Author: Hussein Ibrahim, Xiaoxuan Liu, Nevine Zariffa, Andrew D. Morris, Alastair K. Denniston
Publisher: The Lancet Digital Health
Publication Year: 2021
Summary: The following study identifies how machine learning and advanced technologies can now be used to transform health data, but such innovations could widen the gap of healthcare inequalities for those who may suffer from health data poverty. Health data, or health-related information about a person, can be used to make diagnoses or improve healthcare services at a population level. With healthcare and health data, however, there exist disparities, particularly due to underrepresentation in data of certain demographics and populations. The tools developed are not useful and can cause harm to underrepresented groups, also considered health data poor, because the datasets are derived from and tested on a different subset of the population that may have completely different health issues and diagnostics. Some ways to reduce health data poverty are raising awareness of the equity gap and holding those creating the models accountable for ethical practice; advocating for and requiring that developers create technologies that can perform adequately across diverse populations; investing in representative datasets; encouraging transparency about how the data is collected and any privacy concerns as well as how the data will be used to contribute to safer digital health solutions; and understand the inequity of digital access which is often used for data collection.
