In the era of Big Data, a challenge is how to optimize our use of huge volumes of data. In this paper, we address this challenge in the context of a public health surveillance system which identifies disease outbreaks using individual and population health indicators. Our goal is to automate and improve the accuracy of the selection process of the health indicators, a process which is data-intensive and computationally expensive. The health indicators selection process traditionally has been carried out manually by public health experts in collaboration with health data providers. In particular, we present an approach for identifying sets of over-the-counter (OTC) medicine products whose aggregate sales correlate optimally with aggregate counts of emergency department (ED) visits. Towards this goal, we propose an OTC Analytics Appliance which utilizes a distributed search engine to efficiently generate time series of time-stamped records and supports plug-and-play search and correlation functionalities. Using the OTC Analytics Appliance with the Pearson correlation coefficient function, we evaluate Brute-force search, Greedy search, and Knapsack search for their ability to select the optimal or suboptimal set of OTC products automatically. Our results show that greedy search is the most preferable, producing a set of OTC products whose sales that correlate optimally or near optimally to ED visits, while achieving acceptable search times with large datasets. Also, our evaluations show that our approach using the greedy search can be potentially used to efficiently identify different optimal OTC medicine products for detection of different types of disease outbreaks.