ICU staffing feature phenotypes and their relationship with patients' outcomes: an unsupervised machine learning analysis

Comparing causal random forest and linear regression to estimate the independent association of organisational factors with ICU efficiency

PURPOSE

Parametric regression models have been the main statistical method for identifying average treatment effects. Causal machine learning models showed promising results in estimating heterogeneous treatment effects in causal inference. Here we aimed to compare the application of causal random forest (CRF) and linear regression modelling (LRM) to estimate the effects of organisational factors on ICU efficiency.

METHODS

A retrospective analysis of 277,459 patients admitted to 128 Brazilian and Uruguayan ICUs over three years. ICU efficiency was assessed using the average standardised efficiency ratio (ASER), measured as the average of the standardised mortality ratio (SMR) and the standardised resource use (SRU) according to the SAPS-3 score. Using a causal inference framework, we estimated and compared the conditional average treatment effect (CATE) of seven common structural and organisational factors on ICU efficiency using LRM with interaction terms and CRF.

RESULTS

The hospital mortality was 14 %; median ICU and hospital lengths of stay were 2 and 7 days, respectively. Overall median SMR was 0.97 [IQR: 0.76,1.21], median SRU was 1.06 [IQR: 0.79,1.30] and median ASER was 0.99 [IQR: 0.82,1.21]. Both CRF and LRM showed that the average number of nurses per ten beds was independently associated with ICU efficiency (CATE [95 %CI]: -0.13 [-0.24, -0.01] and -0.09 [-0.17,-0.01], respectively). Finally, CRF identified some specific ICUs with a significant CATE in exposures that did not present a significant average effect.

CONCLUSION

In general, both methods were comparable to identify organisational factors significantly associated with CATE on ICU efficiency. CRF however identified specific ICUs with significant effects, even when the average effect was nonsignificant. This can assist healthcare managers in further in-dept evaluation of process interventions to improve ICU efficiency.

Update on the Epimed Monitor Adult ICU Database: 15 years of its use in national registries, quality improvement initiatives and clinical research

In recent decades, several databases of critically ill patients have become available in both low-, middle-, and high-income countries from all continents. These databases are also rich sources of data for the surveillance of emerging diseases, intensive care unit performance evaluation and benchmarking, quality improvement projects and clinical research. The Epimed Monitor database is turning 15 years old in 2024 and has become one of the largest of these databases. In recent years, there has been rapid geographical expansion, an increase in the number of participating intensive care units and hospitals, and the addition of several new variables and scores, allowing a more complete characterization of patients to facilitate multicenter clinical studies. As of December 2023, the database was being used regularly for 23,852 beds in 1,723 intensive care units and 763 hospitals from ten countries, totaling more than 5.6 million admissions. In addition, critical care societies have adopted the system and its database to establish national registries and international collaborations. In the present review, we provide an updated description of the database; report experiences of its use in critical care for quality improvement initiatives, national registries and clinical research; and explore other potential future perspectives and developments.

Factors that influence critical care nurses' management of sedation for ventilated patients in critical care: A qualitative study

BACKGROUND

Optimising sedation use is key to timely extubation. Whilst sedation protocols may be used to guide critical care nurses' management of sedation, sedation management and decision-making is complex, influenced by multiple factors related to patients' circumstances, intensive care unit design and the workforce.

AIM

To explore (i) critical care nurses' experiences managing sedation in mechanically ventilated patients and (ii) the factors that influence their sedation-related decision-making.

DESIGN

Qualitative descriptive study using semi-structured interviews. Data were analysed using Braun and Clarke's six-step thematic analysis.

SETTING AND PARTICIPANTS

This study was conducted in a 26-bed level 3 accredited ICU, in a private hospital in Melbourne, Australia. The majority of patients are admitted following elective surgery. Critical care nurses, who were permanently employed as a registered nurse, worked at least 16 h per week, and cared for ventilated patients, were invited to participate.

FINDINGS

Thirteen critical care nurses participated. Initially, participants suggested their experiences managing sedation were linked to local unit policy and learning. Further exploration revealed that experiences were synonymous with descriptors of factors influencing sedation decision-making according to three themes: (i) Learning from past experiences, (ii) Situational awareness and (iii) Prioritising safety. Nurses relied on their cumulative knowledge from prior experiences to guide decision-making. Situational awareness about other emergent priorities in the unit, staffing and skill-mix were important factors in guiding sedation decision-making. Safety of patients and staff was essential, at times overriding goals to reduce sedation.

CONCLUSION

Sedation decision making cannot be considered in isolation. Rather, sedation decision making must take into account outcomes of patient assessment, emergent priorities, unit and staffing factors and safety concerns.

IMPLICATIONS FOR CLINICAL PRACTICE

Opportunities for ongoing education are essential to promote nurses' situational awareness of other emergent unit priorities, staffing and skill-mix, in addition to evidence-based sedation management and decision making.

Interprofessional Staffing Pattern Clusters in U.S. ICUs

OBJECTIVES

To identify interprofessional staffing pattern clusters used in U.S. ICUs.

DESIGN

Latent class analysis.

SETTING AND PARTICIPANTS

Adult U.S. ICUs.

PATIENTS

None.

INTERVENTIONS

None.

ANALYSIS

We used data from a staffing survey that queried respondents (n = 596 ICUs) on provider (intensivist and nonintensivist), nursing, respiratory therapist, and clinical pharmacist availability and roles. We used latent class analysis to identify clusters describing interprofessional staffing patterns and then compared ICU and hospital characteristics across clusters.

MEASUREMENTS AND MAIN RESULTS

We identified three clusters as optimal. Most ICUs (54.2%) were in cluster 1 ("higher overall staffing") characterized by a higher likelihood of good provider coverage (both intensivist [onsite 24 hr/d] and nonintensivist [orders placed by ICU team exclusively, presence of advanced practice providers, and physicians-in-training]), nursing leadership (presence of charge nurse, nurse educators, and managers), and bedside nursing support (nurses with registered nursing degrees, fewer patients per nurse, and nursing aide availability). One-third (33.7%) were in cluster 2 ("lower intensivist coverage & nursing leadership, higher bedside nursing support") and 12.1% were in cluster 3 ("higher provider coverage & nursing leadership, lower bedside nursing support"). Clinical pharmacists were more common in cluster 1 (99.4%), but present in greater than 85% of all ICUs; respiratory therapists were nearly universal. Cluster 1 ICUs were larger (median 20 beds vs. 15 and 17 in clusters 2 and 3, respectively; p < 0.001), and in larger (> 250 beds: 80.6% vs. 66.1% and 48.5%; p < 0.001), not-for-profit (75.9% vs. 69.4% and 60.3%; p < 0.001) hospitals. Telemedicine use 24 hr/d was more common in cluster 3 units (71.8% vs. 11.7% and 14.1%; p < 0.001).

CONCLUSIONS

More than half of U.S. ICUs had higher staffing overall. Others tended to have either higher provider presence and nursing leadership or higher bedside nursing support, but not both.

Associations of the staffing structure of intensive care units and high care units on in-hospital mortality among patients with sepsis: a cross-sectional study of Japanese nationwide claims data

OBJECTIVE

The objective was to analyse the associations of intensive care unit (ICU) and high care unit (HCU) organisational structure on in-hospital mortality among patients with sepsis in Japan's acute care hospitals.

