This dissertation examined risk factors for hypertensive disorders of pregnancy (HDP) — specifically whether neighborhood socioeconomic deprivation exacerbates individual socioeconomic disadvantage (deprivation amplification) to increase the likelihood of developing HDP. To select the optimal areal unit at which to investigate HDP, geographic proxies for neighborhoods were explored. A thematic review qualitatively examined nontraditional neighborhood boundaries identified through internet sources. Data from 2008–2012 Miami-Dade County, Florida birth records (n=121,421) and the U.S. Census Bureau were used for the remaining analyses. Ordinary least squares (OLS) and geographically weighted regression (GWR) analysis empirically compared the proportion of HDP prevalence explained by six areal units: census block groups, census tracts, ZIP code tabulation areas (ZCTAs), and three types of natural neighborhood — census units clustered based on an eight-item Neighborhood Deprivation Index. Multilevel logistic regression examined relationships between HDP, neighborhood deprivation, and individual-level factors. Odds ratios (OR) and adjusted odds ratios (aOR) were calculated. The thematic review found 22 potential alternatives to census boundaries developed through techniques such as crowd-sourcing and qualitative research. In the sensitivity analysis, census tracts aggregated at the scale of ZCTAs performed twice as well as any other model (GWR2 = 0.27) and were used as the Aim 3 unit of analysis. In the multilevel logistic regression, HDP was associated with moderate (aOR=1.13; CI: 1.05, 1.21) and high neighborhood deprivation (aOR=1.16; CI: 1.07, 1.26). Compared with mothers with private insurance, uninsured women (aOR=1.69; CI: 1.56, 1.84) and Medicaid recipients (aOR=1.12; CI: 1.05, 1.18) had higher HDP odds. Non-Hispanic Black women’s HDP odds were 1.58 times those of non-Hispanic White women. Cross-level interactions — between neighborhood deprivation and educational attainment and neighborhood deprivation and insurance status — did not reach statistical significance. Private sector neighborhood boundaries hold promise for developing new public health tools. Because they are relatively easy to generate from census data, natural neighborhoods may balance tradition and innovation. While no evidence of deprivation amplification was found, results suggested that individual-level and neighborhood deprivation are HDP risk factors. Interventions that target expectant mothers in deprived neighborhoods — particularly non-Hispanic Black and Hispanic women who lack health insurance — may help reduce HDP prevalence and disparities.

在二十世紀九十年代，學者提出地理加權迴歸（Geographically Weighted Regression;簡稱GWR)。GWR是一個企圖解決空間非穩定性的方法。此方法最大的特性，是模型中的迴歸係數可以依空間的不同而改變，這也意味著不同的地理位置可以有不同的迴歸係數。在係數的估計上，每個觀察值都擁有一個固定環寬，而估計值可以由環寬範圍內的觀察值取得。然而，若變數之間的特性不同，固定環寬的設定可能會產生不可靠的估計值。 為了解決這個問題，本文章提出CGWR(Conditional-based GWR)的方法嘗試修正估計值，允許各迴歸變數有不同的環寬。在估計的程序中，CGWR運用疊代法與交叉驗證法得出最終的估計值。本文驗證了CGWR的收斂性，也同時透過電腦模擬比較GWR, CGWR與local linear法(Wang and Mei, 2008)的表現。研究發現，當迴歸係數之間存有正相關時，CGWR比其他兩個方法來的優異。最後，本文使用CGWR分析台灣高齡老人失能資料，驗證CGWR的效果。 / Geographically weighted regression (GWR), first proposed in the 1990s, is a modelling technique used to deal with spatial non-stationarity. The main characteristic of GWR is that it allows regression coefficients to vary across space, and so the values of the parameters can vary depending on locations. The parameters for each location can be estimated by observations within a fixed range (or bandwidth). However, if the parameters differ considerably, the fixed bandwidth may produce unreliable or even unstable estimates. To deal with the estimation of greatly varying parameter values, we propose Conditional-based GWR (CGWR), where a different bandwidth is selected for each independent variable. The bandwidths for the independent variables are derived via an iteration algorithm using cross-validation. In addition to showing the convergence of the algorithm, we also use computer simulation to compare the proposed method with the basic GWR and a local linear method (Wang and Mei, 2008). We found that the CGWR outperforms the other two methods if the parameters are positively correlated. In addition, we use elderly disability data from Taiwan to demonstrate the proposed method.

Thèse numérisée par la Direction des bibliothèques de l'Université de Montréal.