Bangladesh's dengue epidemic left the cities, and the season moved with it

Dengue control in Bangladesh has been built around a specific mental model: an urban disease, concentrated in Dhaka, peaking with the monsoon. Two papers published within three weeks of each other in June describe how much of that model no longer holds.

The geographic shift

The first analysed national dengue surveillance data from 2000 to 2025, alongside environmental data from the World Bank Climate Portal and national meteorological databases, comparing the period 2000–2018 against 2019–2025 [s1].

The clearest signal is the rural-to-urban incidence ratio. Between 2000 and 2018 it ranged from 0.11 to 0.41 — rural incidence running at a tenth to two-fifths of urban [s1]. Between 2019 and 2025 it ranged from 0.86 to 2.12 [s1].

The upper end of that second range is the part that matters. A ratio above 1.0 means rural incidence exceeded urban incidence. Dengue in Bangladesh is no longer reliably a city disease.

The same analysis found standardised incidence ratios declining in major hotspot cities even as nationwide cases rose sharply [s1]. That combination — falling urban concentration, rising national total — is what a geographic diffusion looks like in surveillance data.

The seasonal shift

After 2019, the seasonal peak shifted toward October and November, and incidence from January through June rose significantly compared with the earlier period [s1].

A later peak and a heavier off-season have practical consequences that a change in total case count does not. Vector control campaigns, hospital surge staffing, blood product stocking and public messaging are all scheduled against an expected curve. A curve that has moved by weeks and flattened at the shoulders means those preparations land at the wrong time.

Associations between dengue incidence and temperature, humidity and rainfall strengthened slightly in the post-2019 period, with correlation coefficients above 0.5 [s1]. That is a moderate correlation, and correlation between an environmental variable and a disease count across a country is a weak instrument for attributing the shift to climate specifically — many things changed between 2018 and 2019, including surveillance intensity.

What happens to patients

The second paper looks at the same epidemic from inside four Dhaka hospitals, following patients through an outbreak year [s2].

Between January and November 2024, the study screened 3,135 dengue patients and enrolled 354 who presented with warning signs — the clinical category for a patient at risk of deterioration but not yet classified as severe [s2]. Patients were followed with clinical data and blood samples at enrolment, at clinical deterioration or discharge, and at convalescence around day 21 [s2].

Sixty-two of the 354, or 17.5%, progressed to severe dengue [s2]. There were two deaths [s2].

Roughly one in six patients admitted with warning signs got worse. That is the number that determines how many hospital beds an outbreak of a given size requires.

The demographics are worth noting: 60% of patients were aged 21 to 40, and 76% were male [s2]. A male-skewed, working-age patient population in a hospital series usually reflects care-seeking patterns as much as infection patterns.

DEN-2 was the predominant serotype at 90% [s2].

The predictors

Multivariable logistic regression, adjusted for confounders, identified six independent predictors of progression to severe dengue [s2]. The published abstract reports confidence intervals rather than point estimates for each:

• Arthralgia (95% CI 1.04–3.98) [s2]
• Anorexia (95% CI 1.40–11.7) [s2]
• Lethargy (95% CI 1.97–10.7) [s2]
• Diabetes mellitus (95% CI 1.01–6.17) [s2]
• Pleural effusion (95% CI 4.97–21.1) [s2]
• Prolonged APTT (95% CI 1.02–1.08) [s2]

Every interval excludes 1.0, which is what makes them independent predictors. But note how wide several of them are: anorexia and pleural effusion span an order of magnitude. These are usable as risk flags, not as calibrated probabilities.

Lethargy is the one with the clearest bedside signal in the raw data: present in 29% of patients who progressed to severe dengue against 11% of those who did not [s2].

The authors' framing is deliberately modest and, for a resource-constrained system, exactly the right ambition: a few simple, easily measurable indicators can support early risk assessment and guide timely admission [s2]. None of the six requires equipment beyond a basic ward.

The two papers together

Read side by side, they describe a compounding problem.

The disease is moving into rural and semi-urban areas [s1] — which is where diagnostic capacity, ICU beds and clinicians experienced in fluid management for severe dengue are thinnest. The severe-progression rate among patients with warning signs is high enough that a meaningful proportion of cases need inpatient care [s2]. And the timing of the epidemic has shifted [s1], so the surge arrives when systems are calibrated for something else.

The first paper also notes, from available surveillance records, rapid clinical deterioration within 24 to 72 hours of diagnosis among a subset of fatal cases [s1]. That window is short, and it is the argument for the second paper's screening indicators.

Limits

Both are observational. The surveillance analysis compares two eras of a reporting system that itself changed over 25 years, and rising rural case counts are partly a function of rural surveillance improving — a bias the study cannot fully separate from real geographic spread [s1]. The hospital study is four facilities in one city in one outbreak year, with a single dominant serotype [s2]; DEN-2 predominance at 90% means the progression rate and predictors may not transfer to a season dominated by a different serotype.

The Dhaka study was funded by USAID, under a crisis modifier grant issued in response to the dengue outbreak in Bangladesh [s2] — a detail worth recording, given what has happened to that funding channel.

What to watch

Whether the rural-to-urban ratio stays above 1.0 in subsequent seasons or was elevated only in particular years [s1], and whether the October–November peak holds. Both are answerable from the same national surveillance system within a season or two, and both determine where Bangladesh needs to put clinicians before the next epidemic year rather than during it.

This article is informational and does not constitute medical advice.

Sources

• [s1] Khan A, Sharif N, Sharif N, Dey SK., "Spatiotemporal trends, evolving seasonality and symptoms of dengue in Bangladesh, 2019-2025," New Microbes and New Infections, published online 11 June 2026. https://doi.org/10.1016/j.nmni.2026.101793
• [s2] Shahrin L, Sarmin M, Alam T, et al., "Progression of dengue during an outbreak (2024) in Dhaka, Bangladesh: a hospital based longitudinal study," The Lancet Regional Health – Southeast Asia, published online 30 June 2026. https://doi.org/10.1016/j.lansea.2026.100809