Early prediction of sepsis in intensive care patients using the machine learning algorithm NAVOY® Sepsis, a prospective randomized clinical validation study.

PURPOSE: To prospectively validate, in an ICU setting, the prognostic accuracy of the sepsis prediction algorithm NAVOY® Sepsis which uses 4 h of input for routinely collected vital parameters, blood gas values, and lab values. MATERIALS AND METHODS: Patients 18 years or older admitted to the ICU at Skåne University Hospital Malmö from December 2020 to September 2021 were recruited in the study. A total of 304 patients were randomized into one of two groups: Algorithm group with active sepsis alerts, or Standard of care. NAVOY® Sepsis made silent predictions in the Standard of care group, in order to evaluate its performance without disturbing the outcome. The study was blinded, i.e., study personnel did not know to which group patients were randomized. The healthcare provider followed standard practices in assessing possible development of sepsis and intervening accordingly. The patients were followed-up in the study until ICU discharge. RESULTS: NAVOY® Sepsis could predict the development of sepsis, according to the Sepsis-3 criteria, three hours before sepsis onset with high performance: accuracy 0.79; sensitivity 0.80; and specificity 0.78. CONCLUSIONS: The accuracy, sensitivity, and specificity were all high, validating the prognostic accuracy of NAVOY® Sepsis in an ICU setting, including Covid-19 patients.

A Machine Learning Algorithm Predicting Acute Kidney Injury in Intensive Care Unit Patients (NAVOY Acute Kidney Injury): Proof-of-Concept Study.

BACKGROUND: Acute kidney injury (AKI) represents a significant global health challenge, leading to increased patient distress and financial health care burdens. The development of AKI in intensive care unit (ICU) settings is linked to prolonged ICU stays, a heightened risk of long-term renal dysfunction, and elevated short- and long-term mortality rates. The current diagnostic approach for AKI is based on late indicators, such as elevated serum creatinine and decreased urine output, which can only detect AKI after renal injury has transpired. There are no treatments to reverse or restore renal function once AKI has developed, other than supportive care. Early prediction of AKI enables proactive management and may improve patient outcomes. OBJECTIVE: The primary aim was to develop a machine learning algorithm, NAVOY Acute Kidney Injury, capable of predicting the onset of AKI in ICU patients using data routinely collected in ICU electronic health records. The ultimate goal was to create a clinical decision support tool that empowers ICU clinicians to proactively manage AKI and, consequently, enhance patient outcomes. METHODS: We developed the NAVOY Acute Kidney Injury algorithm using a hybrid ensemble model, which combines the strengths of both a Random Forest (Leo Breiman and Adele Cutler) and an XGBoost model (Tianqi Chen). To ensure the accuracy of predictions, the algorithm used 22 clinical variables for hourly predictions of AKI as defined by the Kidney Disease: Improving Global Outcomes guidelines. Data for algorithm development were sourced from the Massachusetts Institute of Technology Lab for Computational Physiology Medical Information Mart for Intensive Care IV clinical database, focusing on ICU patients aged 18 years or older. RESULTS: The developed algorithm, NAVOY Acute Kidney Injury, uses 4 hours of input and can, with high accuracy, predict patients with a high risk of developing AKI 12 hours before onset. The prediction performance compares well with previously published prediction algorithms designed to predict AKI onset in accordance with Kidney Disease: Improving Global Outcomes diagnosis criteria, with an impressive area under the receiver operating characteristics curve (AUROC) of 0.91 and an area under the precision-recall curve (AUPRC) of 0.75. The algorithm's predictive performance was externally validated on an independent hold-out test data set, confirming its ability to predict AKI with exceptional accuracy. CONCLUSIONS: NAVOY Acute Kidney Injury is an important development in the field of critical care medicine. It offers the ability to predict the onset of AKI with high accuracy using only 4 hours of data routinely collected in ICU electronic health records. This early detection capability has the potential to strengthen patient monitoring and management, ultimately leading to improved patient outcomes. Furthermore, NAVOY Acute Kidney Injury has been granted Conformite Europeenne (CE)-marking, marking a significant milestone as the first CE-marked AKI prediction algorithm for commercial use in European ICUs.

Self-reported documentation of goals and outcomes of nutrition care - A cross-sectional survey study of Scandinavian dietitians.

