About Us

Predictive Analytics System


The Hotspotting predictive analytics system centers around transitioning overdose prevention efforts from a reactive to a proactive approach. There are three pillars to this effort: a series of data sharing partnerships which enables the consolidation of health data in a data warehouse, the application of a predictive analytics software program to this data set, and then the assignment of at-risk individuals identified by this program to specific treatment partners aligned with the hotspotting program.

The Hotspotting Predictive
Analytics System

Data Consolidation

The Hotspotting program’s predictive analytics system relies on applying predictive software to a medical data warehouse. Existing data warehouse options include large medical providers, government entities with jurisdiction over public programs like Medicaid, and treatment provider coalitions with established information sharing arrangments.

Software Based Predictive Analytics Program Applied to Data Warehouse

The Staten Performing Provider System, via a series of official data sharing arrangements and trusted partnerships has integrated health data from multiple sources into a population health management (PHM) data platform and warehouse. Utilizing this breadth of medical data, the SI PPS and MIT Sloan School of Management’s Initiative for Health Systems Innovation initiated an effort to investigate how the utilization of predictive analytics, when combined with social and healthcare data sets, could be deployed in practice. Simply put, how can cutting edge predictive systems be combined with state of the art intervention models to dramatically reduce overdose related incidents? MIT identified SI- PPS as its principal partner in this effort because of the PPS’s unique system and collection of medical data, electronic health records (EHR) data and social data (emergency health, criminal justice and social agencies). Together the two organizations co-designed an opioid incident predictive
algorithm, which uses machine learning to predict, based on a series of “high risk factors”, who within a community is at the highest risk of overdosing on opioids. This algorithm has been proven to identify a small cohort of at-risk individuals which accounts for a highly significant share of overdose events. The findings from this algorithm development will be published in an article jointly written by MIT and the SI PPS entitled, “Preventing Opioid Overdose: From Prediction to Operationalization,” to be published in the Manufacturing and Service Operations Management Journal.



Lastly, Investments have also been made in a cloud-based software system so as to enable it to automate this algorithm and seamlessly integrate into the program’s machine learning elements. As such, this predictive analytics program processes the data warehouses medical data in an automated fashion, enabling the Hotspotting, program to more easily scale to other locations.

Identification and Assignment of At-Risk Individuals

Upon identification of high-risk individuals by the Hotspotting predictive analytics system, the ‘Hot-Spotting The Overdose Epidemic’ Program then coordinates with the leading treatment providers aligned with the initiative to identify their pre-existing patients/clients in that pool.


This relevant client data is then transferred from the cloud-based data warehouse to the Channels care management platform, which has a solidified client import component from the data warehouse. At this point, client engagement and care coordination is initiated with the principal goal being to reduce overdose rates by substantively treating those most at risk before they overdose, not after.