A comparison of predictors for mortality and bacteraemia in patients suspected of infection.

BACKGROUND: Stratification by clinical scores of patients suspected of infection can be used to support decisions on treatment and diagnostic workup. Seven clinical scores, SepsisFinder (SF), National Early Warning Score (NEWS), Sequential Orgen Failure Assessment (SOFA), Mortality in Emergency Department Sepsis (MEDS), quick SOFA (qSOFA), Shapiro Decision Rule (SDR) and Systemic Inflammatory Response Syndrome (SIRS), were evaluated for their ability to predict 30-day mortality and bacteraemia and for their ability to identify a low risk group, where blood culture may not be cost-effective and a high risk group where direct-from-blood PCR (dfbPCR) may be cost effective. METHODS: Retrospective data from two Danish and an Israeli hospital with a total of 1816 patients were used to calculate the seven scores. RESULTS: SF had higher Area Under the Receiver Operating curve than the clinical scores for prediction of mortality and bacteraemia, significantly so for MEDS, qSOFA and SIRS. For mortality predictions SF also had significantly higher area under the curve than SDR. In a low risk group identified by SF, consisting of 33% of the patients only 1.7% had bacteraemia and mortality was 4.2%, giving a cost of € 1976 for one positive result by blood culture. This was higher than the cost of € 502 of one positive dfbPCR from a high risk group consisting of 10% of the patients, where 25.3% had bacteraemia and mortality was 24.2%. CONCLUSION: This may motivate a health economic study of whether resources spent on low risk blood cultures might be better spent on high risk dfbPCR.

Usability study of a new tool for nutritional and glycemic management in adult intensive care: Glucosafe 2.

The new decision support tool Glucosafe 2 (GS2) is based on a mathematical model of glucose and insulin dynamics, designed to assist caregivers in blood glucose control and nutrition. This study aims to assess end-user acceptance and usability of this bedside decision support tool in an adult intensive care setting. Caregivers were first trained and then invited to trial GS2 prototype on bedside computers. Data for qualitative analysis were collected through semi-structured interviews from twenty users after minimum three trial days. Most caregivers (70%) rated GS2 as convenient and believed it would help improving adherence to current guidelines (85%). Moreover, most nurses (80%) believed that GS2 would be timesaving. Nurses' risk perceptions and manual data entry emerged as central barriers to use GS2 in routine practice. Issues emerged from the caregivers were compiled into a list of 12 modifications of the GS2 prototype to increase end-user acceptance and usability. This usability study showed that GS2 was considered by ICU caregivers as helpful in daily clinical practice, allowing time-saving and better standardization of ICU patient's care. Important issues were raised by the users with implications for the development and deployment of GS2. Integrating the technology into existing IT infrastructure may facilitate caregivers' acceptance. Further clinical studies of the performance and potential health outcomes are warranted.

Clinical- vs. model-based selection of patients suspected of sepsis for direct-from-blood rapid diagnostics in the emergency department: a retrospective study.

Selecting high-risk patients may improve the cost-effectiveness of rapid diagnostics. Our objective was to assess whether model-based selection or clinical selection is better for selecting high-risk patients with a high rate of bacteremia and/or DNAemia. This study involved a model-based, retrospective selection of patients from a cohort from which clinicians selected high-risk patients for rapid direct-from-blood diagnostic testing. Patients were included if they were suspected of sepsis and had blood cultures ordered at the emergency department. Patients were selected by the model by adding those with the highest probability of bacteremia until the number of high-risk patients selected by clinicians was reached. The primary outcome was bacteremia rate. Secondary outcomes were DNAemia rate, and 30-day mortality. Data were collected for 1395 blood cultures. Following exclusion, 1142 patients were included in the analysis. In each high-risk group, 220/1142 were selected, where 55 were selected both by clinicians and the model. For the remaining 165 in each group, the model selected for a higher bacteremia rate (74/165, 44.8% vs. 45/165, 27.3%, p = 0.001), and a higher 30-day mortality (49/165, 29.7% vs. 19/165, 11.5%, p = 0.00004) than the clinically selected group. The model outperformed clinicians in selecting patients with a high rate of bacteremia. Using such a model for risk stratification may contribute towards closing the gap in cost between rapid and culture-based diagnostics.

