The effect of variable demand aggregation areas on modeling primary health care accessibility in Idaho

Title from first page of PDF file (viewed January 6, 2011)Includes bibliographical references (p. 67-69)This study examines the accessibility of primary health care services to rural populations. Specifically, the data used in this study are from the state of Idaho, which provides a good example of a largely rural area striving to provide adequate primary health care accessibility for many of its residents. In this study, demand aggregation was incorporated into an a Maximal Covering Location Problem (MCLP) accessibility model in order, first, to determine the unmet demand for primary health care services in Idaho, based on the allocation of Idaho residents' demand for primary health care to the state's existing primary health care facilities (PHCF). This procedure identified areas without adequate accessibility to PHCF (locations where demand for Primary Health Care Services (PHCS) is unmet). Secondly, the study used this model to determine potential locations (from a set of feasible sites) that could meet unmet demand determined by the initial models. In addition, the study explored the consequences of different demand locations and different driving time constraints on all of these models' results. The representation of demand in location/allocation accessibility modeling is an important consideration in making accurate determinations of the location of demand since different representations of demand have different consequences with regard to model results. Four different demand representations (tract centroids, block group centroids, aggregated block cluster centroids, and weighted aggregated block cluster centroids) and three different driving time constraints (30, 45, and 60 minutes) were used to run the models. This resulted in a total of 12 deterministic models which indicate where demand for PHCS is unmet. These same demand representations and driving time constraints were then used to run 12 predictive models which evaluate the fitness of 20 candidate sites to meet the unmet demand (as determined by the results of the deterministic models). Lastly, one final predictive model was run using only the most robust candidate site(s) from the 20 potential candidate sites. This final model used the weighted aggregated block demand representation at the 30 minute driving time constraint to make a final determination of the candidate site(s) that would maximally alleviate the unmet demand for PHCS. In general, as the models moved from demand representations using tract to block group to aggregated block, the amount of unmet demand for PHCS decreased. The tract model yielded the greatest amount of unmet demand; The aggregated block model yielded the lowest amount (except at the furthest driving time constraint); and the weighted aggregated block model yielded slightly greater amounts of unmet demand than the non-weighted aggregated block model and yielded the least amount of unmet demand at the furthest driving time constraint. The predictive models showed that only nine of the 20 potential candidates sites could alleviate any of the unmet demand for PHCS, Furthermore, they indicated that of these nine sites, only two could alleviate a substantial amount of unmet demand