Nationwide, there is a racial/ethnic disparity in incidence of HIV infection and AIDS mortality, with African Americans and Latinos having disproportionately higher rates of both HIV and AIDS than Whites and Asian/ Pacific Islanders. The racial disparity in late diagnosis of HIV/AIDS reflects that of timely -diagnosed HIV, suggesting that late diagnosis may be one important driver of the widening racial disparities seen in the AIDS epidemic. Late HIV diagnosis is defined as a diagnosis of AIDS simultaneously with or within one year of an initial HIV diagnosis. This dissertation research was conducted in conjunction with a larger mixed -methods study to investigate late diagnosis of HIV among Latinos and African Americans in Oakland, and to collaboratively design, implement and evaluate two interventions to address the problem. To conduct a systematic and critical review of the literature on HIV/ AIDS diagnosis among Latinos and African Americans within the United States in order to identify the multi -level social determinants of racial/ ethnic inequities in late-stage HIV. The overarching goal of this dissertation research was to use a Community -Based Participatory Research (CBPR) approach to identify the individual, interpersonal, community, and societal factors, and their interactions, which facilitate or hinder timely HIV testing by at-risk populations among Latinos and African Americans living in Oakland, California. The dissertation had the following aims: 1) To conduct a systematic and critical review of the literature on HIV/ AIDS diagnosis among Latinos and African Americans within the United States in order to identify the multi-level social determinants of racial/ ethnic inequities in late-stage HIV. 2) To analyze qualitative data from the larger parent study in order to identify and examine multi-level factors that exacerbate or attenuate barriers to timely HIV testing and diagnosis among African Americans and Latinos in Oakland, California. 3) To use reflexive analysis and participant observation to examine the benefits and challenges of using a CBPR framework to guide academic/ community collaboration for research and action goals. In order to achieve these aims, I conducted the three studies included in this dissertation. In the first study, I found that the majority of studies on racial disparities in HIV testing and diagnosis have been either cross-sectional or focused on one racial or ethnic group, often in one geographic location. In all studies that compared racial and ethnic groups (n=17), Latinos and African Americans were more likely to receive a late diagnosis than non-Hispanic Whites or Asian Americans. Furthermore, 95.8% (n= 23) of the reviewed studies focused on individual level risk factors or investigated structural barriers via measurements at the individual level. However, in my second study I found that significant factors identified by participants as relating to the late diagnosis of HIV lay beyond the traditional individual -level elements of attitude, norms, control beliefs, and power, and were located within the interpersonal, community, and/or societal levels of the ecological model. In the third study, we found that use of a CBPR approach both benefited the study and presented challenges in four key areas: 1) inclusion/ exclusion; 2) bridging social capital and the role of bridge people; 3) education, status, race/ ethnicity, privilege and power imbalances; and 4) conflicting priorities. The findings from these three studies demonstrate the importance of including communities in investigations of late -stage HIV inequities. Community engagement added value to this study as all partners contributed to aligning every activity toward the dual goals of increased knowledge and improved practice. The study findings also demonstrated that disparities in HIV testing behaviors are shaped and maintained by ecological factors at multiple levels. Especially when investigating health disparities that cluster in marginalized and oppressed communities, community -academic collaboration and multi -level frameworks can enhance findings.