Article Text
Abstract
Introduction Using a study-based register in the process of systematic reviewing reduces waste and makes it possible to shortcut many processes normally undertaken by review teams. However, this works for simple ‘Intervention X vs Intervention Y for Condition Z’-style reviews, but the challenge is to provide the same shortcuts for systematic reviews of classes of interventions, overviews or network meta-analyses. As one might expect an Information Specialist to say, classification is the answer.
Objectives To report experience and progress with specific classification of healthcare conditions, interventions, and outcomes for the purposes of facilitating systematic reviews.
Methods We used the study-based Register of Cochrane Schizophrenia Group (in MS-Access MeerKat 1.6; holds 25 212 reports of 18 105 studies – 28 Feb 2018). The PICO meta-data (health care problems, interventions, comparisons, and outcomes) of each study have been extracted. We used NLM’s MeSH, The British National Formulary, and WHO ATC classification system.
Results Health care problems: In the 18 105 studies we identified 266 health care problems within schizophrenia trials which were specific focus of the evaluation – amongst which negative symptoms (546 trials), treatment resistance (467 trials), depression (350 trials), tardive dyskinesia (293 trials) and weight gain (260 trials) were the most common.
Interventions: Of the 3910 interventions randomised within these trials, we found 155 classes of drugs with antipsychotics, antidepressants, and benzodiazepines being the most researched. There are 41 additional specific interventions related to some sort of physical/exercise approach. Classifying psychological interventions, and Chinese Traditional Medicine (with its 537 trials with 246 interventions) remains a challenge.
Outcomes: We use seven main classes for outcomes within schizophrenia reviews: Global State, Mental State, Adverse Events, Functioning, Service Use, Quality of Life, and Cost. We propose to use existing classification of outcome tools to clean and curate the 13 187 outcomes. Classification heaven!
Conclusions Better reporting of PICO meta-data would help and improve classification. However, all current classification systems do not really fit the systematic review purpose. New systems, designed with systematic review output in mind, greatly enhance the review process (including prioritisation of titles) and reviewer experience (including prioritisation of effort).