Adjusting for Unmeasured Spatial Confounding with Distance Adjusted\n Propensity Score Matching

Propensity score matching is a common tool for adjusting for observed\nconfounding in observational studies, but is known to have limitations in the\npresence of unmeasured confounding. In many settings, researchers are\nconfronted with spatially-indexed data where the relative locations of the\nobservational units may serve as a useful proxy for unmeasured confounding that\nvaries according to a spatial pattern. We develop a new method, termed Distance\nAdjusted Propensity Score Matching (DAPSm) that incorporates information on\nunits' spatial proximity into a propensity score matching procedure. We show\nthat DAPSm can adjust for both observed and some forms of unobserved\nconfounding and evaluate its performance relative to several other reasonable\nalternatives for incorporating spatial information into propensity score\nadjustment. The method is motivated by and applied to a comparative\neffectiveness investigation of power plant emission reduction technologies\ndesigned to reduce population exposure to ambient ozone pollution. Ultimately,\nDAPSm provides a framework for augmenting a "standard" propensity score\nanalysis with information on spatial proximity and provides a transparent and\nprincipled way to assess the relative trade offs of prioritizing observed\nconfounding adjustment versus spatial proximity adjustment.\n