Housing system and herd size interactions in Norwegian dairy herds; associations with performance and disease incidence
© Simensen et al; licensee BioMed Central Ltd. 2010
Received: 6 August 2009
Accepted: 16 February 2010
Published: 16 February 2010
According to the Norwegian animal welfare regulations, it has been forbidden to build new tie-stall barns since the end of 2004. Previous studies have shown that cow performance and health differ between housing systems. The interaction between housing system and herd size with respect to performance and disease incidence has not been evaluated.
Cow performance and health in 620 herds housed in free-stall barns were compared with in 192 herds housed in tie-stall barns based on a mail survey and data from the Norwegian Dairy Herd Recording and Cattle Health Systems. The housing systems herds were comparable with respect to herd size (15-55 cows). Associations between performance/disease incidence and housing system, herd size and year of building the cow barn were tested in general linear models, and values for fixed herd size of 20 and 50 cows were calculated. On the individual cow level mixed models were run to test the effect of among others housing system and herd size on test-day milk yield, and to evaluate lactation curves in different parities. All cows were of the Norwegian Red Breed.
Average milk production per cow-year was 134 kg lower in free-stall herd than in tie-stall herds, but in the range 27-45 cows there was no significant difference in yields between the herd categories. In herds with less than 27 cows there were increasingly lower yields in free-stalls, particularly in first parity, whereas the yields were increasingly higher in free-stalls with more than 45 cows.
In free-stalls fertility was better, calving interval shorter, and the incidence rate of teat injuries, ketosis, indigestions, anoestrus and cystic ovaries was lower than in tie-stalls. All of these factors were more favourable in estimated 50-cow herds as compared to 20-cow herds. In the larger herd category, bulk milk somatic cell counts were higher, and the incidence rate of mastitis (all cases) and all diseases was lower.
This study has shown that there is an interaction between housing system and herd size, and that performance and health is not universally better in small free-stalls than in tie-stalls.
Modern dairy production in large herds most commonly takes place in free-stall barns. Tie-stalls are, however, still an alternative to consider, particularly in smaller herds. In Norway tie-stalls constitute the most common housing system for dairy cows but, according to the Norwegian animal welfare regulations, it has been forbidden to build new tie-stall barns since the end of 2004. A panel appointed by European Food Safety Authority stated that there currently is limited amount of scientific data linking the period per day of being tied in tie-stalls to the level of disease and overall impact on welfare, so this should be studied. A minority of the panel recommended that dairy cattle should not be routinely kept in tie-stalls .
Previous studies have shown that performance and disease incidences vary between free-stalls and tie-stalls. A higher reproductive performance has been recorded in free-stalls [2–4] whereas milk yield has been reported to be lower [3–6]. Lower disease incidence rates have been recorded in free-stalls, i.e. of mastitis[2, 6], teat injuries [4, 7, 8] and ketosis[2–4, 9], whereas cows in free-stalls have shown poorer claw health [10–12]. Milk somatic cell counts have been found to be higher in free-stalls [5, 13] or at the same level in the two housing categories [2, 14]. When comparing housing systems, results may as well interact with differences in herd size. However, the interaction between housing system and herd size has not been studied and should be evaluated further.
The purpose of the present paper was to study differences in performance and disease incidences between free-stall and tie-stall housing for dairy cows in herds with approximately equal size, and to evaluate the interaction between housing system and herd size. The study is based on mail survey and data from the Norwegian Dairy Herd Recording System (NDHRS) and the Norwegian Cattle Health Recording System (NCHRS).
Materials and methods
An overview of types of dairy cow housing systems (free-stalls or tie-stalls) was obtained by the NDHRS personnel from 11,600 dairy farms, which was 81% of all farms participating in the NDHRS. Fourteen per cent (1,600) of these farms were confirmed to have a free-stall system.
A questionnaire was sent to 2,400 farmers during March 2006. Of these 1,600 were confirmed as having free-stall barns and 800, with larger milk quota (more than 100,000 litres), where type of housing had not been confirmed. The purpose was to verify type of housing (by the farmer) and to obtain information on year of building, housing design and management in free-stall systems as basis for selection of farms for further studies. Questions were of the multiple-choice type, and alternative answers were pre-coded. Completed questionnaire was returned by 1,323 farmers (response rate of 55% after one reminder). Of these were 1036 farms with free-stalls and 287 with tie-stalls.
