What causes the burden of stroke in Scotland? A comparative risk assessment approach linking the Scottish Health Survey to administrative health data.

BACKGROUND

The availability of robust evidence to inform effective public health decision making is becoming increasingly important, particularly in a time of competing health demands and limited resources. Comparative Risk Assessments (CRA) are useful in this regard as they quantify the contribution of modifiable exposures to the disease burden in a population. The aim of this study is to assess the contribution of a range of modifiable exposures to the burden of disease due to stroke, an important public health problem in Scotland.

METHODS

We used individual-level response data from eight waves (1995-2012) of the Scottish Health Survey linked to acute hospital discharge records from the Scottish Morbidity Record 01 (SMR01) and cause of death records from the death register. Stroke was defined using the International Classification of Disease (ICD) 9 codes 430-431, 433-4 and 436; and the ICD10 codes I60-61 and I63-64 and stroke incidence was defined as a composite of an individual's first hospitalisation or death from stroke. A literature review identified exposures causally linked to stroke. Exposures were mapped to the layers of the Dahlgren & Whitehead model of the determinants of health and Population Attributable Fractions were calculated for each exposure deemed a significant causal risk of stroke from a Cox Proportional Hazards Regression model. Population Attributable Fractions were not summed as they may add to more than 100% due to the possibility of a person being exposed to more than one exposure simultaneously.

RESULTS

Overall, the results suggest that socioeconomic factors explain the largest proportion of incident stroke hospitalisations and deaths, after adjustment for confounding. After DAG adjustment, low education explained 38.8% (95% Confidence Interval 26.0% to 49.4%, area deprivation (as measured by the Scottish Index of Multiple Deprivation) 34.9% (95% CI 26.4 to 42.4%), occupational social class differences 30.3% (95% CI 19.4% to 39.8%), high systolic blood pressure 29.6% (95% CI 20.6% to 37.6%), smoking 25.6% (95% CI 17.9% to 32.6%) and area deprivation (as measured by the Carstairs area deprivation Index) 23.5% (95% CI 14.4% to 31.7%), of incident strokes in Scotland after adjustment.

CONCLUSION

This study provides evidence for prioritising interventions that tackle socioeconomic inequalities as a means of achieving the greatest reduction in avoidable strokes in Scotland. Future work to disentangle the proportion of the effect of deprivation transmitted through intermediate mediators on the pathway between socioeconomic inequalities and stroke may offer additional opportunities to reduce the incidence of stroke in Scotland.