Predicting Antibiotic Resistance in Urinary Tract Infection Patients with Prior Urine Cultures.

To improve antibiotic prescribing, we sought to establish the probability of a resistant organism in urine culture given a previous resistant culture in a setting endemic for multidrug-resistant (MDR) organisms. We performed a retrospective analysis of inpatients with paired positive urine cultures. We focused on ciprofloxacin-resistant (cipro(r)) Gram-negative bacteria, extended-spectrum-beta-lactamase (ESBL)-producing Enterobacteriaceae, carbapenem-resistant Enterobacteriaceae (CRE), and carbapenem-resistant nonfermenters (CRNF). Comparisons were made between the frequency of each resistance phenotype following a previous culture with the same phenotype and the overall frequency of that phenotype, and odds ratios (ORs) were calculated. We performed a regression to assess the effects of other variables on the likelihood of a repeat resistant culture. A total of 4,409 patients (52.5% women; median age, 70 years) with 19,546 paired positive urine cultures were analyzed. The frequencies of cipro(r) bacteria, ESBL-producing Enterobacteriaceae, CRE, and CRNF among all cultures were 47.7%, 30.6%, 1.7%, and 2.6%, respectively. ORs for repeated resistance phenotypes were 1.87, 3.19, 48.25, and 19.02 for cipro(r) bacteria, ESBL-producing Enterobacteriaceae, CRE, and CRNF, respectively (P < 0.001 for all). At 1 month, the frequencies of repeated resistance phenotypes were 77.4%, 66.4%, 57.1%, and 33.3% for cipro(r) bacteria, ESBL-producing Enterobacteriaceae, CRE, and CRNF, respectively. Increasing time between cultures and the presence of an intervening nonresistant culture significantly reduced the chances of a repeat resistant culture. Associations were statistically significant over the duration of follow-up (60 months) for CRE and for up to 6 months for all other pathogens. Knowledge of microbiology results in the six preceding months may assist with antibiotic stewardship and improve the appropriateness of empirical treatment for urinary tract infections (UTIs).

The use of venous blood gas in assessing arterial acid-base and oxygenation status - an analysis of aggregated data from multiple studies evaluating the venous to arterial conversion (v-TAC) method.

BACKGROUND: Several methods exist to reduce the number of arterial blood gases (ABGs). One method, Roche v-TAC, has been evaluated in different patient groups. This paper aggregates data from these studies, in different patient categories using common analysis criteria. RESEARCH DESIGN AND METHODS: We included studies evaluating v-TAC based on paired arterial and peripheral venous blood samples. Bland-Altman analysis compared measured and calculated arterial values of pH, PCO2, and PO2. Subgroup analyses were performed for normal, chronic hypercapnia and chronic base excess, acute hyper- and hypocapnia, and acute and chronic base deficits. RESULTS: 811 samples from 12 studies were included. Bias and limits of agreement for measured and calculated values: pH 0.001 (-0.029 to 0.031), PCO2 -0.08 (-0.65 to 0.49) kPa, and PO2 0.04 (-1.71 to 1.78) kPa, with similar values for all sub-group analyses. CONCLUSION: These data suggest that v-TAC analysis may have a role in replacing ABGs, avoiding arterial puncture. Substantial data exist in patients with chronic hypercapnia and chronic base excess, acute hyper- and hypocapnia, and in patients with relatively normal acid-base status, with similar bias and precision across groups and across study data. Limited data exist for patients with acute and chronic base deficits.

Clinical refinement of the automatic lung parameter estimator (ALPE).

