Skip to main content
. Author manuscript; available in PMC: 2023 Feb 2.
Published in final edited form as: Annu Rev Med. 2022 Jul 28;74:401–412. doi: 10.1146/annurev-med-043021-024004

Table 1.

Mechanisms by which bias in medical technologies is introduced and perpetuated

Mechanism Pulse oximetry Algorithms to estimate healthcare needs Pulmonary function testing
Physical aspects of technology design Melanin may cause light scatter leading to performance differences based on skin pigment Not applicable Not applicable
Use of data that do not reflect important patient subgroups Calibration of modulation ratio with saturation may be biased if data lack sufficient diversity Not applicable Not applicable
Conflation of biological and social determinants of health Not applicable Use of a biased outcome: healthcare costs as proxy for healthcare needs Assumption that population-based differences in lung function represent biological differences
Evaluation and regulatory standards Lack of robust standards to ensure equal performance in relevant subgroups Varying standards for regulating software algorithms Delayed appreciation of findings described in peer-reviewed medical literature