- Australian Council for Educational Research

Report
Translational science
fostering integration
The predictive validity of the
AEDI: Predicting later cognitive
and behavioural outcomes.
Assoc Prof Sally Brinkman
ACER Conference August 2014
Presentation Structure
•
•
•
•
•
Background.
Predictive validity – 2 studies.
Inequality in child development
and predictive strength.
Conclusions.
International interlude (if time).
What is the AEDI?
•
•
•
Teacher checklist,
5 Domains
Triennial Census
Sensitive Periods in Early Brain
Development
Pre-school years
High
School years
Numbers
Peer social skills
Language
Symbol
Habitual ways of responding
Emotional control
Vision
Hearing
Low
0
1
2
3
4
Years
5
6
EDI
Graph developed by Council for Early Child Development (ref: Nash, 1997; Early Years Study, 1999; Shonkoff, 2000.)
7
Past reliability and validity studies
•
•
•
•
•
•
•
Teacher to parent inter rater reliability
Teacher to teacher inter rater reliability
Repeat testing intra rater reliability
Construct and concurrent validity
Rasch psychometric property analyses
Indigenous and minority culture validation studies
Schools and the AEDI study
•
Publications downloadable from:
www.aedc.gov.au,
www.offordcentre.com/readiness
How does the
AEDI predict
outcomes through
school?
Predictive Validity
STUDY 1: LSAC – 2004 Wave 1
4year old cohort
STUDY 2: NMHS – 2003 EDI cohort
Predictive Validity – Study 1
Longitudinal Study of Australian Children
• nationally representative sample
• two cohorts of Australian children: 5,104 infants
and 4,976 four year olds
• first wave of the LSAC commenced in May 2004
• face-to-face interviews with parents, parent selfcompleted questionnaires, interviewer observation,
direct child assessment, and teacher completed
questionnaires
Predictive Validity – Study 1
• Of the original 4948 children participating in the 2004
Wave 1 (4 year old cohort), information was obtained
for 89.7% (n=4332) in the Wave 3 2008 data collection.
• AEDI – Nested Sample, children from WA, Vic and QLD
• 717 children with complete data in Wave 1
• 523 children with complete teacher and parent data in
Wave 3 (72.6%).
Predictive Validity – Study 1
•
•
•
•
•
Even gender divide,
5.7% of children had ESL,
1.1% of children were of Aboriginal descent,
4% with medically diagnosed SN status,
Age gap between Wave 1 and Wave 3 ranges
from 3yr 4mths through to 4yr 5mths (avg
gap 3yr 8 mths).
Predictive Validity – Study 1
Instruments collected at ~4 years during Wave 1
– AEDI
– SDQ
– PPVT
– WAI
– PEDS
– PedsQL
– Global Health
Predictive Validity – Study 1
Teacher completed instruments collected at
~8 years during Wave 3
– SDQ
– Academic Rating Scale (literacy)
Sensitivity & Specificity (looking backwards)
• Sensitivity the percentage of sick people who were
correctly identified as having the condition.
• Specificity the percentage of healthy people who were
correctly identified as not having the condition.
