Prediction of Complications and Prognostication in Perioperative Medicine: A Systematic Review and PROBAST Assessment of Machine Learning Tools.

BACKGROUND: The utilization of artificial intelligence and machine learning as diagnostic and predictive tools in perioperative medicine holds great promise. Indeed, many studies have been performed in recent years to explore the potential. The purpose of this systematic review is to assess the current state of machine learning in perioperative medicine, its utility in prediction of complications and prognostication, and limitations related to bias and validation. METHODS: A multidisciplinary team of clinicians and engineers conducted a systematic review using the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) protocol. Multiple databases were searched, including Scopus, Cumulative Index to Nursing and Allied Health Literature (CINAHL), the Cochrane Library, PubMed, Medline, Embase, and Web of Science. The systematic review focused on study design, type of machine learning model used, validation techniques applied, and reported model performance on prediction of complications and prognostication. This review further classified outcomes and machine learning applications using an ad hoc classification system. The Prediction model Risk Of Bias Assessment Tool (PROBAST) was used to assess risk of bias and applicability of the studies. RESULTS: A total of 103 studies were identified. The models reported in the literature were primarily based on single-center validations (75%), with only 13% being externally validated across multiple centers. Most of the mortality models demonstrated a limited ability to discriminate and classify effectively. The PROBAST assessment indicated a high risk of systematic errors in predicted outcomes and artificial intelligence or machine learning applications. CONCLUSIONS: The findings indicate that the development of this field is still in its early stages. This systematic review indicates that application of machine learning in perioperative medicine is still at an early stage. While many studies suggest potential utility, several key challenges must be first overcome before their introduction into clinical practice.

The Sequential Organ Failure Assessment (SOFA) Score: has the time come for an update?

The Sequential Organ Failure Assessment (SOFA) score was developed more than 25 years ago to provide a simple method of assessing and monitoring organ dysfunction in critically ill patients. Changes in clinical practice over the last few decades, with new interventions and a greater focus on non-invasive monitoring systems, mean it is time to update the SOFA score. As a first step in this process, we propose some possible new variables that could be included in a SOFA 2.0. By so doing, we hope to stimulate debate and discussion to move toward a new, properly validated score that will be fit for modern practice.

The oxygen cost of rehabilitation interventions in mechanically ventilated patients: an observational study.

OBJECTIVE: Early rehabilitation is assumed to be a crucial intervention to facilitate weaning from mechanical ventilation in critically ill patients and to limit their long-term functional dependence. However, little is known about the physiological load imposed on patients during such interventions. Without the ability to quantify the exercise intensity of rehabilitation interventions it is impossible to establish a clear separation between usual care and intervention groups in randomised controlled trials. This may explain the lack of definitive benefit of rehabilitation in published trials. We sought to characterise the physiological load, measured as oxygen consumption (V˙O2), of the physical activities carried out during rehabilitation interventions in mechanically ventilated participants. DESIGN: Observational study. SETTING: Single centre medical-surgical university hospital ICU. PARTICIPANTS: 26 mechanically ventilated participants ventilated >7 days, able to participate in a rehabilitation program. INTERVENTION: Oxygen consumption (measured by the Medgraphics Ultima breath-by-breath gas exchange analysis system) and heart rate were measured continuously pre-, during and post-standard rehabilitation sessions. RESULTS: 52 sessions were recorded in 26 participants. There was considerable variation in the oxygen cost of the physical activities between participants. The recovery time for 1 in 4 rehabilitation sessions was longer than the rehabilitation activity time. CONCLUSIONS: Absolute exercise intensity in mechanically ventilated ICU participants, as measured by oxygen consumption, is not activity-dependent.

Early, Goal-Directed Therapy for Septic Shock - A Patient-Level Meta-Analysis.