Comparison between free-stalls and tie-stalls at the herd level was based on NDHRS averages for the calendar year 2005. Studies at the individual cow level were performed using NDHRS data for 2005 and 2006. This material included cows that had terminated lactation during 2006. Cows with lactation periods of more than 483 days were excluded (2.5% of the lactations). Cows from breeds other than Norwegian Red were also excluded.
Data extraction from NDHRS
The following data were extracted from NDHRS at herd level: Milk production per cow-year (estimated at mean production per 365 days), fertility index, herds mean calving interval, the herds annual geometric mean bulk tank milk somatic cell count (BMSCC).
The fertility index is defines as:
a = non return after 60 days (%) + % of 2 or 3 inseminations in same oestrus period,
b = number of services per inseminated cow or heifer,
c = average number of days between calving and last insemination,
d = number of cows culled because of infertility,
e = total number of inseminated cows or heifers.
Disease data at herd level were extracted from available calculated herd incidence rates estimated as cows having at least one event of disease or other events divided by number of cow years at risk, expressed as incidence rate per 100 cow-years. Separate incidences were available for removals, replacements, severe/moderate clinical mastitis, teat injuries, ketosis, milk fever, indigestion, retained placenta, abortion, anoestrus, metritis, cystic ovaries and all diseases.
At individual cow level, parity and calving interval for lactations that ended in removal or new calvings in 2006, altogether 26,276 lactations, was extracted. Furthermore test-days data regarding kg milk and information whether the cow was diseased or not on test-day were extracted. After removing all cows with other breeds than NRF, the data set was reduced to 25,898 lactations, of which 38% were cows in first parity, 26% in second, 18% in third and 19% in higher parities. Mean lactation length was 293 days (median 342 days, 25% with less than 223 days and 10% with less than 78 days and range 1-483 days). Of the included lactations, 14.5% were second lactation within the same cow. The test-day data which were extracted from these lactations included 191,046 observations from 21,235 cows. After removing test-say yields lower than 5 kg and higher than 60 kg, the final material included 185,272 observations.
Herd level models
X consisted of the independent variables stall type (1 = tie-stall, 2 = free-stall), herds size (continuous), the interaction between stall type and herd size and the year of building/reconstruction (continuous).
Differences between herd disease incidence rates were tested by means of general modelling (Proc GenMod in SAS) using the log link function and poisson or negative binomial distribution according to the best fit of the model assessed by deviance. Number of disease events in each herd were estimated from the given incidence rate and herd size (number of cow-years). Logarithm (Ln) of mean number of cows in the herd during one year was used as offset variable. If herd size was significantly associated with the outcome the model was adjusted for herd size. The model fit was evaluated by log likelihood or deviance. Independent variables were number of cases with the diseases mastitis, severe/moderate clinical mastitis, teat injuries, ketosis, milk fever, indigestion, retained placenta, abortion, anoestrus, metritis, cystic ovaries and all diseases. The model was E(Y) = nλ, where λ is a function of the predictors. Ln(λ) = β0 + β1 X where variance = μ+ αμ2 with a poisson distribution α = 1. X consisted of the independent variables stall type (1 = tie-stall, 2 = free-stall), herds size (continuous), the interaction between stall type and herd size or the year of building/reconstruction (continuous).
In all models at herd level non-significant variables were removed one by one, first the interaction between herd size and stall type and thereafter herd size or building year if non-significant. Significance was set at p ≤ 0.05, but the variable was included in the model if p < 0.10.
Test-day level model with milk yield as dependent variable
where DIM = days in milk, X = represent all other fixed affects type of stall (tie- = 1 or free-stall = 2), herd size (continuous), calving interval and interaction by herd size and type of stall, herd size and DIM, type of stall and DIM and calving interval and DIM applied, Zh = the herd random effect, Zl = the lactation repeated effect within herd using AR(1) correlation matrix, and e = random error effect. As this model has well known interaction between parity and DIM and lnDIM, the model was stratified on parity both to avoid random effect at individual and the interaction terms (parity and DIM) which would make the model too complicated to deduce.
