Disease recording in the Nordic countries
According to the legislation in the Nordic countries, veterinarians should record when a cow is treated with any drug with a withdrawal period, e.g. any antibiotics. In NO and SE, veterinarians are required to write a record regardless of whether the cow diagnosed with a disease was treated or not. In SE, the veterinarian needs to submit the record to the Board of Agriculture (
). If the herd participates in milk recording the information is subsequently transferred to the Swedish Dairy Association’s cattle database. In DK and NO, recording of all diseases, i.e. the submission of the disease record to a central database, is mandatory for herds in milk recording. In FI about 90% of the herds in milk recording also participate in disease recording, to have access to herd health and breeding advisory services. In DK, the farmer or the veterinarian submits the records to the central database. In FI, the veterinarian records the data on the cow’s health card and an artificial insemination (AI) technician or the farmer submits the records to the central database. In NO, the farmer, herd management advisor or veterinarian submits the records. In all four countries, farmers can also record diseases themselves with or without involvement of a veterinarian. For a detailed description of the disease recording systems, see, e.g., Olsson et al.
], Bartlett et al.
], Gröhn et al.
], Østerås et al.
] and Mörk et al.
There are two exceptions to the rule that clinically diseased cows may not be treated with prescription drugs without prior clinical examination by a veterinarian. The first is the “common practice” for milder cases of CM in FI. Finnish dairy producers may take a milk sample and send it for bacteriological analysis, and consult the veterinarian when the results are known. Depending on the veterinarian’s judgement, treatment may be prescribed without the veterinarian visiting the farm. The farmer should record any treatment on the cow’s health card; the information is then submitted to the central database as described above. The second exception is the Danish system for veterinary care in dairy cattle that was introduced in 2006 and by 2008 was used by approximately 8% of Danish dairy farmers
]. Farmers may, after signing a contract with their veterinarian, initiate treatment themselves for certain disorders and only with a predetermined drug of choice. Instead of visiting each diseased cow, the veterinarian makes frequent scheduled visits to examine all cows at specific and predetermined stages of lactation and to follow up previously treated cows. In this system, dairy cows can be regarded as veterinary-treated since the diagnostic criteria and follow-up of treatment are under veterinary supervision. Diagnoses and treatments made by the farmer are recorded and submitted to the database by the farmer. We hereafter refer to this system as the “Danish New Herd Health (DNHH)”.
Study herds and data collection
The data collection for CM was conducted as a part of a larger study where data for four major production-related complexes where studied and where data collection was conducted in parallel for all four complexes (mastitis, lameness, and metabolic and reproductive disturbances). The number (%) of dairy herds in milk recording in DK, NO and SE was approximately 4,000 (90%), 11,800 (97%) and 5,000 (76%) herds, respectively, in 2008. In FI, the health surveillance system kept records of approximately 8,700 (70%) herds in the same year, 2008. The target population for this study were cows belonging to herds participating in these national schemes with a herd size of at least 15 cows. Four random samples were selected from these populations. Country-specific sample size calculations were done, with the disease event as the unit of interest. Calculations were based on an expected completeness of 80%, country-specific average herd sizes, the national registered disease incidences of previous years and a maximum width of the confidence intervals of 0.1. To account for the unknown clustering of registered disease events within herds, the sample size was doubled. The target sample size was 150–200 herds per country, which was considered conservative. Taking country-specific expected response rates and practical circumstances into account, 1,000, 900, 800 and 400 herds in DK, FI, NO and SE, respectively, were invited to participate. The owners of these herds were sent a letter explaining the purpose of the study and giving a description of what was expected of participating herds. In FI, NO and SE, participants were offered lottery tickets with prizes ranging from a gift voucher for travel (FI) and weekend travel to an agricultural show (NO) to free embryos (SE). In DK, no incentives were offered to participating farmers.