The automatic lung parameter estimator (ALPE) method was developed in 2002 for bedside estimation of pulmonary gas exchange using step changes in inspired oxygen fraction (FIO2). Since then a number of studies have been conducted indicating the potential for clinical application and necessitating systems evolution to match clinical application. This paper describes and evaluates the evolution of the ALPE method from a research implementation (ALPE1) to two commercial implementations (ALPE2 and ALPE3). A need for dedicated implementations of the ALPE method was identified: one for spontaneously breathing (non-mechanically ventilated) patients (ALPE2) and one for mechanically ventilated patients (ALPE3). For these two implementations, design issues relating to usability and automation are described including the mixing of gasses to achieve FIO2 levels, and the automatic selection of FIO2. For ALPE2, these improvements are evaluated against patients studied using the system. The major result is the evolution of the ALPE method into two dedicated implementations, namely ALPE2 and ALPE3. For ALPE2, the usability and automation of FIO2 selection has been evaluated in spontaneously breathing patients showing that variability of gas delivery is 0.3 % (standard deviation) in 1,332 breaths from 20 patients. Also for ALPE2, the automated FIO2 selection method was successfully applied in 287 patient cases, taking 7.2 ± 2.4 min and was shown to be safe with only one patient having SpO2 < 86 % when the clinician disabled the alarms. The ALPE method has evolved into two practical, usable systems targeted at clinical application, namely ALPE2 for spontaneously breathing patients and ALPE3 for mechanically ventilated patients. These systems may promote the exploration of the use of more detailed descriptions of pulmonary gas exchange in clinical practice.

Decision support for optimized blood glucose control and nutrition in a neurotrauma intensive care unit: preliminary results of clinical advice and prediction accuracy of the Glucosafe system.

Assessment of glycemic control with model-based decision support ("Glucosafe") in neurotrauma intensive care patients in an ongoing randomized controlled trial with a blood glucose (BG) target of 5-8 mmol/L. Assessment of BG prediction accuracy of the model and assessment of the effect that two potential model extensions would have on prediction accuracy in this trial. In the intervention group insulin infusion rates and nutrition are varied based on Glucosafe's decision support. In the control group, the caloric target is 25-30 kcal/kg per day and insulin is regulated according to department rules. BG concentrations, insulin infusion rates, and feed rates are compared from the data of 12 consecutive patients. BG measurements are predicted retrospectively and the mean relative prediction error is calculated using (1) the current model from the trial, (2) the current model modified by using a BG-dependent variable endogenous insulin appearance rate, (3) the current model modified by a patient-specific carbohydrate absorption factor. BG control was improved by Glucosafe. 76 % of BG measurements in Glucosafe patients were in the 5-8 mmol/L band (Controls: 51 %). BG means (log-normal) ± SD were 7.0 ± 1.19 mmol/L in Glucosafe patients compared to 8.0 ± 1.24 mmol/L in controls (P = 0.05). Mean caloric intake was 93.5 ± 15 % of resting energy expenditure in Glucosafe patients (Controls: 129 ± 29 %). The BG-dependent variable insulin appearance rate had no measurable effect on prediction accuracy. The patient-specific carbohydrate absorption factor improved prediction accuracy significantly (P = 0.001). Glucosafe advice reduces hyperglycemia in neurotrauma intensive care patients. Further parameterization can improve model prediction accuracy.

Enabling a learning healthcare system with automated computer protocols that produce replicable and personalized clinician actions.

Clinical decision-making is based on knowledge, expertise, and authority, with clinicians approving almost every intervention-the starting point for delivery of "All the right care, but only the right care," an unachieved healthcare quality improvement goal. Unaided clinicians suffer from human cognitive limitations and biases when decisions are based only on their training, expertise, and experience. Electronic health records (EHRs) could improve healthcare with robust decision-support tools that reduce unwarranted variation of clinician decisions and actions. Current EHRs, focused on results review, documentation, and accounting, are awkward, time-consuming, and contribute to clinician stress and burnout. Decision-support tools could reduce clinician burden and enable replicable clinician decisions and actions that personalize patient care. Most current clinical decision-support tools or aids lack detail and neither reduce burden nor enable replicable actions. Clinicians must provide subjective interpretation and missing logic, thus introducing personal biases and mindless, unwarranted, variation from evidence-based practice. Replicability occurs when different clinicians, with the same patient information and context, come to the same decision and action. We propose a feasible subset of therapeutic decision-support tools based on credible clinical outcome evidence: computer protocols leading to replicable clinician actions (eActions). eActions enable different clinicians to make consistent decisions and actions when faced with the same patient input data. eActions embrace good everyday decision-making informed by evidence, experience, EHR data, and individual patient status. eActions can reduce unwarranted variation, increase quality of clinical care and research, reduce EHR noise, and could enable a learning healthcare system.

Mathematical modelling of the acid-base chemistry and oxygenation of blood: a mass balance, mass action approach including plasma and red blood cells.

