Outreach, epistemology and the evolution of critical care

Epistemology is a branch of philosophy concerned with the study of knowledge [1]. A classical definition of knowledge proposes that knowledge lies in the intersection of truth and belief. Whilst mischievous, it is reasonable to say that around the subject of critical care outreach, there is much belief, but that truth doesn't always intersect with it. However, new hypotheses can arise from the apposition of the uncertain state of knowledge in one area of medicine with observations from another. As ever, it is up to the reader whether to believe new theories and down to the processes of empiricism to establish the truth. With this in mind we would like to share some thoughts that arise out of the experience with outreach and from critical care in general. At the turn of the last century there was much optimism within the ranks of intensivists. Intensive Care Medicine (ICM) achieved specialty status, trainees could look forward to gaining a Certificate of Completion of Training in ICM, and we were starting to get the results of large multicentre randomised controlled trials that, for the first time, brought some degree of order to the complexities of managing the critically ill [2, 3]. There was also a new mantra: 'Critical care without walls'. A seminal paper by McQuillan et al. [4] in the UK, published in 1998, had confirmed what we already suspected: that the ward care of the critically ill patient was often inadequate. At the same time, a new system of caring for the sick ward patient was coming of age in Australasia – the medical emergency team (MET) concept [5, 6]. The MET system replaced the traditional cardiac arrest team with a team of critical care doctors and nurses who would respond to calls by ward staff to patients who fulfilled simple calling criteria, with the aim of improving outcome by early intervention. The MET calling criteria were based on single parameters that often preceded and might lead to cardiac arrest. In the late 1990s, the concept became fashionable in the UK, but with attempts to improve the system's sophistication. Thus, by scoring multiple physiological parameters the so called 'track-and-trigger' systems were developed. Intensivists now felt empowered to take their skills out of the intensive care unit to resuscitate, admit to the intensive care unit where necessary and prevent death. The result was critical care outreach [7]. This promise of improved outcomes (maybe even reduced costs) was based on the concept of early intervention in the process of critical illness. Outcome from critical illness had been repeatedly shown to be a function of the number of failing systems and the duration of those system failures [8, 9]. The well studied inflammatory cascade, neurohumoral response, immunological response and the concept of organ interdependence represent the scientific basis for this observation, irrespective of whether the primary insult was infection, trauma or surgery. Therefore it was logical that the way to improve outcome should be prevention of the initiation of the inflammatory cascade (for example, early diagnosis of sepsis, early surgery, early antibiotics, early resuscitation) and interruption of the inflammatory amplification (using the same interventions) and therapies such as anti-inflammatory drugs and cytokine removal strategies. And indeed, such interventions have an evidence base [10, 11]. Thus, we had proof that there was a problem with the management of the sick ward patient and a ready made solution with seemingly impeccable provenance. Intensive care medicine was ready to hit the big time, and there was talk of hospitals' becoming critical care units with hotels tacked on the side to look after the well patient [12]. In the UK, there was an injunction from the Department of Health, that we should develop 'critical care outreach', and the UK's Intensive Care Society was moved to establish guidelines for the creation of such services. Some of us (SF) became proselytisers for the new quasi-religion 'outreach' [6]. Others were more cautious (BC) [13]. Initially, the believers appeared to be vindicated: intensivists were staggered to find so many sick patients (as identified by the available track-and-trigger systems) in the general wards of their hospitals. However, the more curious asked themselves: what had previously happened to these patients? The intensive care unit is the inevitable destination for any patient who is approaching death, unless an active decision is made not to admit or the patient experiences cardiopulmonary arrest and dies before admission. Therefore the intensivist would ultimately have come across these patients. Except that we weren't. Seemingly, there were very sick patients on the wards that had never previously come to our attention and presumably they just got better. Then came the evidence. Of course the studies were becoming difficult to perform because everyone knew that outreach improved outcomes, so how could one ethically randomly assign patients not to receive outreach care? At a practical level, most trusts were developing outreach systems just as they had been told to do, so that even if you could justify randomising a particular group of patients not to receive outreach care, this control group might be contaminated by the effect of outreach in the neighbouring ward or even hospital. The evidence, however, was striking. Outreach was failing to live up to expectations [14–16]. So the evidence did not support outreach and experience suggested that a large group of sick patients seemed to survive without help from the 'specialists'. What is happening? In attempting to find answers to this question one encounters issues of profound importance for the intensive care community. Firstly, however, there are more questions. Humans have evolved to survive potentially life-threatening insults, principally infection and trauma. For a person to die in response to such insults, a limited number of scenarios must pertain: The insult is overwhelming – for example, severe trauma or infection with highly virulent bacteria. The response to the insult is adaptive but cannot be sustained to the point of recovery. For example, in pneumonia, the respiratory muscles will increase their workload to maintain ventilation of damaged and non-compliant lungs, but eventually the muscles will fatigue and the patient will experience respiratory arrest. The response to the insult may be adaptive initially but becomes maladaptive. For example, the response to localised infection (vasodilation, capillary leak, white cell margination and activation) is adaptive, but becomes maladaptive when the inflammatory response becomes systemic (vasodilatory shock, hypovolaemic shock due to capillary leak, lung and kidney damage from cytokines and activated immune cells). Or consider the response to haemorrhage: the kidneys conserve fluid to restore circulating volume, but the persistent oliguria becomes maladaptive as the patient becomes uraemic. Scenarios 2 and 3 are individually determined, either by co-morbidity or genetics. Thus a patient with chronic obstructive pulmonary disease will experience respiratory muscle fatigue in the setting of pneumonia earlier than a young athlete and we know that the inflammatory