Cross-sectional and case-control studies are popular individual- level designs. Cross-sectional studies measure/determine prevalence of an adverse health condition, while case-control studies compare two groups based on the exposure measured retrospectively. Individuals within a population can be divided into groups based on observed characteristics, and a particular group may be more or less susceptible to an adverse health condition. In contrast to individual-level study designs, disease mapping and ecological regression models are well-known methods for modeling population-level characteristics using aggregated data. Such designs emphasize on group-level characteristics and ignore variability within groups. However, optimal prediction of an adverse health condition requires us to integrate these techniques into a general frame-work for modeling individual- and group-level factors simultaneously. To overcome this methodological gap, we formulate the joint Besag-York-Mollie (BYM2) model for modeling adverse health condition, integrating individual- and group-level factors into a single framework. The individual- and group-level factors are modelled using submodels linked through association parameters in the joint BYM2 model. The group-level submodel can incorporate spatial auto-correlation in the outcome and covariate measurement error. We propose a Bayesian approach for inference and present comparative studies, via both real and simulated data. The simulation results demonstrate better performance of the joint BYM2 model for capturing parameter values with reasonable uncertainty. We demonstrate an application for modeling the risk of developing adverse health condition among Canadian secondary school students using individual-, school-, and neighbourhood-level risk factors.