Mathematical models of the acid-base chemistry of blood based upon mass action and mass balance equations have become popular as diagnostic tools in intensive care. The reference models using this approach are those based on the strong ion approach, but these models do not currently take into account the effects of oxygen on the buffering characteristics of haemoglobin. As such these models are limited in their ability to simulate physiological situations involving simultaneous changes of O(2) and CO(2) levels in the blood. This paper describes a model of acid-base chemistry of blood based on mass action and mass balance equations and including the effects of oxygen. The model is used to simulate the mixing of venous blood with the same blood at elevated O(2) and reduced CO(2) levels, and the results compared with the mixing of blood sampled from 21 healthy subjects. Simulated values of pH, PCO(2), PO(2) and SO(2) in the mixed blood compare well with measured values with small bias (i.e. 0.000 pH, -0.06 kPa PCO(2), -0.1% SO(2), -0.02 kPa PO(2)), and values of standard deviations (i.e. 0.006 pH, 0.11 kPa PCO(2), 0.8% SO(2), 0.13 kPa PO(2)) comparable to the precision seen in direct measurement of these variables in clinical practice. These results indicate that the model can reliably simulate the mixing of blood and has potential for application in describing physiological situations involving the mixing of blood at different O(2) and CO(2) levels such as occurs in the mixing of lung capillary and shunted pulmonary blood.

Performance of the TREAT decision support system in an environment with a low prevalence of resistant pathogens.

OBJECTIVES: To evaluate a decision support system (TREAT) for guidance of empirical antimicrobial therapy in an environment with a low prevalence of resistant pathogens. METHODS: A retrospective trial of TREAT has been performed at Copenhagen University, Hvidovre Hospital. The cohort of patients included adults with systemic inflammation and suspicion of community-acquired bacterial infection. The empirical antimicrobial treatment recommended by TREAT was compared with the empirical antimicrobial treatment prescribed by the first attending clinical physician. RESULTS: Out of 171 patients recruited, 161 (65 with microbiologically documented infections) fulfilled the inclusion criteria of TREAT. Coverage achieved by TREAT was significantly higher than that by clinical practice (86% versus 66%, P = 0.007). There was no significant difference in the cost of future resistance between treatments chosen by TREAT and those by physicians. The direct expenses for antimicrobials were higher in TREAT when including patients without antimicrobial treatment, while there was no significant difference otherwise. The cost of side effects was significantly lower using TREAT. CONCLUSIONS: The results of the study suggest that TREAT can improve the appropriateness of antimicrobial therapy and reduce the cost of side effects in regions with a low prevalence of resistant pathogens, however, at the expense of increased use of antibiotics.

Reporting of adverse events in randomized controlled trials of highly active antiretroviral therapy: systematic review.

OBJECTIVES: Our objectives were to systematically assess the quality of reporting of adverse events (AEs) in publications of randomized trials of highly active antiretroviral therapy (HAART), and to examine whether reporting quality affects the effect estimates reported for AEs. METHODS: We searched the PubMed, Cochrane library and EMBASE electronic databases up to December 2008. We included all published randomized controlled trials assessing HAART for treatment-naive adult HIV-infected individuals, with 48 weeks' follow-up. The quality of AE reporting was extracted according to CONSORT guidelines. We pooled the relative risks for AEs and compared results by sponsorship and different reporting methods. RESULTS: Forty-nine trials, including 19 882 patients, published between 2000 and 2008, met the inclusion criteria. Only one of the trials reported on AE collection methods. Twenty-six trials reported only AEs attributed to drugs, 17 of which did not refer to the attribution methods. AE reporting was nearly always selective and selection criteria were highly variable, based on severity grading or occurrence threshold. Presentation of AEs above an occurrence threshold was more common in studies sponsored by industry (30/31) than in studies sponsored by non-profit organizations (3/18). Moreover, we showed that differences in the methods of reporting AEs may affect the results reported for AEs. No significant improvement in AE reporting was seen over this period. CONCLUSIONS: We found substantial variability in AE reporting. Variability was influenced by sponsor identity and affected outcomes. These facts obstruct our ability to choose HAART based on currently published data.

Learning susceptibility of a pathogen to antibiotics using data from similar pathogens.