BACKGROUND: The documentation of goals and outcomes of nutrition care in Electronic Health Records is insufficient making further exploration of this of particular interest. Identifying common features in documentation practice among Scandinavian dietitians might provide information that can support improvement in this area. AIMS: To explore the associations between clinical dietitians' self-reported documentation of patients' goals and outcomes and demographic factors, self-reported implementation of the systematic framework the Nutrition Care Process 4th step (NCP) and its associated terminology, and factors associated with the workplace. METHODS: Data from a cross-sectional study based on a previously tested web-based survey (INIS) disseminated in 2017 to dietitians in Scandinavia (n = 494) was used. Respondents were recruited through e-mail lists, e-newsletters and social media groups for dietitians. Associations between countries regarding the reported documentation of goals and outcomes, implementation levels of the NCP 4th step, demographic information and factors associated with the workplace were measured through Chi-square test. Associations between dependent- and independent variables were measured through logistic regression analysis. RESULTS: Clinically practicing dietitians (n = 347) working in Scandinavia, Sweden (n = 249), Norway (n = 60), Denmark (n = 38), who had completed dietetic education participated. The reported documentation of goals and outcomes from nutrition intervention was highly associated with the reported implementation of NCP 4th step terminology (OR = 5.26; p = 0.009, OR = 3.56; p = 0.003), support from the workplace (OR = 4.0, p < 0.001, OR = 8.89, p < 0.001) and area of practice (OR = 2.02, p = 0.017). Years since completed dietetic training and educational level did not have any significant associations with documentation practice regarding goals and outcomes. CONCLUSION: Findings highlight strong associations between the implementation of the NCP 4th step terminology and the documentation of goals and outcomes. Strategies to support dietitians in using standardized terminology and the development of tools for comprehensive documentation of evaluation of goals and outcome are required.

The Potential Cost and Cost-Effectiveness Impact of Using a Machine Learning Algorithm for Early Detection of Sepsis in Intensive Care Units in Sweden.

Background: Early diagnosis of sepsis has been shown to reduce treatment delays, increase appropriate care, and reduce mortality. The sepsis machine learning algorithm NAVOY® Sepsis, based on variables routinely collected at intensive care units (ICUs), has shown excellent predictive properties. However, the economic consequences of forecasting the onset of sepsis are unknown. Objectives: The potential cost and cost-effectiveness impact of a machine learning algorithm forecasting the onset of sepsis was estimated in an ICU setting. Methods: A health economic model has been developed to capture short-term and long-term consequences of sepsis. The model is based on findings from a randomized, prospective clinical evaluation of NAVOY® Sepsis and from literature sources. Modeling the relationship between time from sepsis onset to treatment and prevalence of septic shock and in-hospital mortality were of particular interest. The model base case assumes that the time to treatment coincides with the time to detection and that the algorithm predicts sepsis 3 hours prior to onset. Total costs include the costs of the prediction algorithm, days spent at the ICU and hospital ward, and long-term consequences. Costs are estimated for an average patient admitted to the ICU and for the healthcare system. The reference method is sepsis diagnosis in accordance with clinical practice. Results: In Sweden, the total cost per patient amounts to €16 436 and €16 512 for the algorithm and current practice arms, respectively, implying a potential cost saving per patient of €76. The largest cost saving is for the ICU stay, which is reduced by 0.16 days per patient (5860 ICU days for the healthcare sector) resulting in a cost saving of €1009 per ICU patient. Stochastic scenario analysis showed that NAVOY® Sepsis was a dominant treatment option in most scenarios and well below an established threshold of €20 000 per quality-adjusted life-year. A 3-hour faster detection implies a reduction in in-hospital mortality, resulting in 356 lives saved per year. Conclusions: A sepsis prediction algorithm such as NAVOY® Sepsis reduces the cost per ICU patient and will potentially have a substantial cost-saving and life-saving impact for ICU departments and the healthcare system.

A Machine Learning Sepsis Prediction Algorithm for Intended Intensive Care Unit Use (NAVOY Sepsis): Proof-of-Concept Study.

BACKGROUND: Despite decades of research, sepsis remains a leading cause of mortality and morbidity in intensive care units worldwide. The key to effective management and patient outcome is early detection, for which no prospectively validated machine learning prediction algorithm is currently available for clinical use in Europe. OBJECTIVE: We aimed to develop a high-performance machine learning sepsis prediction algorithm based on routinely collected intensive care unit data, designed to be implemented in European intensive care units. METHODS: The machine learning algorithm was developed using convolutional neural networks, based on Massachusetts Institute of Technology Lab for Computational Physiology MIMIC-III clinical data from intensive care unit patients aged 18 years or older. The model uses 20 variables to produce hourly predictions of onset of sepsis, defined by international Sepsis-3 criteria. Predictive performance was externally validated using hold-out test data. RESULTS: The algorithm-NAVOY Sepsis-uses 4 hours of input and can identify patients with high risk of developing sepsis, with high performance (area under the receiver operating characteristics curve 0.90; area under the precision-recall curve 0.62) for predictions up to 3 hours before sepsis onset. CONCLUSIONS: The prediction performance of NAVOY Sepsis was superior to that of existing sepsis early warning scoring systems and comparable with those of other prediction algorithms designed to predict sepsis onset. The algorithm has excellent predictive properties and uses variables that are routinely collected in intensive care units.

