Saskia van Vught
Thursday 10 January 2013
Caring for cough: guiding diagnosis in primary care
Promotor: Prof. dr. Theo Verheij
Defence: 10 January 2013
Acute cough is a very common reason for primary care visits with a broad differential diagnosis. Among the most common causes are influenza, other viral or bacterial bronchial infections, pneumonia and exacerbations of asthma or COPD, which are often grouped together in one disease entity: lower respiratory tract infection (LRTI). In the majority of patients the disease course is self limiting and the GP should inform patients about this. For such wait-and-see policy diagnostic strategies are needed that help to safely exclude the specific conditions warranting targeted therapy. Due to methodological considerations however, the (additional) diagnostic value of several diagnostic tools remains largely unknown. Therefore we aimed to evaluate various commonly used diagnostic tools in patients who present with acute cough or LRTI-like symptoms to primary care in order to guide general practioners (GPs) in their diagnostic decisions.
Data from the observational GRACE-09 project were used for the studies described in this thesis. Between October 2007 and April 2010 participating GPs from 12 European countries recruited 3106 patients who presented with acute cough or a presentation suggestive of LRTI.
Chapter 2 describes the results of a diagnostic study that determined the accuracy of GPs’ clinical judgment based on signs and symptoms to detect radiographic pneumonia in LRTI patients. Immediately after history taking and physical examination, GPs registrated whether they considered pneumonia present. Chest radiography was performed within one week and served as reference standard. A total of 140 out of 2810 patients had radiographic pneumonia (5%), of whom 41 (29%) were clinically identified. GPs diagnosed pneumonia in 31 (1%) patients without radiographic pneumonia. Sensitivity, specificity, positive and negative predictive value of GPs clinical judgment were 29%, 99%, 57% and 96% respectively. Prevalence of ‘classical’ pneumonia symptoms and patients’ symptom severity scores were lower in clinically unrecognised radiographic pneumonia (p<0.05). We concluded that although most cases of radiographic pneumonia were not diagnosed based on clinical judgment alone, the majority of these clinically unidentified radiographic pneumonias showed mild clinical presentation.
In Chapter 3, a cross sectional diagnostic study is presented that assessed the accuracy of previously published diagnostic models for pneumonia in patients who present with acute cough. Secondary aim was to quantify the added value of the inflammatory markers C-reactive protein (CRP) and procalcitonin (PCT) above history and physical examination findings. After consultation, 3106 eligible patients underwent measurement of CRP and PCT. Within 7 days chest radiography was performed and assessed by local radiologists (reference test for pneumonia). After exclusion of 286 patients for missing chest radiograph results, 140 out of 2820 patients (5%) had radiographic pneumonia. Six published ‘symptoms and signs models’ showed moderate discrimination (Receiver Operating Curve (ROC) areas ranging from 0.55 (95% CI 0.50-0.61) to 0.71 (95% CI 0.66-0.76)). The optimal combination of items included: absence of runny nose, breathlessness, crackles and diminished breath sounds on auscultation, tachycardia, and fever, with an ROC area of 0.70 (95% CI 0.65-0.75) after internal validation. Adding CRP increased the ROC area to 0.79 (95% CI 0.75-0.83) with a net reclassification index of 24%. PCT had no additional diagnostic value. We concluded that symptoms and signs had moderate clinical value in discriminating between high and low risk for pneumonia in patients presenting with acute cough. The added value of CRP measurement is merely in excluding pneumonia while PCT adds no clinically relevant information.
Chapter 4 addresses the type and prevalence of incidental chest radiography findings in primary care patients with acute cough who underwent chest radiography as part of a research study work up. Apart from signs of pneumonia and bronchitis, local radiologists were asked to evaluate ‘other’ findings on the radiographs. In the 2823 patients with good quality chest radiographs available, these findings were categorised according to clinical relevance based on literature. In addition they were analysed for type and prevalence by primary care network (PCN), gender, age and smoking behaviour. Incidental findings were recorded for 19% of all patients, depending on the PCN of inclusion (p<0.001). Findings which would require further diagnostic procedures were reported in 3%, with 1.8% being suspected nodules and shadows. Incidental findings were more commonly reported for older and smoking patients (p<0.001). We concluded that the incidence of clinically relevant incidental findings in chest radiographs is low in adult patients with acute cough, with a proportion depending on PCN.
