Search

Objectives: Develop and implement a machine learning algorithm to predict severe sepsis and septic shock and evaluate the impact on clinical practice and patient outcomes., Design: Retrospective cohort for algorithm derivation and validation, pre-post impact evaluation., Setting: Tertiary teaching hospital system in Philadelphia, PA., Patients: All non-ICU admissions; algorithm derivation July 2011 to June 2014 (n = 162,212); algorithm validation October to December 2015 (n = 10,448); silent versus alert comparison January 2016 to February 2017 (silent n = 22,280; alert n = 32,184)., Interventions: A random-forest classifier, derived and validated using electronic health record data, was deployed both silently and later with an alert to notify clinical teams of sepsis prediction., Measurement and Main Result: Patients identified for training the algorithm were required to have International Classification of Diseases, 9th Edition codes for severe sepsis or septic shock and a positive blood culture during their hospital encounter with either a lactate greater than 2.2 mmol/L or a systolic blood pressure less than 90 mm Hg. The algorithm demonstrated a sensitivity of 26% and specificity of 98%, with a positive predictive value of 29% and positive likelihood ratio of 13. The alert resulted in a small statistically significant increase in lactate testing and IV fluid administration. There was no significant difference in mortality, discharge disposition, or transfer to ICU, although there was a reduction in time-to-ICU transfer., Conclusions: Our machine learning algorithm can predict, with low sensitivity but high specificity, the impending occurrence of severe sepsis and septic shock. Algorithm-generated predictive alerts modestly impacted clinical measures. Next steps include describing clinical perception of this tool and optimizing algorithm design and delivery.

Background: On Tuesday, January 12, 2010 at 16:53 local time, a magnitude 7.0 M(w) earthquake struck Haiti. The global humanitarian attempt to respond was swift, but poor infrastructure and emergency preparedness limited many efforts. Rapid, successful deployment of emergency medical care teams was accomplished by organizations with experience in mass disaster casualty response. Well-intentioned, but unprepared, medical teams also responded. In this report, we describe the preparation and planning process used at an academic university department of anesthesiology with no preexisting international disaster response program, after a call from an American-based nongovernmental organization operating in Haiti requested medical support. The focus of this article is the pre-deployment readiness process, and is not a post-deployment report describing the medical care provided in Haiti., Methods: A real-time qualitative assessment and systematic review of the Hospital of the University of Pennsylvania's communications and actions relevant to the Haiti earthquake were performed. Team meetings, conference calls, and electronic mail communication pertaining to planning, decision support, equipment procurement, and actions and steps up to the day of deployment were reviewed and abstracted. Timing of key events was compiled and a response timeline for this process was developed. Interviews with returning anesthesiology members were conducted., Results: Four days after the Haiti earthquake, Partners in Health, a nonprofit, nongovernmental organization based in Boston, Massachusetts, with >20 years of experience providing medical care in Haiti contacted the University of Pennsylvania Health System to request medical team support. The departments of anesthesiology, surgery, orthopedics, and nursing responded to this request with a volunteer selection process, vaccination program, and systematic development of equipment lists. World Health Organization and Centers for Disease Control guidelines, the American Society of Anesthesiology Committee on Trauma and Emergency Preparedness, published articles, and in-country contacts were used to guide the preparatory process., Conclusion: An organized strategic response to medical needs after an international natural disaster emergency can be accomplished safely and effectively within 6 to 12 days by an academic anesthesiology department, with medical system support, in a center with no previously established response system. The value and timeliness of this response will be determined with further study. Institutions with limited experience in putting an emergency medical team into the field may be able to quickly do so when such efforts are executed in a systematic manner in coordination with a health care organization that already has support infrastructure at the site of the disaster.

This prospective study examined whether the integration of acute care nurse practitioners (ACNP) in a "semiclosed" surgical intensive care unit (SICU) model increased compliance with clinical practice guidelines (CPG). Patients were admitted to critical care services with a (a) "semiclosed"/ACNP team or (b) "mandatory consultation"/non-ACNP team. CPG compliance was significantly higher (P < .05) on the "semiclosed"/ACNP team for all 3 CPGs examined in the study.

