Evaluating the Efficacy of Stratified Sampling with Proportional Allocation for Estimation of Trachoma Prevalence in Sudan Open Access

Charania, Sana (Spring 2018)

Permanent URL: https://etd.library.emory.edu/concern/etds/k0698755v?locale=en



Trachoma is the leading infectious cause of blindness in the world. The Carter Center implemented the SAFE strategy to reduce the prevalence of trachoma in Sudan, which incorporated surveillance activities. Sampling methods used in this effort - Probability Proportional to Estimated Size (PPES) - led to large variability in the sampling weights resulting in imprecise estimates. The objective of this study is to obtain a valid and precise estimate of the prevalence of Trachomatous Trichiasis (TT) among individuals ≥15 in Sudan and inform programming efforts for The Carter Center’s trachoma surveillance system. We drew a sample of four districts in Sudan using stratified sampling with proportional allocation. We calculated prevalence estimates and accounted for stratification and clustering when evaluating parameter estimates and their standard errors for each locality. Then, we compared the estimator’s qualities of the new design to those of the PPES study to assess the efficacy of the two sampling protocols and determine whether the stratified random sampling approach added precision to the estimator. The prevalence of TT exceeded the WHO threshold of 0.2% in all four localities: Al Kamleen 1.73% (0.95, 3.14), Al Managle 0.81 (0.45, 1.46), Rife Kassala 0.50 (0.20, 1.29) and Sinnar 1.29% (0.79, 2.11). The average overall Design effect (Deff ) was lower in PPES sampling compared to stratified sampling with Proportional Allocation (1.26). Due to the little variability of Trachoma between clusters, the Deff yields a number that indicates the sample was more efficient than a simple random sample (Deff<1). This is likely due to small prevalence of the outcome rather than efficacy of the sample design. The Deff from the weights were comparable between the two sampling protocols. Evaluating the efficacy of the protocol with a more variable outcome, we calculated the Deff for the prevalence of females. Where there is more variability between clusters, the PPES study yields a larger Deff (2.56) than stratified sampling (1.90). Extremely low prevalence can make assessing sampling designs difficult, but there is evidence that stratified sampling provides benefits where prevalence of the condition is small and strong evidence of improvement where the prevalence is large. 

Table of Contents


Literature Review                                                                                                  1

Trachoma Surveillance                                                                              1

Equal Probability of Selection for Each Member                                       5

Stratification                                                                                              6

Clustering                                                                                                  7

Design Effect                                                                                            7

Probability Proportional to Estimated Size                                                8

Proportional Allocation                                                                             8

Manuscript                                                                                                        10

Introduction                                                                                           10

Methods                                                                                                 11

Results                                                                                                   14

Discussion                                                                                             15

Tables                                                                                                    19

Figures                                                                                                   21

References                                                                                             23

Summary, Public Health Impact, Future Directions                                         25 


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
Partnering Agencies
Last modified

Primary PDF

Supplemental Files