The contribution of population studies to understanding SLCN: The whole is greater than the sum of its parts Sheena Reilly 11 th May 2015 – Born Talking Seminar - Norwich
Jan 19, 2016
The contribution of population studies to understanding SLCN:
The whole is greater than the sum of its partsSheena Reilly
11th May 2015 – Born Talking Seminar - Norwich
2
Outline • Setting the scene about population studies
• Population vs clinical studies: it’s not a competition • What is the value of population studies?
• What have we learned?• Trajectories • Predictors • Associations • The SLI story
• Data from series of longitudinal, population studies
• Why are we frustrated by some of the findings?
3
Bridget Taylor – Seminar 1 Jan 2015
4
• Birth cohorts • Community cohorts• Clinical cohorts or case series
5
6
A population view of language
Population: 5.8 million; 1.2 million children
7
A population view of language
8
• 6 metropolitan local government areas (LGAs)
• ABS - Socioeconomic Indexes for Areas used to select LGAs• Maroondah & Whitehorse (high SES)• Banyule & Brimbank (middle SES)• Whittlesea & Casey (low SES)
Early language in Victoria Study ELVS
Community – Language focused studies
9
Longitudinal Clinical or Case series
• Manchester Language Study – Conti-Ramsden• The largest UK study of individuals with a history of SLI.
• A random sample of all 7 year old children who were attending language units in England in 1995.
• Late Talkers Cohort – Rescorla • 53 mother–child dyads from middle class or upper middle class white families.
10
Population view of low languagePopulations Clinic
presenters
11
12
5081
100
150
Rec
eptiv
e La
ngua
ge
50 85 100 150Non-verbal IQ
Help sought No help sought
4 years of age
5081
100
150
Exp
ress
ive
Lang
uage
50 85 100 150Non-verbal IQ
Help sought No help sought
4 years of age
5081
100
150
Rec
eptiv
e La
ngua
ge
50 85 100 150Non-verbal IQ
Help sought No help sought
7 years of age
5081
100
150
Rec
eptiv
e La
ngua
ge
50 85 100 150Non-verbal IQ
Help sought No help sought
7 years of age
Receptive and expressive language & non-verbal performance at 4 and 7 years of age. Help seeking behaviour shown in the 12 months prior
13
Receptive and expressive language at 4 and 7 years of age: help seeking behaviour in the 12 months prior
Receptive Expressive
14
Summary of similarities/differences Population/community Clinical
All children or representative sample of children with condition
Sub-group of children with condition
May be referred or self selected because of particular traits e.g. higher parent concern
Full range of ability(s) Likely to be more severe or have co-morbidities
Prospective information available Retrospective re early development
Typically Longitudinal Cross sectional or Longitudinal
Inbuilt comparison group Control group* (recruited if available)
Can extrapolate findings to population Findings only relevant to clinical cohort
*information about early development of control group often retrospective
15
Bridget Taylor – Seminar 1 Jan 2015
16
Understanding SLCN• Focus on complex interactions within and between environmental and biological systems• Holistic, not reductionist• Ability to concurrently study speech, language and fluency
Acknowledgments: Jeff Craig
noun!
expressive
receptive!
verb!
pragmatics grammar
17
Access to populations has permitted study of language and its inter-connectedness to:
• Other aspects of communication
• Literacy
• Education
• Psychosocial development
• Samples that will be large enough to permit analysis of gene-environment interactions
18
Population health gains
19
Outline • Setting the scene about population studies
• Population vs clinical studies: it’s not a competition • What is the value of population studies?
• What have we learned?• Trajectories • Associations • The SLI story
• Data from series of longitudinal, population studies
• Why are we frustrated by some of the findings?
