AI is driving one of the great breakthroughs in medicine: improving the lives of patients by improving outcomes, driving faster times to diagnosis and treatment, all while lowering costs. Be they undiagnosed rare disease patients, patients who would benefit from therapy line advancements or those at risk of non-adherence, this new capability is changing the way healthcare is delivered.

Over 7,000 rare diseases have been identified, per America’s Biopharmaceutical Companies, and the biopharma industry continues to innovate in its quest to develop treatments for as many as possible. In 2018, 59 new drugs were launched. Half of those were designated as orphan drugs; a third were first-in-class; and 46% launched with trials of less than 500 people. This environment leads to a major issue: how can physicians be expected to know and diagnose even a fraction of rare diseases, much less stay abreast of the constant innovations in treatments affecting so many small patient populations?

While treatments advance, it often falls on the manufacturers to educate and inform medical professionals not only about the treatment, but of the symptoms of the specific rare disease itself. General physician education is an option, but for the vast majority of physicians who do not and never will encounter a patient with a particular rare disease, carving out enough of an awareness foothold to support patients who do need that diagnosis is nearly impossible.

This leaves individuals who suffer from symptoms that look like a variety of more common conditions in a cycle, bouncing from their general practitioner to a line of specialists for an unspecified amount of time until they hopefully come across a physician familiar with their rare condition. But technology can break this cycle. Patient finding via AI and machine learning can help connect undiagnosed patients to the treatment they need. These algorithms can generate findings that shorten the process that typically leads to a rare disease diagnosis.

In one case study, Swoop and worked with a biotech company to help predict undiagnosed patients with an extremely rare, typically fatal genetic disorder; with less than 6,500 estimated patients in the US, the likelihood of a physician recognizing this disease was extremely low. By leveraging what the manufacturer knew about the 100 existing diagnosed patients, the diagnosed patients’ de-identified health data and Swoop and’s technology, a model was developed that resulted in patients being more likely to receive the correct diagnosis and therapy much earlier. Thanks in large part to Swoop’s actions, the patient population increased from 100 to over 500.

This improvement of patient outcomes does not end with rare diseases. Algorithms can find the most esoteric of patterns that may not make logical sense on the surface, helping predict any number of diagnoses both rare and common. In another case, Swoop and leveraged their technology and health data to identify patients with MS who were close to failing first-line therapy. By helping the client coordinate both direct-to-consumer and healthcare professional outreach programs, we were able to help a subset of patients suffering from a debilitating disease move off of a treatment path that was coming to an end and onto one more appropriate to their particular situations.

As technology advances, we must remember that the goal is to improve patient outcomes. Health care is a business, but one with a laudable purpose: to improve lives, and to shorten the path between illness and recovery. Biopharma manufacturers innovate in the creation of therapies, but educating the right physicians and patients on these new therapies remains a challenge that AI can help solve.