Plasma proteomics of biomarkers for inflammation or cancer cannot predict relapse in chronic myeloid leukaemia patients stopping tyrosine kinase inhibitor therapy.

Several studies have now shown that chronic myeloid leukaemia (CML) patients in deep molecular remission may discontinue tyrosine kinase inhibitor (TKI) treatment with a treatment free remission (TFR) rate of approximately 40-60 %. Some factors influencing the possibility of TFR have been described but better tools are needed for individual prediction of long-term TFR. Herein, two multiplex panels were utilised to analyse a total of 162 different plasma proteins from 56 patients included in the TKI stopping trial EURO-SKI (Saussele et al., 2018). The purpose was to identify possible plasma protein markers for prediction of successful TKI discontinuation and to evaluate effects of TKI discontinuation on plasma protein profiles. No protein biomarkers sampled before TKI discontinuation could separate relapse cases from non-relapse cases but some plasma proteins differed between patients who relapsed and those who remained in TFR when followed over time after TKI cessation. In conclusion, the plasma protein markers in this study could not predict relapse after TKI discontinuation but may be of use to understand the mechanisms involved in maintenance of TFR.

Adherence to a regulation that aims to prevent and treat malnutrition-The case of Swedish elderly care.

Malnutrition constitutes a serious and challenging problem in elderly care. In 2015, a Swedish regulation that aims to prevent and treat malnutrition came into effect. This study set out to explore associations between level of adoption of the regulation reported as: no, started, yes, in a previous survey, and registrations in a national quality registry. Registry data on screening and actions extracted from the first trimester in 2014 (n=18967), 2016 (n=20318) and 2017 (n= 25669) represented 209, 197 and 199 of 290 Swedish municipalities respectively. A repeated measures ANOVA showed that there was no effect on screened nutritional status, Pearson's chi-square that there were minor differences in types of actions, and regression analysis that the number of actions increased on average by 0.3 due to a higher level of adoption of the regulation. Over the years studied, five actions were prominent regardless of level of adoption or screened nutritional status. Hence, to date, no firm conclusions regarding effects of the regulation can be drawn. Despite the regulatory nature, it appear as if the regulation and the level of adoption reported so far is routine in theory, although not yet leveraged to an implemented practice visible in the quality registry but instead decoupled from practice.

Multivariate two-sample permutation tests for trials with multiple time-to-event outcomes.

Clinical trials involving multiple time-to-event outcomes are increasingly common. In this paper, permutation tests for testing for group differences in multivariate time-to-event data are proposed. Unlike other two-sample tests for multivariate survival data, the proposed tests attain the nominal type I error rate. A simulation study shows that the proposed tests outperform their competitors when the degree of censored observations is sufficiently high. When the degree of censoring is low, it is seen that naive tests such as Hotelling's T2 outperform tests tailored to survival data. Computational and practical aspects of the proposed tests are discussed, and their use is illustrated by analyses of three publicly available datasets. Implementations of the proposed tests are available in an accompanying R package.

Reforming foodservice in elderly care: National actions and local outcomes.

AIM: The aim was to explore the outcome, on a local level, of steering, organisation and practices of elderly care foodservice by Swedish municipalities, and changes relative to national actions. METHODS: A survey using a web-based questionnaire about elderly care foodservice targeting all Swedish municipalities (n = 290) was conducted in 2006 and 2013/2014. The questionnaire included the topics: organisation of foodservice, its practice in elderly care and steering devices such as guidelines and policies. Based on the share of a rural population, municipalities were divided into groups: rural (≥50%), urban (<50%) and city (≤20%). RESULTS: The response rate from municipalities was 80% in 2006 and 56% in 2013/2014; 45% participated in both surveys. The results showed increased use of local food policies (P = 0.03) and meal choice (P < 0.001), while access to clinical/community dietitians declined (P = 0.01) between the surveys. In home-help services, daily delivered cook-serve meals declined (P < 0.001) and chilled meals delivered three times a week increased (P = 0.002) between the surveys. City municipalities used private foodservice organisations the most (P < 0.001), and reported reduced use of cook-serve systems in favour of chilled. In rural municipalities, the use of public providers (98%) and a cook-serve system (94%) were firmly established. Urban municipalities were placed between the other groups. CONCLUSIONS: National actions such as soft governance and benchmarking appear largely to determine local level outcomes. However, conditions for adapting these measures vary between municipality groups. While efficiency enhancing trends were prominent, questions remain whether national actions should be expanded beyond performance to also examine their consequences.

