As technology and data become more available on farm, producers have better opportunities to monitor changes in animal health. This study evaluated changes in real-time sensor data in the days leading up to clinical mastitis (CM) infections caused by gram-positive (GP) or gram-negative (GN) pathogens. Historical data detailing milk yield, quality (electrical conductivity and SCC) and composition (lactose, protein, and fat percentage) as well as daily activity (step count, number of rest bouts, total resting time and average duration of rest bout) data were used for analyses. Milk data were collected using an in-line milk analyzer (AfiLab, S.A.E. Afikim, Israel) at the Virginia Tech (VT) Dairy and University of Florida (UF) Dairy for 14 d before and following a CM event (n = 268). Daily activity measures were collected using a novel pedometry system (Afi PedometerPlus) at VT and a traditional pedometer (Afi Pedometer) at UF. A quarter milk sample was aseptically collected for bacteriology upon detection of CM. A control was matched to each clinical case (n = 268) based on breed, lactation number and DIM. For each cow, the slope of each explanatory variable was estimated by linear regression for days ranging between 7 and 5, 4, 3, 2, or 1 d before CM detection. Univariate, mixed effect logit models were fit with farm as a random effect for each explanatory variable to identify directionality and significance of changes in these variables in the days leading up to CM. Daily step count (P = 0.01) and rest duration (P = 0.04) were significant predictors of infections caused by GN bacteria. For GP infections, lactose (P = 0.01) and protein (P = 0.05) percentages, conductivity (P = 0.05), daily steps (P = 0.03), rest bout duration (P = 0.08), number of rest bouts (P = 0.06) and daily milk yield (P = 0.09) were significant predictors. The most accurate predictor for both pathogen types was daily step count, correctly identifying 77.6% of GN infections and 73.4% of GP infections. Changes in cow activity, or milk components measured in-line, can indicate a case of clinical mastitis, potentially enabling earlier intervention.