Good Morning,

Looking for some input into problems and special considerations that need to be taken into consideration when looking to do cross sectional (retrospective) in the primary care setting such as the prevalence of leg ulceration in substance misusers. Any special consideration apart from consent, confidentiality, vulnerability etc that needs to be taken into account?
I understand that anything retrospective relies on people remembering the events. I am thinking maybe mixed methods approach would be suitable for this? Any thoughts would be much appreciated

======= Date Modified 08 May 2012 22:45:32 =======
Hi Leergn

The main issue when trying to identify prevalence of anything from a cross sectional study, is working out how generalisable your sample is to a broader population. For example, if you want an estimate of national prevalence, studying practices just in one region could give you a biased result. But if you're only interested in a region, then a decent number of practices in that region might be enough to give you a good estimate. Some kind of sample size calculation would be a good idea once you have a clearly defined research question.

Are you collecting data yourself, or working with existing sources? Or both? Another issue with cross sectional studies is it is difficult to infer any kind of causal relationships if the outcome and exposure data were captured at the same time, unless there is some kind of clear temporal ordering of things.

If you're using existing primary care data, don't jump in straight away. Take time to 'get to know' your dataset as it is likely that you will need to do data cleaning. Talk to other users and find out what coding issues there are etc. Getting reliable case definition from primary care data can be challenging, particularly in conditions that may be comorbid, or where the condition may have been diagnosed/treated in secondary care and then referred back to primary care for follow up.

======= Date Modified 09 May 2012 08:43:09 =======
============= Edited by a Moderator =============
*Removed by PostGrad Forum team - SPAM*

Good Morning Hazyjane!

Many thanks for your advice and assistance. I will hope to use data already collected in the medical records but construct a questionnaire to be sent to generate quantitative data on the prevalence in a set area. I agree with you point; that bias is a risk in this area but due to it being set to a specific area I think it will help avoid this. Definitely sample size calculations needs to be done using probability sampling perhaps? I think it is important to give everyone the chance to be included in this study to avoid any selection bias; do you agree?
Thanks for your help about data cleaning; I will definitely get in touch with other users about coding problems etc.

Many thanks indeed

Lee :-)

Yes, if you say it's a prevalence estimate in a specific region/context that's fine. I must admit I'm not completely clued up on sampling methods/sample sizes for this kind of thing but I'm sure there will be advice out there.

Even if you want to capture as wide a sample as possible, it is still advisable to state inclusion/exclusion criteria from the outset to show that you've thought things through. Even if they seem like obvious things (e.g. over 16s only) it's worth stating them clearly for any organisations you approach.

I'm not an expert in this area but I suspect that primary care data will underestimate prevalence as I would guess that some of these cases might end up being treated in A&E, bypassing GPs entirely. In fact the likely differences you will find according to who you ask is an interesting point in its own right.

Good luck with your research :)