Sepsis is perhaps not an especially fashionable thing to target, yet it strikes over a million Americans each year, and kills more people per year than breast cancer, prostate cancer, and AIDS combined. So it’s a big deal, and as with most health related issues, the earlier it is diagnosed, the easier it is to treat successfully.
A recent study from Johns Hopkins University uses machine learning to aid in that early diagnosis. The study aims to better predict how diseases, and their treatments, will impact patients, and the risk of the patient developing infections as a result.
Patient data
As with all machine learning based projects, it rests upon good access to data. The system taps into the patient record, and aims to predict the trajectory of the patient’s health. In other words, whether they get better or worse. The researchers also believe that it could help clinicians better predict how different treatments will affect the patient.
The system is currently being piloted at John Hopkins, before hopefully then rolling out further afield. The trial will allow the team to monitor physician behavior, and how it influences practice.
Sepsis is the 11th biggest cause of death, so it’s undoubtedly a serious issue to tackle, and the team believe that the memory capabilities of the system allow it to develop a much richer and more complex range of models than a human can ever achieve. This allows a level of precision that is incredibly difficult to achieve today.
Of course, to do this, such systems need better access to patient data, and for healthcare providers to open up their electronic health records. It’s something I’ve argued for several times, with a fascinating case study of the potential of this emerging in the NHS recently.
Smarter use of data
Patient data company Validic recently announced a partnership with The Salford Royal NHS Foundation Trust of the National Health Service (NHS) to develop a new model of integrated care in the UK.
Validic allow user-generated data from mobile apps and wearable devices to be shared and merged with data contained in electronic patient records.
This kind of flexibility makes it easier for patients to use third party services such as that developed by the Johns Hopkins team, and seamlessly merge that with their hospitals electronic health record.
There is undoubtedly a huge amount of potential for healthcare to be positively disrupted by better use of data in the coming years, but it’s crucial that the process is managed effectively to ensure that patients win out of any future arrangements.