BACKGROUND: After a single-center trial and observational studies suggesting that early, goal-directed therapy (EGDT) reduced mortality from septic shock, three multicenter trials (ProCESS, ARISE, and ProMISe) showed no benefit. This meta-analysis of individual patient data from the three recent trials was designed prospectively to improve statistical power and explore heterogeneity of treatment effect of EGDT. METHODS: We harmonized entry criteria, intervention protocols, outcomes, resource-use measures, and data collection across the trials and specified all analyses before unblinding. After completion of the trials, we pooled data, excluding the protocol-based standard-therapy group from the ProCESS trial, and resolved residual differences. The primary outcome was 90-day mortality. Secondary outcomes included 1-year survival, organ support, and hospitalization costs. We tested for treatment-by-subgroup interactions for 16 patient characteristics and 6 care-delivery characteristics. RESULTS: We studied 3723 patients at 138 hospitals in seven countries. Mortality at 90 days was similar for EGDT (462 of 1852 patients [24.9%]) and usual care (475 of 1871 patients [25.4%]); the adjusted odds ratio was 0.97 (95% confidence interval, 0.82 to 1.14; P=0.68). EGDT was associated with greater mean (±SD) use of intensive care (5.3±7.1 vs. 4.9±7.0 days, P=0.04) and cardiovascular support (1.9±3.7 vs. 1.6±2.9 days, P=0.01) than was usual care; other outcomes did not differ significantly, although average costs were higher with EGDT. Subgroup analyses showed no benefit from EGDT for patients with worse shock (higher serum lactate level, combined hypotension and hyperlactatemia, or higher predicted risk of death) or for hospitals with a lower propensity to use vasopressors or fluids during usual resuscitation. CONCLUSIONS: In this meta-analysis of individual patient data, EGDT did not result in better outcomes than usual care and was associated with higher hospitalization costs across a broad range of patient and hospital characteristics. (Funded by the National Institute of General Medical Sciences and others; PRISM ClinicalTrials.gov number, NCT02030158 .).

Assessment of Clinical Criteria for Sepsis: For the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3).

IMPORTANCE: The Third International Consensus Definitions Task Force defined sepsis as "life-threatening organ dysfunction due to a dysregulated host response to infection." The performance of clinical criteria for this sepsis definition is unknown. OBJECTIVE: To evaluate the validity of clinical criteria to identify patients with suspected infection who are at risk of sepsis. DESIGN, SETTINGS, AND POPULATION: Among 1.3 million electronic health record encounters from January 1, 2010, to December 31, 2012, at 12 hospitals in southwestern Pennsylvania, we identified those with suspected infection in whom to compare criteria. Confirmatory analyses were performed in 4 data sets of 706,399 out-of-hospital and hospital encounters at 165 US and non-US hospitals ranging from January 1, 2008, until December 31, 2013. EXPOSURES: Sequential [Sepsis-related] Organ Failure Assessment (SOFA) score, systemic inflammatory response syndrome (SIRS) criteria, Logistic Organ Dysfunction System (LODS) score, and a new model derived using multivariable logistic regression in a split sample, the quick Sequential [Sepsis-related] Organ Failure Assessment (qSOFA) score (range, 0-3 points, with 1 point each for systolic hypotension [≤100 mm Hg], tachypnea [≥22/min], or altered mentation). MAIN OUTCOMES AND MEASURES: For construct validity, pairwise agreement was assessed. For predictive validity, the discrimination for outcomes (primary: in-hospital mortality; secondary: in-hospital mortality or intensive care unit [ICU] length of stay ≥3 days) more common in sepsis than uncomplicated infection was determined. Results were expressed as the fold change in outcome over deciles of baseline risk of death and area under the receiver operating characteristic curve (AUROC). RESULTS: In the primary cohort, 148,907 encounters had suspected infection (n = 74,453 derivation; n = 74,454 validation), of whom 6347 (4%) died. Among ICU encounters in the validation cohort (n = 7932 with suspected infection, of whom 1289 [16%] died), the predictive validity for in-hospital mortality was lower for SIRS (AUROC = 0.64; 95% CI, 0.62-0.66) and qSOFA (AUROC = 0.66; 95% CI, 0.64-0.68) vs SOFA (AUROC = 0.74; 95% CI, 0.73-0.76; P < .001 for both) or LODS (AUROC = 0.75; 95% CI, 0.73-0.76; P < .001 for both). Among non-ICU encounters in the validation cohort (n = 66 522 with suspected infection, of whom 1886 [3%] died), qSOFA had predictive validity (AUROC = 0.81; 95% CI, 0.80-0.82) that was greater than SOFA (AUROC = 0.79; 95% CI, 0.78-0.80; P < .001) and SIRS (AUROC = 0.76; 95% CI, 0.75-0.77; P < .001). Relative to qSOFA scores lower than 2, encounters with qSOFA scores of 2 or higher had a 3- to 14-fold increase in hospital mortality across baseline risk deciles. Findings were similar in external data sets and for the secondary outcome. CONCLUSIONS AND RELEVANCE: Among ICU encounters with suspected infection, the predictive validity for in-hospital mortality of SOFA was not significantly different than the more complex LODS but was statistically greater than SIRS and qSOFA, supporting its use in clinical criteria for sepsis. Among encounters with suspected infection outside of the ICU, the predictive validity for in-hospital mortality of qSOFA was statistically greater than SOFA and SIRS, supporting its use as a prompt to consider possible sepsis.

