Recommended Interventions. Based on recently published recommendations by an international panel of intensivists (10) and recent major study results, the group of SepNet investigators selected seven interventions with respect to clinical relevance (Table 1). Two of
these, low-tidal volume (Vt) ventilation in patients with acute lung injury (ALI)/ARDS and glycemic control, are applicable throughout the course of the disease. Therefore, we grouped the results of the audit into time and severityindependent and time and severity dependent recommendations.
Data Collection and Analysis. Audits were performed by external intensivists from 18 SepNet regional study centers (see Appendix) after prior central training. The auditors first gathered data on adherence to seven best practice interventions through a semistructured interview of the ICU medical director, i.e., the senior intensivist on site who is permanently in charge of local practice in the German system of closed ICUs. In the absence of the ICU director, the deputy was approached. The ICU directors could answer with “always,” “frequently,” “sometimes,” “rarely,” or “never” to each practice recommendation. Thereafter, charts of patients who fulfilled the inclusion criteria were reviewed together with the resident physician who in some cases was identical to the medical ICU director.
Patient data were anonymized in the case report forms. Electronic data entry, data management, quality control, and analysis were performed centrally. Data collected from patient documentation included maximal blood glucose levels; maximal documented Vt; ad- ministration and dosage of aPC, hydrocortisone, dopamine, and AT, all noted within 24 hrs preceding the audit.
ALI and ARDS as well as predicted body weight (PBW) were defined as described elsewhere (6). Maximum Vt (Vt max) per PBW was calculated over the preceding 24 hrs.
The study was approved by the responsible institutional ethics committees and by the federal data protection commissioner.
Statistical Analysis. We analyzed whether documented practice (i.e., maximal glucose values and Vt max in mL/kg PBW) and perceived adherence (i.e., interview responses) were significantly different across strata (i.e., hospital size). All data were categorical and are reported as absolute or relative frequencies where appropriate. Fisher’s exact test was applied for all group comparisons. p values below 0.05 were considered significant. SPSS 11.0.1 (SPSS Inc., Chicago, IL) was used for data analysis.