SPA Annual Meeting Reviews

Session III: PEDx Talks

By Kamie Yang, MD
Clinical Lecturer
University of Michigan

Augmented Intelligence in Your Pediatric Operating Room

Jorge A. Galvez, MD, MBI

Jorge A. Galvez, MD, MBI (Children’s Hospital of Philadelphia, Executive Director of Pennsylvania Pediatric Medical Device Consortium) began his talk on machine learning by pointing out that we utilize machine learning-based technologies in our everyday lives when we use the map application on our phones to integrate traffic and location data to plan our routes and estimate time of arrival. But, how can machine learning be used in medicine? In healthcare, machine learning can be used to help us identify patients at risk for particular conditions so that timely and effective treatments can be chosen, as well as in the early detection of particular adverse events. 

Large amounts of data are now available, from medications to ventilation data to genomics. This data can be analyzed through an algorithm to potentially obtain answers to questions. These algorithms are initially formulated based on initial “training data” that is validated and deployed for use in the real world. An essential key to machine learning is that a specific, labeled outcome must be identified as the ultimate goal endpoint of the algorithm. 

There are many algorithm types utilized in machine learning.  A few mentioned by Dr. Galvez include:

  1. Decision Trees
  2. Support Vector Machines (SVM) – Where data is divided into two populations and an algorithm is utilized to predict whether a new example falls into one category or the other.
  3. Neural Networks – Model based on a collection of connected “nodes” which loosely model neurons in a brain.  Every possible combination of interactions of variables in the dataset can then be made to examine their relationship to a labeled outcome.

Machine learning algorithms have some pitfalls that should be considered.  “Overfitting” can occur, where an algorithm is fit to match a particular data set so much so that it does not perform well when applied to a new data set.  Dr. Galvez also feels that machine learning, like many other emerging technologies, will also follow the waxing and waning expectations illustrated in the Gartner Hype Cycle and looks forward to when we ultimately find ourselves on the “plateau of productivity”.

When asked if machine learning will create machines that will replace anesthesiologists, Dr. Galvez stated that he feels that we are a long way away from that. He cited the Sedasys machine from Johnson and Johnson, which was programmed to titrate propofol for endoscopic procedures based on vital signs. After investing 400 million dollars to develop this technology, the device was labeled a failure and Johnson and Johnson abandoned the program in 2016 due to poor sales. 

On the other hand, machine learning has been successfully utilized in other clinical settings. In a study of 93 patients with single-ventricle physiology, Ruiz VM, et al (2019) created an algorithm to predict critical events (ex. need for cardiopulmonary resuscitation, emergency endotracheal intubation, and extracorporeal membrane oxygenation) based on “routinely collected” physiologic data. The algorithm was able to predict these events with improving accuracy the closer it was applied to the time of the critical event, with the area under the Receiver Operating Characteristic (ROC) curve reaching 0.88.  Ultimately, Dr. Galvez noted that being able to tell an intensivist that a particular patient “may run into trouble in the next eight hours” based on the patient’s routine clinical data is a very useful message.    

However, applying machine learning in clinical settings is not as simple as building an algorithm and putting it into use at your hospital. The FDA views these algorithms as “Software as a Medical Device” (SaMD). According to the FDA’s proposed regulatory framework (published April 2019), pre-market approval will be necessary for internal use of machine learning algorithms, even if they are not intended for sale to outside parties.  Also, once deployed, models must be logged, tracked, and evaluated so that model re-training can occur (Figure 1).

Fig 1

Figure 1: Overlay of FDA’s Total Product Lifecycle (TPLC) approach to AI/ML workflow
From https://www.fda.gov/media/122535/download
(SaMD = Software as a Medical Device, AI: Artificial Intelligence, ML: Machine Learning)

Dr. Galvez closed his talk by stating that although machine learning is still in its infancy, endless amounts of data are now available to us through electronic medical records.  Soon we may leverage vital signs or other clinical surrogates to map our clinical care “routes” much as we use traffic data and road maps to choose our routes of travel everyday with our phones.  What new algorithms will be created with machine learning-based technologies and what will clinical medicine look like 20 years from now?