Plasma proteomics in CML patients before and after initiation of tyrosine kinase inhibitor therapy reveals induced Th1 immunity and loss of angiogenic stimuli.

BACKGROUND AND AIMS: The simultaneous measurement of many proteins is now possible using multiplex assays. In this pilot study we investigated a total of 124 proteins in plasma from chronic myeloid leukemia (CML) patients with the purpose of identifying proteins that are differently expressed at diagnosis and after tyrosine kinase inhibitor (TKI) treatment initiation. METHODS: Samples were taken from 14 CML patients at diagnosis and after three months of TKI treatment (imatinib or dasatinib). Samples were analyzed by Mesoscale Discovery, Myriad RBM MAP technology and Olink Proseek. RESULTS: Multiple plasma proteins were differentially expressed before and after initiation of TKI therapy. Protein patterns demonstrated a possible shift towards Th1-immunity and reduced angiogenic stimuli. Further, some plasma proteins were identified that can be of potential interest to study further for biologic, prognostic or therapeutic significance such as E-selectin, uPAR, growth hormone and carbonic anhydrase IX. CONCLUSIONS: Plasma proteomics seems feasible and useful in CML patients, both for studying patterns of protein expression and for identifying single proteins differentially expressed before and after treatment. Plasma proteomics may be useful to map disease activity and biological processes. Hence, plasma proteomics can be used to understand drug mechanisms and treatment responses in CML.

Validation of an algorithm for chronic periodontitis risk assessment and prognostication: analysis of an inflammatory reactivity test and selected risk predictors.

BACKGROUND: Patients with severe forms of chronic periodontitis present with varying degrees of decreased inflammatory reactivity. A previously reported algorithm for chronic periodontitis risk assessment and prognostication is based on the analysis of some 20 risk predictors. One of these predictors is a skin provocation test that assesses the individual patient's reactivity to a lipid A challenge. The aim of this report was to analyze results from validation data for the algorithm with respect to the contribution of results of the skin provocation test as a risk predictor for the progression of chronic periodontitis and to compare these results with the contribution from other predictors, namely smoking, angular bony destruction, furcation involvement, abutment teeth, and endodontic pathology. METHODS: Data from a previously reported clinical validation sample were used for the analysis, including the calculation of quality measures and explanatory values using different types of regression analysis and non-parametric testing. RESULTS: Smoking, endodontic pathology, abutment teeth, angular bony destruction, and furcation involvement presented with individual explanatory values for periodontitis progression between 4% and 13% and highly significant parameter estimates. Explanatory values for the results of the skin provocation test ranged between 2.6% and 5.1% depending on the disease severity group, with a positive predictive value of 82% for the identification of high-risk patients. CONCLUSION: The skin provocation test provided a clinically significant contribution to the quality of analysis with the periodontitis risk and prognostication algorithm, in particular in the selection of high-risk patients for in-depth individual tooth analysis.

Validation of an algorithm for chronic periodontitis risk assessment and prognostication: risk predictors, explanatory values, measures of quality, and clinical use.

BACKGROUND: The American Academy of Periodontology has recently stated that, "[risk assessment will become] increasingly important in periodontal treatment planning and should be part of every comprehensive dental and periodontal evaluation." (J Periodontol 2006;77:1608). Unaided risk assessment and prognostication show significant variability because chronic periodontitis is a multifactorial disease. This report summarizes the clinical validation of an algorithm for chronic periodontitis risk assessment and prognostication. The algorithm is a Web-based analytic tool that integrates some 20 risk predictors and calculates scores indicating levels of risk for chronic periodontitis for the dentition (Level I) and, if an elevated risk is found, prognosticates disease progression tooth by tooth (Level II). METHODS: An independent clinical validation sample was generated in an open, prospective clinical trial and analyzed in a predetermined validation plan. RESULTS: The analyses identified two threshold scores above which significant progression of periodontitis was found. Based on these scores, sufficiently high explanatory values with significant and increasing parameter estimates for increasing risk were established in Level I, justifying detailed analysis tooth by tooth in Level II. Subsequent prognostication of chronic periodontitis in Level II was found to be accompanied by clinically relevant measures of quality in relation to rates of disease progression. Three score intervals representing increasing levels of periodontitis progression were identified corresponding to increasing levels of significant annual marginal bone loss. CONCLUSIONS: The predictors included in the algorithm reflect a relevant selection for periodontitis risk assessment. Risk assessment and prognostication with the algorithm provides the clinician with a validated, reliable, consistent, and objective tool supporting treatment planning.

