Year in review 2012: Critical Care -management

Outcomes research plays a key role in defining the effects of medical care in critical care. Last year Critical Care published a number of papers that evaluated patient-centered and policy-relevant outcomes. We present this review article focusing on key reported outcomes associated with severe community-acquired pneumonia, mortality associated with decisions regarding triage to the ICU, and both short-term and long-term mortality associated with ICU admissions. We further analyze the literature, assessing outcomes such as quality of life and the psychological burden associated with critical care. We also reviewed processes of care, and studies looking at cost analysis of treatment associated with critical care.


Introduction
Unlike other medical research, key variables evaluating disease, the patient population and therapies are heterogeneous and diffi cult to defi ne in critical care research [1]. With respect to critical care research, both shortterm outcomes and long-term outcomes hold a key place in understanding the impact of our interventions. Last year Critical Care published a number of highly accessed, original articles in critical care research. We present this review article focusing on key reported outcomes including mortality, patient-assessed outcomes such as quality of life, processes of care and cost analysis.

Outcomes associated with severe communityacquired pneumonia
Estimates indicate that 10% of the 500,000 patients hospitalized with community-acquired pneumonia are admitted to the ICU [2]. Delay in the recognition of severe community-acquired pneumonia is associated with increased mor tality. Clinical prediction to assess severe community-acquired pneumonia is hampered not only by the absence of a unique defi nition for severe community-acquired pneumonia but also by the inclusion of ICU admission as a variable in the prediction model in many cases [3,4]. Marti and colleagues studied 11 diff erent prediction scores based on 20 variables to identify community-acquired pneumonia patients requiring ICU admission [5]. Older prediction scores (Pneumonia Severity Index and CURB-65) were not sensitive in identifying the need for ICU admission. Newer generation prediction rules such as those of the American Th oracic Society/Infectious Diseases Society of America 2007 were much more specifi c in predicting the need for ICU admission. Th e American Th oracic Society/Infectious Diseases Society of America 2007 minor criteria along with newer unvalidated scores such as severe community-acquired pneumonia performed the best and had an area under the curve of 0.85 in predicting ICU admission. Although the new prediction rules performed better, their lack of specifi c test characteristics to help in bedside decisions for ICU admissions makes their use limited in clinical practice [5].
Th e association of statins with the risk of developing pneumonia and any reduction in mortality associated with severe community-acquired pneumonia has had confl icting results [6,7]. Nielson and colleagues used data from the Danish national registry to perform a population-based case-control study to answer this debate [8]. Th ey evaluated 70,914 hospitalized pneumonia cases, and an unadjusted analysis using matching for age and sex showed an odds ratio (OR) of 1.17 (95% confi dence interval (CI) = 1.14 to 1.21) associating current statin use with a risk for developing pneumonia requiring hospitalization. After adjustment for confounders, the OR was reduced to 0.80 (95% CI = 0.77 to 0.83). Th is reduction in the rate of hospitalization was seen in all subgroups of current statin users, but this eff ect was lost in former statin users. Th e authors also analyzed the 30-day mortality of hospitalized patients and revealed a 27% decrease in mortality associated with pneumonia in current statin users (hazard ratio = 0.73, 95% CI = 0.67 to 0.79). Th ese results are in line with a number of other studies, but a few recent studies have not shown this protective eff ect of statin use, and the question of a healthy user eff ect in statin users has been proposed [9]. Th e authors perform a very well-structured study with careful consideration to address any residual con founding, and bring forward the possibility of a pleiotropic eff ect of statin use on both the risk of hospitalization and the overall outcomes associated with pneumonia. Th e impor tant questions of whether current statin users admitted to the ICU should have these drugs continued or whether statins should be initiated at the time of ICU admission ultimately await the results of ongoing trials.
Kojicic and colleagues evaluated a cohort of 596 patients to study the frequency and associated outcomes of acute lung injury [7] in patients with microbiologically proven pneumonia. Of these patients, 28.7% developed acute lung injury and the risk was higher in hospital pneu monia (38.5%) and healthcare-associated pneu monia (31.6%) when compared with community-acquired pneumonia (21.8%) [10]. Th e patients who developed acute lung injury associated with pneumonia had a markedly higher risk of hospital mortality even after adjustment for baseline severity of illness and comorbidities (OR = 9.7, 95% CI = 6.0 to 15.9). Th e authors also performed a nested case-control study with matching for site of isolation and the kind of pathogen to identify variables associated with the development of acute lung injury. After adjustment for baseline diff erences and the Pneumonia Severity Index, the use of inappropriate antimicrobial treatment (OR = 3.2, 95% CI = 1.3 to 8.5) and transfusions (OR = 4.8, 95% CI = 1.5 to 19.6) were independently associated with the development of acute lung injury. Th e authors performed a posthoc analysis that, after adjustment for these variables, showed the use of statins was associated with a decreased risk of developing acute lung injury. Th ese results present potential clinical targets for preventing progression to acute respiratory distress syndrome (ARDS) in patients with pneumonia.

