OR WAIT null SECS
How data science and machine learning can help patients with diseases like multiple sclerosis.
Machine learning can help patients with multiple sclerosis.
Big data analytics and machine learning are rapidly changing the healthcare landscape. These new approaches have the ability to not only accelerate the often long and difficult diagnostic process for chronic autoimmune diseases like multiple sclerosis (MS), but to monitor disease progression and therapeutic efficacy.
Patients who suspect they may have multiple sclerosis typically have to wait for their symptoms to progress for doctors to rule out all other possible diagnoses. During this wait — which averages four to five years, according to the National Multiple Sclerosis Society — they are undergoing expensive and sometimes invasive tests, ranging from spinal taps to MRIs. Historically, there hasn’t been a single test doctors could administer to confirm a diagnosis. For a final MS diagnosis, doctors must find evidence of damage in at least two separate areas of the central nervous system that occurred at different points in time.
Because the diagnostic process for MS can be difficult, the disease can remain undiagnosed or even be misdiagnosed in early stages. Lupus, fibromyalgia syndrome, Lyme disease and even migraines, among others, can present with similar symptoms to MS. The misdiagnosis rate ranges between 5-35 percent, according to a 2013 article published in Current Neurology and Neuroscience Reports by Andrew J. Solomon and Brian G. Weinshenker. Misdiagnosis can send patients down an expensive treatment path for a disease they do not actually have — a double-edged sword, as the patient’s actual disease remains undiagnosed and untreated while the wrong treatments can place patients in unnecessary risk.
Conventional wisdom says early diagnosis and treatment improve outcomes, which, in turn, can reduce the cost of managing patients with chronic disease. So how can a correct diagnosis be made as early as possible?
Big data analytics and machine learning hold the key to answering that question. Fortunately, a wealth of information — including, but not limited to, electronic medical records, clinical research data and payer claims data — can be collected and analyzed across the healthcare ecosystem. There’s also socioeconomic data — unemployment rates, education levels, personal income and crime rates, to name just a few — that can be important predictors of patient health. These data sets help model social determinants of human health and disease.
In the future, software engineers, business intelligence managers and bioinformatics specialists will play a vital role in helping clinicians diagnose and predict development of chronic disease. These teams will create data analytics platforms that combine population-level healthcare and socioeconomic data sets to create a comprehensive view of a patient population over time.
Looking ahead, adding genetic or genomic data sets to these population-level studies could increase the resolution of the insights delivered. By using data science tools like machine learning, patterns can be pinpointed in population data sets often composed of billions of data points, allowing for the detection and monitoring of autoimmune and related diseases — and even predicting an oncoming diagnosis that could arise months or years down the road.
How can this be done? It takes a careful combination of rich data sources and machine learning tools analyzed under the supervision of an experienced data scientist to uncover the subtleties and nuances hidden in the data that lead to meaningful insights. The first step is to look at large populations and find patterns in patient data. Then, as those patterns are investigated and compared to known outcomes, specific profiles emerge that identify someone at risk of developing a chronic disease. For example, the process may uncover a similar diagnostic process, such as several doctor visits over time to discuss numbness, blurred vision, development of urinary tract infections or migraines (all MS symptoms).
Taken individually, one might not see the relation between these symptoms. But a robust analytics platform allows data scientists to gather these data together, look at the entire patient history instead of a snapshot (like individual visits) and observe if other patients who ultimately go on to develop a chronic illness have faced a similar or different diagnostic process. Recognizing patterns in patient data allows for doctors to: 1) identify patients who need to be monitored for a possible diagnosis of a chronic disease; 2) discern the trajectory of the illness; and 3) flag suspected cases of misdiagnosis.
Big data adds credibility to a diagnosis — it provides supporting evidence that doctors do not have now. Additionally, once a diagnosis has been made, patients can be monitored for signs of future relapses, flares or other adverse events commonly seen in inflammatory diseases like MS. This gives physicians a window of opportunity to alter treatment plans and be proactive in the management of these diseases.
As an example of this approach, IQuity launched a pilot study in New York that focused on MS and analyzed healthcare claims for 20 million people — comprising 4 billion data points. The study identified patients who had been correctly diagnosed, as well as those who had been misdiagnosed. With over 90 percent accuracy, the analytics platform showed promising results, predicting the onset of MS within that patient population at least eight months before traditional methods would typically yield a diagnosis.
The healthcare industry has entered a new era in leveraging the power of predictive analytics to give providers accurate information to support an earlier diagnosis and give patients a higher quality of life — and this approach can be expanded to virtually any disease. When patients receive early, correct diagnoses, this translates into significant potential savings for self-insured employers and insurance companies and improved bottom lines for care management companies.
Applications of data science and use of predictive modeling, in particular, have the ability to forever change how care is delivered for the better. Using these approaches to predict, detect and monitor disease is just the tip of the iceberg.
Chase Spurlock, Ph.D., is CEO and co-founder of IQuity, a Nashville-based data science company using genomic and proprietary healthcare data sets to predict, detect and monitor chronic disease. For more information, visit www.iquity.com.
Navigate the digital transformation with confidence. Register for our newsletter.