PAGES:  [BACK to PROTEUS home page]    Back to CKHS100 MAIN PAGE     Outline     Research protocol

SPECIAL TOPICS:   Histograms/Central tendency    Sensitivity/Specificity   Rates   Critical review 

               Evaluation Research   Definitions    Designs  Sample size  Sample methods  Confounding 

REMEMBER THAT IN CASE-CONTROL STUDIES (WHICH REALLY

SHOULD BE CALLED CASE-COMPARISON STUDIES), YOU START

WITH ILL AND WELL GROUPS AND GO BACK TO EXPLORE POSSIBLE

EXPOSURES. 

IN COHORT STUDIES YOU START WITH HEALTHY EXPOSURE AND

NON-EXPOSURE GROUPS, AND FOLLOW THEM TO DISCOVER THE

INCIDENCE OF ILLNESS OVER TIME.

BUT IN A CASE-CONTROL STUDY YOU CAN NEVER "JOIN" THE ILL

AND WELL GROUPS FOR ANY REASON BECAUSE THEY ARE NOT

PROPORTIONAL TO EACH OTHER.  HERE'S AN EXAMPLE.

KAWASAKI SYNDROME IS A VERY RARE CONDITION.  FEWER

THAN 1 IN A THOUSAND INFANTS ARE AFFECTED. SO IS IT

REASONABLE TO ASSUME THAT 14 (11+3) FAMILIES WHERE

CARPET SHAMPOO WAS USED WOULD SEE A CASE?  OR THAT

12 OUT OF 39 FAMILIES WHERE NO CARPET SHAMPOO WAS USED

WOULD ALSO SEE A CASE?  NO OF COURSE NOT.  CASES AND

CONTROLS WERE ARTIFICIALLY SELECTED AND PLACED IN THIS

COMPARISON, AND WE CANNOT ASSUME "INCIDENCE RATES" 

 WHEN WE COMPARE CASE-CONTROL AND COHORT STUDIES, THE C-C SEEMS BETTER IN MOST CATEGORIES EXCEPT IN THE CONSISTENCY AND RELIABILITY OF DATA AND CONTROL OF BIAS (THE RED CIRCLE).