Background: The authors performed a prospective study to determine whether breath test analysis using an electronic nose correlates with a clinical pneumonia score., Methods: Exhaled gas was sampled from the expiratory limb of the ventilator in mechanically ventilated surgical intensive care patients and assayed with the electronic nose. Components of a clinical pneumonia score were recorded concurrently., Results: The score predicted by the electronic nose showed good correlation with the actual pneumonia score (r = 0.81). Bland Altman analysis showed a mean bias of 0.0 (limits +/- 2.6)., Conclusions: The electronic nose is a new biosensor technology that correlates with a clinical pneumonia score.

Unlabelled: The purpose of this study was to determine if rates of telemetry events differ between patients whose monitoring is appropriately "indicated" versus "not indicated" by systematically applying rigorous criteria for appropriateness of electrocardiogram (ECG) telemetry usage. We performed a retrospective cohort study on 1097 telemetry admissions between January 1, 2000 and March 31, 2000. A convenience sample of 218 patients generated 236 telemetry admissions. One-hundred-sixty-two arrhythmic events were detected during 400 "indicated" telemetry days. Nine arrhythmic events were detected during 345 "not indicated" telemetry days. The relative rate for arrhythmic events was significantly different, at P < 0.0001, with the incidence rate ratio of 15 indicating a very large effect size. Consequently, current use of ECG telemetry may not be optimal, and a prospective analysis of the application of rigorous indications for ECG telemetry needs to be undertaken., Implications: The application of standard criteria to electrocardiogram telemetry admissions found that the majority of abnormal heart rhythms were found when patients met appropriate criteria.

Objective: To review the history and current applications of artificial intelligence in the intensive care unit., Data Sources: The MEDLINE database, bibliographies of selected articles, and current texts on the subject., Study Selection: The studies that were selected for review used artificial intelligence tools for a variety of intensive care applications, including direct patient care and retrospective database analysis., Data Extraction: All literature relevant to the topic was reviewed., Data Synthesis: Although some of the earliest artificial intelligence (AI) applications were medically oriented, AI has not been widely accepted in medicine. Despite this, patient demographic, clinical, and billing data are increasingly available in an electronic format and therefore susceptible to analysis by intelligent software. Individual AI tools are specifically suited to different tasks, such as waveform analysis or device control., Conclusions: The intensive care environment is particularly suited to the implementation of AI tools because of the wealth of available data and the inherent opportunities for increased efficiency in inpatient care. A variety of new AI tools have become available in recent years that can function as intelligent assistants to clinicians, constantly monitoring electronic data streams for important trends, or adjusting the settings of bedside devices. The integration of these tools into the intensive care unit can be expected to reduce costs and improve patient outcomes.

Objective: To determine whether the presence of an on-site, organized, supervised critical service improves care and decreases resource utilization., Design: The study compared two patient cohorts admitted to a surgical intensive care unit during the same period of time. The study cohort was cared for by an on-site critical care team supervised by an intensivist. The control cohort was cared for by a team with patient care responsibilities in multiple sites supervised by a general surgeon. The main outcome measures were duration of stay, resource utilization, and complication rate., Setting: Study patients were general surgical patients in an academic medical center., Results: Despite having higher Acute Physiology and Chronic Health Evaluation II scores, patients cared for by the critical care service spent less time in the surgical intensive care unit, used fewer resources, had fewer complications and had lower total hospital charges. The difference between the two cohorts was most evident in patients with the worst APACHE II score., Conclusions: Critical care interventions are expensive and have a narrow safety margin. It is essential to develop structured and validated approaches to study the delivery of this resource. In this study, the critical care service model performed favorably both in terms of quality and cost.