Pairwise analyses by means of the analysis of variance showed that milk production per cow-year was 134 kg lower in free-stalls than in tie-stalls, 6943 vs. 7077 kg (p < 0.05). In free-stalls FS-index was higher than in tie-stalls (78.9 vs. 57.2, p < 0.001), calving interval was lower (369 vs. 384 days, p < 0.001), and the incidence rate of the following diseases were also lower: teat injuries (1.08 vs. 2.12, p < 0.001), ketosis (1.30 vs. 3.39, p < 0.001), indigestions (0.66 vs. 1.46, p < 0.001) and anoestrus (1.19 vs. 2.28, p < 0.002). No difference between housing systems was found for among others mastitis (all cases) all diseases and BMSCC.
Model estimates (with standard error in brackets) from the general linear (GLM) model estimates in 620 free-stalls and 192 tie-stalls using Proc GLM in SAS.
Interaction tie-stall herd size
Year of building
Herd size in 2nd power
Milk per cow-year (kg)
FS status (FS-index)
Calving interval (days)
Model estimates (with standard error in brackets) from the poisson or negative binomial (NB) model estimates in 620 free-stalls and 192 tie-stalls cows using Proc Genmod in SAS.
Interaction tie-stall herd size
Dispersion (If NB fit)
All cases of mastitis
Overview of significant associations between performance/disease incidences and type of housing/herd size (20 and 50 cows) as shown in Table 1 and 2 and Fig. 2 and 3.
Milk per cow-year
Fertility status (FS-index)
Mastitis, all cases
Model estimates (with standard error in brackets) from the mixed models (ProcMixed in SAS) regarding test-day milk yield for cows in parity 1 (73,878 observations) and parity 3 (31,259 observations).
Estimate (β), parity 1
Estimate (β), parity 3
Dim (days in milk)
CI (calving interval)
Dim* cows (herd no of cow-years)
The two herd categories differed with respect to year of building/renovation in that tie-stall buildings were in average 12 years older than the free-stall buildings. Four of the tie-stalls and 33 of the free-stalls were built in 2004 and were thus in a transition phase that year with respect to performance and health. We therefore used NDHRS data for 2005/2006 in the analyses. Other studies [5, 17] have shown that cows changing from one system to another produced less milk the first few months after the transfer, but recovered after a full year. Automatic milking may be a confounding factor as it has been associated with a higher risk for high SCC , but in our material only 16 of the farms had automatic milking.
Usually herd estimates of cow performance are based on milk production per cow-year or disease incidences per cow-year as measures of intensity of milk production and disease occurrence at herd level. These measures are used as descriptive statistics, but they could be biased owing to different combinations of parities and different combinations of days in milk. This again could be due to different removal strategies in different herds. When analyzing the association between milk yield and housing system/herd size we therefore used two models. In addition to the herd level general linear model we also used individual cow test-day mixed models which also included calving interval in different parities. The results from these models confirmed the herd level result in that milk yields were significantly associated with housing system and herd size.
The finding that milk production per cow-year was lower in free-stalls is in agreement with results from previous studies in Norway [3, 4] and Finland , whereas no difference was found between tie-stalls and loose housing in a large-scale study in Denmark . Our result could indicate that free-stall barns have a less optimal function in smaller herds as other factors appear to intervene. Cattle are social animals and readily form dominance hierarchies, especially at areas of access to feed and water and the best resting area . In general, changes in social strategies are related to variations in group size . Sheep in larger groups spent less time queuing at the feed barrier , and in weaned pigs there were significantly more fights in smaller groups . Among dairy calves increased group size reduced conflicts, and calves in the larger groups spent more time feeding . Possibly similar interactions between group size and social behaviour also take place among adult cattle. The difference we found in the shape of the lactation curves, i.e. that yields were lowest among first parity cows primarily in 20-cow herds (Fig. 5) may indicate that these cows are lower in social rank and that this has larger consequences in smaller herds. The design of the cow barns may also be a factor having effect on cow behaviour, and suboptimal designs (area per cow, blind alleys, access to drinking water etc.) may have more unfavourable consequences in small free-stalls. Several of the associations recorded regarding performance and disease incidence were less favourable in the 20-cow herds as compared to the 50-cow herds, and this also indicates that small free-stalls are functioning less optimal.