DESIGN

Multicentre cross-sectional study.

SETTINGS

Patients with sepsis aged ≥18 years who received critical care in acute care hospitals throughout Japan between April 2018 and March 2019 were identified using the National Database of Health Insurance Claims and Specific Health Checkups of Japan (NDB).

INTERVENTIONS

None.

PARTICIPANTS

10 968 patients with sepsis were identified. ICUs were categorised into three groups: type 1 ICUs (fulfilling stringent staffing criteria such as experienced intensivists and high nurse-to-patient ratios), type 2 ICUs (less stringent criteria) and HCUs (least stringent criteria).

PRIMARY OUTCOME MEASURE

The study's primary outcome measure was in-hospital mortality. Cox proportional hazards regression analysis was performed to examine the impact of ICU/HCU groups on in-hospital mortality.

RESULTS

We analysed 2411 patients (178 hospitals) in the type 1 ICU group, 3653 patients (422 hospitals) in the type 2 ICU group and 4904 patients (521 hospitals) in the HCU group. When compared with the type 1 ICU group, the adjusted HRs for in-hospital mortality were 1.12 (95% CI 1.04 to 1.21) for the type 2 ICU group and 1.17 (95% CI 1.08 to 1.26) for the HCU group.

CONCLUSION

ICUs that fulfil more stringent staffing criteria were associated with lower in-hospital mortality among patients with sepsis than HCUs. Differences in organisational structure may have an association with outcomes in patients with sepsis, and this was observed by the NDB.

Unmasking the hidden aftermath: postintensive care unit sequelae, discharge preparedness, and long-term follow-up

A significant portion of individuals who have experienced critical illness encounter new or exacerbated impairments in their physical, cognitive, or mental health, commonly referred to as postintensive care syndrome. Moreover, those who survive critical illness often face an increased risk of adverse consequences, including infections, major cardiovascular events, readmissions, and elevated mortality rates, during the months following hospitalization. These findings emphasize the critical necessity for effective prevention and management of long-term health deterioration in the critical care environment. Although conclusive evidence from well-designed randomized clinical trials is somewhat limited, potential interventions include strategies such as limiting sedation, early mobilization, maintaining family presence during the intensive care unit stay, implementing multicomponent transition programs (from intensive care unit to ward and from hospital to home), and offering specialized posthospital discharge follow-up. This review seeks to provide a concise summary of recent medical literature concerning long-term outcomes following critical illness and highlight potential approaches for preventing and addressing health decline in critical care survivors.

The Association Between Prepandemic ICU Performance and Mortality Variation in COVID-19: A Multicenter Cohort Study of 35,619 Critically Ill Patients

BACKGROUND

During the COVID-19 pandemic, ICUs remained under stress and observed elevated mortality rates and high variations of outcomes. A knowledge gap exists regarding whether an ICU performing best during nonpandemic times would still perform better when under high pressure compared with the least performing ICUs.

RESEARCH QUESTION

Does prepandemic ICU performance explain the risk-adjusted mortality variability for critically ill patients with COVID-19?

STUDY DESIGN AND METHODS

This study examined a cohort of adults with real-time polymerase chain reaction-confirmed COVID-19 admitted to 156 ICUs in 35 hospitals from February 16, 2020, through December 31, 2021, in Brazil. We evaluated crude and adjusted in-hospital mortality variability of patients with COVID-19 in the ICU during the pandemic. Association of baseline (prepandemic) ICU performance and in-hospital mortality was examined using a variable life-adjusted display (VLAD) during the pandemic and a multivariable mixed regression model adjusted by clinical characteristics, interaction of performance with the year of admission, and mechanical ventilation at admission.

RESULTS

Thirty-five thousand six hundred nineteen patients with confirmed COVID-19 were evaluated. The median age was 52 years, median Simplified Acute Physiology Score 3 was 42, and 18% underwent invasive mechanical ventilation. In-hospital mortality was 13% and 54% for those receiving invasive mechanical ventilation. Adjusted in-hospital mortality ranged from 3.6% to 63.2%. VLAD in the most efficient ICUs was higher than the overall median in 18% of weeks, whereas VLAD was 62% and 84% in the underachieving and least efficient groups, respectively. The least efficient baseline ICU performance group was associated independently with increased mortality (OR, 2.30; 95% CI, 1.45-3.62) after adjusting for patient characteristics, disease severity, and pandemic surge.

INTERPRETATION

ICUs caring for patients with COVID-19 presented substantial variation in risk-adjusted mortality. ICUs with better baseline (prepandemic) performance showed reduced mortality and less variability. Our findings suggest that achieving ICU efficiency by targeting improvement in organizational aspects of ICUs may impact outcomes, and therefore should be a part of the preparedness for future pandemics.

ICU registries: From tracking to fostering better outcomes

NA.

National ICU Registries as Enablers of Clinical Research and Quality Improvement

OBJECTIVES

Clinical quality registries (CQRs) have been implemented worldwide by several medical specialties aiming to generate a better characterization of epidemiology, treatments, and outcomes of patients. National ICU registries were created almost 3 decades ago to improve the understanding of case-mix, resource use, and outcomes of critically ill patients. This narrative review describes the challenges, proposed solutions, and evidence generated by National ICU registries as facilitators for research and quality improvement.

DATA SOURCES

English language articles were identified in PubMed using phrases related to ICU registries, CQRs, outcomes, and case-mix.

STUDY SELECTION

Original research, review articles, letters, and commentaries, were considered.

DATA EXTRACTION

Data from relevant literature were identified, reviewed, and integrated into a concise narrative review.

DATA SYNTHESIS

CQRs have been implemented worldwide by several medical specialties aiming to generate a better characterization of epidemiology, treatments, and outcomes of patients. National ICU registries were created almost 3 decades ago to improve the understanding of case-mix, resource use, and outcomes of critically ill patients. The initial experience in European countries and in Oceania ensured that through locally generated data, ICUs could assess their performances by using risk-adjusted measures and compare their results through fair and validated benchmarking metrics with other ICUs contributing to the CQR. The accomplishment of these initiatives, coupled with the increasing adoption of information technology, resulted in a broad geographic expansion of CQRs as well as their use in quality improvement studies, clinical trials as well as international comparisons, and benchmarking for ICUs.

CONCLUSIONS

ICU registries have provided increased knowledge of case-mix and outcomes of ICU patients based on real-world data and contributed to improve care delivery through quality improvement initiatives and trials. Recent increases in adoption of new technologies (i.e., cloud-based structures, artificial intelligence, machine learning) will ensure a broader and better use of data for epidemiology, healthcare policies, quality improvement, and clinical trials.

Measuring the Impact of ICU Strain on Mortality, After-Hours Discharge, Discharge Delay, Interhospital Transfer, and Readmission in Australia With the Activity Index

OBJECTIVES

ICU resource strain leads to adverse patient outcomes. Simple, well-validated measures of ICU strain are lacking. Our objective was to assess whether the "Activity index," an indicator developed during the COVID-19 pandemic, was a valid measure of ICU strain.

DESIGN

Retrospective national registry-based cohort study.

