The study was conducted in the Asante Akyem/Akim North district of the Ashanti Region of Ghana. The district covers an area of 1,099.7 km² with majority of people living in the peri-urban areas. The demography of the district is well characterized with an adjusted study population of approximately 109,840 based on a recent updated census of the study area. We selected ES sites from rural and peri-urban towns in the district. The rural areas are Hwidiem, Domeabra and Juansa while the peri-urban area is Agogo. Agogo is the main administrative area in the district and most populated. Clinical surveillance, supported by the Severe Typhoid in Africa surveillance study (SETA), found an adjusted incidence of febrile typhoid of 112 (84–164) cases per 100,000 persons-years as of 2019 [6]. Currently, there is an ongoing Phase-IV cluster-randomised controlled trial which is evaluating TCV effectiveness in Ghana [12]. The clinical trial is embedded within the typhoid surveillance project and allows for enhanced surveillance of febrile illness at various locations including pharmacy shops, health centres, traditional healing centres and others.

This was a longitudinal study that was conducted over a 14-month period from July 2022 to August, 2023. Samples were collected from 40 sites located in rural and peri-urban areas of the Asante Akim North district of the Ashanti region. Collection of GS and MS was completed for the initial 6-months of the study to determine the sensitivity of both methods for the detection of S. Typhi. Thereafter, GS was dropped and MS sampling was continued for an additional 8 months.

Using a previously published protocol, a geographic information systems-based approach using environmental data was implemented to systematically identify candidate ES sites at wastewater confluence points [13]. Briefly, sewage drainage channels were digitally mapped and geospatially referenced. These data were coupled with a high-resolution digital elevation model (DEM) to execute hydrological reconditioning used to identify and down select candidate ES site locations, delimit catchment boundaries and generate population estimates based on the High-Resolution Settlement Layer (HRSL). ES sites were screened to confirm broad geographic coverage, ensure a variety of catchment population sizes, and capture peri-urban and a limited number of rural areas. Field visits were conducted to assess suitability of candidate ES sites for inclusion in the study. Using our previously published and implemented study protocol [9], we estimated that this study would require 40-45 ES sites, each generating monthly repeat samples for 1 year to allow comparison with other sites in Africa and Asia conducting the same protocol.

Prior to initiating this study, we engaged local communities to provide education about the purpose and importance of this study. The process of sample collection was explained to the local communities, and they were encouraged to collaborate with the research team and offer them assistance in locating the various study sites. To encourage local support for this study, we recruited community members as research assistants and trained them in collection and packaging of the samples for transportation to the laboratory.

Ethics approval for this study was obtained from the Committee on Human Research Publication and Ethics (CHRPE) of the School of Medical Sciences, Kwame Nkrumah University of Science and Technology (KNUST), Kumasi, Ghana. Written informed consent was obtained from participants above 18 years who were enrolled in the typhoid surveillance Child assent and parental/guardian consent were obtained for children less than 18 years. The anonymity and confidentiality of the participants were assured using unique codes. Participants who did not provide consent were excluded from the study.

For the environmental surveillance programme, we did not seek or request for informed consent because this did not include human subjects.

GS and MS sampling methods were used for the collection of wastewater from the study sites. For MS, standardized swabs were deployed in wastewater and sewage 24 hours before collection. At collection, MS were placed in 450 ml of universal pre-enrichment broth (UPE; Fisher Scientific, Heywood, UK,) and grab samples were collected in 1L bottles.

Both GS and MS were transported via cold chain to the Genomic and Infectious Disease Laboratory of KNUST. The protocols and methods for MS and GS have been described previously [14,15]. The methods were developed through an expert advisory group established by the Bill and Melinda Gates Foundation (BMGF).

Upon sample arrival in the laboratory, Moore swabs were incubated overnight at 37 ± 5 ºC for approximately 18-22 hours. About 40 ml of the samples was transferred in a 50 ml falcon tube and filtered through a 0.45 µm membrane (Millipore, Cork, Ireland). The membranes were cut into 5-6 pieces using sterile forceps and scissors and stored in a cryotube for later extraction. GS were similarly filtered immediately upon arrival in the laboratory. Up to 5 filters were used for sequential filtration of the GS with each filtration process lasting about 1 hour for a total of 4-5 hours. The filters were massaged in Ringer’s lactate and the supernatant was centrifuged at 10,000xg for 10 mins. The pellets were stored at -20^oC until DNA extraction.

