Letters to the editor Effect of clearance periods on hospital stroke incidence using linked administrative data In data linkage studies of disease inci- dence, prevalence pooling, the effect of including prevalent cases as first-ever cases can be accounted for by using an appropriate clearance period to check for prior hospitalisations (1–3). Provided the clearance is long enough, the first record for an individual is assumed to reflect the first-ever stroke event. To evaluate the use of different clear- ance periods in identifying first-ever cases of hospitalised stroke, stroke hos- pital admissions in Western Australia (WA) were extracted from the Hospital Morbidity Data System covering the years 1988 to 2003 (4). Data linkage using the WA Data Linkage System (5) allowed separations belonging to particular individuals to be identified and their first stroke admissions to be selected. A 12-year clearance was used to identify incident counts for acute stroke for 2000. This procedure was repeated using hospitalisations cov- ering shorter periods to investigate pre- valence pooling over different clearance periods. A total of 3875 admissions with an acute stroke code, comprising 41% of all cerebrovascular disease admissions, were investigated. Using the 12-year clearance first-ever cases as the stan- dard, 41% of unlinked admissions with acute stroke codes in 2000 were deemed repeat stroke admissions, either due to recurrences or multiple admis- sions per event. The percentage repeat acute stroke admissions reduced to 16%, 7% and 2% when admissions were linked for periods of 1 year, 2 years and 5 years, respectively. A clearance of 10 years identified o1% as prevalent cases. A similar pattern was evident in stroke admission data for 2002 using a 14-year clearance as the standard. Knowledge of the relative benefit of different lengths of clearance time can guide future analyses of hospital-based stroke incidence. These results indicate that after 10 years, o1% of those identi- fied as first-ever will be recurrences, while after 5 years, no more than 3% would be recurrent cases. Jurisdictions in which linked hospital records exist for fewer years than in WA, a clearance of 5 years could be considered adequate for certain types of study. This guideline will be useful as other Australian states introduce or extend their linked data capacity. An alternative ‘backcasting’ method (6) using a retrograde survival model to obtain correction factors for over-ascertainment is a valid alternative, being particularly useful for determining trends without loss of information from the early years of observation (6). However, the method does not identify cases at the unit record level for subsequent survival analysis or identify patient samples for more detailed studies requiring the collection of further information. Where no population-based stroke registers or linked data systems exist, incidence estimates using hospital ad- missions as the starting point will con- tinue to need downward adjustments of acute stroke events to account for re- currences and multiple admissions within episodes. The analysis of the ratio between first-ever to total acute admissions calculated here found about 60% of acute events to be first-ever events, and the proportion was closely similar by age and gender. The extent to which this ratio is generalisable within and beyond Australia is uncertain be- cause of differences in discharge practices and the coding of transfers between and within hospitals. Additionally, the pro- portions of recurrent and incident cases may change with time, reflecting the changing impact of the secondary pre- vention of stroke using new medications and interventions. Judith M. Katzenellenbogen 1à , Peter Somerford 2,3 , James B. Semmens 3 , and James P. Codde 3,4 1 Centre for International Health, Curtin Health Innovation Research Institute, Curtin University of Technology, Perth, WA, Australia 2 Epidemiology Branch, Health Department of Western Australia, Perth, WA, Australia 3 Population Health Research, Curtin Health Innovation Research Institute, Curtin University of Technology, Perth, WA, Australia 4 South Metropolitan Health Service, Health Department of Western Australia, Perth, WA, Australia Correspondence: Judith M. Katzenellenbogen à , Centre for International Health, Curtin Health Innovation Research Institute, Curtin University of Technology, Perth, WA, Australia. E-mail: J.Katzenellenbogen@curtin.edu.au, jkatzenellenbogen@meddent.uwa.edu.au DOI: 10.1111/j.1747-4949.2010.00455.x Repeat acute stroke hospital admissions, as a percentage of all acute stroke admissions, utilising different clearance periods for data linkage, Western Australia 2000 and 2002 Year of admission 2000 2002 Acute stroke admissions % repeat admissions Acute stroke admissions % repeat admissions No linkage 3875 41Á0 3740 41Á3 Linkage within index year 2679 16Á0 2594 15Á3 2-year clearance 2425 7Á1 2367 7Á2 5-year clearance 2310 2Á4 2264 3Á0 10-year clearance 2260 0Á3 2216 0Á9 12-year clearance 2254 0Á0 2204 0Á3 14-year clearance 2197 0Á0 & 2010 The Authors. 336 Journal compilation & 2010 World Stroke Organization International Journal of Stroke Vol 5, August 2010, 336–339