Abstract
Cost-effectiveness of early parenteral versus early enteral nutrition in critically ill patients
MZ Sadique, RD Grieve, DA Harrison, KM Rowan and for the CALORIES Trial Investigators
Duration of antimicrobial therapy for bloodstream infections in critically ill patients in the United Kingdom
TP Hellyer, K Leonard-Bedwell, J Furneval, R Fowler, N Daneman and P Dark
Acute MRI enhances prognostication in traumatic brain injury
Andrew D Kane, Joanna P Simpson, Ben Glocker, Tilak Das, Joanne G Outtrim, David K Menon and Virginia FJ Newcombe
Comparison of clinical sedation scores with the Responsiveness Index (RI): analysis of routine clinical data in the development and evaluation of strategies to improve sedation practice in Intensive Care (DESIST) study
A Stirling, J Antonelli, K Kydonaki, T Quasim, J Ruddy, A Davidson, J Rutherford, P Peltola, MOK Särkelä, K Uutela, TS Walsh and for the Development and Evaluation of Strategies to Improve Sedation practice in inTensive care (DESIST) study investigators
Surviving sepsis: one-year survival following a decade of whole systems audit
M Chikhani, R Sinha, S Wood, E Blyth, A Hutchinson, V Weston and M Simmonds
Cost-effectiveness of early parenteral versus early enteral nutrition in critically ill patients
MZ Sadique1, RD Grieve1, DA Harrison2, KM Rowan2 and for the CALORIES Trial Investigators
1Department of Health Services Research and Policy, London School of Hygiene and Tropical Medicine, London, UK
2Clinical Trials Unit, Intensive Care National Audit & Research Centre, London, UK
Early and appropriate nutritional support is recommended for critically ill patients. Evidence is conflicting regarding the optimum route of delivery and robust cost-effectiveness evidence is lacking. The aim of the study was to evaluate the cost-effectiveness of early nutritional support via the parenteral route versus early nutritional support via the enteral route in critically ill adults.
We undertook a cost-effectiveness analysis (CEA) using data from a large, pragmatic, multi-centre randomised controlled trial, the CALORIES trial, which recruited patients from 33 adult, general critical care units in England.1 Resource use and outcome data on 2388 trial patients were used to report cost-effectiveness at 90 days and at 1 year, and to project lifetime cost-effectiveness. The CEA used information on health-related quality of life at 90-days and at 1 year combined with information on vital status to report Quality-Adjusted Life Years (QALYs). Each QALY was valued using the NICE recommended threshold of willingness-to-pay (£20,000 per QALY) in conjunction with the costs of each route to report the incremental net monetary benefits (INB) of early nutritional support via the parenteral versus the enteral route.
Mean health-related quality of life at 90 days and 1 year was similar between the randomised groups. At 1 year, the parenteral route group had, on average, higher costs (mean difference £1580, 95% confidence interval [CI] −£792 to £3951), similar QALYs (0.013, 95% CI −0.014 to 0.040) and negative INB (−£1,320, 95% CI −£3709 to £1069) compared to the enteral route group. When a lifetime time horizon was taken, the INB of the parenteral route was positive but with considerable uncertainty surrounding the result (£440, 95% CI −£3586 to £4466). The probability that early nutritional support via the parenteral route is more cost-effective than via the enteral route was 14% at 1 year and 58% over the lifetime.
For critically ill adult patients, providing early nutritional support via the parenteral versus the enteral route is unlikely to be cost-effective. However, it should be recognised that considerable uncertainty surrounds the cost-effectiveness results, especially when projected over the patient’s lifetime.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: NIHR Health Technology Assessment Programme (07/52/03).
Reference
Duration of antimicrobial therapy for bloodstream infections in critically ill patients in the United Kingdom
TP Hellyer1, K Leonard-Bedwell2, J Furneval3, R Fowler4,5, N Daneman4 and P Dark2
1Department of Anaesthesia, City Hospitals Sunderland, Sunderland, UK
2Institute for Inflammation and Repair, University of Manchester, Manchester, UK
3Intensive care unit, City Hospitals Sunderland, Sunderland, UK
4Departments of Medicine and Critical Care Medicine, Toronto, Canada
5Sunnybrook Health Sciences Center, Toronto, Canada
The optimum duration of antibiotic therapy for bacteraemia is unknown. Only one trial, in neonates, has compared shorter versus longer therapy.1 A recent report of Canadian practice demonstrated that the median duration of appropriate treatment was 14 days (interquartile range 9–17.5) among patients with a predicted mortality of 46% (actual 27%).2 In contrast, UK data from a single centre demonstrated shorter courses of antibiotics – a median of 5 days.3 We aimed to determine the duration of antibiotic therapy in 2 general UK ICUs.
