Case Study: Patterns and Predictions achieves scalable, real-time veteran suicide-risk prediction with Cloudera

Using Big Data to Predict Veteran Suicide Risk

Patterns and Predictions (P&P) is a predictive analytics firm whose Centiment® linguistics-driven technology powers the Durkheim Project, a research initiative—funded by DARPA and partnered with Dartmouth’s Geisel School of Medicine, Cloudera and others—aimed at identifying suicide risk among veterans. The project addresses a pressing public-health challenge: military and veteran suicide rates are significantly higher than the general population, and the VA has emphasized the need for better data on at-risk veterans to improve prevention programs.

P&P and partners built a Hadoop-based, real-time analytics platform that applies unstructured-text and linguistic machine learning (including lightweight Bayesian counters on Cloudera’s CDH) to clinical notes, social posts and mobile data, securely stored behind a medical firewall. Phase One validated the approach—predicting suicide risk in a veteran control group with about 65% accuracy—and Phase Two scaled ingestion and scoring to support up to 100,000 participants and 1+TB/day, delivering real-time risk scores, lower operational cost and improved computational throughput while the project pursues further validation for interventional studies.

Patterns and Predictions

Chris Poulin

Principal Partner, Patterns And Predictions