Computational Methods in Growth Analysis

Validity of the WHO Cutoffs for Biologically Implausible values of Weight, Height, and BMI in Children and Adolescents in NHANES from 1999 through 2012

This analysis of data on 26,480 children and adolescents taken as part of the NHANES study from 1999–2000 through 2011–2012 showed that the overall prevalence of children with biologically implausible body measurements (determined commonly accepted rules involving modified z scores) was 0.9%. Most of these were high values rather than low. Further analysis that correlated these BIVs with other body measurements suggested that the majority of these seemingly anomalous values were accurate. Using these methods to exclude BIVs tends to underestimate the prevalence of obesity in these data.

Am J Clin Nutr. 2015 Nov; 102(5): 1000–1006.
Freedman 2015 (Link) | PubMed 26377160 | Author Search

Modified Z-scores in the CDC Growth Charts

Computational methods for detecting biologically implausible values in growth data from the Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention. Describes the calculation of z-scores and ‘modified z-scores’ in the CDC growth chart data published in 2000. A modified z value is defined, for values above the mean, as half of the difference between the value corresponding to a z-score of 2 and the mean. For values below the mean, the modified z-score is half of the difference between the value corresponding to a z-score of 2 and the mean. One expresses the modifies z-score in terms of the modified z-value. For example, for a 4-year old (48.5 months old) boy, the mean BMI is 15.62. The BMI value corresponding to a z-score of -2 is 13.74. So the modified z-value is (15.62 – 13.74)/2 = 0.94. A boy that age with a BMI of 12 would have a modified BMI z-score of (12 – 15.62)/0.94 = -3.85.

CDC 2000 (Link)

Automated Identification of Implausible Values in Growth Data from Pediatric Electronic Health Records

Reports the development of an automated method for identifying implausible values in pediatric EHR growth (weight and height) data, tested via data points collected in the primary care environment on over 280,000 patiets. The method compares each measurement’s z-score to a weighted moving average of prior measurements. The method had a sensitivity of 97% and a specificity of 90% for identifying implausible values compared to physician judgment, and identified almost all simulated errors.

Daymont 2017 (JAMIA) | PubMed 28453637 | Author Search