Pneumonia, Meningitis, and Septicemia in Adults and Older Children in Rural Gambia: 8 Years of Population-Based Surveillance

Abstract Background Representative data describing serious infections in children aged ≥5 years and adults in Africa are limited. Methods We conducted population-based surveillance for pneumonia, meningitis, and septicemia in a demographic surveillance area in The Gambia between 12 May 2008 and 31 December 2015. We used standardized criteria to identify, diagnose, and investigate patients aged ≥5 years using conventional microbiology and radiology. Results We enrolled 1638 of 1657 eligible patients and investigated 1618. Suspected pneumonia, septicemia, or meningitis was diagnosed in 1392, 135, and 111 patients, respectively. Bacterial pathogens from sterile sites were isolated from 105 (7.5%) patients with suspected pneumonia, 11 (8.1%) with suspected septicemia, and 28 (25.2%) with suspected meningitis. Streptococcus pneumoniae (n = 84), Neisseria meningitidis (n = 16), and Staphylococcus aureus (n = 15) were the most common pathogens. Twenty-eight (1.7%) patients died in hospital and 40 (4.1%) died during the 4 months after discharge. Thirty postdischarge deaths occurred in patients aged ≥10 years with suspected pneumonia. The minimum annual incidence was 133 cases per 100 000 person-years for suspected pneumonia, 13 for meningitis, 11 for septicemia, 14 for culture-positive disease, and 46 for radiological pneumonia. At least 2.7% of all deaths in the surveillance area were due to suspected pneumonia, meningitis, or septicemia. Conclusions Pneumonia, meningitis, and septicemia in children aged ≥5 years and adults in The Gambia are responsible for significant morbidity and mortality. Many deaths occur after hospital discharge and most cases are culture negative. Improvements in prevention, diagnosis, inpatient, and follow-up management are urgently needed.


Study Population
The Basse Health and Demographic Surveillance System (BHDSS) operates in the half of URR that lies to the south of the Gambia River (population 171 070 in 2012). The population is largely rural, and the major occupation is agriculture. This study included all residents of the BHDSS, the population of which is enumerated every 4 months with records of births, deaths, and migrations. The prevalence of HIV infection among women in antenatal care in 2012 was 1.2% [20]. This report is restricted to patients aged ≥5 years who were confirmed resident in the BHDSS.
Pneumococcal conjugate vaccine (PCV), which was introduced for infants in the URR as 7-valent PCV in August 2009 and as 13-valent PCV in May 2011. A previous analysis indicates that the incidence of invasive pneumococcal disease during this study period showed a possible trend toward herd protection for older patients, but data were insufficient to be certain [21].

Case Definitions and Study Procedures
The methods used in this study have been described previously [21,22]. In brief, we conducted 24 hours a day surveillance for pneumonia meningitis and septicemia from 12 May 2008 to 31 December 2015, screening all individuals who presented to Basse Health Centre or to one of the peripheral healthcare facilities in the region, whether or not they were admitted. Surveillance nurses applied standard criteria to every patient to identify cases for referral to a clinician at Basse Health Centre (Table 1 and Supplementary Appendix 1.1). Clinicians applied standard criteria to generate a surveillance diagnosis of suspected meningitis, septicemia, pneumonia, or another diagnosis (Table 1 and Supplementary Appendix 1.2). The surveillance diagnosis led to standardized investigation with blood culture, lumbar puncture, and chest radiography (Table 1 and Supplementary Appendix 1.3). According to clinical and radiological features, aspiration of pleural fluid or lung aspiration was performed for patients with pleural effusion or dense peripheral pneumonic consolidation. Weight was measured using a digital scale (HD-314, Tanita, Arlington Heights, Illinois) and height using a ShorrBoard (Weigh and Measure, LLC, Olney, Maryland). In 2010, samples were not collected between 5 October and 3 November, and radiographs were not performed between 5 and 12 October, when the Gambia River flooded the Medical Research Council (MRC) Unit Field Station in Basse. Patients were admitted or treated as outpatients according to national guidelines. The outcome of each patient's admission was recorded. Postdischarge mortality was monitored through visits every 4 months to each household in the BHDSS.

