Introduction

The majority of mechanically ventilated patients within the intensive care unit (ICU) receive sedative drugs. Sedation is administered to ensure patient comfort, reduce anxiety, and facilitate treatments. Optimising sedation management is recognised as important in improving patient outcomes [1]. Under-sedated patients may become agitated and distressed and are at risk of adverse events such as extubation [2–4], whereas over-sedation can prolong time to recovery [1, 5].

Assessment of sedation level is carried out mainly by nurses or critical care physicians by assessing patient responses to simple stimuli. Sedation scales such as the Ramsay scale or the Richmond Agitation-Sedation Scale (RASS) are widely used [6–8]. However, there is no universally accepted standard, and this can make comparison between different studies or ICUs difficult [2]. Furthermore, some of these scales have not been fully validated in ICU patients [4]. Recently, devices such as the bispectral index monitor (BIS), which aim to assess sedation levels more objectively, have been introduced. However, most studies of BIS have been performed in surgical settings, and to date its effectiveness is not fully proven [8–10].

Available guidelines on sedation typically provide limited guidance on optimal sedation monitoring and levels. This is at least partly because optimal sedation levels differ between patients according to their clinical circumstances, and therefore sedation practice is ideally individually tailored to each patient, as recommended by several guidelines [2, 11, 12]. However, among guidelines that do recommend an optimal level of sedation, there are discrepancies, indicating a lack of consensus on this issue. For example, of a survey of available guidelines, one [13] recommended a sedation level of 2 or 3 on the Ramsay scale, whereas one recommended a goal of RASS -3 for an intubated patient [14] and a second recommended a goal of RASS 0 to -2 [15]. A number of guidelines stress the importance of establishing a set protocol for the sedation of ICU patients [16, 17] but do not set out such a protocol in detail, leaving it to individual institutions, and more recent guidelines recognise the benefit of regular (daily) interruption of sedation for eligible patients [11, 14, 18, 19] within sedation protocols.

It is recognised that optimising sedation practice is a recognised quality marker for intensive care treatment, and procedures designed to optimise patient sedation state, such as daily sedation breaks and more frequent monitoring, are key elements of recent quality improvement initiatives. However, despite these recent efforts to improve the quality of sedation practice in the ICU, the epidemiology of sedation, and specifically the prevalence of over- or under-sedation, is unclear. To investigate this further, we carried out a systematic review of the publicly available literature to identify the reported incidence of sub-optimal sedation.

Materials and methods

Searching

Medline, Embase and the Cumulative Index to Nursing and Allied Health Literature (CINAHL) databases were searched from 1988 to 15 May 2008 using terms for sedation, ICU, sedation quality management, and sub-optimal sedation. The standard Scottish Intercollegiate Guidance Network (SIGN) filters for randomised controlled trials (RCTs), economic studies and observational studies [20] were combined to capture all study designs relevant to the study question. Full details of the search strategy used are available from the authors on request. Conference proceedings from 2005 through 2008 were hand-searched for relevant studies. All results were uploaded into a bespoke internet SQL (structured query language)-based database.

Selection criteria

Inclusion of studies was according to a predetermined set of criteria. To be included, studies had to be in adult humans who were sedated and undergoing mechanical ventilation within the ICU and furthermore had to report the incidence of sub-optimal sedation, over- or under-sedation, or of optimal sedation, as defined by the study. Studies that reported the impact of sedation practice on outcomes were also included; these data are reported separately. In addition, short-term studies (including only patients sedated less than 24 hours) were excluded. Only English-language studies were included. To check that they met the review inclusion criteria, all abstracts were reviewed twice by two independent reviewers, with all conflicts resolved by a third reviewer. Full papers of all included studies were retrieved and were again reviewed twice to ensure that they met inclusion criteria. Studies included at this stage were classified as to which aspect of the review question they met, and appropriate data were extracted, summarised and analysed.

Data extraction

Data were extracted by two reviewers and checked by a third reviewer against the original studies. For all studies, the following data were extracted: country, sponsor, study design, patient population, objective, number of patients in the study, details of comparisons made (such as between different treatment arms or between different sedation monitoring systems), and the proportion of measurements, patients, or time in which patients were judged to be optimally sedated, sub-optimally sedated, over-sedated, or under-sedated.