SETTING

One hundred seventy-five public and private hospitals in Australia (June 2020 through March 2022).

SUBJECTS

Two hundred seventy-seven thousand seven hundred thirty-seven adult ICU patients.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Data from the Australian and New Zealand Intensive Care Society Adult Patient Database were matched to the Critical Health Resources Information System. The mean daily Activity index of each ICU (census total of "patients with 1:1 nursing" + "invasive ventilation" + "renal replacement" + "extracorporeal membrane oxygenation" + "active COVID-19," divided by total staffed ICU beds) during the patient's stay in the ICU was calculated. Patients were categorized as being in the ICU during very quiet (Activity index < 0.1), quiet (0.1 to < 0.6), intermediate (0.6 to < 1.1), busy (1.1 to < 1.6), or very busy time-periods (≥ 1.6). The primary outcome was in-hospital mortality. Secondary outcomes included after-hours discharge from the ICU, readmission to the ICU, interhospital transfer to another ICU, and delay in discharge from the ICU. Median Activity index was 0.87 (interquartile range, 0.40-1.24). Nineteen thousand one hundred seventy-seven patients died (6.9%). In-hospital mortality ranged from 2.4% during very quiet to 10.9% during very busy time-periods. After adjusting for confounders, being in an ICU during time-periods with higher Activity indices, was associated with an increased risk of in-hospital mortality (odds ratio [OR], 1.49; 99% CI, 1.38-1.60), after-hours discharge (OR, 1.27; 99% CI, 1.21-1.34), readmission (OR, 1.18; 99% CI, 1.09-1.28), interhospital transfer (OR, 1.92; 99% CI, 1.72-2.15), and less delay in ICU discharge (OR, 0.58; 99% CI, 0.55-0.62): findings consistent with ICU strain.

CONCLUSIONS

The Activity index is a simple and valid measure that identifies ICUs in which increasing strain leads to progressively worse patient outcomes.

Descriptive Comparison of Two Models of Tele-Critical Care Delivery in a Large Multi-Hospital Health Care System

Introduction: The Society of Critical Care Medicine Tele-Critical Care (TCC) Committee has identified the need for rigorous comparative research of different TCC delivery models to support the development of best practices for staffing, application, and approaches to workflow. Our objective was to describe and compare outcomes between two TCC delivery models, TCC with 24/7 Bedside Intensivist (BI) compared with TCC with Private Daytime Attending Intensivist (PI) in relation to intensive care unit (ICU) and hospital mortality, ICU and hospital length of stay (LOS), cost, and complications across the spectrum of routine ICU standards of care. Methods: Observational cohort study at large health care system in 12 ICUs and included patients, ≥18, with Acute Physiology and Chronic Health Evaluation (APACHE) IVa scores and predictions (October 2016-June 2019). Results: Of the 19,519 ICU patients, 71.7% (n = 13,993) received TCC with 24/7 BI while 28.3% (n = 5,526) received TCC with PI. ICU and Hospital mortality (4.8% vs. 3.1%, p < 0.0001; 12.6% vs. 8.1%, p < 0.001); and ICU and Hospital LOS (3.2 vs. 2.4 days, p < 0.001; 9.8 vs. 7.2 days, p < 0.001) were significantly higher among 24/7 BI compared with PI. The APACHE observed/expected ratios (odds ratio [OR]; 95% confidence interval [CI]) for ICU mortality (0.62; 0.58-0.67) vs. (0.53; 0.46-0.61) and Hospital mortality (0.95; 0.57-1.48) vs. (0.77; 0.70-0.84) were significantly different for 24/7 BI compared with PI. Multivariate mixed models that adjusted for confounders demonstrated significantly greater odds of (OR; 95% CI) ICU mortality (1.58; 1.28-1.93), Hospital mortality (1.52; 1.33-1.73), complications (1.55; 1.18-2.04), ICU LOS [3.14 vs. 2.59 (1.25; 1.19-1.51)], and Hospital LOS [9.05 vs. 7.31 (1.23; 1.21-1.25)] among 24/7 BI when compared with PI. Sensitivity analyses adjusting for ICU admission within 24 h of hospital admission, receiving active ICU treatments, nighttime admission, sepsis, and highest third acute physiology score indicated significantly higher odds for 24/7 BI compared with PI. Conclusion: Our comparison demonstrated that TCC delivery model with PI provided high-quality care with significant positive effects on outcomes. This suggests that TCC delivery models have broad-ranging applicability and benefits in routine critical care, thus necessitating progressive research in this direction.

Interpretable machine-learning model for Predicting the Convalescent COVID-19 patients with pulmonary diffusing capacity impairment

INTRODUCTION

The COVID-19 patients in the convalescent stage noticeably have pulmonary diffusing capacity impairment (PDCI). The pulmonary diffusing capacity is a frequently-used indicator of the COVID-19 survivors' prognosis of pulmonary function, but the current studies focusing on prediction of the pulmonary diffusing capacity of these people are limited. The aim of this study was to develop and validate a machine learning (ML) model for predicting PDCI in the COVID-19 patients using routinely available clinical data, thus assisting the clinical diagnosis.

METHODS

Collected from a follow-up study from August to September 2021 of 221 hospitalized survivors of COVID-19 18 months after discharge from Wuhan, including the demographic characteristics and clinical examination, the data in this study were randomly separated into a training (80%) data set and a validation (20%) data set. Six popular machine learning models were developed to predict the pulmonary diffusing capacity of patients infected with COVID-19 in the recovery stage. The performance indicators of the model included area under the curve (AUC), Accuracy, Recall, Precision, Positive Predictive Value(PPV), Negative Predictive Value (NPV) and F1. The model with the optimum performance was defined as the optimal model, which was further employed in the interpretability analysis. The MAHAKIL method was utilized to balance the data and optimize the balance of sample distribution, while the RFECV method for feature selection was utilized to select combined features more favorable to machine learning.

RESULTS

A total of 221 COVID-19 survivors were recruited in this study after discharge from hospitals in Wuhan. Of these participants, 117 (52.94%) were female, with a median age of 58.2 years (standard deviation (SD) = 12). After feature selection, 31 of the 37 clinical factors were finally selected for use in constructing the model. Among the six tested ML models, the best performance was accomplished in the XGBoost model, with an AUC of 0.755 and an accuracy of 78.01% after experimental verification. The SHAPELY Additive explanations (SHAP) summary analysis exhibited that hemoglobin (Hb), maximal voluntary ventilation (MVV), severity of illness, platelet (PLT), Uric Acid (UA) and blood urea nitrogen (BUN) were the top six most important factors affecting the XGBoost model decision-making.

CONCLUSION

The XGBoost model reported here showed a good prognostic prediction ability for PDCI of COVID-19 survivors during the recovery period. Among the interpretation methods based on the importance of SHAP values, Hb and MVV contributed the most to the prediction of PDCI outcomes of COVID-19 survivors in the recovery period.

Organisation, staffing and resources of critical care units in Kenya

OBJECTIVE

To describe the organisation, staffing patterns and resources available in critical care units in Kenya. The secondary objective was to explore variations between units in the public and private sectors.

