Measuring dairy herd heat detection: the use of inter-service intervals

Good oestrus detection is essential for efficient reproductive performance in dairy herds using artificial insemination. Veterinary surgeons and farmers use a variety of tools to measure and monitor oestrus detection, including the analysis of inter-service intervals (ISIs), which rely on an estimation of the normal oestrous cycle length. The normal oestrous cycle length of the cow is also important more broadly when managing and monitoring dairy herd fertility. Whilst the normal inter-ovulatory interval is typically considered to be 21 days, some studies have found alternative intervals to be more prevalent. The aim of this project was to evaluate and develop the use of inter-service interval analysis for the assessment of oestrus detection in dairy herds. In the first study, insemination records from 167 dairy herds from across the UK were used to generate ISI profiles for each calendar year of each herd. Intervals between serves were categorised as short irregular (2-17 days), short regular (18-24 days), long irregular (25-35 days), long regular (36-48 days) or extended (>48 days). Herd calendar year performance was ranked by oestrus detection efficiency, the mean of the top quartile of herd-years had 6%, 40%, 16%, 19% and 19% of intervals in each interval category respectively. The results show a substantial difference to accepted ISI profile targets such as achieving a 6:1 ratio of short regular to long regular ISI, and will be of use when interpreting herd data and target setting for UK dairy herds. The next study used ISI data from 159 herds across England and Wales. Univariable analysis of the subset of 114,572 intervals between 15 and 30 days (a range covering the increased frequency of ISIs occurring at the expected time of the first return to oestrus following an insemination) revealed a modal ISI of 22 days with 90% of the inseminations in this range occurring between 18 and 28 days. Multilevel regression modelling was used to evaluate the associations between cow factors and inter-service interval, whilst accounting for clustering at the herd and cow level. This revealed significant associations between predicted ISI and insemination number, days in milk, lactation 305 day milk yield, and month and year of insemination. These findings suggest the “normal” range of ISI for modern UK dairy cows is longer than expected and that there is a large amount of unexplained variation in cycle length within individual animals over time. The subsequent study used ISI data from 312 herds to identify associations between ISI and conception risk. Conception risk varied significantly with ISI, with a period of increased conception risk broadly corresponding to the period of increased frequency of ISIs identified in the previous study. Unsupervised machine learning was used to identify latent classes of ISIs using conception risk and frequency data. A class of ISIs was identified at 19-26 days that was associated with increased conception risk and increased numbers of inseminations. This 19-26 day period was considered a plausible candidate as a revised “normal” range for an expected ISI. This was further developed in a study, using data from the first study to evaluate the impact of accounting for the 19-26 day period of ISIs on common ISI related metrics. The proportion of re-inseminations occurring at 18-24 and 19-26 days was compared across all herd-years as well as the relative ranking of each herd-year using both the 18-24 and 19-26 day interval. There was a significant difference in the apparent ISI profile performance of the herd-years when using the modified interval compared to the standard interval. There was a significant association between herd-year 305 day milk yield and the change in apparent performance when using the new metric. A series of almost identical random effects regression models explaining 200 day in calf rate, each with a different oestrus detection metric as an explanatory variable, were compared. The model in which return oestrus detection was represented by mean ISI explained the highest proportion of variation in 200 day in calf rate. The final study used regression modelling on a subset of data from 28 herds identified as having good recording of clinical mastitis and lameness incidents. Logistic regression was used to identify associations between disease incidences and changes in the probability of re-insemination at either 18-24 or 19-26 days. Lameness 0-28 days after the first insemination of the interval decreased the odds of a re-insemination at an appropriate time by approximately 20%. Clinical mastitis 1-28 days prior to the first insemination of the interval increased the odds of insemination at the expected time by approximately 20%. The associations were similar for either outcome. The effect appeared less important at a population level when the population attributable risk was calculated. This project has used several “big data” techniques applied to a large set of on farm records to improve understanding of inter-service intervals. New target levels are suggested and a new expected or ‘normal’ range is proposed. These should be useful to advisors helping to improve oestrus detection and therefore reproductive performance on farms