Protocolised Management In Sepsis (ProMISe): a multicentre randomised controlled trial of the clinical effectiveness and cost-effectiveness of early, goal-directed, protocolised resuscitation for emerging septic shock.

BACKGROUND: Early goal-directed therapy (EGDT) is recommended in international guidance for the resuscitation of patients presenting with early septic shock. However, adoption has been limited and uncertainty remains over its clinical effectiveness and cost-effectiveness. OBJECTIVES: The primary objective was to estimate the effect of EGDT compared with usual resuscitation on mortality at 90 days following randomisation and on incremental cost-effectiveness at 1 year. The secondary objectives were to compare EGDT with usual resuscitation for requirement for, and duration of, critical care unit organ support; length of stay in the emergency department (ED), critical care unit and acute hospital; health-related quality of life, resource use and costs at 90 days and at 1 year; all-cause mortality at 28 days, at acute hospital discharge and at 1 year; and estimated lifetime incremental cost-effectiveness. DESIGN: A pragmatic, open, multicentre, parallel-group randomised controlled trial with an integrated economic evaluation. SETTING: Fifty-six NHS hospitals in England. PARTICIPANTS: A total of 1260 patients who presented at EDs with septic shock. INTERVENTIONS: EGDT (n = 630) or usual resuscitation (n = 630). Patients were randomly allocated 1 : 1. MAIN OUTCOME MEASURES: All-cause mortality at 90 days after randomisation and incremental net benefit (at £20,000 per quality-adjusted life-year) at 1 year. RESULTS: Following withdrawals, data on 1243 (EGDT, n = 623; usual resuscitation, n = 620) patients were included in the analysis. By 90 days, 184 (29.5%) in the EGDT and 181 (29.2%) patients in the usual-resuscitation group had died [p = 0.90; absolute risk reduction -0.3%, 95% confidence interval (CI) -5.4 to 4.7; relative risk 1.01, 95% CI 0.85 to 1.20]. Treatment intensity was greater for the EGDT group, indicated by the increased use of intravenous fluids, vasoactive drugs and red blood cell transfusions. Increased treatment intensity was reflected by significantly higher Sequential Organ Failure Assessment scores and more advanced cardiovascular support days in critical care for the EGDT group. At 1 year, the incremental net benefit for EGDT versus usual resuscitation was negative at -£725 (95% CI -£3000 to £1550). The probability that EGDT was more cost-effective than usual resuscitation was below 30%. There were no significant differences in any other secondary outcomes, including health-related quality of life, or adverse events. LIMITATIONS: Recruitment was lower at weekends and out of hours. The intervention could not be blinded. CONCLUSIONS: There was no significant difference in all-cause mortality at 90 days for EGDT compared with usual resuscitation among adults identified with early septic shock presenting to EDs in England. On average, costs were higher in the EGDT group than in the usual-resuscitation group while quality-adjusted life-years were similar in both groups; the probability that it is cost-effective is < 30%. FUTURE WORK: The ProMISe (Protocolised Management In Sepsis) trial completes the planned trio of evaluations of EGDT across the USA, Australasia and England; all have indicated that EGDT is not superior to usual resuscitation. Recognising that each of the three individual, large trials has limited power for evaluating potentially important subgroups, the harmonised approach adopted provides the opportunity to conduct an individual patient data meta-analysis, enhancing both knowledge and generalisability. TRIAL REGISTRATION: Current Controlled Trials ISRCTN36307479. FUNDING: This project was funded by the National Institute for Health Research (NIHR) Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 19, No. 97. See the NIHR Journals Library website for further project information.

