Ventilation and sedation is a procedure that is commonly seen in ITUs and other critical care establishments. It can be use for a number of different therapeutic reasons and to help to treat a number of different clinical situations.(Marx WH et al 1999)
The need for mechanical ventilation is one of the main criteria for admission to and ITU. This is usually administered by endotracheal intubation (occasionally by tracheostomy) and sedation. Sedation is clearly needed for the intubation to be tolerated, but the degree of sedation is more often then not a matter of professional clinical judgement on the part of the responsible anaesthetist. It is a common observation that optimal sedation levels can both improve the quality of care and also reduce its duration in terms of time.(Rainey TG et al. 1998). It also therefore follows that this may well also reduce the time spent in intensive care. (Kollef et al 1998 (I)
Equally inappropriate or excessive amounts of sedation can increase the length of time spent on the ventilator which has it’s own disadvantages and harmful effects. (Kollef et al 1998) (II)
The purpose of this literature review is to examine the rationales and evidence base behind this combination of therapy. (Berwick D 2005)
A good starting point for our considerations is the paper by Brattebø G (et al 2002). It takes as read, our original premise that optimal levels of sedation are imperative for optimal levels of outcome on ventilated patients. It accepts the need for devising some kind of protocol to allow for an improved sedation strategy. The authors devised an observational study which not only introduced guidelines for sedation control, but interestingly also took the opportunity to observe and analyse the actual methods of adoption as the medical and nursing staff tried to implement it. It is this second part of the study design that perhaps sets it apart from many of the prospective studies that we will be reviewing here. (Cochran and Cox. 1957)
The NHS has learned many lessons relating to change management. In an organisation as large and cumbersome as the NHS, the effective introduction of new (or even good) ideas needs careful implementation management. Over the years the NHS has witnessed (and suffered from) the inept management of a newly introduced concept. (Nickols F.2004)
The concept may be good, but the method of its introduction can render it useless. (Berwick D. 1996)
Clinicians who have been in practice for more than twenty five years will remember the debacle of the disastrous introduction of the Griffiths Report (Griffiths Report 1983).
Many would say that, in principal, it was a good workable system, but it’s introduction and implementation was so inept that it was withdrawn before it could reach a fraction of its full potential. (Davidmann 1988) Brattebø’s paper takes stock of this potential for poor implementation and both analyses and assesses the protocol’s introduction to be able to draw appropriate conclusions on the matter.
The study was not on a large scale but this allowed for more careful and intimate observation of the issues under investigation. It was set in one ITU in a University hospital and encompassed all staff and patients (over 18) passing through the unit over an 11 month period.
One interesting strategy was that the authors considered and evaluated different methods of achieving staff conformance as the study progressed, and progressively adopted those methods that were found to be the most effective. (Carey et al. 1995)
The paper is both long and involved. The important outcomes of the investigation were that by optimising the sedation given to the patients, the mean ventilator time decreased by 2.1 days (from an average 7.1 days before the trial to 5.3 days after it). This also allowed a reduction in the mean period of time spent in the ITU by an average of one day (9.3 days to 8.3 days). The authors also record that these reductions and efficiencies were not associated with any increase in adverse incidents (accidental extubations etc.). One particularly significant comment made by the authors was:-
Lessons learnt: Relatively simple changes in sedation practice had significant effects on length of ventilator support. The change process was well received by the staff and increased their interest in identifying other areas for improvement.
The authors describe their methodology as using the “breakthrough method” (Marx et al. 1999) in which they make use of multiple short cycle improvements (Plesk PE 1999). They describe the method as entailing:
Setting goals, choosing appropriate small changes, and measuring whether the changes do lead to improvements; if so, the changes are incorporated in the departmental routines (Rainey et al 1998).