The higher reproductive performance (higher fertility index, lower calving interval and lower treatments rates of anoestrus and cystic ovaries) which was found in free-stalls in this study is in agreement with previous results [2, 3, 9], and may possibly be related to easier oestrus detection when the cows are allowed to move freely. However, earlier oestrus detection and higher conception rates at first service have been observed in tie-stall barns as compared to free-stalls .
No significant difference between free-stalls and tie-stalls was found in average herd incidence rate of mastitis and BMSCC. This is in agreement with previous findings [3, 4, 7, 25]. In another study, cows in tie-stalls had lower cell counts, but the proportion of cows having new high SCC for the first time and the incidence rate of mastitis did not differ between housing systems . Lower mastitis incidence rate in free-stalls was found but no difference in somatic cell counts [2, 6].
Teat injuries were found to be less frequent in free-stalls than in tie-stalls, and this is in accordance with previous results [4, 7, 8] Other studies have shown poorer claw health in free-stalls [10–12], but the present material did not provide sufficient information on claw health to make a comparison between housing systems possible.
The incidence rate of ketosis was 2.6 times higher in tie-stalls than in free-stalls, and this result is in agreement with previous findings [2–4, 9]. The main risk factors for ketosis are related to nutrition and feeding. We found that the incidence of indigestions was higher in tie-stalls and that the shape of the lactation curve differed between herd categories, particularly in the first and last part of lactation for cows in parity 3 in the 20-cow herds. These results indicate that feeding strategies were different in the two housing categories.
Several of the performance and health related factors were more favourable in free-stalls than in tie-stalls, and more so in estimated 50-cow herds as compared to 20-cow herds. Milk yields were increasingly lower in free-stall herds with less than 27 cows. Our study has shown that performance and health is not universally better in small free-stalls than in tie-stalls.
The authors would like to thank the participating farmers for their help and support during the study. The access to the data was given by the Norwegian Dairy Herd Recording System (NDHRS) and the Norwegian Cattle Health Services (NCHS) (for health data) in agreement 3/2006. The study was financially supported by grants from The Research Council of Norway (15%), Agricultural Agreement Research Fund (16%) and Foundation for Research Levy on Agricultural Products (69%).
- Scientific Option of the Panel on Animal Health and welfare on request from European Commission on welfare of dairy cows. The EFSA Journ. 2009, 1143: 1-38.Google Scholar
- Valde JP, Hird DW, Thurmond MC, Østerås O: Comparison of ketosis, clinical mastitis, somatic cell count, and reproductive performance between free stall and tie stall barns in Norwegian dairy herds with automatic feeding. Acta Vet Scand. 1997, 38: 181-192.PubMedGoogle Scholar
- Bakken G, Røn I, Østerås O: Clinical disease in dairy cows in relation to housing systems. Proceedings VIth Int Congress on Animal Hygiene: 14-17 June 1988; Skara. 1988, 18-22.Google Scholar
- Østerås O: Sykdomsforekomst hos kyr i båsfjøs og løsdriftsfjøs (Disease incidence in cows in tie-stalls and free-stalls). Husdyrforsøksmøtet 1990; Ås. Aktuelt fra Statens fagtjeneste for landbruket nr. 4. 1990, 232-237.Google Scholar
- Hovinen M, Rasmussen MD, Pÿrälä S: Udder health of cows changing from tie stalls to free stalls with conventional milking to free stalls either with connventional or automatic milking. J Dairy Sci. 2009, 92: 3696-3703. 10.3168/jds.2008-1962.View ArticlePubMedGoogle Scholar
- Bakken G: Environment and bovine udder diseases in the loose housing system for dairy cows with reference to relevant data from the cowhouse system. Acta Agric Scand. 1981, 31: 445-451. 10.1080/00015128109435726.View ArticleGoogle Scholar