20
Knowledge From Our Longitudinal Studies• Typical and disrupted phenotypes (trajectories) of language development• Environmental and biological factors predicting variation
• Social, psychological and educational development
• Health care, education, welfare and societal costs
• Potential for intervention and factors influencing efficacy
21
22
Data collection points
8mth 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Q Q Q Q Q Q Q Q Q Q Q T T T T A A A A A A A
C C
Q : parent-report questionnaireA : face-to-face assessment (child and/or adult)T: Teacher reportC: child self report
child & parent
child
child
child
child
23
ELVS is measuring• Language & communication
• General development & health
• Family history
• Socio-demographic details
• Mental health & family stress factors
• Parent-child interactions
• Child behaviour & temperament
24
ELVS Cohort
• N = 1910
• 50.5% male, 49.5% female
• 3.1% (60) premature (<36 weeks)
• 2.8% (53) non-singletons
• 6% (127) speak a language other than English in the home (~ 50 different languages spoken)
287.7234.7
Language: Expressive vocabulary at 2 years
(Reilly et al Pediatrics 2007; Reilly et al IJSLP 2009)
* MB-CDI: Fenson et al, 1994
O 679
261.3 words
n =1742
Mean SD Range
Total 261.3 162 0 - 679
Girls 287.7 159.7 0 - 679
Boys 234.7 160.6 0 - 679
< 79 < 119
(Reilly et al Pediatrics 2007; Reilly et al IJSLP 2009)
* MB-CDI: Fenson et al, 1994
O 679
words
Late talkers* at 2 years
19.7% (n = 333)
Average Words
65
Average words
39
261.3
27
28
81%
5%Impaired
2 years 4 years
typical talkers
late talkers
impaired
19%
6%
75%typical
typical14%
79.4%
14%Impaired
4 years 7 years
typical talkers
impaired
impaired 6%
73.4%typical
typical
6%
81%
4%Impaired
2 years 7 years
typical talkers
late talkers
impaired
19%
6%
75%typical
typical15%
20.6%
29
Typical - development in the typical range at each age
Precocious (late) - typical development in infancy followed by high probabilities of precocity from 24 mths onwards
Impaired (early) – delayed development in infancy followed by typical language development thereafter
Impaired (late) -Typical development in infancy but delayed from 24 mths onwards
Precocious (early) - high probabilities of precocity in early life followed by typical language by 48 mths
Ukoumunne et al 2011
Five substantive classes
30
Characteristics indicative of social advantage were more commonly found in the classes with improving profiles.
Okoumunne et al 2012Characteristics indicative of social advantage were more commonly found in the classes with improving profiles.
And between 4 and 7 years6
08
01
001
20
CE
LF
-4 R
ece
ptiv
e L
ang
uage
Sco
re
4 5 6 7
Child's age in years
High (32.7%) Medium (53.1%) Low (14.2%)
Receptive Language
60
80
100
120
CE
LF
-4 E
xpre
ssiv
e L
angu
age
Sco
re
4 5 6 7
Child's age in years
High (27.1%) Medium (57.9%) Low (15.0%)
Expressive Language
Mean score and 95% confidence interval presented, groups derived by Latent Class Analysis
32
CELF core age 4 (z score) Change from 4-7 (z score)
Mean diff
95% CI p Mean diff
95% CI p
Disadvantage (1 sd) -0.04 (-0.13, 0.04) 0.31 -0.03 (-0.13, 0.07) 0.61
Family language ability (1 sd)
0.19 (0.13, 0.25) 0.00 0.11 (0.04, 0.18) 0.00
Maternal Education
16-18 vs. post school -0.06 (-0.25, 0.12) 0.51 -0.16 (-0.38, 0.05) 0.13
Less than 16 0.14 (-0.05, 0.34) 0.14 -0.07 (-0.30, 0.16) 0.54
Young Mum2 -0.24 (-0.46, -0.01) 0.04 0.10 (-0.17, 0.37) 0.46
Non English Speaking Background
-0.84 (-1.09, -0.58) 0.00 0.61 (0.31, 0.91) 0.00
1adjusted for child’s gender, IQ, autism, developmental delay, and birth order2up to age 24 at child’s birth
Language at 4 and change from 4 to 7
33
