2. The 3C cohort study of LRTI in primary care
Predicting adverse outcome from lower respiratory tract infection in primary care:
The 3C cohort study of LRTI in primary care
Moore, M.
Little, P., Stuart B, Smith S, Thompson MJ, Knox K, van den Bruel A, Lown,M., Mant,D
01 June 2015

3. Family Practice in Salisbury
University of Southampton
No financial or other conflicts of interest

4. Where?

5. Overview
LRTI- why prescribe?
Methods
Results
Clinical Implications

6. Background Lower Respiratory Tract Infection LRTI: Is common
Antibiotics often prescribed
NAPCRG 2016

7. Respiratory infection account for the majority of outpatient prescribing
UK database data 568 practices
Median prescribing rates -48% for ‘cough and bronchitis’ -60% for ‘sore throat’ -60% for ‘otitis-media’
Median for RTI 54% (39-69%) lowest/highest

8. Background Drivers of prescribing in LRTI: Relief of symptoms?
Worries about pneumonia?
Worries about adverse outcome?

9. Background- symptom relief?
Cochrane review acute bronchitis -17 trials 2936 participants
No difference in clinical improvement
Some reduction in cough and night cough
Modest reduction in cough duration (-0.46 days)
Cochrane library 2014

10. Background- symptom relief?
GRACE study whole population n=2061

11. Background- symptom relief?
GRACE study age over 60
NAPCRG 2016

12. Background- symptom relief?
GRACE study: Green sputum
NAPCRG 2016

13. Background- symptom relief?
GRACE study: Smokers
NAPCRG 2016

14. Background- symptom relief?
GRACE study: Missed radiological pneumonia

15. Background- symptom relief?
In Summary: For symptoms:
Overall cochrane review modest treatment effect
GRACE study subgroups: Only those with missed radiological pneumonia derived significant benefit
Worth thinking about the diagnosis of pneumonia

16. Background- worry about adverse outcome?
Re-consultation with new or worsening illness
Admission
Death
Late onset pneumonia

17. Methods
Adults presenting in UK general practice with LRTI had symptoms, signs, and antibiotic prescribing strategies recorded.
Re-consultation with new or non–resolving symptoms, or hospitalisation or death, were documented within 30 days.
NAPCRG 2016

18. Methods
Inclusion Patients aged 16 or over presenting in UK general practice with acute LRTI. We used a pragmatic: cough as the main symptom, judged to be infective in origin by the GP.
Exclusion: other cause of acute cough (e.g. heart failure, acid reflux, fibrosing alveolitis etc); patients unable to fill out the diary; immune compromised; previously presented with the same episode of illness

19. Results
28867 adult patients with acute cough were recruited with informed consent by 522 general practices in the UK General Practice Research Network between October 2009 and April 2013

20. Baseline characteristics
Age > %
Female 59%
Co-morbidity 30%
Hx of fever 38%
Chills 32%
Fever > %
Sats <95% 6%
Low BP 8%

21. Results -28867 37 same day admission to hospital
6484 patients re-consulted within 30 days
258 hospitalisations in total
720 were referred for a chest x-ray in the first week
30 patients died
7349 (25%) no antibiotic (61%) immediate antibiotic (13%) delayed antibiotic prescription

22. Results- re-consultation
Based on notes review including all settings
6,484 re-consultations
268 of these were adverse reactions
More severe symptoms at baseline- earlier re-consultation
NAPCRG 2016

23. Results- re-consultation

24. Results- re-consultation
Factors predicting re-consultation (No adjusted RR >1.5)
Age >60, Female, Pneumovax
Lung co-morbidity, on steroid/inhaler
SOB, Fever, Chest pain, coryza (-) Muscle aches, rusty sputum (-)
Severity assessment, RR, Fever, pulse>100, low sats, crackles, wheeze, bronchial breathing
SAPC 2016

25. Results- re-consultation
Multi-varate model factors predicting re-consultation/death:
There are 10 variables that predict hospitalisation or death with a RR of 1.5 or higher
Age 60+, comorbidity, shortness of breath, chest pain, crackles, higher severity score, high pulse, high temperature, low oxygen saturation and low blood pressure
SAPC 2016

26. - hospitalisation death
Results
- hospitalisation death
258 hospitalisations -37 same day
30 deaths, of which
15 deaths in patients not hospitalised
None of those admitted on same day died
Total of 273 hospitalisations or death.
Case review
233 relevant hospitalisations
13 relevant deaths
Total of 245 relevant admissions or deaths
NAPCRG 2016

27. Results timing of admission

28. Results- pneumonia
In the cohort of 28,883 participants, there are 1782 chest xrays of whom 720 x-rayed within the first 7 days. The results of these are:
Outcome
114/720 (16%) pneumonia in first 7 days
111/1062 (10%) late diagnosis pneumonia
11 clinical diagnosis without xray
NAPCRG 2016

29. Results timing of x-ray diagnosis pneumonia

30. Results- to summarise
Risk of serious adverse outcome after presentation with LRTI is low in this cohort of 28883
6484 (22%) re-consultation
230 (0.8%) x-ray pneumonia
Admissions 258, 234 (0.8%) related
Deaths 30, 13 related (0.04%)
SAPC 2016

