Critical issues in longitudinal and observational studies: purpose, short versus long term, selection of study instruments, methods, outcomes, and biases.

Longitudinal observational studies (LOS) provide key information about outcomes and treatment effectiveness that are not available from other types of investigations, including randomized controlled trials. Although LOS are easy to perform, they are difficult to perform correctly. Major problems include recruitment, retention, and relevance, but the central problems in LOS are bias, understanding the nature of the biases, and reporting the biases. To advance the quality and validity of LOS there must be a uniform requirement for reporting detailed data that includes details about the sampled population and the sampling methods, the rationale for the selection of control subjects, the probable biases, and estimates of the extent and consequences of the biases. Sufficient covariates should be collected so that where possible statistical adjustment for bias can be made.