Pragmatic randomized trials using routinely collected health data: meta-epidemiological study

Lars Hemkens, Kimberly McCord

Status: Abgeschlossen Dauer: Feb. 2016 - Jan. 2017

Many highly important healthcare questions are poorly addressed by randomized controlled trials (RCTs). Often RCTs are too short and/or too small to evaluate patient-relevant outcomes (such as mortality), even when treatments are evaluated that aim to improve survival (e.g. in cancer research). The main reason appears to be the high costs for adequately powered RCTs. However, large and simple pragmatic mega-trials may be conducted with only a fraction of costs. This includes simplification of trials and inclusion of routinely collected health data (RCD) to collect outcome information (e.g. patient mortality). RCD includes disease registries, electronic health records, prescription data or increasingly mobile health data collected by wearable devices. Examples of successful use include the use of national computerized record-linkage systems to obtain 10-year follow-up data of 12000 patients in a study that initially collected only 5 year data. Recently, our working group completed a cluster RCT in 3000 Swiss physicians treating more than 90% of the Swiss population using RCD to measure intervention effects. Various initiatives promote the use of RCD with enormous public investments. For example in the US the National Patient-Centered Clinical Research Network received $93.5 million by the Patient-Centered Outcomes Research Institute (PCORI) for establishing RCDs.nHowever, the validity of RCD may be reduced by errors and biases during collection and processing of data, linkage of databases, and misclassification bias and underreporting. The relevance of such issues in the trials has not been explored and there is no empirical research on this emerging research field. It is unknown which outcomes and interventions are explored using RCD, where this research is used (e.g. in which nations, or setting, for example university hospitals or primary care), how specific sources of bias using RCD are recognized, discussed and addressed by trial investigators, and which potential strengths and weaknesses are described.We therefore aim to conduct the first systematic review of RCTs using any type of RCD for outcome assessment. The objectives are to (1) describe the current research agenda in this novel field, (2) systematically extract challenges reported by the primary investigators and issues of RCD-related biases, (3) identify potential solutions, (4) provide empirical evidence for the development of specific reporting guidelines to improve this highly innovative research field.

Zur Projektübersicht