Optimizing Cluster Survey Designs for Trachomatous Inflammation-Follicular in Amhara Region, Ethiopia Open Access

Gallini, Julia (Spring 2019)

Permanent URL: https://etd.library.emory.edu/concern/etds/1g05fc70w?locale=en


Background: Trachoma is the leading infectious cause of blindness globally, with some of the highest rates observed in Amhara National Regional State, Ethiopia. An international effort led by the World Health Organization (WHO) aims to reduce the prevalence of the trachomatous inflammation-follicular (TF) among children ages 1 to 9 years to below 5% globally by 2020. A key component of the strategy to eliminate trachoma as a health problem is the mass drug administration (MDA) of the antibiotic azithromycin. MDA decisions are made based on prevalence estimates from two stage cluster surveys. Work remains to formally mathematically evaluate the WHO recommended trachoma survey design.

Objective: Characterize the effects of the number of clusters and the number of households sampled on the precision and accuracy of TF estimates as well as costs of surveying in Amhara.


Methods: We simulated a population from which samples of varying numbers of clusters and households were selected. Bootstrapping techniques were used for variance estimation. Sampling schemes were evaluated on the following metrics: precision (through 95% uncertainty intervals), proportion of incorrect and low MDA decisions made (less MDA prescribed than warranted), design effect, effective sample size, and percent of sample efficiently used. Costs for each design were estimated based on previous work.


Results: The number of clusters sampled has a greater impact on the precision of the estimate than the number of households. Increasing households sampled yields a diminishing return in precision beyond 30 households. In low prevalence regions, cluster number has a lesser impact on precision than in higher prevalence regions. Samples are most mathematically and cost efficient used when sampling less than 30 clusters and households. The number of clusters drives survey cost more than the number of households sampled.


Conclusions: We recommend that past data on a region inform the survey design decision. For lower prevalence areas (less than 10%) we recommend 20 clusters of 20-30 households; for moderate to high prevalence regions (greater than 10%) we recommend 15 clusters of 20-30 households. Efficient use of surveying funds now will allow sustained surveying as Amhara approaches TF elimination as a public health problem.

Table of Contents

I.     Introduction. 1

II.   Background. 2

III.  Methods. 6

Generation of Population Dataset 6

Methods for Cluster Level Analysis. 13

Methods for Household Level Analysis. 14

Statistical Methods for Cost Analysis. 15

Bootstrapping Methodology. 17

Proportion Incorrect and Low Methodology. 17

Design Effect and Effective Sample Size Methodology. 18

IV.  Results: Gott Level 20

Proportion of Incorrect and Low Decisions Results. 25

Design Effect Results. 26

V.   Results: Household Level 28

Design Effect and Effective Sample Size Results. 32

VI.  Cost Analysis Results. 35

VII.   Future Directions. 38

VIII.  Discussion. 39

Cluster Level Analysis. 39

Household Level Analysis. 41

Reappearance of Trachoma. 43

Cost Analysis. 44

Recommendations. 45

Limitations. 45

Conclusion. 47

IX.  Appendix. 52

Supplemental Figures to Gott Level Results. 52

Supplemental Figures for Household Level Results. 57

SAS Macros. 82

Population Simulation Macro. 82

Drawing Samples of 30 Households (Using Segment Structure) Macro. 84

Drawing Samples of Varying Numbers of Households (Ignoring Segment Structure) Macro 88

About this Master's Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
  • English
Research Field
Committee Chair / Thesis Advisor
Committee Members
Last modified

Primary PDF

Supplemental Files