Decisions regarding triage to the ICU and associated mortality
Th ere has been a signifi cant increase in the demand for critical services worldwide [11]. Th is increased demand places a huge constraint on bed provision for patients requiring these services [12]. Many patients are transferred to diff erent hospitals based on capacity reasons [13]. Earlier studies reported signifi cant adverse events associated with interhospital transfer [14], but recent data have shown that adverse events are extremely rare during interhospital transfer of critically ill patients [15]. Previous studies have also shown that transferred patients have longer ICU stays and a higher mortality when compared with nontransferred patients [13]. Barratt and colleagues performed a propensity-matched analysis of patients transferred to a diff erent ICU for nonclinical reasons defi ned as lack of bed capacity within 48 hours of admission to predict the OR for hospital mortality [16]. Despite the availability of guidelines to contain non clinical transfers in the United Kingdom, the authors found a signifi cant degree of variation in practice across the 28 critical care networks in England and Wales that they included in their analysis (range of adjusted OR = 0.29 to 4.90). When they matched controls based on propensity to these nonclinical transfers, however, they did not fi nd a diff erence in mortality among these two groups (relative risk = 1.01, 95% CI = 0.87 to 1.16; P = 0.94). Patients who were transferred to other hospitals did receive an extra 3 days of critical care as compared with patients who were not transferred [16].
An alternative strategy to interhospital transfer is to manage patients on site or in other locations such as operating theaters until an ICU bed becomes available. In a study from the United Kingdom, O'Callaghan and associates studied the eff ect of delays in ICU admission when patients were cared for in operating theater suites [17]. Th ey evaluated 1,609 patients, of which 149 patients (9.3%) had a delayed admission to the ICU. Th e median duration of delay was 6 hours (interquartile range = 4.5 to 10 hours). Th e authors did not see any diff erences in mortality and lengths of stay in the ICU and hospital as a result of delays in the operating suites. But they did see much higher rates of respiratory failure (25.5% vs. 15.6%, P <0.01) in this group of patients when compared with the no-delay group in the fi rst 24 hours after ICU referral. Lack of ICU beds is becoming a big problem in the United Kingdom and other western countries; Metcalfe and associates reported that 14% of ICU admissions in the United Kingdom were refused because of lack of availability of beds [18]. Owing to these issues around bed constraints, novel interventions such as suggested by O'Callaghan can help provide better care for critically ill patients. A large reason for the lack of diff erence in outcomes in this study might be due to the fact that, unlinked with previously published data, two anesthetic trainees managed the patients with senior input from an ICU specialist while they were awaiting ICU transfer. Th ese results might therefore be a refl ection of the importance of appropriate care for critically ill patients even in non-ICU settings. Of course, the safety of this approach depends on access to appropriate clinicians outside the ICU, which may not exist everywhere. Th e authors did not evaluate the eff ects of using alternate locations for intensive care on provid ing other hospital operative or emergency services.
Th e impact of bed occupancy for critically ill patients is perhaps seen most dramatically in disease outbreaks, such as the recent H1N1 pandemic [19]. Th is need is more pronounced when a critical threshold is achieved and usual ICU care for elective procedures and so forth has to be discontinued [20]. Richard and associates used an online screening registry to study the patterns of ICU bed occupancy in diff erent ICUs in France during the H1N1 pandemic [21]. Th ey defi ned the critical threshold as an occupancy rate >15% by patients with the fl u at any given time. Th e study showed that there was a great variation in the peak occupancy based on the geographi cal region, on the time and on whether a hospital was university or nonuniversity affi liated during the infl uenza pan demic. Studies of this nature may help inform the development of electronic triaging and communication systems that will assist with real-time data analysis to help distribute ICU resources much more eff ectively, and can be an eff ective tool in data notifi cation for public health surveillance agencies.
Medical emergency teams are used to provide expert critical care services for clinical deterioration promptly in non-ICU wards. Th ere is a large variability in the roles of, composition of and services provided by these teams. Recent studies have questioned the impact of such teams on clinical outcomes [22]. Implementation of medical emergency teams is a complex, multistep process, however, and it is thought that a large driver of the negative results in studies is associated with diffi culty in changing provider behavior and processes of care in hospital settings [23]. Tobin and associates studied the impact of introduction of a medical emergency team on hospital mortality and evaluated individual hospital admissions over a 10-year period in 12 diff erent hospitals in Australia [24]. Th ey found that the benefi t of medical emergency teams was delayed, and the benefi t became apparent only after a medical emergency team had been active for more than 2 years in a hospital setting. After accounting for con founders, the mortality was not statistically diff erent from that in the period before the medical emergency team in the fi rst 2 years (OR = 0.99, 95% CI = 0.97 to 1.02), but hospital mortality decreased the longer the team had been in place in any institution (OR = 0.93, 95% CI = 0.91 to 0.96 for 2 to 4 years; and OR = 0.90, 95% CI = 0.88 to 0.92 where the system had been in place for more than 4 years). Similar to previously published data, this study again high lighted the fact that there is a large variation in the composition and role of the medical emergency teams. But more importantly the authors have shown that changes in outcomes associated with these teams took time to become apparent, but these improvements persist once hospital processes and culture are changed.

