This was confirmed by the sensitivity analysis, removing the most common pain-related adverse events (tachycardia and Veliparib Sigma hypertension). Use of analgesics may decrease stress response in critically ill patients [46,47]. In our study, the main SAE observed were oxygen desaturation and ventilator distress (Table (Table4).4). The rate of these SAE decreased throughout the study, although ventilator management or oxygenation practices were not changed, contrary to pain management practices.This study constitutes an improvement in quality and safety in healthcare. Such processes are fundamental to improving our healthcare, by changing our systems, avoiding overuse of ineffective care and underuse of effective care [48].

Quality improvement methods, such as the Plan-Do-Check-Adjust cycle, seek to apply proven treatments and recommended strategies to “real world” patients, allowing the integration of “best evidence” and “clinical evidence” [20,22]. To our knowledge, there are no published data regarding the feasibility of a quality improvement process for moving ICU patients. Changing practices is challenging in an ICU setting, with necessary education of a large team [49,50]. Moreover, a multidisciplinary approach is essential, placing responsibility with the team rather than with individuals. Differences in pain appreciation among physicians, nurses and assistant nurses are well known in the ICU setting [34] and were found again in our questionnaire. It has been previously reported that ICU physicians under-evaluated patients’ pain compared to nurses [51], and that ICU nurses under-evaluated patients’ pain compared to assistant nurses [52].

Our study has several limitations. First, there were less missing data in the third phase (adjusted intervention P-D-C-A phase) than in the two first phases. This could be explained by a high workload during February and September 2010, much higher than in April 2011. Indeed, one-third of the unit had to be closed unexpectedly in April after Phase 3 had begun. To deal with missing data and to avoid a possible bias due to more frequently evaluating patients in pain in the two first phases, patients were randomly enrolled in Phase 4. This phase (consolidation of P-D-C-A-steps) was aimed to reinforce the results observed in the previous phase [21,23].

Second, pain was evaluated by the bedside RN (BPS) or by the patient with the help of the bedside RN (NRS), and Batimastat not by an independent investigator. However, this design is appropriate in a quality improvement process of routine care because self-evaluation of the caregiver is part of the improvement process [22,53]. Moreover, even if it was not possible to have an independent investigator at the bedside for all 16 patients during the turning every morning, the presence of an observer could have introduced another bias leading to more accurate care [32].