It has been over two years since I have written on this subject and it continues to be something that I get excited about whenever a publication comes my way on the topic. The last time I looked at this topic it was after the publication of a randomized trial comparing in which one arm was provided automated FiO2 adjustments while on ventilatory support and the other by manual change. Automated adjustments of FiO2. Ready for prime time? In this post I concluded that the technology was promising but like many new strategies needed to be proven in the real world. The study that the post was based on examined a 24 hour period and while the results were indeed impressive it left one wondering whether longer periods of use would demonstrate the same results. Moreover, one also has to be wary of the Hawthorne Effect whereby the results during a study may be improved simply by being part of a study.
The Real World Demonstration
So the same group decided to look at this again but in this case did a before and after comparison. The study looked at a group of preterm infants under 30 weeks gestational age born from May – August 2015 and compared them to August to January 2016. The change in practice with the implementation of the CLiO2 system with the Avea ventilator occurred in August which allowed two groups to be looked at over a relatively short period of time with staff that would have seen little change before and after. The study in question is by Van Zanten HA The effect of implementing an automated oxygen control on oxygen saturation in preterm infants. For the study the target range of FiO2 for both time periods was 90 – 95% and the primary outcome was the percentage of time spent in this range. Secondary outcomes included time with FiO2 at > 95% (Hyperoxemia) and < 90, <85 and < 80% (hypoxemia). Data were collected when infants received respiratory support by the AVEA and onlyincluded for analysis when supplemental oxygen was given, until the infants reached a GA of 32 weeks
As you might expect since a computer was controlling the FiO2 using a feedback loop from the saturation monitor it would be a little more accurate and immediate in manipulating FiO2 than a bedside nurse who has many other tasks to manage during the care of an infant. As such the median saturation was right in the middle of the range at 93% when automated and 94% when manual control was used. Not much difference there but as was seen in the shorter 24 hour study, the distribution around the median was tighter with automation. Specifically with respect to ranges, hyperoxemia and hypoxemia the following was noted (first number is manual and second comparison automated in each case).
Time spent in target range: 48.4 (41.5–56.4)% vs 61.9 (48.5–72.3)%; p<0.01
Hyperoxemia >95%: 41.9 (30.6–49.4)% vs 19.3 (11.5–24.5)%; p<0.001
< 90%: 8.6 (7.2–11.7)% vs 15.1 (14.0–21.1)%;p<0.0001
< 85%: 2.7 (1.4–4.0)% vs 3.2 (1.8–5.1)%; ns
Hypoxemia < 80%: 1.1 (0.4–1.7)% vs 0.9 (0.5–2.1)%; ns
What does it all mean?
I find it quite interesting that while hyperoxemia is reduced, the incidence of saturations under 90% is increased with automation. I suspect the answer to this lies in the algorithmic control of the FiO2. With manual control the person at the bedside may turn up a patient (and leave them there a little while) who in particular has quite labile saturations which might explain the tendency towards higher oxygen saturations. This would have the effect of shifting the curve upwards and likely explains in part why the oxygen saturation median is slightly higher with manual control. With the algorithm in the CLiO2 there is likely a tendency to respond more gradually to changes in oxygen saturation so as not to overshoot and hyperoxygenate the patient. For a patient with labile oxygen saturations this would have a similar effect on the bottom end of the range such that patients might be expected to drift a little lower then the target of 90% as the automation corrects for the downward trend. This is supported by the fact that when you look at what is causing the increase in percentage of time below 90% it really is the category of 85-89%.
Is this safe? There will no doubt be people reading this that see the last line and immediately have flashbacks to the SUPPORT trial which created a great deal of stress in the scientific community when the patients in the 85-89% arm of the trial experienced higher than expected mortality. It remains unclear what the cause of this increased mortality was and in truth in our own unit we accept 88 – 92% as an acceptable range. I have no doubt there are units that in an attempt to lessen the rate of ROP may allow saturations to drop as low as 85% so I continue to think this strategy of using automation is a viable one.
For now the issue is one of a ventilator that is capable of doing this. If not for the ventilated patient at least for patients on CPAP. In our centre we don’t use the Avea model so that system is out. With the system we use for ventilation there is also no option. We are anxiously awaiting the availability of an automated system for our CPAP device. I hope to be able to share our own experience positively when that comes to the market. From my standpoint there is enough to do at the bedside. Having a reliable system to control the FiO2 and minimize oxidative stress is something that may make a real difference for the babies we care for and is something I am eager to see.