MATERIALS AND METHODS

An online cross-sectional survey was used to collect data on organisational characteristics (model of care, type of unit, quality- related activities, use of electronic medical records and participation in the national ICU registry), staffing and available resources for monitoring, ventilation and general critical care.

RESULTS

The survey included 60 of 75 identified units (80% response rate), with 43% (n = 23) located in government facilities. A total of 598 critical care beds were reported with a median of 6 beds (interquartile range [IQR] 5-11) per unit, with 26% beds (n = 157) being non functional. The proportion of ICU beds to total hospital beds was 3.8% (IQR 1.9-10.4). Most of the units (80%, n = 48) were mixed/general units with an open model of care (60%, n = 36). Consultants-in-charge were mainly anesthesiologists (69%, n = 37). The nurse-to-bed ratio was predominantly 1:2 with half of the nurses formally trained in critical care. Most units (83%, n = 47) had a dedicated ventilator for each bed, however 63% (n = 39) lacked high flow nasal therapy. While basic multiparametric monitoring was ubiquitous, invasive blood pressure measurement capacity was low (3% of beds, IQR 0-81%), and capnography moderate (31% of beds, IQR 0-77%). Blood gas analysers were widely available (93%, n = 56), with 80% reported as functional. Differences between the public and private sector were narrow.

CONCLUSION

This study shows an established critical care network in Kenya, in terms of staffing density, availability of basic monitoring and ventilation resources. The public and private sector are equally represented albeit with modest differences. Potential areas for improvement include training, use of invasive blood pressure and functionality of blood gas analysers.

Clinical characteristics and outcomes of patients with COVID-19 admitted to the intensive care unit during the first and second waves of the pandemic in Brazil: a single-center retrospective cohort study

OBJECTIVE

To describe and compare the clinical characteristics and outcomes of patients admitted to intensive care units during the first and second waves of the COVID-19 pandemic.

METHODS

In this retrospective single-center cohort study, data were retrieved from the Epimed Monitor System; all adult patients admitted to the intensive care unit between March 4, 2020, and October 1, 2021, were included in the study. We compared the clinical characteristics and outcomes of patients admitted to the intensive care unit of a quaternary private hospital in São Paulo, Brazil, during the first (May 1, 2020, to August 31, 2020) and second (March 1, 2021, to June 30, 2021) waves of the COVID-19 pandemic.

RESULTS

In total, 1,427 patients with COVID-19 were admitted to the intensive care unit during the first (421 patients) and second (1,006 patients) waves. Compared with the first wave group [median (IQR)], the second wave group was younger [57 (46-70) versus 67 (52-80) years; p<0.001], had a lower SAPS 3 Score [45 (42-52) versus 49 (43-57); p<0.001], lower SOFA Score on intensive care unit admission [3 (1-6) versus 4 (2-6); p=0.018], lower Charlson Comorbidity Index [0 (0-1) versus 1 (0-2); p<0.001], and were less frequently frail (10.4% versus 18.1%; p<0.001). The second wave group used more noninvasive ventilation (81.3% versus 53.4%; p<0.001) and high-flow nasal cannula (63.2% versus 23.0%; p<0.001) during their intensive care unit stay. The intensive care unit (11.3% versus 10.5%; p=0.696) and in-hospital mortality (12.3% versus 12.1%; p=0.998) rates did not differ between both waves.

CONCLUSION

In the first and second waves, patients with severe COVID-19 exhibited similar mortality rates and need for invasive organ support, despite the second wave group being younger and less severely ill at the time of intensive care unit admission.

Clustering using unsupervised machine learning to stratify the risk of immune-related liver injury

BACKGROUND AND AIM

Immune-related liver injury (liver-irAE) is a clinical problem with a potentially poor prognosis.

METHODS

We retrospectively collected clinical data from patients treated with immune checkpoint inhibitors between September 2014 and December 2021 at the Nagoya University Hospital. Using an unsupervised machine learning method, the Gaussian mixture model, to divide the cohort into clusters based on inflammatory markers, we investigated the cumulative incidence of liver-irAEs in these clusters.

RESULTS

This study included a total of 702 patients. Among them, 492 (70.1%) patients were male, and the mean age was 66.6 years. During the mean follow-up period of 423 days, severe liver-irAEs (Common Terminology Criteria for Adverse Events grade ≥ 3) occurred in 43 patients. Patients were divided into five clusters (a, b, c, d, and e). The cumulative incidence of liver-irAE was higher in cluster c than in cluster a (hazard ratio [HR]: 13.59, 95% confidence interval [CI]: 1.70-108.76, P = 0.014), and overall survival was worse in clusters c and d than in cluster a (HR: 2.83, 95% CI: 1.77-4.50, P < 0.001; HR: 2.87, 95% CI: 1.47-5.60, P = 0.002, respectively). Clusters c and d were characterized by high temperature, C-reactive protein, platelets, and low albumin. However, there were differences in the prevalence of neutrophil count, neutrophil-to-lymphocyte ratio, and liver metastases between both clusters.

CONCLUSIONS

The combined assessment of multiple markers and body temperature may help stratify high-risk groups for developing liver-irAE.

The Intersection Between Heart Failure and Critical Care Cardiology: An International Perspective on Structure, Staffing, and Design Considerations

The overall patient population in contemporary cardiac intensive care units (CICUs) has only increased with respect to patient acuity, complexity, and illness severity. The current population has more cardiac and noncardiac comorbidities, a higher prevalence of multiorgan injury, and consumes more critical care resources than previously. Patients with heart failure (HF) now occupy a large portion of contemporary tertiary or quaternary care CICU beds around the world. In this review, we discuss the core issues that relate to the care of critically ill patients with HF, including global perspectives on the organization, designation, and collaboration of CICUs regionally and across institutions, as well as unique models for provisioning care for patients with HF within a health care setting. The latter includes a discussion of traditional and emerging models, specialized HF units, the makeup and implementation of multidisciplinary team-based decision-making, and cardiac critical care admission and triage practices. This article illustrates the ways in which critically ill patients with HF have helped to shape contemporary CICUs throughout the world and explores how these very patients will similarly help to inform the future maturation of these specialized critical care units. Finally, we will critically examine broad, contemporary, international models of HF and cardiac critical care delivery in North America, Europe, South America, and Asia, and conclude with opportunities for the further investigation and generation of evidence for care delivery.

End of life decisions in immunocompromised patients with acute respiratory failure

PURPOSE

To identify patient, disease and organizational factors associated with decisions to forgo life-sustaining therapies (DFLSTs) in critically ill immunocompromised patients admitted to the intensive care unit (ICU) for acute respiratory failure.

MATERIAL AND METHODS

We performed a secondary analysis of the international EFRAIM prospective study, which enrolled 1611 immunocompromised patients with acute respiratory failure admitted to 68 ICUs in 16 countries between October 2015 and June 2016. Multivariate logistic analysis was performed to identify independent predictors of DFLSTs.