In Chapter 5 two observational studies on undetected chronic obstructive lung disorders are presented. Chapter 5.1 describes the prevalence of airway obstruction and bronchodilator responsiveness in adults presenting with acute cough in primary care. Patients known to have physician diagnosed asthma or chronic obstructive pulmonary disease (COPD) were excluded. Complete case analysis of spirometry results (28-35 days after inclusion) was performed in 1947 patients who consulted with acute cough. Bronchodilator responsiveness was diagnosed if there were recurrent airway complaints AND an increase of the forced expiratory volume in one second (FEV1) ≥12% after bronchodilation. Airway obstruction was diagnosed according to two thresholds for the (post bronchodilator) ratio of the FEV1 to the forced vital capacity (FEV1/FVC ratio): below 0.7 and below the lower limit of normal (LLN). There were 240 patients who showed bronchodilator responsiveness (12%). In 193 patients (10%) the FEV1/FVC ratio was below 0.7, and in 126 (7%) below LLN. Spearman’s correlation between the two definitions of obstruction was 0.71 (p<0.001). Discordance was most pronounced among patients under 30 and in older patients. We concluded that both bronchodilator responsiveness and persistent airway obstruction are common in adults without established asthma or COPD who present with acute cough, which suggests undiagnosed asthma and COPD. As both conditions benefit from appropriate and timely interventions, GPs should be aware and responsive to this under-diagnosis. Therefore, in Chapter 5.2 the diagnostic value of two previously developed diagnostic models for identifying asthma and COPD in persistent cough was determined in our patients with acute cough. Between day 28 and 35 after inclusion, 2532 adult patients without established asthma or COPD underwent spirometry. Asthma was defined by recurrent wheezing, cough or dyspnoea and reversibility of FEV1. COPD was defined by spirometric obstruction. The existing diagnostic models showed moderate discrimination (ROC area 0.56 (95%CI 0.53-0.59) and 0.64 (95%CI 0.58-0.69), respectively) and poor calibration (p<0.001). In addition, the diagnostic value of signs and symptoms for asthma and COPD in patients with acute cough was determined. Items with highest diagnostic value were age, ≥20 pack years of smoking, a history of wheezing, nocturnal cough and ≥2 other cough episodes in the past year. We concluded that it is difficult to identify patients with a substantially increased risk of chronic obstructive lung disorders among patients who present with acute cough, because the prevalence of the strongest diagnostic signs and symptoms was low. Patients have to be instructed to revisit after several weeks for re evaluation if symptoms persist.
In Chapter 6 the results of a diagnostic validation study of a diagnostic model (‘Flu Score’) to detect influenza are described. Therefore, we analysed data of 3105 adult primary care patients suspected of LRTI. After history taking and physical examination, nasopharyngeal swabs were taken for polymerase chain reaction (PCR) tests to detect Influenza A and B (reference test). Diagnostic accuracy of the Flu Score to identify influenza was measured. We demonstrated that 358 patients had influenza on PCR (12%). The ROC area of the Flu Score during winter months, documented influenza and peak influenza seasons were 0.66 (95%CI 0.63-0.69), 0.69 (95%CI 0.65-0.73) and 0.72 (95%CI 0.67-0.77) respectively, but calibration was poor. The Flu Score identified 72% as ‘low risk’ and 8% of these patients had influenza. 7% were classified as ‘high risk’ and 30% had influenza (positive likelihood ratio (LR+) 3.3). During peak influenza season the proportions in these two risk categories rose to 9% and 47% (LR+ 4.5). We concluded that influenza is common among patients presenting to primary care with LRTI during winter months. The value of the Flu Score is merely in ruling in, rather than reliably ruling out influenza in LRTI patients, with a diagnostic accuracy increasing with higher influenza prevalence.
Finally, in Chapter 7, the implications for clinical practice and further research of the main findings presented in this thesis are discussed. First, there are several difficulties affecting implementation of the suggested diagnostic models. Validation, recalibration and prevention of complex arithmetic models are possible strategies to increase effective use in daily practice. Second, because diagnosis and prognosis do not completely overlap, prognostic models are warranted to be used in combination with diagnostic models in order to optimize treatment in LRTI patients. Finally, although CRP could empower GPs with an educational as well as a diagnostic tool in LRTI, adequate training in both correct interpretation of the test as well as communication skills to manage patients’ expectations would be necessary.
Despite the use of diagnostic models, that may serve as a cornerstone streamlining physicians’ thoughts and considerations, the diagnostic process in LRTI patients remains complex and will challenge GPs also in the future.