Dataset Information

Development and validation of a reinforcement learning algorithm to dynamically optimize mechanical ventilation in critical care.

ABSTRACT: The aim of this work was to develop and evaluate the reinforcement learning algorithm VentAI, which is able to suggest a dynamically optimized mechanical ventilation regime for critically-ill patients. We built, validated and tested its performance on 11,943 events of volume-controlled mechanical ventilation derived from 61,532 distinct ICU admissions and tested it on an independent, secondary dataset (200,859 ICU stays; 25,086 mechanical ventilation events). A patient "data fingerprint" of 44 features was extracted as multidimensional time series in 4-hour time steps. We used a Markov decision process, including a reward system and a Q-learning approach, to find the optimized settings for positive end-expiratory pressure (PEEP), fraction of inspired oxygen (FiO₂) and ideal body weight-adjusted tidal volume (Vt). The observed outcome was in-hospital or 90-day mortality. VentAI reached a significantly increased estimated performance return of 83.3 (primary dataset) and 84.1 (secondary dataset) compared to physicians' standard clinical care (51.1). The number of recommended action changes per mechanically ventilated patient constantly exceeded those of the clinicians. VentAI chose 202.9% more frequently ventilation regimes with lower Vt (5-7.5?mL/kg), but 50.8% less for regimes with higher Vt (7.5-10?mL/kg). VentAI recommended 29.3% more frequently PEEP levels of 5-7?cm H₂O and 53.6% more frequently PEEP levels of 7-9 cmH₂O. VentAI avoided high (>55%) FiO₂ values (59.8% decrease), while preferring the range of 50-55% (140.3% increase). In conclusion, VentAI provides reproducible high performance by dynamically choosing an optimized, individualized ventilation strategy and thus might be of benefit for critically ill patients.

SUBMITTER: Peine A

PROVIDER: S-EPMC7895944 | biostudies-literature | 2021 Feb

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Development and validation of a reinforcement learning algorithm to dynamically optimize mechanical ventilation in critical care.

Peine Arne A Hallawa Ahmed A Bickenbach Johannes J Dartmann Guido G Fazlic Lejla Begic LB Schmeink Anke A Ascheid Gerd G Thiemermann Christoph C Schuppert Andreas A Kindle Ryan R Celi Leo L Marx Gernot G Martin Lukas L

NPJ digital medicine 20210219 1

The aim of this work was to develop and evaluate the reinforcement learning algorithm VentAI, which is able to suggest a dynamically optimized mechanical ventilation regime for critically-ill patients. We built, validated and tested its performance on 11,943 events of volume-controlled mechanical ventilation derived from 61,532 distinct ICU admissions and tested it on an independent, secondary dataset (200,859 ICU stays; 25,086 mechanical ventilation events). A patient "data fingerprint" of 44 f ...[more]

PMID: 33608661

Similar Datasets

Project description:ObjectivesTo assess the reliability of codes relevant to critically ill patients in administrative data.DesignRetrospective cohort study linking data from Acute Physiology and Chronic Health Evaluation Outcomes, a clinical database of ICU patients with data from Medicare Provider Analysis and Review. We linked data based on matching for sex, date of birth, hospital, and date of admission to hospital.SettingForty-six hospitals in the United States participating in Acute Physiology and Chronic Health Evaluation Outcomes.PatientsAll patients in Acute Physiology and Chronic Health Evaluation Outcomes greater than or equal to 65 years old who could be linked with hospitalization records in Medicare Provider Analysis and Review from January 1, 2009, through September 30, 2012.Measurements and main resultsOf 62,451 patients in the Acute Physiology and Chronic Health Evaluation Outcomes dataset, 80.1% were matched with data in Medicare Provider Analysis and Review. All but 2.7% of Acute Physiology and Chronic Health Evaluation Outcomes ICU patients had either an ICU or coronary care unit charge in Medicare Provider Analysis and Review. In Acute Physiology and Chronic Health Evaluation Outcomes, 37.0% received mechanical ventilation during the ICU stay versus 24.1% in Medicare Provider Analysis and Review. The Medicare Provider Analysis and Review procedure codes for mechanical ventilation had high specificity (96.0%; 95% CI, 95.8-96.2), but only moderate sensitivity (58.4%; 95% CI, 57.7-59.1), with a positive predictive value of 89.6% (95% CI, 89.1-90.1) and negative predictive value of 79.7% (95% CI, 79.4-80.1). For patients with mechanical ventilation codes, Medicare Provider Analysis and Review overestimated the percentage with a duration greater than 96 hours (36.6% vs 27.3% in Acute Physiology and Chronic Health Evaluation Outcomes). There was discordance in the hospital discharge status (alive or dead) for only 0.47% of all linked records (κ = 1.00).ConclusionsMedicare Provider Analysis and Review data contain robust information on hospital mortality for patients admitted to the ICU but have limited ability to identify all patients who received mechanical ventilation during a critical illness. Estimates of use of mechanical ventilation in the United States should likely be revised upward.

Dataset Information

Development and validation of a reinforcement learning algorithm to dynamically optimize mechanical ventilation in critical care.

Publications

Development and validation of a reinforcement learning algorithm to dynamically optimize mechanical ventilation in critical care.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets