ALS is a clinical diagnosis of exclusion with the average delay in ALS diagnosis being 1 year after the appearance of the first symptom. This prolonged diagnostic time can be detrimental as it delays initiating approved treatments and may preclude patients from enrolling in clinical trials.

The main objective was to provide actionable insight to better identify ALS patients earlier in their disease progression. The team used identified sets of medical predictors to analyze if patients are actually symptomatic prior to their diagnosis. From a medical perspective, the sooner that people get treated, even if there is no cure to date, the better their quality of life.

Insurance claims databases contain detailed patient histories including diagnoses given, procedures performed, drugs prescribed, providers seen, and treatment facilities visited over the course of many years for millions of patients. When aggregated and de-identified, these claims databases allow large populations of patients with specific diseases to be studied using big data analysis methods. A global government-grade, data analytics platform was used to sort through and apply multiple analytic and machine learning methodologies to analyze the pre-ALS histories of thousands of ALS patients to search for signals that could point to an eventual ALS diagnosis.

To our knowledge, this was the largest set of ALS patients ever studied for pre-diagnosis signals of ALS. The analysis methods used identified numerous clinically relevant signals of ALS commonly observable in ALS patients before diagnosis.

Two hundred million health records were analyzed to identify 13,000 ALS patients. The data demonstrated some patients had symptoms up to 5 years before diagnosis. Features that differentiate ALS patients before diagnosis in claims data can allow clinicians to diagnose ALS earlier in the disease progression, improve the speed of referral to an ALS specialist, and allow for quicker initiation of appropriate therapy.

Applying big data analytics to a large claims database and utilizing dissimilar Real World Data such as EHR, genomics, consumer, social media, claims, census and environmental data can help to enable physicians to identify and treat patients unknowingly suffering from diseases like cancer or diabetes sooner. This could improve, or lengthen the lives of, thousands more people.

ALS, or Lou Gehrig's Disease, weakens muscles to a point where the person cannot walk, swallow, or even breathe. The only outcome is death, usually within 3 years after diagnosis. The medical community believed patients were symptomatic only 12-18 months before diagnosis. Now, through a new method of healthcare analytics we can prove that patients actually have symptoms up to 5 years before diagnosis. This new breakthrough shows trackable symptoms that could help physicians diagnose ALS earlier than previously believed, helping improve or lengthen the lives of thousands of patients.

May 2016 - Analysis begins on ALS

June 2016 - Patient definition complete and data set identified and acquired

August 2016 - Demonstration of proof of concept in a pilot study

September 2016 - Expanded the project to a nationwide study

March 2017 - Nationwide study completed

March 27-30, 2017 - Results of initial pilot study showing ALS patients in 4 U.S. states displayed signals of ALS in their claims histories up to 60 months before initial ALS diagnosis were presented at the Academy of Managed Care Pharmacy (AMCP) Managed Care & Specialty Pharmacy Annual Meeting.

Digital transformation of the healthcare sector has resulted in an unprecedented volume of available healthcare data. These data offer great potential to provide new insights into debilitating and life-threatening diseases, which could improve the lives of patients living with these diseases. To harness the potential of these data, technological methods and tools capable of extracting meaningful information from large aggregated datasets must be developed for, and applied to, the healthcare sector.

In one such application, predictive analytics and machine learning methods were applied to a 200-million patient longitudinal healthcare claims database to search for signals in the pre-diagnosis histories of patients with amyotrophic lateral sclerosis (ALS) that could enable earlier identification and treatment of these patients, and ultimately lead to life-altering advances in the treatment of the disease.