Wireless Connectivity Is A Data Integrity Issue, Not IT
By John Oncea, Chief Editor, Clinical Tech Leader
When clinical operations teams think about wireless connectivity, the mental model is usually binary: either the device has a signal, or it doesn’t. Yasir Shafiullah, whose career started in 3G and 4G field optimization before moving into RF and analog IC design at the University of Oulu’s CWC-RF department, thinks that framing undersells what’s actually happening, and what’s actually at stake.
“From an RF perspective, when we send data over a poor network connection, packets can be lost over the air,” Shafiullah said. “If we send a one-gigabyte file, the receiving side may not receive the full file because packets are lost due to poor reception.” A connection can register as “working” on a status bar while still quietly failing to deliver complete, decodable data, which, for a clinical trial, means the difference between a usable data point and a gap nobody notices until analysis.
When Data Arrives But Can’t Be Trusted
Shafiullah described a technical threshold that matters more than most people realize: error vector magnitude. A signal can be received with some error and still be decoded correctly up to a point. Past that point, the data, even if it technically “arrived,” can’t be reliably decoded at all. Poor Wi-Fi, inconsistent local providers, and lack of satellite backup all push a connection toward that threshold. In a decentralized trial spread across mountainous terrain or rural coverage gaps, some units will always be closer to that edge than others.
Multipath interference adds another layer. In dense environments, cities with tall buildings, or terrain with reflective surfaces, a signal can arrive at the receiver more than once, echoing off obstacles with slightly different delays. Reconstructing a clean signal from those overlapping reflections is a signal-processing problem trial sponsors rarely think about, but it’s constantly happening in the background of any wireless monitoring device.
Geography Can’t Be Fixed, But It Can Be Engineered Around
Asked how engineers compensate for geographic connectivity gaps, Shafiullah didn’t pretend there’s a workaround for physics: additional network infrastructure, signal-processing techniques to reconstruct multipath-affected signals, and, increasingly, satellite connectivity options like Starlink for genuinely remote sites. He also flagged something worth remembering: Bluetooth, the protocol most wearables rely on, has an inherently short range, and disconnections when a patient simply walks too far from a receiver are a normal, expected limitation rather than a device defect.
Battery Life Is A Connectivity Decision, Not Just A Convenience Feature
The tradeoffs don’t stop at signal strength. Battery constraints shape every other engineering decision in a wearable device, according to Shafiullah: how often a sensor samples, how much processing happens on device versus offloaded to a phone or the cloud, and how much redundancy is built into transmission. A device engineered to extend battery life by sampling less frequently or transmitting at lower power is, by definition, making a data-quality tradeoff, even if that tradeoff is invisible in the vendor’s marketing.
He was candid that he hasn’t tracked which battery technologies are currently medically graded or best suited for clinical use, a fair reminder that this is a fast-moving area worth asking vendors about directly rather than assuming.
What This Means For Evaluating Vendors
The practical takeaway for sponsors and CROs: “good Wi-Fi” is not a sufficient answer to a connectivity question. Worth asking instead: what’s the acceptable packet loss threshold before data is considered unusable, how does the device handle multipath-heavy environments, what happens during a Bluetooth disconnection, and what battery-driven sampling or power tradeoffs were made in the device’s design. Each of these affects the completeness and reliability of a dataset in ways a signal-strength indicator won’t reveal.
Connectivity and battery tradeoffs sit downstream of the sensor and synchronization issues covered in “Why Trustworthy Clinical Trial Data Starts as Analog, Not Digital” and “The Synchronization Problem Clinical Research Isn’t Watching;” a well-designed, well-synchronized sensor still depends on the network layer to get that data home intact. Once the data does arrive, the final question is what AI can and can’t do with it, which is where What AI Can’t Fix in Decentralized Clinical Trial Data picks up, coming August 17.