Connected Doesn’t Mean Cared For

The following is a guest article by Ryan Carlson, Technology Evangelist at Soracom

A hospital full of sensors is not a safer hospital. It is a hospital with sensors.

That distinction has taken the healthcare industry about a decade and a significant amount of capital to absorb. Device count is the wrong scoreboard. The measure of a healthcare IoT deployment isn’t whether the signal reached the cloud. It’s whether the right person received the right information in time to do something that positively changed what happened to a patient.

Achieving this outcome requires more than IoT connectivity. It requires a care redesign program that happens to involve technology, and in 2026, the path between those two things is still more complicated than anyone who drafted the original interoperability legislation anticipated.

FHIR Works, Until It Doesn’t

The healthcare interoperability story is technically true and practically frustrating. Yes, Fast Healthcare Interoperability Resources (FHIR) adoption is up. In 2024, about 9 in 10 hospitals enabled API-based patient access. The Centers for Medicare & Medicaid Services (CMS) has extended FHIR-based requirements into payer workflows, covering patient access, prior authorization, and payer-to-payer exchange. Progress is real.

But “FHIR-supported” is not the same as “plug-and-play,” and anyone who has tried to build a third-party clinical IoT application in the last five years knows exactly what that gap looks like in practice.

FHIR APIs tend to expose easy data: demographics, medications, labs, allergies, and basic notes. The clinically useful workflows, remote monitoring, care management, longitudinal care plans, and clinical decision support are harder to reach, more inconsistent across implementations, and frequently still dependent on HL7 v2 data standard messages, Clinical Document Architecture (CDA) documents, flat files, and manual workarounds that politely ignore the existence of the standard. Two hospitals running a system from the same Electronic Health Records (EHR) vendor may still require completely different integration work because local customization broke whatever standardization the transport layer offered. Terminology collisions between RxNorm, SNOMED, LOINC, ICD-10, local codes, and payer codes mean a FHIR resource can carry data cleanly while the underlying meaning is still up for interpretation.

The walled garden problem is real and underacknowledged. EHR vendors have implemented enough of FHIR to satisfy regulatory requirements and enough to keep appointments scheduled, records requests managed, and payers processed. What parts of FHIR would make it easier for a third-party developer to build something clinically meaningful without a vendor contract, a security review, a custom interface build, and an ongoing maintenance relationship? That progress moves slower because ease of integration has never been a priority for entrenched ecosystems. This means meaningful market penetration often requires supporting multiple dominant EHR ecosystems simultaneously, each with its own APIs, security reviews, implementation processes, data models, and operational requirements.

This is the operational reality of building connected solutions in healthcare. But the existence of workarounds is not a sign that innovation has failed. Workarounds are often the bridge that allows clinically valuable solutions to reach patients years before the healthcare ecosystem achieves the level of interoperability maturity and standardization already expected in most other industries.

What Actually Works

The IoT deployments that produce measurable outcomes in 2026 are not the ones that assumed clean data flows and direct integrations would be available. They’re the ones that built a care model around these constraints.

Remote patient monitoring is the clearest example of overcoming interoperability hurdles. When data doesn’t move cleanly into Electronic Health Records (EHR), the programs that succeed build a care operations layer: remote nurses, pharmacists, care coordinators, or monitoring teams who review alerts, contact patients, document actions, and surface only what requires a provider’s attention. The data flows imperfectly. The care team fills the gap. That’s not a workaround. That’s a care redesign.

The Accountable Care Organization community, focused on value-based care arrangements, has quietly been doing the most interesting work. When financial incentives shift from volume to outcomes, the business case for prevention and early intervention becomes real. A connected scale that catches a heart failure patient’s fluid retention before it becomes a readmission is worth something specific and calculable in that model. A cellular glucose monitor that reaches a rural patient who doesn’t have reliable broadband and wasn’t previously reachable by traditional care coordination is not a nice-to-have. It is, for some patients, the difference between managed disease and avoidable crisis.

