Discharge recommendation based on a novel technique of homeostatic analysis.

OBJECTIVE: We propose a computational framework for integrating diverse patient measurements into an aggregate health score and applying it to patient stability prediction. MATERIALS AND METHODS: We mapped retrospective patient data from the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) II clinical database into a discrete multidimensional space, which was searched for measurement combinations and trends relevant to patient outcomes of interest. Patient trajectories through this space were then used to make outcome predictions. As a case study, we built AutoTriage, a patient stability prediction tool to be used for discharge recommendation. RESULTS: AutoTriage correctly identified 3 times as many stabilizing patients as existing tools and achieved an accuracy of 92.9% (95% CI: 91.6-93.9%), while maintaining 94.5% specificity. Analysis of AutoTriage parameters revealed that interdependencies between risk factors comprised the majority of each patient stability score. DISCUSSION: AutoTriage demonstrated an improvement in the sensitivity of existing stability prediction tools, while considering patient safety upon discharge. The relative contributions of risk factors indicated that time-series trends and measurement interdependencies are most important to stability prediction. CONCLUSION: Our results motivate the application of multidimensional analysis to other clinical problems and highlight the importance of risk factor trends and interdependencies in outcome prediction.

Effect of a machine learning-based severe sepsis prediction algorithm on patient survival and hospital length of stay: a randomised clinical trial.

INTRODUCTION: Several methods have been developed to electronically monitor patients for severe sepsis, but few provide predictive capabilities to enable early intervention; furthermore, no severe sepsis prediction systems have been previously validated in a randomised study. We tested the use of a machine learning-based severe sepsis prediction system for reductions in average length of stay and in-hospital mortality rate. METHODS: We conducted a randomised controlled clinical trial at two medical-surgical intensive care units at the University of California, San Francisco Medical Center, evaluating the primary outcome of average length of stay, and secondary outcome of in-hospital mortality rate from December 2016 to February 2017. Adult patients (18+) admitted to participating units were eligible for this factorial, open-label study. Enrolled patients were assigned to a trial arm by a random allocation sequence. In the control group, only the current severe sepsis detector was used; in the experimental group, the machine learning algorithm (MLA) was also used. On receiving an alert, the care team evaluated the patient and initiated the severe sepsis bundle, if appropriate. Although participants were randomly assigned to a trial arm, group assignments were automatically revealed for any patients who received MLA alerts. RESULTS: Outcomes from 75 patients in the control and 67 patients in the experimental group were analysed. Average length of stay decreased from 13.0 days in the control to 10.3 days in the experimental group (p=0.042). In-hospital mortality decreased by 12.4 percentage points when using the MLA (p=0.018), a relative reduction of 58.0%. No adverse events were reported during this trial. CONCLUSION: The MLA was associated with improved patient outcomes. This is the first randomised controlled trial of a sepsis surveillance system to demonstrate statistically significant differences in length of stay and in-hospital mortality. TRIAL REGISTRATION: NCT03015454.

A computational approach to early sepsis detection.

OBJECTIVE: To develop high-performance early sepsis prediction technology for the general patient population. METHODS: Retrospective analysis of adult patients admitted to the intensive care unit (from the MIMIC II dataset) who were not septic at the time of admission. RESULTS: A sepsis early warning algorithm, InSight, was developed and applied to the prediction of sepsis up to three hours prior to a patient's first five hour Systemic Inflammatory Response Syndrome (SIRS) episode. When applied to a never-before-seen set of test patients, InSight predictions demonstrated a sensitivity of 0.90 (95% CI: 0.89-0.91) and a specificity of 0.81 (95% CI: 0.80-0.82), exceeding or rivaling that of existing biomarker detection methods. Across predictive times up to three hours before a sustained SIRS event, InSight maintained an average area under the ROC curve of 0.83 (95% CI: 0.80-0.86). Analysis of patient sepsis risk showed that contributions from the coevolution of multiple risk factors were more important than the contributions from isolated individual risk factors when making predictions further in advance. CONCLUSIONS: Sepsis can be predicted at least three hours in advance of onset of the first five hour SIRS episode, using only nine commonly available vital signs, with better performance than methods in standard practice today. High-order correlations of vital sign measurements are key to this prediction, which improves the likelihood of early identification of at-risk patients.

