Weighted prevalence (infections ÷ examined) fell from ~46% in 1985 to ~8% by 2020. Read the line alongside the bars: years with very few surveys (e.g. 1992, 1996–97) rest on little data. The 2008–2010 and 2015/2020 peaks in survey volume reflect national school-based survey programmes.
Bars: surveys started that month · Line: weighted prevalence
Rural vs urban prevalence by year
Weighted prevalence of infections
Microscopy vs RDT by year
Weighted prevalence of infections
Seasonality is visible even in this survey-based data: fieldwork and reported prevalence are highest in the rainy-season months (Feb–May) and lowest from July and November — consistent with Kenya's bimodal transmission pattern. Rural sites carry roughly twice the prevalence of urban sites in most years.
Geography of transmission
1,607 sites with coordinates
Transmission is highly uneven in space. High prevalence clusters in western Kenya (Lake Victoria basin) and along the coast (Kilifi, Kwale, Malindi), while central, eastern and northern sites sit near zero — matching the documented ecology of malaria in Kenya.
Survey sites across Kenya
Point colour = prevalence (%) · point size = number examined · hover for details
0%25%50%75%100%
Prevalence by approximate zone
Zones derived from coordinates; hover for survey counts
Rapid diagnostic test brands recorded in the dataset
Test
Surveys
Share
Breakdowns
Setting · Method · Age · Species
Individuals examined by setting
Share of examined population
Individuals examined by method
Microscopy vs rapid diagnostic tests
Surveyed population by age group
Who each survey sampled
Prevalence by diagnostic method
Distribution of per-survey prevalence (box = IQR, whiskers = min–max)
Prevalence by setting
Distribution of per-survey prevalence
Prevalence by latitude band
Altitude proxy — lowlands vs highlands gradient
Prevalence by survey sample size
Weighted prevalence per examined-size band
All — surveys report Plasmodium falciparum as the species. Microscopy surveys show a higher median prevalence than RDT surveys in this dataset, but the two methods are also used in different places and years — they are not directly interchangeable (see the model, below).
Sample size & reliability
Why small surveys are risky
Sample size vs estimated prevalence
Each point is one survey; orange line is the smoothed trend
Survey summary statistics
Per-survey distribution (current filter)
Surveys examining fewer than ~20 people produce the most extreme estimates — both the highest and lowest prevalence values. This is why weighted summaries (by individuals examined) are preferred over simple averages of sites.
Statistical model
Binomial logistic regression
Model: positive/examined ~ year + setting + method (RURAL and Microscopy as reference levels). Fitted to the — surveys with a recorded setting. Each odds ratio is adjusted for the other variables.
Adjusted odds ratios (95% CI)
Characteristic
OR
95% CI
p-value
Read with care: survey sites are not a random sample of Kenyan communities, and sites cluster spatially (e.g. Kilifi, western Kenya). These are associational, not causal, estimates.
What the model says
Odds of infection per additional year
—
≈ —% decline per year, holding setting & method constant
Urban vs rural
—
≈ half the odds of rural sites
RDT vs microscopy
—
adjusted for year & setting
Model fit: deviance —, AIC —
Full dataset explorer
—
All surveys (filtered)
Search any site name; rows follow the slicers above