Abstract M 7 Use of producer-recorded health data in determining incidence risks and relationships between health events and culling J. B. 1 Animal 1, Cole A. H. , 1, Sanders* Improvement Programs Laboratory, Agricultural Research Service, USDA, Beltsville, MD 20705 -2350; Dairy Records Management Systems, Raleigh, NC. Introduction Through Dairy Herd Improvement (DHI) programs, data from over 4 million cows are submitted each year for use in US genetic evaluations of dairy cattle. Farm use of computerized record keeping is increasing. This can provide direct benefit to farmers, and improved the efficiency of DHI data collection. Milk production records, pedigree, calving and breeding records, and cow disposal records originating from on-farm computer systems are all used in USDA evaluations. Studies of disease incidence among dairy cows have typically relied on data collected in a research setting, or by veterinarians. Differences in on-farm recording systems were thought to make producer-recorded data unsuitable for the study or evaluation of disease incidence and susceptibility. This study was undertaken to investigate characteristics of producer-recorded health data collected through on-farm computer record-keeping systems. Health traits are of increasing importance to producers. Health data which can be easily collected through DHI may be used to enhance existing genetic evaluation procedures, or develop evaluations for particular health traits of interest like susceptibility to metabolic disease. Data Dairy Records Management Systems (Raleigh, NC) provided producer-recorded health data for 1834 herds from 1997 through 2003. Records included cow ID, date, health event code (4 -character), and a comment field (up to 16 characters of supplemental information). A herd-based dataset included codes used by each farm with a 12 character code definition. Of 3. 7 million event records, 34% were categorized as health disorder events (HDE; 36 categories) and 59% as health maintenance or management events (e. g. vaccinations, hoof trims not associated with lameness, dry treatments, estrus synchronization, etc. ). Health disorders are the focus of this study. Distribution of parities within calving years was fairly consistent, except that in earlier years, a higher percentage of HDE lactations were first parity (41% in 1998 vs. 32% in 2001). In fact, contrary to expectation the percentage of HDE lactations within parity was highest for first parity for 1998. In earlier years, culling without recording a precipitating HDE may have been more common. Since HDE are more likely to precipitate culling in older animals, this skews the distribution of recorded HDE toward first parity (i. e. since first parity HDE-affected animals may be kept, their HDE actually get recorded). General Abnormalities bloat, bovine leukosis virus (BLV) positive, displaced abomasum (DA) or DA surgery, diarrhea, general digestive disorder, fever, hardware, infectious bovine rhinotracheitis (IBR), Johne's positive, neospora, pinkeye, misc. injury, misc. respiratory disease or treatment, misc. abnormal health condition. Locomotion Disorders (LOCO) foot abscess, hoof block, foot injury or wrap, foot rot, lameness, laminitis, warts. Mammary Disorders edema or treatment, mastitis or treatment. Metabolic Disorders metabolic abnormality, acidosis, ketosis and/or treatment, milk fever 1 (MF), hypocalcemia 1. Reproductive Disorders abortion, cystic ovary (CO), dystocia, uterine infusion, metritis or pyometria or treatment 2, retained placenta (RP) or treatment 2, abnormal reproductive cond. , abnormal uterine cond. (e. g. prolapse). 1 Milk fever reported >7 DIM was converted to hypocalcemia. 2 Retained placenta reported >7 DIM was converted to metritis. Herd-years reporting ≥ 1 incidence 703 Overall Reported Incidences 17, 616 Affected Lactations 9, 365 Event All general abnormalities LOCO DA Mastitis Ketosis 533 511 736 287 17, 244 3, 256 34, 320 5, 735 10, 384 2, 712 12, 027 2, 739 The master dataset included 43, 489 HDE lactation records and 135, 659 additional lactations from the same 1244 herd years. Records included production data, and the number and date of first occurrence of each HDE reported with an indication of whether this category of HDE was reported for any lactation in the herd year. Milk Fever RP Metritis/pyometria CO 98 68 674 812 330 215 15, 651 15, 747 287 178 9, 756 10, 120 HDE lactations 15827 11800 7679 4415 2254 1048 466 Calving Year HDE lactations 1997 2074 1998 4028 1999 5459 2000 9962 2001 15428 2002 (Jan-May) 6538 After van Dorp et al. (1999), diseases with a short period of risk (e. g. DA, Ketosis, MF, metritis, and RP) lactational incidence rate (LIR) was calculated as: For diseases with a longer period of risk (e. g. locomotion disorders, mastitis, cystic ovary) the true rate must account for the declining number of at-risk lactations (ARL) during the risk period (as lactations become affected). The incidence density (ID) was calculated as: The HDE included in the master dataset were: Lactation records from the national dairy database were matched with HDE records. Calvings of parity 1 -7, from 1997 through May, 2002 were considered. Lactations having test days in multiple herds were excluded. The HDE were matched with lactation records when they occurred during reported lactation DIM and ≤ 305 