Physicochemical parameters for each sampling site were also measured using Aquaprobe Model AP-2000 (Aquareed, UK). Parameters measured include temperature (air and water), pH, salinity, seawater specific gravity, dissolved oxygen (DO), turbidity (TUR), electrical conductivity (EC), and Oxidation-reduction potential (ORP). The calibration protocol for all sensors of the probe was carried out before the use of the probe in the field according to the manufacturer’s protocol. We also measured the depth of each of the ES sites and classified this as shallow (<5 cm), medium (5-50 cm) and deep ( > 50 cm). The flow speed of the ES sites were determined by observation.

DNA extraction was performed using Qiagen QIAamp PowerFecal Pro DNA extraction kit (Qiagen, Hilden, Germany). An extraction control (Cy5-QXL670; Eurogentec) was adjusted to cycling thresholds of between 30-33 and 1µl was used for all extraction processes. The DNA extraction was carried out according to the manufacturer’s protocol and finally eluted in a volume of 50µl.

Quantitative Realtime PCR was performed on the extracts using CFX 96 RT-PCR machine (Biorad, Singapore). The qPCR was performed using primers and probes targeting the ttr, tviB and staG genes. A positive signal for all three targets were considered as positive. However, if the sample was positive for ttr and only one of either staG or tviB, the target that was negative was repeated in a singleplex reaction. We also set up separate reactions for the detection of HF183 (human restricted faecal indicator organism Bacteroides dorei) in all samples. All study protocols including primer sequences can be found at https://www.protocols.io/workspaces/typhoides.

Concurrent with the ES study, blood culture surveillance was conducted on patients presenting with febrile illness with a history of fever or acute fever. Patients were included in our analysis if they resided within the catchment area of the ES. Fever was measured objectively using standardized thermometer via tympanic membrane or axillary.

For all patients who met the inclusion criteria, blood culture samples were collected into BACTEC BD bottles (BD, Franklin Lakes, NJ USA) and incubated in FX40 BD blood culture system (BD, Franklin Lakes, NJ USA) at 35°C for five days or until positive. Presumptive positive blood culture vials signalled by the BACTEC were removed and processed for bacterial identification. Bacterial cultures were performed by plating/streaking the blood unto blood agar (BA – Columbia agar base supplemented with 5% sheep blood), chocolate agar (CA) and MacConkey agar (Becton and Dickinson, USA). All the plates were incubated aerobically overnight at 35-37°C except for CA plates that were incubated in 5% CO₂ for microaerophilic condition. Isolates suspected to be Salmonella were identified based on colonial morphology on the various agar, microscopic presentation, latex agglutination test and biochemical tests including Sulphur Indole Motility (SIM) test and Analytical Profile Index (API) 20E.

Data capture system was designed for field and laboratory activities. For field activities, a standard questionnaire was developed, and this was used to capture study information. Laboratory data were captured on laboratory forms. Both field and laboratory data were entered on an electronic database system (REDCap) and cleaned for errors. Data was exported in a comma separated format and imported into R statistical software (version 4.4.1) for analysis. Descriptive statistics were computed for continuous and categorical variables. Statistical comparisons between subgroups of categorical variables were analyzed using the Fischer’s exact test or Chi-square test where appropriate. We determined the various ES site characteristics (speed of flow, depth of wastewater, width of wastewater, level of dissolved oxygen, pH etc) associated with S. Typhi detection by performing univariate analysis for the first 6 months of comparative sampling of both MS and GS. Variables that were significant at the univariate level were fitted in two multivariate mixed effect logistic regression models for subsets of MS data and GS data, using site-specific random effect. Results were expressed as odds ratio with 95% confidence interval (CI). In order to determine the association between environmental S.Typhi detections and blood culture detections, we matched the proportion of monthly detections of environmental S.Typhi detections by the blood culture S.Typhi detection and computed this using generalized linear logistic regression model. We have included this in data analysis as an additional information. For all analysis, a p-value of less than 0.05 was considered statistically significant.

Of the 40 sites selected, 35 were located in Agogo (peri-urban)) and the remainder from rural sites; Hwidiem (1), Domeabra (1) and Juansa (3). The 40 ES sites were drawn from 30 clusters which were defined as part of the typhoid phase-IV randomized controlled trial [12]. Table 1 describes the population, ES characteristics and catchment areas included in this study.

Seven hundred and twenty (720) samples comprising 240 (33%) GS and 480 (67%) MS were collected. During the initial six months of the study, 240 each of GS and MS were collected. The GS collections were suspended after it was observed that the sensitivity for detecting S. Typhi was low compared to MS. MS collections continued for the subsequent 8 months generating an additional 240 MS samples. Most of the samples were collected in peri-urban areas (Agogo) due to previous reports of high endemicity of typhoid in those communities. The rural area of Domeabra recorded the fewest number of S. Typhi detections while Agogo recorded the highest (Table 2).

