Loss and representativeness in a 53 year follow up of a national birth cohort (The 1946 Birth cohort) Dr Gita Mishra MRC National Survey of Health and Development Department of Epidemiology and Public Health Royal Free Hospital, University College Medical School, London www.nshd.mrc.ac.uk
25
Embed
Dr Gita Mishra MRC National Survey of Health and Development
Loss and representativeness in a 53 year follow up of a national birth cohort (The 1946 Birth cohort). Dr Gita Mishra MRC National Survey of Health and Development Department of Epidemiology and Public Health Royal Free Hospital, University College Medical School, London www.nshd.mrc.ac.uk. - PowerPoint PPT Presentation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Loss and representativeness in a 53 year follow up of a national birth cohort (The 1946 Birth cohort)
Dr Gita MishraMRC National Survey of Health and Development
Department of Epidemiology and Public Health
Royal Free Hospital, University College Medical School, London
www.nshd.mrc.ac.uk
Acknowledgments
MRC NSHD Team membersProfessor Michael Wadsworth (Study director)Professor Diana KuhDr Marcus RichardsDr Rebecca HardySuzie ButterworthStephanie BlackRachel CooperImran ShahWarren Hilder
Benefits of the longitudinal or life course design
• Known sequence and chronology of development and ageing, and of exposures
• Near contemporaneous data collections– only short period of recall
• Ability to describe the diversity of underlying pathways to later health outcomes
The costs of the design in a long-running study
• The fixed sample structure
• Each data collection is informed by contemporary scientific ideas and methods
• Risk over time of increased sample loss and the consequent potential for bias
MRC National Survey of Health and Development (MRC NSHD)
AimsOriginally the study aimed to address 2 specific issues in the years before the establishment of the NHS– Reasons for falling fertility– Effectiveness of obstetrics and midwifery on
premature births, infant mortality, and on promotion of health of mother’s and infants
Has evolved into a life course study investigatingpathways to physical and cognitive ageing
MRC National Survey of Health & Development1946 birth cohort
Birth Registrations3rd - 9th March 1946 (N=16,695)
Population of the maternity survey (N=13,687)
Selection of follow-up sample of all single born, legitimatechildren of fathers in non-manual and agricultural employment
and 1 in 4 of all other single born legitimate children (N=5,362)
Study of cohort first born offspring at ages 4 years and 8 years (N=2,205)
Study of women’s health, annually, at ages 47-54 years (N= 1,572)
MRC National Survey of Health & Development1946 birth cohort
Birth Registrations1 week 1946 (N=16,695)
Population of the maternity survey 1 data collection at birth (N=13,687)
Selection of follow-up sample of all single born, married womenwith husbands in non-manual and agricultural employment
and 1 in 4 of all other comparable births (N=5,362) 20 data collections from age 2 to 53 years
Study of cohort first born offspring at ages 4 years and 8 years (N=2,205)
Study of women’s health, annually, at ages 47-54 years (N= 1,572)
8 data collections
Maternity study of all births in 1 week in England, Wales & Scotland in 1946
Follow-up of a class-stratified sample (N=5362)
In infancy at 2 & 4 years
During school years at 6, 7, 8, 9, 10, 11, 13, 15 years
In early adulthood at 19, 20, 22, 23, 25, 26, 31 years
In middle adulthood at 36, 43, 53 years
Years Cohort ages National policy problems
1946 Birth Costs of maternity, reason forfalling fertility
1947-1950 1-4 years Social class differences in maternaland child mortality and morbidity.Value of health visitors’ work.
1951-1961 6-15 years Increasing the national level ofeducational attainment. The ‘waste of talent’ problem.
1962-1976 16-30 years Outcomes of education in terms ofoccupational choice and skills.Delinquency.
1976- 30 years onwards Ageing processes, self care ofhealth, receptivity to healthpromotion.
