REGISTRY AND ANALYSIS OF COMPUTER TRIALS IN PRIMARY CARE

Computerized information systems could greatly improve primary health care delivery. However, claims of benefits are often based on non-experimental evidence. Randomized controlled clinical computer trials (RCTs) can rigorously assess the value of computer technology in practice. A review of these trials would provide a useful evaluation of the current status of this technology. Unfortunately, several obstacles exist that make a review of computer RCTs difficult. These include the large variety of medical and information science publications in which reports may appear, non-standard descriptors, incomplete indexing, and submission/selection bias against the null hypothesis. The purpose of this project is to provide a comprehensive database and review of computer RCTs to assist the selection of computer services with demonstrated ability to improve the quality of patient care. A database of computer RCTs is proposed to register published reports of trials, unpublished trials, and reported ongoing or planned trials. The registration will follow a stepwise protocol: 1) Systematic search for trial reports. Several methods will be combined to ensure completeness: (a) MEDLINE and other database searches with various strategies, (b) hand search of selected books, before 1966 journals, citations in articles reviewed, and proceedings (abstracts), (c) other informal methods (e.g. letters, electronic mail, meetings). Only reports on clinical trials evaluating effects in the process or outcome of patient care will be collected. 2) Registration of retrieved trial reports. Using specific eligibility criteria a decision about registration (inclusion/exclusion) of the retrieved trial reports will be made. In case of ambiguity the authors will be contacted. Registered reports will be entered into a database supporting linked access to repeat publications and trials testing similar interventions, measuring similar outcome parameters, or using similar subjects. 3) Blinded classification of the reported trial features. Considering the potential bias of conventional reviewing, a blinded review method will be applied to analyze the methods and results of the large variety of computer trials. The procedure will be based on independent evaluations and consensus development. The review will result in a classification of computer RCTs according to the method of intervention, site of the trial, study population, treated condition, and the outcome of the trial.