The backbone of the study was the concept that, in cases of respiratory failure of any cause, patients are generally given both sedation and analgesia most often by the method of continuous infusion. (Brook AD et al 1999)
Commonly, the responsible clinician tends to err on the side of safety and prescribe heavier doses than may actually be necessary, with the result that the patient becomes more heavily sedated than may be required. (Kreiss et al 2000), and the corollary of this is that the patient will tend to spend more time on a ventilator than is absolutely necessary (Kollef M et al. 1998 (I)).
The alternative, say the authors, is to introduce a scoring system which can quantify when the sedation is “sufficient but not excessive” so that the lowest rates of sedative infusion can be achieved with the use of occasional supplemental bolus doses (Barr et al 1995). The natural inference with this system is that this will reduce the time spent on the ventilator. As we have discussed, this was one of the outcome measures of the study.
The actual process is described in detail in the paper and therefore we will not present it here, but it hinged on the application of the MAAS scale (Devlin et al 1999) and correlation with the degree of sedation that was required to produce a particular score on that scale.
The protocol and introduction of the scheme appears to have been managed meticulously well with multiple staff presentations, wall posters, feedback questionnaires, and other management tools. (Shortell SM et al. 1998),
The key note to absorb from this study is that the application of this system resulted in a 30% reduction in ventilator time for the patient on this particular ITU. A fully critical appraisal would have to observe that this could simply be a predictable result from the introduction of a fixed protocol in a system of lax and unmanaged clinical practice which allowed for a 30% overuseage in ventilator time in the unit. There is no evidence to support or refute this point of view, but it would be charitable to assume that the clinical staff were doing what they thought best at the time. This method allowed them to have a more precise control over the sedation levels achieved. (Henriksen and Kaplan 2003)
It is interesting to note that the authors stated that as the intervention was aimed primarily at the clinical staff, it was not considered necessary to get patient consent. Although clearly it would have been difficult to obtain direct patient consent, we have to express surprise that proxy consent was not obtained on behalf of the patients from the relatives. (Sugarman J & Sulmasy 2001)
We can now turn our attentions to a paper that we have quoted in support of a point in the Brattebø paper. Kreiss (et al 2000) produced a paper that took a similar approach but with different methodology. This study arose from the observation that continuous sedation levels tended to impede the clinical assessment of the patient and that it was common practice to allow “windows” of lighter sedation to allow neurological assessments together with assessments of the mental state. (Kong et al 2003)
The study compared the outcomes of one group who received reduced sedation at the discretion of the responsible clinician, who determined when appropriate neurological examination was necessary, with the outcomes from a second group who were “woken” on a daily basis irrespective of clinical need. (Brock et al 1998) A critical assessment would have to conclude that the cohort was not particularly large (with about 60 patients in each group), although it is accepted that the numbers of patients available for such a study are, by necessity, limited. (Grimes et al. 2002)
The results however, showed surprisingly different outcomes in the two groups. The mean time on ventilation and sedation in the control group was 7.3 days but was only 4.9 days in the intervention group. It also showed an even more marked reduction in average time in ITU than the Brattebø study, of 9.9 days (control group) to 6.4 days (intervention group).
It should be noted that in this study a number of adverse incidents occurred, including three patients in the intervention group who removed their endotracheal tube. (Vassal et al 1998)
This actually compared favourably with four who removed their tubes in the control group (see on)
In the aftermath of the Kreiss paper, Gorman (et al 2004) published a similar study, which considered the effect of interruption of the continuous levels of sedation when patients were undergoing mechanical ventilation in the ITU.
Their object was to assess whether this did have any effect on the length of time that the patient spent on the ventilator. The study was not particularly large (by study standards) but was large in comparison with other ITU based studies with an initial entry cohort of 150 patients.
In contract to the Kollef study (Kollef et al 1998 – see on), this study was truly randomised once the inclusion criteria had been met into two groups. One group had continuous infusion of sedation and the other group were woken daily and then restarted on a regime which entailed recommencing half the dose of medication and then retitrating until the required level of sedation was reached which was defined as level 3-4 on the Ramsay scale. (Brock WA et al. 1998)
In addition both groups were randomly subdivided further to be sedated with either midazolam or propofol, so there were actually four investigation groups.
