Location-allocation model is one of the best methods to find the optimal location of public service facilities. Traditional p-median model takes efficiency as a major criterion and applies the adjacency rule—that is, allocating the residents of every demand site to the closest facility, neglecting difference in facilities’ capacity (scale). Hence it is difficult to adapt such model to certain distribution requirements of urban comprehensive hospitals that are moderate equilibrium—residents select comprehensive hospital stochastically with certain probability and location and capacity calculation of the facilities should be solved synchronously due to the spatial interaction between supply and demand. In order to address this type of location-allocation problem, we take the P-median model as the basic framework and discuss the development and applicability of the model, then construct a gravity P-median model based on the spatial interaction between the residents (demand) and urban comprehensive hospitals (supply). The new model makes some improvements in a number of aspects. First, spatial equality, that is, all residents can conveniently reach at least one comprehensive hospital, can be ensured by incorporating the highest travel cost (from the demand site to the adjacent hospital) factor. Second, spatial allocation efficiency is guaranteed through the pursuit of minimizing total weighted travel cost. Third, facility location decision and scale configuration are solved simultaneously by incorporating a facilities’ capacity factor. Fourth, facility scale efficiency and fairness of service quality are ensured by incorporating the minimum scale factor. Furthermore, through the empirical test in Wuxi City comprehensive hospital spatial configuration, the new model is validated and considered effective and practical. After optimization using the new model, compared with the current distribution the new spatial allocation of urban comprehensive hospitals is more equitable and more convenient for residents in the service areas to access; the collaborative distribution of comprehensive hospitals and community health service institutions can be achieved, and therefore the spatial distribution of the health facilities is more reasonable. Instead of practical applications, this study focused on the theoretical approach of model building, so some parameters need to be adjusted based on the supply and demand change when such model is applied to practical planning. It should be noted that such new model gives a relative optimal distribution result, which can support certain decision making for future public facility distribution adjustments or new town construction, meanwhile enrich research on public facility location allocation both in China and abroad.