[Confounding bias due to indication and severity in observational studies]

Abstract

Observational studies are subject to biases that may lead to misinterpretation of the results. This study aimed to determine the influence of omeprazole treatment on the duration of sick leave in patients with ankle sprains treated with non-steroidal anti-inflammatory drugs. We used the Ibermutuamur database. Contrary to our expectations, sick leave was longer in patients who received omeprazole than in those who did not. These findings were probably due to the influence of a bias due to confounding by severity, given that patients who received omeprazole had a worse kind of ankle sprain; however, a bias due to confounding by indication cannot be excluded. To avoid the influence of these systematic errors, biases should be monitored from the design stage to the data analysis stage.