Health care stands on the brink of a revolutionary transformation as an influx of new technologies, prominently Artificial Intelligence (AI), promise enormous improvements in diagnosis, treatment, and patient care. However, the world is witnessing the unsettling trend of these computational technologies magnifying and perpetuating health disparities, embedded in the societal fabric. The root cause of this issue tracks back to the heavy reliance on ‘big data’. It is thus imperative that the health care stakeholders recognize the critical role that small data plays in balancing the scales and ensuring that the AI in health care contributes to health equity.
Big data, globally, has become synonymous with success in the AI landscape. In health care, monumental volumes of data from electronic health records, medical imaging, genetic sequencing, and wearable technologies, among others, are being leveraged to build AI models. These models are then used for various purposes such as predicting disease onset, diagnosing conditions, personalizing treatments, and improving patient care. However, there lies an inherent risk in this reliance on big data.
A major limitation that big data introduces is that it skews heavily towards the representation of the majority or the most privileged in society. Put simply, if the data predominantly consists of a certain population group, the AI models will inevitably have a bias towards that population group. This results in marginalized or minority groups getting underrepresented in big data and consequently being disadvantaged by AI models in health care. In fact, studies have shown that AI systems built on such health data can lead to racially biased outcomes when it comes to diagnosis and treatment suggestions.
These disparities, embedded deeply in datasets, are only destined to widen if the data ingested by AI continues to remain unrepresentative. This is where small data can assume a prominent role; by focusing on individual databases and confidentially collecting personal health information from a wider breadth of the population, including those in underrepresented groups. Small data involves collecting less information, but from a broader and more representative sample.
Leveraging small data allows health care providers to grasp the specific details of a patient’s condition, their history, lifestyle, and environmental variables – factors that often escape big data. In addition, most real-world data is small data, and research has substantiated that AI trained on real-world small data performs better and is more reliable in clinical prediction than its big data counterparts.
Moreover, small datasets also make it possible to solicit insights that are directly applicable to individuals, supporting the drive towards personalized medicine. This would ensure a more equitable distribution of the benefits offered by AI in health care.
Therefore, it is pivotal to incorporate small data into the AI paradigm. Bridging the disparity in AI with robust small data solutions also requires stringent checks and regulatory control over data collection processes, ensuring that the data is truly representative of the societal spectrum. Additionally, integrating lessons from fields such as humanities and social sciences, which have historically dealt with smaller sample sizes and qualitative data, could assist in shaping the AI technology that serves every section of the society equally.
Summing up, although big data has shown promising benefits, an over-reliance on this can unintentionally create a bias in health care AI that perpetuates disparities. To truly harness the benefits of AI and to curtail these discrepancies, emphasis should be placed on using small, but representative data. This would ensure that AI in health care is a boon for all, contributing to a more equitable and inclusive future in health care