In broad terms, the outcome measures were the length of time that the patients spent on the ventilator together with the overall time they spent in the ITU and how these outcomes correlated with the method of administration of the appropriate sedative.
Before we consider the results, we must consider the structure of the trial. And compare it with the seemingly similar Kollef trial. There are two major differences in this trial’s design which make it significantly different from Kollef’s trial. One very significant point in the design of the Koleff trial was the fact that the control group was sedated to Ramsay 3-4 levels in order to match the required levels in the intervention group.
Although this clearly makes for easier and more direct comparison, it is possible that this level of sedation may actually have been greater than was actually required for that particular patient. This may have had the effect of being a significant confounding factor in the eventual findings with regard to ventilation time.
The second problem was that the Koleff study was not truly “blinded” as although the treating physicians did not know which group their patients had been assigned to, the structure of the study required a recording nurse to sit with the patient to record their levels of sedation at all times. This could not fail to have been noticed by the clinicians in charge of the case, and may have introduced a source of bias into the results.
The Gorman trial took account of both of these features so that they were not potential confounding factors in the newer study. The results were still similar although now possibly more statistically valid. The interruption group required less time on the ventilator and less time in hospital than the continuous infusion group. The authors note that these recommendations of interrupted sedation and retitration, have now become standard recommendations for the guidance of clinicians in the ITU setting. (Jacobi et al.2002)
We have refered to the complication of unintentional extubation (UEX) of the patient. This is an unusual but occasional consequence of lightening the sedation level of a patient who is being mechanically ventilated. (Esteban et al 1999 (I)
It is enlightening to therefore consider the study by Boulain (1998) who reviewed this phenomenon. The author took the potentially rather unpromising step of designing a prospective cohort trial to examine the occurrence. He followed nearly 450 patients over a two month period. in the study he recorded that over 10% of patients had at least one episode of unexpected extubation during the study period, which is rather higher than some of the other authorities have reported. (Krinsley JS & Barone JE 2005) (Epstien et al 2000). The direct relevance to our considerations comes from the sentence:
By use of multivariate analysis, we identified four factors contributing to unexpected extubation : chronic respiratory failure, endotracheal tube fixation with only thin adhesive tape, orotracheal intubation, and the lack of intravenous sedation.
The third factor is obviously artificial, as it is an obvious prerequisite of unexpected extubation in the first place. (Betbese et al. 1998)
It was noted that, at the moment of unexpected extubation over 61% of the patients were showing signs of clinical agitation. This is therefore almost certainly an indication that sedation levels were not sufficiently high. The converse of this argument is this later on in the paper the author quotes that of the 46 occasions when this happened, only 28 were reintubated. The implication here is that the 18 that were not reintubated, presumably were judged not to need further assistance with their breathing and therefore, almost by definition, were oversedated just prior to the moment of unexpected extubation .
It is worth considering the other two papers quoted above, in passing as they add to our discussion of the issue. The Krinsley (et al 2005) paper has just been published with the results of a fairly substantial cohort of patients (100) who had unexpected extubation and compared them to a control group of 200 patients (all mechanically ventilated)
This paper did not specifically look at the factors that were associated with unexpected extubation, but concerned itself mainly with the factors that were the consequence of unexpected extubation. The authors note that there was a mortality associated with unexpected extubation of 20% which is at distinct variance to the Boulain study (he cited only one death in the unexpected extubation group). This paper cites a reintubation rate of 56% which approximates to the rate of the Boulain study.
Interestingly, the author performed multiple logistic regression calculations on the results and found that:
age was the only predictor of the need for reintubation after unexpected extubation and that age and the need for reintubation were the only predictors of mortality after unexpected extubation.