Quantitative data synthesis

Due to the wide range of included study types, no studies were suitable for quantitative data synthesis.

Results

Systematic review study flow

The flow of studies through the systematic review is documented in the QUOROM (Quality of Reporting of Meta-Analyses) diagram in Figure 1. Seventy-five primary and seven secondary studies met the inclusion criteria. Of these, 18 did not provide any data; either they did not contain data on the outcomes extracted in this review or they did not provide these data in quantitative form. Thirty-six studies reported data on the incidence of sub-optimal sedation. The remainder reported the impact of sedation practice on outcomes; these data are reported separately. Of the included studies, three were cohort studies that specifically investigated the epidemiology of sedation, 23 were studies investigating anaesthetic drugs (of which 19 were RCTs and four were observational studies), six studies compared sedation monitoring devices or scales (of which one was an RCT and the remainder were observational studies), three studies investigated the introduction of sedation guidelines, and one did not fit any of these categories. The majority of studies (20) were published after 2002, indicating the increasing interest in the practice of sedation quality in recent years, in particular following the publication of updated sedation guidelines from the American College of Critical Care Medicine [2, 6].

Figure 1
figure 1

The QUOROM (Quality of Reporting of Meta-Analyses) diagram illustrates the flow of studies through the systematic review.

Full size image

Definitions of adequate sedation

To assess the incidence of sub-optimal sedation, it is necessary to consider the definition of what constitutes optimal sedation. We used the definition of optimal sedation (and consequently of what constituted sub-optimal sedation) provided by individual studies due to the fact that optimal sedation levels will vary according to study setting (for example, between neurological ICU and medical ICU).

Across all of the studies, 13 different sedation scales were used to assess sedation quality; additionally, nurse assessment of sedation quality simply as over-sedated, under-sedated, or adequate was used three times (Table 1). The Ramsay scale was the most commonly used scale, in 14 studies, with a variant used in a further 7 studies. This is illustrated in Figure 2.

Figure 2
figure 2

The frequency with which each sedation scale was used in the studies included in our systematic review. ICU, intensive care unit; MAAS, Motor Activity Assessment Scale; OAAS, Observer's Assessment of Alertness/Sedation Scale; RASS, Richmond Agitation Sedation Scale; SAS, Riker Sedation-Agitation Scale.

Full size image
Table 1 Incidence of optimal and sub-optimal sedation in included studies
Full size table

In addition to the variation in scales used to assess sedation, there was variation in the recommended range of optimal sedation levels stated. Sedation requirements obviously differ among patients; nevertheless, the variation in recommended ranges in included studies indicates some uncertainty in what constitutes optimal sedation. Of the studies using the Ramsay scale, recommended ranges were 2 to 3 (recommended in two studies [21, 22]), 2 to 4 (two studies [23, 24]), 2 to 5 (two studies [25, 26]), 3 to 4 (two studies [27, 28]) and 4 to 5 (one study [29]), while three studies did not recommend specific levels but recommended that levels be optimised for each individual patient [30–32]. This variation was reflected in the other scales used; for studies recommending a modified Ramsay scale, recommended ranges were 1 to 4 [33], 3 to 4 [34], 4 [35], and 5 to 6 (the last range being specifically for seriously injured patients [36]) or targets optimised for each patient [37, 38]. The stated SAS (Riker Sedation-Agitation Scale) target level was 1 to 3 [39], 4 [40, 41], or 3 to 4 [42]. Due to the number of studies recommending that optimal sedation state be determined individually for each patient, there was no comparison possible for other scales.

Incidence of sub-optimal sedation

Table 1 lists the study design, sedation assessment scale or tool used, and incidence of sub-optimal sedation reported by studies. As stated above, we used individual study definitions of optimal and sub-optimal sedation because of the fact that optimal sedation levels are likely to vary by study setting.

The three observational studies that investigated the epidemiology of sedation were considered to be the most relevant to the study question as their specific aim was to investigate clinical sedation practice rather than practice within the confines of a trial, where more frequent monitoring and the Hawthorne effect could contribute to improving standards.

