The Northern Ireland Longitudinal Study (NILS) BSPS Conference September 2007 David Marshall (NISRA) PRESENTATION CONTENT 1. 2. 3. 4. 5. 6. What is a Longitudinal Study? What is the NILS? Why does NI need an LS? How Matching was completed Security/Confidentiality Projects 1. WHAT IS A LONGITUDINAL STUDY? A Study of the same group of people over time Example 27-year old Female in 1991 Address Change Address Change Marriage Birth of Child Census | | 1991 Stillbirth | | | | Birth of Child | | | Divorce Cancer Census | | | 2001 Death Census | | | | | | | | | | 2011 | 2. WHAT IS THE NILS? Database of individuals and their life events that can be accessed (in a controlled way) by researchers Began work in 2003 Launched in December 2006 Funded by the Health Department in NI Steering Group 2. WHAT IS THE NILS? cont’d 104 NI birth dates (4 LS, 20 SLS) 2001 Census-based Main Data Sources Demographic Data from Health Registrations 2001 Census Births/Deaths Registrations Property Valuations Number Dates Size Start Date Number of Censuses Data on all 3 LSs E&W LS SLS NILS 4 20 104 500,000 (1%) 1971 4 300,000 (5%) 1991 2 500,000 (28%) 2001 1 Births to sample mothers Births of sample members Deaths of sample members Immigrants Embarkations Widow(er)hoods E&W LS Births to Sample Fathers Stillbirths/ Infant Deaths Internal Migration Cancer Registrations Hospital Episodes Education Marriages Claimant Count SLS ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ NILS ✔ 2. NILS Mortality Dataset Additional exercise All deaths post 2001 linked to Census 65,000 deaths linked Population at risk is 2001 Census Compared to 20,000 in NILS main sample Many of death-based analyses will probably be carried out on this until further information included in NILS 3. WHY DOES NI NEED AN LS? Different People Different issues Different Policies 4. HOW MATCHING WAS PROCESSED Data Preparation Automatic Matching & scoring • • Exact Matching ‘Fuzzy’ Matching Manual • • Confirmation Searching 4. MATCHING LEVELS Approximately : 50% of matches were exact 30% very good fuzzy matches 10% manual confirmation on fuzzy matches 10% Manual searches of database 87.4% unadjusted match rate 96.4% adjusted for Census imputation, accuracy of health service records 5. SECURITY/CONFIDENTIALITY Processed Under Census Legislation Personal Data & Sensitive Personal Data Research Approvals Group and application process Researchers only allowed access to anonymised records Data Held in Secure Environment Closed network Biometric access to PCs No output media allowed (CDs, Data Sticks, Print-outs) All output checked for Disclosure 6. NILS Projects Temperature-related mortality and housing Religion / denomination differences in mortality Impact of non-linkage using NILS mortality Some test findings so far…. Migration ‘Stickiness’ Cultural and educational variables in fertility analyses important Carers have lower mortality Mortality Rates by Religion Suicide risk is strongly related socioeconomic disadvantage and social support at an individual level & area factors have no independent effect 6. ESRC/RDO Money for Researchers £300,000 to for short research projects raise awareness of NILS Application Process (2 stage) SUMMARY 28% Sample of NI Population Includes 2001 Census, birth/death and heath service registration data Data held securely within legislative boundaries Projects started and first paper accepted in academic journal The Northern Ireland Longitudinal Study (NILS) BSPS Conference September 2007 David Marshall (NISRA) The Northern Ireland Longitudinal Study (NILS) nils.nisra@dfpni.gov.uk www.nisra.gov.uk/nils Contextual Data 2001 Census Database 1991 Census Babies linked to Birth Reg. Pre 2001 NILS Core Data Events Deaths Central Services Agency Key demographic information on NILS members (514,000 live) Births to Mothers Births to Fathers Stillbirths & Infant Deaths @ Census Date Births Data (baby linked to Birth Registration) 2001 births onwards VLA/Rating Data New members (36,000) POINTER Address Database Migration (Immigrants, Emigrants, Re-Entrants & Within NI Movers)