Individual language & literacy trajectories for 20 children selected at random
Taylor et al 2014
34
332 low
Language and literacy patterns
4 years
6 years
8 years
318 start low 2474 start middle-high
2,792 CHILDREN
308 low 2484
2460
10 years 381 finish low 2411 finish middle-high
120
112
110
2286
2264
2189
198
196
222
188
220
271
34
35
5 most common language and literacy patternsfrom 16 possible patterns for 2792 children
Age 4Languag
e
Age 6Languag
e
Age 8Languag
e
Age 10Literacy
n %
Middle-High Middle-High Middle-High Middle-High
1915 69
Middle-High Middle-High Middle-High Low 202 7
Low Middle-High Middle-High Middle-High 118 4
Low Low Low Middle-High 27 1
Low Low Low Low 26 1
• Start on-track and stay on-track is the most common pattern
• Start behind and stay behind is the least common pattern
36
• Language & Learning Group, Division of Mental Health, Norwegian Institute of Public Health• Synnve Schjolberg, Group Leader
• Imac Zambrana
• Eivind Ystrom
• Norwegian Research Council
• Norwegian Ministry for Education
• Dept Psychology, University of Oslo• Francisco Pons
Acknowledgements
37
Trajectories of Language Delay from age 3 to 5Language Delay at 5 years
Language Delay at 3 years
Yes No Total
Yes 318 (3%) 529 (5%) 847 (8%)
No 688 (6.5%) 9 052 (85.5%) 9 740 (92%)
Total 1 006 (9.5%) 9 581 (90.5%) 10 587 (100%)
Language Delay at 5 years
Language Delay at 3 years
Yes No Total
Yes 318 (3%) 529 (5%) 847 (8%)
No 688 (6.5%) 9 052 (85.5%) 9 740 (92%)
Total 1 006 (9.5%) 9 581 (90.5%) 10 587 (100%)
Persistent; Transient; Late-Onset
38
What about the recovered the late talkers?
Dale et al (2014) Am J Speech-Language Pathology 2014
Resolved late talkers: No more at risk of later language imp.than age and gender matched controls
Longitudinal study of twins from age 2 years
Tracked language development at 4, 7 and 12 years
Recommend:Periodic monitoring of recovered late talkers &Screening child in low normal range at school
entry for signs of late language difficulties
39
Predictors
40
Predicting Outcomes at 2 & 4 years
2 years:
12 putative risk factors/predictors did NOT strongly predict outcomes
Variation explained (4.3% & 7.0%) by any 1 risk factor was small.
4 years:
Variance explained at 4 years was around 20%Addition of late talking status (2 years) helped explain 23.6% (rec) and 30.4% (exp) language status.
Reilly et al Pediatrics 2007 & 2010
41
Persistent Language Difficulties at 5 years3% (n=318) of overall sample had persistent language difficulties
38% of children with language delay at 3 years had language difficulties at both time points
Odds of having a persistent language problem:
• Doubled - for boys
• Doubled - family history of late talking
• Doubled - poor comprehension skills at 18mths
• Increased – lower paternal educationEadie et al 2014
42
Transient Language Difficulties at 5 years5% (n=529) of overall sample had transient
language difficulties between 3 and 5 years
Odds of having a transient problem:
• Increased - family history of • late talking or • speech difficulties
• Increased - poor comprehension skills at 18mths
• Doubled – lower levels of maternal education
• Increased – higher birth order
Eadie et al 2014
43
Across international studies4, 5, & 8 year old findings corroborate that
More than half of the late talkers do not present with language difficulties at school entry
Trajectories that broadly represent persistent, transient and late onset language impairment exist across languages
Poor early comprehension skills are a strong & consistent predictor of persistent problems, particularly for girls
Family history of speech, language & literacy difficulties is important & may be a discriminating factor regarding language trajectories
