L6 cohort studies

Question 1

LO2: comp IR/MR btw 2 grps within chort by calc IRR. INT COMPAR.

Answer 1

• rate for exposed and unexp subcohorts. IRR. ef as subj to rand var. use obs and ef to estim likely true rel risk. incr no exposed= big decr ef. just incr no unexp= small decr ef. 95% CI.
  probs: subcoh can be diff sizes eg exposed minority. so often req large study. subcoh may not be comparab wrt other vars-confounding.

Question 2

LO3: comp IR/MR in coh with that of ref pop using stand meths eg SMR. EXT COMPAR.

Answer 2

-SMR=O/E. ef.
calc E cases- coh studies us lt. peop age dur and age specif I/MR in ref pop change.
-lexis diag- calc sep E for each age grp in each time period. obt age specif rates ref pop for each time period from routine data. x these rates by approp cells p-y=estim E cases in each cell. E eachc ell summed over all age grps and for all time periods=tot E.
can also add variabs eg age sex specif rates at each time period. lim to vars in routine sources.
ADJ for age and time!
-lims: usef when cant use subcoh BUT-
often lim data av for ref pop. often no incid data, us use MR. study and ref pop may not be comparab- selec bias= distort of ev/data fromw ay its collec. eg healthy worker eff. many occup coh SMR under 100 as employ healthy.

Question 3

LO1: desc logist basis of, and practic probs inv in cohort studies.

Answer 3

-comp incid based on roeutinely av data. eg study aetiol. LIMS.
ideal- 2 identic grps diff only in expos of int. but expos linked to other facs.
next best rand controlled trial- force with consent.
next cohort- no force. meas and rec non ident feats. adj for.
-cohort study- exposed and unexp count outc events and p-y. longit.
conhort study= id grp and foll over time see if expos aff outc. norm eff of susp RF that cant be exp contr eg smok.
-p-p= sum tot time of ev foll up in study. count events. account for diff foll up periods.
-cohort study start with dis free. lt foll up. comp risk acc to expos./personal chars. anal.
cohort= defined grp known expos/personal chars. risk estim as IR each grp. rel risk.
-compar by- subdiv expos levs int to cohort. IRR.
with ext ref pop. SMR for rel risk.
-adv cohort over routine data- can study expos and perosnal chars not routinely collected. detailed info on outc and expos. addit data on confound facs.
-concurr or prospec cohort study- data collec immed or later. to make sure no one gets dis eg first 5yr post recruit ie peop undiag.
-historic or retropspec- foll up data from past. recruit out free indivs and classif init expos status and subseq outc status using historic recs. fol pu- count p-y and events.
-expos data- binary/ sev categs/ contin.