Fetal Myelomeningocele Repair: Post MOMS Era

Jina Sinskey, MD (University of California, San Francisco) started her talk by introducing 'design thinking', a method that can be used to solve complex problems with a multidisciplinary approach. There are five basic steps to design thinking:

  1. Empathy – The most important step of the process, where we focus on the patients and providers we are attempting to help.
  2. Define – Defining a problem with a statement (i.e.  “How might we _____”).   The statement must be broad enough to encourage creativity while still being narrow enough to generate a list of actionable goals.
  3. Ideate – Brainstorming potential solutions.
  4. Prototype – Creating solutions.
  5. Test solutions.

Dr. Sinskey then continued by discussing the groundbreaking Management of Myelomeningocele Study (MOMS) trial published in 2011 in the NEJM. In this randomized controlled trial, the outcomes of prenatal repair of myelomeningocele were compared to postnatal repair. The trial was stopped early due to the overwhelmingly improved outcomes seen in the prenatal group, where prenatally repaired patients were:

  1. Half as likely to need a VPS in the first 12 months of life (40% vs. 82% P<0.001).
  2. Twice as likely to be able to walk without orthotics or devices (42% vs. 21%, P=0.01)

However, since the MOMS trial, many other improvements have occurred in the field of fetal myelomeningocele repair. Dr. Sinskey continued to discuss how design thinking was applied to make improvements in three aspects of this procedure: (1) surgical approach, (2) anesthetic approach, and (3) fetal monitoring technique. 

Surgical Approach.  Due to the large hysterotomy needed to perform fetal surgery during the MOMS trial, prenatal repair of myelomeningocele not only increased the maternal risks of preterm labor, but also put the mother at increased risk of uterine dehiscence and precluded all figure vaginal births. In efforts to decrease maternal risk, alterations were made to the surgical approach.  Instead of performing an open laparotomy, open hysterotomy, dissecting the dura mater, and performing a multilayer closure as done in the original MOMS trial, various combinations of laparoscopic maternal incisions and fetoscopic hysterotomies were later studied. In the percutaneous fetoscopic approach, the surgical procedure was simplified so that a patch was used to close the defect with a simplified single-layer skin closure. Though PROM and premature delivery still occurs with this technique, short and long term maternal morbidity were much improved. 

Anesthetic Approach.  The goals of anesthesia during fetal repair of myelomeningocele are to maintain fetal cardiac function, maintain uteroplacental perfusion, and maintain uterine relaxation. To achieve uterine relaxation, the MOMS trial utilized 2-3 MAC of inhaled anesthetic, resulting in significant maternal hypotension. More recently, TIVA (remifentanil and propfol ggts), have been added by some institutions to decrease the MAC of inhalation agent required. At UCSF, Dr. Sinskey routinely utilizes 1 MAC of inhalation agent and supplements this with a remifentanil infusion, resulting in less maternal hypotension and less need for vasopressor infusions.  

Fetal Monitoring Technique.  In the MOMS trial, continuous fetal ECHO was performed to monitor fetal well-being. However, due to the ongoing surgery, continuous monitoring was not always possible. Since the MOMS era, umbilical artery (UA) doppler monitoring has developed as another means to monitor the fetus.  Normally, forward flow is seen throughout the fetal cardiac cycle in the UA as blood is delivered from the fetus to the placenta. The absence of end diastolic flow and reversal of end diastolic flow (EDF) indicates increasingly poor fetal well-being. After UA dopplers started to be monitored, Dr. Sinskey’s group noticed that a majority of the episodes of absent or reversal of EDF occurred with uterine incisions, concurrent with the loss of amniotic fluid. In response to this, they began instilling warm normal saline to counteract the fluid loss and saw half of the worrisome UA signals improve. 

Dr. Sinskey closed her interesting talk by reiterating that design thinking has allowed us to improve our outcomes in the fetal repair of myelomeningoceles. She anticipates the future will yield other problems that can also be successfully solved with this multidisciplinary technique. 

Pediatric Sedation for Dental Procedures: Why is it Different?

Rita Agarwal, MD, FAAP

Caleb Sears, after whom Caleb’s Law is named, was a healthy six year-old boy with an extra tooth that an oral surgeon recommended be removed.  During the removal procedure, Caleb was sedated with midazolam, nitrous, fentanyl, propofol, and ketamine and eventually stopped breathing. No reversal drugs, CPR, oxygen, or BVM was given. The oral surgeon attempted to intubate, but was unsuccessful. Reportedly, no CPR was occurring when EMS arrived to the oral surgeon’s office and Caleb suffered irreversible brain damage and died in 2015.  Rita Agarwal MD, FAAP (Stanford and Lucille Packard Children's Hospital) recounted this and many other equally alarming stories throughout her talk. 

