Related Experiment Videos
Sampling designs for xerophthalmia prevalence surveys
1Department of International Health, Johns Hopkins School of Hygiene and Public Health, Baltimore, MD 21205-2103, USA.
International Journal of Epidemiology
|November 18, 1997
Summary
The Expanded Program on Immunization (EPI) sampling design can accurately estimate xerophthalmia prevalence with minimal bias. EPI sampling with 15 children per cluster is efficient for travel distance and mean square error, especially in Nepal.
Area of Science:
- Epidemiology
- Public Health
- Statistical Sampling
Background:
- Xerophthalmia remains a public health concern, necessitating accurate prevalence estimation.
- The Expanded Program on Immunization (EPI) sampling design is widely used but its suitability for estimating specific health outcomes like xerophthalmia requires evaluation.
- Understanding the efficiency and potential biases of EPI sampling is crucial for resource allocation in health surveys.
Purpose of the Study:
- To assess the bias and design effects of the EPI sampling strategy for estimating xerophthalmia prevalence.
- To quantify the potential travel distance savings using the EPI design compared to traditional methods.
- To evaluate the impact of cluster size and house selection methods on survey efficiency and accuracy.
Main Methods:
- Computer simulations were employed using survey data from 40 wards in Sarlahi district, Nepal.
- The EPI sampling strategy was simulated with varying fixed cluster sizes (7, 10, 15, 20, 25).
- Bias was calculated by comparing estimated prevalence to true prevalence, and design effects were compared to stratified random sampling (SRS).
Main Results:
- Xerophthalmia prevalence was estimated at 2.8%, with EPI design showing an overestimation between 0.27% and 1.16%.
- Design effects ranged from 0.73 to 1.35, with no significant relationship to distance between households or cluster size.
- EPI designs with cluster sizes of 7 or 10 reduced travel distance within wards; smaller cluster sizes (≤15) with nearest neighbor selection minimized travel and mean square error.
Conclusions:
- The EPI sampling design, particularly with cluster sizes of 15 or fewer and optimized house selection, offers an efficient method for estimating xerophthalmia prevalence.
- The choice of sampling strategy should balance survey accuracy (bias, design effect) with logistical considerations like travel time and cost.
- EPI sampling with 15 children per cluster is favored over complete ward surveys unless inter-ward travel exceeds two days.