Improving the Estimation of Population Iodine Status: A New Method for Calculating the Precision and Subgroup Differences of Median Urinary Iodine Concentration from Spot Urine Specimens

BACKGROUND: Urinary iodine concentrations (UIC) from spot urine specimens are typically not normally distributed. As such, survey managers use the median UIC, rather than the mean, as the measure of central tendency. Non-normal distributions prevent the comparison of sample means to a standard or subgroup-specific means using parametric tests. OBJECTIVES: We used Box-Cox transformations of UIC data to obtain normal distributions to accurately calculate measures of precision and statistical significance. We compared this with the conventional method using bootstrapping to calculate 95% confidence intervals (CIs) around the median. METHODS: We applied the Box-Cox transformation to UIC data from national surveys containing nonpregnant women in The Gambia, India, Kyrgyzstan, Lebanon, Senegal, Sierra Leone, and Uzbekistan. We calculated subgroup-specific means and CIs and used parametric statistics (i.e., t-tests and analysis of variance) to assess the statistical significance of subgroup differences while accounting for complex sampling. Means and CIs were then back-transformed into the original units (μg/L). The back-transformed mean of the transformed UIC values can be interpreted as an estimate of the population median, given the near symmetry of the transformed values. RESULTS: The crude UIC data from all countries were not normally distributed. Box-Cox λ values ranged from 0.128 to 0.454, and all transformed datasets had skewness values between -0.1 and +0.1, confirming approximate normality. The Box-Cox method allowed detection of statistically significant differences in the back-transformed mean UICs by region and household-salt iodization status in nearly all countries. CONCLUSIONS: The Box-Cox transformation enables the calculation of results that account for complex sampling, enables the estimation of the statistical significance of apparent subgroup differences, and offers a more accurate approach to analyzing skewed UIC data. Program planners can use this approach to prioritize population groups with insufficient iodine status and tailor nutrition programs to address inequities in iodine nutrition.

Перевод пока недоступен