King's College London

Research portal

Monitoring Physiological Trajectories

Research output: Other contribution

Original languageEnglish
TypePhD Transfer Report
StatePublished - 2014

Documents

  • PhD Transfer Report

    PCharlton_PhD_Transfer_Report.pdf, 2 MB, application/pdf

    3/06/2016

    Final published version

King's Authors

Impacts

  • Development of a system for automated processing of physiological monitoring data

    Impact: Health Impacts

Abstract

Clinical deteriorations of hospital patients must be recognised early to maintain patient safety and minimise treatment costs. Currently early warning scores are calculated several times each day to warn of potential deteriorations. A score is calculated using parameters measured at one particular time. In contrast, clinicians often use physiological trends over time to improve their
assessment. Therefore, we hypothesised that:

Deterioration of inpatients could be detected earlier by monitoring their physiological trajectories.

To test this hypothesis, we have constructed a database of patients’ physiology throughout their hospital stay after cardiac surgery, including continuous ECG and pulse oximetry signals. Data timestamps were misaligned during acquisition, so an algorithm has been developed to correct for this. Artefactual data was removed using signal quality indices. Initial physiological trajectories were calculated using Gaussian processes.

We have implemented algorithms to estimate respiratory rate, a key indicator of deteriorations, from these signals. We have evaluated their precision in a cohort of healthy subjects. Preliminary results suggest that they are more precise in younger subjects. Therefore, further work is required to determine whether they are sufficiently precise for use with the patient population.

We have identified indices of cardiovascular function which can be derived from these signals for detection of deteriorations. We hypothesised that the variability in cardiovascular state in the hours after surgery may indicate the class of trajectory which a patient is likely to follow. Therefore, we derived and evaluated the precision of cardiac output algorithms to determine their precision during changes in vascular state. Preliminary results suggest that more complex
algorithms are required.

Finally, we have identified the remaining steps required to test this hypothesis.

Download statistics

No data available

View graph of relations

© 2015 King's College London | Strand | London WC2R 2LS | England | United Kingdom | Tel +44 (0)20 7836 5454