From clean water supplies to the polio vaccine, the most effective public health interventions are typically preventative policies that help stop a crisis before it starts. But predicting the next public health crisis has historically been a challenge, and even interventions like chlorinating water or distributing a vaccine are in many ways reactive. Thanks to predictive analytics, we are piloting new ways to predict public health challenges, so we can intervene and stop them before they ever begin.
We can use predictive analytics to leverage seemingly unrelated data to predict who is most susceptible to birth complications or chronic diseases or where and when a virulent outbreak is most likely to occur. With this information, public health officials should be able to respond before the issue manifests itself – providing the right prenatal treatments to mitigate birth complications, identifying those most likely to be exposed to lead or finding food establishments most at risk for violations. With this information, data becomes actionable. Predictive analytics has the potential to transform both how government operates and how resources are allocated, thereby improving the public’s health.
While the greatest benefits have yet to be realized, at the Chicago Department of Public Health (CDPH), we are already leveraging data and history to make smarter, more targeted decisions. Today, we are piloting predictive analytic models within our food protection, tobacco control policy, and lead inspection programs.
Recently, CDPH and the Department of Innovation and Technology engaged with local partners to identify various data related to food establishments and their locations – building code violations, sourcing of food, registered complaints, lighting in the alley behind the food establishment, near-by construction, social media reports, sanitation code violations, neighborhood population density, complaint histories of other establishments with the same owner and more.
The model produced a risk score for every food establishment, with higher risk scores associated with a greater likelihood of identifying critical violations. Based on the results of our pilot and additional stakeholder input, we are evaluating the model and continue to make adjustments as needed. Once it is proven successful, we plan to utilize the model to help prioritize our inspections, and by doing so, help improve food safety.
To be clear, this new system is not replacing our current program. We continue to inspect every food establishment following our current schedule, ensuring the entire food supply remains safe and healthy for our residents and tourists. But predictive analytics is allowing us to better concentrate our efforts on those establishments more likely to have challenges. In time, this system will help us work more closely with restaurateurs so they can improve their business and decrease complaints. In short, businesses and their customers will both be happier and healthier.
Building on the work of the food protection predictive model, we developed another key partnership with the Eric & Wendy Schmidt Data Science for Social Good Fellowship at University of Chicago (DSSG) to develop a model to improve our lead inspection program.
Exposure to lead can seriously affect a child’s health, causing brain and neurological injury, slowed growth and development, and hearing and speech difficulties. The consequence of these health effects can be seen in educational attainment where learning and behavior problems are often the cause of lower IQ, attention deficit and school underperformance. Furthermore, we’ve seen a decrease in federal funding over the past several years for our inspectors to go out and identify homes with lead based paint and clearing them. But thanks to data science, we are now engaging on a project where we can apply predictive analytics to identify which homes are most likely to have the greatest risk of causing lead poisoning in children – based on home inspection records, assessor value, past history of blood lead level testing, census data and more.
Predictive models may help determine the allocation of resources and prioritize home inspections in high lead poisoning risk areas (an active approach), instead of waiting for reports of children’s elevated blood lead levels to trigger an inspection (the current passive approach). An active predictive approach shortens the amount of time and money spent in mitigation by concentrating efforts on those homes that have the greatest risk of causing lead poisoning in children.
Incorporating predictive models into the electronic medical record interface will serve to alert health care providers of lead poisoning risk levels to their pediatric and pregnant patient populations so that preventive approaches and reminders for ordering blood lead level lab tests or contacting patients lost to follow-up visits can be done.
There is a great opportunity in public health to use analytics to promote data-driven policies. We need to use our data better, share it with the public and our partners, and then leverage that data to create better policies, systems and environmental changes.
Public institutions should increasingly employ predictive analytics to help advance their efforts to protect the health of their residents. Furthermore, large, complex data sets should be analyzed using predictive analysis for improved pattern recognition, especially from diverse data sources and types, ultimately leading to significant public health action. For the Chicago Department of Public Health, predictive analytics is not the future, it is already here.
By: Bechara Choucair, Jay Bhatt and Raed Mansour
Originally published at blogs.hbr.org
 |
 |
 |
