drowning in data a day on the icu
play

Drowning in Data A day on the ICU Dr Dan Harvey Consultant - PowerPoint PPT Presentation

Drowning in Data A day on the ICU Dr Dan Harvey Consultant Critical Care, Nottingham University Hospitals Consultant Critical Care, Nottingham University Hospitals Hon. Associate Prof. University of Nottingham FICM Professional


  1. Drowning in Data – A day on the ICU Dr Dan Harvey Consultant Critical Care, Nottingham University Hospitals

  2. Consultant Critical Care, Nottingham University Hospitals Hon. Associate Prof. University of Nottingham FICM Professional Standards NIHR Critical Care Specialty Lead - East Midlands

  3. Objectives • What does the data look like ? • Why is it difficult for us to analyse it ? • The Result … . • The Opportunity

  4. Patients whose observations meet criteria for High Risk Red Sepsis (NICE, 2016), who have an infection* (blue), of which have an EWS of 3 or more (red). Data from 1.11.17-3.1.18. 100 90 80 70 60 50 Matched condition Yes Of which EWS >2 40 30 *Based on FreeText analysis of eHandover 20 10 0 01-Nov 03-Nov 05-Nov 07-Nov 09-Nov 11-Nov 13-Nov 15-Nov 17-Nov 19-Nov 21-Nov 23-Nov 25-Nov 27-Nov 29-Nov 01-Dec 03-Dec 05-Dec 07-Dec 09-Dec 11-Dec 13-Dec 15-Dec 17-Dec 19-Dec 21-Dec 23-Dec 25-Dec 27-Dec 29-Dec 31-Dec 02-Jan

  5. Adult emergency admission patients with initial EWS of 4 or more on arrival at hospital NUH | Observations for arrivals between 1 November 2017 and 3 January 2018 Source: Nervecentre and Medway PAS 45 40 35 30 25 20 15 Patients with EWS>=4 10 5 0

  6. Resource 1 Occ 8 Occ 15 Occ 2 Occ 9 Occ 16 Occ 3 Occ 10 Occ 17 Occ 4 Occ 11 Occ 18 Occ 5 Occ 12 Occ 19 Occ 6 Occ 13 Occ 20 Occ 7 Occ 14 Occ 21 Occ (Recovery + 2, ED Resus + 1, Theatres + 1)

  7. Planned Workload Hepatectomy – 3 hours Mr. A Cancer Colectomy – 3 hours Mr. B Cancer Laryngectomy & Flap – Mrs. C Cancer 8 hours Mr. D Spinal Fusion Cancelled x 1 Mrs. E Aortic Aneurysm Repair

  8. Mortality among Patients Admitted to Strained Intensive Care Units. Gabler et al. Am J Respir Crit Care Med. 2013 Oct 1; 188(7): 800 – 806.

  9. • Present data rather than analyses it • Fails to recognise relationships • Fails to use models for prediction • … assumes mental models & patterns are correct

  10. Could we predict this ? 70 60 50 40 30 20 10 0 Blood Sugar Insulin Rate NG Feed

  11. 14 12 10 8 6 4 2 0

  12. Time series analysis as input for clinical predictive modeling: Modeling cardiac arrest in a pediatric ICU Curtis E Kennedy & James P Turley Theoretical Biology and Medical Modelling 2011 8 :40

  13. • Harm Rx • Benefit Control • Harm • Benefit

  14. Mechanical Support Cardiovascular Respiratory Pharmacology Metabolism Cellular Function

  15. Precision digital monitors Personalised Model Model Machine Recommend Learning Action Effect

Recommend


More recommend