- Østerås O, Vågsholm I, Lund A: Teat lesions with reference to housing and milking management. J Vet Med A. 1990, 37: 520-524.View ArticleGoogle Scholar
- Regula G, Danuser J, Spycher B, Wechler B: Health and welfare of dairy cows in different husbandry systems in Switzerland. Prev Vet Med. 2004, 66: 247-264. 10.1016/j.prevetmed.2004.09.004.View ArticlePubMedGoogle Scholar
- Ekesbo I: Disease incidence in tied and loose housed dairy cattle. Thesis. Acta Agric Scand. 1966, Suppl 15Google Scholar
- Cook NB: Prevalence of lameness among dairy cattle in Wisconsin as a function of housing type and stall surface. J Am Vet Med Ass. 2003, 223: 1324-1328. 10.2460/javma.2003.223.1324.View ArticleGoogle Scholar
- Sogstad ÅM, Fjeldaas T, Østerås O, Forshell K: Prevalence of claw lesions in Norwegian dairy cattle housed in tie stalls and free stalls. Prev Vet Med. 2005, 70: 191-209. 10.1016/j.prevetmed.2005.03.005.View ArticlePubMedGoogle Scholar
- Cramer G, Lissemore KD, Guard CL, Leslie KE, Kelton DF: Herd- and cow-level prevalence of foot lesions in Ontario dairy cattle. J Dairy Sci. 2008, 91: 3888-3895. 10.3168/jds.2008-1135.View ArticlePubMedGoogle Scholar
- Bielfeldt C, Badertscher R, Tölle KH, Krieter J: Factors influencing somatic cell score in Swiss dairy production systems. Schweizer Arc Tierheilk. 2004, 146: 555-560. 10.1024/0036-72220.127.116.115.View ArticleGoogle Scholar
- Hultgren J: Foot/leg and udder health in relation to housing changes in Swedish dairy herds. Prev Vet Med. 2002, 53: 167-189. 10.1016/S0167-5877(01)00279-3.View ArticlePubMedGoogle Scholar
- Østerås O, Solbu H, Refsdal AO, Roalkvam T, Filseth O, Minsaas A: Results and evaluation of thirty years of health recordings in the Norwegian dairy cattle population. J Dairy Sci. 2007, 90: 4483-4497. 10.3168/jds.2007-0030.View ArticlePubMedGoogle Scholar
- Wilmink JBM: Comparison of different methods of predicting 305-day milk yield using means calculated from within-herd lactation curves. Livestock Prod Sci. 17: 1-17. 10.1016/0301-6226(87)90049-2.Google Scholar
- Norell RJ, Appleman RD: Change of milk production with housing system and herd expansion. J Dairy Sci. 64: 1749-1755.Google Scholar
- Mark T, Lassen J: Genotype by housing interaction in Danish hostein. Proc Interbull Meeting Dublin, Ireland 23.08.2007-26.08.2007. (bulletin no 37)Google Scholar
- Friend TH, Polan CE, McGillard ML: Free stall and feed bunk requirements relative to behavior, production and individual feed intake in dairy cows. J Dairy Sci. 1977, 60: 108-116.View ArticleGoogle Scholar
- Estevez I, Andersen IL, Nævdal E: Group size, density and social dynamics in farm animals. Appl Anim Behavi Sci. 2007, 103: 185-204. 10.1016/j.applanim.2006.05.025.View ArticleGoogle Scholar
- Jørgensen Meisfjord GH, Andersen IL, Berg S, Bøe KE: Feeding, resting and social behaviour in ewes housed in two different group sizes. Appl Anim Behav Sci. 2009, 116: 198-203. 10.1016/j.applanim.2008.08.014.View ArticleGoogle Scholar
- Andersen IL, Nævdal E, Bakken M, Bøe KE: Aggression and group size in domesticated pigs, Sus scrofa: 'when the winner takes it all and the loser is standing small'. Anim Behav. 2004, 68: 965-975. 10.1016/j.anbehav.2003.12.016.View ArticleGoogle Scholar
- Færevik G, Andersen IL, Jensen MB, Bøe KE: Increased group size reduces conflicts and strengthens the preference for familiar group mates after regrouping of weaned dairy calves (Bos taurus). Appl Anim Behav Sci. 2007, 108: 215-228. 10.1016/j.applanim.2007.01.010.View ArticleGoogle Scholar
- Hackett AJ, Batra TR, McAllister AJ: Estrus detection and subsequent reproduction in dairy cows continuously housed indoors. J Dairy Sci. 1984, 67: 2446-2451.View ArticlePubMedGoogle Scholar
- Østerås O, Walland F, Lunder T: Løsdrift i mjølkeproduksjonen, lønner det seg? (Free-stalls in dairy production, is it profitable?). Husdyrforsøksmøtet 1994; Ås. Norges landbrukshøgskole-Fagtjenesten, Faginfo nr 6. 1994, 185-191.Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.