Potential economic consequences of a cardioprotective agent for patients with myocardial infarction: modelling study.

OBJECTIVE: To investigate the cost-effectiveness of a hypothetical cardioprotective agent used to reduce infarct size in patients undergoing percutaneous coronary intervention (PCI) after anterior ST-elevation myocardial infarction. DESIGN: A cost-utility analysis using a Markov model. SETTING: The National Health Service in the UK. PATIENTS: Patients undergoing PCI after anterior ST-elevation myocardial infarction. INTERVENTIONS: A cardioprotective agent given at the time of reperfusion compared to no cardioprotection. We assumed the cardioprotective agent (given at the time of reperfusion) would reduce the risk and severity of heart failure (HF) after PCI and the risk of mortality after PCI (with a relative risk ranging from 0.6 to 1). The costs of the cardioprotective agent were assumed to be in the range £1000-4000. MAIN OUTCOME MEASURES: The incremental costs per quality-adjusted life-year (QALY) gained, using 95% CIs from 1000 simulations. RESULTS: Incremental costs ranged from £933 to £3820 and incremental QALYs from 0.04 to 0.38. The incremental cost-effectiveness ratio (ICER) ranged from £3311 to £63 480 per QALY gained. The results were highly dependent on the costs of a cardioprotective agent, patient age, and the relative risk of HF after PCI. The ICER was below the willingness-to-pay threshold of £20 000 per QALY gained in 71% of the simulations. CONCLUSIONS: A cardioprotective agent that can reduce the risk of HF and mortality after PCI has a high chance of being cost-effective. This chance depends on the price of the agent, the age of the patient and the relative risk of HF after PCI.

Trial of early, goal-directed resuscitation for septic shock.

BACKGROUND: Early, goal-directed therapy (EGDT) is recommended in international guidelines for the resuscitation of patients presenting with early septic shock. However, adoption has been limited, and uncertainty about its effectiveness remains. METHODS: We conducted a pragmatic randomized trial with an integrated cost-effectiveness analysis in 56 hospitals in England. Patients were randomly assigned to receive either EGDT (a 6-hour resuscitation protocol) or usual care. The primary clinical outcome was all-cause mortality at 90 days. RESULTS: We enrolled 1260 patients, with 630 assigned to EGDT and 630 to usual care. By 90 days, 184 of 623 patients (29.5%) in the EGDT group and 181 of 620 patients (29.2%) in the usual-care group had died (relative risk in the EGDT group, 1.01; 95% confidence interval [CI], 0.85 to 1.20; P=0.90), for an absolute risk reduction in the EGDT group of -0.3 percentage points (95% CI, -5.4 to 4.7). Increased treatment intensity in the EGDT group was indicated by increased use of intravenous fluids, vasoactive drugs, and red-cell transfusions and reflected by significantly worse organ-failure scores, more days receiving advanced cardiovascular support, and longer stays in the intensive care unit. There were no significant differences in any other secondary outcomes, including health-related quality of life, or in rates of serious adverse events. On average, EGDT increased costs, and the probability that it was cost-effective was below 20%. CONCLUSIONS: In patients with septic shock who were identified early and received intravenous antibiotics and adequate fluid resuscitation, hemodynamic management according to a strict EGDT protocol did not lead to an improvement in outcome. (Funded by the United Kingdom National Institute for Health Research Health Technology Assessment Programme; ProMISe Current Controlled Trials number, ISRCTN36307479.).

Prevalence and outcome of cirrhosis patients admitted to UK intensive care: a comparison against dialysis-dependent chronic renal failure patients.