Cellular IoT connectivity matters here in a way that gets underappreciated. The populations with the greatest need for remote monitoring are often the same populations least likely to have reliable home internet. Devices that work on cellular reach patients that portal-based solutions never will, and AI helps extract meaning from the data flowing in from sensors

On AI: Useful, Not Magic

AI can have a positive impact in clinical settings. It can be used to flag deteriorating trends, prioritize alert queues, identify patients at elevated readmission risk, and surface patterns across populations. Those are legitimate uses.

What AI cannot do is compensate for a care model that wasn’t designed around clinical intent in the first place. A sepsis risk score delivered to a system that a nurse checks once per shift is not an early warning. It is documentation. The value is in the workflow, not the model.

Healthcare providers already spend more time looking at EHR screens, entering data, and answering qualifying questions than they spend looking at patients. AI that automates the administrative burden and returns clinical attention to the patient is useful. AI that adds another dashboard to monitor, another alert to evaluate, and another system to reconcile is not innovation. It’s overhead with better branding.

The FDA’s own guidance on AI in healthcare emphasizes the performance of the human-AI team rather than the model in isolation. That framing is right. The question is never “What can the algorithm detect?” It’s “What does a clinician do differently because of it, and does that change the patient’s outcome?”

The Actual Measure

Healthcare IoT done right is a care delivery story with a technology chapter setting the plot in motion.

The sensor connects. The data moves. The care coordinator calls. The medication adjusts. The patient doesn’t get readmitted. That chain of events, not the device count, is the outcome. Designing it requires honesty about where the infrastructure resides, respect for the clinicians being asked to use the system, and enough humility to build the care model around reality rather than the version of interoperability the standards bodies promised.

We’ll get there. Probably not all at once, and probably not without continuing to invent workarounds that shouldn’t be necessary. That’s healthcare technology in 2026. It’s also, when it works, genuinely worth the effort.

About Ryan Carlson

Ryan Carlson is the Technology Evangelist at Soracom, a company specializing in cloud-native cellular connectivity for IoT applications. Ryan has helped pioneer connected products in energy, healthcare, transportation, and commercial services as a product owner, solutions architect, researcher, and principal IoT consultant. He has first-hand experience in product design, health data interoperability, user research, IoT corporate strategy, and overseeing product development and go-to-market strategies.

The following is a guest article by Ryan Carlson, Technology Evangelist at Soracom

A hospital full of sensors is not a safer hospital. It is a hospital with sensors.

That distinction has taken the healthcare industry about a decade and a significant amount of capital to absorb. Device count is the wrong scoreboard. The measure of a healthcare IoT deployment isn’t whether the signal reached the cloud. It’s whether the right person received the right information in time to do something that positively changed what happened to a patient.

Achieving this outcome requires more than IoT connectivity. It requires a care redesign program that happens to involve technology, and in 2026, the path between those two things is still more complicated than anyone who drafted the original interoperability legislation anticipated.

FHIR Works, Until It Doesn’t

The healthcare interoperability story is technically true and practically frustrating. Yes, Fast Healthcare Interoperability Resources (FHIR) adoption is up. In 2024, about 9 in 10 hospitals enabled API-based patient access. The Centers for Medicare & Medicaid Services (CMS) has extended FHIR-based requirements into payer workflows, covering patient access, prior authorization, and payer-to-payer exchange. Progress is real.

But “FHIR-supported” is not the same as “plug-and-play,” and anyone who has tried to build a third-party clinical IoT application in the last five years knows exactly what that gap looks like in practice.

FHIR APIs tend to expose easy data: demographics, medications, labs, allergies, and basic notes. The clinically useful workflows, remote monitoring, care management, longitudinal care plans, and clinical decision support are harder to reach, more inconsistent across implementations, and frequently still dependent on HL7 v2 data standard messages, Clinical Document Architecture (CDA) documents, flat files, and manual workarounds that politely ignore the existence of the standard. Two hospitals running a system from the same Electronic Health Records (EHR) vendor may still require completely different integration work because local customization broke whatever standardization the transport layer offered. Terminology collisions between RxNorm, SNOMED, LOINC, ICD-10, local codes, and payer codes mean a FHIR resource can carry data cleanly while the underlying meaning is still up for interpretation.

