BACKGROUND:

The Pawtucket Heart Health Program was a community-based research and demonstration project in cardiovascular disease prevention in the town of Pawtucket, Rhode Island and in the non-intervention comparison town of New Bedford, Massachusetts. The study was initiated by the NHLBI in August, 1980.

See also Study 1022.

DESIGN NARRATIVE:

There were four major areas of analysis: 1) gender differences in trends from 1980 to 1991 in total event rates, hospital discharge rates, in-hospital case-fatality and out of hospital deaths for coronary heart disease (CHD) and stroke; 2) gender specific analyses of methodological issues which may affect the interpretation of morbidity and mortality surveillance data; 3) gender differences in the presentation, diagnosis, and treatment of CHD and stroke which may influence morbidity and mortality trends; and 4) gender differences in the association between population trends in risk factors and morbidity and mortality trends.

The study was conducted using morbidity and mortality data collected by the Pawtucket Heart Health Program between 1980 and 1993, and risk factor data obtained through biennial cross-sectional household health surveys conducted in random samples of residents in the two study communities. Morbidity and mortality data were obtained according to standardized epidemiologic surveillance methods. Hospitalized cases of coronary heart disease and stroke were identified through hospital discharge records and were validated using a standardized diagnostic algorithm based on data abstracted from medical records.

Out-of-hospital deaths identified through death certificate data were validated using an algorithm based upon medical history data and information concerning the circumstances of death obtained from informant interviews.

Gender specific trends in both age-adjusted and age specific morbidity and mortality were compared. Trends in rates estimated from events classified according to ICD-9 discharge and death certificate codes were contrasted with those estimated from cases validated by standardized diagnostic algorithms.

Data regarding circumstances of death and in-hospital treatment were assessed in relation to potential influences on morbidity and mortality trends.

Absolute changes over time and average annual percent change over time were evaluated using linear and log-linear regression regression techniques.

Other analytic methods included logistic regression and analysis of variance. Expected annual event rates in men and women were estimated from risk factor data using equations based upon accelerated failure time Weibull models, and the impact of risk factors on mortality rates was assessed by comparing predicted with actual rates.