This study was approved by the institutional ethical review boards of Patan Hospital and The Nepal Health Research Council. All enrollees were required to provide written informed consent for the collection and storage of all samples and subsequent data analysis. In the case of those under 18 years of age, a parent or guardian was asked to provide written informed consent.

Patan Hospital is a 318-bed government hospital providing emergency and elective outpatient and inpatient services located in Lalitpur Sub-metropolitan City (LSMC) within the Kathmandu Valley (Figure 1). Enteric fever is common at the outpatient clinic at Patan Hospital, which has approximately 200,000 outpatient visits annually. The population of LSMC is generally poor, with most living in overcrowded conditions and obtaining their water from stone spouts or sunken wells (Figure 1).

All febrile patients attending the outpatient or emergency department between April and October 2011 with a non-focal fever lasting 3 or more days, aged between 2–65 years and providing informed consent were eligible for this study. All individuals received a blood culture and only those with a blood culture positive for S. Typhi or S. Paratyphi A were enrolled.

Community-based controls were matched for age, sex and residential ward and enrolled at a ratio of 3∶1. Controls were identified in households neighboring cases. If the case lived in a stand-alone house, the household to the right of the “case-household” was approached by a community medical assistant (CMA) within 2 weeks of the case enrollment. If this control refused, the household to the left was approached, followed by the house parallel across the street. If the case lived in multi-story building, the household above the “case-household” was approached. If this control refused, a household a storey either below or two stories below was approached. Controls were required to be within 5 years of the age of the case and must not have had fever, gastrointestinal disturbances or history of enteric fever in the month before administration of the questionnaire. If an approached control failed to meet the enrollment criteria or refused participation, the house to the left or the storey below the “case-household” was approached for enrollment. The trained CMAs administered a 129-question questionnaire to each enrolled individual.

Anti-coagulated blood samples were collected from all febrile patients upon arrival in the outpatient department. For those over the age of 12 years, 10 ml of blood sample was collected; 5 ml was collected from those aged 12 years or less. The blood samples were inoculated into tryptone soya broth and sodium polyethanol sulphonate up to 50 ml. The inoculated media was incubated at 37°C and examined daily for bacterial growth over seven days. On observation of turbidity, the media was sub-cultured onto MacConkey agar. Any bacterial growth presumptive of S. Typhi or S. Paratyphi was identified using serotype specific antisera (Murex Biotech, Dartford, UK).

ELISAs to measure IgG against the Vi and O:2 antigens (NVGH, Siena, Italy) [23], [24] were performed on 795 plasma samples that were age-stratified and randomly selected from a serum bank comprised from the blood of patients attending Emergency Department of the Patan Hospital for reasons other than typhoid treatment or those relating to a febrile illness between January 2009–December 2011. Ages of enrollees ranged from 0–65 and were resident in the same demographic area as the case/control enrollees. All plasma samples were subjected to ELISA detecting IgG antibodies to both Vi and O:2. Briefly, plasma was diluted into 1∶200 and aliquoted into ELISA plates (Nunc, Sigma-Aldrich Co, UK) independently coated with Vi and O:2 antigens. After washing, bound IgG was detected using an alkaline phosphatase–conjugate antiserum (Sigma-Aldrich Co, UK). Antibody levels were quantified using standard curves. The standard curve for Vi-antibody was created using a prepared anti-human Vi-antibody standard. The standard curve for O:2-antibody was created using a pool of plasma from S. Paratyphi A confirmed patients who had high levels of antibody to O:2, which had been screened prior to the serum bank. The cutoff value of the ELISAs was defined as the optical density of blank control wells plus two standard deviations.

Data were imported into STATA v9.2 (College Station, TX, USA). The association between the outcome of enteric fever (defined as infection with either S. Typhi or S. Paratyphi A) and each exposure of interest was examined using a matched univariate analysis through conditional logistic regression. A conceptual model was generated to develop a biologically plausible set of covariates a-priori that were thought likely to influence the outcome, including those related to poor socioeconomic status and poor water quality. Variables were included in a matched multivariate analysis through a-priori selection or were associated (p<0.25) with enteric fever in the univariate analysis [25]. Model fit was assessed through log likelihood and relative AIC value. Co-linearity among variables was assessed but no strong associations were found between the variables included in the final model. The presence of effect modification was evaluated between each of the salient exposures and confounders of interest through the χ² test for homogeneity. Several sets of interactions were found to be significant. Each interaction term was included in the final model of interest and model fit was assessed. However, due to small numbers of patients in many of the strata, inclusions of these interactions led to an unstable estimate and were thus discarded. One interaction, however, household size affecting the risk of water storage on outcome of enteric fever, improved model fit and led to stable estimates so was included. The categorical variable household size was generated through assessing whether the house was the same size or smaller than the median number of people (12) in this dataset or larger than the median. All P values are two-sided.

During the period of investigation 103 febrile patients with culture confirmed enteric fever were enrolled, 48% (49/103) were positive for S. Typhi and 52% (54/103) were positive for S. Paratyphi A. Baseline characteristics of the enteric fever cases are described in Table 1. Briefly, those with culture confirmed enteric fever were more often male (64%; 66/103) and had a median age of 18 years (interquartile range (IQR) 10–23 years). Males and females with S. Paratyphi A did not differ significantly in age (median: 18 years, IQR: 5–55 and 20 years, IQR: 6–28, respectively), but female S. Typhi cases were significantly older (median: 21.5 years, IQR: 8–50) than male S. Typhi cases (median: 16 years, IQR: 7–32) (P = 0.03, Mann-Whitney U test). A total of 294 controls were enrolled, leading to a final case-control ratio of 1∶2.85.