Efficacy and safety of quetiapine in combination with lithium or divalproex for maintenance of patients with bipolar I disorder (international trial 126).

BACKGROUND: This study examined the efficacy and safety of quetiapine in combination with lithium or divalproex compared with placebo with lithium or divalproex in the prevention of recurrent mood events in bipolar I patients, most recent episode mania, depression, or mixed. METHODS: Patients received open-label quetiapine (400-800 mg/day; flexible, divided doses) with lithium or divalproex (target serum concentrations 0.5-1.2 mEq/L and 50-125 microg/mL, respectively) for up to 36 weeks to achieve at least 12 weeks of clinical stability. Patients were subsequently randomized to double-blind treatment with quetiapine (400-800 mg/day) plus lithium/divalproex or placebo plus lithium/divalproex for up to 104 weeks. The primary endpoint was time to recurrence of any mood event. RESULTS: Treatment with quetiapine in combination with lithium/divalproex significantly increased the time to recurrence of any mood event compared with placebo plus lithium/divalproex. The proportion of patients having a mood event was markedly lower in the quetiapine than in the placebo group (18.5% versus 49.0%). The hazard ratio for time to recurrence of a mood event was 0.28 (P<0.001), a mania event 0.30 (P<0.001), and a depression event 0.26 (P<0.001) corresponding to risk reductions of 72%, 70%, and 74%, respectively. During the randomization phase, the most common adverse events occurring in > or =5% in the quetiapine group were somnolence, nasopharyngitis, and headache. Insomnia was more common in the placebo group. During the randomization phase, there was an increase in weight of 0.5 kg in the quetiapine group and a reduction of 1.9 kg in the placebo group. The incidence and incidence density of a single emergent fasting blood glucose value> or =126 mg/dL was higher with quetiapine than with placebo (9.3% versus 4.1%; 17.6 versus 9.5 patients per 100 patient-years). LIMITATIONS: This was an enriched sample of patients with bipolar I disorder responding to treatment with quetiapine plus lithium/divalproex. CONCLUSIONS: Maintenance treatment with quetiapine in combination with lithium/divalproex significantly increased time to recurrence of any event (mania, depression, or mixed) irrespective of the polarity of the index episode compared with placebo with lithium/divalproex. Long-term treatment with quetiapine was generally well-tolerated. Quetiapine with lithium/divalproex can provide an effective long-term treatment option for bipolar I disorder to prevent recurrences not only of mania but also depression.

Prevention of schizophrenia relapse with extended release quetiapine fumarate dosed once daily: a randomized, placebo-controlled trial in clinically stable patients.

INTRODUCTION: This long-term, randomized, double-blind, placebo-controlled study examined the efficacy of extended release quetiapine fumarate (quetiapine XR) in preventing psychotic relapse in schizophrenia. METHODS: Three hundred twenty-seven clinically stable patients with schizophrenia were switched to open-label quetiapine XR (300mg on Day 1, 600mg on Day 2, followed by flexible dosing [400-800mg/day]) for a 16-week stabilization phase. Thereafter, patients who were clinically stable for four months were randomized to flexible doses of quetiapine XR (400-800mg/day) or placebo. Primary endpoint was time to first schizophrenia relapse after randomization. Secondary endpoints included risk of relapse at six months. Interim analyses were planned after 45 and 60 relapses and final analysis after 90 relapses. Maximal treatment time was one year. RESULTS: The study was terminated after the first interim analysis showed a significant difference between randomized treatment groups. Time to relapse was significantly longer in quetiapine XR-treated patients versus placebo (hazard ratio 0.16 [95% confidence interval 0.08, 0.34]; p=0.001). Fewer quetiapine XR-treated patients relapsed versus those receiving placebo (10.7% vs. 41.4%, respectively). Estimated risk of relapse at six months was significantly lower with quetiapine XR (14.3%) compared with placebo (68.2%; p=0.0001). The incidence of treatment-related adverse events (AEs) was similar between quetiapine XR and placebo groups (18% and 21% of patients, respectively) and only one percent of patients in each group withdrew because of AEs. CONCLUSION: Once-daily quetiapine XR (400-800mg/day) was effective in preventing relapse in patients with clinically stable schizophrenia. Quetiapine XR was well tolerated during longer-term use.