MedCity Influencers

Does big data hold the secret to controlling sepsis?

A  big data, machine learning-based system could analyze the profiles of patients who have suffered from sepsis and build a profile that would show the specific data readings for each data point, alone and as they interact with other data points.

Abstract 3D network in future

As doctors well know, sepsis is among the most devastating conditions hospital patients experience. It’s the third biggest killer of hospital patients, and in the U.S. it’s the biggest single most expensive condition for hospitals to treat. But a solution is likely a long way off; the rate of septicemia and related conditions among hospital patients worldwide has been growing for at least a decade, and the growth shows no signs of letting up.

Why isn’t more being done to stem the problem?  One important reason is that sepsis is very difficult to diagnose as the symptoms can be similar to those of other conditions. Despite the wide variety of sepsis diagnosis tools available, including rule-based surveillance algorithms and smart alerts, current sepsis detection and notification systems are far from perfect. In the disease’s early stages – when catching it would be most effective – sepsis’s symptoms often are mild, gradually worsening. It’s only later that sepsis-specific symptoms develop – but by that time the disease will be much more difficult to battle. As a result, sepsis is one of the most misdiagnosed – and deadly – conditions among hospital patients.

But the successful treatment of sepsis may already be within reach – based on information in the hospital’s database. While the symptoms of sepsis are commonly conflated – or confused – with symptoms of other conditions, those symptoms do have their own profile: fever level, the effect on organs/bodily functions, blood pressure, and a thousand other factors. The data on that is in the patient’s electronic profile, and if analyzed properly, it could be used to identify the patients suffering from sepsis conditions.

According to the Agency for Healthcare Research and Quality, sepsis cost $20 billion to treat in 2011, $5 billion more than the runner-up condition, osteoarthritis. And as medical costs continue to rise, the costs of treating sepsis does as well. Beyond the costs, sepsis contributes to as many as half of all hospital deaths, according to JAMA, while other experts attribute even more deaths to the condition. And, again according to JAMA, the mortality rate from sepsis has been increasing in recent years.

Changing those numbers is where big data comes in. To understand the relationship between symptoms and sepsis, you need a system that could examine thousands of data points to find connections and conditions that would indicate a diagnosis of sepsis, as opposed to something else. The only way to do that is with a  big data, machine learning-based system; by analyzing the profiles of patients who have suffered from sepsis, you could build a profile that would show the specific data readings for each data point, alone and as they interact with other data points, associated with sepsis.

The data to accomplish this is already at hand. Data is being collected constantly from a variety of sources in the hospital. The monitors record the information available at any given moment. Analyzed properly, that information could provide the keys needed to save a patient. For example, a drop in blood pressure is a symptom associated with sepsis. But it is also associated with a host of other medical issues, including heart problems, anaphylaxis, or dehydration. It’s also found among pregnant women, and it can also be due to poor nutrition. However, the drop in blood pressure may be different for each of these conditions – and the blood pressure drop for those suffering from septicemia would coincide with other symptoms that would not necessarily be associated with any of those other conditions.

The analysis system would look at the data gathered about a patient and compare it with the profile of an early-stage sepsis sufferer. The closer the data matches that profile, the more likely a patient is to be suffering from sepsis. And because of the machine learning aspect of the system, it is constantly incorporating new data into the profile, allowing it to more accurately identify potential sepsis victims.

A system to detect sepsis and enable doctors to treat it at its early stages will help bring down mortality rates from the disease – something hospitals are increasingly going to be required to do. New York was the first state to implement regulations requiring hospitals to immediately treat sepsis when it is diagnosed. Under “Rory’s Regulations,” hospitals have three hours to administer a “care bundle” to treat septicemia. The regulations were implemented in 2013, and researchers have found that the immediate administration of the care bundle has led to “increased chances of survival for sepsis patients,” according to the New England Journal Of Medicine. While other states have not implemented such regulations yet, if mortality rates from sepsis decline in New York,  it’s likely just a matter of time.

As with many other diseases, early detection is the key to saving lives – and only a big data analysis system, analyzing the many data points, can differentiate between whether a patient is suffering from sepsis, or something else. The possibilities are vast, and the data is involved and often confusing. With time of the essence, any and every resource must be utilized in order to save lives. As it turns out, the most important resource doctors can have about a patient – information – is there for the taking. Utilizing a machine learning-based big data analytics system can help put that data in context, ensuring that it is available when it is needed to reduce the huge cost – financial, and especially human – that sepsis exacts from us.

Photo: from2015, Getty Images
















This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.