The life saving potential of the IoT

Hospital error is the sixth leading cause of preventable death in the US. It kills over 50,000 people every year in the US alone, and likely ten times more worldwide. It harms one in seven of those hospitalised and it frustrates doctors and nurses every day.

This problem is not new. Thirty years ago, the last major change in healthcare system technology changed hospital care through a simple realisation - monitoring patients improves outcomes. That epiphany spawned the dozens of devices that populate every hospital room, like pulse oximeters, multi-parameter monitors, electrocardiogram (ECG) monitors and more. Over the ensuing years, technology and intelligent algorithms improved many other medical devices, from infusion pumps (IV drug delivery) to ventilators. Healthcare is much better today because of these advances. But errors persist. Why?

Today, these devices all operate independently. There’s no way to combine the information from multiple devices and intelligently understand patient status. Devices therefore issue many nuisance alarms. Fatigued healthcare staff members silence the alarms, misconfiguration goes unnoticed and dangerous conditions go unaddressed. And as a result, people die.

Hazards during heart surgery
For instance, there are 14 infusion pumps, each administers a different drug to a single patient. As seen in Figure 1 (above), the pumps are completely independent from each other and the other devices and monitors. This picture is of an intensive care unit (ICU) - an operating room (OR) needs a similar array. During heart surgery, for instance, drugs sedate the patient, stop the heart, start the heart and more. Each drug needs its own delivery device, and there are many more devices, including monitors and ventilators. During surgery, a trained anesthesiologist orchestrates delivery and monitors status. The team has their hands full.

After surgery, the patient must transfer to the ICU. This is a key risk moment. The drug delivery and monitor constellation must be copied from the operating room to the ICU. Today, the OR nurse calls the ICU on the phone and reads the prescription from a piece of paper. The ICU staff must then scramble to find and configure the correct equipment. The opportunity for small slips in transcription, coupled with the time criticality of the change, is fertile ground for a deadly error.

Consider if instead these systems could work together in real time. The OR devices, working with a smart algorithm processor, could communicate the exact drug combinations to the Electronic Medical Record (EMR). The ICU system would check this data against its configuration. Paper and manual configuration produce far too many errors – this connected system eliminates dozens of opportunities for mistakes.

Connectivity for post operation
Once in post op, the danger is not over. Many patients use patient controlled analgesia (PCA) systems (see Figure 2). The PCA system allows the patient to self-administer doses of painkiller medication by pressing a button. The idea is that a patient with sufficient pain medication will not press the button, and therefore be safe from overdose. PCA is efficient and successful, and millions of patients use it every year. Still, PCA overdose kills one to three patients every day in the US. This seemingly simple system suffers from visitor interference, unexpected patient conditions, and especially false alarm fatigue.

Connectivity can also help here. For instance, low oximeter readings cause many alarms. They are only likely to be real problems if accompanied by a low respiratory rate. A smart alarm that checks both oxygen (SPO2) and carbon dioxide (CO2) levels would eliminate many distracting false alarms. An infusion pump that stopped administering drugs in this condition could save many lives. These are only a few examples. The list of procedures and treatments that suffer from unintended consequences is long. Today’s system of advanced devices that cannot work together is rife with opportunity for error. The common weakness? Each device is independent. Readings from one device go unverified, causing far too many false alarms. Conditions easily detected by comparing multiple readings go unnoticed. Actions that can save lives require clinical staff interpretation and intervention.

Data Distribution Service (DDS) standard
The leading effort to build such a connected system is the Integrated Clinical Environment (ICE) standard, ASTM F2761. ICE combines standards, it takes the data definitions and nomenclature from the IEEE 11073 (x73) standard for health informatics. ICE data communications leverage the Data Distribution Service (DDS) standard. ICE then defines control, datalogging and supervisory functionality to create a connected, intelligent substrate for smart clinical connected systems. For instance, the supervisor combines oximeter and respirator readings to reduce false alarms and stop drug infusion to prevent overdose. The DDS DataBus connects all the components with appropriate real time reliable delivery (Figure 3).

DDS is an IoT protocol from the Object Management Group (OMG). While there are several IoT protocols, most focus on delivering device data to the cloud. The DDS DataBus architecture understands and enforces correct interaction between participating devices. DDS focuses on the real-time data distribution and control problem. It can also integrate with the cloud, or connect to other protocols to form a complete connected system. Its unique capabilities fit the medical device connectivity problem well.

Clinical challenges
While the above examples and scenarios are simple, networking medical devices in a clinical environment is quite challenging. Information flow mixes slow data updates with fast waveforms. Delivery timing control is critical. Integration with data from the EMR must provide patient parameters such as allergies and diagnoses. Appropriate monitor readings and treatment history must also be written to the EMR. Large hospitals must match data streams to patients, even as physical location and network transports change during transfers between rooms. Devices from many different manufacturers must be coordinated.

ICE leverages DDS to address these clinical challenges. DDS models the complex array of variables as a simple global data space, easing device integration. Within the data space, the data-centric model elevates programmes to exchange the data itself instead of primitive messages. It can sift through thousands of beds and hundreds of thousands of devices to find the right patient, despite moves. DDS is fast and operates in real time. It easily handles heart waveforms, image data and time critical emergency alerts.

Dr. Julian Goldman leads ICE. He is a practicing anesthesiologist with an appointment at Harvard Medical School, and he is also the director of Bioengineering at the Partners hospital chain. His Medical Device Plug-n-Play (MDPnP) project at the Center for Integration of Medicine and Innovative Technology (CIMIT) connects dozens of medical devices together. MDPnP offers a free open source reference platform for ICE. There are also commercial implementations, including one from a company called DocBox. The CIMIT lab uses these to prototype several realistic scenarios, including the PCA and OR-to-ICU transfer scenarios described here. It demonstrates what is possible with the IoT.

New connected implementation
Hospital error today is a critical healthcare problem. Fortunately, the industry is on the verge of a completely new connected implementation. Smart, connected systems can analyse patient conditions from many different perspectives. They can aid intelligent clinical decisions, in real time. These innovations will save lives. This technology is only one benefit of the IoT future, which will connect many billions of devices together into intelligent systems. It will change every industry, every job and every life. And one of the first applications will extend those lives.