First, the kappa coefficient is a measure of agreement independent of reference to the gold standard and therefore does not measure accuracy of the test. Secondly, ROC analysis examined accuracy based on a sum score that weighted equally positive results by any of the techniques, whereas the kappa calculation requires agreement on result for each technique among examiners. Finally, kappa coefficients are affected by prevalence as are positive and negative predictive values, so that even with a high percentage of observed agreement, the kappa coefficient can be low. We have shown that AusP adds to the sensitivity of the standard examination of the chest, especially when used in populations with a high prevalence of pulmonary abnormalities. The likelihood ratio also shows superiority of AusP over ConA and ConP. No abnormalities were detected by ConP that were not detected by AusP. Given this information, we can say that the ideal screening physical examination of the lung in this population would consist of ConA and AusP rather than ConA and ConP. canada health and care mall
Of the 308 single LEs by AusP, results of 30 were abnormal with no correlating radiographic findings (falsepositive examinations). Twenty-two of these also had abnormal results of ConA or ConP examinations. Given these data, it is possible that pulmonary abnormalities were present in these lungs either in an early stage or undetectable by chest radiograph. Other explanations of false-positive examinations include poor technique, onset or advance of disease during the time between radiograph and examination, or failure by radiologist to recognize disease, or normal chest radiograph in the presence of diseased lung. Normal chest radiographs have been reported in 5 to 14 percent of patients with confirmed PCP.
False-negative examinations by AusP may have been due to a variety of reasons. Likely factors include interobserver variability, inadequacy of technique, minimal disease in the lung, or improvement in condition of lung between radiographic examination and physical examination.
Of interest is that the combined sensitivity of AusP in PCP (confirmed by positive silver stain of either induced sputum or bronchial washings) was 64.8 percent vs 38.9 percent for ConA and 14.8 percent for ConP. Of the 166 pooled abnormal LE, AusP detected 31 more abnormalities than ConP and ConA combined, 14 of the 31 being diagnosed with PCP.
The technique of AusP is not without limitations. Our study involved a highly select population with a high prevalence of lung disease and the generalizability of AusP remains to be determined. In addition, expectation bias may have been introduced by the presence of physical signs (ie, tachypnea or cough), parenteral fluids, or equipment such as oxygen delivery systems.
The preliminary results with the bedside diagnosis of PCP are very encouraging. Further research is needed in looking at the performance characteristics of AusP in other populations and settings. Further study into the transmission of sound waves through the chest is also needed. Whether AusP would be useful in the evaluation of outpatients with or without pulmonary symptoms should also be studied.