Notebook AtlasSURG0002 · Research methods: go to the course map

AtlasSession 3L7: Systematic errorsp.30–33, 78–79

Selection bias

  1. Study designs primer
  2. What is systematic error
  3. Selection bias

Fig. 1

Retrospective knee cartilage registry: KOOS (0–100) at 2 years, collagen scaffold A vs B

Fig. 1. Two groups of 60. 12 early scaffold A failures were never scored. Average score 78.00 vs 72.00.

Scaffold A average
78.00
Scaffold B average
72.00
Left out of scaffold A
12 of 60

Trial Design Board · First look

12

Level 1 · First look · Lecture 7 · p.30–33, 78–79

See how leaving out the people who did badly in one group flatters that group.

Recognise selection bias (eligible people, early follow-up or outcome-related events excluded selectively) and tell it apart from a generalisability limit applied to both groups.

  1. Exit-survey trap (the step you are on)

    Asking only customers still subscribed whether they like a service; the unhappy ones already left.

    Ask only the customers who stayed whether they like a service and it looks loved: the unhappy ones already left. Here 12 of 60 scaffold A patients failed early and were never scored.

step 1/3