Assessing improvements in survival for stroke patients in the Northern Territory 1992–2013: a marginal structural analysis

Abstract

Objective The aim of the present study was to investigate changes in stroke survival among Indigenous and non-Indigenous patients in the Northern Territory (NT). Methods A longitudinal study was undertaken of stroke patients admitted to NT public hospitals between 1992 and 2013. The Kaplan–Meier method and proportional hazards regression were used for survival analysis. A marginal structural model was applied to adjust for time-dependent confounders and informative censoring. Results There were 4754 stroke in-patients over the period, with 3540 new cases and 837 stroke deaths. Mean age of onset for Indigenous patients (51.7 years) was 12.3 years younger than that for non-Indigenous patients. After adjustments for confounders and loss to follow-up, in-hospital deaths were more likely among Indigenous patients (hazard ratio (HR) = 1.56; P < 0.01) and less likely among males (HR = 0.86; P < 0.05) and patients from remote areas (HR = 0.72; P < 0.01). There was a 3% decrease annually in mortality hazard from 1992 to 2013. Renal disease, cancer and chronic obstructive pulmonary disease had deleterious effects on stroke survival. Conclusions Stroke survival has improved in the NT over the past two decades. The marginal structural models provide a powerful methodological tool that can be applied to hospital administrative data to assess changes in quality of care and the impact of interventions. What is known about the topic? Stroke-related mortality has fallen in the past 30 years in Australia. Indigenous Australians have much worse health outcomes than other Australians, including higher stroke incidence and mortality, but it is not known whether stroke survival has improved for Indigenous stroke patients. What does this paper add? This study measured long-term survival for Indigenous and non-Indigenous patients after hospital admissions for stroke care, using hospital admission data analysed by marginal structural models. The present study demonstrates the usefulness of this approach to the quality assessment of health care interventions. What are the implications for practitioners? Hospital administrative data can, and should, be routinely used for monitoring long-term outcomes of acute care. Although survival has improved for Indigenous stroke patients over the past two decades, their survival remains lower than that of non-Indigenous patients. Coordinated acute and primary care plays a vital role in management of chronic diseases to improve stroke survival.