so/visuals/immortal-time-bias-rate-ratios-biased-vs-correct/immortal-time-bias-rate-ratios-biased-vs-correct.png" alt="Bar chart of rate ratios for death: 0.73 and 0.78 under a biased cohort design, 0.98 and 0.94 once person-time is allocated correctly." width="1200" height="630" /></a>
<p>Source: <a href="https://www.koji.so/visuals/immortal-time-bias-rate-ratios-biased-vs-correct">Rate ratio for death under a biased and a correct analysis</a>, Koji</p>