Look around you. Technology is increasingly becoming pervasive in our everyday lives both at home and at work. The promise of technology in the home is to make our lives easier. Automating tasks such as when the lights turn on or what music plays while you eat dinner (all scripted) are offered by several competitors. In the workplace, technology offers hopes of reducing medical error and thereby enhancing safety and accuracy of patient care. The electronic health record while being a nuisance to some does offer protection against incorrect order writing since the algorithms embedded in the software warn you any time you stray. What follows is a bit of a story if you will of an emerging technology that has caught my eye and starts like many tales as a creative idea for one purpose that may actually have benefits in other situations.
In 2012 students in Australia rose to the challenge and designed a digital stethoscope that could be paired with a smartphone. The stethoscope was able to send the audio it was receiving to the smartphone for analysis and provide an interpretation. The goal here was to help diagnose childhood pneumonia with a stethoscope that would be affordable to the masses, even “Dr. Mom” as the following video documents. Imagine before calling your health line in your city having this $20 tool in your hands that had already told you your child had breath sounds compatible with pneumonia. Might help with moving you up the triage queue in your local emergency department.
Shifting the goal to helping with newborns
Of course breath sounds are not the only audio captured in a stethoscope. Heart sounds are captured as well and the speed of the beats could offer another method of confirming the heart is actually beating. Now we have ECG, pulse oximetry, auscultation and palpation of the umbilical stump to utilize as well so why do you need another tool? It comes down to accuracy. When our own heart rates are running high, how confident are we in what we feel at the stump (is that our own pulse we are feeling?). In a review on measurement of HR by Phillipos E et al from Edmonton, Alberta, auscultation was found to take an average of 17 seconds to produce a number and in 1/3 of situations was incorrect. The error in many cases would have led to changes in management during resuscitation. Palpation of the umbilical cord is far worse. In one study “cord pulsations were impalpable at the time of assessment in 5 (19%) infants, and clinical assessment underestimated the ECG HR with a mean (SD) difference between auscultation and palpation and ECG HR of − 14 (21) and − 21 (21) beats min –1″. In another study, 55% of the time providers were incorrect when they thought the HR was under 100 BPM. This leaves the door open for something else. Might that something be the digital stethoscope?
How does the digital stethoscope fare?
Kevac AC et al decided to look at the use of the Stethocloud to measure HR after birth in infants >26 weeks gestational age at birth. The opted to use the ECG leads as the gold standard which arguably is the most accurate method we have for detecting HR. The good news was that the time to signal acquisition was pretty impressive. The median time to first heart rate with the stethoscope was 2 secs (IQR 1-7 seconds). In comparison the time for a pulse oximeter to pick up HR is variable but may be as long as one minute. In low perfusion states it may be even longer or unable to pick up a good signal. The bad news was the accuracy as shown in the Bland Altman plot. The tendency of the stethoscope was to underestimate the EKG HR by about 7 BPM. Two standard deviations though had it underestimate by almost 60 BPM or overestimate by about 50 BPM. For the purposes of resuscitation, this range is far to great. The mean is acceptable but the precision around that mean is to wide. The other issue noted was the frequent missing data from loss of contact with the patient. Could you imagine for example having a baby who has a heart rate of 50 by the stethoscope but by EKG 100? Big difference in approach, especially if you didn’t have EKG leads on to confirm. The authors note that the accuracy is not sufficient and felt that an improvement in the software algorithms might help.
Another go at it
So as suggested, the same group after having a new version with improved software decided to go at it again. This time Gaertner VD et al restricted the study to term infants. Forty four infants went through the same process again with the stethoscope output being compared to EKG lead results. This time around the results are far more impressive. There was virtually no difference between the ECG and the stethoscope with a 95% confidence interval as shown in the graphs with A being for all recordings and B being those without crying (which would interfere with the acquiring of HR). A maximal difference of +/- 18 BPM for all infants is better than what one gets with auscultation or palpation in terms of accuracy and let’s not forget the 2 second acquisition time!
Should you buy one?
I think this story is evolving and it wouldn’t surprise me if we do see something like this in our future. It certainly removes the element of human error from measuring. It is faster to get a signal than even the time it takes to get your leads on. Where I think it may have a role though is for the patient who has truly no pulse. In such a case you can have an EKG HR but the patient could be in pulseless electrical activity. Typically in this case people struggle to feel a pulse with the accuracy being poor in such situations. Using a device that relies on an actual heart contraction to make a sound provides the team with real information. Concurrent with this technology is also the rise of point of care ultrasound which could look at actual cardiac contractions but this requires training that makes it less generalizable. Putting a stethoscope on a chest is something we all learn to do regardless of our training background.
