At present, there is no reliable way to consistently identify people who are on track to develop the neurological disorder. This can make it difficult to get them the treatment they need early on, while subjecting patients to multiple tests as doctors search for the root of the problems they report. Researchers at the Washington University School of Medicine in St. Louis think they’ve come up with a solution, however — by developing algorithm that’s able to accurately predict whether a patient will be diagnosed with Parkinson’s. To do this, it looks for signs in a patient’s medical history, which may otherwise be overlooked.
“We started [developing our algorithm by using] demographic factors already known to be associated with the development of Parkinson’s disease — age, sex, race, and tobacco smoking history — and then added to it a simplified medical history,” Brad Racette, professor of neurology at Washington University School of Medicine, told Digital Trends. “Specifically, we started with 26,468 codes for diagnoses or medical procedures they had received in the past five years. Then we identified which of those codes also were predictive of PD. The final stage of the model development allowed us to compress this list down to the 536 codes that were both the most informative in our dataset, and also the most likely to be useful in future datasets.”
The algorithm proved 73 percent accurate when it came to identifying individuals who would be diagnosed with the disease, based on a test data set from 2009. It was also 83 percent accurate at predicting who would not be diagnosed with the disease. Factors in a person’s medical history which can indicate Parkinson’s include gastrointestinal problems, sleep disturbances, fatigue, weight loss, and more — many of which may not immediately make physicians think of Parkinson’s as a potential cause.
Washington University’s study isn’t the only high tech approach to diagnosing and treating Parkinson’s disease that we’ve covered. With 1 million Americans living with Parkinson’s disease, and an estimated 10 million worldwide, this is a disease that has attracted plenty of research — from attempts to treat it by reprogramming cells in the brain to the use of sensor-packed belts to help patients overcome the symptoms. Hopefully, an early diagnosis tool such as this one can play a valuable role alongside these other treatments.
A paper describing the work was recently published in the journal Neurology.
