An analytics platform called Rulex has been chosen by Massachusetts General Hospital to be used within the Down Syndrome Program for its unique prescriptive capabilities, which offered the ability to not only build models for predicting obstructive sleep apnea (OSA) diagnosis—predictive analytics—but to develop a novel model that explains why OSA occurs—prescriptive analytics.
There is a 50% to 85% incidence of obstructive sleep apnea in individuals with Down Syndrome (DS), according to Rulex, with an overall incidence growing with patients’ age. Sleep-disordered breathing has been shown to affect cognitive abilities, behavior, growth rate, and the more serious consequences of pulmonary hypertension (abnormally high blood pressure in the arteries of the lungs) and cor pulmonale (failure of the right side of the heart).
Unfortunately, the ability of parents to observe sleep abnormalities in their children with DS has been shown to be poor. A polysomnogram continues to be the gold standard test from which to evaluate sleep-disordered breathing. However, sleep studies are often expensive, not readily available, poorly tolerated, and inconvenient, Rulex states.
In conducting the OSA analysis with Rulex, Massachusetts General Hospital has been able to build a new model for determining when a polysomnogram is absolutely needed for diagnosing OSA in individuals with DS. In fact, this model is able to predict with a high confidence (90%), if the children with DS is not affected by moderate or severe OSA. That is, this model can predict with high certainly when a person with DS might not need a sleep study.
To achieve this result, the Rulex’s LLM core technology has been adopted, which solves classification problems by providing a model described through a set of multi-variable if-then rules. Besides obtaining a very good accuracy when classifying a previously unseen patient, the physician can immediately understand which motivations have caused the prediction
One of the unique features of Rulex’s LLM is its ability of determining the set of relevant variables for the given problem even when the dataset includes a large number of attributes. In particular, in the analysis for diagnosing OSA, more than 450 variables have been considered and Rulex was able to detect the subset of them that are important for classifying individuals with DS.
Brian Skotko, MD, MPP, principal investigator and co-director of the Down Syndrome Program at Massachusetts General Hospital, says, in a release. “The statistical analyses were able to predict with high certainty when patients with DS were less likely to have moderate to severe obstructive sleep apnea and thus may not need a diagnostic sleep study.”
Andrea Ridi, Rulex CEO, says: “Rulex allows many diseases to be prevented and made more tolerable, simply by investigating the available data in an innovative way. Rulex prescriptive analytics provides researchers with answers they are seeking for individual patients. With that knowledge, individuals with DS and their families can make lifestyle changes to avoid risks and improve their quality of life.”