I think they could be on to something here but perhaps a little more evidence and in particular a study in the preterm infant would be helpful to demonstrate similar accuracy.
It has been some time since I wrote on the topic of point of care ultrasound (POC). The first post spoke to the benefits of reducing radiation exposure in the NICU but was truly theoretical and also was really at the start of our experience in the evolving area. Here we are a year later and much has transpired.
We purchased an ultrasound for the NICU in one of our level III units and now have two more on the way; one for our other level III and one for our level II unit. The thrust of these acquisitions have been to reduce radiation exposure for one but also to shorten the time to diagnosis for a number of conditions. No matter how efficient x-ray technologists are, from the time a requisition is placed to the arrival of the tech, placement of the baby and then processing of the film, it is much longer than using a POC at the bedside. Having said that though is it accurate? There are many examples to choose from but when thinking about times when one would like an answer quickly I can’t think of anything much better than a pneumothorax.
Chest X-ray vs POC for Diagnosis of Pneumothorax
The diagnosis of a pneumothorax is easily diagnosed by ultrasound when there is an absence of lung sliding as seen in this video. In the majority of cases employing POC we are looking at ultrasound artifacts. In the case of pleural sliding which is best described as ants marching, it’s absence indicates the presence of a pneumothorax. The “lung point” sign as shown in this video marks the transition from pleural sliding to none and in a mode called “M” appears as a bar code when the pneumothorax is present.
Using such signs Raimondi F et al as part of the LUCI (Lung Ultrasound in the Crashing Infant) group compared traditional x-ray diagnosis as the gold standard to POC for diagnosis of pneumothorax. This study is important as it demonstrated two very important things in the 42 infants who were enrolled in the study. The first was the accuracy of POC. In this study each patient had both an ultrasound and an x-ray and the results compared to determine how accurate the POC was. Additionally in cases where there was no time for an x-ray to confirm the clinical suspicion the accuracy of the study was determined based on the finding of air with decompression along with abrupt clinical improvement. In case people are wondering infants as small as 24 weeks were included in the study with an average weight of 1531 +/-832 g for included infants.
The accuracy was stunning with a sensitivity and specificity of 100% each. Comparing this with clinical evaluation (transillumination, assessment of breath sounds) was far less accurate with a sensitivity of 84% (65-96) and specificity 56% (30-80).
Adding to the accuracy of the test is the efficiency of the procedure. “After clinical decompensation, lung ultrasound scans were completed in a mean time of 5.3 +/- 5.6 minutes vs a mean time of 19 +/- 11.7 minutes required for a chest radiograph (P < .001).” In short, it is very accurate and can be done quickly. In an emergency, can you think of a better test?
If efficiency weren’t enough what about the reduction in radiation exposure?
This was the focus of a recent paper by Escourrou G & Deluca D entitled Lung ultrasound decreased radiation exposure in preterm infants in a neonatal intensive care unit. The authors in this study chose to examine retrospecitively the period from 2012 – 2014 as in 2013 they rolled out a program of teaching POC ultrasound to clinicians. The purpose of this paper was to see if practitioners educated in interpretation of ultrasound would actually change their practice and use less ionizing radiation.
Their main findings are indicated in the table
|Min 1 x-ray during admission
|Mean x-rays per patient
|Mean radiation dose (microGy)
As they predicted use of ionizing radiation dropped dramatically. I should also mention that they tracked outcomes such as IVH, mortality and BPD to name a few and found no change over time. In conclusion the use of ultrasound did not affect major outcomes but did spare each neonate ionizing radiation.
Now before anyone hits the panic button I still think the amounts of radiation here are safe for the most part. In Canada the maximum allowed dose for the public per year is 1 mSv which is the equivalent of 1000 microGy. This was obtained from the Nuclear Safety agency in Canada in case you are interested in finding out more about radiation safety limits.
Back in 2012 at least in this study, 2 standard deviations from the mean would have put the level received at a little over a third of what the annual limit is but it is the outliers we need to think of. What about kids getting near daily x-rays while on high frequency ventilation or for monitoring pleural fluid collections? There certainly are many who could receive much higher dosages and it is for those kids that I believe this technology is so imperative to embrace.
It will take time to adopt and much patience. With any new roll out there is a learning curve. Yes there will be learners who will need to handle patients and yes there will be studies done at times to obtain the skills necessary to perform studies in an efficient and correct manner but I assure you it will be worth it. If we have a way of obtaining faster and accurate diagnoses and avoiding ionizing radiation don’t we owe it to our patients and families to obtain such skill? I look forward to achieving a centre of excellence utilizing such strategies and much like this last study it will be interesting to look back in a year an see how things have changed.
