Association Between Urine Output and Mortality in Critically Ill Patients: A Machine Learning Approach.

OBJECTIVES: Current definitions of acute kidney injury use a urine output threshold of less than 0.5 mL/kg/hr, which have not been validated in the modern era. We aimed to determine the prognostic importance of urine output within the first 24 hours of admission to the ICU and to evaluate for variance between different admission diagnoses. DESIGN: Retrospective cohort study. SETTING: One-hundred eighty-three ICUs throughout Australia and New Zealand from 2006 to 2016. PATIENTS: Patients greater than or equal to 16 years old who were admitted with curative intent who did not regularly receive dialysis. ICU readmissions during the same hospital admission and patients transferred from an external ICU were excluded. MEASUREMENTS AND MAIN RESULTS: One hundred and sixty-one thousand nine hundred forty patients were included with a mean urine output of 1.05 mL/kg/hr and an overall in-hospital mortality of 7.8%. A urine output less than 0.47 mL/kg/hr was associated with increased unadjusted in-hospital mortality, which varied with admission diagnosis. A machine learning model (extreme gradient boosting) was trained to predict in-hospital mortality and examine interactions between urine output and survival. Low urine output was most strongly associated with mortality in postoperative cardiovascular patients, nonoperative gastrointestinal admissions, nonoperative renal/genitourinary admissions, and patients with sepsis. CONCLUSIONS: Consistent with current definitions of acute kidney injury, a urine output threshold of less than 0.5 mL/kg/hr is modestly predictive of mortality in patients admitted to the ICU. The relative importance of urine output for predicting survival varies with admission diagnosis.

Quantifying relative within-host replication fitness in influenza virus competition experiments.

Through accumulation of genetic mutations in the neuraminidase gene, the influenza virus can become resistant to antiviral drugs such as oseltamivir. Quantifying the fitness of emergent drug-resistant influenza viruses, relative to contemporary circulating viruses, provides valuable information to complement existing efforts in the surveillance of drug-resistance. We have previously developed a co-infection based method for the assessment of the relative in vivo fitness of two competing viruses. We have also introduced a model of within-host co-infection dynamics that enables relative within-host fitness to be quantified in these competitive-mixtures experiments. The model assumed that fitness differences between co-infecting strains were mediated by strain-dependent viral production rates from infected epithelial cells. Here we extend the model to enable a more complete exploration of biological processes that may differ between virus pairs and hence generate fitness differences. We use the extended model to re-analyse data from competitive-mixtures experiments that investigated the fitness of oseltamivir-resistant (OR) H1N1 pandemic 2009 ("H1N1pdm09") viruses that emerged during a community outbreak in Australia in 2011. Results are consistent with those of our previous analysis, suggesting that the within-host replication fitness of these OR viruses is not compromised relative to that of related oseltamivir-susceptible (OS) strains, and that potentially permissive mutations in the neuraminidase gene (V241I and N369K) significantly enhance the fitness of H1N1pdm09 OR viruses. These results are consistent regardless of the hypothesised biological cause of fitness difference.

Reducing uncertainty in within-host parameter estimates of influenza infection by measuring both infectious and total viral load.

For in vivo studies of influenza dynamics where within-host measurements are fit with a mathematical model, infectivity assays (e.g. 50% tissue culture infectious dose; TCID50) are often used to estimate the infectious virion concentration over time. Less frequently, measurements of the total (infectious and non-infectious) viral particle concentration (obtained using real-time reverse transcription-polymerase chain reaction; rRT-PCR) have been used as an alternative to infectivity assays. We investigated the degree to which measuring both infectious (via TCID50) and total (via rRT-PCR) viral load allows within-host model parameters to be estimated with greater consistency and reduced uncertainty, compared with fitting to TCID50 data alone. We applied our models to viral load data from an experimental ferret infection study. Best-fit parameter estimates for the "dual-measurement" model are similar to those from the TCID50-only model, with greater consistency in best-fit estimates across different experiments, as well as reduced uncertainty in some parameter estimates. Our results also highlight how variation in TCID50 assay sensitivity and calibration may hinder model interpretation, as some parameter estimates systematically vary with known uncontrolled variations in the assay. Our techniques may aid in drawing stronger quantitative inferences from in vivo studies of influenza virus dynamics.