I realize this question may sound weird (to say the least)... but I was always wondering if one took all the reported literature cases for a specific disorder / disease (assuming that there would be 1,000's of them) -- would the collected sample actually approach the true disease population? Would there still be significant reporting bias? What do you think?
Very interesting question, Stanislaw!
I would argue (based on statistical theory), that if the records (what you refer to as literature cases) were randomly drawn from the greater population, then with increasing size they should approximate the true population's values.
That said, the question is if the cases drawn from the literature are truly representative of the overall population's value? Assume that some (if not all) studies had inclusion/exclusion criteria that would limit generalizability? Thus, aggregating all the cases would still not represent the true population's values.
Interesting question! Do a review about it!
Case reports are usually written to point out special cases - therefor the 1,000 cases won't represent the population, but will show the full range of appearences.
Please let me know about the publication of your review.
There would be reporting/selection bias for sure, but you'd come closer to approximating the true disease population with a larger sample size. What are you thinking about specifically?
It's always been on my mind to create large databases of case reports, focusing specifically on clinical presentations, management, and outcome(s). I realize there would be potential biases, but I don't think it's actually been done before. Also, combining this with large series would, in my estimation, result would start approximate population after several hundred cases... just a thought...
I don't believe that several hundred people would approximate a diseased population. If you consider the complications associated with the target condition and then add comorbidities, it would likely take several thousand patients to encompass the entire array of possible states.
All that being said, this entire endeavor is currently possible given that data is now being entered in electronic medical records (not every practice as of yet, but it will be the standard within a few years, and meaningful use criteria will force the issue). Theoretically (and practically) the relevant clinical data should be recorded in the EMR and retrievable.
Thus, I anticipate that soon enough we'll know a tremendous amount about diseases that we don't know currently
Ariel,
I was thinking along exactly the same lines, with an added caveat that pre-EMR reports could then be retrospectively added to the overall database to illuminate the particular disease state in question in even greater detail...
Thanks!
Case reports are picked for publicaton bias. They do not always represent a disease but usually an unusual occurence of the disease process.
Case reports are presented to clinicians accidentally, sometimes could be reported before with another diagnosis, we had seen such conditions before, so case reports represents itself, not the prevalence of the disease.
- Badges
- Science topic