Gender and ICU mortality
Previously published data have shown that that women needing ICU care are less likely to receive aggressive treatment compared with men [25]. Studies from both Canada and Europe have shown that there are gender-based diff erences in mortality favoring men [26,27]. Mahmood and colleagues analyzed the Acute Physiology and Chronic Health Evaluation (APACHE) IV database to evaluate whether similar diff erences existed in American ICUs [28]. After adjustment for potential con founders, there was no statistical diff erence in mortality between the two groups (adjusted OR = 0.98, 95% CI = 0.95 to 1.02). Interestingly the authors found a statistically signifi cant interaction between gender and age, so they performed stratifi ed analyses. After the stratifi cation they found for patients <50 years of age that women had a lower mortality compared with men (adjusted OR = 0.83, 95% CI = 0.76 to 0.91). Among patients ≥50 years of age there was no statistically signifi cant diff erence in the ICU mortality between men and women. Similar to previous studies, however, they did see that men were more likely to receive mechanical ventilation, emergent surgery and coronary artery bypass grafting. Th is study points out that even though there is no diff erence in mortality between genders, there is variability in the provision of care for women when compared with men. Future research therefore needs to focus on understanding why this apparent diff erence in treatment intensity exists.

Light and the ICU
In a very interesting study Castro and associates evaluated whether exposure to light both prior to and during critical illness might infl uence mortality [29]. Th ey studied the impact of variation in regional photoperiod and insolation on winter immunoenhancement in these patients. After adjustment for confounders a 1-hour decrease in a cumulative 28-day photoperiod was associated with a signifi cant decrease in hospital mortality (adjusted OR = 0.997, 95% CI = 0.994 to 0.999; P = 0.03). Patients in the lowest quartile of the cumulative 28-day photoperiod had lower mortality risk compared with patients in the highest quartile (adjusted OR = 0.69, 95% CI = 0.47 to 1.01; P = 0.06). Th ere was no diff erence in the exposure to light after a patient was admitted to the ICU. Th e authors postulated that in the ICU the exposure light is of low intensity with minimal diurnal variation, and perhaps future studies need to manipulate light insolation and wavelength to study any biological eff ects in these artifi cial settings.