PURPOSE: Patients with decompensated liver cirrhosis who are admitted to intensive care units (ICU) are perceived, within the UK, as having a particularly poor prognosis. METHODS: We performed a descriptive analysis of cirrhosis patients admitted to general critical care units 1995-2008 compared to patients admitted with pre-existing chronic renal failure. Data were obtained from the Intensive Care National Audit and Research Centre Case Mix Programme Database incorporating 192 adult critical care units in England, Wales and Northern Ireland. RESULTS: Cirrhosis accounted for 2.6 % (16,096 patients) of total admissions with mean age 52.5 years and male preponderance (~60 %). Hospital mortality was high (>55 %) although this improved 5 % in recent years, and median length of stay was short (2.5 days). Mortality in cirrhotics with severe sepsis requiring organ support was 65-90 %, compared to 33-39 % in those without. Conversely, patients with chronic renal failure had lower mortality (42 %) despite similar characteristics and higher acute physiology and chronic health evaluation (APACHE) II scores. The APACHE II score under-predicted mortality in cirrhotics. CONCLUSIONS: Cirrhosis patients exhibit worse outcomes compared to pre-existing renal failure patients, despite similar characteristics. Survival worsens considerably with organ failure, especially with sepsis. They represent a small number of admissions, albeit increasing over recent years, and, in general, have a short ICU stay. Patients with single organ failure have acceptable survival rates and mortality has improved; although we have no data on those refused ICU admission potentially causing survival bias. Given the extremely high mortality in patients with multi-organ failure, support should be limited/withdrawn in such patients.

Case mix, outcome, and activity for admissions to UK critical care units with severe acute pancreatitis: a secondary analysis of the ICNARC Case Mix Programme Database.

INTRODUCTION: Severe acute pancreatitis (SAP) requiring admission to a critical care unit is associated with high mortality and long lengths of stay. We describe the case mix, outcome, and activity of admissions with SAP who were identified from a high-quality clinical database. METHODS: We conducted a secondary analysis of the ICNARC (Intensive Care National Audit & Research Centre) Case Mix Programme Database of 219,468 admissions to 159 adult, general critical care units in England, Wales, and Northern Ireland for the period of 1995 to 2003 to identify admissions with SAP. The ability of the modified Glasgow criteria to discriminate hospital survivors from non-survivors was compared to that of the Acute Physiology and Chronic Health Evaluation (APACHE) II score and a number of individual physiological parameters. RESULTS: A total of 2,677 admissions with SAP were identified (1.2% of all admissions). Mortality for these admissions was 31% in the critical care unit and 42% in hospital. The median length of stay in the critical care unit was 3.8 days and was similar for survivors and non-survivors. Increasing numbers of modified Glasgow criteria were associated with increasing hospital mortality, but better discrimination was provided by the APACHE II score and by several physiological parameters. CONCLUSION: SAP requiring critical care is associated with high mortality and long length of stay. The modified Glasgow criteria represent a simple measure of severity but were not designed to predict hospital mortality. It may be possible to develop a specific model for risk prediction in patients with SAP requiring critical care.

Assessment of the clinical effectiveness of pulmonary artery catheters in management of patients in intensive care (PAC-Man): a randomised controlled trial.

BACKGROUND: Over the past 30 years the pulmonary artery catheter (PAC) has become a widely used haemodynamic monitoring device in the management of critically ill patients, though doubts exist about its safety. Our aim was, therefore, to ascertain whether hospital mortality is reduced in critically ill patients when they are managed with a PAC. METHODS: We did a randomised controlled trial to which we enrolled 1041 patients from 65 UK intensive care units. We assigned individuals to management with (n=519) or without (n=522) a PAC. The timing of insertion and subsequent clinical management were at the discretion of the treating clinician. Intensive care units decided a priori to have the option of using an alternative cardiac output-monitoring device in control patients. FINDINGS: 1014 patients were eligible for analysis. We noted no difference in hospital mortality between patients managed with or without a PAC (68% [346 of 506] vs 66% [333 of 507], p=0.39; adjusted hazard ratio 1.09, 95% CI 0.94-1.27). We noted complications associated with insertion of a PAC in 46 of 486 individuals in whom the device was placed, none of which was fatal. INTERPRETATION: Our findings indicate no clear evidence of benefit or harm by managing critically ill patients with a PAC. Efficacy studies are needed to ascertain whether management protocols involving PAC use can result in improved outcomes in specific groups if these devices are not to become a redundant technology.