The single operator model: The crux of the problem lies in the fact that dentistry utilizes sedation practices and standards that are well below those utilized in other fields. Specifically, the “single operator model” is common practice in dentistry and oral surgery. In this model, one “Anesthesia Permit Holder” directs and performs the anesthesia while also performing the surgical procedure. Two assistants are present during the procedure, one to assist with the procedure and one to “watch the monitors”.

Furthermore, the national training standards for dental sedation assistants are alarmingly low. The American Association of Oral and Maxillofacial Surgeons (AAOMS) recommends that dental assistants monitoring sedation be DAANCE certified (Dental Anesthesia Assistant National Certification Examination).  However, this certification requires only six months of practice, 36 hours of online education, and a national exam. Even worse, a non-DAANCE certified dental sedation assistant could have no more than a high school education.  The California Dental Board has more strict requirements than most other states, where dental sedation assistants in California are required to have 12 months of practice and 110 hours of on-site and online education.  However, when taken into perspective, these increased requirements are still severely lacking when you consider that a dental hygienist – the person who cleans your teeth – has 2-4 years of training and an associate or bachelor degree.

Even more, the regulation of dental sedation falls onto individual state dental boards, yielding significant variability between states. The definition of an “Anesthesia Permit Holder” is determined at the state level and usually includes OMFS, Anesthesiologists, CRNAs, Certified Anesthesia Assistants, Dentist Anesthesiologists, to even a Dentist with an “Anesthesia Permit”.  The last group includes dentists that have done some weekend anesthesia courses.

Returning back to Caleb’s story, Dr. Agarwal recounted that with the help of the American Academy of Pediatrics and the California Society of Anesthesiology, the first part of Caleb’s Law was passed in 2016, requiring the Dental Board of California to establish a committee to study the safety of pediatric anesthesia in dental offices and to collect epidemiologic data of adverse events.  It also required consent for general anesthesia, requiring providers to discuss with parents the differing practice models and safety precautions being utilized. 

The Current State of Affairs: However, though adverse events or deaths must be reported to the state, most states do not track or respond to these events. To get a sense of the magnitude of the problem, Dr. Agarwal highlighted that the OMS National Insurance Company (OMSNIC) (the company that insures a majority of OMFS providers) found a death or serious neurological injury rate of 1:350,000 among adults and children. This compares poorly to the death and serious adverse event rate of:

0: >2,000,000 among healthy pediatric surgical patients in the Wake Up Safe database
0: 500,000  among patients in the Pediatric Sedation Research Consortium (PSRC)

Dr. Agarwal pointed out that adverse events can occur during all anesthetics, but the key is to appropriately treat them to prevent worsening of the patient’s clinical status.

More recently, the second part of Caleb’s Law required that a separate qualified anesthesia provider be available for children under the age of seven years receiving sedation. Additionally, it required pediatric outcome data collection, an update in the dental sedation terminology to match those utilized by the ASA, and restructuring of the California dental sedation and anesthesia permit system.   Unfortunately, in 2017, following strong opposition by the dental lobby, the second part of Caleb’s Law was not passed. The dental lobby stated that they felt the law would inhibit access to care and increase costs. 

New Guidelines for Dental Sedation: In light of the fight for Caleb’s Law, in June 2019, the American Academy of Pediatrics (AAP) and the American Academy of Pediatric Dentistry (AAPD) published updated guidelines for sedation for dental procedures in the journal Pediatrics. The report, “Guidelines for Monitoring and Management of Pediatric Patients Before, During, and After Sedation for Diagnostic and Therapeutic Procedures (Cote)” highlighted dental sedation and recommended that for deep sedation and general anesthesia, a qualified anesthesia provider must serve as an independent observer and remain separate from performing (or assisting with) the dental procedure. Furthermore, both the independent observer and the operating dentist must be PALS certified.

The American Association of Oral and Maxillofacial Surgeons (AAOMS) was not happy with the published guidelines and wrote a letter in the journal Pediatrics expressing their displeasure with not being included in the formulation of these guidelines. They also felt that the new guidelines were “overly restrictive and based on hyperbole, opinion, and fueled by emotion” and urged withdrawal of AAP support.  They felt that the guidelines would negatively affect patients by reducing access to care and increasing costs and that the “single operator model” was safe and effective in a “vast majority of cases”. Dr. Agarwal responded to this by stating that providing anesthesia accounts for the bulk of the income of many oral surgery procedures and that oral surgery without a charge for sedation often loses money. 

