This is the first study on hospitalised children with radiologically confirmed pneumonia in DR Congo in the PCV13 post-vaccine era. It includes both clinical and microbiological aspects. A majority (73%) of the children were diagnosed as having severe pneumonia. The most common symptoms for both severe and non-severe pneumonia cases were fever, cough or abnormal auscultation, similar to that described in a Vietnamese [24] and a Tanzanian study [25]. High white cell counts (> 20,000 cells/µL) and high CRP levels (> 75 mg/dL) were associated with severe pneumonia at admission. Similar findings of elevated white cell counts in hospitalised children with pneumonia were reported from Senegal [26], as well as high CRP levels (> 80 mg/dL) also being associated with radiologically confirmed pneumonia in Tanzanian children [16].
The case fatality rate in our study was 9.5%, similar to findings from a Tanzanian district hospital (11%) [25] but higher than those reported from hospitalized children in Cambodia (3.2%) [27]. Up to 80% of deaths in children with severe respiratory infections may occur outside hospitals in low-income countries [28]. It is possible that some children first receive self-medication by parents or guardians, first seeking care at a private pharmacy, traditional practitioner [29] or at any other non-appropriate health care provider before attending professional health care facilities [5, 28, 30]. This is supported by the fact that less than half (42%) of under-five children with suspected pneumonia in DR Congo were found to have been treated by a trained health care provider [8].
Eighty-seven percent of the children included in our study received pre-hospitalisation antibiotic treatment, in most cases amoxicillin, ampicillin or penicillin (59%) followed by trimethoprim-sulfamethoxazole (20%). In 34% of the cases, information about pre-hospitalisation medication was obtained from patient or guardian reports only. Earlier reports show that parent history of antibiotic treatment in children is not always reliable [31–33]. Bioassays for detection of antibiotic activity in urine or serum may be of value for assessment of pre-hospitalisation antibiotic use [34–36] but these were not performed in the present study. However, from earlier studies it is known that pre-hospitalisation medication is very common in low-income countries, especially when there is limited health care access and high availability of antibiotics [5, 8]. Fifteen years ago, it was found that trimethoprim-sulfamethoxazole and amoxicillin were frequently used in self-medication in Congolese children having cough with or without fever [5]. We have recently showed a high level of resistance against trimethoprim-sulfamethoxazole in pneumococci colonizing healthy children in this area of DR Congo [18] and similar results were shown for healthy children in northern Tanzania [37]. Here we found that pre-hospitalisation medication by the peroral broad spectrum antibiotic trimethoprim-sulfamethoxazole was more common in children with severe pneumonia than in children with non-severe disease, and pre-hospitalisation treatment with trimethoprim-sulfamethoxazole was also associated with fatal outcome.
Recently the DR Congo introduced clinical guidelines for the management of pneumonia. However, the antibiotic management of severe pneumonia was not included. For optimizing antibiotic treatment regimens antimicrobial resistance testing and surveillance are needed, especially considering the high level of resistance to commonly used antibiotics in the country [18]. In the present study, ceftriaxone and gentamicin were the main antibiotics used for treating severe pneumonia. This corroborates the findings of a study in Senegal where more than 55% of pneumonia cases were treated with ceftriaxone [26, 38].
The low isolation rate of live pneumococcus (0.9%) by culture in the present study may be explained by the high frequency of pre-hospitalisation antibiotic treatment. The pneumococcal isolation rate was higher (21%) among the 794 healthy Congolese children included in our previous study; Also, 7% of these children had been treated with antibiotics in the last month according to their parents or guardians [18]. If carried out before specimen collection, this is known to reduce the culture isolation rate of bacteria but has less impact on the detection rate by PCR [36, 39]. Here we detected pneumococci by PCR in almost all (96%) nasopharyngeal secretions and H. influenzae in 54%. This result corroborates findings from Zanzibar, Tanzania, in which pneumococci were found in 87% and H. influenzae in 77% of febrile children [40]. When employing a more stringent cut-off level, (i.e. Ct-value < 30) in which only high amounts of microbial nucleic acids are identified, we found S. pneumoniae in 53%, and H. influenzae in 20% of the cases. We could not show any significant differences between the age groups and presence of pathogens as reported in other studies [41, 42] nor an association between severity of disease and specific viruses or bacterial species as identified in the study. However, high levels of bacterial or viral nucleic acids in the nasopharynx (irrespective of the species or types) were associated with more severe pneumonia at admission, compared to less severe cases. This supports the importance of not only bacteria but also viruses in the development of severe pneumonia.
Surprisingly, our detection rate of 7% RSV was much lower than the 31% reported in a multicentre study in Africa and Asia [14]. This discrepancy might be explained by the geographic variability of pathogens [14]. However, although only a few cases were detected, RSV was more prevalent in children that died, also having high nucleic acid levels of pneumococci. In the two fatal RSV-positive cases, pneumococci were also detected, supporting the evidence of RSV being associated with pneumococci in critical cases.
The frequent co-occurrence of viruses and bacteria in childhood pneumonia [14] has been associated with disease severity [10, 14]. Infection with RSV facilitates colonisation with bacteria such as S. pneumoniae and H. influenzae in the nasopharynx of young children [43]. In addition, invasive pneumococcal infection increases during the seasonal peak of various respiratory viruses, including influenza virus and RSV [14].
Since we did not include a hospital based control group, and most of these pathogens are also prevalent in healthy children, it was not possible to discern the etiology of the pneumonia cases in the present study, not even when only microbial nucleic acids in large amounts were considered [16, 40, 44, 45].
PCV13 serotypes/serogroups were more commonly identified than non-PCV13 serotypes/groups (63% versus 37%). This was in contrast to a study performed in Mozambique, in which half of the identified serotypes were included in the vaccine [46]. However, we may have under-estimated the frequency of non-PCV13 serotypes. This is because our method could neither distinguish between the 6ABCD serotypes, (of which only 6A and 6B are in PCV13), or between 9A and 9V (of which only 9V is in PCV13). Moreover, the assay measures all the serotypes included in the pneumococcal vaccines but few of the additional ones. This suggests that, because those 50 samples were positive for pneumococci according to our PCR assay, but were negative for the serotypes included in the detection panel, they may have contained non-vaccine serotype pneumococci. Thus, non-vaccine serotypes were most likely much more prevalent than we detected here. As seen in other studies non-vaccine serotypes are continually emerging after the introduction of PCV13 [47, 48]. Here, serotype 19F was the most frequent PCV13 serotype, whereas serotypes 15BC, 11A and 10A were the most prevalent non-PCV13 serotypes. This was similar to our findings among the healthy Congolese children [18]. In this way, similar information about serotype distribution was obtained from healthy children as from children with pneumonia requiring hospital stay in the same area. However, the previous study was based on cultured pneumococcal isolates rather than direct detection in the nasopharyngeal sample as was done here [18].