Predictive Validity: Outcome is Literacy
Best predictors:
• AEDI (all domains)
• WAI
Worst predictors:
• PEDS
• SDQ
Predictive Validity: Outcome is Maths
Best predictors:
• WAI
• AEDI (all domains)
Worst predictors:
• PEDS
• SDQ
Predictive Validity: Outcome is Behaviour
Best predictors:
• WAI
• AEDI (all domains)
Worst predictors:
• PEDS
• SEIFA
Predictive Validity – Study 1
Outcome Measures at ~ 8 years of age
AEDI Measure
SDQ
ARS
ARS
Language and Mathematics
Literacy
Vulnerable on one or Spec = 0.86
Spec = 0.88
Spec = 0.88
more of the AEDI
Sens = 0.34
Sens = 0.65
Sens = 0.65
Domains. (Australian NPV = 0.94
NPV = 0.94
NPV = 0.94
National Progress
PPV = 0.20
PPV = 0.48
PPV = 0.45
Measure)
Predictive Validity – Study 2
• North Metro Health Service
– Population wide
– Pre-primary (avg 5.6 years)
– 2003
– Original EDI
• Individually data linked to education records (DET WA)
– Govt schools only
– Biased (transience)
– WALNA yr3, NAPLAN yr5 and NAPLAN yr7
Predictive Validity – Study 2
70
% performing poorly on NAPLAN in yr 7
NAPLAN Reading yr 7
60
NAPLAN Numeracy yr 7
50
40
30
20
10
0
None
One
Two
Three
Four
Number of AEDI domains vulnerable on in first year of school
Five
Source:
Brinkman et. al. in
Child Indicators
(2013)
Predictive Validity – Study 2
NAPLAN
Year 3
EDI Domains
NAPLAN
Year 5
NAPLAN
Year 7
Numeracy Reading Numeracy Reading
Numeracy Reading
Physical well-being
.23**
.22**
.25**
.22**
.24**
.24**
Social competence
.24**
.24**
.22**
.24**
.24**
.27**
Emotional maturity
.17**
.16**
.12**
.16**
.15**
.19**
Language and cognitive
development
Communication skills and
general knowledge
Total Score
.42**
.40**
.37**
.40**
.39**
.40**
.36**
.34**
.30**
.34**
.28**
.39**
.36**
.35**
.32**
.35**
.32**
.38**
How do perinatal
factors predict the
AEDI?
Predictive Validity – Perinatal onto the AEDI
• SA Linked Data set at the individual level
– 2003 to 2004 birth population
– Developmentally vulnerable on the 2009 AEDI
• Strongest predictors at birth:
– Childs gender
AUC=0.72
– Gestational age
– Mothers occupational status (ASCO)
– Fathers occupational status (ASCO)
– Mothers smoking status
Predictive Validity – Perinatal onto the AEDI
Sensitivity: % of cases of poor development identified according to the number of risk
factors present perinatally
% of total population of children according to the number of risk factors they have
Relationship between
the AEDI and
Socio-Economic
Position (SEP)
Social Inequality in Child Health and Development in South Australia
2009
Developmental vulnerability
High
Targeted Programs
by high
developmental
Proportionate
Universal
Universal
Programs
Programs
vulnerability
Barriers
to uptake addresses barriers across the social gradient
that increasingly
Targeted Programs
by high social
disadvantage
Low
High
Social Disadvantage
Low
Changes in South Australian Community (LGA) AEDI results
Vulnerable on 1 or more domain from 2009 - 2012
Developmental vulnerability
High
Low
High
Social Disadvantage
Low
So - How does the
AEDI predict school
outcomes considering
SEP?
WHAT WE PREDICTED TO SEE.
The famous Feinstein graph – 1970 British Birth Cohort
Feinstein, L. (2003). Inequality in the
Early Cognitive Development of British
Children in the 1970 Cohort.
Economica, 70 (73–97)
WHAT DO WE SEE?
Feinstein Replication with Australian Data – 2003 Perth AEDI Cohort
Source: Brinkman, Sincovich, Gregory 2013
Reflections
The pertinent questions to ask
• What has happened differently to the
cohort born in 2003/2004 to the cohort
born in 2006/2007 to the cohort born in
2009/2010?
• How do we reduce inequality in child
development?
Conclusions:
• The AEDI has shown to be a
moderate to strong
predictor of school based
outcomes
• Take away message –
improve school readiness
for all with a progressive
universalist approach from
birth to school age.
Translational science, fostering integration.
International
interlude
International Interlude
• Licensing
• Costs
• Protection’s around programing
Vs
• Greater good / Public ownership
• Improving local systems
• Local capacity building
• International comparable and locally relevant
Tonga – locally mapped TeHCI

similar documents