Determining the Optimal Methodology for Identifying Incident Stroke Deaths Using Administrative Datasets Within...

Area of Science:

• Public Health
• Epidemiology
• Health Services Research

Background:

• Accurate stroke incidence and mortality quantification are vital for public health surveillance and health system planning.
• Administrative data present a cost-effective alternative to traditional population-based studies for tracking stroke.
• Optimal methods for identifying stroke deaths within administrative data require clarification.

Purpose of the Study:

• To identify the most effective method for determining stroke-related deaths in administrative datasets.
• To compare stroke death counts derived from underlying cause of death (CoD) versus all causes of death (CoD).
• To validate administrative data findings against established 'gold standard' population-based stroke incidence studies.

Main Methods:

• Utilized linked, whole-population hospital and death datasets from South Australia, Northern Territory, and Western Australia (2012-2015).
• Identified first-ever stroke events and calculated fatal stroke counts using both underlying CoD and all CoD.
• Determined 28-day case fatality rates and compared results with Australian population-based stroke incidence studies.

Main Results:

• Incident stroke events totaled 16,150 (underlying CoD) and 18,074 (all CoD).
• Twenty-eight-day case fatality rates were 24.7% (underlying CoD) and 32.7% (all CoD).
• Case fatality using underlying CoD closely matched 'gold standard' population-based studies (20%-24%).

Conclusions:

• The underlying cause of death method yields fatal incident stroke estimates more consistent with population-based studies.
• Estimates derived from all-cause fields in death registers were less aligned with established benchmarks.
• Underlying CoD is recommended for more accurate stroke surveillance using administrative data.