A prediction model for community-acquired Chlamydia pneumoniae pneumonia in hospitalized patients.

Background: The objective of the study was to identify factors that help to predict community-acquired Chlamydia pneumoniae pneumonia in hospitalized patients.

Patients And Methods: Clinical data of 83 patients with serologically confirmed C. pneumoniae pneumonia were compared with the data obtained from 72 patients with bacterial pneumonia. The criteria of bacterial pneumonia included positive blood and/or sputum cultures and negative serology for Mycoplasma pneumoniae, Chlamydia psittaci, Chlamydia pneumoniae and Coxiella burnetii. The data collected included demographics, chronic diseases, pre- and post-hospitalization course of pneumonia, clinical data on admission and laboratory findings. Descriptive statistical analysis, involving numerous variables, was followed by univariate and multivariate logistic regression analysis.

Results: Two different situations, one including demographic data and information on the pre-hospitalization course of pneumonia, and another based on clinical information on admission and on laboratory results, were modeled using multivariate logistic regression. Several variables selected from these two models were incorporated into the third model, and the following four variables were found to have the highest predictive value of C. pneumoniae pneumonia: nursing home residence (odds ratio [OD] 3.73, 95% confidence interval [CI] 1.39-10.06), low c-reactive protein (CRP) levels (OD 5.99, 95% CI 1.82-19.67), nonproductive cough (OD 0.32, 95% CI 0.14-0.73), and a normal urinalysis (OD 0.38, 95% CI 0.17-0.83).

Conclusion: Our findings seem to allow for a more reliable differentiation between C. pneumoniae pneumonia and other bacterial pneumonias, but further investigations will be needed to validate the proposed model.