However, the single operator model is not the only problem and several children have died recently even though a physician anesthesiologist was caring for them.  The new guidelines would not have helped them. Also, dental caries in children is now an epidemic in this country. What can we do in this environment?

Advocacy.  First, it is important to educate pediatricians, parents, and dentists about pediatric dental sedation. Additional information on dental sedation can be found on the healthychildren.org website, the parent portal of the American Academy of Pediatrics. Publicizing this problem on social medial will also help get the word out. Dr. Agarwal also highlighted that we need to establish good data collection methods so we can better understand the problem and recommended that since dental sedation is regulated at the state level, getting involved in our respective state anesthesia or pediatric societies is of the upmost importance.  Children’s lives depend on it.

Virtual Reality and the Art of Distraction

James J. Thomas, MD

Dr. James J. Thomas (Children’s Hospital of Colorado) started his talk by mentioning Cody, a patient with SMA who required general anesthesia for routine wound care. He was one of the first patients Dr. Thomas tried to help with Virtual Reality (VR) technology. 

A systemic review and meta analysis from 2019 looked at the effectiveness of VR on reducing pain and anxiety in pediatric patients undergoing medical procedures.  This study looked at procedures like dental care, burn wound care, port access and lumbar puncture and found that VR was better than routine care at decreasing pain and anxiety.

Dr. Thomas then showed a video of a patient with ALL who required lumbar punctures once a month. In the video, he described how VR was able to distract her from her pain during the procedure and allowed her to totally avoid getting frequent anesthetics.  

Immersion and presence are two terms frequently utilized when discussing VR.  Immersion is an objective characteristic of a VR environment that exists as a spectrum and defines how “real” the system feels. Presence, on the other hand, is a subjective, psychological experience of the user describing how “present” they feel inside the virtual system. It is important because the more “present” a patient feels in a particular VR environment, the more likely they will be distracted and less anxious about their situation. The distraction helps to mitigate their pain experience.

VR does not work on the traditional system of pain perception (from a delta and c fibers connecting to the brain), but instead works by grabbing our attention to lessen the brain’s focus on painful stimuli. Others believe that VR stimulates the visual cortex with input, thus distracting patients from the experience of noxious stimuli.

Furthermore, a 2018 study comparing standard of care (watching TV) during blood draw vs. a VR distraction group found that patients with higher “anxiety sensitivity” levels (i.e. were more likely to be anxious during the procedure) showed the most benefit when they received VR distraction. This suggests that anxious patients with higher levels of baseline anxiety are more likely to benefit from VR techniques.   

VR distraction can even be used to decrease the amount of anesthesia a patient requires. Dr. Thomas discussed an obese patient that required regular lumbar punctures that was helped by VR. This patient typically required 200mcg fentanyl, 4mg versed, and a 40 min procedure time. By combining VR with sedation medications, they were able to cut his procedure time and medication requirements by half.  They also found that the patient was more calm and able to participate in the procedure.

Dr. Thomas went on to discuss another study that looked at using VR during the induction of anesthesia.  Three groups were studied: parental presence, midazolam premed, and VR distraction. The parental presence and midazolam premed groups both showed increased anxiety during the induction of anesthesia while no change in anxiety level was seen in the VR group. 

Virtual reality has also been used to simulate parental presence.  Dr. Thomas discussed how he has utilized a “social” VR environment where both a mom and patient could interact with each other as avatars.  He successfully utilized such a system to calmly induce anesthesia while allowing a patient to talk to his mom in virtual reality.

Potential barriers to implementing a VR program in a hospital setting include:

  1. Infection control: The headsets are not easy to keep clean.  Replaceable covers and straps that can be wiped down can be purchased (VRcover.com).
  2. Network connectivity issues: Especially when traveling within hospitals.
  3. Patient appropriate content: Starlight Children’s Foundation has vetted content.
  4. Orientation issues: Lateral positioning can be difficult for VR programs and require reorientations of the view.

However, unlike some may think, seizures, headache, and nausea/vomiting do not seem to be frequent problems with VR systems.  When asked about age appropriateness, Dr. Thomas stated that even a 2-3 year old could probably figure a VR system out. Furthermore, eye tracking and voice recognition technologies are also starting to be incorporated, so that patients like Cody (who could not move a muscle due to his SMA) may also one day be helped by this new technology.

Back to top