Maggie McNally, James Curtain, Kirsty O’Brien, Borislav D Dimitrov, and Tom Fahey HRB Centre for Primary Care Research Department of General Practice Royal.

Maggie McNally, James Curtain, Kirsty O’Brien, Borislav D Dimitrov, and Tom Fahey

HRB Centre for Primary Care Research

Department of General Practice

Royal College of Surgeons in Ireland

Predicting severity of pneumonia

in general practice: a meta-

analysis of the CRB-65 criteria

Outline

• What is a clinical prediction rule?

• Assessment of clinical prediction rules

• CRB-65: a clinical prediction rule

• Statistical methods in meta-analysis

• Results

• Conclusions

• Future work

Clinical Prediction Rule

• Clinical tool that quantifies contribution of:– History– Examination– Diagnostic tests

• Stratify patients according to probability of having target disorder

• Outcome can be in terms of diagnosis, prognosis, referral or treatment

Ottawa Ankle Rule: an example of a CPR

Stages of assessment of a Clinical Prediction Rule

Step 1: Derivation

identification of factors with predictive power

Step 2: Validation

evidence of reproducible accuracy

Narrow Broad

Step 3: Impact Analysis

evidence of rule changing behaviour and improving

outcome

Level of Evidence

4 3 2 1

ConfusionRespiratory rate ≥ 30/min Blood pressure (SBP≤ 90 or DBP≤60)Age ≥ 65

0 1 or 2 3 or 4

Low Risk

mortality 1.2%

Intermediate Risk

mortality 8.13%

High Risk

mortality 31%

CRB-65: a clinical prediction rule

Likely suitable for home treatment

Consider hospital referral

Urgent hospital admission

Studies included in meta-analysis (n=13)study setting participants

Barlow et al 2007 inpatients 419

Bauer et al 2006 outpatients + inpatients 1959

Bont et al 2008 outpatients 314

Buising et al 2007 emergency department 740

Capelastegui et al 2006 outpatients + inpatients 1776

Chalmers et al 2008 inpatients 1007

Kruger et al 2008 inpatients 1404

Man et al 2007 inpatients 1016

Menendez et al 2009 inpatients 447

Myint et al 2006 inpatients 192

Schaaf et al 2007 inpatients 105

Schuetz et al 2008 emergency department 373

Zuberi et al 2008 inpatients 137

TOTAL 9889

Level of evidence for CRB-65

Step 1: Derivation

identification of factors with predictive power

Step 2: Validation

evidence of reproducible accuracy

Narrow Broad

Step 3: Impact Analysis

evidence of rule changing behaviour and improving

outcome

Level of Evidence

4 3 2 1

Statistical Methods

• Derivation study used as predictive model

• Results presented as ratio measurement:

predicted deaths by CRB-65 rule

observed deaths in validation study

Results

ConfusionRespiratory rate ≥ 30/min Blood pressure (SBP≤ 90 or DBP≤60)Age ≥ 65

0 1 or 2 3 or 4

Low Risk

mortality 1.2%

Intermediate Risk

mortality 8.13%

High Risk

mortality 31%

CRB-65

RR 1.25 (CI 0.60 – 2.59)

RR 9.63 (CI 1.23 – 75.63)

n = 799

events = 0 (0%)

n = 1887

events = 14 (0.74%)

ConfusionRespiratory rate ≥ 30/min Blood pressure (SBP≤ 90 or DBP≤60)Age ≥ 65

0 1 or 2 3 or 4

Low Risk

mortality 1.2%

Intermediate Risk

mortality 8.13%

High Risk

mortality 31%

CRB-65

n = 647

events = 10 (1.5%)

n = 5674

events = 455 (8.0%)

RR 4.92 (CI 2.39 – 10.11)

RR 0.99 (CI 0.80 – 1.23)

ConfusionRespiratory rate ≥ 30/min Blood pressure (SBP≤ 90 or DBP≤60)Age ≥ 65

0 1 or 2 3 or 4

Low Risk

mortality 1.2%

Intermediate Risk

mortality 8.13%

High Risk

mortality 31%

CRB-65

n = 26

events = 5 (19.2%)

n = 869

events = 257 (29.6%)

RR 1.58 (CI 0.59 – 4.19)

RR 1.04 (CI 0.88 – 1.23)

Conclusions

ConfusionRespiratory rate ≥ 30/min Blood pressure (SBP≤ 90 or DBP≤60)Age ≥ 65

0 1 or 2 3 or 4

Low Risk

mortality 1.2%

Intermediate Risk

mortality 8.13%

High Risk

mortality 31%

Hospital Based Patients

Community Based Patients

• General trend towards over-prediction

• However,– Low cohort numbers– Low event numbers

Future Work

Step 1: Derivation

identification of factors with predictive power

Step 2: Validation

evidence of reproducible accuracy

Narrow Broad

Step 3: Impact Analysis

evidence of rule changing behaviour and improving

outcome

Level of Evidence

4 3 2 1

Acknowledgements

• RCSI Research Institute

• Grainne McCabe, RCSI Library