RESULTS

The main causes of immunosuppression were hematological malignancies (50%) and solid tumor (38%). Patients had a median age of 63 yo (54-71). A pulmonologist was involved in the patient management in 38% of cases. DFLSTs had been implemented in 28% of the patients. The following variables were independently associated with DFLSTs: 1) patient-related: older age (OR 1.02 per one year increase, 95% confidence interval(CI) 1.01-1.03,P < 0.001), poor performance status (OR 2.79, 95% CI 1.98-3.93, P < 0.001); 2) disease-related: shock (OR 2.00, 95% CI 1.45-2.75, P < 0.001), liver failure (OR 1.59, 95% CI 1.14-2.21, P = 0.006), invasive mechanical ventilation (OR 1.79, 95% CI 1.31-2.46, P < 0.001); 3) organizational: having a pulmonologist involved in patient management (OR 1.85, 95% CI 1.36-2.52, P < 0.001), and the presence of a critical care outreach services (OR 1.63, 95% CI 1.11-2.38, P = 0.012).

CONCLUSIONS

A DFLST is made in one in four immunocompromised patient admitted to the ICU for acute respiratory failure. Involving a pulmonologist in patient's management is associated with less non beneficial care.

Association between nursing cost and patient outcomes in intensive care units: A retrospective cohort study of Belgian hospitals

INTRODUCTION

Hospitals with better nursing resources report more favourable patient outcomes with almost no difference in cost as compared to those with worse nursing resources. The aim of this study was to assess the association between nursing cost per intensive care unit bed and patient outcomes (mortality, readmission, and length of stay).

METHODOLOGY

This was a retrospective cohort study using data collected from the intensive care units of 17 Belgian hospitals from January 01 to December 31, 2018. Hospitals were dichotomized using median annual nursing cost per bed. A total of 18,235 intensive care unit stays were included in the study with 5,664 stays in the low-cost nursing group and 12,571 in the high-cost nursing group.

RESULTS

The rate of high length of stay outliers in the intensive care unit was significantly lower in the high-cost nursing group (9.2% vs 14.4%) compared to the low-cost nursing group. Intensive care unit readmission was not significantly different in the two groups. Mortality was lower in the high-cost nursing group for intensive care unit (9.9% vs 11.3%) and hospital (13.1% vs 14.6%) mortality. The nursing cost per intensive care bed was different in the two groups, with a median [IQR] cost of 159,387€ [140,307-166,690] for the low-cost nursing group and 214,032€ [198,094-230,058] for the high-cost group. In multivariate analysis, intensive care unit mortality (OR = 0.80, 95% CI: 0.69-0.92, p < 0.0001), in-hospital mortality (OR = 0.82, 95% CI: 0.72-0.93, p < 0.0001), and high length of stay outliers (OR = 0.48, 95% CI: 0.42-0.55, p < 0.0001) were lower in the high-cost nursing group. However, there was no significant effect on intensive care readmission between the two groups (OR = 1.24, 95% CI: 0.97-1.51, p > 0.05).

CONCLUSIONS

This study found that higher-cost nursing per bed was associated with significantly lower intensive care unit and in-hospital mortality rates, as well as fewer high length of stay outliers, but had no significant effect on readmission to the intensive care unit. .

Clustering phosphate and iron-related markers and prognosis in dialysis patients

BACKGROUND

Hyperphosphatemia in patients undergoing dialysis is common and is associated with mortality. Recently, the link between phosphate metabolism and iron dynamics has received increasing attention. However, the association between this relationship and prognosis remains largely unexplored.

METHODS

We conducted an observational study of patients who initiated dialysis in the 17 centers participating in the Aichi Cohort Study of the Prognosis in Patients Newly Initiated into Dialysis. Data were available on sex, age, use of phosphate binder, drug history, medical history and laboratory data. After excluding patients with missing values of phosphate, hemoglobin, ferritin and transferrin saturation, we used the Gaussian mixture model to divide the cohort into clusters based on phosphate, hemoglobin, logarithmic ferritin and transferrin saturation. We investigated the prognosis of patients in these clusters. The primary outcome was all-cause death. In each cluster, the prognostic impact of phosphate binder was also studied.

RESULTS

The study included 1175 patients with chronic kidney disease who initiated dialysis between October 2011 and September 2013. Among them, 785 were men and 390 were women, with a mean ± SD age of 67.9 ± 13.0 years. The patients were divided into three clusters, and mortality was higher in cluster c than in cluster a (P = 0.005). Moreover, the use of phosphate binders was associated with a lower risk of all-cause death in two clusters (a and c) that were characterized by older age and higher prevalence of diabetes mellitus, among other things.

CONCLUSIONS

We used an unsupervised machine learning method to cluster patients, using phosphate, hemoglobin and iron-related markers. In two of the clusters, the oral use of a phosphate binder might improve prognosis.

Eventos adversos em pacientes de terapia intensiva: estudo transversal

RESUMO Objetivo: identificar a prevalência de eventos adversos e a necessidade de cuidado do paciente crítico em uma unidade de terapia intensiva. Método: estudo transversal, realizado de janeiro a março de 2020. Os eventos adversos investigados foram: lesão por pressão, extubação orotraqueal acidental, queda, perda de acesso venoso central e infecção relacionada à assistência à saúde. O número de horas necessárias para o cuidado do paciente foi mensurado pela Nursing Activities Score. As variáveis independentes categóricas foram descritas por frequências absoluta e relativa, e as contínuas, por tendência central. A medida de magnitude foi a razão de chance e considerou-se intervalo de confiança de 95%. Resultados: dos 88 pacientes avaliados, 52,3% apresentaram eventos adversos, os quais foram associados à maior necessidade de cuidados, gravidade e ao maior tempo de internação. O Nursing Activities Score médio foi 51,01% (12 h 24 min), sendo identificado um déficit de 20% a 30% de pessoal de enfermagem na unidade. Conclusão: a prevalência dos eventos adversos na unidade é alta e o déficit de pessoal de enfermagem na unidade revelou a necessidade de dimensionamento adequado de pessoal para reduzir os danos ocasionados pelos cuidados prestados aos pacientes críticos.

Adverse events in critically ill patients: a cross-sectional study

Abstract Objective: To identify the prevalence of adverse events and the critically ill patient’s need for care in an intensive care unit. Method: This is a cross-sectional study, carried out from January to March 2020. The adverse events investigated were pressure injury, accidental orotracheal extubation, fall, loss of central venous access, and healthcare-associated infection. The number of hours required for patient care was measured by the Nursing Activities Score. The categorical independent variables were described by absolute and relative frequencies, and the continuous ones, by central tendency. The magnitude measure was the odds ratio and a confidence interval of 95% was considered. Results: of the 88 patients evaluated, 52.3% had adverse events, which were associated with a greater need for care, severity, and longer hospital stay. The mean Nursing Activities Score was 51.01% (12 h 24 min), with a deficit of 20% to 30% of nursing staff in the unit being identified. Conclusion: The prevalence of adverse events in the unit is high and the shortage of nursing staff in the unit revealed the need for adequate staffing to reduce the damage caused by the care provided to critically ill patients.

PHarmacist Avoidance or Reductions in Medical Costs in CRITically Ill Adults: PHARM-CRIT Study

To comprehensively classify interventions performed by ICU clinical pharmacists and quantify cost avoidance generated through their accepted interventions.

DESIGN

A multicenter, prospective, observational study was performed between August 2018 and January 2019.

SETTING

Community hospitals and academic medical centers in the United States.

