not sure if you have data on the source population that generated the deaths or just  a clinical series of death cases.

if the former you could do retrospective cohort or case control or both (case control within cohort).

if the latter the only approach would be to compare mean risk factor levels or risk factor proportions between the abusive deaths cases and non abusive deaths cases..

They are all deaths that have autopsies at the Chief Medical Examiner and all the deaths that we review as part of the State's Fatality Review Panel. So unless an infant dies out of state it is likely that we have pretty much all infant deaths in the state of Connecticut during that 6 year time period.

So it sounds like I could do a retrospective cohort study, correct?

Thanks

yes. but you would have to collect source population exposure/confounder data from the whole of Connecticut over the relevant time period..

a case control study would require selection of controls ( live infants) from Connecticut,,preferably age-sex matched ,and ascertaining exposur/confounder data on the cases and these controls.

each type of death would be a separate case control  study

What are you trying to answer? Given that this is a "rare" outcome the case control may be better.

I meant that you could do a retrospective cohort study (in answer to the final question in Kirsten Bechtel's  second post  viz "  so it sounds like I could do a retrospective cohort study,correct?" ,

However this  would require more data collection.

In a case control study, the non case controls  effectively "sample" the source population exposure distribution abd hence this approach might be more feasible.

Sorry, I have no experiences in this field

Rotman says " a cohort study is not always perspective and can also be retrospective, also know historical cohort studies ( Rothman K; An introduction. 2002. page71)

 KIrsten

If you are collecting data on dead infants only without having information on the reference population, Then this is not a retrospective cohort study.  I guess you have data on dead infants  classified as abused and non-abused. Then I suggest that you have three alternative designs to do:

1. Compare the abused with the non-abused dead according to various characteristics and conditions surrounding their death including their age and sex.

2. Use a case-control study in a similar way as in 1 above but you need to make additional inquiry about other possible risk factors which were not recorded . This is in my view is the best one at your disposal at this stage of the study. Mind you, you may add a third group as second control from infants who did not die yet. 

3. The difficult design is to make a retrospective cohort study. You need to identify infants born in the population under the catchment area of dead children. 

Catchment population---Live births- during a given period of time ------Fate within a time span (dead or alive and abused death or non-abused death.

With best wishes