Computers Can Identify Human Risks Before They Happen

In the world of heavy industry, preventative maintenance is a must. Ignore a milling machine or truck engine for too long and problems that could have been easily corrected turn into more expensive breakdowns.

Computers Can Identify Human Risks Before They Happen
Computers Can Identify Human Risks Before They Happen

The same is true with human health. The infection spreads, the disease progresses, the physical weakness turns someone into an invalid. The need for preventative treatment, and its ultimate cost effectiveness, is why the Affordable Care Act mandates such care. But there is an additional problem. Some people are statistically more likely to be subject to various conditions than others. Initial screening can make resource allocation and effective policy and operations more possible. Computer scientists can help.

An example is the development of a new mathematical model that can identify soldiers at high risk for suicide, as the site LiveScience reported. More than 40,000 Army soldiers who had faced psychiatric hospitalization between 2004 and 2009 provided the body of data for researchers.

At question was the risk of suicide among these people. People hospitalized with a psychiatric diagnosis are typically at higher risk than the general populace, and that was true for this Army group. Sixty-eight soldiers, 264 per hundred thousand, committed suicide within a year of being discharged from the hospital. That is far more than the 18.5 per hundred thousand rate of soldiers in general, which is slightly higher than adults between 25 years and 64 years of age, according to the CDC.

However, even at those rates, suicide among the formerly hospitalized soldiers was rare. Providing cost- and time-intensive suicide prevention programs for everyone in the group would be virtually impossible. So researchers used computers to analyze the data and find factors that could be predictive of potential suicide.

Eventually, the researchers narrowed the list to 131 factors that had some positive correlation with suicide attempts, including age, gender, access to a firearm, previous treatment for a psychiatric illness, current case of post-traumatic stress disorder, age of enlistment, and anti-depressant prescriptions filled within the previous 12 months. A computer allowed analysis of many more variables than a doctor or other healthcare professional could consider at the same time.

After building a model and running additional analysis, the researchers found that half the suicides in the study came from the 5 percent of Army patients identified as being most at risk. The same group was also at risk for accidental death, a suicide attempt, or readmission to a hospital.

The model isn’t perfect, but it could help target potential suicides. There is a world of other health-related conditions and problems that could also benefit from a rigorous analysis and application of computing.