Delirium in the ICU
Delirium in critically ill patients is associated with increased morbidity and mortality [30,31]. Unfortunately standard clinical evaluation underestimates the prevalence of this diagnosis and recognition of delirium remains a diagnostic challenge in this patient population. Th e Confusion Assessment Method for the Intensive Care Unit and the Intensive Care Delirium Screening Checklist are the most commonly used scales for delirium evaluation in the ICU [32]. In a recent meta-analysis the pooled values of sensitivity (80.0%) and specifi city (95.9%) for the Confusion Assessment Method for the Intensive Care Unit to diagnose delirium in critically ill patients were very impressive. Th e area under the summary receiver operating characteristics curve was 0.97, which suggests excellent accuracy for this test. Th e equivalent value for the Intensive Care Delirium Screening Checklist was 0.89, which suggests good accuracy for this test as well [33]. Th e results of this meta-analysis thus support the use of the Confusion Assessment Method for the Intensive Care Unit or the Intensive Care Delirium Screening Checklist as screening tools for delirium in critically ill patients. Also, based on the high specifi city of the Confusion Assessment Method for the Intensive Care Unit, this is an excellent diagnostic tool for evaluating delirium in ICU patients. Of course, the clinical value of identifying delirious patients by screening has yet to be demonstrated because a specifi c intervention that off ers the best therapy, particularly for quiet delirium, has not been identifi ed.
Owing to the risks associated with the development of delirium in critically ill patients, addressing modifi able risk factors such as sedation management, minimization of sleep disruption and early mobilization of patients has become part of ICU management [34,35]. Van Rompaey and colleagues performed a randomized trial to see whether the provision of earplugs to ICU patients could aff ect the development of confusion or delirium in critically ill patients. Th ey studied 136 patients and followed the intervention and control groups over a period of 5 days [36]. Th e authors evaluated the development of confusion or delirium using the Neelon and Champagne Confusion Scale score in patients of both groups. Th ey found similar rates of delirium (19% vs. 20%) in the two groups, but the patients with earplugs developed mild confusion at a much lower rate (15% vs. 40%) compared with the control group. Th ese patients also reported signifi cantly better sleep during their fi rst night in the ICU. Th e onset of delirium in the intervention group also occurred later during the hospital course. A Cox regression analysis revealed that the use of earplugs reduced the rate of delirium by 53% (hazard ratio = 0.47, 95% CI = 0.27 to 0.82). In a multivariable model the rate of delirium was associated with age, Sequential Organ Failure Assessment score and a history of smoking. Th is study points out to the fact that targeted prevention can have a signifi cant impact on outcomes in critically ill patients.
Th ere are convincing data available that prove analgesia and sedation protocols improve the duration of mechanical ventilation, the length of ICU stay and survival in critically ill patients [34,37]. However, there is a wide variability in the implementation and perhaps even in the content of these protocols in diff erent ICU settings [38,39]. Dale and colleagues were able to demonstrate that higher volume teaching hospitals that were members of a larger hospital network were more likely to develop sedation, analgesia and delirium order sets, and also were more likely to follow appropriate guidelines to develop these protocols [40].

Long-term outcomes
Survivors of critical illness are at risk for both physical and psychological morbidity even after hospital discharge [41]. Th ese complications are a manifestation of the under lying injury, associated multiorgan failure and debilitation from the long ICU and hospital stay [42].

