Identifying COPD in primary care: targeting patients at the highest risk
What is COPD?
Chronic obstructive pulmonary disease orCOPD is a long-term inflammatory diseaseof the lungs, primarily caused by cigarettesmoking.
It includes emphysema and chronicbronchitis and overlaps with asthma.
It results in increasing respiratorysymptoms such as shortness of breath andin its severe stage can require long-termoxygen therapy.
It can have a profound impact on quality oflife and often results in premature death.
What is COPD?
Chronic obstructive pulmonary disease orCOPD is a long-term inflammatory diseaseof the lungs, primarily caused by cigarettesmoking.
It includes emphysema and chronicbronchitis and overlaps with asthma.
It results in increasing respiratorysymptoms such as shortness of breath andin its severe stage can require long-termoxygen therapy.
It can have a profound impact on quality oflife and often results in premature death.
The problem
COPD is the fifth leading cause of death inthe UK.
COPD costs the NHS ~£1 billion per yearand is a major cause of hospitaladmissions.
Much of the disease burden remainsundiagnosed with an estimated 2 millionundiagnosed cases in the UK.
There is a national drive to identifypatients with undiagnosed COPD so thatthey get better access to the healthcarethey need.
The problem
COPD is the fifth leading cause of death inthe UK.
COPD costs the NHS ~£1 billion per yearand is a major cause of hospitaladmissions.
Much of the disease burden remainsundiagnosed with an estimated 2 millionundiagnosed cases in the UK.
There is a national drive to identifypatients with undiagnosed COPD so thatthey get better access to the healthcarethey need.
Current
smoker
Ex-smoker
Cough
Shortness
of breath
Wheeze
Antibiotics
Salbutamol
Prednisolone
Asthma
Chest infection
Our solution
Developed a risk prediction model that canbe used by general practitioners to helpidentify patients with undiagnosed COPD.
Compared risk factors routinely recorded inprimary care between 9397 patients withCOPD and 17829 patients without COPD.
We used a statistical method calledmultiple logistic regression to develop arisk prediction model that incorporates riskfactors significantly associated with COPD(see figure- the numbers indicate the sizeof the association between the risk factorand COPD).
We tested the accuracy of the model on aseparate group of patients consisting of4615 with COPD and 8756 without COPD.
Accuracy of the model was assessed usinga statistic called the area under thereceiver operator characteristic curve (orAUC). An AUC of 1 indicates perfectaccuracy while an AUC of 0.5 indicates anindiscriminate model.
Our model had an AUC of 0.87 whichdemonstrates a promising ability to identifypatients with undiagnosed COPD.
Our model could detect up to 80% ofpatients with undiagnosed COPD.
We aim to further develop and validate thismodel using data from a large randomisedcontrolled trial (Birmingham LungImprovement Studies, TargetCOPD)
Our solution
Developed a risk prediction model that canbe used by general practitioners to helpidentify patients with undiagnosed COPD.
Compared risk factors routinely recorded inprimary care between 9397 patients withCOPD and 17829 patients without COPD.
We used a statistical method calledmultiple logistic regression to develop arisk prediction model that incorporates riskfactors significantly associated with COPD(see figure- the numbers indicate the sizeof the association between the risk factorand COPD).
We tested the accuracy of the model on aseparate group of patients consisting of4615 with COPD and 8756 without COPD.
Accuracy of the model was assessed usinga statistic called the area under thereceiver operator characteristic curve (orAUC). An AUC of 1 indicates perfectaccuracy while an AUC of 0.5 indicates anindiscriminate model.
Our model had an AUC of 0.87 whichdemonstrates a promising ability to identifypatients with undiagnosed COPD.
Our model could detect up to 80% ofpatients with undiagnosed COPD.
We aim to further develop and validate thismodel using data from a large randomisedcontrolled trial (Birmingham LungImprovement Studies, TargetCOPD)
Application
Simple scoringsystem used by aclinician with orwithout the aid ofsoftware.
Incorporation intoprimary careinformationsystems to estimateCOPD risk scoresfor an entirepopulation.
Allow clinicians inprimary care toidentify patients athigh risk of COPDwho might benefitfrom further clinicalassessment andtreatment.
Application
Simple scoringsystem used by aclinician with orwithout the aid ofsoftware.
Incorporation intoprimary careinformationsystems to estimateCOPD risk scoresfor an entirepopulation.
Allow clinicians inprimary care toidentify patients athigh risk of COPDwho might benefitfrom further clinicalassessment andtreatment.
Shamil Haroon (s.haroon@bham.ac.uk), Rachel Jordan, Tom Marshall, Richard Riley, Robert Lancashire & Peymane Adab.
Public Health, Epidemiology & Biostatistics, University of Birmingham
