Home > Research > Publications & Outputs > Model-based geostatistics enables more precise ...

Text available via DOI:

View graph of relations

Model-based geostatistics enables more precise estimates of neglected tropical-disease prevalence in elimination settings: mapping trachoma prevalence in Ethiopia

Research output: Contribution to journalJournal articlepeer-review

E-pub ahead of print
Close
<mark>Journal publication date</mark>13/11/2021
<mark>Journal</mark>International Journal of Epidemiology
Publication StatusE-pub ahead of print
Early online date13/11/21
<mark>Original language</mark>English

Abstract

Background
As the prevalences of neglected tropical diseases reduce to low levels in some countries, policymakers require precise disease estimates to decide whether the set public health targets have been met. At low prevalence levels, traditional statistical methods produce imprecise estimates. More modern geospatial statistical methods can deliver the required level of precision for accurate decision-making.

Methods
Using spatially referenced data from 3567 cluster locations in Ethiopia in the years 2017, 2018 and 2019, we developed a geostatistical model to estimate the prevalence of trachomatous trichiasis and to calculate the probability that the trachomatous trichiasis component of the elimination of trachoma as a public health problem has already been achieved for each of 482 evaluation units. We also compared the precision of traditional and geostatistical approaches by the ratios of the lengths of their 95% predictive intervals.

Results
The elimination threshold of trachomatous trichiasis (prevalence ≤ 0.2% in individuals aged ≥15 years) is met with a probability of 0.9 or more in 8 out of the 482 evaluation units assessed, and with a probability of ≤0.1 in 469 evaluation units. For the remaining five evaluation units, the probability of elimination is between 0.45 and 0.65. Prevalence estimates were, on average, 10 times more precise than estimates obtained using the traditional approach.

Conclusions
By accounting for and exploiting spatial correlation in the prevalence data, we achieved remarkably improved precision of prevalence estimates compared with the traditional approach. The geostatistical approach also delivers predictions for unsampled evaluation units that are geographically close to sampled evaluation units.