P0034 - Utilizing Big Data to Identify the Prevalence of and Risk Factors for Obstructive Sleep Apnea in a Pediatric Population

Background: Obstructive sleep apnea (OSA) is a somewhat common chronic condition among the pediatric population. While current literature estimates the prevalence of OSA in a pediatric population between 1.2–5.7%, these studies utilize small sample sizes. This study aims to utilize big data tools and techniques through aggregate electronic health records to characterize the prevalence of OSA with attention to differences in racial, ethnic, gender, and age groups, as well as by status of hypertension and obesity.

Methods: An analysis of de-identified aggregate electronic health record data from the IBM Watson Health Explorys platform. Subgroups included: age cohorts (age =95, overweight 85-94, healthy weight 5-84). Diagnoses were defined as any encounter diagnoses with ICD codes of Obstructive Sleep Apnea Syndrome and Essential hypertension. Difference between groups was calculated using two-proportion z-test or chi-square test for independence.

Results: Among 10,565,170 children under 18, OSA prevalence was 0.65%. Among all race groups, rates of OSA were significantly different (p<0.0001) (Table 1). Among all age groups, males had significantly higher OSA prevalence compared to females: ages 0-4: 0.38% to 0.28% (p<0.0002), ages 5-9: 0.77% to 0.60% (p<0.0002), ages 10-14: 0.84% to 0.72% (p<0.0002), ages 15-19: 0.62% to 0.52% (p<0.0002). OSA prevalence increased significantly with increasing BMI percentile: healthy weight 1.4% (n=2,315,730), overweight 1.9% (n=1,110,420) (p<0.0002), obese 2.1% (n=1,216,910) (p<0.0002). In every sub-cohort, hypertensive patients experienced significantly higher rates of OSA than their non-hypertensive counterparts (Table 2).

Conclusion: This study utilized aggregate EHR data to characterize the prevalence of OSA using a large population sample. The results indicate that pediatric providers should have a higher index of suspicion for assessing OSA among groups with certain characteristics, particularly children with hypertension, obesity, and those of African American race and male gender. Further, this study underscores the utility of big data tools in better characterizing the prevalence and risk factors for diseases which are more challenging to identify due to their low population rates, like many chronic diseases in the pediatric population.