A survey of practice across 44 ICUs in France also found a high incidence of deep sedation, in 41% to 57% of readings over a 6-day period [43]. This study highlighted the risks of prolonged deep sedation, which, however, was not specifically defined as over-sedation. Results from these three studies indicate that 30% to 60% of sedation assessments indicate 'deep' or 'over' sedation, although precise description of the prevalence is confounded by imprecise definition or health care worker perceptions. These studies clearly indicate an excess of over-sedation compared with under-sedation.

Martin and colleagues [30] conducted a postal survey of 220 ICUs in Germany. This study found that 42.6% of patients sedated between 24 and 72 hours and 39.5% of patients sedated over 72 hours were over-sedated; the incidence of under-sedation was much lower (<6%).

In the US-based study of Weinert and colleagues [44], the aim was to compare subjective and objective ratings of sedation. Subjects provided 12,414 sedation assessments and were judged by nurses to be sub-optimally sedated in 17% of assessments, over-sedated in 2.6%, and under-sedated in 13.9%. Critically, however, patients were unrousable or minimally rousable just under one third of the time, indicating a high incidence of deep sedation. This finding illustrates the importance of the perception of the health care worker or assessor or both in describing the prevalence of sub-optimal sedation.

The remaining included studies comprised studies of sedative drugs [21–26, 29, 31–34, 36, 37, 39–42, 45–50], studies investigating different sedation devices or scales [27, 28, 38, 51, 52], and studies looking at the introduction of a sedation guideline or protocol [3, 35, 53, 54]. Studies varied by design and aim, by sedatives used, by scales and definitions of sub-optimal sedation used, and by the way incidence was reported (as a proportion of measurements, patients, or time). While these studies did not necessarily have the incidence of sub-optimal sedation as their primary focus, the data in such studies were considered to be of interest to the inclusive scope of this review. Although studies of sedative drugs or of the introduction of guidelines or protocols may not give an accurate estimate of the incidence of sub-optimal sedation within routine clinical practice, they nevertheless show that it does occur and can give an impression of the extent to which it may be a problem, even in settings that could be reasonably expected to be more controlled than in routine practice. The incidence of sub-optimal sedation reported in these studies is summarised in Figure 3 (separated by study and treatment arm where relevant). The reported incidence varied from 1% [39] to 75% [28], with the majority reporting an incidence of over 20%. The incidence of over- and under-sedation was similarly variable, and figures of between 2.8% and 44% for over-sedation [28, 33, 51] and between 2% and 31% for under-sedation [23, 51] were reported. A further study [2] that looked at the introduction of a sedation guideline did not record the incidence of sub-optimal sedation but recorded the median Ramsay scale values. These were 4 during the day and 5 at night, in contrast to the study's stated aim of Ramsay levels of 2 to 3 during the day and 3 to 4 at night; this study again noted a possible tendency toward over-sedation of patients. Importantly, there was no change in this tendency before and after reinforcement of the guideline, suggesting that this was insufficient to improve sedation practice [3].

Figure 3
figure 3

Incidence of sub-optimal sedation across included studies. The plot shows the percentage of measurements, patients, or time in which patients were sub-optimally sedated according to each included study's definition of optimal sedation and measurements reported. Studies are grouped by study design. Where more than one group was reported by a study (for example, a comparison of two different treatment arms), separate points are shown for each group. RCT, randomised controlled trial.

Full size image

Discussion

Our systematic review identified few studies that specifically described the epidemiology of sedation during ICU care. Description of the incidence of sub-optimal sedation and over- and under-sedation was difficult due to variation in the use of these terms within individual studies. Overall, available data suggest a high incidence of over-sedation in ICUs, potentially present at 40% to 60% of assessments. A lower reported incidence of sub-optimal sedation across most studies suggests that health care workers consider deep levels of sedation appropriate for many patients.

