Predictive models that use data from individuals are an important source of information in medical settings. Predictive modelling, or the use of electronic algorithms to forecast future events, makes it possible to harness the power of big data to improve people’s health and reduce the cost of healthcare. However, this opportunity can raise policy, ethical, and legal questions. An important area for the Prediction Modelling Group to explore is the involvement of service users, carers and other stakeholder in discussions about ethical issues and to develop guidelines for the development and use of prediction models in mental health.
Case study example: predicting suicide
There is huge potential value in developing new approaches to modelling suicide risk, given the current relative difficulty of predicting suicide. Early work with electronic health records is promising, but presents important ethical questions around triggering suicide prevention interventions. This paper describes the utilitarian, Kantian, and personal rights ethical models that can be used to evaluate the potential impacts of using a predictive model in this context, and offers recommendations for navigating these.
Some of the major legal, policy, and ethical issues raised by predictive analytics are as follows:
1) Consent and privacy: Model developers should be allowed to use patient data that have already been collected without explicit consent, provided the developers comply with federal regulations regarding research on human subjects and privacy of health information.
2) Equitable Representation: Predictive analytics models require vast amounts of data that should be representative of the whole population.
3) The importance of patient-centred perspectives: The development of a predictive analytics model inevitably involves choices about which problems to make high priorities, how the model will be used, what clinical interventions will be provided to patients at risk for adverse outcomes, and what outcomes will be measured in the model’s evaluation. Patient representation should be included in the governance of organizations that develop and implement predictive analytics models in medicine.
4) Standards for validation and transparency before a predictive analytics model is used in clinical care: Model should be carefully evaluated for effectiveness and any adverse consequences.
5) Outcomes assessed in model validation: The primary outcomes for any predictive analytics model should be “hard” clinical end points.
6) Testing the model in real world settings: Once a predictive analytics model has been developed and validated internally, the model must be tested in the field. This raises significant questions regarding consent, liability, and choice architecture
Source: The Legal And Ethical Concerns That Arise From Using Complex Predictive Analytics In Health Care (Glenn Cohen et al, Health Affairs, 2014)