Use of Patient-Specific Information for Randomization in Clinical Research: A Randomized Trial

Le résumé fourni par la source

Objectives: Technology companies conduct tens of thousands of automated experiments each year using A/B testing to optimize their products. Their methods rely on field experiments incorporated into customer-facing products. In medicine, pragmatic trials do the same by performing experiments alongside usual care. This study aims to assess the effectiveness of using patient-specific information for study group randomizations. Methods: We developed an electronic medical records-based randomization method (EMRRM) that used parity of encounter identification (ID) and patient ID numbers to separate patients into 2 random groups. This method was retrospectively applied to patients in the outpatient, inpatient, and emergency department settings. To assess the randomness of group assignments, we used the National Institute of Standards and Technology's Special Publication 800-22 (NIST) statistical package for randomness testing, and A/A testing for group similarities. The main outcome measure was the EMRRM's ability to perform group randomization as assessed by NIST and A/A test results. Results: For each of the clinical settings, encounter ID and patient ID methods passed either 10 or 11 out of 11 applicable NIST tests. On A/A testing, both methods successfully randomized patients into similar groups with equivalent demographics, laboratory tests, lengths of stay, and hospital admission rates. Conclusions: This study validated the use of encounter ID and patient ID-based EMRRMs in pragmatic studies. Patient information-based EMRRM can be used for automatic random study group assignments in pragmatic trials.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.