Saturday, May 7, 2016
I wanna get technical: notes on writing a systematic review
Just recently I had the pleasure of clicking "submit" on the longest academic paper I have written so far-nearly 30,000 words long. It is a systematic review of cognition and biomarkers of stress in dementia caregivers. It's an important topic, but today I'd like to talk specifically about the method of doing a systematic review and what that entails.
The process of systemically reviewing the literature allows you to take steps to ensure that a review of research literature is not an incomplete or partisan exercise. However, writing a systematic review, particularly a large systematic review, can really take the pleasure out of reading a research article. Interesting discussion sections are raced past when one has over 100 articles from which to extract key methodological details. Here's what you do:
Firstly, one needs to see what literature is out there. After making searches for caregiver stress and a host of stress biomarkers (e.g. cortisol) on a number of research databases, the second reviewer and I found ourselves with circa 4,000 abstracts and titles to look though. Some of these were almost immediately identifiable as not really relevant to the review in question. However, some required a pretty careful reading of the abstract, and about for about 250 we had to look through the full text of the article. For example, we were interested in the impact of dementia caregiving stress on cognition, but it might not be immediately apparent if cognitive assessment was performed on the carers or just the patients with dementia.
Once you've whittled things down to just the articles you want to include in the review, it's time for data extraction. You get an Excel sheet with a whole host of bits of info (e.g. inclusion criteria, sample size, age/gender of the participants, location where the study was run, etc. etc.) and fill in this form for each study. I should say that in the case of this review we had 151 full papers to go through. A mountainous (albeit physically sedentary) task. The second reviewer blasted through the task quite a bit faster than I did-I'm going to assume it's because she is more experienced at conducting systematic reviews, as the alternatives are not so good for my ego.
Anyways, when the data extraction is done you THEN need to assess the quality of the research literature. Another template is put together in an Excel file and again you wade through the articles, ticking boxes about whether or not the articles have fulfilled certain criteria (did they control for potential confounding variables, was the statistical analysis appropriate). In fairness, there's often some overlap here between the data extraction, although not enough in this case to avoid going back through the articles again!
THEN you need to write the thing. When you have results from over 150 research articles to summarise, the mere process of checking the formatting within your tables becomes a dull and onerous task. Uughhhh. Still, when it's finished you can say that you've gone to lengths to avoid being selective in either assessing or reporting the published research in the field.
Obviously this is one account-there are more/less complex ways of doing a systematic review. Generally there will not be quite so many articles included in a systematic review, and if the outcome is more focused one may wish to do a meta-analysis. I think as time goes by more journals will start to strongly encourage or demand that review papers are systematic. All I can say about my near future is the next couple of reviews I write will not be systematic in nature.
If/when the article of this systematic review is published I'll be letting you know about the results ;)