Roos Leslie L, Walld Randy
Manitoba Centre for Health Policy, Department of Community Health Sciences, University of Manitoba, Winnipeg, Canada.
Can J Public Health. 2007 Jul-Aug;98 Suppl 1(Suppl 1):S54-61. doi: 10.1007/BF03403727.
The effects of family and place on health outcomes may be seriously misestimated by standard analytic techniques. The information-rich settings in several Canadian provinces can provide appropriate designs to minimize biases resulting from omitted variables and measurement error. This paper compares siblings with children living in the same neighbourhood (but not in the same family) in terms of health care utilization and health care costs. A complete history of residential mobility since birth permits an estimate of the effects of exposure to different environments.
Registry data from a Manitoba cohort born between 1978 and 1985 and continuously resident in the province generated a large sample of same-sex siblings and neighbours (N = 18,280). Administrative information on physician billings, hospital inpatient stays, and costs provided data on utilization between ages 12 and 17.
Large effects on the outcome variables were associated with being in the same family (correlations up to 0.50), whereas the correlations representing upper limits on neighbourhood effects were usually small. These neighbourhood correlations typically shrank slightly after adjustment for family effects. Higher neighbour correlations with utilization (particularly ambulatory visits) occurred in rural Manitoba and probably reflect variation in access to care. Higher correlations are associated with relatively small neighbourhoods and with families remaining in the neighbourhood for at least 17 years.
Although specific variables taken from administrative data are only marginally predictive, our results emphasize the importance of "family" in affecting health care utilization in Manitoba. The minimal effects of neighbourhood differ from those found by investigators generally using weaker designs and emphasizing the significance of neighbourhood.
标准分析技术可能会严重低估家庭和居住环境对健康结果的影响。加拿大几个省份信息丰富的环境能够提供适当的设计，以尽量减少因变量遗漏和测量误差导致的偏差。本文比较了兄弟姐妹与居住在同一社区（但非同一家庭）的儿童在医疗保健利用和医疗保健成本方面的情况。自出生以来完整的居住流动历史记录有助于估计接触不同环境的影响。
从1978年至1985年出生且一直居住在曼尼托巴省的队列登记数据中，生成了大量同性兄弟姐妹和邻居的样本（N = 18280）。关于医生账单、住院天数和费用的行政信息提供了12至17岁期间的利用数据。
对结果变量有很大影响的因素与处于同一家庭有关（相关性高达0.50），而代表社区影响上限的相关性通常较小。在调整家庭影响后，这些社区相关性通常会略有下降。在曼尼托巴省农村地区，邻居与医疗利用（特别是门诊就诊）的相关性较高，这可能反映了获得医疗服务的差异。较高的相关性与相对较小的社区以及在社区中居住至少17年的家庭有关。
尽管从行政数据中提取的特定变量预测性有限，但我们的结果强调了“家庭”在影响曼尼托巴省医疗保健利用方面的重要性。社区的影响极小，这与一般使用较弱设计并强调社区重要性的研究人员的发现不同。
