Associate Professor University of Kentucky Lexington, United States
Background: Health disparities resulting from digital factors can impact individuals of all backgrounds, but certain groups are at a higher risk of poor health outcomes. Digital inequalities only worsen these already existing disparities, yet are seldom analyzed using an outlier detection method.
Objectives: In an effort to identify high-risk populations, this study aimed to examine the prevalence and variations of digital disparities among adults in the United States of America.
Methods: The study used the Health Information National Trends Survey (HINTS) by the U.S. National Cancer Institute as its data source. The sample consisted of 12,227 respondents from the combined data of HINTS 5-Cycle I (2017), II (2018), and III (2019). Only adults aged 18 or older were included in the study. The one-class Support Vector Machine (SVM) classification model was utilized to identify high-risk populations. The parameters of the model, such as the kernel type, the cost parameter, and the type of normalization applied to the data, were carefully selected to ensure the best possible results. The data used in the study consisted of demographic, socioeconomic, health, and preventive behavior variables. The data was pre-processed to ensure its suitability for the one-class SVM analysis, including missing data imputation and feature scaling.
Results: The results of the one-class SVM analysis indicated that the most vulnerable digital disparity groups were young, female, African American, low-educated, and low-income respondents. The anomaly detection showed a racial disparity gap between White and African American within and outside normal proportions (White: +9.28%, African American: -6.6%). The dataset classified "less than 10 percent" of the high-risk groups as high-risk. The findings of this study provide important insights into the extent and diversity of digital disparities among adults in the USA. The results suggest that young, female, African American, low-educated, and low-income respondents are the most vulnerable to digital disparities and need special attention and support. The racial disparity gap between White and African American highlights the need for further research in this area and the importance of addressing digital disparities in a culturally sensitive manner.
Conclusions: In conclusion, this study highlights the importance of addressing digital disparities and their impact on health outcomes. The one-class SVM analysis provides valuable insights into the extent and diversity of digital disparities among adults in the USA and identifies the most vulnerable populations.