Microbiology
Blood, lung aspirate, cerebrospinal fluid (CSF), pleural fluid, and other microbiological samples were processed at the MRC Basse Field Station using standard methods [23]. Aerobic and anaerobic blood cultures were taken for each adult patient, with 5 mL of blood added to each bottle. Children had 2-3 mL added to a single pediatric blood culture bottle. Bottles were weighed before and after blood was added to ensure adequate filling, with weights reported to clinicians. Biochemical tests, including the Analytical Profile Index (bioMérieux, UK), and serological tests were used to confirm suspected pathogens. Latex agglutination tests were used on all CSF samples to identify Streptococcus pneumoniae, Neisseria meningitidis, and group B streptococci. Results were aggregated with culture-positive cases. Pneumococcal isolates were serotyped at the MRC Fajara laboratory using a latex agglutination assay employing factor and group-specific antisera (Statens Serum Institut, Copenhagen, Denmark) and polymerase chain reaction [24]. The laboratories in Basse and Fajara submitted to external quality assurance throughout the study (UK National External Quality Assessment Service, World Health Organization [WHO] reference laboratory in Denmark, and the Royal Australasian College of Pathologists).

Radiology
Chest radiographs were performed using consistent methods producing digital images in accordance with WHO recommendations [25]. Two readings of each radiograph were undertaken by 3 independent readers and discordant results were resolved by a pediatric radiologist with experience in reading adult films. All readers were trained and used the WHO panel of standardized radiographs for calibration to the WHO standard for radiological pneumonia, achieving κ agreement scores of ≥0.8 before reading study radiographs.

Ethics Approval
The study was approved by The Gambia Government/MRC Institutional Ethics Committee (number 1087) and the ethics committee of the London School of Hygiene and Tropical Medicine. Participants, or their parents or guardians, gave written informed consent.

Statistical Analysis
Categorical data were compared using the χ 2 or Fisher exact test. Continuous data were compared using t tests with appropriate distributional assumptions. We calculated age-specific annual minimum incidence of disease by dividing the number of cases by the midpoint population estimates from the BHDSS and multiplying by 100 000. We did not adjust our figures for the sensitivity of diagnostic tests, study enrollment rates, or healthseeking behavior. Age groups were prespecified as 5-9 years, 10-14 years, 15-54 years, and ≥55 years. Incidence and 95% confidence intervals (CIs) were calculated assuming a Poisson distribution. Statistical significance was set at a P value < .05.

RESULTS
A total of 1921 patients were screened for inclusion and 1657 met inclusion criteria; 19 had invalid records, so 1638 were included in the analysis ( Figure 2). The median age of patients was 10.8 years (interquartile range [IQR], 6.8-29 years) with a left-skewed age distribution similar to that of the background population; this was particularly evident in the age group 5-9 years, in which a disproportionate number of cases occurred ( Figure 3).
The number of cases of suspected pneumonia (n = 1392) was much greater than those of suspected meningitis (n = 135) or septicemia (n = 111) (Tables 2 and 3). The substantial number of suspected meningitis cases was associated with an epidemic One or more of the following has been present for ≤14 days: history of cough and difficulty breathing; history of cough and pleuritic chest pain; history of cough and supraclavicular/sternal recession or nasal flaring; history of productive cough and fever; history of rigors; history of seizure; impaired consciousness; altered mental state; axillary temperature of at least 38°C or <36°C in a patient admitted or being admitted; photophobia; neck stiffness; local musculoskeletal swelling or tenderness; irrespective of residential location, any patient with suspected meningitis Diagnosis All 3 suspected illnesses defined as physician diagnoses and also to be considered if: • For meningitis: ≥2 of the following are present: axillary temperature ≥37.5°C; meningism (neck stiffness and/or photophobia); altered mental state (Glasgow Coma Scale <14) • For suspected pneumonia: an illness of ≤14 days' duration where ≥2 of the following are present: cough; hemoptysis; pleuritic chest pain; breathlessness; axillary temperature of ≥38.0°C • For suspected septicemia: at least 1 of: clinician diagnosis of focal sepsis (including but not limited to: septic arthritis, osteomyelitis, endocarditis, peritonitis, liver abscess, soft tissue abscess or cellulitis) or of generalized septicemia; axillary temperature <36.0°C or ≥38°C in a patient admitted or being admitted; history of rigors Investigation All patients are to have blood culture and a rapid diagnostic test for malaria.
• Patients with suspected meningitis are to have lumbar puncture.
• Patients with suspected pneumonia are to have chest radiograph.
• Chest radiograph should also be considered in patients with meningitis or septicemia if the clinician's impression is of coexisting pneumonia or if it is judged that a chest radiograph will assist in management. • Lung aspirate should be considered for a patient if peripheral consolidation has been demonstrated, preferably by radiograph.
• Other investigations including pleural tap and joint aspirate may be considered according to the clinical indication.
of meningococcal meningitis in the region in 2012 [26]. The nutritional status of patients was poor, with a median body mass index of 14 kg/m 2 (Table 3). Approximately half of the patients with suspected pneumonia were admitted to hospital.
The minimum estimated annual mortality rate associated with suspected meningitis, septicemia, or pneumonia was 7 per 100 000 PY with the greatest mortality being 20 per 100 thinsp;000 PY in those aged ≥55 years. Based on counts of all-cause mortality in the BHDSS during the study period, we estimated that suspected meningitis, septicemia, or pneumonia was associated with a minimum of 3.6% of all deaths in 5-to 9-year-olds, 1.8% in 10-to 14-year-olds, 3.5% in 15to 54-year-olds, 1.5% in those aged ≥55 years, and 2.7% of all deaths in those aged ≥5 years.