These are results which we have not seen in any other published work. The author also comments that there is a statistical correlation between the actual event of unexpected extubation and an increase in the length of stay in the ITU but interestingly, there is also a statistical correlation with a reduction in the mortality rate. (Chevron et al 1998)
Further statistical analysis showed that the differences in the mortality rate corresponded with the perceived need for reintubation. The patients who were not judged to need reintubation universally had “remarkably good outcomes”. The paper concludes with the comment that:
It remains incumbent on ITU teams to institute protocols for regular identification of patients ready to be liberated from mechanical ventilation.
The Epstein paper (Epstien et al 2000), is ostensibly on the same issue of unexpected extubation, but it appears to be much more clinically orientated than the other two. For this reason it certainly merits consideration. The study itself was rather smaller than the Krinsley study, with 75 study patients and a control cohort of 150. Significantly, in this case, the controls were matched for “acute Physiology and Chronic Health Evaluation II score, presence of comorbid conditions, age, indication for mechanical ventilation, and sex.” This makes for better statistical analysis of the overall results.
The authors here found similar results in terms of the patients who required reintubation after UEX had significantly longer stays in ITU and higher mortality rates.
A significant features of this study is a particularly insightful resume of the situation pertaining to unexpected extubation in which the authors review the published literature (obviously this does not include the Krinsey figures). They point to the fact tha unexpected extubation occurs in 3-16% of mechanically ventilated patients. (Zwillich et al 1999). The authors present the rather pragmatic view that:
Successfully managed unplanned extubation has the potential of improving outcome by shortening the duration of intubation, thereby reducing the patient's exposure to complications of mechanical ventilation.
They also state the obvious corollary to this argument that a failure to tolerate unexpected extubation has the potential to reduce the chances of a good outcome by subjecting the patient “to the complications of a premature removal of (demonstrably) needed ventilatory support. It should therefore not be a matter of surprise that some studies have shown an increase in mortality in that group of patients who effectively failed and episode of unexpected extubation when directly compared to those who did not need reintubation. (Atkins et al 1997).
As we have observed with other articles in this review, the authors comment that only limited conclusions can be drawn from the comparison between these two groups simply because most patients who failed an unexpected extubation (and therefore needed reintubation) had been on full ventilatory support and this contrasts to the group who successfully tolerated unexpected extubation (no reintubation) who were generally those patients who were in the process of weaning trials.
They follow this up with the eminently sensible suggestions that:
Therefore, the unique impact of unplanned extubation on outcome is better studied by comparison with controls not experiencing unplanned extubation. The majority of previous studies, including the only published case-control analysis, suggest no increase in mortality when comparing patients with and without unplanned extubation.
If we consider the figures obtained in the Boulain (1998) study and analyse then in this way then we find that there is actually no increase in mortality if these two groups are directly compared.
This study shows that the mortality that was reported in the other papers was actually almost independent of the fact of the unexpected extubation and directly statistically related to the severity of the underlying illness, the actual cause of the respiratory failure in the first place and also the presence of any other co-existing morbidity. (Torres, A et al 1995).
The authors also point to the fact that two other recent studies (Epstein et al 1998) and (Esteban et al 1999) have shown that patients who had been reintubated within 12 hours of a planned extubation had actually a lower mortality rate than those who were reintubated later than 12 hours.
Taking an overview of this paper, the only practical positive outcome that it presents (apart from the demolition of findings in other papers)
Is that a successfully tolerated unexpected extubation reduced the duration of weaning trials but “had no other measurable beneficial effect on outcome”
One of the reasons for presenting the data contained in these last few papers in comparative detail, is that it demonstrated clearly the difficulties encountered in making a sufficiently uncritical appraisal of the data and statistical analysis of the figures presented. If one makes a sufficiently critical analysis of the data and study design then some facts can be seen as virtually unchallengable and others can be seen as seriously flawed. (Christie, J. M. et al 1996),
A number of papers that we have read in preparation for this review are centred on the development of a clinically applicable scale that can be used to accurately and reproducibly ascertain the level of sedation. This is clearly a fundamental problem in the consideration of the thrust of this review as correlating the amount of sedation given with the length of time spent on a ventilator is of no value at all as the effect of a given amount of sedation on a 30 stone fit man will clearly be considerably less than on a frail 8 stone elderly lady. Equally there is a huge spectrum of sedatives and indeed, sedative cocktails, that are in common current usage.