A retrospective analysis was performed on prospectively collected data from Salford Royal Hospital. Data was retrospectively collected from City Hospital Sunderland from electronic records and paper case notes, using the BALANCE trial4 case report form. Patients who had a positive blood culture with a pathogenic organism while admitted in a critical care area were included. Patients were excluded if the positive culture was a single organism representing a probable contaminant (coagulase negative Staphylococci, Bacillus spp., Corynebacterium spp., Propionbacterium, spp., Aerococcus spp. or Micrococcus spp.) or if there was a diagnosis with an established need for prolonged antibiotic treatment (infective endocarditis, osteomyelitis, septic arthritis, undrainable abscess or unremovable prosthetic material). Adequate antibiotic therapy was defined as at least one antibiotic with activity against all pathogens cultured. Duration of therapy was measured as consecutive days of adequate therapy from the date of taking the index culture without interruption of >24 hours.
Among 84 consecutive bacteraemic patients (37 from Sunderland, Dec 2011–Oct 2014 and 47 from Salford, Aug 2010–Jan 2013), the mean age was 58 years (SD 15), 44 (52%) were male, 66% were medical admissions, 25% surgical and 10% neurological/neurosurgical admissions. Median duration of ICU stay was 17 days (IQR 8–28 days) and median duration of hospital stay was 28 days (IQR 15–48 days). The index culture was monomicrobial in 83.3% of patients and polymicrobial in 16.7%. Bacteraemia occurred while on antibiotics in 38 (45%) patients. The median duration of adequate antimicrobial therapy was 6 days (IQR 2–10 days). Excluding patients who died within 10 days while receiving antibiotics and thereby accounting for immortal time bias, the median duration was 7 days (IQR 4–11 days). Two patients (2.4%) had breakthrough bacteraemia. ICU mortality was 38% and hospital mortality was 42%.
These data demonstrate practice variability in bacteraemia treatment in the UK, and in comparison to other countries among patients with high overall mortality but low rates of breakthrough bacteraemia. This variability highlights the need for a trial to determine optimal duration of therapy for bacteraemia in critically ill patients.
References
Acute MRI enhances prognostication in traumatic brain injury
Andrew D Kane1, Joanna P Simpson1, Ben Glocker2, Tilak Das3, Joanne G Outtrim1, David K Menon1 and Virginia FJ Newcombe1
1University Division of Anaesthesia, University of Cambridge
2Faculty of Engineering, Imperial College London
3Dept. of Radiology, Addenbrooke’s Hospital, Cambridge
Traumatic brain injury (TBI) is the highest cause of death under 40 years of age.1 Improvements in road safety, intensive and neurosurgical care enhance survival but may increase unfavourable outcomes (e.g. DECRA).2 Therefore early prognostic tools are essential to allow informed decisions about continued treatment. Current TBI outcome databases, e.g. IMPACT, predict outcome from initial clinical, lab and CT data with no adjustment for injury progression during intensive care.3 Recent advances in magnetic resonance imaging (MRI) provides an exciting opportunity to refine and individualise TBI prognostication. Here we hypothesised that a quantitative acute MRI was superior to IMPACT at differentiating between favourable vs. unfavourable outcome and survival vs. death at 6 months from injury.
Moderate and severe TBI patients requiring neuro-critical care and able to undertake a research MRI within 7 days of injury were recruited. Traumatic lesion volumes (core contusion and oedema in the hemispheres, basal ganglia and brain stem and subdural blood) were quantified by manually drawing regions of interest on FLAIR and gradient echo MRI. Clinical variables required for the IMPACT model were collected. Multiple analysis of variance (MANOVA) was performed to determine if MRI added statistical explanation of variance (i.e. partial η2) in patient outcome. These data generated multivariate linear models of prediction which were tested against the IMPACT model for superiority.
One-hundred-and-twenty-five patients were recruited between 2006 and 2014 (Median age 33 y, range 16–72 y; 94 M:29 F). MANOVA revealed that MRI approximately doubled the partial η2 from 18.2% to 30.0% for unfavourable outcome and from 22.1% 44.9% for death compared to IMPACT model variables in this population. Core hemispheric contusion (P = 0.002) and oedema (P = 0.014) were associated with unfavourable outcome whereas brainstem (P = 0.019) and basal ganglia oedema (P = 0.001) were associated with death. A linear prediction model with MRI-derived lesion volumes and clinical data significantly increased the area under the ROC curve compared against IMPACT for the risk of unfavourable outcome (Figure 1), but not for death.
(a) ROC curves for favourable vs. unfavourable outcome constructed from the published IMPACT core model and MRI model in this cohort. (b) Mean ± SEM area under ROC curves, *P < 0.05 vs. IMPACT.
Incorporation of acute MRI lesion volumes increased statistical explanation of variation in outcome and generates a linear prediction model superior to IMPACT, a current gold standard. Development of automated lesion detection software4 will allow straightforward quantification of brain injury enabling prospective validation and translation of the model into the clinical environment.
