The Internet of Things (IoT) and artificial intelligence (AI) are increasingly ubiquitous in our everyday lives. By some estimates, the IoT in healthcare will grow from a market size of $60 billion in 2014 to $136 billion by 2021. The potential applications and benefits to healthcare are vast. Value-based care, where patient outcomes are improved while the corresponding resources are reduced, is an especially exciting application for the IoT and AI.
To date, capturing data on patient outcomes to support the delivery of value-based care has been challenging for a number of reasons. Patients are individuals, each with their own genetic makeup and co-morbidities. Finding a common language, or ontology, to describe patients’ diagnoses, treatments and outcomes has also been difficult. Using an ontology is necessary to encourage consistency in documentation not only within, but also across organizations. In this way, insights may be gleaned about patient’s diagnoses, treatments, and outcomes.
From a resource perspective, it’s been difficult to get a true sense of how devices are being used in the field on a large scale. Healthcare technology management professionals (HTMs) get a snapshot of device use when issues are reported by users and error logs are reviewed and analyzed, but undertaking these activities on a large scale isn’t usually possible for HTMs on an ongoing basis.
The IoT holds promise in this area. IoT is the term used to describe technology connected to the internet for which data about its operation (e.g. various states, errors, inputs and outputs) may be harnessed. When this data is captured it holds the potential for insight, however, capturing the data is just the beginning; analyzing this data, along with the data of other similar devices is where things start to get interesting.
For example, consider the challenges associated with alarm management. Capturing data about the states of multiple devices, how clinicians are interacting with those devices (e.g. cancelling alarms) and real-time patient data can support our understanding of how to appropriately set alarm parameters, identify sensor issues and accommodate cases where clinically unimportant changes are happening to the patient (e.g. the patient rolls over).
Although the information required to glean these insights is available now, we often still have trouble capturing and analyzing data in a meaningful way simply because there is so much information available to us. This is where AI can help. AI refers to an algorithm’s ability to first describe and then predict or optimize future states based on a known set of data. Through AI, it’s possible for many data sets to be analyzed without humans having to know a priori exactly which questions to ask.
In the future, a hospital administrator using AI might receive a notification suggesting the sensitivity of infusion pump alarms in the surgical intensive care unit (SICU) be decreased by 10% based on patient factors, as well as the number of alarms and corresponding responses by clinicians (e.g. cancelling alarms). The resulting decrease in the frequency of alarms allows nurses to respond more effectively and efficiently to patients requiring assistance. Because there are fewer alarms, they’re also more likely to respond each time an alarm sounds. Finally, the average patient gets more uninterrupted sleep, has lower blood pressure and is discharged half a day earlier because the SICU is quieter.
For all the potential of the IoT and AI, there are still many questions, challenges, and hurdles to overcome. Patient privacy, data sharing and ownership, social license, data quality, security, technology standards, ontology standards and change management are just some of the areas requiring advancement. However, if we can get these things right, huge strides in value-based care can be achieved.
Melissa Kozak is a clinical engineer at the Techna Institute with University Health Network in Canada.