Paediatric sleep apnea event prediction using nasal air pressure and machine learning.

Sleep-disordered breathing is an important health issue for children. The objective of this study was to develop a machine learning classifier model for the identification of sleep apnea events taken exclusively from nasal air pressure measurements acquired during overnight polysomnography for paediatric patients. A secondary objective of this study was to differentiate site of obstruction exclusively from hypopnea event data using the model. Computer vision classifiers were developed via transfer learning to either normal breathing while asleep, obstructive hypopnea, obstructive apnea or central apnea. A separate model was trained to identify site of obstruction as either adeno-tonsillar or tongue base. In addition, a survey of board-certified and board-eligible sleep physicians was completed to compare clinician versus model classification performance of sleep events, and indicated very good performance of our model relative to human raters. The nasal air pressure sample database available for modelling comprised 417 normal, 266 obstructive hypopnea, 122 obstructive apnea and 131 central apnea events derived from 28 paediatric patients. The four-way classifier achieved a mean prediction accuracy of 70.0% (95% confidence interval [67.1-72.9]). Clinician raters correctly identified sleep events from nasal air pressure tracings 53.8% of the time, whereas the local model was 77.5% accurate. The site of obstruction classifier achieved a mean prediction accuracy of 75.0% (95% confidence interval [68.7-81.3]). Machine learning applied to nasal air pressure tracings is feasible and may exceed the diagnostic performance of expert clinicians. Nasal air pressure tracings of obstructive hypopneas may "encode" information regarding the site of obstruction, which may only be discernable by machine learning.

The Positive Effect of Adenotonsillectomy on Drooling in Children.

OBJECTIVES: (1) To investigate the prevalence and severity of drooling among healthy young children referred for adenotonsillectomy; (2) to evaluate the effect of adenotonsillectomy on drooling. STUDY DESIGN: Prospective study. SETTING: Tertiary referral center. METHODS: Healthy typically developed children aged 18 to 48 months referred to adenotonsillectomy for upper airway obstruction (UAO) were recruited. Age-matched children recruited from the community served as controls. Drooling frequency and severity were assessed at baseline and 2 months following surgery based on 2 subjective scales: the Drooling Infants and Preschoolers Scale (DRIPS) and Thomas-Stonell and Greenberg Saliva Severity Scale (TSGS). RESULTS: Eighty-seven children aged 18 to 48 months were included in the study. Forty-three children referred to adenotonsillectomy (study group) and 44 age-matched controls. There were significant differences in almost all of the DRIPS items between children in the presurgery group compared to controls. Drooling severity and frequency were greater in the former compared to the latter (TGF-s severity: 1.4 ± 1.0 vs 0.6 ± 0.8, P = .001; TGF frequency: 1.3 ± 0.9 vs 0.5 ± 0.6, P < .001). After surgery, the scores for all DRIPS and TSGS items decreased significantly and were comparable to the control group. CONCLUSIONS: The frequency and severity of drooling among otherwise young children referred for adenotonsillectomy were greater than those for healthy controls. Following surgery, both the frequency and severity significantly improved and became comparable to those of controls. These findings suggest that a major improvement in drooling is one of the benefits of a surgical intervention in a child with UAO.

Harnessing the power of electronic health records and open natural language data mining to capture meaningful patient experience during routine clinical care.

INTRODUCTION: Electronic health records (EHR) are a rich data source for both quality improvement and clinical research. Natural language processing can be harnessed to extract data from these previously difficult to access sources. OBJECTIVE: The objective of this study was to create and apply a natural language search query to extract EHR data to ask and answer quality improvement questions at a pediatric aerodigestive center. METHODS: We developed a combined natural language search query to extract clinically meaningful data along with International Statistical Classification of Diseases (ICD10) and Current Procedural Terminology (CPT) code data. This search query was applied to a single pediatric aerodigestive center to answer key clinical questions asked by families. Data were extracted from EHR data from first clinic visit, operative note, microbiology lab report, and pathology report for all new patients from 2020 to 2021. Included as three queries were: 1) if I bring my child to a pediatric aerodigestive center, how often will my child obtain a medical diagnosis without needing an intervention? 2) if my child has a diagnostic procedure, how often will a diagnosis be made? 3) if a diagnosis is made, can it be addressed during that endoscopic intervention? RESULTS: For the 711 new patients coming to the pediatric aerodigestive center from 2020 to 2021, only 26-32% required an interventional triple endoscopy (rigid/flexible bronchoscopy with esophagoduodenoscopy). Of these triple endoscopies, 75.7% resulted in a positive finding that enabled optimization of that child's care. Of the 221 patients who underwent diagnostic triple endoscopies, 40.7% underwent intervention at the same time for laryngeal cleft (injection or suture, dependent upon age). CONCLUSION: Here we created an effective model of open language search query to extract meaningful metrics of patient experience from EHR data. This model easily allows the EHR to be harnessed to create retrospective and prospective databases that can be readily queried to answer clinical questions important to patients. Such databases are widely applicable not just to pediatric aerodigestive centers but to any clinical care setting using an EHR.