DISCUSSION
Eight years of population-based surveillance among older children and adults in rural Gambia revealed a significant burden of disease due to pneumonia, meningitis, and septicemia. The findings, generated using standardized procedures in a wellcharacterized population under demographic surveillance, provide a representative description of the frequency, etiology, and mortality associated with meningitis, septicemia, and pneumonia in age groups for which there are few available data in Africa. Please see the Supplementary Material for more detail on the context of this discussion and a comparison of publications referenced in Supplementary Material Table 1.
The minimum incidence and mortality of suspected pneumonia, meningitis, or septicemia was substantial. We found that suspected pneumonia was >5 times more prevalent than suspected meningitis or suspected septicemia and that the greatest burden of any illness occurred in the age group 5-9 years.
Our observed minimum incidence of suspected pneumonia of 133 cases per 100 000 PY is a similar figure to Cohen et al [13] studying severe respiratory infection in Johannesburg, and our minimum incidence for suspected meningitis of 13 cases per 100 000 PY is similar to Traore et al [18] and Soeters et al [15] studying meningitis in West Africa. Thriemer et al in Zanzibar [17] reported a bacteremia incidence in the age groups 5-15 years and >15 years of 6 and 7 per 100 000 PY, respectively, approximately half our value of 13, whereas Sigauque et al in Mozambique [14] reported a rate of 49 cases of bacteremia per 100 000 PY in 5-to 15-year-olds, a rate not dissimilar from rates we report when the age discrepancy is taken into account. Final rates from Thriemer et al [17], adjusted for study nonenrollment and blood culture sensitivity, are 20 times higher, and from Verani et al [19] (the only similar study to adjust for healthcare-seeking behavior), 2-3 times higher than their unadjusted rates. This emphasizes the point that nearly all such estimates are of a minimum incidence. However, the fact that our surveillance operated in all primary healthcare centers in the surveillance area, the high enrollment rate in surveillance, and the use of standardized physician diagnoses rather than positive  laboratory tests are likely to contribute to a robust figure that omits only a healthcare-seeking behavior adjustment. Our data suggest that pneumonia, meningitis, and septicemia were together associated with at least 2.7% of all deaths in older children and adults in the study population. Case fatality risk (CFR) was highest in cases of suspected meningitis (16.3%), although due to its higher prevalence across all ages, suspected pneumonia accounted for double the deaths of suspected meningitis and was particularly lethal in older patients (CFR of 15.2% at 4 months). Cohen et al [13] reported an in-hospital CFR of 5% in HIV-uninfected individuals with pneumonia, compared to our risk of 2% for admitted cases of suspected pneumonia in our low-HIV-incidence cohort. Traore et al [18] reported a CFR for pneumococcal meningitis of 45% in 5-to 14-year-olds and 44% in subjects aged >14 years, whereas our mortality risk in these 2 age groups was 19% and 6%, respectively, for suspected meningitis, most of whom had no microbiological diagnosis.
Across all diseases and ages, we found a concerningly high mortality in the 4-month period after discharge from hospital, with 40 postdischarge deaths compared to 28 inpatient deaths. Postdischarge mortality was particularly associated with suspected pneumonia and with consolidation on chest radiograph: 32 of the 40 postdischarge deaths were associated with suspected pneumonia and 16 with radiological pneumonia. Postdischarge mortality was similar across all 4 age groups [27]. These observations suggest that postdischarge death may relate to inadequate diagnosis and triage and/or therapy rather than a group-specific predisposing factor. Other plausible explanations, some of which have been highlighted in younger children [27], include unaddressed comorbid conditions such as malnutrition, anemia, HIV, tuberculosis, sickle cell disease, or diabetes, or issues such as nonmedical discharge or nosocomial infection. Pneumococcal antimicrobial resistance is not significant in our setting [28].   