3 things digital health developers can learn from the Industrial Internet of Things

In “3 practices Industrial IoT can learn from digital health and the Internet of Medical Things,” I shared some thoughts on what IIoT developers could learn from their healthcare counterparts – digital health or Internet of Medical Things (IoMT) developers. We saw how a “patient first/user first” focus is giving digital health applications a leg-up on adoption and achieving outcomes. Now we will look at it from the other side and see how IIoT practices can help developers in the digital health realm.

I started my career at Honeywell and was putting prototypes of WiFi-connected, body-worn systems onto operators in oil refineries in the mid-1990s. Like many who come from the industrial automation and controls (IAC) space, I can legitimately say that I’ve been working with the IIoT for more than 20 years.

Not surprisingly, the learnings I have to share with digital health developers come from the longer experience with and maturity of IoT technology in industrial applications. With that backdrop, let’s look at three ways developers and product managers in the healthcare space might learn from their counterparts in the industrial space.

Automation applied to everything

In the same way digital health developers are very patient-focused, IIoT developers tend to be very operationally focused. Two outcomes dominate their world – uptime and cost reduction – and automation has long been the solution to both. Automation takes many forms – acquisition, aggregation, analysis – but for IIoT developers, automation is all about getting action from data. They deploy sensors to automate and improve the integrity of data aggregation.

One IIoT colleague used to say “Look for clipboards. When you see someone carrying a clipboard there’s an opportunity for IoT.”  Gateways automate the aggregation of the sensors and networks automate the movement, storage, analysis and presentation of the data.

Recently, action from the data has come from getting structured data into the cloud. Centralization of information and control has long been a leading architecture for industrial operators. When I visited refineries in the early 1990s, we always stopped by the Main Control Room – the operational center of the refinery with control as the base function.

But today industrial automation and control is about the NOC – the Network Operating Center – because the data that the networks provide is the most important part of operations. Industrial NOCs are no longer on site. They span all sites of the operation and they maintain the automated acquisition, aggregation, and distribution of data to operational actions for the entire enterprise.

I have yet to see or discuss a NOC with a healthcare company. I am sure there are those who have already arrived at this point, but most digital health discussions are still at the individual patient or clinic level. This is appropriate as patients are key to outcomes, but the same was true at one point in time for industrial developers when they were focused on one machine or process site.

Success at the individual level must be scaled to the enterprise level and this will be a result of data automation and network management. In order to scale the cost savings targeted by the Quadruple aim, digital health IT organizations will have to extend their internal Electronic Health Record (EHR) focus to include and automate the wave of data that is coming from external networks of wearable sensors.

Digital twins and prediction machines help operations embrace data

Uptime is a big winner from automation. The opposite of uptime is, of course, downtime and in automated systems downtime usually comes from something breaking. Industrial developers attack downtime today with two key data-based tools – digital twins and prediction machines.

Digital twins are mathematical models for a piece of equipment or process. Formal Methods was the name the systems engineers used at Honeywell in the 1990’s – “digital” was only used by engineers then, not marketers. Digital twins, when accompanied by the right sensor data from subject asset, allow operators to better understand and, in the advanced case, predict the state of the machine, e.g. machine health. With the right mathematical or CAD (Computer Aided Design) models the software can automatically generate alerts when the equipment needs maintenance or is performing out of spec. Data leads to action.

When one combines a digital twin-based state machine with combinatorial probability, the systems designer now has a prediction machine that can anticipate the future of the machine. Prediction machines can be a priori, algorithmic-based, or can be automated and intelligent by watching sensor data along with operational outcomes of the machine. IIoT developers’ long time focus on data for automation and elimination of downtime for cost reduction have made them early and aggressive adopters of both digital twins and prediction machines.

Digital health developers should not ignore these practices because the “health of patients is very different from the health of equipment.”  True, a patient can be much more complex, but a digital twin does not need to be an all-encompassing model for the subject.

Consider the value of a simple measure of weight, heart rate, and breath rate over time for a Chronic Obstructive Pulmonary Disease (COPD) patient. Consider the value of knowing the physical and culinary activity of a diabetic when combined with glucometer readings. Models for patients are unlikely to be a priori, rather they can be derived from statistical, multi-parameter data sets.

Medical device developers are very familiar with this kind of analysis, but clinical trial data will be miniscule compared to the billions of readings per day that will come from patients with multi-sensor wearables like the Apple Watch. Digital health developers and clinicians will have to turn to the massive Patient Generated Health Data (PGHD) data lake to empirically create models of both patients and their behavior.

Patient activity sensors make prediction machines in healthcare much more interesting. The prediction could be analogous to industrial machines and do probability-based forecasting of the patient-state based on physical measurements. But if a data-driven system can predict patient behavior, then it could also affect that behavior with motivations and triggers to help patients avoid bad behavior and thus achieve their own better outcomes. This could be a new type of preventative care – take action before a malcondition presents.

Both data tools will improve outcomes and motivate the further acquisition and use of data by both care givers and patients, particularly PGHD as it becomes the dominant type of data driving the digital health platform. Digital twin use in digital health will drive an exponential adoption curve as it has in the industrial space.

Retrofit is a thing – a serious thing

As IIoT applications have gained recognition for their impact, developers have begun to look backwards in time to previous generations of systems and equipment already in the field.

“Retrofitting IoT technology is particularly relevant for services that involve expensive things that have long lifecycles including plant, cars, trucks and industrial machines. These things will be in operation long into future technological generations but the economic rationale for replacing them with IoT-enabled hardware isn’t there. “ – George Malim

A million-dollar investment for a new digital service on a piece of equipment that sells 1500 units per year is usually a non-starter at the CFO office. But most industrial equipment is expected to last 10 years and some as long as 30 years. When one of these durable equipment manufacturers considers retrofitting a digital service on the installed base, the return can go from service on 5000 units over the next five years to 50,000 new contracts over the next 12 months – a 50x improvement in ROI and the associated new service revenue has much higher margins.  Timothy Chou describes the IoT digital service retrofit opportunity in CFO terms – “How do we double our revenues and quadruple our margins using software?”

Chronic care is, by definition, care provided post-diagnosis. When a healthcare system considers a new digital health therapy or proposition, they must immediately look to their EHR system and ask, “How many patients do we already have under care with this situation?” Retrofitting in healthcare is potentially easier than in the industrial space because the patients are already under care. Providers can extend their retrofit market by sharing the proposition with payers, health insurance providers, who will have patients from multiple systems all of whom are accessible with a connected, remote monitoring therapy.

Devices-under-management, the key metric for industrial digital services, becomes Patients-under-management. Interestingly, the Centers for Medicare and Medicaid Services understand the value of continued service in chronic situations and has issued new remote patient monitoring codes to help providers move to this model and capture benefits patients, providers, and the enterprise.

Scott Nelson

While the two groups may not think they have much in common, IIoT and digital health developers must deal with the same innovation fundamentals to be successful with this new technology. Indeed, the Consumer Electronics Show (CES) is the only trade show where I see both in attendance and it is perhaps the best show case of IoT technologies. But “feasible” technology is just one part of an innovation using design thinking. What is interesting to me is that digital health teams have a keen focus on users, “desirability”, whereas IIoT teams focus on productivity and efficiency, “viability.” 

Together they have it all and can find the sweet spot of innovation.