Long-term outcomes and health-related quality of life after ICU discharge
Neuropsychological and cognitive assessments and quality-of-life measures indicate that there is persistent morbidity after discharge from the ICU [41], but there are few data on long-term outcomes and health-related quality of life (HRQoL) in patients undergoing specifi c ICU-related procedures such as renal replacement therapy and extracorporeal life support [43,44]. Vaara and associates looked at the HRQoL and 6-month mortality in critically ill patients in Finland who underwent renal replacement therapy, as compared with the general ICU population. Th e HRQoL was assessed using the EQ-5D instrument [45]. Hospital mortality (35.0% vs. 15.5%, P <0.001) and 6-month mortality (49.4% vs. 27.8, P <0.001) of patients undergoing renal replacement therapy were signifi cantly higher compared with nonrenal replacement therapy patients. But the quality of life based on the EQ-5D at 6 months was similar in both groups. Th e severity of disease and overall mortality rates that were reported were much lower than those reported in previous studies. Also the results for the HRQoL have to be interpreted carefully, as follow-up for these patients was reported to be <50%.
Functional disability after ARDS is well documented [42]. Th e use of extracorporeal membrane oxygenation in ARDS with refractory hypoxemia has been associated with improved outcomes [46]. Th ere are limited data for the eff ects of extracorporeal membrane oxygenation on long-term outcomes and HRQoL in critically ill patients [44]. Hodgson and colleagues compared the HRQoL of Australian extracorporeal membrane oxygenation sur vivors with published data on HRQoL in patients with ARDS with and without the use of extracorporeal membrane oxygenation [47]. Th ey evaluated 21 patients, who received extracorporeal membrane oxygenation for a median of 10.6 (3.6 to 15.8) days: 86% survived to hospital discharge, and 15 patients consented to follow-up for long-term quality-of-life assessment and outcomes. Similar to previous data in ARDS patients, this cohort had signifi cantly lower HRQoL scores on a 36-item Short Form assessment compared with the general Australian popu la tion. When these patients were compared with other studies looking at these outcomes in ARDS patients, the individuals that underwent extracorporeal membrane oxygenation reported reduced HRQoL in vitality and social functions. Further studies with adequate sample sizes and appropriate follow-up are needed before we can fully understand the impact of extracorporeal membrane oxygenation or dialysis on HRQoL and whether the observed eff ect is due to the treatment or to the underlying severity of illness.

Psychological morbidity after ICU discharge
Survivors of critical illness have a high incidence of posttraumatic stress disorder (PTSD), depression and anxiety in the months following intensive care [48,49]. Th ese poor psychological outcomes might result due to a number of clinical, psychological and chronic health risk factors. Wade and colleagues used a validated questionnaire on 100 ICU survivors to study the prevalence of PTSD, anxiety and depression 3 months after ICU admission [50]. At follow-up, 27.1% of the patients had probable PTSD, 46.3% probable depression and 44.4% anxiety. Duration of sedation was the strongest clinical risk factor associated with PTSD, while use of benzo diazepines was associated with the development of depression. Use of inotropes or vasopressors correlated with anxiety at 3 months. Th e analysis also pointed towards acute psychological reactions as potential mediators for these clinical risk factors. After adjustment, psychological history and the mood of the patient in the ICU were associated with both depression and PTSD [50].
Strikingly high rates of psychological morbidity were found in this cohort of intensive care survivors. Th ese fi ndings suggest that psychological interventions, along with pharmacological modifi cations, could help reduce poor outcomes, including PTSD, after intensive care. Schandl and colleagues studied the gender diff erences in long-term symptoms of post-traumatic stress anxiety and depression in an ICU follow-up program with those not off ered such help. Th e authors used the Impact of Event Scale for post-traumatic stress and the Hospital Anxiety and Depression Scale for anxiety and depression [51]. Women without a follow-up had higher Impact of Event Scale scores than men without a follow-up. After adjustment for confounders, women in the intervention groups reported signifi cantly lower Impact of Event Scale scores when compared with the control group. Women in the control group reported higher Impact of Event Scale scores than men in the control group (median 31 ppoints vs. 10 p-points, P <0.01). Th is study highlights that women are more likely to develop psychological problems after ICU admission, but are also more likely than men to seek therapy for their symptoms [51].

