This cross-sectional study was conducted from November 1 to December 31 of 2020 via the online survey tool LimeSurvey^® (LimeSurvey GmbH, Hamburg, Germany).

The questionnaire with a total of 123 questions was created based on a study from the Netherlands (36). Depending on answers given in the first few questions, the further questions differed in amount and content between VHHM participants and non-VHHM participants. Therefore, the time frame needed to answer the questionnaire was evaluated by a two-step pretest and took 20 min for VHHM participants and 12 min for non-VHHM participants.

The included questions were closed single-choice questions, questions with 5-point Likert scales, open-ended questions, and ranking questions. Page 1 contained details of goal and process of the survey and a privacy notice from the conducting Institute of Freie Universität Berlin as well as a data processing consent form was added.

While the first block included questions on general farm data, the second section covered available workforce on the farm. Further on, participants were asked about their subjective definition of VHHM, as participants were intentionally not given a definition of VHHM to prevent bias. Therefore, the classification for participation in a VHHM was left to the participants themselves. This question was followed by a key question on individual participation in a VHHM program, which decided about the further questions asked: Those who denied participation in a VHHM program at the time of the survey followed up with questions about possible potential on their farm and their willingness to pay for veterinary consultation while participants who stated to receive VHHM support on their farm were asked about the detailed design of the service. Eventually, questions regarding the demographics of the participants were asked.

A two-phase pretesting consisted of two phases: Phase I, where three dairy farmers were asked to sample the questionnaire in presence of the first author for understanding of questions and answer options. Subsequently, a few questions were adapted in terms of understanding. In phase II, three different dairy farmers were selected to complete the online format of the questionnaire without prior explanation, whereas the first author recorded the time required. Comprehension problems no longer existed in this phase, but a few questions were shortened, so the time limit of the survey was realistic.

The voluntary participation in the survey was only possible online, and no regional limitations were given. To disseminate the study among the target group, farmers' associations were asked to spread the link among their members (“Deutscher Bauernverband”: 18 associations led by the head association). Moreover, other farmers' associations were contacted by mail and asked for assistance (“Bundesverband der Maschinenringe e.V.” (with all subassociations), “Bund Deutscher Milchviehhalter e.V.,” “Deutscher Raiffeisenverband e.V.,” “Bund der deutschen Landjugend e.V.”). The six largest dairy associations as well as organic associations were likewise included. The willingness of support was given among the sought-after as most replied on first contact.

The data were imported into IBM SPSS Statistics 27 (SPSS for Windows, IBM^®, Armonk NY, USA) for descriptive data analysis and additionally transferred to R Studio [version 4.0.3. “Bunny-Wunnies Freak Out”; R Core Team, 2020 using the packages “dplyr” (37); “car” (38), “desctools” (39), “lmertest” (40), and “corrplot” (41)] for further analysis.

The amount that did not complete the survey and exited before the limit we set was due to analytical reasons: answered questionnaires to at least page 3, including questions of general farm data, available labor force and relative importance of VHHM definition, animal health decision-making, and satisfaction with veterinarian, were included in the analysis. The replies included were examined for duplication using the SPSS function and then subjected to further plausibility checks. No duplicates were identified, missing values were not filled in, and implausible values were removed but not replaced. Frequency tables were created for categorical variables. Continuous variables were checked for normal distribution using histograms and boxplots. To test the stochastic independence of the variables, the Wilcoxon rank test was performed in the part of the descriptive farm data.

The mean values of the variable blocks “advantages,” “disadvantages,” “fulfillment of expectations by vet,” “cooperation with vet,” and “improvements of VHHM” (matrix questions with Likert scale) were used to calculate a new variable (mean value of all equally weighted Likert scales), describing the overall satisfaction with the current VHHM.

Furthermore, to determine the scope of a farm's VHHM program, each VHHM component was scored based on its subquestions (e.g., VHHM component “udder health” included the subquestions “evaluation of herd performance data,” “milk sampling,” “assessment of parlor routine”). Components were weighted equally, and according to the number, the weight of individual subquestions was adjusted. Agreement on all subquestions of all VHHM components would have resulted in a scope of 100%.

Correlation coefficients were used to determine undirected correlations, and in case of normally distributed, metrically scaled data, Bravais-Pearson method was used, whereas, if one variable was at least ordinally scaled, Spearman rank correlation coefficient was calculated.

