Background: Serum C-reactive protein (CRP), blood white cell count (WBC), serum procalcitonin (PCT) and erythrocyte sedimentation rate (ESR) were measured in 132 children hospitalized for community-acquired pneumonia. Serological evidence for viral infection was found in 38 cases and for pneumococcal infection in 41 cases, and the infiltrate was alveolar in 46 cases and interstitial in 86 cases. The aim of the present paper was to determine if there is a combination of these four host response markers and chest radiograph findings suitable for differentiating pneumococcal from viral etiology of pneumonia.

Methods: The 50th, 75th and 90th percentiles of CRP, WBC, ESR and PCT in the total group of 132 patients were calculated. By using these cut-off limits, the likelihood ratios of a positive test result were calculated for the possible combinations of CRP, WBC, ESR and PCT, and the likelihood ratio was 1.50 or more for six combinations.

Results: The highest likelihood ratio (1.74) was achieved with the combination CRP > 90th (80 mg/L) or WBC > 75th (17.0 x 10(9)/L) or PCT > 75th (0.84 microg/L) or ESR > 90th (63 mm/h) percentile. For this combination, the sensitivity was 61% and the specificity 65%. When the 90th percentile cut-off limit was applied also for WBC (>22 x 10(9)/L) and PCT (>1.8 microg/L), the specificity increased to 76%, but the sensitivity decreased to 37%. When the presence of an alveolar infiltration was included in the combination, the likelihood ratio was 1.89; the specificity was as high as 82% and the sensitivity as low as 34%.

Conclusions: CRP, PCT, WBC and ESR have only limited value in differentiating pneumococcal or other bacterial pneumonia from viral pneumonia. If there was a high value in at least one of the markers (CRP > 80 mg/L, PCT > 1.8 microg/L, WBC > 22 x 10(9)/L or ESR > 60 mm/h), viral infections were rare. There was no combination of these markers which was sufficiently sensitive and specific to be used in clinical pediatric practice.