The value of biomarkers in predicting the diagnosis of pneumonia in elderly frail patients admitted to an acute medical ward: is procalcitonin really useful as opposed to C-reactive protein?

Background – Biomarkers, such as C-reactive protein (CRP) and procalcitonin, have been proposed in clinical practice for an early diagnosis of pneumonia in patients admitted to hospital, especially when chest X-ray and physical examination findings are inconclusive. Procalcitonin is generally thought to be more sensitive and specific than CRP, but studies in elderly frail patients with multiple comorbidities are still lacking. Aims – Our aim was to investigate the possible role of CRP and procalcitonin levels at admission to hospital as predictors of pneumonia in a cohort of elderly frail patient with a high comorbidity burden. Methods – At Internal Medicine and Critical Subacute Care Unit of Parma University Hospital (94 beds managed by care intensity) we carried out a retrospective evaluation of the clinical records of all patients over 65 years old admitted to the Acute Care Ward from the Emergency Department during a 8-month period (January-August 2013). For each patient we recorded main diagnosis, chronic comorbidities, diagnosis of pneumonia at discharge, CRP and procalcitonin levels at admission. We also recorded inhospital mortality. For each patient indexes of clinical complexity such as CIRS (Cumulative Index Rating Scale) Comorbidity Score and CIRS Severity Index were also calculated. Data were analyzed with Mann-Whitney’s U, Spearman’s rho and odds ratio test. A Receiver Operating Characteristic (ROC) curve was built to verify PCR and procalcitonin predictive value for pneumonia diagnosis and death. Results – 1199 patients (546 M, 653 F, mean age 78 ± 13 years) were evaluated. 239 patients (20%, 138 M, 101 F) were discharged with a diagnosis of pneumonia. Median levels of CRP at admission were significantly higher in patients with pneumonia (116 mg/L, IQR 46.0-181.0) than in other patients (24 mg/dl, IQR 5.2-84.2, p<0.0001), while procalcitonin median levels were not (respectively, 0.24 ng/ml IQR 0.13-1.79 vs 0.26 ng/ml IQR 0.09-0.88). The ROC analysis (Figure 1) showed that CRP values were significantly predictive of pneumonia (AUC 0.76, IC95% 0.72-0.79, p<0.0001) and death (AUC 0.74, IC95% 0.69-0.80, p<0.0001), while procalcitonin lacked sufficient sensitivity and specificity to be discriminant for pneumonia diagnosis. Patients with a CRP level > 61 mg/L had a 3.5-fold increased risk of having pneumonia than patients with levels ≤ 61 mg/L (OR 3.5, IC95% 2.6-4.7, p<0.0001). CRP levels < 5 mg/L had a 98% negative predictive value in excluding the diagnosis of pneumonia. Both CRP and procalcitonin levels were directly correlated with CIRS comorbidity score and CIRS severity index (Spearman’s rho CIRS comorbidity score vs CRP 0.22, p<0.0001). Conclusions – In elderly frail patients with multiple chronic comorbidities, procalcitonin does not seem to be useful as a biomarker for the diagnosis of pneumonia. Normal CRP values may help excluding pneumonia in most cases, while a CRP elevation is useful for clinical suspicion but lacks specificity for this disease.