PARTICIPANTS

ICU clinical pharmacists.

INTERVENTIONS

Recommendations classified into one of 38 intervention categories (divided into six unique sections) associated with cost avoidance.

MEASUREMENTS AND MAIN RESULTS

Two-hundred fifteen ICU pharmacists at 85 centers performed 55,926 interventions during 3,148 shifts that were accepted on 27,681 adult patient days and generated $23,404,089 of cost avoidance. The quantity of accepted interventions and cost avoidance generated in six established sections was adverse drug event prevention (5,777 interventions; $5,822,539 CA), resource utilization (12,630 interventions; $4,491,318), individualization of patient care (29,284 interventions; $9,680,036 cost avoidance), prophylaxis (1,639 interventions; $1,414,465 cost avoidance), hands-on care (1,828 interventions; $1,339,621 cost avoidance), and administrative/supportive tasks (4,768 interventions; $656,110 cost avoidance). Mean cost avoidance was $418 per intervention, $845 per patient day, and $7,435 per ICU pharmacist shift. The annualized cost avoidance from an ICU pharmacist is $1,784,302. The potential monetary cost avoidance to pharmacist salary ratio was between $3.3:1 and $9.6:1.

CONCLUSIONS

Pharmacist involvement in the care of critically ill patients results in significant avoidance of healthcare costs, particularly in the areas of individualization of patient care, adverse drug event prevention, and resource utilization. The potential monetary cost avoidance to pharmacist salary ratio employing an ICU clinical pharmacist is between $3.3:1 and $9.6:1.

Application Scenarios for Artificial Intelligence in Nursing Care: Rapid Review

Background

Artificial intelligence (AI) holds the promise of supporting nurses’ clinical decision-making in complex care situations or conducting tasks that are remote from direct patient interaction, such as documentation processes. There has been an increase in the research and development of AI applications for nursing care, but there is a persistent lack of an extensive overview covering the evidence base for promising application scenarios.

Objective

This study synthesizes literature on application scenarios for AI in nursing care settings as well as highlights adjacent aspects in the ethical, legal, and social discourse surrounding the application of AI in nursing care.

Methods

Following a rapid review design, PubMed, CINAHL, Association for Computing Machinery Digital Library, Institute of Electrical and Electronics Engineers Xplore, Digital Bibliography & Library Project, and Association for Information Systems Library, as well as the libraries of leading AI conferences, were searched in June 2020. Publications of original quantitative and qualitative research, systematic reviews, discussion papers, and essays on the ethical, legal, and social implications published in English were included. Eligible studies were analyzed on the basis of predetermined selection criteria.

Results

The titles and abstracts of 7016 publications and 704 full texts were screened, and 292 publications were included. Hospitals were the most prominent study setting, followed by independent living at home; fewer application scenarios were identified for nursing homes or home care. Most studies used machine learning algorithms, whereas expert or hybrid systems were entailed in less than every 10th publication. The application context of focusing on image and signal processing with tracking, monitoring, or the classification of activity and health followed by care coordination and communication, as well as fall detection, was the main purpose of AI applications. Few studies have reported the effects of AI applications on clinical or organizational outcomes, lacking particularly in data gathered outside laboratory conditions. In addition to technological requirements, the reporting and inclusion of certain requirements capture more overarching topics, such as data privacy, safety, and technology acceptance. Ethical, legal, and social implications reflect the discourse on technology use in health care but have mostly not been discussed in meaningful and potentially encompassing detail.

Conclusions

The results highlight the potential for the application of AI systems in different nursing care settings. Considering the lack of findings on the effectiveness and application of AI systems in real-world scenarios, future research should reflect on a more nursing care–specific perspective toward objectives, outcomes, and benefits. We identify that, crucially, an advancement in technological-societal discourse that surrounds the ethical and legal implications of AI applications in nursing care is a necessary next step. Further, we outline the need for greater participation among all of the stakeholders involved.

Using data envelopment analysis to perform benchmarking in intensive care units

BACKGROUND

Studies using Data Envelopment Analysis to benchmark Intensive Care Units (ICUs) are scarce. Previous studies have focused on comparing efficiency using only performance metrics, without accounting for resources. Hence, we aimed to perform a benchmarking analysis of ICUs using data envelopment analysis.

METHODS

We performed a retrospective analysis on observational data of patients admitted to ICUs in Brazil (ORCHESTRA Study). The outputs in our data envelopment analysis model were the performance metrics: Standardized Mortality Ratio (SMR) and Standardized Resource Use (SRU); whereas the inputs consisted of three groups of variables that represented staffing patterns, structure, and strain, thus resulting in three models. We compared efficient and non-efficient units for each model. In addition, we compared our results to the efficiency matrix method and presented targets to each non-efficient unit.

RESULTS

We performed benchmarking in 93 ICUs and 129,680 patients. The median age was 64 years old, and mortality was 12%. Median SMR was 1.00 [interquartile range (IQR): 0.79-1.21] and SRU was 1.15 [IQR: 0.95-1.56]. Efficient units presented lower median physicians per bed ratio (1.44 [IQR: 1.18-1.88] vs. 1.7 [IQR: 1.36-2.00]) and nursing workload (168 hours [IQR: 168-291] vs 396 hours [IQR: 336-672]) but higher nurses per bed ratio (2.02 [1.16-2.48] vs. 1.71 [1.43-2.36]) compared to non-efficient units. Units from for-profit hospitals and specialized ICUs presented the best efficiency scores. Our results were mostly in line with the efficiency matrix method: the efficiency units in our models were mostly in the "most efficient" quadrant.

CONCLUSION

Data envelopment analysis provides managers the information needed to identify not only the outcomes to be achieved but what are the levels of resources needed to provide efficient care. Different perspectives can be achieved depending on the chosen variables. Its use jointly with the efficiency matrix can provide deeper understanding of ICU performance and efficiency.

The epidemiology of sepsis in paediatric intensive care units in Brazil (the Sepsis PREvalence Assessment Database in Pediatric population, SPREAD PED): an observational study

BACKGROUND

Data on the prevalence and mortality of paediatric sepsis in resource-poor settings are scarce. We aimed to assess the prevalence and in-hospital mortality of severe sepsis and septic shock treated in paediatric intensive care units (PICUs) in Brazil, and risk factors for mortality.

METHODS

We performed a nationwide, 1-day, prospective point prevalence study with follow-up of patients with severe sepsis and septic shock, using a stratified random sample of all PICUs in Brazil. Patients were enrolled at each participating PICU on a single day between March 25 and 29, 2019. All patients occupying a bed at the PICU on the study day (either admitted previously or on that day) were included if they were aged 28 days to 18 years and met the criteria for severe sepsis or septic shock at any time during hospitalisation. Patients were followed up until hospital discharge or death, censored at 60 days. Risk factors for mortality were assessed using a Poisson regression model. We used prevalence to generate national estimates.