Eadie et al 2014
Associations
The association between child language problems and social, emotional & behavioural difficulties from 4-7 yearsA population-based longitudinal study
Associations
In a population-based sample of 4-7 year old children
To examine cross-sectional relationship between Low Language and SEB Difficulties @ ages 4, 5 and 7
To describe the pathways of LL and SEB Difficulties over time
47
Measures of key constructs
Construct
4yo 5yo 7yo
Language CELF-P2*
CELF-4^
CELF-4
SEB Difficulties
SDQ# SDQ SDQ
*Clinical Evaluation of Language Fundamentals – Preschool Edition (2nd)^ Clinical Evaluation of Language Fundamentals – 4th Edition – Australian Edition# Strengths and Difficulties Questionnaire
48
Strengths & Difficulties QuestionnaireSDQ Domain
Hyperactivity/inattentionConduct problemsPeer problemsEmotional problemsProsocial behaviour
Total Difficulties
Score
• 25 items • Parent report• 3-point scale - not true, somewhat true,
certainly true
49
Movement between groups over time
Typical Language & Typical SEB
Only Low Language
Only SEB Difficulty
Low Language & SEB Difficulty
50
610 (79%)
76 (10%)
59 (8%)
26 (3%)
4 years of age n (%)
84%
9%
5%
2%
5 years of age %
78%
13%
6%
3%
7 years of age %
Typical Language & Typical SEB
Only Low Language
Only SEB Difficulty
Low Language & SEB Difficulty
51
610 (79)
10
26 (3)
4 years of age %
84
9
5
2
5 years of age %
78
13
6
3
7 years of age %
No Low Language & No SEBD
Low Language only
SEBD only
Low Language & SEBD
8%
45%
45%
3%
52
79
10
8
3
4 years of age %
84
9
5
2
5 years of age %
78
13
6
3
7 years of age %
Typical Language & Typical SEB
Only Low Language
Only SEB Difficulty
Low Language & SEB Difficulty
6 (8)
568 (93)
12 (2)
1 (0)
38 (64)
1 (2)
19 (32)1 (2)
5 (19)
8 (31)5 (19)
8 (31)
29 (5)
34 (45)
34 (45)2 (3)
53
79
10
8
3
4 years of age %
84
9
5
2
5 years of age %
78
13
6
3
7 years of age %
6 (8)
568 (93)
12 (2)
1 (0)
38 (64)
1 (2)
19 (32)1 (2)
5 (19)
8 (31)5 (19)
8 (31)
29 (5)
34 (45)
34 (45)2 (3)
50 (8)
28 (4)
6 (1)
21 (29)
3 (4)6 (8)
19 (50)
3 (8)
12 (32)
4 (11)
2 (13)
5 (31)3 (19)
6 (38)
42 (58)
561 (87)
Typical Language & Typical SEB
Only Low Language
Only SEB Difficulty
Low Language & SEB Difficulty
54
Summary
• Strong relationship between LL & SEB problems in children aged 4, 5 & 7 years of age
• LL children experience more SEB difficulties
• Great fluidity & complexity in both language and SEB development over time
55
The SLI story
56Reilly et al 2014
Introduction of the term ‘specific’ and SLI
57
Descriptions of clinical cases or series of cases
Case control studies
Epidemiology
Medicine, Paediatrics, Speech Pathology, Linguistics, Developmental Psychology
Observation has driven theoretical approaches
Reilly et al 2014
58
Typical LanguageSLI
NSLI
SLI
Non-Specific LI
Tomblin and Nippold 2014
Low Language versus SLI (expressive)50
81
100
150
Expre
ssiv
e L
anguage
50 86 100 150non-verbal IQ
Specific Language Impairment Typical development
Developmental Delay typical language - low non-verbal score
Expressive language versus non-verbal IQ
• Typical language + Low NV• Low Language + Low NV
• Typical language + Typical NV• Low Language + Typical NV
20.6% with Low Language
60
Receptive and Expressive Language standard scores and non-verbal performance 4 years 7 years
^ Language & NV IQ within normal range; ☐ Low NV IQ & Language within normal range; X – SLI; - Low Language and NV IQ
61
• Two cohorts - Two countries
• Different language measures
• Iowa - children with SLI and NSLI continuously distributed across range of scores
• The two categories derived from recognised cutpoints are somewhat arbitrary.
