OBJECTIVES: The promise of big data has raised great expectations in healthcare research for drug discovery, as well as for personalized patient treatment and monitoring. This study aims to examine digital health applications that collect patient data in order to identify potential sources of bias in the resulting databases.

METHODS: A literature search was conducted on Google Scholar using the keywords "digital health applications" and "demographics." Variables investigated included the surveyed population (country), sample size, and application type. Demographic variables analyzed within the surveyed population encompassed income level, education level, age, gender, disease status (ongoing diseases or reported healthy), and presence of ethnic disparities (including the number of analyzed ethnic groups, if applicable).

RESULTS: Out of 250 scanned articles, 34 met the inclusion criteria of at least 4 demographic variables present. The most prevalent applications were focused on healthy lifestyle (8/34) and diabetes care (5/34), with the United States being the primary country with more studies, accounting for 10 out of the 34 articles. The majority of users exhibited characteristics such as high purchasing power (15/34 vs. 2/34), high education level (27/34 vs. 1/34), age under 40 (16/34 vs. 10/34), female gender (19/34 vs. 9/34), and reported unhealthy status (14/34 vs. 8/34). Among articles including ethnicity data (12/34 vs. 23/34), all demonstrated imbalances in minority group representation.

CONCLUSIONS: This study highlights the presence of various biases in databases constructed from health applications, which may negatively impact the development of secure and inclusive algorithms. Further research is necessary to expand the range of reviewed applications and encompass a broader spectrum of potential sources of bias.