Data collection was done during two study periods in 2008, from 15 February to 15 April and from 15 September to 15 November. In DK, the first study period started and ended 2 weeks later for practical reasons. The study periods were chosen to avoid recording during the most labour intensive months for the farmers, but to still capture possible seasonal variation. Participating farmers were provided with written information, recording sheets, and prepaid envelopes for returning the recording sheets. Recording sheets could also be sent by fax or e-mail. In FI, an online recording option was available. All farmers were given a reminder by telephone or text message after 1 month and at the end of each study period. Where no disease events were detected, farmers reported “no events” either by returning a recording sheet with only the herd identification (ID), or by sending a text message, fax or email.
Before the second period all farmers were contacted by mail or phone and provided with new recording sheets, instructions and envelopes. Farmers who did not want to continue participating after the first study period, or who had failed to return their recording sheets or report absence of disease events were excluded. In DK and NO, a few farmers participated only during the second study period.
A number of parameters for the participating herds and target populations of herds were investigated to assess how representative the study herds were. The selected parameters represented milk production, udder health, reproductive performance and indicators of metabolic status. Due to limited data access, official country statistics
] and only mean values without measures of spread were used for DK instead of target population data.
The farmer-recorded data
The diagnosis and clinical signs of mastitis, i.e. ‘redness, soreness or swelling of the udder’ and ‘visible changes of the milk’, as well as systemic signs, e.g. “fever’ and ‘poor appetite’, were listed on the recording sheets with a tick-box next to each item. Farmers had to fill in the herd ID, cow ID, date of detection of symptoms, dates of veterinary visits, who diagnosed the cow (farmer or veterinarian), and any treatments the cow was given or absence of treatment. All CM events were recorded at the cow-level (quarter-specific data were not recorded). A copy of the recording sheet, in English or any of the four Nordic languages concerned, is available from the main author upon request.
Instructions were printed on the back of each recording sheet. Only cows were included in the study, i.e. no heifers or young stock. If a cow recovered, but had a relapse, a new recording sheet was filled out. In general, the same recording sheet was used for all revisits by a veterinarian for the same disease event, and all therapy for that event, etc. Only events of clinical disease were recorded, i.e. where the cow was not well according to the farmers’ judgement. Events of subclinical mastitis, e.g. high somatic cell counts (SCCs) at test milking or at California Mastitis Test (CMT) testing, were not recorded as ‘CM’. No additional cow-side test was used, unless such tests were part of the normal routines in the herd.
Data from the submitted recording sheets were manually entered into an Access database (DK and SE) or electronically scanned (FI and NO). In FI, web-based recording was also available. All records were manually checked for data entry errors. Records were removed if the date of detection was outside the study period (n
25), if no dates were noted (n
36), or if the herd ID or cow ID was missing or unknown (n
14). All data management and analysis was done using the statistical package SAS 9.2 (SAS Institute Inc., Cary, NC, USA).
A disease event was defined as “CM” if one of the following was indicated by the farmer on the recording sheet: “mastitis”, “changed appearance of the milk“or “swelling, redness or soreness of the udder” following the International Dairy Federation’s definition
]. A CM event was defined as “veterinary-visited” if one of the following was indicated: “veterinary-examined”, “diagnosed by a veterinarian”, “treated by a veterinarian” or if a date for a veterinary visit was stated. For herds participating in the DNHH system (n
10), CM events were also defined as “veterinary-visited” if the farmer had recorded “treatment with antibiotics or non-steroid anti-inflammatory drugs by the farmer”, because such treatments were under veterinary supervision. For the same reason, the definition for “veterinary-visited” CM events in the Finnish data was extended to antibiotics given by the farmer, to include cows that were treated after analysis of a milk sample, and after a telephone consultation and prescription by the veterinarian.
All CM events that occurred within 8 days of the first event were considered one CM case
]. The date used for defining a case was the first visit date even though each recording sheet could have several dates including date of detection and several visits dates. If no visit date was given on a recording sheet, the date of detection was used. If any of the events within a case were defined as “veterinary-visited”, the entire case was defined as “veterinary-visited”. For the case date, the date of the first event was used.