• Children with SLI only differed in language severity scores - significantly higher mean language scores than children with NSL
62
Marked social gradient for language outcomes:
Three large scale population studies: - Millenium Cohort Study (MCS) British Abilities Scale - Naming Vocabulary at 5 years by Index of Multiple Deprivation quintile
- Growing up in Scotland (GUS) British Abilities Scale – Naming Vocabulary at 5 years by Index of Multiple Deprivation quintile
- Early Language in Victoria Study (ELVS) Clinical Evaluation of Language Fundamentals (CELF-P2) Core Language at Five years by SEIFA Quintiles
Law et al 2013; Reilly et al 2013; Reilly et al 2014
63
60
80
100
120
140
| C
EL
F C
OR
E L
AN
GU
AG
E s
td s
core
1 2 3 4 5
ELVS 5yo CELF-P2 Core Score by SEIFA quintiles
MCS GUS
ELVS
64
Are outcomes different depending on classification
Findings from three longitudinal population studies:
- Dollaghan (2004): 3 -4 year olds- Tomblin et al 2013: 10 and 16 year olds
- Law et al 2009: 34 year olds
65
Dollaghan (2004)
620 participants drawn from a larger study (n=6000) of otitis media in a socio-demographically diverse population in Pittsburgh, USA.
At 3-4 years Language scores were evenly distributed
- No evidence of an SLI taxon. - Children with SLI were not a qualitatively
distinct group
66
SLI & NSLI significantly greater levels of behaviour problems than typical controls
SLI & NSLI - similar patterns in psychosocial outcomes at both ages.
Tomblin et al – Iowa study
*Achenbach Child Behavior Checklist (CBCL) Teacher Report Form
16 years: children with poor language - less socially skilled regardless of performance IQ.
Psychosocial outcomes* for 6 year olds (SLI and NSLI) at 10 and 16 years of age
Conclusion: poor language skills at school entry confers elevated risk for psychosocial problems both in the
middle and end of the school years.Risk NOT altered by the child’s performance IQ.
67
Law et al 2009
Adult literacy difficulties: N-SLI group (OR: 4.35); SLI group (OR 1.59)
Adult mental health difficulties:
Low employment: SLI group (OR 2.24) than for the N-SLI group (OR1.88).
Having SLI and N-SLI at 5 years was associated with:
Long term risk of early language difficulties is importantIn each case significant predictors of adult outcomes were social factors
68
Outline • Setting the scene about population studies
• Population vs clinical studies: it’s not a competition • What is the value of population studies?
• What have we learned?• Trajectories • Predictors • Associations • The SLI story
• Data from series of longitudinal, population studies
• Summary
• Why are we frustrated by some of the findings?
69
Summary
• Trajectories
70
Preschool (3 years)
71
Kindergarten (4 years)
72
Primary school (5 years)
73
Primary school (6 years)
74
Primary school (7 years)
75
Set priorities for research into Language ImpairmentDevelopment of practical tools:
Risk prediction tools that will zero in on children destined for lasting Language Impairment
76
• Not linear • The way language develops is complex and can accelerate,
plateau and sometimes go backwards.• These fluctuating developmental pathways make it hard to
accurately predict persistent Language Impairment
• A strong biological trajectory dominant in the early years; social disadvantage helps explain more variance in outcome by 4 years.
• Gap may widen by 4 years - possibly because of cumulative exposure to less rich language environments
77
Summary
• Contrasting with fluidity to age 4, language ability across the child population is better delineated from ages 4 to 7
• But there is potential for change in the individual child• Family language environment was the most salient social risk
factor
78
• Activation and acceleration rates vary
• Surveillance rather than screening approach may be required
• Language development - vulnerable to further disruption by social disadvantage in the later preschool years.
• While sobering, offers a fairly prolonged window of early childhood during which these impacts could be genuinely prevented, rather than simply ameliorated.
Clinical and Public Health implications
79
Summary
• Predictors
80
Predictors
• Unlikely to be helpful in screening for language delay in the earlier years (< 2 years)
• More helpful in identifying children with Low Language by 4 years
Recommendation• Language promotion activities in infants younger than 24
months – targeted and based on the level of communication skills displayed
81
Summary
• Associations
82
Summary
• Strong relationship between LL & SEB problems in children aged 4, 5 & 7 years of age
• LL children experience more SEB difficulties
• Great fluidity & complexity in both language and SEB development over time
83
Summary
• SLI
84
• Remove ‘specific’ and use the term Language Impairment (LI)
• Abandon the exclusionary criteria*
• Whilst they are convenient for experimental research they do not reflect the real world where symptoms and conditions may overlap and co-morbidity may emerge over time.