FINDINGS

Of 241 PICUs invited to participate, 144 PICUs (capacity of 1242 beds) included patients in the study. On the day of the study, 1122 children were admitted to the participating PICUs, of whom 280 met the criteria for severe sepsis or septic shock during hospitalisation, resulting in a prevalence of 25·0% (95% CI 21·6-28·8), with a mortality rate of 19·8% (15·4-25·2; 50 of 252 patients with complete clinical data). Increased risk of mortality was associated with higher Pediatric Sequential Organ Failure Assessment score (relative risk per point increase 1·21, 95% CI 1·14-1·29, p<0·0001), unknown vaccination status (2·57, 1·26-5·24; p=0·011), incomplete vaccination status (2·16, 1·19-3·92; p=0·012), health care-associated infection (2·12, 1·23-3·64, p=0·0073), and compliance with antibiotics (2·38, 1·46-3·86, p=0·0007). The estimated incidence of PICU-treated sepsis was 74·6 cases per 100 000 paediatric population (95% CI 61·5-90·5), which translates to 42 374 cases per year (34 940-51 443) in Brazil, with an estimated mortality of 8305 (6848-10 083).

INTERPRETATION

In this representative sample of PICUs in a middle-income country, the prevalences of severe sepsis or septic shock and in-hospital mortality were high. Modifiable factors, such as incomplete vaccination and health care-associated infections, were associated with greater risk of in-hospital mortality.

FUNDING

Fundação de Amparo à Pesquisa do Estado de São Paulo and Conselho Nacional de Desenvolvimento Científico e Tecnológico.

TRANSLATION

For the Portuguese translation of the abstract see Supplementary Materials section.

Intensive care patients receiving vasoactive medications: A retrospective cohort study

BACKGROUND

Vasoactive medications are high-risk drugs commonly used in intensive care units (ICUs), which have wide variations in clinical management.

OBJECTIVES

The aim of this study was to describe the patient population, treatment, and clinical characteristics of patients who did and did not receive vasoactive medications while in the ICU and to develop a predictive tool to identify patients needing vasoactive medications.

METHODS

A retrospective cohort study of patients admitted to a level three tertiary referral ICU over a 12-month period from October 2018 to September 2019 was undertaken. Data from electronic medical records were analysed to describe patient characteristics in an adult ICU. Chi square and Mann-Whitney U tests were used to analyse data relating to patients who did and did not receive vasoactive medications. Univariate analysis and Pearson's r² were used to determine inclusion in multivariable logistic regression.

RESULTS

Of 1276 patients in the cohort, 40% (512/1276) received a vasoactive medication for haemodynamic support, with 84% (428/512) receiving noradrenaline. Older patients (odds ratio [OR] = 1.02; 95% confidence interval [CI] = 1.01-1.02; p < 0.001) with higher Acute Physiology and Chronic Health Evaluation (APACHE) III scores (OR = 1.04; 95% CI = 1.03-1.04; p < 0.001) were more likely to receive vasoactive medications than those not treated with vasoactive medications during an intensive care admission. A model developed using multivariable analysis predicted that patients admitted with sepsis (OR = 2.43; 95% CI = 1.43-4.12; p = 0.001) or shock (OR = 4.05; 95% CI = 2.68-6.10; p < 0.001) and managed on mechanical ventilation (OR = 3.76; 95% CI = 2.81-5.02; p < 0.001) were more likely to receive vasoactive medications.

CONCLUSIONS

Mechanically ventilated patients admitted to intensive care for sepsis and shock with higher APACHE III scores were more likely to receive vasoactive medications. Predictors identified in the multivariable model can be used to direct resources to patients most at risk of receiving vasoactive medications.

Competence in Decision Making: Setting Performance Standards for Critical Care

BACKGROUND

Health care professionals must be able to make frequent and timely decisions that can alter the illness trajectory of intensive care patients. A competence standard for this ability is difficult to establish yet assuring practitioners can make appropriate judgments is an important step in advancing patient safety. We hypothesized that simulation can be used effectively to assess decision-making competence. To test our hypothesis, we used a "standard-setting" method to derive cut scores (standards) for 16 simulated ICU scenarios targeted at decision-making skills and applied them to a cohort of critical care trainees.

METHODS

Panelists (critical care experts) reviewed digital audio-video performances of critical care trainees managing simulated critical care scenarios. Based on their collectively agreed-upon definition of "readiness" to make decisions in an ICU setting, each panelist made an independent judgment (ready, not ready) for a large number of recorded performances. The association between the panelists' judgments and the assessment scores was used to derive scenario-specific performance standards.

RESULTS

For all 16 scenarios, the aggregate panelists' ratings (ready/not ready for independent decision making) were positively associated with the performance scores, permitting derivation of performance standards for each scenario.

CONCLUSIONS

Minimum competence standards for high-stakes decision making can be established through standard-setting techniques. We effectively identified "front-line" providers who are, or are not, ready to make independent decisions in an ICU setting. Our approach may be used to assure stakeholders that clinicians are competent to make appropriate judgments. Further work is needed to determine whether our approach is effective in simulation-based assessments in other domains.

Adequacy of workforce - are there enough critical care doctors in the US-post COVID?

PURPOSE OF REVIEW

The ICU is a complex ecosystem in which intensive care physicians, advanced practice providers (APPs), pharmacists, and respiratory therapists work in concert to take care of critically ill patients. The SARS COV2 pandemic highlighted weaknesses in the American healthcare system. This article explores the ability of American healthcare to adapt to this challenge.

RECENT FINDINGS

With the COVID-19 pandemic, intensivists, and ventilators have been identified as the most critical components leading to shortages in ICU capacity. Anesthesiologists play a unique role in being able to provide 'flex capacity' with critical care staffing, space, and equipment (post-anesthesia care units, operating rooms, and ventilators). With the advent of APPs, intensive care physician staffing ratios may potentially be increased to cover patients safely in a physician-led team model. Tele-medicine expands this further and can allow hospital coordination for optimizing ICU bed use.

SUMMARY

Although intensivists have been able to take care of the increased ICU caseload during the COVID-19 pandemic through recruiting other specialties, the question of what is the appropriate staffing model for the future is yet to be elucidated. Creating stronger multidisciplinary care teams that have the capacity to flex up critical care capacity may be the most prudent longer-term solution.

Machine learning-based patient classification system for adult patients in intensive care units: A cross-sectional study

AIM

This study aimed to develop a patient classification system that stratifies patients admitted to the intensive care unit based on their disease severity and care needs.

BACKGROUND

Classifying patients into homogenous groups based on clinical characteristics can optimize nursing care. However, an objective method for determining such groups remains unclear.

METHODS

Predictors representing disease severity and nursing workload were considered. Patients were clustered into subgroups with different characteristics based on the results of a clustering algorithm. A patient classification system was developed using a partial least squares regression model.

RESULTS

Data of 300 patients were analysed. Cluster analysis identified three subgroups of critically patients with different levels of clinical trajectories. Except for blood potassium levels (p = .29), the subgroups were significantly different according to disease severity and nursing workload. The predicted value ranges of the regression model for Classes A, B and C were <1.44, 1.44-2.03 and >2.03. The model was shown to have good fit and satisfactory prediction efficiency using 200 permutation tests.

CONCLUSIONS

Classifying patients based on disease severity and care needs enables the development of tailored nursing management for each subgroup.

IMPLICATIONS FOR NURSING MANAGEMENT

The patient classification system can help nurse managers identify homogeneous patient groups and further improve the management of critically ill patients.