We and Sigauque et al [14] show similar trends in the bacteria causing bacteremia in older children and adults in sub-Saharan Africa; S. pneumoniae, S. aureus, and Salmonella species featured prominently in each study. We have fewer gram negatives in total: 32% vs 39% of isolates, and 8% vs 20% for Salmonella species. This may reflect our screening criteria, which did not include screening for abdominal or diarrheal causes of sepsis or the differing burden of HIV or Salmonella species disease in the 2 communities. Our study and all studies reviewed show that most patients investigated with blood, CSF, and lung cultures have no microbiological diagnosis made, even in patients who died from meningitis and who are likely to have had a bacterial infection. In our study, culture-positive S. pneumoniae was due to serotypes 1 and 5 in 78% of cases, while 90% of cases were caused by serotypes included in 13-valent PCV, the PCV currently used in The Gambia's infant immunization program. Although the effects of herd protection from infant pneumococcal vaccination during our study period are uncertain [21], this finding may have implications for the proportion of invasive pneumococcal disease potentially preventable by protection related to PCV.
The strengths of our study include the standardization of screening, clinical investigation, case definitions, and laboratory practices, with investigation of >97% of eligible patients. Our surveillance included outpatients and inpatients and aimed to detect all cases presenting to any government health facility. Conduct of the study in a geographically and demographically defined population allowed follow-up after discharge, calculation of incidence, and estimation of the proportion of all deaths that were related to pneumonia, meningitis, or septicemia. Limitations included the inability to detect all cases that occur in the population, particularly in those who seek care outside the government healthcare sector (eg, from a traditional healer or a pharmacy) or who die at home, and absence of healthcare-seeking behavior data to correct for these patients; less-than-optimal sensitivity of conventional microbiology to detect all bacterial pathogens; and no testing for mycobacterial, viral, or other pathogens.
Our study demonstrates the extent to which sepsis syndromes remain major causes of death and disease in older children and adults in rural Africa and the many gaps that exist in understanding the causes of these syndromes and determinants of their outcome in older children and adults. High rates of postdischarge mortality in Africa have been observed in young children [27], but to our knowledge this is the first description of postdischarge mortality associated with acute infections  Abbreviations: CXR, chest radiograph; NA, not applicable. a Inpatient mortality is calculated as a percentage of the whole cohort, not as a percentage of those admitted, and 4-month mortality is also calculated as a percentage of the whole cohort and therefore includes inpatient and outpatient deaths. b One of the 2 patients in this category had coexisting suspected meningitis. c Two patients with suspected meningitis refused admission and survived to 4 months. exceeding inpatient mortality in older children and adults. This finding underlines the importance of follow-up of pneumonia patients, apparently cured, with even closer attention paid to those with consolidation on chest radiograph. The determinants of postdischarge mortality in older patients should be formally studied in other African settings to establishing prevention strategies. The inadequacy of conventional diagnostics was brought into sharp relief, highlighting the need for better diagnostics to determine the etiology of pneumonia, meningitis, or septicemia and prevalent comorbidities. We call for a greater focus on the prevention, diagnosis, treatment, and follow-up of sepsis patients by the international community. Data availability. Deidentified participant data will be made available