1. Главная
  2. Ресурсы
  3. Библиотечный поиск
SELECTION OF CONTROLSFOR HOSPITAL-BASEDCASE-CONTROL STUDIESUSING RETROSPECTIVEDATA ON THEGEOGRAPHICDISTRIBUTION OF CASESAND CONTROLS Giovanni LeonardiTony FletcherKvetoslava Koppová2004 год

SELECTION OF CONTROLS FOR HOSPITAL-BASED CASE-CONTROL STUDIES USING RETROSPECTIVE DATA ON THE GEOGRAPHIC DISTRIBUTION OF CASES AND CONTROLS
статья из журнала

Страница публикацииПубликация в OpenAlex
Аннотация: ISEE-542 Introduction: A potential problem in hospital-based case control studies arises if the catchment populations for cases and controls for the same hospital are different. If the exposure is distributed spatially and varies between the two catchment populations, then selection bias may occur. As an example, we refer to a large hospital-based study looking at the incidence of three cancers (non-melanoma skin cancer, bladder cancer and kidney cancer) in relation to exposure to a contaminant in drinking water. In this paper, information on the retrospective distribution of cases and controls for the study areas is used to develop a system for control selection aimed at minimising selection bias. Methods: Data on the incidence of the three cancers of interest and the control diagnosis, including location of patients’ residence, were derived from hospital records in the study area. Each settlement of residency was geo-coded (longitude/latitude) using an on-line global gazetteer. The incidence rate for each cancer diagnosis and control diagnosis were assigned to each geo-coded settlement using census data as denominator. Events were also grouped into quartiles of the population with respect to distance from the hospital and the incidence rate calculated for each quartile. The potential exposure (mg contaminant L-1 water) in each of these quartiles was also estimated using data from a previous water survey. The magnitude of potential bias was computed for a range of approaches to controls selection. Results: Spatial distribution of past control diagnoses indicates that there is spatial heterogeneity of control referral, especially between countries. We are aware there is spatial heterogeneity of exposure to the water contaminant. If control access to hospital in the past is related to control ascertainment in a future study, the conditions of potential bias are met, as shown in Table 1. By rotating sampling of control diagnoses across all hospitals in the study area, and weighting the selection in proportion to the population resident in each subdivision of the study area, we could minimise the expected size of selection bias. Conclusions: We estimated the extent of the possible bias in a variety of conditions in relation to access of controls in a hospital based case-control study. We also outlined a practical strategy for minimising the possibility of selection bias in this type of study. We suggest that the increased logistic effort in implementing the suggested procedure is more than compensated by the increased validity of any results.
Год издания: 2004
Авторы: Giovanni Leonardi, Tony Fletcher, Kvetoslava Koppová, Rupert Hough, Péter Rudnai, Eugen Gurzău
Издательство: Lippincott Williams & Wilkins
Источник: Epidemiology
Ключевые слова: Data-Driven Disease Surveillance
Другие ссылки: Epidemiology (HTML)
journals.lww.com (HTML)