Background/Purpose: Gout is considered the most common inflammatory arthritis in the US with an estimated prevalence of 3.9% based on the National Health and Nutrition Examination Survey (NHANES). However, misclassification hampers understanding of the true population burden of gout. For example, the range of global prevalence varied from < 1% to 6.8% based on one recent published Nature Reviews Rheumatology. We aimed to examine the impact of misclassification on the prevalence of gout in the US.

Methods: We conducted Bayesian misclassification correction analysis using the data from the NHANES 2015-2016. This survey asked the question “Has a doctor or other health professional ever told you that you had gout?” To inform the sensitivity and specificity of this questionnaire item, we identified two most relevant studies (J Rheumatol 2011;38:135 for sensitivity; Arthritis Care Res 2016;68:1894 for specificity) by systematically searching the PubMed (– March 1, 2022). The specificity study was from a secondary care setting. Thus, we further ranged the specificity information to the higher values, considering the spectrum effect (i.e., specificity of an instrument can be higher in the general population than in secondary care patients who likely included more “mimics”). A flat prior was used for the true prevalence parameter. Bayesian hierarchical regression and post-stratification models were used to account for the misclassification and the known differences (age, sex, race, and income) between the NHANES sample and the US population (data from United States Census Bureau, https://www.census.gov/). We obtained the posterior estimates of the US population prevalence of gout as well as the sensitivity and specificity of the questionnaire item, combining the prior studies and the NHANES.

Results: We found 5,076 participants with 237 gout cases in the unweighted NHANES 2015-2016 sample. The external sensitivity study provided a sensitivity of 84.2% (165 positives / 196 cases). The external secondary-care specificity study provided a specificity of 72.0% (236 negatives / 328 non-cases). The model did not converge with the prior specificity of 72%. Our four modified specificities were the following: 80%, 85%, 90%, and 95%. The posterior means of the misclassification-corrected US population prevalence of gout ranged from 2.7% to 4.1%, increasing with a higher prior specificity (Table 1). The posterior mean of the sensitivity was approximately 83%. The posterior means of the specificity ranged from 96.7% to 98.8%, likely representing the better specificity in the general population.

Conclusion: Even with a pessimistic prior assumption on the questionnaire item specificity (80%), the posterior estimate of the misclassification-corrected US prevalence of gout remained high. A validation study of the gout questionnaire item in the general population can better inform the true prevalence in the US.

« Back to ACR Convergence 2022

ACR Meeting Abstracts - https://acrabstracts.org/abstract/impact-of-misclassification-on-the-us-prevalence-of-gout-bayesian-sensitivity-analysis-of-the-national-health-nutrition-examination-survey-nhanes/