This module relates to Section 5.7 of the Reviewers' Handbook and you should read this now.

Sift and sift again

We saw in the previous module, Searching for studies, that in a systematic review we attempt to find every study that has ever been done addressing our question. We try to do this by running sensitive searches. Inevitably, when we do this, we find lots of reports of studies that could be relevant, and we then have to decide which ones are relevant to our question.

It's important to remember that our decisions about which studies to include should be based on the design of those studies, and not the results. If we allow ourselves to be swayed by the results of the studies, we might exclude a perfectly eligible study because we don't like or believe the results.

If you have already run some searches, you will know that these sensitive searches usually turn up hundreds or thousands of records. Most of these will have come from an electronic database, and you'll only have limited information about the study, like the example below.

ID:CN-00240563
TI:Antibiotic prophylaxis of wound infections in skin surgery [see comments].
AU:Bencini, P. L., Galimberti, M., Signorini, M., and Crosti, C.
SO:Archives of Dermatology
YR:1991
VL:127
NO:9
PG:1357-60
AB:A controlled prospective study of 2165 outpatients undergoing skin surgery was performed to evaluate the utility and the effects of several antibiotic schedules for prophylaxis of wound infections. The patients were divided into four groups. Twenty-three of the 541 group A patients, given no antibiotics, had wound infections. Eight of the 542 group B patients, given systemic antibiotics from immediately after surgery until the third day, had wound infections. Four of the 540 group C patients, treated only with local sterile antibiotic powder sprinkled into the wound during surgery, had wound infections develop, and only one infection occurred in the 542 group D patients given systemic antibiotics from 2 days before surgery until the second day after surgery. This last schedule was the best for prophylaxis of wound infections in contamination-prone regions. Local antibiotic administration is a simple method for prevention of infections in routine skin surgery.
KY:Administration, Cutaneous; Adolescence; Adult; Bandages; Cefazolin; Ad [Administration & Dosage]; *Cefazolin; Tu [Therapeutic Use]; Female; Human; Incidence; Injections, Intramuscular; Italy; Ep [Epidemiology]; Male; Middle Age; Powders; Premedication; Prospective Studies; Skin; Pa [Pathology]; *Skin; Su [Surgery]; Skin Neoplasms; Pa [Pathology]; *Skin Neoplasms; Su [Surgery]; Staphylococcal Infections; Ep [Epidemiology]; Surgical Wound Infection; Ep [Epidemiology]; *Surgical Wound Infection; Pc [Prevention & Control]; Time Factors
DE:RCT.
CC:SR-HANDSRCH, SR-SKIN

It isn't usually possible to be certain from this sort of record whether it will turn out be included in the review or not. There usually simply isn't enough information to be quite sure. Try to list the extra information you might want to help you decide if this was a randomised controlled trial relevant to a review of antibiotic use for preventing wound infection in skin surgery.

Here are a few suggestions

• Was the study randomised and how (ie how were the patients divided into different groups)?
• What type of surgery was performed?
• How was resultant infection determined?
• Were all patients followed up
• Did they have any other type of intervention or co-morbidity that may have influenced the results?

What is normally possible by looking at these records is to tell whether the study might be relevant or is very unlikely to be relevant. You can then order the full paper copy of the ones that you think might be relevant. This saves you getting full paper copies of hundreds of articles that turn out to be irrelevant.

Once you have these hard copies of the probably relevant studies, you can get on with comparing them with your review's inclusion criteria. This second stage is where you make definite decisions about whether studies are included or excluded from the review.