Variability Test of Estimates of Korea Community Health Survey Data in Gyeongsangnam-do (2008-2018)

Abstract

Objectives: In this study, we aimed to determine the regional and yearly variability of estimates from the Gyeongsangnam-do Korea Community Health Survey conducted from 2008 to 2018. Methods: Line graphs, box plots and dendrograms were analysed to determine the distribution of data by indicator in the Korea Community Health Survey data. The intraclass correlation coefficient (ICC) was used to check the consistency of the data by indicator, GLM to check for differences by time of measurement, and Run-test to check whether the variation was random. Results: The data distribution of the indicators varied within the 30% range suggested by the KDCA (line graphs). Outliers were found in certain regions, suggesting that the results are not consistent (box plots). Using dendrograms, we can see the similarities and differences in the results at once, and identify 2008 and 2009 as the first group, 2010-2014 as the second group, and 2015-2018 as the third group. The results of ICC and GLM for inter-annual variability and consistency show that most indicators were not consistent across years. Finally, the results of the Run-test to determine the randomness of the fluctuations also show that most of the indicators have random fluctuations. Conclusions: The Korea Community Health Survey has been conducted in the same way every year from the conception of the statistics to the time of the survey to ensure the comparability and consistency of the survey results from year to year, but there has been no quality assessment of the period when stable performance appears after policy and system changes. Therefore, this study suggests that followup research on the causality of the variability with the historical background, environment, events and accidents is needed.