Spatial variation in time to diagnosis of visceral leishmaniasis in Bihar, India
Background Visceral leishmaniasis (VL) is a debilitating and—without treatment—fatal parasitic disease which burdens the most impoverished communities in northeastern India. Control and ultimately, elimination of VL depends heavily on prompt case detection. However, a proportion of VL cases remain undiagnosed many months after symptom onset. Delay to diagnosis increases the chance of onward transmission, and poses a risk of resurgence in populations with waning immunity. We analysed the spatial variation of delayed diagnosis of VL in Bihar, India and aimed to understand the potential driving factors of these delays.
Methods The spatial distribution of time to diagnosis was explored using a Bayesian hierarchical model fit to 4270 geo-located cases notified between January 2018 and July 2019 through routine surveillance. Days between symptoms meeting clinical criteria (14-day fever) and diagnosis were assumed to be Poisson-distributed, adjusting for individual- and village-level characteristics. Residual variance was modelled with an explicit spatial structure. Cumulative delays were estimated under different scenarios of active case detection coverage.
Results The 4270 cases analysed were found to be prone to excessive delays in areas outside existing endemic ‘hot spots’. After accounting for differences associated with age, HIV status and mode of detection (active versus passive surveillance), cases diagnosed within recently affected (≥ 1 case reported in the previous year) blocks and villages experienced shorter delays on average (by 13% [2.9–21.7%] (95% credible interval) and 7% [1.3–13.1%], respectively) than those in non-recently-affected areas.
Conclusions Delays to VL diagnosis when incidence is low could influence whether transmission of the disease could be interrupted or resurges. Prioritising and narrowing surveillance to high-burden areas may increase the likelihood of excessive delays in diagnosis in peripheral areas. Active surveillance driven by observed incidence may lead to missing the risk posed by as-yet-undiagnosed cases in low-endemic areas, and such surveillance could be insufficient for achieving and sustaining elimination.