Maintaining the study’s momentum
Sources of information on sample loss
By age 53 years
• Unavoidable losses – Death (n=469 8.7%)– Emigration (n=461 8.6%)– Living abroad (n=119 2.2%)
• Avoidable or potentially avoidable– Permanent refusal (n=640 12.4%) *only 28 new cases
– Temporary refusal for this data collection only (first
classified at age 43 y) (n=280 5.2%)– Failure to trace (n= 330 6.1%)
Data collections & contacts with the sample (n=5362)
Year Age in Years
Respond-ent
Data Collector
Target Sample
% achieve
1946 -50 0-4 Mother HV 4993 95%
1951-61 5-15 Mother & child
SN or SD or HV & T
4900 89%
1962-77 16-31 All CMs P, HV, I 4858 78%
1965-71 19-25 Mothers of first born
RN 1783 94%
1982 36 All CMs RN 4838 86%
1989 43 All CMs RN 4826 87%
1993-2000 47-54 Women P 84-90%
1999 53 All CMs RN 3673 3035 (83%)
CM cohort member, HV health visitor, SN school nurse, SD school doctors, T teacher, I interview, RN research nurse, P postal
Attrition
The greatest overall attrition occurred in early adult years (16-31 years)– Cohort member could, for the first time, choose
whether to respond
– 5 out of the 7 data collections were by postal questionnaire
– Name and address changes were particularly frequent
– There may have been an adverse effect on response due to blurring of focus of the study aims during this period
Attrition con’t
The 3 later data collections (at ages 36, 43, and 53 years) have focussed strongly on health and obtained higher response rates than those earlier years
– Clear re-focus on health
– The employment of research nurses to collect data
– Introduction of summary feed-back of findings with a birthday card
– A clear explanation of the study’s aims in letters requesting each data collection
Birthday cards
– Birthday cards were introduced at age 16 years to encourage response after leaving school, which requests notification of changes of name and/or address
– They have been continued ever since but now include details of recent work, with references to recent publications
60th Birthday
Sample characteristics of avoidable losses (refusals or failure to trace)
Raised risk on avoidable loss were found in key variables
Childhood1. Shortness at age 4 years2. Experience of serious illness3. Late achievement of bladder control 4. Childhood social class of crowding5. Paternal manual social class, low cognitive test scores6. Low parental interest in education 7. Teachers’ ratings during adolescence of frequent problems
with discipline, disobedience and aggressionAdulthood
1. Adult social circumstances of low educational attainment2. Manual social class employment, 3. Not owning the home at 26 years 4. Not belonging to clubs or association 5. Being obese at 36 years was also associated with
avoidable loss from the study at age 53 years
Attrition from avoidable causes by quartiles of educational and cognitive score at 8 years
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
Year
Pro
po
rtio
n r
em
ain
ing
in
stu
dy
Low (Q1)
Quartile 2
Quartile 3
High (Q4)
Missing data and multiple imputation
• The importance of checking completeness is strongly emphasised at the nurse training sessions
• 73 % of those who provided data at 53 years, were also successfully contacted on 17 or more of the 20 data collections
• Only 7% of them had taken part in 10 or fewer previous data collections
• Multiple imputation is now used in analysis together with sensitivity analysis to deal with missing items/contacts– Growth and breast cancer risks– Diaries of alcohol consumption
Representativeness
• Representation is important not only for extrapolation, but also for estimating true prevalence, and for maintaining policy relevance
• There are some limitations on the representativeness of this sample– Selection predated the major immigration flows– Excludes births out of wedlock– Excludes multiple births
• Nevertheless it remains representative in most respects of the native population born in the early post war years
Sample Representativeness: Comparison of the weighted sample at age 53 years with 1991(50-54 yrs) census data
Full tim e em ploym ent
0
20
40
60
80
100
Males Females
Per
cen
tag
e
Social class
0
20
40
60
80
100
Non manual Manual Non manual Manual
Per
cen
tag
e
Highest educational qualifications - degree or higher
0
20
40
60
80
100
Males Females
Perc
enta
ge
Gender
0
20
40
60
80
100
Males Females
Per
cen
tag
e Census50-54 yrs NSHD NSHD
CensusAges 45 and up
Census16+ yrs
Sample Representativeness: Comparison of the weighted sample at age 53 years with 1991(50-54%) census data
Marital Status
0
10
20
30
40
50
60
70
80
90
100
Single Married Widow ed Separated/ Divorced
Single Married Widow ed Separated/ Divorced
Per
cen
tag
e
Males Females
Conclusions I
• A high rate of contact can be maintained
• Data collection with direct contact, such as home visits by a research nurse
• Provide information about the work of the study to the study members
– Introduction of summary feed-back of findings with the birthday card and website
– A clear explanation of the study’s aims in a letter requesting each data collection
Conclusions II
The responding sample continues in most respects to be representative of the national population of a similar age
Consistency of response over the study’s 20 data collections has been high.
The size of the sample responding in adulthood is adequate for the study of the major costly diseases and for the study of functional ageing and its precursors.
Although the problems inherent in the prospective design are unavoidable they are not, in the study described, a barrier to scientific and policy value
References
Wadsworth MEJ et al. JECH 1992; 46:3000-304Wadsworth MEJ et al. Soc Sci Med 2003; 57:2193-205Wadsworth MEJ et al. IJE 2006; 35:49-54Longford NT et al. JRSSA 2000; 163:381-402De Stavola et al. AJE 2004; 159:671-682