The quality of published studies was low. There was wide variation in the method used to assess sedation state, the frequency of measurement, and the stated response to evaluations. In addition, the completeness of data in relation to entire ICU populations was usually not stated, introducing the potential for selection bias. Only three cohort studies were found. The importance of selection or inclusion bias was lowest with this study design. All of these indicated a substantial incidence of sub-optimal sedation, with over-sedation being more common (33% to 57%). Notably, one study reported that nurse assessment of sedation found a low incidence of over-sedation, which appeared at odds with the fact that in one third of measurements patients were unrousable or minimally rousable. A difference in perceptions of what constitutes optimal sedation between different health care worker groups and between individual health care workers is also likely to affect the reported incidence of sub-optimal sedation. This finding emphasises the importance of using sedation-assessment methods that have high validity and low inter-rater variability. Many of the scales in current use have not been subjected to formal evaluation. The RASS has been shown to have good construct validity and high levels of consistency among health care workers [55] but was used infrequently in the published literature.

When non-cohort studies were considered, a wide range of incidence of sub-optimal sedation levels was reported (from 1% to 75%). These studies were randomised or non-randomised trials of the efficacy of sedative drugs [21–26, 29, 31–34, 36, 37, 39–42, 45–50], or were studies that evaluated sedation devices or scales [27, 28, 38, 51, 52] to monitor sedation levels, or looked at the introduction of sedation guidelines [3, 35, 53, 54]. Overall, the chance of selection and investigator bias, and of study effects, in these studies was high. Drug trials in particular were considered less relevant to our study question as practice within a clinical trial may differ from standard care and is more likely to be controlled. Despite these factors, the majority of studies found an incidence of sub-optimal sedation, using study-specific definitions, of greater than 20%. These studies confirm the findings of the observational cohort studies and suggest high levels of sub-optimal sedation during routine care.

Improving sedation management through sedation protocols and interventions such as daily interruption of sedation is an increasing focus of quality improvement initiatives in critical care in some health care systems [1, 56, 57]. Sedation protocols and scales are increasingly, though not universally, used. A review of German hospitals by Martin and colleagues [6] showed increases in the use of sedation protocols (from 21% of hospitals to 46%) and in the use of sedation scales (from 8% to 51%) over the period of 2002 to 2006. In Finland, Tallgren and colleagues [3] reported that reinforcing a sedation guideline increased the percentage of expected Ramsay scale recordings made, but only to 71% of expected recordings that were actually made, indicating that formal sedation assessments were still not carried out as regularly as they should have been. These studies suggest that, despite recent evidence supporting the avoidance of over-sedation [57, 58], the use of systematic approaches to measure sedation state and optimise sedation for individual patients is not universal. The high prevalence of sub-optimal sedation and high incidence of over-sedation in published studies indicate potential for significant quality improvement in this aspect of care. This is likely to translate into substantial patient benefit.

There are several limitations to our review. As with any systematic review, studies may have been missed; this review was confined to English-language publications and therefore may be biased toward the US and UK in focus. Despite the inclusion of conference searching, there may be relevant grey literature that we did not search. As previously discussed, many of the included studies did not investigate the quality of sedation practice as their primary aim, limiting the relevance of the information provided.

Conclusions

Our review indicates the poor quality of epidemiological data concerning current sedation practice and the incidence of sub-optimal sedation. A key issue is the standardisation of methods of assessment and definitions of optimal sedation. Despite this, available data suggest that many patients in ICUs are considered sub-optimally sedated and, specifically, that the incidence of over-sedation remains high. The strong associations between sedation practice, especially over-sedation, and adverse patient outcomes suggest that a more uniform approach to monitoring depth and quality of sedation will improve quality of care.

Key messages

  • The literature shows that, within the intensive care unit (ICU), a substantial proportion of patients experience inappropriate levels of sedation (that is, under- or over-sedation).

  • There is a greater tendency toward over-sedation in particular.

  • There is a lack of consensus in the literature as to what constitutes optimal sedation practice within the ICU, with little standardisation of either assessment methods or definitions.

  • Improvements in the definition and measurement of optimal sedation may have a positive impact on patient outcomes.

Authors' information

TW is a professor of anaesthetics and critical care at Edinburgh University. DLJ is head of health economics, EMEA (Europe, the Middle East and Africa), at GE Healthcare. CWP is a consultant at Heron Evidence Development Ltd, a health outcomes research consultancy. KFC is a health outcomes analyst at Heron Evidence Development Ltd.