DIM. Herd years were required to average ≥ 20 cows in milk and 80% of cows with production records in the national database (i. e. passing evaluation edits). Parity 1 2 3 4 5 6 7 and J. S. Clay 2 Incidence Risk and Density Disease incidence for each HDE category was calculated within herd years having at least one reported incidence. For HDE with low true incidence, this may result in inflated estimates of incidence rate. In this study, however, it is not know whether codes included in the herd-code file are actively used by that herd, or simply available in the on-farm system. Restricting the herd-code file to those selected as ‘in-use’ by the herd could improve the usability of farm data. For example in a herd year with 200 calvings and 70 lactations affected by mastitis, the mastitis ID is: 70∕((200 + 130)∕ 2) = 42. 4 Overall, LIRs for DA, ketosis, milk fever, RP, and metritis/ pyometria were 4%, 7%, 3%, 4%, and 10%, respectively; and IDs for LOCO, mastitis, and CO were 21%, 13%, and 12%, respectively. Over time (1997 to 2002), however, rates for these HDE increased significantly within herds. The effects were small, and similar across all HDEs suggesting an overall improvement in record keeping, rather than true increases in disease rates. Calvings from 2001 (the most current complete year) in 314 herds were selected for further analysis. Event DA Ketosis Milk Fever RP Metritis/pyometria LOCO Mastitis Cystic ovary Lactational Incidence Rate 0. 04 0. 06 SD 0. 036 0. 066 0. 03 0. 11 0. 037 0. 031 0. 112 Incidence Density SD Production and Culling Another approach to validating producer recorded HDE data was to evaluate relationships of HDE with associated production and culling parameters that are part of standard lactation data currently collected through DHIA. Mastitis HDE are expected to be associated with higher SCS and termination code (TC) ‘ 7’ (culled for mastitis), CO HDE are expected to be associated with higher days open (DO) and reproductive HDE are generally expected to be associated with TC ‘ 4’ (culled for reproduction). For 2002 calvings in parities grouped 1 st and >1 st, LSM for SCS difference from herd-parity group were 0. 57 ± 0. 056 and 0. 49 ± 0. 034 (P<0. 0001) for 1 st and >1 st parity cows with ≥ 1 mastitis HDE and LSM for DO difference from herd-parity group were 37. 8 ± 6. 44 and 35. 7 ± 4. 54 (P<0. 0001) for 1 st and >1 st parity cows with ≥ 1 CO HDE. Relative culling risks within herd were calculated with PROC NLMIXED (SAS, 2005). Estimated risk of culling for mastitis (TC=7) was 3. 7 ± 0. 36 times greater for cows with at least one mastitis HDE. Estimated risk of culling for reproduction (TC=4) was 1. 9 ± 0. 14 times greater for cows with at least one RP, metritis, or CO HDE. Discussion Currently, less than 2% of herds report that DHI test data were recorded electronically (including all the herds in this study), however, these herds account for 6% of all cows. Computer use on farms will continue to increase, particularly on larger farms where careful ongoing analysis of management data is critical. Three firms provide almost on-farm management software in use today. Efforts are underway to standardize data collected in progeny test herds. Strategies should be developed now to maximize the usefulness of all producer recorded data in the future. 0. 22 0. 13 0. 338 0. 170 0. 161 Zwald et al. (2004) also calculated LIR from producer recorded data for DA (0. 03), ketosis (0. 10), metritis (0. 21), lameness (0. 10), mastitis (0. 20), and CO (0. 08). Metritis in that study included most reported uterine disorders, while lameness did not include some conditions in LOCO. The lower incidence of mastitis found in this study may be due to less reporting of subclinical mastitis, however this difference bears further investigation. Rates (LIR or ID) for HDE increased with increasing parity. LOCO Mastitis Cystic ovary DA Ketosis Milk fever RP Metritis The raw data used in this study included over 2600 different event codes. Data were carefully inspected to assign records to HDE categories. An on-going project is studying ways in which this process can be automated using regular expression mapping to maximize data conservation and quality. A data exchange format which includes a set of standard codes for identifying common health problems has been developed using information available through the Dairy Herd Improvement program (http: //aipl. arsusda. gov/formats/fmt 6. html). These results demonstrate that producer-recorded health data have similar characteristics to data collected in controlled research settings. Further research on use of producer-recorded health data in genetic evaluations is warranted. References SAS Institute Inc. , SAS Online. Doc® 9. 1. 3, Cary, NC: SAS Institute Inc. , 2002 -2005. van Dorp, R. T. E. , S. W. Martin, M. M. Shoukri, J. P. T. M. Noordhuizen, and J. C. M. Dekkers. 1999. An epidemiologic study of disease in 32 registered Holstein dairy herds in British Columbia. Can. J. Vet. Res. 63: 185— 192. Zwald, N. R. , K. A. Weigel, Y. M. Chang, R. D. Welper, J. S. Clay. 2004. Genetic selection for health traits using producer-recorded data. I. Incidence rates, heritability estimates, and sire breeding values. J. Dairy Sci. 87: 4287— 4294.
Скачать презентацию Abstract M 7 Use of producer-recorded health data
005764f1b78f001fa12cfe3ea5523bd8.ppt