Concurring with many previous reports
[8–11, 21, 22], the present study confirmed that AEs are frequently experienced by ICU patients. Our study showed clearly that risk-index scores for drug-related AEs decreased as a consequence of the multifaceted intervention strategy on medication that we applied. Despite the intrinsic limitations of self-reporting methodologies for evaluating performance, we think that a risk-index score analysis represents a useful tool to evaluate targeted improvement strategies.
AE incidence rates are difficult to compare among studies, in part because of different AE reporting strategies. Despite implementation of a single methodology, the Sentinel Events Evaluation (SEE) 1 study reported medication errors by unit staff at a rate of 10.5 per 100 patient days
, whereas a value of 74.5 per 100 patient days was documented in the SEE 2 study
. This difference is probably related to the fact that the SEE 1 study captured errors in five selected AE categories, whereas the SEE 2 study focused on medication errors only.
As documented in other similar studies, our reporting rates differed among healthcare worker groups
[10, 11]. ICU nurses reported most of the AEs reflecting the different numbers and types of activities performed by distinct caregiver groups. Contrary to popular belief, most of the AEs in our study occurred during elective procedures, as in the multicenter SEE 2 study
. Elective procedures may be considerably improved by adopting educational interventions and protocol implementation. Single-center and multicenter studies have exhibited similar diurnal distributions of AEs, with a peak in morning caregiver activities
[8, 17]. This observation highlights the need to recognize these variations throughout the day and to identify the corresponding opportunities for error, which will promote better planning for and allocation of resources to decrease risk.
The characteristics of the AEs in our study are consistent with previous reports that suggest a common pattern of AEs in the ICU
[8–11, 23]. Drug-related AEs are the most common; these errors are frequently detected and reported because medication administration is a multistep procedure that requires correct prescription and administration of the right drug to the right patient at the right dose via the right route at the right time
[21, 22, 24]. Our most frequent medication-based AEs were related to incorrect prescriptions. The practice of handwritten and sometimes only verbally communicated medical prescription in our ICUs may partly explain this finding. Furthermore, the prescription was annotated by a trainee in most of the AE cases. This calls for a better supervision when prescriptions are ordered verbally. As previously reported
[21, 22, 24], many of these medication-related AEs can potentially be prevented using adequate information and communication technologies, which in turn, however, may introduce new hazards requiring further investigation
. In a stepwise multivariate logistic regression analysis, Valentin et al. found that physician electronic prescribing was associated with reduced odds for the occurrence of a medication-related AE
. A basic computerized prescribing tool can be implemented in combination with clinical decision support for correct drug dosing, a drug/laboratory value check, and a drug/drug interaction check. Finally, the involvement of a pharmacist in the direct care of ICU patients has been shown not only to positively impact patient outcome, but also to reduce costs
[26, 27]. Noncompliance with internal guidelines, policies, and checklists was frequently reported in our study as a personal-related contributing factor for AEs, underlying the urgent need to improve continuous education of our ICU personnel. Poor communication among the care staff has been repeatedly identified as a contributing factor to many AEs
[28, 29] and also has been observed in our study. Possible solutions may be achieved by improving interdisciplinary communication during bedside rounds in the ICUs, which was previously associated with reduced AE rates
. Furthermore, communication skills training is needed to create a favorable atmosphere of trust and respect allowing open interaction as well as creating a positive unit climate, which has recently been shown to maximize patient safety
Our multifaceted, medication-focused intervention was associated with reduced risk-index scores for drug-related AEs, although we cannot ultimately confirm any causal relationship despite the consistency, plausibility, and temporality. Moreover, the unchanged risk-index scores for airway-related AEs and indwelling lines-related AEs supports the hypothesis that our multifaceted intervention plan positively affected medication-related AE risk-index scores. Because we did not introduce any specific strategy for mitigating communication-related AEs, we assume that the new electronic prescribing process affected indirectly the observed improved risk-index score in this category. Computerized physician order entry performed during the multidisciplinary clinical round allowed a better communication and integration of treatment plans between caregivers. In a cluster-randomized, crossover study, Garrouste-Ogeas et al. demonstrated that multifaceted safety programs decreased insulin administration errors
. Contrary to our study, their trial focused on two selected medications (insulin and anticoagulant) at two selected stages (prescription and administration). Furthermore, AE reporting resulted from a combination of an external observer with medical chart review and significant Hawthorne effects (bias following a change in behavior under observation) were observed by the authors. Compared with our study, the cluster-randomized, crossover design used by Garrouste-Ogeas et al. has obviously several strengths (no bias arising from regression to the mean and no historical bias).
We also have demonstrated in agreement with a previous report
 that AE reporting increased after structured meetings with the care staff that addressed the staff’s perspectives on AE reporting. Our improved post-meeting reporting rate decreased in the following months, suggesting that regular meetings at closer intervals are needed to maintain a high and constant reporting rate.
This trial has several limitations. First, basic patient demographic characteristics were not mentioned in the AE report to maximize anonymity and confidentiality. As a consequence, we did not document the number of patients suffering an AE, but only the total number of events assigned to different categories. This limitation precludes the calculation of odds ratios and a more sophisticated statistical analysis. Second, the survey was performed in four multidisciplinary ICUs of non-University teaching hospitals using a locally developed AE questionnaire. Therefore, our results may not be generalizable to other ICUs with different characteristics (patient types, severity of illness, staffing pattern). Third, the nurses reported most of the AEs, inducing a possible reporting bias. However, this observation is consistent with previous reports
[10, 11], reflecting the larger time specific caregivers spend in direct patient care. A before-and-after study design such as ours that evaluates a multifaceted implementation strategy has the potential risk of historical bias and other possible external confounders
. These limitations could be avoided with a randomized study design, but this was impractical in our ICUs. Finally, our study focused on AE reporting rather than on patient safety. In this context, it is important to remember that AE reporting is the first step in safety-improvement strategies
; the SEE 2 study revealed that an existing critical incident reporting system was an independent predictor for a decreased risk of parenteral medication-based AEs at the administration stage
The strengths of our study include the multicenter design, the long study period, the prospective self-reporting by caregivers during hospital stay, the utilization of a detailed questionnaire allowing better characterization of AE occurrence, and the use of the risk-index score methodology. This methodology may enable monitoring of the results of a specific implementation plan. Reporting of the crude AE rates before and after an implementation cannot be used to evaluate progress in patient safety for several reasons, including nonrandom sample reporting from an unknown probability distribution, a reporting bias of unknown magnitude and direction as well as an unknown at-risk population
. A risk-index score as a function of time may represent a useful tool for the assessment of targeted improvement interventions, given its use of a non-rate-based measure. Despite defining consequences as objectively as possible, it is inevitable that scoring the consequences of some risks will involve a degree of subjectivity
. It is important that effective, practical training and relevant examples form a part of the implementation of any risk-assessment system to maximize scoring consistency across the organization.
As in other investigations of AEs in the ICU, we had no “gold-standard” method to detect AEs. The use of external observers
[8, 19, 21] may be considered a reference method for capturing AEs, but this strategy appears to consume a large number of resources and suffers from the Hawthorne effect. The AE self-reporting method provides a detailed description of the AE and identifies a large number of preventable incidents, but it has the risk of selection bias, underreporting and some degree of Hawthorne effect, whereas retrospective medical chart review provides less contextual information for an AE and identifies fewer preventable incidents. A combination of different data-gathering methods is likely to be optimal. However, it is usually accepted that it is preferable to reduce the quantity but not the quality of data collection