Premium
Individual growth curve models for assessing evidence‐based referral criteria in growth monitoring
Author(s) -
van Dommelen P.,
van Buuren S.,
Zandwijken G. R. J.,
Verkerk P. H.
Publication year - 2005
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.2234
Subject(s) - growth curve (statistics) , referral , gestational age , medicine , pediatrics , growth model , sensitivity (control systems) , area under the curve , longitudinal study , bayley scales of infant development , statistics , mathematics , pregnancy , biology , psychiatry , mathematical economics , family medicine , pathology , electronic engineering , engineering , genetics , cognition , psychomotor learning
Abstract The goal of this study is to assess whether a growth curve model approach will lead to a more precise detection of Turner sydnrome (TS) than conventional referral criteria for growth monitoring. The Jenss–Bayley growth curve model was used to describe the process of growth over time. A new screening rule is defined on the parameters of this growth curve model, parental height and gestational age. The rule is applied to longitudinal growth data of a group of children with TS ( n =777) and a reference ( n =487) group. The outcome measures are sensitivity, specificity and median referral age. Growth curve parameters for TS children were different from reference children and can therefore be used for screening. The Jenss–Bayley growth model, which uses all longitudinal measurements from birth to a maximum age of 5 years with at least one measurement after the age of 2, together with parental height and gestational age can achieve a sensitivity of 85.2 per cent with a specificity of 99.5 per cent and a median referral age of 4.2 (the last measurement between the age of 2 and 5 of each child is considered to be the moment of referral). Sensitivity increases by 2 percentage points when decreasing the specificity to 99 per cent. The Jenss–Bayley growth model from birth to a maximum age of 8 years with at least one measurement after the age of 2, together with parental height results in a sensitivity of 89.0 per cent with a specificity of 99.5 per cent and a median referral age of 6.1. For a specificity of 98 per cent, we obtain a sensitivity of 92.3 per cent. In comparison to conventional rules applied to the same data, sensitivity is about 11–30 percentage points higher at the same level of specificity for the Jenss–Bayley growth rule. We conclude that from the age of 4, growth curve models can improve the screening on TS to conventional screening rules. Copyright © 2005 John Wiley & Sons, Ltd.