31. Results- modelling
Is it possible to predict those at risk of serious adverse outcome (admission death)
For the clinician- Well if I don’t decide to admit you today- how can I tell who might do badly?
273 hospitalisations or deaths
Exclude those - admitted on the day -unrelated to index consultation
122 late diagnosis pneumonia (clinical +xray)

32. Results- modelling
Risk factors at first consultation for death or hospitalization from LRTI complications within 30 days or late-onset or unresolved pneumonia confirmed by x-ray or re-consultation 8-30 days after first consultation (n=325)
Analyses controlled for antibiotics at index consultation
Prior probability of serious adverse outcome
325/28830= (1.1%)

33. Patient characteristics

34. Presenting Symptoms

35. Examination findings

36. Results- clinical score
Items carried forward significant at the 1% level Nine items combined into a total score which ranges from 0 (none of these) to 9 (all of these).
The AUROC of this score is (Bootstrapped 95% CI 0.69, 0.75).

37. Results- clinical score
Risk Ratio (95% CI)
p-value
O2 sat < 95%
2.37 (1.80, 3.12)
<0.001
Age 60+ years
2.02 (1.59, 2.57)
SBP< 90 or DBP < 60 mmHg
1.65 (1.16, 2.33)
0.005
Temp > 37.8°C
1.81 (1.29, 2.55)
0.001
Any co-morbidity
1.61 (1.23, 2.11)
Chills
1.37 (1.07, 1.74)
0.011
No coryza
1.49 (1.16, 1.91)
0.002
Sputum: purulent
0.74 (0.58, 0.95)
0.015
Severity assessment > 5/10
1.51 (1.15, 1.97)
0.003
Crackles
1.30 (0.95, 1.78)
0.098
Shortness of breath
1.32 (0.98, 1.79)
0.070
Chest pain
Headache
0.81 (0.62, 1.06)
0.119
NAPCRG 2016

38. Results- clinical score
Risk Ratio (95% CI)
p-value
O2 sat < 95%
2.37 (1.80, 3.12)
<0.001
Age 60+ years
2.02 (1.59, 2.57)
SBP< 90 or DBP < 60 mmHg
1.65 (1.16, 2.33)
0.005
Temp > 37.8°C
1.81 (1.29, 2.55)
0.001
Any co-morbidity
1.61 (1.23, 2.11)
Chills
1.37 (1.07, 1.74)
0.011
No coryza
1.49 (1.16, 1.91)
0.002
Sputum: purulent
0.74 (0.58, 0.95)
0.015
Severity assessment > 5/10
1.51 (1.15, 1.97)
0.003
Crackles
1.30 (0.95, 1.78)
0.098
Shortness of breath
1.32 (0.98, 1.79)
0.070
Chest pain
Headache
0.81 (0.62, 1.06)
0.119

39. Score Items Age 60+, Co-morbidity No coryza
History of chills/shivering
Presence of chest pain
Severity score 6 or over (out of 10)
Low BP (systolic <90 diastolic <60)
Temp>37.8
O2 saturation <95% ,

40. Distribution of clinical score
N (%) of total cohort with each score
None
1293 (5.5%) 1
4071 (17.3%) 2
6456 (27.4%) 3
6119 (25.9%) 4
3693 (15.7%) 5
1444 (6.1%) 6
435 (1.8%) 7
80 (0.3%) 8
9 (0.04%) 9
1 (0.00%)

41. Predictive value of clinical score
Cut off score to use
N (%) of total cohort
Sensitivity
Specificity
NPV
PPV
1 or more
22,308 (94.5%)
99.2%
5.3%
99.8%
1.2%
2 or more
18,237 (77.3%)
96.2%
22.9%
1.4%
3 or more
11,781 (49.9%)
82.4%
50.4%
99.6%
1.8%
4 or more
5,662 (24.0%)
53.4%
76.3%
99.3%
2.5%
5 or more
1,969 (8.3%)
27.1%
91.9%
99.1%
3.6%
6 or more
525 (2.2%)
11.8%
97.9%
99.0%
5.9%
7 or more 90
(0.4%)
3.8%
99.7%
98.9%
11.1%

42. Applying in practice?
A score of 3 or less (76% of population) NPV 99.6% No treatment
A score of 4 (16% of population) PPV 2.5% NPV 99.3% Delayed?
A score of 5 or more 8.3% of population PPV 3.6% Immediate prescription

43. Other scores Some common factors with other risk scores
Fever, Low Sats (CCC cohort pneumonia score)
Age, Blood Pressure (CRB 65 mortality risk score)
Absence of coryza and temp>37.8 (GRACE pneumonia model)
Low sats might be a proxy for SOB in GRACE score
NAPCRG 2016

44. Strengths and Limitations
Observational data residual confounding
Large numbers of patients in real life practice
Over fitting of model
No validation sample

45. Implications for Practice
A clinical score can be used to predict the risk of hospitalisation/death/late onset pneumonia
Although complex (9 item) it could be used to direct antibiotic strategy

46. Wrapping up Overall little symptomatic benefit from antibiotics
It is worth spotting ‘missed pneumonia’
Adverse outcomes are rare after LRTI
Use a score to identify those at low risk of adverse outcome
This low risk group have little or nothing to gain from antibiotics

47. Any Questions?