A Cross-Cutting Approach to Surveillance and Laboratory Capacity as a Platform to Improve Health Security in Uganda

Baseline Capacity Assessment, Training, and Site Activation

Surveillance for clinical and demographic data was ongoing at 5 of the 6 sites via the inpatient malaria surveillance system. Thus, routine and standard data collection for purposes of surveillance was a new activity for only 1 site. Site activation reflected collection of expanded information for surveillance, improved electronic surveillance infrastructure, and laboratory diagnostic capacity. Activation of sites occurred between July 2016 and October 2017. The 6 sentinel hospital sites and their associated system activation dates were: Jinja Regional Referral Hospital (Jinja RRH) (July 2016), Arua RRH (July 2016), Mubende RRH (August 2016), Kabale RRH (April 2017), Apac District Hospital (July 2017), and Tororo District Hospital (October 2017) (Figure 1).

OPEN IN VIEWER

Figure 1.

Geographic location of 6 sentinel site hospitals engaged in case-based surveillance and laboratory capacity-building efforts in Uganda

Before activation of this system at each selected government hospital, a baseline rapid assessment was conducted to determine: (1) existing on-site laboratory capacity for microbiology; (2) potential for expansion of on-site microbiologic capacity in terms of workforce, training needs, equipment needs, and available space; (3) biosafety and biosecurity practices and gaps; (4) infection prevention and control practices and gaps; and (5) existing organizational and management structures and their associated ability to support modified practices.

Briefly, assessment was conducted using the Potter and Brough Capacity Pyramid approach, which assesses capacities at various structural levels, all within the local context, including tools, skills, staff and infrastructure, structures, systems, and roles. ²⁰ Standards and tools used or adapted included those of the WHO SLIPTA program (Stepwise Laboratory Improvement Process Towards Accreditation), the WHO laboratory assessment tool, and CDC infection control assessment tools.^21-23 These assessments revealed extensive gaps including inactive hospital-based infection prevention and control programs, lack of supplies and infrastructure to promote basic hand hygiene, lack of microbiology services, poor access control to laboratory rooms, inadequate sample transportation for microbiologic samples to a central reference laboratory, lack of defined systems for pharmaceutical tracking, weak information technology and data management, and inadequate interface between clinicians and hospital laboratories. Prior to activation, ongoing routine diagnostic testing was unavailable for infectious conditions other than malaria and HIV.

The process of surveillance site activation included inception meetings with varied hospital and department administrators and staff; provision of equipment and supplies for phlebotomy, infection control, clinical specimen processing and storage, and internet connectivity according to needs identified on assessment. At each surveillance site, multifaceted training was aimed at pediatricians, nurses, and laboratory technicians, but it included additional cadres of Ugandan government healthcare workers staffed in each hospital. This training was delivered through interactive lecture and hands-on practice. Content included aseptic technique, appropriate sample volumes for laboratory tests, and the importance of good record keeping for surveillance and improved clinical care. Training also included the importance of investigating nonmalarial causes of illness and incorporation of bacterial blood culture into routine clinical care. Initial emphasis focused on improving diagnosis of nonmalarial illnesses by reserving blood culture for febrile children without evidence of malaria by microscopy or rapid diagnostic test (RDT). However, clinicians were also encouraged to use blood culture as a diagnostic tool in any circumstance deemed appropriate.

Continued regular site support visits occurred following activation to provide ongoing training in phlebotomy, troubleshoot staff concerns, identify areas for improvement, and provide summary data and feedback. Concerns and issues that arose were addressed and incorporated into ongoing training efforts at all surveillance sites. When clustering of contaminated cultures in space and time occurred, investigation and intervention followed. Areas dedicated to phlebotomy were created on all wards to improve the likelihood of aseptic draws in otherwise difficult hygienic conditions.

Laboratory System Strengthening

Blood cultures were made available at the time of surveillance site activation. With a focus on improving surveillance and diagnostic capacity for severe illness in the pediatric population, most blood culture bottles purchased and provisioned to hospitals were aerobic pediatric bottles for blood volumes up to 3 mL according to patient age and weight. Before installation of BD BACTEC^™ (Becton Dickinson, Franklin Lakes, NJ, USA) machines in sentinel hospital laboratories, all inoculated blood culture bottles were transported daily at ambient temperature to the Department of Medical Microbiology (DMM) at Makerere University in the capital city of Kampala for processing. Upon successful installation of BACTEC machines in sentinel site hospitals and associated training, those hospital laboratories incubated all blood culture bottles locally for up to 7 days. Negative results were provided directly back to hospital clinicians. Bottles read as positive by the machines were usually subjected to Gram stain at the hospital laboratory in order to provide timely and clinically relevant information before transportation to DMM for subculturing, identification, and antimicrobial susceptibility screening. Identification at DMM was performed using conventional, phenotypic identification algorithms down to species and subspecies or serotype when appropriate reagents were available. Classification of isolated organisms as pathogens or likely contaminants was made presumptively by DMM based on general knowledge regarding pathogenicity of various organisms and presented accordingly here. Most antimicrobial susceptibility testing was performed using the Kirby-Bauer disk diffusion method according to CLSI standards and interpreted according to 2014 breakpoints.

