Relationship Between Levels of Digital Health Literacy Based on the Taiwan Digital Health Literacy Assessment and Accurate Assessment of Online Health Information: Cross-Sectional Questionnaire Study

Abstract

Background: The increasing amount of health information available on the internet makes it more important than ever to ensure that people can judge the accuracy of this information to prevent them from harm. It may be possible for platforms to set up protective mechanisms depending on the level of digital health literacy and thereby to decrease the possibility of harm by the misuse of health information.

Objective: This study aimed to create an instrument for digital health literacy assessment (DHLA) based on the eHealth Literacy Scale (eHEALS) to categorize participants by level of risk of misinterpreting health information into high-, medium-, and low-risk groups.

Methods: This study developed a DHLA and constructed an online health information bank with correct and incorrect answers. Receiver operating characteristic curve analysis was used to detect the cutoff value of DHLA, using 5 items randomly selected from the online health information bank, to classify users as being at low, medium, or high risk of misjudging health information. This provided information about the relationship between risk group for digital health literacy and accurate judgement of online health information. The study participants were Taiwanese residents aged 20 years and older. Snowball sampling was used, and internet questionnaires were anonymously completed by the participants. The reliability and validity of DHLA were examined. Logistic regression was used to analyze factors associated with risk groups from the DHLA.

Results: This study collected 1588 valid questionnaires. The online health information bank included 310 items of health information, which were classified as easy (147 items), moderate (122 items), or difficult (41 items) based on the difficulty of judging their accuracy. The internal consistency of DHLA was satisfactory (α=.87), and factor analysis of construct validity found three factors, accounting for 76.6% of the variance. The receiver operating characteristic curve analysis found 106 people at high risk, 1368 at medium risk, and 114 at low risk of misinterpreting health information. Of the original grouped cases, 89.6% were correctly classified after discriminate analysis. Logistic regression analysis showed that participants with a high risk of misjudging health information had a lower education level, lower income, and poorer health. They also rarely or never browsed the internet. These differences were statistically significant.

Conclusions: The DHLA score could distinguish those at low, medium, and high risk of misjudging health information on the internet. Health information platforms on the internet could consider incorporating DHLA to set up a mechanism to protect users from misusing health information and avoid harming their health.