Influenza hospitalizations are positively associated with poverty. Therefore, individuals in lower socioeconomic brackets are considered to be members of at-risk populations. With the goal of improving situational awareness, we developed a framework for combining multiple data sources to predict at-risk hospitalizations. The data sources considered were: emergency departments, primary health care providers, and Google Flu Trends. We demonstrate that out-of-sample performance was lowest in the most at-risk zip codes, which identifies a key data blindspot, highlights the importance of understanding the dynamics of influenza in at-risk populations, and reveals the far-reaching public health consequences of restricted access to health care.