The clinical presentations of S. Typhi and S. Paratyphi A were largely indistinguishable, with most cases exhibiting a progressive fever (77%; 79/103), nausea (50%; 51/103), and a limited number having an abdominal rash (2%; 2/103) or constipation (6%; 6/103). Patients with S. Typhi were more likely to present with abdominal pain (61%; 30/49) than those with S. Paratyphi A (33%; 18/54) (P = 0.005, χ² test) and to have diarrhea (25%; 12/49 and 9%; 5/54, respectively (P = 0.038, χ² test)).

To identify important associations between various exposures and the outcome of enteric fever due to either S. Typhi or S. Paratyphi A, we performed a series of univariate analyses and, to control for confounding, built a multivariate model. As shown in Table 2, a variety of exposures were found to be protective against enteric fever in a univariate analysis. These protective variables included, an awareness of enteric fever, reporting a previous enteric fever episode, recent contact with an enteric fever case, using a metal cover on a household water storage container, and the consumption of pani puri. The counterintuitive protective effects of recent contact with an enteric fever case and consumption of pani puri were considered to be a result of study design. As cases and controls were geographically matched, it is likely that controls were aware of local cases. Additionally, as pani puri has the potential to be fecally contaminated [26], the protective association is likely to be explained by uncollected information. These exposures were not included in the final model. Enteric fever risks from the univariate analysis included a household monthly income of <$125, the duration of residence in Kathmandu, the use of stone spout water, storing water in the household, the use of a household latrine, and the recent consumption of street food. From the multivariate model, an awareness of enteric fever (AOR: 0.25, 95%CI: 0.1–0.5, p<0.001) and the use of a metal cover on household water storage (AOR: 0.28, 95%CI: 0.1–0.7, p = 0.006) remained significantly protective against infection. The use of a household latrine, compared to a community latrine (AOR: 4.10, 95%CI: 1.4–12.4, p = 0.013), as well as recent street food consumption (AOR: 2.85, 95%CI: 1.4–6.0, p = 0.006) remained as strong risk factors for infection. Notably, the risk of storing water was found to vary by household size. Those living in households with more than 12 people were at significant risk of infection if they stored water (AOR: 9.35, 95%CI: 1.7–51.0, p = 0.010), while those who lived in smaller houses were unaffected by water storage habits.

From the series of univariate analyses performed on S. Paratyphi A cases and matched controls only, many of the identified protective and risk factors, as shown in Table 3, were similar to the overall enteric fever analysis shown in Table 2. However, the final multivariate model demonstrated that the use of a metal cover for water storage (AOR: 0.19, 95%CI: 0.1–0.7, p = 0.014) remained strongly protective against infection with S. Paratyphi A, whereas the consumption of street food within the two weeks preceding illness remained a significant risk factor (AOR: 2.95; 95%CI: 1.1–7.8, p = 0.028). Furthermore, residing in Kathmandu for less than two years (AOR: 2.29, 95%CI: 0.9–5.7, p = 0.077) remained a weakly significant risk factor for S. Paratyphi A infection only.

Many of the important protective and risk factors for S. Typhi infection only were again similar to the overall enteric fever analysis, however some S. Typhi-specific exposures emerged. Firstly, the use of stone spout water and household water storage were strong risk factors for S. Typhi infection in the univariate analysis, although the use of stone spout water was only mildly significant in the multivariate model (AOR: 4.17, 95%CI: 0.9–20.5, p = 0.078). Water storage was a risk but was not significantly associated with S. Typhi infection in the multivariate model (AOR: 2.56, 95%CI: 0.8–8.6, p = 0.129). Living in a house with more than 12 people was also a risk that did not remain strongly significant in the multivariate model (AOR: 2.11, 95%CI: 0.8–5.5, p = 0.127), although reporting a monthly income of less than $125 was strongly and independently associated with S. Typhi infection (AOR: 5.21, 95%CI: 1.5–18.4, p = 0.010). Awareness of enteric fever was strongly protective, specifically against infection with S. Typhi (AOR: 0.28, 95%CI: 0.1–0.8, p = 0.014).

To assess exposure to S. Typhi and S. Paratyphi A in the local population, we measured IgG against Vi antigen (S. Typhi) and O:2 antigen (S. Paratyphi A) in 795 age-stratified (0–65 years) serum samples derived from the same population as our case/control enrollees (Figure 2). The resulting data demonstrated a consistently high level of IgG against Vi and O:2 in all age groups. IgG against S. Typhi was highest at birth and then declined with a secondary peak at the age of 17–18 years. Antibody against S. Paratyphi A was lowest at birth and peaked at the age of 11–12 years and subsequently declined; yet persisted into old age. There was a weak correlation between levels of IgG to Vi and to O:2 (Spearman's ρ: 0.30, P<0. 001), this was most apparent within the group aged 11–20 years (Spearman's ρ: 0.50, P<0.001) (Figure 2).