Deep learning technology has numerous innovative medical applications, one that stands out in the field of precision medicine is the ability to utilise electronic health records to create better models for patient representation. The potential for this technology is enormous.

At present, clinicians are already capturing a staggering amount of patient data. The traditional function of this data collection has always been for empowering doctors and clinical workers with on-hand knowledge, in order to provide informed care for individual patients. This will always be the primary function of health records.

The new opportunity presented here however is in the aggregation of patient records to create broadly applicable overviews of public health trends and statistics. The application of these overviews is two-fold. Firstly, granting unparalleled insights into public health trends that can inform and advise public health policy on a government level. Secondly providing invaluable scientific data to the medical community to assist in research and development for both community and individual care planning.

The real power of machine learning however comes from the snowball-like process of increasing data availability. An algorithmic approach is vastly improved by an increase in data input, but the data output of a system like this feed back into it as well. The output accelerates the rate at which new health data is made available which in turn feeds research applications which can inform machine learning development. The technology is iterating upon itself by working in tandem with human researchers.

There are challenges in addressing potential systemic biases and potential for data outliers to influence outcomes, but over time these issues are increasingly ironed out through repetition and rigorous examination of outcomes.