Temperature, pulse oximetry help predict pneumonia in primary setting
In adults presenting to primary care, radiographic confirmation of pneumonia is very uncommon, according to a recent study. In this setting, pulse oximetry and clinical signs may aid the diagnosis of pneumonia.
The cohort study included 28,883 adults with lower respiratory tract infections (LRTI). Those with other causes of acute cough were excluded. Clinical data of the participants, such as symptoms and severity, smoking history, and examination results, were collected and included in the analysis.
Chest radiography was available for 1,782 of the 28,883 participants within 30 days. For 720 participants, chest radiography was available within 7 days. Only 115 of these were able to accurately confirm pneumonia, resulting in a very low confirmation rate of 0.4 percent.
Four clinical examination findings were shown to be diagnostically useful for pneumonia, the most powerful of which was having a temperature >37.8oC (risk ratio [RR], 2.6; 95 percent CI, 1.5 to 4.8). The other signs were crackles on auscultation (RR, 1.8; 1.1 to 3.0), oxygen saturation <95 percent (RR, 1.7; 1.0 to 3.1) and pulse >100 per minute (RR, 1.9; 1.1 to 3.2).
Area under the receiver operating curve (AUC) analysis showed that temperature alone was poorly predictive of pneumonia (AUC, 0.59; 0.55 to 0.63). Including crackles (AUC, 0.65; 0.60 to 0.70; p=0.001) then oxygen saturation (AUC, 0.67; s0.61 to 0.73; p=0.095) substantially increased predictive power.
In contrast, the final inclusion of pulse (AUC, 0.68; 0.62 to 0.74) did not appreciably increase the predictive power of the model.
In terms of sensitivity, using temperature alone did not produce a sufficiently sensitive predictive model. To achieve a sensitivity of 83.5 percent, all four factors needed to be considered.