Acute stroke incidence estimated using a standard algorithm based on electronic health data in various areas of Italy

AIM: to define an algorithm and implement it in various areas of Italy, in order to evaluate acute stroke incidence through current databases. SETTING: Lazio, Tuscana , Venezia AULSS 12, Torino ASL 5. PARTICIPANTS: resident-based population in the above mentioned 4 areas during 2002-2004. MAIN OUTCOME: Annual and triennal incidence rate (crude and standardized per 100,000 inhabitants with 95% CI) by sex and age classes (0-14, 15-34, 35-54, 55-64, 65-74, 75-84, 85+), standardized rate of mortality by sex and areas. METHODS: acute stroke incident cases during 2002-2004 in the 4 Italian areas were identified through hospitalization databases (SDO) and death causes (CM). The selection was made including hospitalization cases (no outpatients) and deceased people with a discharge or death code ICD9-CM 430*, 431*, 434*, 436* with no hospitalization for stroke diagnosis in the previous 60 months. Moreover, patients with 438* codes in secondary diagnoses and patients with hospital discharge from rehabilitation or long-hospital units were excluded. RESULTS: men have a higher crude incidence rate than women (+30%). The age-specific rates show a large variability among the areas for elderly people (65+ for men and 75+ for women), with higher rates in Toscana in both genders (cases per 100,000 inhabitants: 260.1 men; 193.1 women). Intermediate values were found in Torino and in Lazio; the lowest values are reported in Venezia (men: 182.5; women: 1368). Standardized mortality rates also present higher mortality levels in the two regional areas (Lazio and Toscana) and lower levels in the two urban areas (Torino and Venezia). CONCLUSIONS: It is not easy to evaluate the algorithm. Results seem compatible enough with other studies and show a certain consistency with current mortality data. Different socio-economical characteristics could account for differences in the estimated incidence among areas. However, diferences in the quality indicators suggest that a validation study with standardized diagnostic criteria will make quality evaluation of the algorithm possible.