Statistical and geographic analysis of out-of-hospital cardiac arrest using registry data

Abstract

Background: Out-of-hospital cardiac arrest (OHCA) is the most critical health event that occurs in the community. When the heart stops pumping blood, if resuscitation is not started, biological death will occur within minutes. It is known that patient-level factors significantly affect OHCA incidence and outcome however, area-level variation is often observed. The aim of this thesis was to investigate if area-level grouping or area-level characteristics could be identified that influence OHCA incidence and outcome, and their impact quantified.

Methods: Using data from the Irish OHCA registry, descriptive and geographical analysis of OHCA incidence was performed. Cases were geocoded to Electoral Division (ED) level and combined with national census data, and area-level deprivation data, and classified by urban-rural category. The impact of urban-rural grouping was quantified using multilevel linear regression. To adjust for the impact of a small number of cases at ED-level, and the spatial properties of EDs, Bayesian conditional autoregression (CAR) was used to estimate the relative risk of OHCA. Swedish and Irish registry data was compared using logistic regression to identify the predictors of outcome, and to quantify variation measured. Finally, multilevel logistic regression analyses of outcomes in international airports was performed to allow for a differing effect of predictor variables between countries.

Results: The incidence of OHCA where resuscitation was performed was higher in City and Town EDs (51/100,000 population per year; 95% confidence interval [CI], 46 to 55) than in Rural EDs (35/100,000 population per year; 95% CI, 28 to 42). However, urban-rural grouping accounted for only 2% of variation. Bayesian CAR modelling showed that a one-point increase in relative deprivation was associated with an 11% increased risk of OHCA that occurred at home. Logistic regression analysis of the Utstein comparator group (adults, bystander-witnessed, initial shockable rhythm, presumed medical cause) explained only 17% of outcome variation between Sweden and Ireland, with a 4-fold ‘country effect’ in favour of Sweden. Country-level differences in survival in international airports were also evident, particularly when adjusted for age, gender, and attempted bystander defibrillation (median odds ratio 3.0; 95% credible interval, 1.6 to 14.3]).

Conclusions: Findings did not support changes in provision of resuscitation services based on area-level differences, and only a small proportion of between-country variation was explained by routinely collected variables. As patient-level factors are likely to explain the greater proportion of variation in OHCA outcome, it is recommended that there is international collaboration to ensure comparability of data collection and data interpretation, and to promote comprehensive case capture and maximise data quality. It is also recommended that more explanatory variables are incorporated into OHCA registry data collection. Finally, improvements in survival cannot be achieved without cooperation from local communities, but community preparedness should include: discussion on the inevitability of cardiac arrest as part of life; the prospect of patient survival; and, the need for innovative thinking to make sure that pre-hospital resuscitation is initiated efficiently and effectively.