190 likes | 351 Views
Social, biological and lifestyle risk factors for hospital admission: Analysis of the Scottish Health Survey linked to hospital records. Matt Sutton. Contributors. Richard Lawder (Information Services, NHS NSS) Andrew Elders (Information Services, NHS NSS)
E N D
Social, biological and lifestyle risk factors for hospital admission: Analysis of the Scottish Health Survey linked to hospital records Matt Sutton
Contributors • Richard Lawder (Information Services, NHS NSS) • Andrew Elders (Information Services, NHS NSS) • David Clark (Information Services, NHS NSS) • David Walsh (NHS Health Scotland) • Phil Hanlon (University of Glasgow) • Bruce Whyte (NHS Health Scotland) • Laura Kelso (NHS Health Scotland)
Outline of talk • Project aims • Overview of Scottish Health Surveys • The linked file • Analysis • Results
Project aims • To determine the relative impact of a range of risk factors on the likelihood of hospital admission • To test whether deprivation is an independent predictor of hospital admission • To model the impact on future hospital utilisation of current targets for the reduction of behavioural, biological and social risk factors
Survey data • 1995 Scottish Health Survey • 7,932 respondents aged 16-64 years • 7,363 (93%) consented to follow-up studies • 1998 Scottish Health Survey • 9,047 respondents aged 16-74 years • 8,305 (92%) consented to follow-up studies
Survey content • Individual and household life circumstances • Area deprivation • Individual lifestyles (smoking, drinking, activity) • Self-reported health • Self-reported diagnoses • Biomedical measures (nurse visit)
Record linkage • Respondents linked to the September 2004 version of ISD’s linked SMR01 catalog • 11,396 (73%) respondents linked to • SMR01 hospital admissions • SMR04 psychiatric admissions • GRO death records up to 31 March 2004 • Cancer registrations up to 31 December 2001 • Total of 58,913 administrative records
Linkage results • 11,325 (72.4%) have experienced at least one hospital episode (acute or psychiatric) • 502 (3.2%) have experienced at least one psychiatric hospital admission • 1,003 (6.4%) have had at least one cancer registration • 743 (4.7%) have died
1980s 1998 2004 * * * * * * * * * hospitalisation migration death survey Following respondents
Outmigration • May bias results – ‘healthy migrant effect’ • Based on link to CHI registrations • 331 (4%) respondents excluded from 1998 follow-up • Independent predictors (in multivariate analysis) • Age – highest in 20-24 age group • Higher social class • Degree qualification • Privately rented accommodation • Least deprived areas • Urban/City Centre residence • Country of birth outside Scotland • Ex-drinkers(!)
Analysis • Follow-up of 1998 survey respondents, 16-74 years • Survival analysis of time to event • First hospital admission • First ‘Costly and Complex’ hospital admission • Death • Models allow for: • Survey weights • Clustering by PSU • Competing risk of death in admission models • Different start times
Outcome Hospital admission(s) Death Factors considered Social Area deprivation Income Employment Education Housing Access Summary health stock Self-assessed health Limiting longstanding illness GHQ-12 Prior hospital admissions Biological Body Mass Index Waist-Hip Ratio Blood pressure Total cholesterol HDL cholesterol Gamma-GT Fibrinogen C-Reactive Protein Forced Expiratory Volume Behavioural Smoking Drinking Physical activity Diet
Inventory of models • Three types of event • Any hospital admission • ‘Costly and Complex’ hospital admission • Death • Univariate versus multivariate models • Including and excluding ‘health stock’
Summary of results • Each risk factor has expected results for admissions and deaths in single risk factor models • Health stock is important and reduces significance of lifestyle and social risk factors substantially • Smoking is important in multiple risk factor models • Biological risk factors have little marginal effect in multiple risk factor models (except FEV) and many missing values • Individual/household social factors are more important than area deprivation and influence deaths more than admissions
Self-assessed health and risk of ‘Costly and Complex’ admission *** p<0.001 ** p<0.01 * p<0.05
Smoking status and risk of ‘Costly and Complex’ admission *** p<0.001 ** p<0.01 * p<0.05
Housing tenure and events • Influences risk of death but no significant effect on risk of hospital admission *** p<0.001 ** p<0.01 * p<0.05
Economic activity and events • Unemployment influences risk of death but no significant effect on risk of hospital admission *** p<0.001 ** p<0.01 * p<0.05
Summary • The linked Scottish Health Survey is a valuable resource for examining social, behavioural and biological effects at the individual level • Main strength is for follow-up analysis • Future work will: • focus on admissions for specific diagnoses • simulate the effect of risk factor changes