This is an equally valid point of consideration as the type or mixture of sedatives is not nearly as important as the degree of sedation that the particular prescription actually produces. In this next section of the review we shall consider papers that are primarily concerned with this particular point – the ability to quantify the degree of sedation produced by any particular dosage regime in any given patient.(Sessler W 2004)
Many of the papers that we have assessed thus far have used competitively crude tools as a measure of the degree of sedation experienced by a patient. Some groups have been exploring more subtle and easily reproducible methods of determining the degree of sedation.
One eye-catching recent paper (Haenggi et al 2004) considered and investigated the possibility of assessing the degree of sedation by producing evoked potentials with auditory stimuli. This particular study was done using 10 volunteers who agreed to be sedated to clinical levels with a number of different sedatives. The paper itself is both long and technical, but the results can be condensed into the statement:
Our results suggest that long latency auditory evoked potentials provide an objective electrophysiological analogue to the clinical assessment of sedation independent of the sedation regime used.
The clinical implications of this study are considerable. The authors were able to demonstrate that acoustic stimuli are capable of producing distinct and discrete changes in the electroencephalogram (EEG). These changes can be used without demonstrable harmful effect to monitor the degree of sedation virtually continuously in clinical conditions and independently of the type of drug that is actually being used to produce the sedative level. (Roustan, J.-P et al 2005).
The authors point out that considerably more work k needs to be done in order to calibrate and fully implement such observations in a clinical setting, but an initial assessment would suggest that there may well be considerable clinical potential for this particular application.
Sadly, the prognostic predictions of the Haenggi paper are not confirmed by a paper from Rundshagen (et al 2002). His group also considered the possibility of using auditory evoked potentials to assess the depth of sedation. They specifically targeted the midlatency auditory evoked potentials (MLAEP) Na Pa & Nb potentials and found them not to be of any clinical significance in assessing the degree of sedation. Obviously this does not negate the findings of the Haenggi group, but simply underlines the specificity of their findings.
It is important to consider this work in the chronological context. Thornton (et al 1998) also considered the evoked response in the Nb range and were able to conclude only that the abolition of the AER three wave pattern was indicative of the attainment of the level of sedation where auditory awareness is lost
In another article, Pockett S (1999) reviews the situation further and concludes that although Auditory evoked potentials in the midrange may seem to have the potential for assessment of the level of sedation, he believes that work needs to be done in the high and low range of evoked potentials to fully ascertain the true potential of the method.
Another recently published paper (Roustan et al 2005) takes the points explored here and raises the bar further with the evaluation of a full EEG analysis to determine any correlation with the overall degree of sedation.
The authors point to the fact that there are no truly (or sensitively) reliable clinically based scales to assess the degree of sedation of a patient.
(De Jong, M. M. J. et al 2005)
This particular trial is based on the fact that EEG (spectral and bispectral analysis) parameters have been calibrated and used in the past in the form of an index to monitor the depth of operative anaesthesia. They also point to the fact that there have been no published attempts to correlate these findings with a similar application relating to the degree of sedation in an ITU. The trial comprised over 160 EEG recordings from 40 patients. All of the patients were sedated with either midazolam alone or a combination of midazolam and morphine.
The authors set out to attempt to try to correlate any of the possible measurable parameters of the EEG with two of the most well established clinical sedation scales – the Ramsay and Comfort scales
The authors attempted to correlate which of the parameters were directly related with either too light a state of sedation (Ramsay 1 or 2) or too deep a state of sedation (Ramsay 5 or 6). The paper presents, in great detail, the analysis of the parameters and their correlation (or otherwise) with the level of sedation achieved. For our purposes here we will observe that two of the parameters (called ratio10 and SEF 95) were closely related with the level of sedation, and the relationship was marginally better if the two results were added together.