OBJECTIVES: Selection of empirical antibiotic therapy relies on knowledge of the in vitro susceptibilities of potential pathogens to antibiotics. In this paper the limitations of this knowledge are outlined and a method that can reduce some of the problems is developed. METHODS: We propose hierarchical Dirichlet learning for estimation of pathogen susceptibilities to antibiotics, using data from a group of similar pathogens in a bacteremia database. RESULTS: A threefold cross-validation showed that maximum likelihood (ML) estimates of susceptibilities based on individual pathogens gave a distance between estimates obtained from the training set and observed frequencies in the validation set of 16.3%. Estimates based on the initial grouping of pathogens gave a distance of 16.7%. Dirichlet learning gave a distance of 15.6%. Inspection of the pathogen groups led to subdivision of three groups, Citrobacter, Other Gram Negatives and Acinetobacter, out of 26 groups. Estimates based on the subdivided groups gave a distance of 15.4% and Dirichlet learning further reduced this to 15.0%. The optimal size of the imaginary sample inherited from the group was 3. CONCLUSION: Dirichlet learning improved estimates of susceptibilities relative to ML estimators based on individual pathogens and to classical grouped estimators. The initial pathogen grouping was well founded and improvement by subdivision of the groups was only obtained in three groups. Dirichlet learning was robust to these revisions of the grouping, giving improved estimates in both cases, while the group-based estimates only gave improved estimates after the revision of the groups.

Correlation between acid-base parameters measured in arterial blood and venous blood sampled peripherally, from vena cavae superior, and from the pulmonary artery.

OBJECTIVE: In intensive care units arterial blood sampling is routine for analysing acid-base and oxygenation status. In nonintensive departments arterial blood sampling is seldom performed. Venous blood sampling is routine but not usually analysed for acid-base and oxygenation status. This study describes the correlation between arterial and peripheral, central and mixed venous pH, PCO2 and PO2 in a wide range of adult patients. METHODS: Arterial and venous blood samples were taken anaerobically and simultaneously. The values of pH, PCO2 and PO2 were compared using Bland-Altman plots. RESULTS: A total of 103 patients were included. The arteriovenous difference (bias+/-SD) for pH was 0.026+/-0.023 and for PCO2 -0.60+/-0.57 kPa (peripheral venous blood), 0.036+/-0.014 and -0.79+/-0.26 kPa (central venous blood) and 0.026+/-0.010 and -0.67+/-0.22 kPa (mixed venous blood). The arteriovenous difference for PO2 for peripheral, central and mixed venous blood was 6.27+/-4.36, 8.33+/-3.94 and 11.00+/-4.87 kPa, respectively. CONCLUSION: The venous values of pH, corrected for bias, can give arterial values which are within reasonable laboratory and clinical acceptance criteria. For PCO2 this is also true, except for peripheral blood, where the standard deviation is outside laboratory acceptance criteria but within clinical acceptance criteria. For PO2 the arteriovenous differences are not randomly distributed and even for PO2

Generic causal probabilistic networks: a solution to a problem of transferability in medical decision support.

Causal probabilistic networks provide a natural framework for representation of medical knowledge, allowing clinical experts to encode assumptions about causal dependencies between stochastic variables. Application in medical decision support has produced promising results. However, model features and parameters may vary geo- or demographically. Therefore methods are needed that allow for easy adjustment of the model to a change in conditions. We present a method to represent causal probabilistic networks generically that maximizes the transferability of a models relevance and completeness, when moved from one environment to another, and illustrate application of the method with an example from a medical decision support system.

A decision support system for suggesting ventilator settings: retrospective evaluation in cardiac surgery patients ventilated in the ICU.

Selecting appropriate ventilator settings decreases the risk of ventilator-induced lung injury. A decision support system (DSS) has been developed based on physiological models, which can advise on setting of tidal volume (Vt), respiratory frequency (f) and fraction of inspired oxygen (FiO2). The aim of this study is to assess the feasibility of the DSS by comparing its advice with the values used in clinical practice. Data from 20 patients following uncomplicated coronary artery bypass grafting (CABG) with cardiopulmonary bypass was used to test the DSS. Ventilator settings suggested by the DSS were compared to the settings selected by the clinician. When compared to the clinician the DSS suggested: lowering FiO2 (by median 7%, range 2-17%) at high SpO2 and increasing FiO2 (by median 2%, range 1-5%) at low SpO2; lowering ventilation volume (by median 0.57 l min(-1), range 0.2-1.1 l min(-1)) at high pHa and increasing ventilation volume (by median 0.4 l min(-1), range 0.1-0.9 l min(-1)) at low pHa. Suggested changes in ventilation volume were such that simulated values of PIP were < or = 22.9 cmH2O and respiratory frequency < or = 18 breaths min(-1). In all cases, computer suggested values of FiO2, Vt or f were consistent with maintaining sufficient oxygenation, normalising pH and obtaining low values of PIP.