AKI Treated with Renal Replacement Therapy in Critically Ill Patients with COVID-19

BACKGROUND

AKI is a common sequela of coronavirus disease 2019 (COVID-19). However, few studies have focused on AKI treated with RRT (AKI-RRT).

METHODS

We conducted a multicenter cohort study of 3099 critically ill adults with COVID-19 admitted to intensive care units (ICUs) at 67 hospitals across the United States. We used multivariable logistic regression to identify patient-and hospital-level risk factors for AKI-RRT and to examine risk factors for 28-day mortality among such patients.

RESULTS

A total of 637 of 3099 patients (20.6%) developed AKI-RRT within 14 days of ICU admission, 350 of whom (54.9%) died within 28 days of ICU admission. Patient-level risk factors for AKI-RRT included CKD, men, non-White race, hypertension, diabetes mellitus, higher body mass index, higher d-dimer, and greater severity of hypoxemia on ICU admission. Predictors of 28-day mortality in patients with AKI-RRT were older age, severe oliguria, and admission to a hospital with fewer ICU beds or one with greater regional density of COVID-19. At the end of a median follow-up of 17 days (range, 1-123 days), 403 of the 637 patients (63.3%) with AKI-RRT had died, 216 (33.9%) were discharged, and 18 (2.8%) remained hospitalized. Of the 216 patients discharged, 73 (33.8%) remained RRT dependent at discharge, and 39 (18.1%) remained RRT dependent 60 days after ICU admission.

CONCLUSIONS

AKI-RRT is common among critically ill patients with COVID-19 and is associated with a hospital mortality rate of >60%. Among those who survive to discharge, one in three still depends on RRT at discharge, and one in six remains RRT dependent 60 days after ICU admission.

Prognostic Assessment of COVID-19 in the Intensive Care Unit by Machine Learning Methods: Model Development and Validation

BACKGROUND

Patients with COVID-19 in the intensive care unit (ICU) have a high mortality rate, and methods to assess patients' prognosis early and administer precise treatment are of great significance.

OBJECTIVE

The aim of this study was to use machine learning to construct a model for the analysis of risk factors and prediction of mortality among ICU patients with COVID-19.

METHODS

In this study, 123 patients with COVID-19 in the ICU of Vulcan Hill Hospital were retrospectively selected from the database, and the data were randomly divided into a training data set (n=98) and test data set (n=25) with a 4:1 ratio. Significance tests, correlation analysis, and factor analysis were used to screen 100 potential risk factors individually. Conventional logistic regression methods and four machine learning algorithms were used to construct the risk prediction model for the prognosis of patients with COVID-19 in the ICU. The performance of these machine learning models was measured by the area under the receiver operating characteristic curve (AUC). Interpretation and evaluation of the risk prediction model were performed using calibration curves, SHapley Additive exPlanations (SHAP), Local Interpretable Model-Agnostic Explanations (LIME), etc, to ensure its stability and reliability. The outcome was based on the ICU deaths recorded from the database.

RESULTS

Layer-by-layer screening of 100 potential risk factors finally revealed 8 important risk factors that were included in the risk prediction model: lymphocyte percentage, prothrombin time, lactate dehydrogenase, total bilirubin, eosinophil percentage, creatinine, neutrophil percentage, and albumin level. Finally, an eXtreme Gradient Boosting (XGBoost) model established with the 8 important risk factors showed the best recognition ability in the training set of 5-fold cross validation (AUC=0.86) and the verification queue (AUC=0.92). The calibration curve showed that the risk predicted by the model was in good agreement with the actual risk. In addition, using the SHAP and LIME algorithms, feature interpretation and sample prediction interpretation algorithms of the XGBoost black box model were implemented. Additionally, the model was translated into a web-based risk calculator that is freely available for public usage.

CONCLUSIONS

The 8-factor XGBoost model predicts risk of death in ICU patients with COVID-19 well; it initially demonstrates stability and can be used effectively to predict COVID-19 prognosis in ICU patients.

Elderly patients with cancer admitted to intensive care unit: A multicenter study in a middle-income country

BACKGROUND

Very elderly critically ill patients (ie, those older than 75 or 80 years) are an increasing population in intensive care units. However, patients with cancer have encompassed only a minority in epidemiological studies of very old critically-ill patients. We aimed to describe clinical characteristics and identify factors associated with hospital mortality in a cohort of patients aged 80 or older with cancer admitted to intensive care units (ICUs).

METHODS

This was a retrospective cohort study in 94 ICUs in Brazil. We included patients aged 80 years or older with active cancer who had an unplanned admission. We performed a mixed effect logistic regression model to identify variables independently associated with hospital mortality.

RESULTS

Of 4604 included patients, 1807 (39.2%) died in hospital. Solid metastatic (OR = 2.46; CI 95%, 2.01-3.00), hematological cancer (OR = 2.32; CI 95%, 1.75-3.09), moderate/severe performance status impairment (OR = 1.59; CI 95%, 1.33-1.90) and use of vasopressors (OR = 4.74; CI 95%, 3.88-5.79), mechanical ventilation (OR = 1.54; CI 95%, 1.25-1.89) and renal replacement (OR = 1.81; CI 95%, 1.29-2.55) therapy were independently associated with increased hospital mortality. Emergency surgical admissions were associated with lower mortality compared to medical admissions (OR = 0.71; CI 95%, 0.52-0.96).

CONCLUSIONS

Hospital mortality rate in very elderly critically ill patients with cancer with unplanned ICU admissions are lower than expected a priori. Cancer characteristics, performance status impairment and acute organ dysfunctions are associated with increased mortality.

Organizational factors associated with adherence to low tidal volume ventilation: a secondary analysis of the CHECKLIST-ICU database

Background

Survival benefit from low tidal volume (V_T) ventilation (LTVV) has been demonstrated for patients with acute respiratory distress syndrome (ARDS), and patients not having ARDS could also benefit from this strategy. Organizational factors may play a role on adherence to LTVV. The present study aimed to identify organizational factors with an independent association with adherence to LTVV.

Methods

Secondary analysis of the database of a multicenter two-phase study (prospective cohort followed by a cluster-randomized trial) performed in 118 Brazilian intensive care units. Patients under mechanical ventilation at day 2 were included. LTVV was defined as a V_T ≤ 8 ml/kg PBW on the second day of ventilation. Data on the type and number of beds of the hospital, teaching status, nursing, respiratory therapists and physician staffing, use of structured checklist, and presence of protocols were tested. A multivariable mixed-effect model was used to assess the association between organizational factors and adherence to LTVV.

Results

The study included 5719 patients; 3340 (58%) patients received LTVV. A greater number of hospital beds (absolute difference 7.43% [95% confidence interval 0.61–14.24%]; p = 0.038), use of structured checklist during multidisciplinary rounds (5.10% [0.55–9.81%]; p = 0.030), and presence of at least one nurse per 10 patients during all shifts (17.24% [0.85–33.60%]; p = 0.045) were the only three factors that had an independent association with adherence to LTVV.

Conclusions

Number of hospital beds, use of a structured checklist during multidisciplinary rounds, and nurse staffing are organizational factors associated with adherence to LTVV. These findings shed light on organizational factors that may improve ventilation in critically ill patients.