The results were found to be highly indicative of the degree of sedation in any one individual patient but there was a large degree of interindividual variability. It was also described that the authors found that bispectral evaluation improved the sensitivity over simple spectral analysis. It follows from this that the technique can be used to great effect once it has been calibrated to the individual patient but it is unlikely that one all-encompassing index could be achieved to apply to all patients.
Another paper (Ely et al 2003) considered how to tackle the problem of assessing the depth of sedation in the ITU patient. At present there are a number of different methods (we have already described some) and a number of different scales that can be used to assess sedative levels. The practical difficulty is that it is difficult to interpret the results of one scale in relation to another. The Ely paper set about a direct comparison between several of the commonly used scales and a common measuring point – the Richmond Agitation – Sedation Scale (RASS). The methods and scales compared included:
RASS, Glasgow Coma Scale (GCS), and Ramsay Scale (RS); validity of the RASS correlated with reference standard ratings, assessments of content of consciousness, GCS scores, doses of sedatives and analgesics, and .bispectral electroencephalography.
The study was designed as a prospective cohort study of over 300 patients. The design was quite ingenious as each patient was independently assessed by two nurses, each using one particular assessment tool. In some cases the nurses assessed the same patient using the same tool independently (but blinded to each other) in order to assess interrater reliability.
Again, the statistical analysis presented in the paper is quite formidable and not appropriate to present here but the important overall conclusions were that the RASS gave the best degree of correlation with all of the available indices. The authors felt able to state, at the end of their paper:
The RASS demonstrated excellent interrater reliability and criterion, construct, and face validity. This is the first sedation scale to be validated for its ability to detect changes in sedation status over consecutive days of ICU care, against constructs of level of consciousness and delirium, and correlated with the administered dose of sedative and analgesic medications.
The argument is examined further, in a slightly different direction in the De Wit (et al 2003) paper. These authors performed a similar exercise to the Ely team (above) insofar as they compared the assessments of sedation level in patients who were receiving sedatives both by continuous infusion and also by other (intermittent) means. They looked at the Sedation-Agitation scale and the bispectral index as evaluations of the sedative level. The cohort was only 19 patients and they analysed 80 assessments on those 19 patients. As a general finding, the authors point out that those patients receiving continuous sedation were more deeply sedated than those who received bolus sedation. This is actually the only really practically useful finding in the paper as the general conclusions that the authors reach are:
Objective and subjective assessments of sedation are highly correlated. Use of continuous infusions is associated with deeper levels of sedation, and patients receiving continuous infusions are more likely to be oversedated. Sedation therapy should be guided by subjective or objective assessment.
Which, in reality, does little more than to simply confirm the results of many other published works on the subject.A more recent paper by Watson & Kane-Gill (2004) also covers effectively the same ground and does not provide any further help in the assessment of the validity of the various scales
If we consider the relevance of the agents used, there are a number of excellent papers on the subject.
There are clearly a number of sedatives that are commonly used in the ITU setting for the sedation of the ventilated patient. They obviously all have the same macro-effect, but are all subtly different in their mode of operation and their side effect profiles. Their use is determined largely by the personal preference of the lead clinician and his experience together with any protocols that may be in force in any one particular ITU.
In our consideration of the length of time patients spend on mechanical ventilation, we should also properly consider the effect that the choice of the sedative agent has on the overall clinical outcome.
To that end we can start with a very recent paper by Arroliga (et al 2005) which considered the effects of sedatives and neuro-muscular blockers (NMBs) on patients who were on mechanical ventilators.
In effect, this study was a meta-analysis of the use of these agents as the entry cohort was enormous (over 5,000 patients).
The first (and probably self evident) result that the study draws is the fact that those patients who were sedated for their mechanical ventilation spent a statistically significantly longer time on both ventilation and weaning periods than those patients who did not receive sedation. They also spent longer in the ITU than those who did not