Predictors of Pneumococcal Co-infection for Patients with Pandemic (H1N1) 2009

We conducted a systematic investigation of pneumococcal co-infection in patients with a diagnosis of pandemic (H1N1) 2009 and any risk factor for complications or with severity criteria. We found 14% prevalence, with one third of patients having nonpneumonic infections. A severity assessment score >1 and high C-reactive protein levels were predictors of pneumococcal co-infection.

criteria, were sent to an outpatient infectious diseases clinic and were included in a protocol described below.
The protocol included the collection of oropharyngeal and nasopharyngeal swab samples, tested as previously described (9). All patients were asked for a urine sample for pneumococcal antigen detection. A sputum sample was obtained if spontaneously expectorated, and 2 blood cultures were collected when low systolic blood pressure, hypo-or hyperthermia, signs of sepsis, or a score >2 obtained by using the British Thoracic Society's CURB-65 assessment tool (10) were present. Patients had chest radiographs taken, C-reactive protein (CRP) measured, and the severity of illness calculated by using the CURB-65 score.
The CDC real-time reverse transcription PCR (rRT-PCR) protocol for detection and characterization of the pandemic (H1N1) 2009 virus was used for the diagnosis of infl uenza cases. The BinaxNOW S. pneumoniae urinary antigen test (Inverness Medical Diagnostics, Princeton, NJ, USA) was performed and read as previously described (11). Criteria for diagnosis of S. pneumoniae infection required isolation of the microorganism from blood, isolation of the predominant organism from a qualifi ed sputum sample (11), or a positive urinary antigen test result.
A total of 418 patients with an infl uenza-like illness were evaluated, of whom 179 were confi rmed as having cases of pandemic (H1N1) 2009 virus by rRT-PCR. Of these, 99 (55.3%) patients provided a urinary sample for pneumococcal antigen detection, 37 (20.7%) a sputum sample, and 48 (26.8%) blood cultures. There were no signifi cant differences in demographic or clinical data among patients with or without a valid urinary sample for diagnostic testing, with the exception of pneumonia, which was more frequent among patients with an available sample (32.3% vs. 3.8%, p<0.001). Of 239 patients with rRT-PCR negative for infl uenza, pneumococcal infection was investigated in 171, of whom 43 (25.1%) had pneumococcal disease.
Of 100 patients who had infl uenza, a test available for pneumococcal detection, and no other bacterial pathogens identifi ed, 14 had pneumococcal infection diagnosed (14%, 95% confi dence interval 7.2-20.8; p = 0.03, compared with patients without infl uenza). Of these, 13 had a positive urinary antigen test result; 8 (57.1%) of the 14 had infection diagnosed only on the basis of this result. Of the remaining 6 patients, S. pneumoniae was isolated from blood in 2 and in sputum from 4. Demographic characteristics did not differ between pneumococcal-co-infected and non-co-infected patients (Table). Compared with patients with infl uenza infection only, those with pneumococcal co-infection more frequently had pneumonia (p<0.001), were more frequently admitted to hospital (p<0.001) and to the intensive care unit (p = 0.034), had lower O2 saturation (p = 0.006) and higher axillary temperature (p = 0.009), and more frequently had the following CURB-65 score criteria: confusion (p<0.001), respiratory rate >30 breaths/min (p = 0.009), and systolic blood pressure <90 mm Hg (p = 0.03) (Table). CURB-65 score was >1 for 35.7% of patients with pneumococcal coinfection but only 3.5% of those with infl uenza infection only (p<0.001). Levels of CRP were signifi cantly higher in patients with infl uenza plus pneumococcal disease (190.7 mg/L vs. 26.6 mg/L; p <0.001).
Pneumococcal infection characteristics were also compared between patients with and without (n = 43) pandemic (H1N1) 2009 infection included in the study. Confusion according to CURB-65 criteria was more frequent among patients with both infections (p = 0.003), while other clinical data did not differ between groups.

Conclusions
We found that the prevalence of concurrent pneumococcal infection was 14% in patients who had pandemic (H1N1) 2009 virus infection and any risk factor for infl uenza-related complications or who met severity  ‡Asthma, chronic lung disease, heart disease, neurologic and neurodevelopmental conditions, blood disorders, endocrine disorders (such as diabetes mellitus), kidney disorders, liver disorders, metabolic disorders, weakened immune system, people <19 y of age who are receiving long-term aspirin therapy, persons who are morbidly obese (body mass index >40). §Measured in patients with signs of sepsis; n values as shown.
criteria. Of note, infection in more than half these patients would not have been diagnosed if a pneumococcal urinary antigen test had not been performed. We evaluated the frequency of pneumococcal disease in patients with the pandemic (H1N1) 2009 virus through a systematic investigation by using validated diagnostic methods. Although a recent study identifi ed a high frequency of S. pneumoniae and Haemophilus infl uenzae in nasopharyngeal swabs from patients with infl uenza A (H1N1) tested by using molecular techniques (12), their clinical signifi cance and positive predictive value remain undetermined. We characterized the clinical spectrum of pneumococcal infection accompanying infl uenza pandemic. Moreover, the systematic calculation of the CURB-65 score provided us a simple, objective, and useful tool for categorization of severity and comparison between patients. Although the prevalence of pneumococcal infection in patients with infl uenza might have been overestimated by more frequent urinary sampling among those with pneumonia, we found that more than one third of the patients had nonpneumonic pneumococcal infections. Compared with patients with infl uenza only, pneumococcal co-infected patients showed a higher severity of disease as defi ned by a higher frequency of CURB-65 criteria, lower O 2 saturation, and more frequent admission to the intensive care unit. Furthermore, a CURB-65 score >1 was found to be a predictive factor of pneumococcal co-infection. Additionally, levels of CRP were also much higher in patients with pneumococcal infection. Although many co-infected patients also had pneumonia, a fact which might have explained the above-mentioned fi ndings, the same predictive factors distinguished between infl uenza pneumonia and infl uenza plus pneumococcal pneumonia.
Pneumococcal infection was more prevalent among patients with a negative test result for infl uenza. This prevalence could be explained because, presumably, many of the noninfl uenza patients met the infl uenza-like illness defi nition due to bacterial infection. However, it is unknown whether some of the infl uenza-attributable pneumococcal infection might have been underestimated, because S. pneumoniae might follow infl uenza after a lag period, and a delay in the request for medical attention might diminish the sensitivity of diagnostic tests for infl uenza. Nonetheless, data from treated patients have shown a median duration of viral shedding of 5 to 9 days and slower viral clearance with delayed antiviral drug administration (13).
In summary, the prevalence of pneumococcal coinfection during the infl uenza A (H1N1) 2009 pandemic was noteworthy, and it was associated with a higher severity of disease. In one third of the cases the clinical signs and symptoms did not indicate pneumonia, and more than one half could only be diagnosed with the urinary antigen test. A CURB-65 score >1 and CRP levels proved to be useful tools to identify patients at higher risk for pneumococcal co-infection for whom physicians should adopt additional diagnostic and therapeutic measures.