This blog post is about the difference between predictive vs. prescriptive analytics in healthcare. What are these tools? Does one have a better use case than the other in healthcare?
What is predictive analytics?
Predictive analytics can be described as using statistical modeling to predict an outcome. It can fall under using historical data and machine learning techniques (i.e. decision trees classifiers, neural networks) to train a model and then using this model to predict an outcome.
Predictive analytics can be used in several industries such as the financial industry to predict cases of fraud detection and can also be used in the insurance industry to predict payout rates.
Although predictive analytics is quite new in the field of healthcare, there are potential benefits such as predicting readmission rates for patients. In Canada, In a study in 2018 done by the Canadian Institute for Health Information, 1 in 11 patients were readmitted within a month of leaving the hospital and readmissions cost more than $2.1 billion CAD dollars a year. Being able to flag patients who might be at risk for readmission and improve patient monitoring before they leave and after, when it comes to follow-up care, might result in significant cost savings in the future .
What is prescriptive analytics?
Prescriptive analytics can be described as using data driven decisions to output several outcomes and determine what should be done.
One of the differences between these types of analytics is that predictive analytics is usually focused on predicting mainly one outcome. For instance, a breast cancer risk scoring tool can take into account the historical clinical data of patients and using advanced machine learning tools, can estimate your risk of developing breast cancer, and this would fall under predictive analytics.
Whereas with prescriptive analytics the output is a range of options and the best option is chosen from a multitude of results. Google’s self-driving car would be an example of prescriptive analytics because the car is being operated based on taking in new information as it travels, such as: crowds of people, buildings, traffic signals and so on and then it makes the best decision based on these inputs.
I visualize prescriptive analytics as a feedback loop constantly taking in relevant information to influence you final decision where I see predictive analytics as more of a straight line, you have your historical data you train your model using this historical data and predict a final outcome from this data
Although applying prescriptive analytics to the field is newer than even predictive analytics, there are far reaching benefits, if we go back to the example of hospital readmission rates, instead of just flagging patients who are at risk. Wider scale intervention programs could be developed with several options such as linking up high risk patients with physicians, providing patients with a hotline to contact the hospital and advanced care for these patients and analyzing out of these outcomes, which is the most valuable.
In my own opinion, I think that prescriptive analytics will be of more value in the future over predictive analytics in healthcare organizations because I think one of the drawbacks with predictive models in healthcare is the lack of explainability and potential black box models that are the foundation of some predictive models, specifically neural networks.
I think that if physicians and patients want to have predictive models integrated into their diagnosis, physicians will have to explain these complex mathematical models to patients and this might be a barrier to usage. Whereas, I think with prescriptive analytics because it is a more integrated model where the clinician is providing their own input when it comes to building the model and is presented with a range of options where they use their own clinical knowledge and background to present the best option to the patient, this model might be more widely accepted. As always, it will be interesting to see how analytics will drive clinical metrics and improve patient outcomes.
What do you think? Will prescriptive analytics overtake predictive analytics in the health informatics space? Is there room for both? If you are interested in this content, please consider subscribing to the Health Analytic Insights Podcast, wherever you listen to podcasts! 
0 Comments