boyle

Document Sample
boyle Powered By Docstoc
					The Scottish Longitudinal Study
 A New Source for Scottish Research



             Paul Boyle
The (Scottish) Longitudinal Study
   The England and Wales Longitudinal Study
    (LS) established following 1971 Census
    •   To study occupational mortality and fertility
    •   Scotland included originally
    •   Withdrew for funding / sample size reasons
    •   Original files destroyed


   Re-establishing the SLS
    • Funded by SHEFC and CSO (£1.5 million)
    • Working in close collaboration with GRO(S)
    • Borrowing as much as possible from ONS
   People
    •   Director: Paul Boyle
    •   Project Manager (Technical): Lin Hattersley
    •   Project Manager (Staff): Katherine Chisholm
    •   Programmer: Zengyi Huang
    •   Programmer: Joan Nolan
    •   20 form pickers / clerical assistants
    •   Research Fellow (Andy Cullis)
    •   Research Fellow (Vernon Gayle)
   Management committee
    •   Paul Boyle (University of St Andrews)
    •   Allan Findlay (University of Dundee)
    •   Robin Flowerdew (University of St Andrews)
    •   Sally Macintyre (University of Glasgow)
    •   Steve Platt (University of Edinburgh)
   Steering committee
    •   David Orr (GROS)
    •   Paul Boyle (SLS)
    •   Ian Mate (GROS)
    •   Muriel Douglas (NHSCR)
    •   Lin Hattersley (SLS)
    •   Rod Muir (ISD, PAC)
    •   Louisa Blackwell (ONS)

    • Secretary:     Katherine Chisholm (SLS)
What is the SLS?

   Provides linked data from the Scottish Census
    and administrative records
   Sampling based on 20 ‘semi-random’
    birthdays
   Initial sample drawn from the 1991 Census
   Similar sample drawn from 2001 Census
    Data sources

   Census                                      Vital statistics
    • 1991 Census                                •   Live births
    • 2001 Census                                •   Stillbirths
    • Including data on occupation,              •   Infant mortality
      economic activity, housing,                •   Deaths
      ethnicity, age, sex, marital status,       •   Widow(er)hoods
      health, education, religion etc.
                                                 •   Marriages

   Population data
    • Immigration                               Health data
    • Emigration                                 • Cancer registrations
    Data sources

   Census                                      Vital statistics
    • 1991 Census                                •   Live births
    • 2001 Census                                •   Stillbirths
    • Including data on occupation,              •   Infant mortality
      economic activity, housing,                •   Deaths
      ethnicity, age, sex, marital status,       •   Widow(er)hoods
      health, education, religion etc.
                                                 •   Marriages

   Population data
    • Immigration                               Health data
    • Emigration                                 • Cancer registrations
                                                 • Hospital admissions
Strengths

   Sample size much larger than most surveys
    • BHPS has ~10,000 people in GB
    • SLS has ~278,000 members + ~518,000
      household members in Scotland (1991)
   The census is compulsory
   Linkage and trace rates are high
   Includes those in communal establishments
   Ability to link hospital admissions data to
    socio-economic characteristics
Weaknesses

   Restricted range of variables
    • Smoking
    • Income
   Census information only collected every
    decade
   Not possible to return to the sample to ask
    extra questions
   The data are highly confidential
How does the SLS differ from the LS?
   Sample percentage larger (5.5% vs 1%)
   20 SLS birthdays, but includes the four LS
    birthdays
   Fewer censuses captured
    • SLS 1991 & 2001 (currently planned)
    • LS 1971, 1981, 1991, 2001
   Some variables in the LS not coded in the SLS
    • e.g. 1991 place of work
   Some variables in the SLS not coded in the LS
    • e.g. hospital admissions and marriages
   The SLS is cheaper!
How far have we got?
   Identification of 1991 sample
    • Electronic records extracted from 1991 Census
    • Forms have been ‘picked’
    • Flagging data passed to NHSCR


   Coding 1991 ‘difficult to code’ information
    •   Only originally coded for 10% Census
    •   Designed interface for data input
    •   Implementing occupation and industry coding software
    •   62,000 basic coding completed
    •   5,000 occupation and industry coding completed


