In many ways, internet of things (IoT) is a double-edged sword: connected devices are capturing huge volumes and varieties of data that can be mined for everything from potentially life-saving health care information to guidance toward peak athletic performance, but it is incredibly difficult to convert that raw data into truly meaningful and actionable insights.

IDC projects that, by 2025, IoT devices will generate more than 73 zettabytes of data globally – that’s 73 billion terabytes – and 152,200 IoT devices will connect to the internet every minute. In 2019, we saw 86 percent of health care organizations using some form of IoT technology, including a whopping 646 million IoT devices. That, coupled with increasing demand for remote monitoring and telemedicine created by the pandemic is driving a global IoT market that is expected to exceed $158 billion by the end of this year – and more than $534 billion by 2025.

The massive data contains metrics that can be used for everything from treating traumatic injuries to remotely monitoring vital signs. The real challenge is identifying, analyzing, and delivering the right data to the clinician, patient, or insurance provider in an understandable, clinically relevant, and actionable way.

Therein lies the challenge.

Driven in large part by the proliferation of connected devices for clinical and consumer use, the volume of medical data doubles every 73 days. Unless that information can be analyzed and converted into something actionable, it only amplifies the amount of data generated by IoT.

I have spent a significant part of my pharmaceutical research and development career. Discovering or developing a successful pharmaceutical is a daunting task, and the pharmaceutical industry has and continues to wrestle with the same fundamental issue regarding large datasets as does IoT. There are basically two ways to design a drug.

One approach is the combinatorial method, which parallels the IoT method of analyzing large data sets; grabbing everything we can get our hands on, dumping it into the test tubes, and testing it repeatedly until we (hopefully) isolate the part that will produce the desired effect.

Conversely, the rational drug design approach is based on studying the structures, properties, and functions of the desired molecule and its binding site. We start with a body of scientific knowledge and work methodically toward the answer.

As a founder of an AI-driven biotechnology company, I am working with my data science team to combine these two approaches to rationally generate and mine relevant and actionable large datasets – to sort out the informational “noise” and get us down the path toward meaningful results, “AI IQ.”

In order to move health care analytics forward and produce impactful results, that data must be properly categorized, generated, and analyzed, or we will be simply doing work for work’s sake.

Categorize the data and build the model

Whether IoT devices capture temperature, sound, movement, or other quantifiable or qualitative information, the problem lies in white noise. That is where the categorization of the data becomes a necessary first step. At this juncture, it is pertinent to decide what is the process for evaluation and identify what data is necessary for an optimal endpoint. Once that framework is developed, and the data is categorized, the model for analysis can be constructed.

Generate relevant data

After the data is identified and categorized, an organization can then determine the method to capture the data – a process that must be precise, accurate, and reproducible.

The proper AI software allows relevant data to be analyzed from any source. For example, in health care, accessible, clinically relevant, and actionable functional motion data ultimately enables virtually any clinical specialty to perform medically necessary tests at the appropriate frequency as part of a comprehensive diagnosis and treatment plan and achieve the quality outcomes necessary in a true value-based model of care.

Analyze the data using AI software

Once all the significant information is collected, it can then be streamed instantly back to the app, or cloud-based analytics software capable of crunching the data and converting it into something actionable for the individual recipient or to a broader audience.

AI software can then analyze relevant data points and relationships quickly and robustly to provide relevant predictive analytics about the processes and outcomes.

At this juncture, non-optimal parts of the process can be identified and changed if necessary, and the process constantly be reassessed to determine the efficacy of that change.

Utilizing the proper processes will be transformational for the health care industry. As artificial intelligence/machine learning and mobile technology are applied to IoT-generated data, health care organizations can deliver precise, accurate, and reproducible results that will optimize every aspect of the health care industry.

However, we need to have the cooperation of all the elements of the health care system if we are to fully capitalize on and optimize the data-rich IoT environment by educating organizations on the proper way to analyze IoT data. We need to be very thoughtful and spend time and resources on both the type of data generated by the IDC-estimated 75 billion IoT devices expected to be on the market by the end of 2025 and the predictive analytics tools that will give IoT a realistic chance to advance health care.

Frank Fornari is chairman and founder, BioMech, a leading-edge biotechnology company that develops and distributes real-time motion analytics and artificial intelligence/machine learning solutions, including BioMech Lab^TM and Coretex^TM that quantify and improve outcomes in health care, sports/wellness, and industrial sectors.

Image credit: Shutterstock.com