Researchers at the Washington University School of Medicine (WUSM) in St. Louis developed a new computational method that can help predict if a patient will be diagnosed with Parkinson’s disease based solely on medical records.
This new predictive algorithm was revealed in the report “A predictive model to identify Parkinson disease from administrative claims data,” which was published in the journal Neurology.
The lack of reliable disease biomarkers has made Parkinson’s disease difficult to diagnose. In the majority of cases it is detected only when symptoms are already settled and the disease is in advanced stages. Based on the patient’s clinical history, this method has the potential to help diagnose the disease before the onset of symptoms, promoting more suitable patient care.
“Using this algorithm, electronic medical records could be scanned and physicians could be alerted to the potential that their patients may need to be evaluated for Parkinson’s disease,” Brad A. Racette, MD, the Robert Allan Finke Professor of Neurology and senior author of the study, said in a WUSM press release, written by Tamara Bhandari.
“One of the most interesting findings is that people who are going to develop Parkinson’s have medical histories that are notably different from those who don’t develop the disease,” he added. “This suggests there are lifelong differences that may permit identification of those likely to develop the disease decades before onset.”
Led by Susan S. Nielsen, PhD, assistant professor in neurology at the WUSM in St. Louis, the research team analyzed Medicare claims data collected from 2004 to 2009, from a total of 89,790 cases of patients diagnosed with Parkinson’s disease and 118,095 matched individuals who had not been diagnosed with the disease in 2009 or prior years.
To build the algorithm, the team considered patients’ age, sex, race or ethnicity, history of tobacco smoking, and several clinical factors, including psychiatric conditions, trauma, cognitive impairment, fatigue, sleep disorders, weight loss, and diabetes, among others.
The model was able to predict with 85.7 percent accuracy if a patient would develop Parkinson’s. Specifically, the algorithm identified 73 percent of patients who would be diagnosed with Parkinson’s, and 83 percent of patients who would not.
Factors known to be associated with the disease, such as tremors, posture abnormalities, and psychiatric or cognitive dysfunction, helped the researchers identify which patients would develop the disease. But weight loss, gastrointestinal problems, chronic kidney disease, sleep disturbances, fatigue, and trauma (including falls) also were positive predictors of Parkinson’s.
On the other hand, other health factors like obesity-related conditions, cancer, cardiovascular disease, gout, and allergy, were not risk factors for Parkinson’s.
“We want to be able to catch people as early as possible,” Racette said. “If I know someone may be in the beginning stages of Parkinson’s disease, I would evaluate their gait and balance to determine if they have unrecognized impairments that could lead to falls, or whether they have difficulty performing activities of daily living. Either of these scenarios may benefit from treatment.”
In the 18 months preceding their diagnosis, Parkinson’s patients usually require several doctors’ appointments and medical tests to determine the cause of their symptoms. By recognizing the early signature of the disease, the new algorithm could improve early Parkinson’s diagnosis, while reducing the number of time-consuming and expensive diagnosis tests.
This study was funded by the National Institute for Environmental Health Sciences of the National Institutes of Health (NIH), the Michael J. Fox Foundation, and the American Parkinson Disease Association.