A doctor once told my grandmother “let’s get you well as soon as possible – a hospital is no place for sick people.” It’s a doctor’s primary goal to treat a patient so they can return home, and stay home as quickly as possible. In the past, doctor’s were limited to the amount of information they had when deciding on a course of treatment – typically the patient’s historical information and the doctor’s training and past experiences. However, real-time machine learning based solutions can analyze limitless amounts of historical information and make deductions about a patient in real-time. Systems like this can help doctors diagnose and treat illnesses faster to get customer’s well and back home as fast as possible.

So what would a machine learning solution look like? Well, one could be using historical information to reduce hospital stay.  Hospital systems have been consistently collecting patient data for years. By having a large set of data that include symptoms, medical history, and outcomes, machine learning based solutions can identify patients that have a higher likelihood of staying the in the hospital for a longer than average time. For example, Patient X , a 50-year female, has just checked into the ER and is complaining of pressure in her jaw and nausea, has elevated blood pressure and pulse, a history of high cholesterol, is diabetic, and is 10 pounds underweight. At triage, a healthcare professional would see that 87% of people that were “similar” typically stayed in the hospital for 2 additional days over the average patient. The system could also tell the healthcare professional that 75% of these similar patients were having a mini-heart attack. Action could be taken immediately to treat the heart attack and possibly lead to a shorter hospital stay. Of course, this was an easy one, but data models can be trained to identify patients that have a high-risk of other illnesses that are commonly misdiagnosed or not identified quickly or even patients who are susceptible to contracting a hospital acquired infection. Overall, this system would be able to let doctors treat patients faster and get them well more quickly than before.

Another way to improve customer satisfaction is to reduce re-admissions. The cost of unplanned readmissions in the US is $41 billion every year. This is a huge number. However, traditionally it’s difficult to reduce this number since healthcare professionals typically don’t have any control over a patient’s behavior after the patient is discharged. Often high-risk patients don’t make follow up appointments, forget to take their medications and don’t monitor their vital signs.  However, with wearables becoming more common and 77% of Americans having smartphones, a patient’s activities can be easily monitored and patients can be alerted in real-time to take medication or schedule a doctor’s appointment. Also, the patient’s doctor could be alerted in real-time if certain events occur, like a spike in blood pressure or vital sign changes that could signal a heart attack or stroke. When these indications are detected, the doctor’s staff could preemptively call the patient to request an appointment or tell them to go to the ER immediately. Solutions like this could significantly lower the instances of re-admission.

Being in a hospital is rarely a fun experience. With machine learning based solutions hospitals can get patients correctly diagnosed, treated and discharged faster. For more information on creating a real-time machine learning based solution, check out our website or e-mail us at info@entrigna.com