Cumulative dose of hypertension predicts outcome in intracranial hemorrhage better than American Heart Association guidelines.

BACKGROUND: Hypertension is common after intracranial hemorrhage (ICH) and may be associated with higher mortality and adverse neurologic outcome. The American Heart Association recommends that blood pressure be maintained at a mean arterial pressure (MAP) less than 130 mm Hg to prevent secondary brain injury. OBJECTIVES: To prospectively evaluate whether a new method of assessing hypertension in ICH more accurately identifies patients at risk for adverse outcomes. METHODS: The authors prospectively studied all patients presenting to two University of California, San Francisco hospitals with acute ICH from June 1, 2001, to May 31, 2004. Factors related to acute hospitalization were recorded in a database, including all charted vital signs for the first 15 days. Patients were followed up for one year, with their modified Rankin Scale (mRS) score at 12 months as primary outcome. Hypertension dose was determined as the area under the curve between patient MAP and a cut point of 110 mm Hg while in the emergency department (ED). The dose was adjusted for time spent in the ED (dose/time(ed) [d/t(ed)]). Hypertension dose was divided into four categories (none, and progressive tertiles). Multivariate logistic regression was used to calculate the odds ratio for adverse mRS by tertiles of d/t(ed). RESULTS: A total of 237 subjects with an ED average (+/-SD) length of stay of 3.42 (+/-3.7) hours were enrolled. In a multivariate logistic regression model controlling for the effects of age, volume of hemorrhage, presence of intraventricular hemorrhage, race, and preexisting hypertension, there was a 4.7- and 6.1-fold greater likelihood of an adverse neurologic outcome (by mRS) at one and 12 months, respectively, in the highest d/t(ed) tertile relative to the referent group without hypertension. CONCLUSIONS: Hypertension after acute ICH is associated with adverse neurologic outcome. The dose of hypertension may more accurately identify patients at risk for adverse outcomes than the American Heart Association guidelines and may lead to better outcomes if treated when identified in this manner.

Influence of data resolution and interpolation method on assessment of secondary brain insults in neurocritical care.

Continuous monitoring of physiologic vital signs is routine in neurocritical care. However, this patient information is usually only recorded intermittently (most often hourly) in the medical record. It is unclear whether this is sufficient to represent the occurrence of secondary brain insults (SBIs) or whether more frequent data collection will provide more comprehensive information for patient care. In 16 patients, physiologic data were acquired concurrently via two methods: per clinical routine, usually hourly, in the medical record (MR) and every minute via a custom data acquisition system (DA). SBIs were defined as a mean arterial pressure<90 mmHg, an intracranial pressure>20 mmHg or a temperature>37.5 degrees C. Number of events, cumulative duration of events and area under the curve (AUC) were compared between the two methods and 95% limits of agreement were assessed for various methods of MR data interpolation. For all three parameters, analysis of the DA and MR data frequently differed with regard to number of events, total duration of events and AUC. MR data tended to underestimate the number of total events. 95% limits of agreement were most narrow for trapezoidal interpolation of MR data, but even these limits were fairly broad. Assessment of secondary brain insults is highly dependent on (1) the temporal resolution of the method used to acquire patient data and on (2) the interpolation method if data are acquired intermittently. High frequency data acquisition may be necessary for more precise evaluation of secondary brain injury in neurocritical care.

A novel method of evaluating the impact of secondary brain insults on functional outcomes in traumatic brain-injured patients.

OBJECTIVES: Prior studies suggest that the emergency department (ED) occurrence of secondary brain insults (SBIs), such as systemic hypotension and hypoxia, worsens outcome in patients with traumatic brain injury. However, previous methods of assessing SBIs have been relatively crude, generally only determining the incidence and duration of events. The authors hypothesized that a new method that accounts for the cumulative depth and duration of SBIs would provide a more informative measure that better correlates with outcome. METHODS: The authors developed a computer algorithm to calculate the total "dose" of an SBI (in this case, hypotension and hypoxia) as the area under the curve between a cut-point value and a measured vital sign over time. To test this method, the authors used an existing data set of head trauma patients for whom occurrence in the ED of any hypotension had been shown to be associated with in-hospital mortality. The authors applied the algorithm using the cut-point values from the prior study (systolic blood pressure