A P Polednak1. 1. Office of Policy, Planning and Evaluation, Connecticut Department of Public Health, Hartford 06134, USA.
Abstract
OBJECTIVE: This study assessed the agreement in coding of race and Hispanic ethnicity for the same patients in a hospital discharge database and another database. METHODS: Race-ethnicity coding was examined for 72,276 cancer patients discharged from Connecticut hospitals (1992-1997) who were linked with the statewide cancer registry that included information on birthplace (for 76% of patients) and surname (for all patients). Surnames in the cancer registry were also linked with a list of Spanish surnames, to improve ascertainment of ethnicity among patients of probable Hispanic origin. RESULTS: Kappa coefficients (kappa) indicated substantial agreement (beyond that expected by chance) for White (kappa = .74) and Black (kappa = .93) race, Hispanic ethnicity (kappa = .73) and non-Hispanic White race-ethnicity (kappa = .83) categories. Kappa was moderate (ie, .52) for the Asian-Pacific Islander race category. Only 42% of Asian-Pacific Islanders and 62% of Hispanics in the cancer registry were similarly coded in the discharge database. CONCLUSIONS: Although both databases are imperfect, the findings suggest that ascertainment of certain racial-ethnic groups in hospital discharge databases could be improved if birthplace and surname were available.
OBJECTIVE: This study assessed the agreement in coding of race and Hispanic ethnicity for the same patients in a hospital discharge database and another database. METHODS: Race-ethnicity coding was examined for 72,276 cancerpatients discharged from Connecticut hospitals (1992-1997) who were linked with the statewide cancer registry that included information on birthplace (for 76% of patients) and surname (for all patients). Surnames in the cancer registry were also linked with a list of Spanish surnames, to improve ascertainment of ethnicity among patients of probable Hispanic origin. RESULTS: Kappa coefficients (kappa) indicated substantial agreement (beyond that expected by chance) for White (kappa = .74) and Black (kappa = .93) race, Hispanic ethnicity (kappa = .73) and non-Hispanic White race-ethnicity (kappa = .83) categories. Kappa was moderate (ie, .52) for the Asian-Pacific Islander race category. Only 42% of Asian-Pacific Islanders and 62% of Hispanics in the cancer registry were similarly coded in the discharge database. CONCLUSIONS: Although both databases are imperfect, the findings suggest that ascertainment of certain racial-ethnic groups in hospital discharge databases could be improved if birthplace and surname were available.