First off I should state that while I generally love Apple products and have owned many, I have no financial interest in the company so this is not a plug with a hidden objective. Rather I was tipped off by a friend who is co-founder of Kindoma (a brilliant piece of software I would add that I also previously wrote about here). She was watching the Apple Keynote address and texted me after she saw something that she knew might pique my interest. No it wasn’t a bigger or faster iPhone, an iPad or even the Watch itself but rather a new capability using an Watch that I believe will revolutionize how we physicians and other health care providers interact with each other.
The short 4 minute presentation was by the founders of a piece of software called Airstrip and thankfully I was able to isolate just that presentation
I like others work in a busy NICU that at any given time has multiple babies receiving myriad blood and radiological tests. In tandem there are babies in need of physical exams, discharges to prepare, lectures to hear and paperwork to sign. After testing is ordered there is no way to predict exactly when results will be ready but when they are, we rely on our memory or the assistance of the bedside nurses to remind us that we need to follow-up. In the case of testing that is slow to come back, such as those that are only batched on certain days or sent to an outside laboratory the potential for missed follow-up is high.
Do Physicians Actually Miss Tests?
This was in fact the subject of a systematic review The safety implications of missed test results for hospitalised patients: a systematic review. This study examined the results of 12 studies, each of which sought to determine how commonly results were missed by staff physicians. One of the studies included found that 28.8% of the time results that were considered urgent were never accessed by the physician. Interestingly 5.1% of the time they did attempt to see them via a login terminal but before the results were ready so in other words were either too early or too late (not at all before the patient left). When looking at only emergency room settings, seven studies quantified the extent of failure to followup in EDs This ranged from 1.0% to 75% of tests and 0% to 16.5% of patients treated in the ED. Test types included: radiology with failure to follow-up ranging from none to 5.6%; microbiology with failed follow-up ranging from 3.0% to 75% and urgent biochemistry with 44.7% not followed up. One can see the comparison between a busy NICU and an ED so to think that so many tests can be missed due to lack of follow-up is frightening. Another concerning finding from the above analysis was that whether the hospital used paper, paper/electronic charting or purely electronic did not affect the rate of missed results.
Another question though is how quickly do physicians respond to a critical result.
This was the question that Kuperman GJ and colleagues tried to answer in their paper from 1998. In the chart review of a 9 day period, 99 test results were identified as being critical (CLR). Among these 99 CLRs, the median time interval until an appropriate treatment was ordered was 2.5 hours. This interval was 1.8 hours when the CLR met the laboratory’s criteria and a phone call was made, and 2.8 hours when the CLR met more complex criteria not requiring a phone call (p = 0.07). Shockingly, for 27 (27%) of the CLRs, an appropriate treatment was ordered only after five or more hours. The use of a phone call system does not seem to truly improve the reaction time even for these critical results.
The following year the same group published their findings using an automated alerting system to notify health care providers of such abnormal results. The laboratory equipment is capable of identifying an abnormal result and then automatically generating a page to the responsible physician. In this study, the main goal was to see if their previous response times could be improved. After exclusions, 192 alerting situations (94 interventions, 98 controls) were analyzed. The intervention group had a 38 percent shorter median time interval (1.0 hours vs. 1.6 hours, P = 0.003; mean, 4.1 vs. 4.6 hours, P = 0.003) until an appropriate treatment was ordered. There was no significant difference between the two groups in the number of adverse events.
You might be surprised by the minimal increase in efficiency with such a paging system but I can’t say I am shocked. Having a system that still requires the provider to call back or check a computer is great if the person is free at the time but what if they are with a patient or being pulled in three different directions in the ED. Will they still remember to answer the page? In many cases I imagine they might forget and then recall afterwards.
How Will Airstrip Resolve These Issues?
It is unclear to me whether the data from an electronic patient record needs to be pushed by a nurse or whether it can be automated but if not now I am certain the future will have this capability. Information can be pushed to the Watch as soon as it is reported so the clinician need only glance down at their watch to see the results. Guessing as to whether the results are ready is eliminated as are wasted minutes each time they sit down at a terminal to check if they are done. Imagine as well ordering an x-ray and a message appearing on the watch to inform you it is processed and ready for viewing. Then there is the ripple effect to consider. At some point perhaps even now the bedside nurse or unit clerk need to waste time finding the doctor to remind them to check. They could focus on the patient which is something they would rather do I imagine anyway!
In terms of privacy issues the technology is able to recognize when the physician is wearing the watch and push the data to them as long as it is on their wrist. Once removed (if taken off in a washroom and forgotten) it is disabled and in addition is HIPPA compliant with hospital compliance requirements.