   Programming derived variables
   Linkage and ‘flagging’ through NHSCR
    •   278,359 have been actioned
    •   241,591 have been flagged
    •   2,316 of these are new births
    •   611 are Scottish duplicates
    •   10,258 have been sent to CR Southport to flag on
        the English/Welsh database
    •   12 are English duplicates
    •   3,268 are dummy records
    •   22,583 are in the process of being actioned
        following further patient information
    •   36 are no trace
   Vital statistics
    • Specifications completed
    • 1991 test data received


   2001 Census information
    • Commissioned ‘top up’ coding of 65-74 year olds and 10-
      year occupational coding
    • Received pre-one number census download, to allow
      flagging of imputed data
    • Received post-one number census download, which includes
      imputed characteristics


   Hope to ‘complete’ the job in 2004
Potential uses of the SLS

   Source data for academic research / social
    policy / government departments etc.

   Analysis of successive census data
    • The links between social and geographical mobility
    • The changing geographical distribution of the
      ageing population
    • Work patterns of men and women through the
      lifecourse
   Analysis of successive event/health data
    • Studies of changes in birth spacing
    • Associations between fertility and later diseases
    • The changing importance of different cancers



   Analysis of census and event/health data
    • Occupational mortality and morbidity
    • Economic status and diabetes
    • Socio-economic factors and teenage pregnancy
    • Marital status differences in self-reported illness
    • Survival analysis of cancer by area deprivation and
      occupation
    • Housing tenure and respiratory disease
   Why are Scotland’s fertility rates significantly lower
    than the rates in the rest of Britain?
   How do in-migrants fare after arrival in Scotland?
   Are older people becoming healthier in Scotland?
   Are health inequalities widening between the better
    and worse off in Scotland?
   Given that Scotland has some of the highest lung
    cancer rates in the world, what are the
    characteristics of those who succumb to the disease?
   Do unemployed people in Scotland ‘get on their bikes’
    and move to places where unemployment rates are
    low, or not?
Accessing the SLS
   A culture of data sharing
   The data will be kept in a ‘secure environment’
   A team will be established to provide access to the
    data
   A ‘data dictionary’ will be released once the dataset is
    completed
   Researchers will not receive individual-level SLS data
    directly
   Data will only be released as tabulations, statistical
    summaries or aggregated data
   In-house ‘safe-setting’ modelling of individual-level
    data (by support team)
Why does Scotland need the SLS?




 1991 deprivation in Scotland compared to England & Wales
1991 deprivation census variables in Scotland and England & Wales
1996 age-standardised all cause mortality per 100,000 in Europe
1996 age-standardised mortality for all malignant neoplasms per
100,000 in Europe
1996 age-standardised mortality for malignant neoplasm of the
trachea, bronchus and lung per 100,000 in Europe
Brief examples of LS research
   Does migration exaggerate the relationship between
    deprivation and self-reported illness?
    • Cross-sectional studies assume deprivation influences health
      outcomes
    • However, people move around
    • Migration is selective, not random
    • Health may influence migration
    • Are the ill more likely to move towards deprived places, and
      the well to move away from them?



    Norman P, Boyle PJ and Rees P (forthcoming) Selective migration, health and
       deprivation: a longitudinal analysis Social Science and Medicine
Age distribution of cohorts 1971, 1981 and 1991
SIRs 1991
SIRs 1971   SIRs 1991
Other related activities
   Adding local-area geographical data to
    longitudinal datasets (ESRC)
    • Ideally small area information would be available
    • Raises disclosure risk and confidentiality problems
    • A strategy for adding geographical variables /
      identifiers which does not cause disclosure
      problems

   Training in longitudinal methods for the social
    sciences (ESRC)
    • Collaborative project with the University of Stirling
    • Integrated programme of training activities
    • Traditional training and distance learning package
The future…?

   Linkage of additional data into the SLS
    • Historical IQ tests?
    • Benefits data?
    • Educational data?


   New forms of data access
    • Web-based project design
    • Teaching package with a single dataset
   British LS (BLS)
    • Matching variables
    • Creation of derived variables


   Future funding
    • We only have resources to create the database
    • Research and technical support funding will be
      required
    • Bid currently being considered by ESRC / MRC /
      Scottish Executive
    • Research group will provide longitudinal analysis
      support

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:18
posted:12/1/2011
language:English
pages:32