ECMO-/ECLS

Database studies range in their objectives and can include: descriptive studies of
cohorts of patients with a specific disease; longitudinal natural history studies of specific
patient populations; resource utilization studies reporting on costs, length of stay, or
other variables at a single institution or across institutions; studies of practice variation
for a particular disease or treatment across institutions; benchmarking studies
comparing rates of specific procedures, outcomes, or complications across institutions;
or comparative effectiveness studies comparing two treatments across all patients in the
database (single institution or multi-institutional). In all database studies, groups of
patients, treatments or outcomes of interest must be identified. It is critical that the
identification and grouping of patients, treatments, and outcomes be described and
validated as completely as possible. This is where the reliability and validity of these
studies must be carefully evaluated. For example, most administrative databases are
based on ICD-9 coding; the determination of the presence of a disease, receipt of a
treatment, or occurrence of an outcome in a patient is based on an ICD-9 code for that
factor being including in the database record for that patient. Therefore, the patients
included in a study and the study’s results depend on how well coding is performed at
each institution and how many ICD-9 codes are included in the various fields of the
database (e.g. diagnoses, procedures). Each database will have varying levels of
reliability with different rates of misclassification of variables and missing data. These
limitations should be addressed and reported as completely as possible in each study.
Prospective observational studies or prospective registries represent slightly
higher levels of evidence. These studies identify variables to be collected and then
prospective collecting the data. These are less biased because the data is defined and