response is determined by our genes: for example, different tumour necrosis factor alpha alleles confer different responses to infection [17]. What is the implication of this discourse? Perhaps, put crudely, some people are destined to live despite critical illness and perhaps some will inevitably die despite our best efforts, either through circumstance, co-morbidity or heredity. The proportion of the critically ill population whose outcome is likely to be modifiable by our interventions is perhaps smaller than we think. Yes. Many of the original track-and-trigger systems were based on crude, back-of-an-envelope guesswork, whereby readily measured parameters were lined up, assigned a weighting (or not) and scored. Such systems could tell you that a patient with a systolic blood pressure of 70 mmHg, an arterial blood oxygen saturation of 81%, a urine output of 5 ml.h−1, a pulse rate of 131 beats.min−1, and a respiratory rate of 35 breaths.min−1 was sick. For the healthcare assistants who perform the observations, this may have been useful, but for intensivists, however, it wasn't. Further refinement by means of large datasets and the application of regression analyses led to systems that are able to predict a high likelihood of death. Because the ultimate aim of outreach was to prevent cardiac arrest and death, this would appear to be a logical and noble aim of any track-and-trigger system. Except that they don't seem to work! Or at least, the combination of the system used to identify patients and the subsequent response doesn't seem to work. This leads to the next question, perhaps the most important for intensivists. William Osler wrote that 'patients do not die of their disease but rather of the physiological consequences of their disease'. Wise words indeed, and the logical response to this proposition supports, if not the very creation of the specialty of intensive care, at least to its main article of faith: 'normalisation of physiology'. For example, it is surely self-evident that in order to avert death, an apnoeic patient should be rendered non-apnoeic – for example, by means of tracheal intubation and artificial respiration. If the blood pressure is low then give pressor agents until the mean arterial pressure is normal; if the patient is anaemic give blood until the haemoglobin level is normal; if the patient is hypoxic then use mechanical ventilation to achieve normal blood gas values. If the patient is oliguric because of hypovolaemia then give intravenous fluid to promote urine flow. If there is overproduction of nitric oxide causing vasodilatory shock then block its production. Millions of lives have indeed been saved by intensive care treatments whose aim is also to normalise physiology. And yet 25% of patients still die on our intensive care units and application of this approach to patients identified by outreach doesn't seem to improve outcome. What is wrong with our interventions and treatments? All treatments have side effects and risks. For many intensive care interventions the risks are easily understood. For example, inotropes such as adrenaline increase the force of contraction of the heart and restore cardiac output, but in making the heart work harder, the heart's metabolic demands increase and this can precipitate myocardial ischaemia. For some intensive care interventions the risks are less obvious. Thus, whilst fluid resuscitation is mandatory in the setting of hypovolaemia due to sepsis, we know the increased capillary permeability caused by sepsis means that pulmonary oedema is likely to occur at lower pulmonary capillary pressures than in healthy subjects. This leads to one of the most enduring critical care dilemmas – do we sacrifice the kidneys by restricting fluid in order to keep the lungs dry, or restore urine flow and accept that we may need to provide respiratory support? Some of the risks of intervention are relatively arcane. For example, mechanical ventilation of the lungs can induce what is termed 'biotrauma', whereby lung tissue damage due to high shear forces induces an inflammatory response and subsequent systemic inflammatory response (with promotion of multi-organ failure) [3]. In the case of bacterial pneumonia, ventilation can furthermore cause bacterial translocation into the circulation. Thus, 'life saving' ventilation can precipitate multiple organ failure where only respiratory failure existed [18]. Where such risks exist, of course the intensivist will try to militate against them. For example, there are now consensus guidelines on fluid management that suggest that some cases of postoperative oliguria (being part of the adaptive response to trauma) should be managed by watchful waiting rather than aggressive fluid resuscitation [19]. In the case of mechanical ventilation, the recognition of harm prompted research that led to low tidal volume ventilatory strategies that reduce biotrauma and improve outcome [3]. However, for some patients it is entirely possible that critical care intervention (either specific therapies or procedures, or non-specific over-treatment) will do more harm than good. A study by Levy et al. [20], published last year, suggested just that: critical care management involved more interventions and resulted in worse outcomes than non-critical care. Whilst it is highly likely that this finding is spurious due to inadequate adjustment of confounding factors, nonetheless it tends to support the argument that aggressive treatment may not always be appropriate. Perhaps the 'poor' management of the critically ill that was exposed by McQuillan et al. [4] represented a process of selection (albeit unwittingly) by the admitting teams. That is, critically ill patients on the wards were receiving 'minimal' interventions and were eventually referred to the intensivist only when they had failed to improve spontaneously (although at this stage many patients would be heading towards multi-organ failure). This approach, whether by accident or design, meant that subjects with the wherewithal to recover were allowed to do so without being subjected to unnecessary intervention and over-treatment. So now we can return to the question of why critical care outreach doesn't seem to work. Perhaps one of the answers is that the track-and-trigger systems are based on a false premise, because at best they derive from application of multiple logistic regression to large datasets in order to identify parameters that predict death. We do not want to predict death; rather, we want to predict who will benefit from intensive care and who will survive without intensive care intervention and may actually be harmed by it. In addition, we need to know who will die despite all intervention; such people should not be denied as 'normal' a death as possible. At present, such systems do not exist, and until they do, it entirely likely that outreach will not fulfil its promise. H.C. Lassen and Bjorn Ibsen instigated the first epoch of critical care medicine by the mass ventilation of the lungs of patients with polio, with the second epoch coming with the large multicentre studies that have brought order to the chaos of how to treat patients on the intensive care unit. Perhaps the next epoch will come when we know whom to treat–with the development of targeted critical care.