We instituted serologic tests for patients with a history of febrile illness of at least 4 days' duration, a restriction designed to assure improved sensitivity for detection of IgM antibodies for certain conditions. Tests were instituted at each site as they became available in Uganda and following adequate staff training, independent of site activation. A lateral flow assay to detect IgM antibody to a variety of Leptospira spp. was provided to all sentinel hospitals (Test-It^™, Life Assay, Cape Town, South Africa). Remaining sera were transported to the Uganda Virus Research Institute and screened for IgM reactivity to multiple mosquito-borne arboviruses, specifically West Nile, yellow fever, dengue serotypes 1-4, Zika, and chikungunya viruses. IgM positive samples were considered presumptive positive flavivirus or alphavirus infections. When sample volume allowed, IgM screening was followed with confirmatory plaque reduction neutralization tests (PRNT). This allowed for confirmation of specific infecting virus and differentiation of chikungunya and o'nyong nyong virus infections among those that screened positive on chikungunya IgM screening. Additionally, sera were tested for antibodies to Brucella abortus, B. suis, and B. melitensis by the microagglutination test, ²⁴ with minor modifications including use of U-bottom plates, incubation at 28°C, and discontinued use of safranin, using antigen produced by the US Department of Agriculture National Veterinary Services Laboratory (Ames, Iowa, USA). Brucella antibody titers were interpreted as follows: >160 considered positive, 20-80 considered borderline, and <20 was considered negative. Remaining sera were stored for ultimate shipment to US CDC. Testing at CDC provided quality assurance of tests newly implemented in Uganda and allowed for investigation into other vector-borne and zoonotic conditions causing human illness in Uganda for which local laboratory diagnostic capacity had not yet been developed. Accordingly, sera were screened for human antibody reactivity to Rickettsia africae because of this antigen's apparent broad specificity for spotted-fever group rickettsiae (SFGR) in this geographic region (unreported data). Due to limits of available assays, screening was conducted using an immunofluorescence assay (IFA) that primarily detects IgG antibodies. Thus, this rickettsial screening was most likely to define past exposure rather than cause of current infection, nevertheless informing potential need for development of appropriate laboratory capacity in Uganda. A titer of 1:64 was considered the cut-off for significant antibody response.

Surveillance and Laboratory Data Capture and Management

In each sentinel hospital pediatric ward, a standardized medical record form previously developed for malaria surveillance ²⁵ was adapted to incorporate additional variables relevant to expanded diagnostic capacity with stakeholder and clinician input. This form serves as a medical chart for all admissions, allowing providers to record symptoms, signs, treatments, and outcome at discharge.

A full-time surveillance data officer was stationed at each hospital. The primary functions of this person were to engage with clinicians and laboratory staff to ensure that samples were collected, labeled, and transported properly and that clinical and laboratory records were complete; to enter demographic, clinical, and laboratory data into the electronic database; and to facilitate ongoing communication among the surveillance team and hospital staff and administrators. Initially, data were entered into a Microsoft Access^® database, but following improvements in connectivity, data entry was subsequently transitioned to a web-based data management system on the District Health Information System (DHIS-2) platform (www.dhis2.org). DHIS-2 is a flexible, open-sourced information system used by Uganda's Ministry of Health for aggregate public health data management and by the Public Health Emergency Operations Center for recording and tracking case-based data related to outbreak response and specimen collection, transportation, and testing. The system was tailored to allow case-based routine surveillance data transmission, with multiple data entry points on the same individual record to enable linking of patient data with central laboratory results, including antimicrobial resistance patterns. The system permits site-based staff, central laboratory staff, and ministry of health personnel to view data on individual patients and associated laboratory results in real-time, thereby improving data quality and use for case management, specimen tracking and testing, and disease surveillance.

Clinical and Demographic Data

Between July 2016 and December 2017, patient data were collected for 30,500 admissions in 6 hospitals (Table 1). Median patient age was 2 years (range: 1 day to 14 years); 56% were male. Of admitted patients, 74% (n = 22,684) had on-schedule or complete childhood immunizations as assessed by providers on site. Most admissions (n = 26,573; 87%) were associated with documented fever (≥37.5°C) or history of fever. In accordance with national and international guidelines, most admissions for acute febrile illness (87%) were tested for malaria by blood smear, rapid diagnostic test, or both (n = 26,526). Overall prevalence of malaria among these hospitalized patients was 53%, ranging between 4% and 79% across hospitals (Table 1). Cough was reported for 62% (n = 18,839) of admitted patients (range: 58%-68% across hospitals), and diarrhea was reported for 27% (n = 8,151) patients (range: 20%-47% across hospitals). Gentamicin was the most commonly prescribed antibiotic, documented as treatment in more than 10,000 admissions; ceftriaxone was used in more than 8,500 admissions and ampicillin in more than 5,000 admissions.