A plot of the geographic distribution of the S. Typhi positives at the ES sites show wide distribution of detections (Fig 1).

Similar to S. Typhi detections, we also recorded a higher prevalence of human faecal contamination (HF183) in the first 6-months in Moore swabs (225/240; 93.7%; 95%CI 88.8-97%) compared to grab samples (158/240; 65.8%; 95%CI 56-74%) and the difference was statistically significant (p < 0.001). We also observed a statistically significant (p < 0.01) higher concentration of faecal contamination in Moore swab (146,194; IQR = 29508.7-658056.1) compared to grab samples (1163.4; IQR = 279-8805.1). Similarly, the concentration of faecal contamination in S. Typhi positive MS samples (269,728.4; IQR = 44692.9-839123.8) was significantly higher (p = 0.009) than S. Typhi negative MS samples (67882.2; IQR = 18,180.8–503,587.1) (Fig 3). However, for GS, there was no significant difference in the concentration of faecal contamination levels for S. Typhi positive and negative subjects.

This figure describes the log concentrations of the HF183 detections in MS and GS sampling. Concentrations in MS are generally higher than GS.

Based on a univariate analysis that compared S. Typhi detections associated with the various wastewater characteristics, we performed a multivariate analysis for MS and GS separately. The univariate variables analyzed include towns of ES site, flow speed, depth of wastewater, dissolved oxygen, width of wastewater, pH, electrical conductivity, total dissolved solids and salinity. For MS, the variables that were significant and included in the multivariate model were pH, dissolved oxygen and flow speed. In the multivariable logistic regression, we found that the odds of S. Typhi detection for MS was significantly associated with high pH (>7) (OR = 3.08; 95%CI 1.69–5.62) and having width of wastewater site being more than 1m (OR = 2.14; 95%CI 0.67–46.84). We did not find an association between any of the predictor variables with S. Typhi detection in grab samples.

A univariate analysis of HF183 detections identified pH as being significant in MS. For GS, we observed the Flow speed, dissolved oxygen, width of wastewater, electrical conductivity, total dissolved solids and salinity were significant. Inclusion of these variables in a multivariate model identified only pH (OR = 6.56; 95%CI 1.4–30.8) as independently associated with HF183 detection in MS. For GS, only dissolved oxygen (OR = 0.38; 95%CI 0.16–0.90) was identified as independently associated with HF183 detection.

For the 14 months study period, we recorded an S. Typhi detection rate of 42.1% (202/480; 95%CI 35-50%) based on MS collections. During the same study period, we also collected a total of 5,576 blood cultures in the catchment area of the ES study. The median age of study subjects was 7 years (IQR: 2-19). Of the 5,576 subjects recruited, 2,538 (45.5%) were 5 years and below, 1,426 (25.6%) were between 6 and 15 years and 1,612 (28.9%) were more than 16 years. A total of 12 S. Typhi were isolated in blood cultures; 5 (41.6%) from subjects 5 years and below, 5 (41.6%) from those between 6 and 15 years and 2 (16.7%) in those above 15 years of age. The crude incidence per 100,000 population for S. Typhi in the blood culture surveillance was 11 (95%CI 4.8–20.0). Generally, the prevalence of S. Typhi in the environment was higher than among hospital-based blood cultures. Analysis of this relationship showed a negative association between S. Typhi detection in the environment and clinical cases of typhoid (b = -0.33; p = 001). Fig 4 shows the variation of environmental S. Typhi detections and blood culture-based detections.

To better understand the temporal variability, and potential impact of climate variables on prevalence of S. Typhi in ES samples, we examined the monthly rainfall patterns over the 14-month period based on MS detections (Fig 5).

Overall, we observed statistically significantly higher rates of detection of S. Typhi during the wet or rainy seasons (June-August and September-October) (p = 0.012). Conversely, low levels of detections occurred predominately during the dry season.

Logistic regression models were further used to investigate the effect of daily rainfall on S. Typhi detection rates using separate measures of cummulative precipitation and wet day classification for multiple lag period. For Moore swabs, crude analysis reveals that measures of both cummulative precipitation and wet day designation were significantly positively associated with S. Typhi detection at the 0-, 1- and 2-day lag scales of analysis (Fig 6A and 6B). The strength of association was relativily consistent for both analyses with mean ORs of 1.06 (95% CI: 1.02-1.10) and 2.01 (95%CI: 1.40-2.88) for precipitation and wet day classification, respectively. For GS, we oberved significant positive associations between precipitation and wet day and S. Typhi detection at the 1-day lag analysis scale only (Fig 6C and 6D).