Reflection: How do I think health analytics can contribute to improving patient care and healthcare outcomes?

Health analytics, particularly through the use of advanced techniques like machine learning and prescriptive analytics, has the potential to revolutionize patient care and healthcare outcomes. Risk segmentation or scoring through predictive modelling can support health systems distribute appropriate resources to case management, social services and government programmes (Tan et al., 2020). While predictive analytics helps anticipate potential health risks and patient outcomes, prescriptive analytics takes it a step further by suggesting actionable steps to optimize these outcomes. This proactive approach is especially valuable in the context of precision medicine, where tailored interventions are needed to provide the right treatment for the right patient at the right time (Mosavi & Santos, 2020).

One example of this in practice is C-Path, a machine learning tool that goes beyond traditional methods by identifying new features in pathology samples to improve the detection of high-risk breast cancer cases. C-Path demonstrates how machine learning can enhance the feature selection process, discovering novel patterns that even experienced pathologists might overlook (Mosavi & Santos, 2020). These new features are then used to build models that predict patient outcomes more accurately than standard clinical methods. This approach highlights the power of machine learning in feature selection, where the identification of previously unknown predictive variables can lead to significant improvements in patient care.

However, successful application of health analytics requires more than just advanced algorithms and computational power. It also demands a deep understanding of the context in which these models operate. Machine learning models are often built on static datasets that may not capture the dynamic nature of health and its determinants. Social and behavioral determinants of health (SBDH), for example, are factors that influence patient health outcomes but can vary significantly over time. Machine learning methods typically struggle to account for these changes as they address stationary distributions of SBDH factors (Kino et al. 2021). When the data distribution changes—such as the determinants of antiretroviral therapy (ART) discontinuation—additional modelling techniques like time-series analysis or event-driven simulations are necessary to ensure accurate predictions (Tan et al., 2020).

Moreover, the inclusion of SBDH in predictive models has shown mixed results. While research supports the positive association between SBDH and health outcomes, studies have found that these factors do not consistently add predictive value when integrated into traditional forecasting models. For instance, Bhavsar et al. (2020) used random survival forest methods with poverty status and electronic health records (EHR) data but did not observe significant improvements in predicting health service utilization. This suggests that while SBDH is critical to understanding health outcomes, its integration into machine learning models needs further refinement and a more nuanced approach (as cited in Kino et al. 2021).

Ultimately, prescriptive analytics offers a way to bridge these gaps by providing recommendations that take into account the limitations and complexities of predictive models. By incorporating contextual knowledge and adjusting for changing data patterns, prescriptive analytics can help healthcare providers develop more effective interventions and treatment strategies. This approach aligns with the goal of precision medicine, which aims to tailor healthcare to the individual, ensuring that each patient receives care that is not only evidence-based but also personalized to their unique health profile and circumstances.

---

References

Nong, P., & Adler-Milstein, J. (2021). Socially situated risk: challenges and strategies for implementing algorithmic risk scoring for care management. JAMIA Open, 4(3). https://doi.org/10.1093/jamiaopen/ooab076

Sadat Mosavi, N., & Filipe Santos, M. (2020). How Prescriptive Analytics Influences Decision Making in Precision Medicine. Procedia Computer Science, 177, 528-533. https://doi.org/https://doi.org/10.1016/j.procs.2020.10.073

Tan, M., Hatef, E., Taghipour, D., Vyas, K., Kharrazi, H., Gottlieb, L., & Weiner, J. (2020). Including social and behavioral determinants in predictive models: trends, challenges, and opportunities. JMIR medical informatics, 8(9), e18084.