Prognostication scores and sepsis
Prognostic application of scoring systems in critical care is limited [52]. Th ese systems rely on average risk associated with a population, so their ability to predict outcomes in individual patients is inadequate [53]. Recent studies suggest that soluble urokinase plasminogen activator receptor may be a marker of poor outcomes in septic patients [54]. Based on this research, a multicenter study from Greece analyzed the use of soluble urokinase plasminogen activator receptor in association with APACHE II to develop a new prognostication score to provide information regarding the risk of death in septic patients [55]. Th e authors found that the addition of this receptor to APACHE II did not change the receiver operator characteristics for prognosis of adverse outcomes in septic patients. But an APACHE II score ≥17 and a soluble urokinase plasminogen activator receptor ≥12 were independently associated with un favor able outcomes in these patients. Th e authors there fore used regression analysis to develop a prognostication rule. Th ey developed four strata based on these values, and found that the utilization of soluble urokinase plasminogen activator receptor was seen in high-risk patients not detected by APACHE II scores. Among patients with APACHE II score ≤17 the OR for death in patients with soluble urokinase plasminogen activator receptor ≥12 was 3.62. An independent cohort of Swedish patients con fi rmed prognostication by this rule [55].

Resource allocation
Nurse practitioners and physician assistants are staffi ng an ever-increasing number of ICUs in the USA [56]. Physicians supervise these physician assistants and nurse practitioners and in turn they provide patient care service including, but not limited to, history-taking, physical examination, implementing protocols and performing procedures [56]. In an interesting study by Carpenter and colleagues the authors evaluated how these practitioners allocated their time between patient care activities and nonpatient care activities [57]. Th e study looked at the pre-interven tion characteristics of the nurse practitioners and physician assistants, and based on billing information the average practitioner spent 53% (range 21 to 95%) of their time on patient care. Th e authors provided an incentive to these physician assistants and nurse practitioners, and data collection in the follow-up period revealed a 31% absolute improvement in documentation for patient care time. Th e authors postulate that this study demonstrates that these affi liates spend up to 85% of their time on patient care-related activities [57]. Th e rest of the time is spent rounding, providing sign-outs and performing chart reviews. Th is study provides some vital information for ICUs that plan to introduce nurse practitioners and physician assistants in their ICUs. Th ese data need to be put into context, however, because this study was carried out in a wide variety of coverage models, and nurse practi tioner and physician assistant staffi ng was depen dent in many cases on the availability of residents. Also these data are self-reported, and there was either inadequate or missing documentation for most of these practitioners until an incentive was introduced. Th erefore it is diffi cult to say that these roles would remain unchanged over longer periods of time.

Cost analysis of total parenteral nutrition
Th e results from the Impact of Early Parenteral Nutrition Completing Enteral Nutrition in Critical Illness (EPaNIC) trial showed that an early parenteral nutrition strategy was less eff ective than late parenteral nutrition because it resulted in higher rates of new ICU infections, a prolonged duration of organ replacement therapy and consequently an increased length of stay in the ICU [58]. Vanderheyden and associates carried out a cost analysis exploring the magnitude and drivers of the cost diff erence that resulted from the early parenteral nutritional strategy [59]. Th e investigators developed a cost tree based on the treatment allocation for the study patients and reported healthcare costs for the excess costgenerating events (new infections and prolonged ICU stay) that were identifi ed. Th ese two groups represented one-quarter of the ICU population each but almost onehalf of the costs each. Th e investi gators found that late parenteral nutrition reduced mean total incremental health costs by €1,110. Th e higher costs associated with the early parenteral nutrition group were mostly brought about by the higher costs for pharmacy, blood products and clinical chemistry. While the value of a formal costeff ectiveness analysis of early parenteral nutrition when the eff ectiveness analysis showed harm is questionable, the study provides signifi cant detailed information on the cost structures of intensive care.