The TREAT project: decision support and prediction using causal probabilistic networks.

TREAT is a decision support system for antibiotic treatment in inpatients with common bacterial infections. It was tested in a randomised controlled trial in three countries and shown to improve the percentage of appropriate empirical antibiotic treatments, while at the same time reducing hospital stay and the use of broad-spectrum antibiotics. TREAT is based on a causal probabilistic network and uses a cost-benefit model for antibiotic treatment, including costs assigned to future resistance. In the present review we discuss the advantages of using causal probabilistic models for prediction and decision support, and the various decisions that were taken in the TREAT project.

Variation in the PaO2/FiO2 ratio with FiO2: mathematical and experimental description, and clinical relevance.

INTRODUCTION: Previous studies have shown through theoretical analyses that the ratio of the partial pressure of oxygen in arterial blood (PaO2) to the inspired oxygen fraction (FiO2) varies with the FiO2 level. The aim of the present study was to evaluate the relevance of this variation both theoretically and experimentally using mathematical model simulations, comparing these ratio simulations with PaO2/FiO2 ratios measured in a range of different patients. METHODS: The study was designed as a retrospective study using data from 36 mechanically ventilated patients and 57 spontaneously breathing patients studied on one or more occasions. Patients were classified into four disease groups (normal, mild hypoxemia, acute lung injury and acute respiratory distress syndrome) according to their PaO2/FiO2 ratio. On each occasion the patients were studied using four to eight different FiO2 values, achieving arterial oxygen saturations in the range 85-100%. At each FiO2 level, measurements were taken of ventilation, of arterial acid-base and of oxygenation status. Two mathematical models were fitted to the data: a one-parameter 'effective shunt' model, and a two-parameter shunt and ventilation/perfusion model. These models and patient data were used to investigate the variation in the PaO2/FiO2 ratio with FiO2, and to quantify how many patients changed disease classification due to variation in the PaO2/FiO2 ratio. An F test was used to assess the statistical difference between the two models' fit to the data. A confusion matrix was used to quantify the number of patients changing disease classification. RESULTS: The two-parameter model gave a statistically better fit to patient data (P < 0.005). When using this model to simulate variation in the PaO2/FiO2 ratio, disease classification changed in 30% of the patients when changing the FiO2 level. CONCLUSION: The PaO2/FiO2 ratio depends on both the FiO2 level and the arterial oxygen saturation level. As a minimum, the FiO2 level at which the PaO2/FiO2 ratio is measured should be defined when quantifying the effects of therapeutic interventions or when specifying diagnostic criteria for acute lung injury and acute respiratory distress syndrome. Alternatively, oxygenation problems could be described using parameters describing shunt and ventilation/perfusion mismatch.

A stochastic model of susceptibility to antibiotic therapy--The effects of cross-resistance and treatment history.

OBJECTIVE: Selection of antibiotic therapy is a complicated process, depending on, among others, the effect of cross-resistance between antibiotics. We propose a model, which incorporates information about treatment history in the form of information on the success or failure of the current treatment and which combines this with data on cross-resistance to predict the susceptibility to future antibiotic treatments, thus providing a systematic basis for revision of antibiotic treatment. METHODS AND MATERIAL: The stochastic model was built as a causal probabilistic network (CPN). Data used in the model were based on a bacteriology database including data on patient and episode unique pathogens cultured from a microbiological sample. RESULTS: In this paper, we develop a CPN that can exploit knowledge about cross-resistance between two consecutive treatments, explore the properties of this CPN and consider how the CPN can be integrated into a complete decision support system for selection of antibiotic therapy. CONCLUSION: The model presented may be useful both as a theoretical tool describing cross-resistance between antibiotics and as a part of complete decision support system for selection of antibiotic therapy.