• Agree definition and criteria for research and test these in existing population studies to inform clinical services and policy
* All of them?
85
Outline • Setting the scene about population studies
• Population vs clinical studies: it’s not a competition • What is the value of population studies?
• What have we learned?• Trajectories • Predictors • Associations • The SLI story
• Data from series of longitudinal, population studies
• Summary
• Why are we frustrated by some of the findings?
86
New knowledge
• Access to populations has revealed levels of complexity not recognisable in case-control designs.
• More complex than persistent vs resolution
• Appreciation of continuities and fluctuations with
fluidity -continuing into school years
• IQ discrepancy not relevant to long term language outcomes
• Exclusion criteria create convenient ‘research’ groups • Maybe important for imaging studies
• Social gradient strong – language hyper sensitive to disadvantage
87
Fitting policies and services to language patterns
• The job of fitting policies and services to language patterns is easier if language patterns are stable and predictable
• The job of fitting policies and services to language patterns is NOT easy when language patterns are unstable and unpredictable
• Based on the patterns we see, we have to ask the question, “How well do policies and services fit these language patterns?
Acknowledgement to Cate Taylor CRE conference 2014
88
Fitting population patterns to clinical context
• Speech pathologists• The patterns you observe in children in specialist service systems
may not fit the patterns we see in the general population
• Your job is to change growth trajectories
• Unstable growth patterns provide more scope for change than stable growth patterns (e.g., height)
Acknowledgement to Cate Taylor CRE conference 2014It’s called curiosity
89
Shifting to personalised and population medicine
• “….. clinicians have a responsibility to the population they serve, to the patients they never see, as well as to the patients who have consulted/been referred.
• “….. clinicians, while still focused on the needs of the individual …… when in the consultation, also make decisions about the allocation and use of resources to maximise value for all the people….they serve
• “This is different from management of a service for the patients who present to the service”.
Gray, A. (2013). The art of medicine: The shift to personalised and population medicine. The Lancet, 382, 200-2001.
90
‘The new responsibilities for the clinician practicing population “speech pathology”* not only includes maximising value by getting the right outcomes for the right patients in the right place with the least use of resources, but also ensuring the prevention of inequity related to age or gender or race or social class’.
‘Population “speech pathology”* is not a new specialty, it is a new paradigm that I believe every clinician will sooner or later adopt, with a proportion of clinicians being allocated explicit time for working for the whole population’.
* Inserted
LANGUAGE DEVELOPMENT & DISORDER:
“There are known knowns. These are things we know that we know. There are known unknowns. That is to say, there are things that we know we don't know. But there are also unknown unknowns. There are things we don't know we don't know.”
91
92
Population health gains
93
Acknowledgements• Centre for Research Excellence in Child Language -
Gold, Goldfeld, Law, McKean, Mensah, Morgan, Tomblin, Wake,
• Early Language In Victoria Study - Bavin, Bretherton, Carlin, Eadie, Gold, Prior, Mensah, Okoumunne, Wake.
• Hearing Language and Literacy group - Cini, Conway, Pezic
• Cate Taylor
Thankyou
95
Research Snapshots: late talking
www.mcri.edu.au/CREchildlanguage
96
97
98
Population health gains
99
Morbidities of language delay
SDQ subscale Language impairmenta
M (SD)
No language impairmentb
M (SD)
p value Effect size
Emotional problems
2.0 (2.1) 1.6 (1.7) 0.006 0.2
Conduct problems 1.8 (1.7) 1.3 (1.5) <0.001 0.3
Hyperactivity-inattention
4.1 (2.7) 2.8 (2.3) <0.001 0.5
Peer problems 1.2 (1.7) 0.9 (1.3) 0.005 0.2
Prosocial behaviour
8.1 (1.9) 8.4 (1.6) 0.05 -0.2
Total difficulties 9.0 (4.7) 6.6 (4.7) <0.001 0.5
Social, emotional & behavioural difficulties with language impairment @ 7 yearsa N=189; b N=881