Regression models were developed using R Studio. First, single-factorial linear regression models were calculated with the performance parameters (dependent variable) and participation in VHHM (influencing variable). For those models where the participation had a p <0.2 and showed normal distribution of residuals, a multifactorial model including relevant confounders was calculated. Therefore, the impact of the participation in VHHM on performance parameter was assessed adjusting for the confounders herd size, region, breed, conventional or organic farming, husbandry system, and staffing ratio. Because of multiple testing, a Bonferroni correction was applied leading to a level of significance of 0.05/5 = 0.01.

To assess if a higher satisfaction or higher scope with VHHM or more frequent visits were associated with a higher 305-day milk yield (305dMY) or a lower age at first calving (AFC), multifactorial linear regression models were calculated including the mentioned variables and confounders (herd size, conventional or organic farming, husbandry system).

The requirements regarding linearity, homoscedasticity, and multicollinearity were tested using the QQ plot of residuals for visual inspection of normal distribution, the Breusch-Pagan test, and Cramer V and variation inflation coefficients, respectively.

With 57,322 registered German dairy farms at the time of evaluation and 216 evaluable questionnaires, the response rate results in 0.38%. Two hundred sixteen of 434 questionnaires were either fully (166×) or partially (50×) completed and were included in the analysis, whereas the remaining 218 did not complete the survey and were excluded from evaluation. Approximately half of the study participants (n = 106) participate in VHHM (Table 1). The VHHM farms kept higher numbers of animals (total stock of milk production including young stock) (mean = 491) compared with non-VHHM farms (mean = 360) (p = 0.0793). The mean 305dMY on VHHM farms differed by 1,218 kg compared with non-VHHM farms (10,195 vs. 8,977 kg; p < 0.0001). Mean AFC on VHHM farms was 26 months, whereas AFC on non-VHHM farms was 27 months (p = 0.0020). The mortality of cows in the period up to 60 days in milk (MORT60DIM) was slightly higher on VHHM farms (mean = 5.47%) than on non-VHHM farms (mean = 4.89%) (p = 0.2527). VHHM farms had a mean staffing ratio of 94 animals/staff member, compared with non-VHHM farms with a mean of 84 animals/staff member (p = 0.3507).

More than three-quarters of the VHHM farms “always/often” discussed decisions with their veterinarian, whereas a little over half of the non-VHHM farms did so (Table 2). The comparison of means revealed that VHHM farms reporting taking important decisions “always” with their veterinarian had numerically the highest 305dMY compared with farmers taking decisions less frequently with their veterinarian. VHHM farms that “always” discussed decisions had numerically the lowest bulk tank somatic cell count (BTSCC). The replacement rate (RR) reached the numerical maximum value within those who “occasionally” made decisions with the veterinarian, regardless of VHHM participation. The AFC was numerically lowest within the VHHM farms that answered “always” with 25 months.

Results in Table 3 demonstrate that more than one-third of the VHHM farms and approximately one-quarter of the non-VHHM farms rated the question about satisfaction with their veterinarian with “very good.” Numerically lowest 305dMY had non-VHHM farms that selected “sufficient” or “unsatisfactory. Analysis showed that non-VHHM farms that indicated “very good” had the numerically lowest BTSCC values. Non-VHHM farms reporting “sufficient” or “unsatisfactory” had the numerically highest BTSCC values, respectively. Regardless of the level of satisfaction with the veterinarian, the numerical values for AFC were between 25 and 26 months for VHHM farms and between 26 to 28 months for non-VHHM farms.

The mean comparison of the ranking question about their relative importance of the term “Veterinary Herd Health Management” (Table 4) showed that the numerically highest 305dMY was reached by the subgroup that ranked “herd data/strategy planning/economy” first. VHHM farms, which considered “problem-solving” in their definition of VHHM, had numerically lowest 305dMY.

In single regression models, only 305dMY and AFC showed significant associations (p ≤ 0.001), and thus, multiple regressions were created for these variables. In multifactorial modeling for 305dMY, adjusted for herd size, region, breed, farm management, housing type, and staffing ratio, VHHM farms had a 660-kg higher 305dMY than non-VHHM farms (adjusted R² = 0.6335) (Table 5). Moreover, AFC on VHHM farms was 0.8 months lower than on non-VHHM farms when adjusted for the confounders written above (p = 0.045).