References
Comparison of clinical sedation scores with the Responsiveness Index (RI): analysis of routine clinical data in the development and evaluation of strategies to improve sedation practice in Intensive Care (DESIST) study
A Stirling1, J Antonelli1, K Kydonaki1, T Quasim2, J Ruddy3, A Davidson4, J Rutherford5, P Peltola6, MOK Särkelä6, K Uutela6, TS Walsh1 and for the Development and Evaluation of Strategies to Improve Sedation practice in inTensive care (DESIST) study investigators
1Anaesthetics, Critical Care and Pain Medicine, University of Edinburgh, Edinburgh
2University Department of Anaesthetics, Glasgow University, Glasgow Royal Infirmary, Glasgow, Scotland
3Department of Anaesthetics, Monklands Hospital, NHS Lanarkshire, Scotland
4Department of Anaesthetics, Victoria Infirmary, NHS GGC, Glasgow, Scotland
5Department of Anaesthetics, Dumfries hospital, NHS Dumfries and Galloway, Scotland
6GE Healthcare Finland Oy, Kuortaneenkatu 2, Helsinki, Finland
DESIST was a quality improvement trial comparing the impact of education, sedation quality feedback data, and the introduction of a novel sedation-monitoring technology (Responsiveness Index (RI)) in eight Scottish ICUs.1 RI continually acquires and analyses facial electromyelogram (fEMG) data, and uses an algorithm that utilises the previous 60 minutes of data to calculate an RI index (range 0–100) representing the frequency and intensity of patient arousals.2 Data are colour coded RED (RI 0–20; high probability of deep sedation); AMBER (RI 20–40); intermediate risk); and GREEN (>40; low risk).3 Four ICUs utilised RI monitoring in the trial. This study reports the comparison between RI data and bedside nurse RASS/SAS score assessments undertaken intermittently as part of routine practice. Validation studies have demonstrated correlation between the two systems, although they are not directly comparable.3,4
Paired clinical sedation score and RI data were available for 46 patients. Clinical scoring used the RASS or SAS score according to local practice. We developed data extraction rules to select discrete periods with apparently stable clinical status, namely: a stable sedation status for three sequential hours, minimum 12 hours between eligible data pairs and the higher or midpoint of any split/ranged scores. For clinical sedation scores we defined three states as: clinically unresponsive (SAS 1–3; or RASS −5 to −3), optimum sedation state (SAS 4–5; or RASS −2 to +1) and agitated states (SAS 6–7; or RASS +2 to +4). We analysed data with both RI and clinical sedation score as independent variable. We reported percent concordance between categories.
Comparison of RI monitor results with RASS/SAS scores and comparison of RASS/SAS scores with RI monitor results.
These routinely collected data highlight the lack of direct comparability between RI values and clinical sedation scores, and represent poorer correlation than found in previous controlled studies.3,4 RI is a composite of sedation status, the intensity of clinical stimulation, and underlying brain function over 60 minutes of care, unlike clinical assessments at specific time points. Another possible explanation is poor reliability of clinical scores in routine practice. Our data highlight the need for good education and protocols to ensure appropriate use of RI technology in the clinical setting.
References
Surviving sepsis: one-year survival following a decade of whole systems audit
M Chikhani1, R Sinha1, S Wood2, E Blyth2, A Hutchinson2, V Weston2 and M Simmonds2
1Anaesthesia and Critical Care, Division of Clinical Neuroscience, School of Medicine, University of Nottingham, UK
2Queen’s Medical Centre, Nottingham University Hospitals NHS Trust, Nottingham, UK
Following the first recommendations of the Surviving Sepsis Campaign guidelines,1 the multidisciplinary Sepsis Action Group at the Queen’s Medical Centre, Nottingham University Hospitals NHS Trust, was created to explore the treatment of patients with severe sepsis and compliance to treatment recommendations. Each iteration of the international guidelines has provided the opportunity for expansion of our audit. The first loop of the audit was carried out in 2005/6, the second in 2009/10, the third in 2013/14. Interim interventions to improve quality of care were carried out between each cycle including dissemination of results and specific education seminars.
Institutional ethical waiver was granted by the Caldicott Guardian. Data were collected using a bespoke and previously validated sepsis audit tool.2 In 2005/6 and 2009/10 adult patients were identified with significant positive blood cultures and included if they fulfilled the pathophysiological criteria for severe sepsis. A more streamlined approach was taken in 2013/14, identifying patients as those admitted to critical care with the diagnosis of severe sepsis, prior to microbiological diagnosis. The duration of the audit was over a 5-month period between November and March for the first two cycles, and over a 27-month period for the third cycle. The period between November 2013 and March 2014 will be used for survival comparison.
There were a total of 46 patients who met the criteria for study in 2005/6, 89 in 2009/10 and 101 in 2013/14. Number of patients with full adherence to the recommended treatment items at 3 and 6 hours was 7 (15%) and 4 (9%) in 2005/6, 13 (15%) and 9 (11%) in 2009/10 and 77 (77%) and 55 (55%) in 2013/14. Figure 1 demonstrates survival in days from the onset of severe sepsis.
The data show consistently high rates of death immediately after onset of severe sepsis; this seems to plateau at 60 days. The apparent improvement in survival demonstrated between 2005/6 and 2009/10 appears to have been maintained in 2013/14 despite a streamlined selection method, which may have selectively included those who are more unwell, and possibly without bacterial infection. However, the massive improvement in bundle item compliance has not been associated with similar marked improvements in mortality in 2013/14. This is in keeping with recent prospective studies3 and may represent differences in the focus of post-resuscitation care, or patient-specific disease susceptibility.