There is additional functionality with being able to communicate with the nurse or family of a patient via the linked patient, family, nurse and lab results to the physician. Imagine getting the results of a head ultrasound and while walking from viewing the images sending the family a note via secure text letting them know you are coming to the unit if they want to hear the results. Embedded within this technology is also the ability to send orders to the EPR via a touch interface on the iWatch. Get an abnormal set of electrolytes that you believe is dilutional? Simply tap the electrolytes order on the Watch and the EPR notifies the bedside nurse or lab to recollect. I could go on with the many potential improvements to workflow that this technology may bring but thus far as you can tell I am quite impressed.
The chief problems that we face as providers in a busy NICU are failing to follow-up on results, and acting on these results. Airstrip on the iWatch would certainly go a long way to helping with these issues. I imagine a study to prove such efficiency will soon follow and if the results are as I expect I look forward to seeing an Watch on all of our medical staff in the near future. Better care is always our goal and this could be one efficient way to achieve it.
Leonard Nimoy passed away February 27th and shortly afterwards a paper was published that his character Spock would have been proud of. For those of you who know me you will surely appreciate that I try to minimize blood work as much as possible in our infants. That is not to say that other colleagues don’t but in my case it may border on an obsession. I am always on the lookout for technology which helps to achieve this goal that is both accurate and safe to use in our population of neonates.
We are all familiar with Masimo as the providers of the Radical-7 Co-Oximetermonitors that we use in our units. They are highly accurate and provide our histogram analysis which we employ to minimize oxygen use in the units. Some time ago an additional capability was developed which allows the non invasive continuous measurement of hemoglobin (SpHb), along with carboxyhemoglobin and methemoglobin.http://www.masimo.com/hemoglobin/ This technology had been validated in adults and infants > 3000g limiting its widespread applicability in many Neonatal Units. In adults however it had been shown to reduce likelihood of transfusion as a significant benefit.
The publication referred to above is entitled Validation of noninvasive hemoglobin measurement by pulse co-oximeter in newborn infants by Nicholas et al. J of Perinatology; March 5, 2015:1-4. (http://1.usa.gov/18zmEqV) What drew me to this observational study comparing continuous measurement of HgB to serum HgB samples was that the population studied were infants all less than 3000g. Sixty one patients were enrolled and at three different time points between 2-10 days, serum HgB was compared to SpHb. The patients had a birth weight mean of 1177 +/- 610g with 48% of the subjects having a weight < 1000g.
The results were quite interesting to me. The mean difference between HgB and SpHb was -0.9 +/- 16.7 g/L for the whole group and for infants < 32 weeks -2.3 +/- 16 g/L. The correlation coefficient for the whole sample was 0.66 and for the <32 week group 0.69. What does this mean? The correlation coefficients suggest that the relationship between the two variables is not quite linear, as a true linear relationship exists as you get closer to a correlation coefficient of 1. The standard deviations are quite large for these samples so for example a patient with a true serum HgB of 100 could have a measured HgB of either 67 or 137 at the extremes but on average would have a SpHb of 97. This may not seem to be that great a correlation but it may be the trend that is more important than the actual number. This is similar to the concept of using trends in transcutaneous or end tidal CO2 sampling in ventilated patients. What will likely need to be teased out in a larger cohort though is how much agreement there is in an individual patient over time. If a patient is reading 10 g/L below on the first serum HgB measurement what will it read later in the patient’s course as HgB levels change. Also as the authors state it is unclear what effect polycythemia and anemia or low perfusion state might have on accuracy since these infants were studied only over a maximum 8 day period.
So where does this leave us? The technology has now been studied in a population of infants that comprises the largest cohort of infants that have HgB monitored frequently. If the technology is capable of trending HgB for us then I think there may be some use. Would we need to draw a CBC every two weeks as an example if the SpHb was reading > 100 and a first and second CBC showed that for that patient the result was within 10 or 15 g/L of the true value? Could we use such a continuous reading to warn us of a severe IVH or pulmonary hemorrhage as it is happening? Certainly such information if accurate would be useful. A trial of this technology may be warranted!
While this post has focused on one advancement, the fictional Tricorder is not far off from reality but it is currently separated into many different devices. We now have contact lenses and subcutaneous devices that can measure glucose continually (http://bit.ly/1awuqQI), endotracheal tubes that can provide continuous pCO2 measurement and the ability to measure bilirubin non-invasively just to name a few. As the technology becomes more sophisticated with time I have no doubt that we will see a drop worldwide in the number of skin breaks we see in our patients and that for me is a step in the right direction.