Table 6 shows that, in the single regression analysis of the variables (VHHM scope, VHHM satisfaction, visit frequency) with the performance data (305dMY, AFC, BTSCC), only a significant correlation between 305dMY and VHHM scope could be found (p = 0.039). With each percent more VHHM scope the 305dMY raised by 18 kg. However, multifactorial modeling of the variables visit frequency, VHHM satisfaction, VHHM scope, herd size, farm management, and housing type showed no significant impact on the 305dMY and the AFC.

The fact that VHHM farms were significantly larger in animal numbers is congruent with a previous study (24). However, it is conceivable that farms with more animals are more interested in current issues and more likely to participate in surveys (45). Another consideration may be that either farms with greater animal numbers are more likely to participate in VHHM, or the structure gained through VHHM participation allows an expansion, but that remains unclear in the final extent.

Moreover, in multiple regression, VHHM farms had a statistically significantly higher 305dMY of more than 600 kg adjusted for region, breed, farm management, housing type, and staffing ratio. Previous studies showed that VHHM-supported farms have significantly higher milk yields than non-VHHM farms (8, 24, 28, 33). The reasons for that can be manifold, as described herein, such as the appreciation of benefits of prophylaxis and a strategic approach.

In addition, a 0.8-month lower AFC on the VHHM farms also suggests better farm management: The AFC has been shown to be influenced by young stock rearing practices, for example, the amount and kind (no waste milk) of milk fed preweaning (46). Several studies show that heifer growth is essentially influenced by the feeding management at calf age, and rearing conditions have an immense effect on the potential of the future dairy cow (47, 48). It would be conceivable that this reflects the merit of VHHM on those farms.

Given the results of the performance parameters, the approach described in a previous study (15) fits: The main characteristic of VHHM is the integrated farm assessment based on valid, collected data and in consideration of economic interests. This approach, they argue, serves to prevent disease and to increase performance. The effect of increased overall performance with VHHM participation has additionally been demonstrated in other studies (28). Nevertheless, this key finding offers potential for further research to determine whether increased performance is a cause or effect of VHHM participation.

For the higher MORT60DIM on VHHM farms, no significant correlation could be shown further on. Still, it could be a consequence of higher milk yields and the purported occurrence of production diseases (11, 49–51) during the critical transition period and subsequently increased involuntary cull of animals (52, 53). Data from Israel explain the occurrence of these diseases through a deficient management and a reduction of said after an intensified VHHM program (50). For intensified dairy farming, it is even more important to put animal welfare at the forefront. Intensification that ignores animal welfare is not a promising or sustainable approach to future farming. The veterinarian is obliged to emphasize this in the interest of the animals within the framework of a VHHM program, which in turn will pay off (54).

The staffing ratio, relating to the total number of animals, might reflect a more intensive form of farming. With 94 animals per staff member on VHHM farms and 84 animals per staff member on non-VHHM farms, the results of this study are in the range of the staffing ratio of US farms, where values between 80 and 100 animals per employee have been described (55–57). Regardless of participation in VHHM, our study found that with each additional animal per staff member, BTSCC increased by 300 cells/mL of milk. Relatively less staff may lead to a less optimized parlor routine and/or routine of bedding maintenance, which implies advantages of a closer ratio.

The more intensive collaboration between veterinarian and farmer in decision-making on VHHM farms is also reflected in the results of previous studies (58). Indirectly, the scope of collaboration appears to be related to farm performance, as farms that consult more frequently with their veterinarian and farms that are more satisfied with their veterinarian both performed better in the mean comparison. As the veterinarian counts as one of the most important advisors to the farm (24, 59), the relationship with the veterinarian is an important contributor to success (28).

Furthermore, survey participants who understood that VHHM was “herd data/strategy planning/economy,” that is, the most encompassing management approach among the definitions given, had the highest farm performance parameters, regardless of their participation in VHHM. This contrasted with the non-VHHM participants, who defined VHHM primarily as problem-solving and had the lowest milk performance of all. The once exclusively therapeutic task of a veterinarian over the years has been increasingly replaced by a disease prophylactic task, so that the commonly known “firefighting veterinarian” is replaced (8, 21).

Certain associations, such as the presence of an effect of VHHM scope, VHHM satisfaction, and frequency of visits or support in a specific area (like udder health) combined with performance parameters, could not be verified in our study. The reason for this could be that the sample size was too small, at least for certain combinations of risk factors. This result may give reason to believe that the extent and design of a VHHM appear to be secondary to farm performance data and are well-worthy subject to further research.