Among 30,500 total admissions electronically captured between July 2016 and December 2017 at all 6 hospitals, 21,199 admissions (70%) occurred after blood culture capacity had been established in each respective hospital and additional associated variables were being collected. Accordingly, this reflected all admissions from the first sites activated in July and August 2016 and only a fraction from sites activated later in July and October 2017 (Table 1). Among these 21,199 admissions, 20% (n = 4,209) self-reported antimalarial use and 15% (n = 3,167) antibiotic use in the week before admission.

Laboratory Data and Mortality

A total of 5,001 single bottle blood cultures were performed at sentinel site hospitals through December 2017. Among these, 412 (8%) yielded bacterial growth; 187 were deemed pathogenic bacteria by the processing laboratory. Of these 187 cultures, 160 had species identities available in the web-based surveillance system at the time of writing (Table 2). An additional 225 bottles grew likely contaminants, specifically Bacillus spp., coagulase-negative staphylococci, Rhodococcus spp., and Micrococcus spp.

A broad range of pathogens was isolated. The most commonly isolated pathogens were Salmonella spp. (34%), Staphylococcus aureus (14%), Citrobacter spp. (8%), Escherichia coli (6%), Enterococcus spp. (5%), and Klebsiella pneumoniae (5%) (Table 2). Notably, although the emphasis of the system was to improve diagnosis for nonmalarial patients, and thus relatively few patients with malaria had blood cultures taken, pathogens were more commonly isolated from blood of malaria-positive children (35/578; 6%) than from malaria-negative children (142/4258; 3%) (p = 0.001).

Jinja RRH, east of Kampala, is the largest sentinel site in terms of patient volume and yielded the greatest diversity of pathogens (Table 2). Several of the pathogens identified to date originated commonly or exclusively from Jinja RRH, including E. coli, S. aureus, Pseudomonas aeruginosa, Enterococcus spp., Citrobacter spp., Acinetobacter spp., and Klebsiella pneumoniae. In contrast, among the 42 positive cultures isolated to date from Arua RRH, in the far northwest of Uganda, more than 70% yielded Salmonella spp. (Table 2).

Following initiation of hospital-based leptospirosis testing, 6% of patient samples screened showed IgM reactivity on rapid test, with some potential for geographic clustering (Table 1). Arboviral screening of 622 sera yielded 18 (3%) confirmed and presumptive-positive samples (Table 1). Fourteen samples were presumptive positive for recent flavivirus infection (eg, dengue, West Nile, yellow fever, or Zika). However, remaining sample volume was insufficient for PRNT confirmation of infecting virus. Positive samples originated from Jinja, Mubende, Arua, and Kabale. Of 4 samples with evidence of acute alphavirus infection, 2 were confirmed as o'nyong nyong virus infection, both originating from Arua. None of the sera screened for antibodies to Brucella spp. were positive. Among 293 patient sera screened in a Rickettsia africae IFA, 168 (57%) demonstrated some degree of positive reactivity, with 22% of samples having reactivity at titers at or above 1:256 (data not shown).

Based on data from the processing laboratory, resistance to commonly used antimicrobials was observed in several bacterial species; a subset of these findings is presented in Table 3. Among S. aureus isolates tested, 71% were susceptible to clindamycin, 40% susceptible to erythromycin, 12% susceptible to penicillin; 0 of 10 isolates screened were susceptible to cotrimoxazole (Table 3). All 15 S. aureus isolates screened against vancomycin were susceptible, but 25% screened were deemed methicillin resistant (MRSA) based on cefoxitin screen. S. pneumoniae isolates were susceptible to erythromycin, clindamycin, and chloramphenicol, but resistant to β-lactams (based on oxacillin screen). Salmonella spp. isolates demonstrated resistance to several agents, including ampicillin (39% susceptible), cotrimoxazole (40% susceptible), chloramphenicol (37% susceptible), and ciprofloxacin (51% susceptible) (Table 3); resistance to ceftriaxone was infrequently observed and only in nontyphoidal Salmonella. Other Enterobacteriaceae screened demonstrated resistance to third-generation cephalosporins, as well as to imipenem (Table 3).

Overall, all-cause mortality among pediatric admissions at sentinel hospitals following availability of blood culture was 4%, with a range across hospitals of 1% to 9% (Table 1). The case-fatality rate was 10% among those with positive blood cultures, 3% among those that had negative blood cultures, and 4% among those with confirmed malaria.