The quality of sample surveys in a developing nation

Paul A Bourne¹, Christopher AD Charles^2,3, Neva South-Bourne⁴, Chloe Morris¹, Denise Eldemire-Shearer¹, Maureen D Kerr-Campbell<sup>5

1</sup>Department of Community Health and Psychiatry, Faculty of Medical Sciences, University of the West Indies, Mona, Kingston, Jamaica; ²King Graduate School, Monroe College, Bronx, New York, USA; ³Center for Victim Support, Harlem Hospital Center, New York, USA; ⁴Research assistant for Paul A Bourne; ⁵Systems Development Unit, Main Library, Faculty of Humanities and Education, University of the West Indies, Mona, Kingston, Jamaica

Background: In Jamaica, population census began in 1844, and many intercensal ratios obtained from the census data showed that there is a general high degree of accuracy of the data. However, statistics from the Jamaican Ministry of Health showed that there are inaccuracies in health data collected from males using sample surveys.

Objectives: The objectives of the present research are to 1) investigate the accuracy of a national sample survey, 2) explore the feasibility and quality of using a subnational sample survey to represent a national survey, 3) aid other scholars in understanding the probability of using national sample surveys and subnational sample surveys, 4) assess older men’s ­evaluation of their health status, and 5) determine whether dichotomization changes self-evaluated health status.

Methods: For the current study, the data used in the analysis were originally collected from 2 different sources: 1) the Jamaica Survey of Living Conditions (JSLC) and 2) Survey of Older Men (SOM). Cross validation of self-evaluated data of men in Jamaica was done with comparable samples of the complete JSLC data and the SOM data, where men older than 55 years were selected from each sample.

Results: In study 1, 50.2% of respondents indicated at least good self-evaluated health status compared with 74.0% in study 2. Statistical associations were found between health status and survey sample (Χ² [df = 5] = 380.34, P < 0.001), self-reported illness and study sample (Χ² [df = 1] = 65.84, P < 0.01, Φ = 0.16), and health care – seeking behavior and study samples (Χ² [df = 1] = 21.83, P < 0.05, Φ = 0.10). Substantially more respondents reported an illness in study 1 (34.3%) than in study 2 (ie, 17.5%).

Conclusion: Clearly, inconsistencies exist in the health data, which indicates that care should be taken in using sample surveys.

Keywords: validation, health data, males, public health, national surveys, subnational surveys