Smart Strategies In Medical Emergency

Teams: A clinical perspective

Dana P. Edelson, MD, MS, FAHA, FHM

Assistant Professor, Section of Hospital Medicine

Executive Medical Director of Inpatient Quality & Safety

University of Chicago Medicine

10/13/2015 ©2015, American Heart Association 2

Smart Strategies In

Medical Emergency Teams:

A clinical perspective

Dana P. Edelson, MD, MS, FAHA

Executive Medical Director for Inpatient Quality & Safety

University of Chicago Medicine

4

Disclosures• Research funding:

– National Heart Lung Blood Institute

– Philips Healthcare

• Ownership interests:

– QuantHC

• Intellectual property over the algorithms described

• Patent pending, ARCD.P0535US.P2

5

Hospitalization Time Course

6

Three step model to improving outcomes

7

Recognition and activation are not physiologic

Galhotra, CCM, 2006

8

Adjusting to subtle changes in physiology

9

Automating recognition and activation

10

Traditional MET activation criteriaHillman, Lancet, 2005

Cardiac arrest AUC: 0.63

11

Modified Early Warning Score (MEWS)

3 2 1 0 1 2 3

Respiratory rate (RPM) — ≤ 8 — 9-14 15-20 21-29 ≥ 30

Heart rate (BPM) — ≤ 40 41-50 51-100 101-110 111-129 ≥ 130

Systolic BP ≤ 70 71-80 81-100 101-199 ≥ 200

Temperature (°C) — ≤35 — 35.0-38.4 — >38.5 —

AVPU — — — AlertReact to Voice

React to Pain

Unresp

Subbe, QJM, 2001

Cardiac arrest AUC: 0.76

12

Paper-based EWS scoring

10/13/2015 ©2013, American Heart Association

13

Leveraging electronic data

14

Deriving a computer-based early warning score

• Data set:

– Over 58,000 admissions

– 109 cardiac arrests on the ward

– 2,543 ICU transfers

• Methods:

– Multinomial logistic regression

– Competing risk model

– Longitudinal analysis

– Normal imputation

Churpek, CCM, 2014

15

eCART model for cardiac arrest Churpek, CCM, 2014

Cardiac arrest AUC:

0.88

16

Risk score change over timeCardiac arrest

ICU transfer

Control

Ca

rdia

c a

rre

st

sco

re

Time preceding event (hours)

55

45

50

40

-48 -12-24-36 0

Churpek, CCM, 2014

17

Risk Stratification using eCART

• Top 5% sickest patientsHigh

• Next 10%Moderate

Normal

eCART 54

eCART 50

18

Real-time implementation

DemographicsVital Signs Lab

Values

eCART

125 bpm

Alert120/80 Potassium

98% O2 Hemoglobin65 yo

ICU transfer

F(x)=…

19

0% 20% 40% 60% 80% 100%

Percent of Events Captured

RRT Called

ICU

Transfer

[n=383]

eCART

Cardiac

Arrest

[n=10]

(High/Mod Risk)

Silent phase comparison to standard care

20

0%

10%

20%

30%

40%

50%

60%

0612182430364248Per

cen

t o

f IC

U T

ran

sfer

s Id

enti

fied

Time Before ICU Transfer, Hour

eCART (High Risk) Standard Practice (RRT Called)

Threshold Timing for ICU Transfers

RRT median

1.7 hours

eCART (High Risk) Median

32.9 hours

Δ 31 hours (p<0.0001)

21

Mortality increases linearly with ICU transfer delay

22

In-Hospital Mortality by ICU transfer delay

p<0.001

< 6 hrs > 6 hrs

23

11 Days13 Days

Length of Stay by ICU transfer delay

p<0.001

< 6 hrs > 6 hrs

24

Real-Time Patient Dashboard

25

Real-Time Risk Trend

26

Garbage in… Tweaking future inputs

Respiratory rate

Mental status

Clinical judgment

MENTAL STATUS:

WHAT ARE THE OPTIONS?

28

AVPU scale

29

Glasgow Coma Scale

30

Richmond Agitation Sedation Scale (RASS)

31

RESPIRATORY RATE:

THE MISSING VITAL SIGN

33

Variable importance

38

40

41

43

48

51

63

66

77

100

0 10 20 30 40 50 60 70 80 90 100

GlucoseWhite blood cell count

Blood urea nitrogenTemperature

Pulse pressure indexDiastolic blood pressureSystolic blood pressure

AgeHeart rate

Respiratory rate

34

Respiratory rate is poorly recorded

Semier, Chest 2013

35

“RR” vs RR: which one is more predictive?

QUANTIFYING CLINICAL

JUDGEMENT

37

Introduction of PAR

38

The Patient Acuity Rating

• How likely is this patient to suffer a cardiac arrest or require emergent transfer to the

ICU in the next 24 hours?

1 2 3 4 5 6 7

Extremely unlikely

Extremely likely

Neither likely nor unlikely

Edelson, JHM, 2011

Cardiac arrest or ICU transfer AUC: 0.82

39

PAR sensitivities and specificities

PAR Sensitivity Specificity

7 16% 99%

≥ 6 41% 95%

≥ 5 62% 85%

≥ 4 82% 68%

≥ 3 93% 41%

≥ 2 99% 12%

≥ 1 100% 0%

Edelson, JHM, 2011

Copyright © 2015 American Medical

Association. All rights reserved.

From: The Value of Clinical Judgment in the Detection of Clinical Deterioration

JAMA Intern Med. 2015;175(3):456-458. doi:10.1001/jamainternmed.2014.7119

PAR vs MEWS: effect may be additive

RETHINKING NIGHTTIME

VITALS

From: A Prospective Study of Nighttime Vital Sign Monitoring Frequency and Risk of Clinical DeteriorationJAMA Intern Med. 2013;():-. doi:10.1001/jamainternmed.2013.7791

44

Example Clinical Decision Support Matrix by Acuity

High

• Automatic MET/RRT

• Twice daily bedside rounding with Attending

Moderate

• Bedside Rounding between RN/MD

• Proactive rounding by MET/RRT

Normal• Standard care

Low• No night-time vital signs

45

Conclusions

• Clinical deterioration on the wards is largely predictable

• Statistically derived algorithms, such as eCART, have improved accuracy over traditional

MET activation criteria

• Early identification and transfer to the ICU is associated with decreased mortality and

shorter length of stay

• Algorithms are likely to be strengthened by reliable input regarding respiratory rate, mental

status and clinical judgment

46

Questions?

10/13/2015 ©2013, American Heart Association