Data from the national cattle databases
Data from the four national cattle databases were retrieved in May 2009 and rearranged into one common disease database (CoDD). The retrieved information included cow-level information on entries and removals, reproductive events, milk recording and disease records. Disease records included events during the time from 7 days before to 7 days after the two study periods. Diagnostic codes differed between the countries. For CM, seven, three, two and 14 codes were used in DK, FI, NO and SE, respectively, which represented varying degrees of clinical signs or other characteristics of CM (Table
). To make a comparison between countries possible, all diagnostic codes for CM were re-coded to “clinical mastitis”. Disease events had a diagnostic date according to the veterinarian’s (or farmer’s) recording. Also in the CoDD, all events with the diagnostic code CM within 8 days of the first event were viewed as one mastitis case, with the case date being the date of the first event.
The country-specific diagnostic codes for clinical mastitis in 2008
The CM cases in the CoDD were matched to CM cases in the farmer-recorded data. A match was defined as the same country, herd ID and cow ID with either the exact same case date (n
1597), or allowing a discrepancy of 7 days between the case dates (n
223). Larger date discrepancies were evaluated. Completeness was calculated by country as the proportion (p) of farmer-recorded cases (n) that could be matched to a CoDD case, i.e., referring to Figure
b, as a/(a
c). The completeness was calculated separately for all cases in the farmer-recorded data and for cases defined as “veterinary-visited” (Figure
). Double-sided confidence intervals for the completeness at a significance level of 95% (z
1.96) were calculated as follows:
Illustration of the data flow from the diseased dairy cow to the national cattle databases in the four Nordic countries included in the study.
To account for the unknown clustering of cases within herds, the confidence intervals were increased post hoc
] by factor 2.
Cases of CM that were present only in the CoDD and not in the farmer-recorded data were considered not properly recorded by farmers in the study, rather than being considered errors in the CoDD. In an alternative calculation of completeness these cases were reclassified as if they had also been recorded by the farmers, i.e. completeness
b). We hereafter refer to the farmer-recorded data (FRD) with the excess CoDD CM cases added as “adjusted FRD”.
If the number of CM cases in the FRD divided by the number of CM cases in the CoDD was >0.9 the herd was classified as a herd where the farmer was a good reporter of findings. Consequently, for herds with many excess CM cases in the CoDD, farmers were classified as poor reporters. The completeness calculations were repeated, this time including only good reporters.
For the Finnish FRD, an additional completeness calculation was done where antibiotics administered by the farmer were not included as a criterion for a CM case to be defined as “veterinary-visited”, i.e. using the same definition for “veterinary-visited” as in NO, SE and DK (excluding herds in the DNHH).
Incidence rate calculation
The time-at-risk for CM was defined as follows: in the CoDD, cows were given one record per lactation and disease case, meaning that cows with no mastitis events had one record per lactation. When a cow left or re-entered the herd during the study period the observation period for that cow ended or started on that date. If a cow calved during the study period, the new lactation, and consequently a new observation period, started on the calving date. The previous observation period ended the day before the calving date, i.e. the dry period was included in the time-at-risk. Heifers were included from the date of first calving. Hence, if a cow both entered and left the herd on the same day she contributed with 1 day at risk during which she could have a registered a disease event.
Country- and database-specific IRs for CM were estimated by Poisson regression (PROC GENMOD, SAS 9.2) with a deviance scale parameter to account for overdispersion in the data. The different dependent variables were the number of CM cases per herd in the CoDD, or veterinary-visited cases or all cases in the farmer-recorded data. Days at risk per herd was the offset variable. The days at risk were the number of cow-days per herd during the two study periods reduced by 8 days per CM case in the FRD or CoDD. Negative binomial models were also evaluated. The changes in point estimates and confidence intervals compared to the Poisson models were minor and did not affect the differences in IRs between countries or between datasets within countries.