Parkinson’s Disease shows a predictable degenerative process of tremors, falls and trouble sleeping. But that’s all in hindsight. By the time it is actually diagnosed, it’s already progressed to a point where it can’t be slowed or mitigated.

Big data algorithms scanning population-level health data may hold a clue in diagnosing the disease early, predicting what profile of patient is likely to develop the disease, according to a new study in the journal Neurology.

“Using only demographic data and selected diagnosis and procedure codes readily available in administrative claims data, it is possible to identify individuals with a high probability of eventually being diagnosed with (Parkinson’s),” write the scientists from Washington University of St. Louis, University of Pennsylvania and University of the Witwatersand in South Africa.

About 200,000 patients who were on Medicare in 2009 were incorporated in the data set. The 89,790 who were diagnosed with Parkinson’s were mixed in with more than 118,000 control patients who did not have the disease.

By further incorporating the health records of the patients from 2004 through 2009, the scientists developed an “elastic net algorithm” that scored key hallmarks in the medical histories.

The Parkinson’s factors included the trademark tremors, but also weight loss, various chronic kidney diseases, posture abnormalities, psychiatric and cognitive problems, gastrointestinal aberrations, sleep disturbances, fatigue, and injuries (especially from falls).

The factors that appeared to weigh against a Parkinson’s diagnosis included obesity conditions like gout, smoking, cancer, cardiovascular disease, allergies, and injuries that were related to wear-and-tear of activity (like hip replacements).

Overall, the algorithm successfully predicted a majority of Parkinson’s cases.

About 73 percent of the diagnosed in 2009 were picked out by the computer model. At the same time, the machine selected 83 percent of the people who would not be diagnosed.

Brad Racette, the senior author from Washington University, said the algorithm could be used to identify Parkinson’s patients and their unique medical records trail long before the disease has gotten too far.

“Using this algorithm, electronic medical records could be scanned and physicians could be alerted to the potential that they patients may need to be evaluated for Parkinson’s disease,” he said, in a school statement.