Recent developments in growth assessment: the LMS method, GAMLSS and beyond

 
Share Download as iCal file
Click the icons above to Share, Tweet or add this event to your calendar (iCal)
 
Medical Section

Thursday 11 February 2016, 02:00pm - 04:30pm

 
Location Royal Statistical Society, 12 Errol Street, London, EC1Y 8LX

To celebrate Professor Tim Cole being awarded the RSS Bradford Hill medal in 2015, there will be an afternoon of talks on the LMS method, and its developments in recent years, followed by a group discussion.

Confirmed speakers (and talks) are:

Prof. Tim Cole (Institute of Child Health, University College London): LMS methods - origins and applications 

Prof. Mikis Stasinopoulos (Director of STORM, London Metropolitan University): To be confirmed

Philip Gichuru (Mathematics & Statistics department, Lancaster University): Developing Robust Scoring Methods for use in Child Assessment Tools 

Prof. Stef van Buuren (Department of Methodology & Statistics, University of Utrecht): Curve matching for personalized prediction of growth and development 

Abstracts of talks:

Prof. Tim Cole: The LMS method – origins and applications - slides (ppt)
The LMS method is a statistical technique for constructing age-related reference ranges. It estimates the distribution of measurements in terms of the age-varying median, variability and skewness, each summarised as a natural cubic spline curve, assuming an underlying normal distribution with a Box-Cox transformation. First described by Cole (JRSS A 1988) and refined by Cole and Green (Stat Med 1992), it has been widely used to construct growth reference centile charts. The talk will summarise the history of the LMS method and illustrate its use with examples of weight in infancy and height in puberty.

Prof. Mikis Stasinopoulos: LMS and GAMLSS - Flexible Regression and Smoothing slides (pdf)

Phillip Gichuru: Developing Robust Scoring Methods for use in Child Assessment Tools - slides (pdf)
Developmental assessment in early childhood is important to ensure that a child is progressing well and that diagnosis of any disability is detected and acted upon early. This is of particular relevance in low and middle income countries, where a number of child assessment tools have been utilised. Unlike anthropometric measures of growth, most developmental assessments return categorical (binary) data (e.g. pass/fail) for each item. We sought to develop robust scoring methods for child assessment tools which would ensure more accurate and therefore timelier intervention of detected delayed development. We also needed to create a framework to suitably correct or account for age. Generally, two main scoring approaches have been used; item by item scoring which creates developmental ‘norms’ for each item and total scoring that utilizes all the responses of the child to give one total score across the entire domain.

Using data from 1,446 normal children from the recent Malawi Development Assessment Tool (MDAT) study (Gladstone et al, 2008, 2010), we reviewed and extended classical total scoring methods. We highlight weaknesses of the classical scoring approaches under the Generalized Linear Model framework and considered extensions of Cole’s (1990) Lambda-mu-sigma (LMS) method. We explored the use of Generalized Additive Models for Location Scale and Shape (GAMLSS) as described by Rigby et al. (2007) and Shape Constrained Additive Models (SCAM) as described by Pya, N. and Wood, S. N. (2015). We also reviewed the current methods of creating Z-scores and again used extensions of Cole’s (1990) work in line with our research objectives both for smoothing and age correction purposes. The results show that smoothing of score values is especially beneficial when the sample data has sporadic coverage over some age groups. The extensions of Cole’s (1990) Lambda-mu-sigma (LMS) method produce more reliable and more generalizable normative scores where classical approaches are not suitable. The sensitivity analysis shows that the classical methods perform well only in ideal situations and the suggested extensions can be used when the former methods fail.

Prof. Stef van Buuren: Curve matching for personalized prediction of growth and development - slides (pdf)
The key idea is to find a small number of children in the existing data who are ‘similar’ to the child for which we want prediction. The realized growth patterns of the matched children suggest how the target child might evolve in future.

Curve matching can predict the future growth and development of an individual child for datasets of sufficient size, say 100,000+ children. An appealing feature of curve matching is that each matched growth trajectory represents real growth of real children. The spread between the matched curves provides a natural indication of the uncertainty of the prediction.

In this lecture, I will outline the principles of technique, discuss potential extensions, and demonstrate how curve matching can be used in practice.

Contact Please register by email to This email address is being protected from spambots. You need JavaScript enabled to view it.

Organiser Name Annie Herbert

Email Address This email address is being protected from spambots. You need JavaScript enabled to view it.

Organising Group(s) RSS Medical Section and RSS Primary Health Care Special Interest Group

 

 

 

 

Join the RSS

Join the RSS

Become part of an organisation which works to advance statistics and support statisticians

Copyright 2019 Royal Statistical Society. All Rights Reserved.
12 Errol Street, London, EC1Y 8LX. UK registered charity in England and Wales. No.306096

Twitter Facebook YouTube RSS feed RSS feed RSS newsletter

We use cookies to understand how you use our site and to improve your experience. By continuing to use our site, you accept our use of cookies and Terms of Use.