Go to any NICU these days and you will no doubt hear about the toxicity of oxygen. Oxygen as we say is a drug and like any such product has both good and bad effects. On the good side is the ability to increase the fraction of inspired oxygen to deal with transient changes in observed saturations. In extreme cases where the patient is desaturating into the 20’s and 30’s with accompanying bradycardia this can be life saving. On the other hand the “bad” is related to toxicity from oxygen free radicals which can increase rates of ROP, BPD and in the case of resuscitation with O2 vs room air may increase the chances of death.
It is for this reason that NICU teams pay particular attention to saturation targeting. The optimal goal remains elusive as oxygen restriction to 85-89% has been linked to higher mortality as in the Support Trial (full article) while higher saturations may increase the rate of ROP and BPD. Many units are falling somewhere in the middle such as goals of 90-95% or as in our units 88-93%. If during your visit to an NICU you were to observe the nurses at the bedside you would see them or the Respiratory Therapists tweaking the oxygen up and down many times a day as they strive to keep the saturations within these ranges. As a Neonatologist I greatly appreciate the efforts of everyone on the team who try to keep maintain these goals but in the end how do they really do?
This was the subject of a 2006 paper by Hagadorn et al. Eighty four infants from 14 centres all less than 28 weeks were enrolled in this study looking at how successful staff were at keeping infants within a desired range. The findings were somewhat discouraging in that 16-64% of the time saturations were in range, while 20-73% of the time they were above range depending on the centre studied. In a different study by Laptook et al published in the same year, the results were not much better with saturations higher than goal about 15% of the time and lower by nearly 26% of the time. What impact might swings in oxygen saturation have on the brain when the saturations are low and similarly on rates of BPD and ROP when the tendency is to overshoot the goals? There is no doubt that everyone is trying to do a good job but how deflating is it to nursing and other staff members when despite their best efforts they are only in range about 60% of the time?
Fortunately I believe change is coming. With improvements in technology it is now possible to have a closed loop system in which a patient who is receiving oxygen has their saturation measured and the information via a feedback loop triggers an automated response. This response either raises or lowers the FiO2 in an automatic way which eliminates the need for health care staff to make such changes. Such technology is actually not that new as it was tested in 2001 by Claure et al and was found to be at least as effective as manual changes by a dedicated nurse. Several other small studies followed, all demonstrating an improved accuracy in maintaining target saturations. This July the same group published the following article Automated versus Manual Oxygen Control with Different Saturation Targets and Modes of Respiratory Support in Preterm Infants. This study compared the accuracy in maintaining saturations using a target of 91-95% and another of 89-93% with 80 patients participating. Patients in this study received both invasive and non-invasive ventilation. When comparing the two target ranges the automatic adjustments showed greater benefit with the results being 62 +/- 17% vs 54 +/- 16% for the 91-95% range and 62 +/- 17% vs 58 +/- 15% in the lower range. Both of these results were statistically different with p values of <0.001. When looking at episodes of significant hypoxemia as defined by a saturation < 80% the findings were interesting showing in the high range 13 vs 4 and in the low range 15 vs 4 favouring the use of autosaturations. The authors were surprised at the lack of large disparity in the accuracy between manual and auto adjustment of FiO2 but this may be related to the Hawthorne effect. Since the nurses in this study were not blinded to the intervention they may have changed their behaviour in essence to try and prove that they were just as good. How this would translate into a real world situation when a study is not being performed I suspect would favour the automated system more.
Perhaps the most interesting part of the study though was the number of Manual FiO2 adjustments per 24 hours between the two groups. For the 89%-93% SpO2 target range this was 1 [0-3] vs 102 [73-173], P < .001 and for the 91%-95% SpO2 target range 1 [0-3] vs 109 [79-156], P < .001. For me this is the crux of the discussion. In a twenty four hour period there is a reduction of about 100 interventions using the saturation range similar to our own. Take note as well that some patients required over 150 interventions in 24 hours as 100 was the mean! While the targeting is improved somewhat with the use of the auto FiO2 adjustment this is where the biggest benefit to the patient lies as I see it. Nurse and respiratory therapists are very busy on any given shift. One hundred adjustments a day translates into 4 changes in FiO2 per hour on average. Although not measured I can’t help but wonder what impact these interruptions have on the rest of the care for a given patient. What is the “ripple effect” of such interruptions in a person’s train of thought. Could this translate into a med error or delay care for another patient to whom the nurse is responsible for as well? If one didn’t have to pay as much attention to the FiO2, what other goals could their attention be put towards? Might this provide them with more time to educate parents, do skin care, or a host of other responsibilities which in the current state get interrupted every time an alarm goes off.
As this technology is improved I see this being a welcome addition to the NICU. As with anything though that promises a hands free environment it will be essential to have an override built in. Furthermore we can never take our eyes off the patient as no matter how safe these systems may be there is always the chance that a computer will freeze as we all know. The impact of such a “glitch” in the algorithm that these systems use could be catastrophic to outcome so as good as these may be we will always need the human presence to ensure that what we think we are getting from the system is actually there.