Tropical Animal Health and Production Animal Science and Zoology
Short-term milk-flow trajectories preceding recorded herd exit in dairy cows: a risk-set analysis of routine herd records
Abstract
Routine milking records may characterize production changes before recorded herd-management events. This retrospective, single-herd risk-set study compared 30-day milk-flow trajectories for cows with a recorded herd exit and contemporaneous cows at risk. We analyzed 114,011 milking events from 201 cows (15 March 2020-31 July 2026). The exposure was the ordinary least-squares slope of daily average milk flow during days − 30 to -1; conditional logistic regression jointly handled multiple same-date exits. Of 330 recorded exits, 113 (34.2%) met eligibility criteria, contributing to 55 date strata. The adjusted odds ratio (OR) per standard-deviation increase in flow slope was 0.749 (95% CI 0.592–0.948; P = 0.016). However, 64 of 113 eligible exits occurred on dates with at least three exits. Removing these dates attenuated the date-tied association (49 exits; OR 0.817; 95% CI 0.589–1.134; P = 0.226). Recent milking-data requirements excluded many cases. The author-confirmed physical-departure list agreed with many case dates; some later milking rows retained departed cow identifiers, for which responder reuse is a possible explanation. Individual reassignment and software-entry times were unavailable. Milk-flow slope was associated with an administrative herd-exit record in this selected cohort, but shared exit dates, incomplete lactation and health histories, and incomplete exit-date coverage and possible milking-record identity misattribution limit interpretation. Prospective prediction was not evaluated.