The walled garden problem is real and underacknowledged. EHR vendors have implemented enough of FHIR to satisfy regulatory requirements and enough to keep appointments scheduled, records requests managed, and payers processed. What parts of FHIR would make it easier for a third-party developer to build something clinically meaningful without a vendor contract, a security review, a custom interface build, and an ongoing maintenance relationship? That progress moves slower because ease of integration has never been a priority for entrenched ecosystems. This means meaningful market penetration often requires supporting multiple dominant EHR ecosystems simultaneously, each with its own APIs, security reviews, implementation processes, data models, and operational requirements.

This is the operational reality of building connected solutions in healthcare. But the existence of workarounds is not a sign that innovation has failed. Workarounds are often the bridge that allows clinically valuable solutions to reach patients years before the healthcare ecosystem achieves the level of interoperability maturity and standardization already expected in most other industries.

What Actually Works

The IoT deployments that produce measurable outcomes in 2026 are not the ones that assumed clean data flows and direct integrations would be available. They’re the ones that built a care model around these constraints.

Remote patient monitoring is the clearest example of overcoming interoperability hurdles. When data doesn’t move cleanly into Electronic Health Records (EHR), the programs that succeed build a care operations layer: remote nurses, pharmacists, care coordinators, or monitoring teams who review alerts, contact patients, document actions, and surface only what requires a provider’s attention. The data flows imperfectly. The care team fills the gap. That’s not a workaround. That’s a care redesign.

The Accountable Care Organization community, focused on value-based care arrangements, has quietly been doing the most interesting work. When financial incentives shift from volume to outcomes, the business case for prevention and early intervention becomes real. A connected scale that catches a heart failure patient’s fluid retention before it becomes a readmission is worth something specific and calculable in that model. A cellular glucose monitor that reaches a rural patient who doesn’t have reliable broadband and wasn’t previously reachable by traditional care coordination is not a nice-to-have. It is, for some patients, the difference between managed disease and avoidable crisis.

Cellular IoT connectivity matters here in a way that gets underappreciated. The populations with the greatest need for remote monitoring are often the same populations least likely to have reliable home internet. Devices that work on cellular reach patients that portal-based solutions never will, and AI helps extract meaning from the data flowing in from sensors

On AI: Useful, Not Magic

AI can have a positive impact in clinical settings. It can be used to flag deteriorating trends, prioritize alert queues, identify patients at elevated readmission risk, and surface patterns across populations. Those are legitimate uses.

What AI cannot do is compensate for a care model that wasn’t designed around clinical intent in the first place. A sepsis risk score delivered to a system that a nurse checks once per shift is not an early warning. It is documentation. The value is in the workflow, not the model.

Healthcare providers already spend more time looking at EHR screens, entering data, and answering qualifying questions than they spend looking at patients. AI that automates the administrative burden and returns clinical attention to the patient is useful. AI that adds another dashboard to monitor, another alert to evaluate, and another system to reconcile is not innovation. It’s overhead with better branding.

The FDA’s own guidance on AI in healthcare emphasizes the performance of the human-AI team rather than the model in isolation. That framing is right. The question is never “What can the algorithm detect?” It’s “What does a clinician do differently because of it, and does that change the patient’s outcome?”

The Actual Measure

Healthcare IoT done right is a care delivery story with a technology chapter setting the plot in motion.

The sensor connects. The data moves. The care coordinator calls. The medication adjusts. The patient doesn’t get readmitted. That chain of events, not the device count, is the outcome. Designing it requires honesty about where the infrastructure resides, respect for the clinicians being asked to use the system, and enough humility to build the care model around reality rather than the version of interoperability the standards bodies promised.

We’ll get there. Probably not all at once, and probably not without continuing to invent workarounds that shouldn’t be necessary. That’s healthcare technology in 2026. It’s also, when it works, genuinely worth the effort.

About Ryan Carlson

Ryan Carlson is the Technology Evangelist at Soracom, a company specializing in cloud-native cellular connectivity for IoT applications. Ryan has helped pioneer connected products in energy, healthcare, transportation, and commercial services as a product owner, solutions architect, researcher, and principal IoT consultant. He has first-hand experience in product design, health data interoperability, user research, IoT corporate strategy, and overseeing product development and go-to-market strategies.