Predicting Salmonella Typhi incidence using prevalence metrics from sentinel studies of community-onset bloodstream infections

BackgroundTyphoid fever incidence estimates are central to policy decisions on vaccine introduction and investments in non-vaccine prevention and control but are often unavailable. We explored whether prevalence metrics from sentinel studies of community-onset bloodstream infections could accurately predict local Salmonella Typhi (S. Typhi) incidence. MethodsUsing a previous systematic review (January 2018-December 2024), we identified studies reporting both typhoid incidence and prevalence of community-onset bloodstream infections from sentinel sites. From authors, we requested data on blood culture isolates and analysed four metrics: (i) S. Typhi prevalence among probable pathogens, (ii) S. Typhi rank order, (iii) S. Typhi to Escherichia coli ratio, and (iv) S. Typhi to stably endemic organisms ratio. Typhoid incidence was categorized as low (<10), medium (10-100) or high (>100) per 100,000 person-years. We used univariate ordinal regression to assess the association between each metric and typhoid incidence level. The model performance was evaluated by the c-statistic, sensitivity, and specificity. FindingsAnalysis of 29 study sites (20 Africa, 9 Asia) yielded 4,625 probable pathogens. The median (IQR) typhoid incidence was 140 (28-319) per 100,000 person-years. All metrics were associated with increased typhoid incidence level: for each 1% increase in S. Typhi prevalence OR 1.07 (95%CI 1.02-1.15); rank order OR 0.25 (95%CI 0.06-0.64); log S. Typhi to E. coli ratio OR 2.91 (95%CI 1.45-7.42); log S. Typhi to stably endemic organisms ratio OR 3.69 (95%CI 1.69-11.3). A parsimonious model using S. Typhi prevalence alone achieved c-statistics of 0.87 (0.58-0.97), 0.76 (0.51-0.91), and 0.88 (0.69-0.96) for low, medium, and high incidence, respectively. InterpretationSentinel prevalence metrics from bloodstream infections, particularly S. Typhi prevalence among probable pathogens, could be useful for inferring local typhoid fever incidence where direct data are unavailable. FundingGates foundation Research in contextO_ST_ABSEvidence before this studyC_ST_ABSGlobally, annual deaths from typhoid fever were estimated at 71,954 (95% uncertainty interval 38,051 to 118,560) in 2023. Typhoid conjugate vaccines (TCV) are recommended for regions with high typhoid incidence. Implementation, however, can be challenging due to a lack of local incidence data. Generating community incidence estimates requires expensive and time-consuming large prospective or hybrid surveillance studies, or novel techniques such as serology or environmental surveillance. Our previous study proposed that metrics from sentinel healthcare facilities such as the prevalence of Salmonella Typhi (S. Typhi) among all bloodstream pathogens or its rank order relative to other pathogens could serve as proxy for community incidence. However, contemporaneous incidence and prevalence data from the same time and location were limited in our previous study. To explore typhoid incidence estimation strategies, we searched PubMed and MEDLINE on January 8, 2026 with search terms including keywords of "typhoid fever", "incidence", and "prediction" without restrictions to language or publication date. Previous studies estimated incidence based on complex country-level covariates and disease modelling that lack ease of applicability for policy decisions. Recognising the need for pragmatic tools, we explored whether prevalence metrics from sentinel studies of community-onset bloodstream infections could accurately predict local S. Typhi incidence. Added value of this studyOur study was based on typhoid incidence studies that had available data for isolates of bloodstream infections. Of 29 sites across Africa and Asia with 4,625 probable pathogens, we found that all four sentinel metrics were significantly associated with typhoid incidence level. We demonstrated that a parsimonious model using S. Typhi prevalence alone achieved good discriminative performance in identifying high incidence settings. Implications of all the available evidenceWhen typhoid incidence estimates are unavailable, prevalence metrics from sentinel studies of community-onset bloodstream infections could help policymakers infer typhoid incidence and optimise resource allocation in water, sanitation, and hygiene, and TCV introduction.