Objective: To validate a new system of continuous cardiac output monitoring., Design: Multicenter, prospective, nonrandomized clinical study., Setting: Four university hospitals., Patients: Forty-seven adult intensive care unit patients., Interventions: Pulmonary artery catheterization., Measurements and Main Results: Continuous and bolus cardiac output measurements were obtained over 72 hrs. The 327 continuous cardiac output measurements compared favorably with bolus cardiac output measurements (bias = 0.12 L/min, precision = +/-0.84). The continuous cardiac measurement was not adversely affected by temperatures of <37 degrees C or >38 degrees C, high (>7.5 L/min) or low (<4.5 L/min) cardiac output values, or duration (72 hrs) of the study., Conclusions: This continuous cardiac output system provides a reliable estimate of cardiac output for clinical use if applied in conditions similar to this study. The combination of a continuous measure of cardiac output with other continuous physiologic monitoring (arterial and mixed venous oxygen saturation, oxygen consumption, etc.) may provide important information that no single parameter could achieve.

Metabolic acidosis is a pathophysiologic state that is associated with serious morbidities and mortality. The diagnosis of metabolic acidosis is perplexing for novice and expert advanced practice nurses for many reasons. Its differential diagnosis is broad and includes common and rare, complex disease. The diagnosis of metabolic acidosis is also difficult because it is frequently associated with mixed, acid-base disorders. Its clinical manifestations are often nonspecific or subclinical, which means that its diagnosis is made from laboratory and other diagnostic tests. Timely diagnosis of metabolic acidosis is needed to institute appropriate therapy to avoid negative physiologic effects.

Background: Inhaled nitric oxide (NO), a selective vasodilator, improves oxygenation in many patients with adult respiratory distress syndrome (ARDS). Vasoconstrictors may also improve oxygenation, possibly by enhancing hypoxic pulmonary vasoconstriction. This study compared the effects of phenylephrine, NO, and their combination in patients with ARDS., Methods: Twelve patients with ARDS (PaO2/FIO2 180; Murray score 2) were studied. Each patient received three treatments in random order: intravenous phenylephrine, 50-200 micrograms/min, titrated to a 20% increase in mean arterial blood pressure; inhaled NO, 40 ppm; and the combination (phenylephrine+NO). Hemodynamics and blood gas measurements were made during each treatment and at pre- and posttreatment baselines., Results: All three treatments improved PaO2 overall. Six patients were "phenylephrine-responders" (delta PaO2 > 10 mmHg), and six were "phenylephrine-nonresponders." In phenylephrine-responders, the effect of phenylephrine was comparable with that of NO (PaO2 from 105 +/- 14 to 132 +/- 14 mmHg with phenylephrine, and from 110 +/- 14 to 143 +/- 19 mmHg with NO), and the effect of phenylephrine+NO was greater than that of either treatment alone (PaO2 from 123 +/- 13 to 178 +/- 23 mmHg). In phenylephrine-nonresponders, phenylephrine did not affect PaO2, and the effect of phenylephrine+NO was not statistically different from that of NO alone (PaO2 from 82 +/- 12 to 138 +/- 28 mmHg with NO; from 84 +/- 12 to 127 +/- 23 mmHg with phenylephrine+NO). Data are mean +/- SEM., Conclusions: Phenylephrine alone can improve PaO2 in patients with ARDS. In phenylephrine-responsive patients, phenylephrine augments the improvement in PaO2 seen with inhaled NO. These results may reflect selective enhancement of hypoxic pulmonary vasoconstriction by phenylephrine, which complements selective vasodilation by NO.

Objectives: To compare data derived from a computer model of the pulmonary circulation with data from a case series of patients with chronic obstructive pulmonary disease (COPD). To evaluate the specific factors contributing to CO2 retention due to oxygen therapy in patients with acute exacerbations of COPD., Design: Data from a computer model of the pulmonary circulation were compared with a previous case series., Patients: Patient data were derived from previous case series., Interventions: Simulated application of oxygen therapy., Measurements and Main Results: The computer model of the pulmonary circulation generates data comparable with those data from a series of patients with COPD treated with supplemental oxygen and permits identification of the causes for hypercarbia. Therapy with supplemental oxygen alters hypoxic pulmonary vasoconstriction and modulates the Haldane effect, resulting in changes in physiologic deadspace., Conclusion: Changes in physiologic deadspace are sufficient to account for the hypercarbia developed by patients with acute exacerbations of COPD when treated with supplemental oxygen.