Electronic coupling is the transfer of energy from one circuit or medium to another. Sometimes it is intentional and sometimes not (crosstalk). I hope that this column, by mixing technology and general observations, is thought provoking and “couples” with your thinking. Most of the time I will stick to technology but occasional crosstalk diversions may deliver a message closer to home.
Medically Deficient Technology
It has been a very challenging month helping my best friend who has been in an intensive care unit (ICU) following a stroke. It has been very difficult emotionally seeing him incapacitated as he makes a slow recovery with many ups and downs.
As expected, in a top-rated Silicon Valley hospital, technology abounds and permeates all aspects of patient care. However, I’ve observed many examples where the technology is insufficient, poorly implemented, or undone by human error. I’ve also seen several instances where a little bit of technology would go a long way to insure consistency in operational processes. Delivering a burger at a fast-food restaurant, albeit simpler, in some ways is more tightly managed than the delivery of medical care.
As a technologist I am disappointed seeing this gap between what technology should deliver and does deliver. At the same time it is encouraging in terms of the opportunities for improvement. More technology is not always the answer as we struggle to contain out of control medical costs that may not produce better clinical outcomes.
There is another aspect of the technology equation. In the commercial world we strive for “cost effective” solutions – the right technology with the proper return on investment (ROI). Hospitals are often not profit driven and in hospitals the mantra must be “patient effective” as human life is priceless. However, politics, salesmanship, and turf protection along with government regulations and the fear of lawsuits often create detours in effective patient care and implementation of new technology.
It is very unlikely that the ICU staff will engage in what a project manager would call a retrospective analysis during my friend’s stay unless something goes terribly wrong. And they certainly will not ask for the inputs of the slew of engineers and project managers present (him, his wife, colleagues, and friends). So even though the medical community is still “practicing medicine”, they are not using an overall continuous improvement process with a formal review and feedback system.
Central to the hospital experience are people: patients and staff. This provides challenges in all areas of hardware, software, and processes. It is very difficult to monitor people well. Even with several hundreds of thousands of dollars of instrumentation and equipment in his ICU room, technology has failed him more than once. Interfacing hardware, such as sensors, to people is difficult and measuring more than the “basics” is challenging.
Like most patients in the ICU, he had continuous monitoring of his “vital” signs. His electrocardiogram (EKG/ECG), blood pressure, respiration rate, pulse rate, saturated oxygen, and body temperature where continuously measured and data logged. Not only was this data visible bedside, it was displayed at the nurses’ workstations and on additional monitors throughout the ICU. When any of these parameters or calculated relationships between the parameters was out of range, alarms would sound for the nurses to respond. This constant electronic monitoring permits the staff to simultaneously juggle the care of multiple patients.
My friend has previously had seizures for which he takes medicine to control. While in the ICU he had additional seizures. However, none of the seizures were sufficient to disturb his vital signs to trigger an alarm. The neurosurgeon only learned of these seizures (which required an adjustment to the anti-seizure medicine) because his wife personally observed them. If not for this direct human observation, the seizures would have gone unnoticed possibly until very severe. There is a long way to go before medical electronics can continuously monitor more than just the basic vitals of the patient. The danger is the perception that today’s electronic monitoring is sufficient leads to complacency that leaves critical patients unobserved for long stretches of time while the staff is otherwise occupied.
Seizures can be detected by an electroencephalogram (EEG) which measures brain electrical activity using 16 to 25 leads attached to the scalp. These measurements are typically taken while the patient is at rest under study conditions. Therefore an EEG is not a practical tool for continuous patient monitoring. There is work underway to detect some seizures electronically based on algorithmically evaluating EKG waveforms to detect patterns and anomalies. Measuring an EKG is typically easier than an EEG in terms of leads and sensors.
For various imaging and waveforms such as EKGs, EEGs, X-rays, computerized tomography (CT) scans, magnetic resonance imaging (MRI), etc. “reading” is still a very highly specialized skill that may be more art than science. Ask several experts to read the same image and there may be a wide variance of interpretations. Efforts to automate reading EKGs for basic cardiac functions have made significant progress. However, software still has not proven the accuracy and range of detection especially of rare conditions required to replace the eyes and interpretation of a human specialist.
With seizures using an EKG for evaluation is even more subjective so the automation efforts appear to be further behind basic cardiac readings. I will be excited to hear about progress in building a smartphone based or connected sensor for monitoring cardiac health or seizures at an upcoming TSensors Summit (www.tsensorssummit.org). However, it is clear that not only must there be a sensing technology breakthrough (perhaps to provide the equivalent of a 12-wire EKG measurement wirelessly and non-invasively at low cost) there also needs to be significant progress in algorithm development and automated data interpretation. It may be several years before the dozen or so seizure tracking (via user data input) smartphone applications are replaced with a fully automated tracking device that is widely available.
There are many other areas in which automated sensing technology could be applied especially in terms of laboratory measurements. Even though it sounds simple to send a sample of blood or urine to the lab for analysis, these are multistep processes. At a minimum three people are involved beyond the patient: nurse collecting the sample, lab technician performing the analysis, and doctor interpreting the results. In reality, this number is likely to be higher. Where this breaks down is when a person does something wrong or fails to act. Repeat around the clock for hundreds of patients and thousands of samples and there are too many opportunities for failure: the nurse forgets to collect the sample, the doctor neglects to review the data, the lab technician transposes a sample or number, etc. I personally witnessed some of these human failures.
Many of these measurement are candidates for a point of care (POC) real time measurement system. By fully automating simple measurements, the primary sources of error (people) can be eliminated. In addition, a higher quality data stream can be generated with more frequent measurements. This would allow the improved feedback to adjust the treatment protocol with greater sensitivity and enable computerized alarms for changes in levels. The challenge is to design a product with sufficient accuracy and robustness that is also easy to implement and operate at a reasonable price. With all the manual process steps involved today, the ROI of an automated solution should hopefully be clear.
On a positive note is the performance of the highly visible technology the hospital staff uses constantly: hands-free one-button voice-dial clip-on communicators. Think shirt-pinned communicators like those from Star Trek: The Next Generation. With our current technology they are slightly bigger than the Starfleet communicator pins and they only work within certain areas of the hospital (excellent WiFi coverage is required). They are however rather efficient and appear to work well to improve communication and responsiveness of the staff.
Yes, one can do hands-free voice dialing using a Bluetooth headset on current smartphones. What is different is the hospital system simply works and does so intuitively. The staff doesn’t fuss with them, the interface is simple, and it performs as designed. Bounding the system complexity by having a fixed number of user names, known job functions, etc. helps the system perform well. To make the technology work as well as envisioned in science fiction requires product management to focus the product on essential core functionality and a fair amount of behind the scenes infrastructure. This success demonstrates that engineers and product managers should not give up hope since it is possible to overcome the additional hurdles and implement successful advanced technology in a medical setting.
Addressing software and process/procedural challenges I’ll leave for future columns. (If you wish to discuss them sooner, please let me know.) I am headed back to the hospital to visit my friend now. Who knows what else I will learn or see while there?
As always, I look forward to hearing your comments directly. Please contact me to discuss your thoughts or if I can be of any assistance.