Effects of different nesting materials on crushing risk and sow posture changes in a farrowing system with temporary confinement

The provision of nesting material around farrowing is considered an important measure to support sow welfare, yet its practical implementation under commercial farrowing conditions remains limited. This study evaluated the effects of different nesting materials on piglet crushing risk and sow activity in a commercial farrowing system with temporary crating for approximately 9 to 10 days around farrowing. A total of 263 farrowings were assigned to one of three material treatments: hay, straw, and a jute bag as minimal enrichment control. Nesting material use, farrowing performance, and crushing losses were recorded. In addition, sow activity during the first 72 h postpartum was quantified using a computer vision-based approach. Although substantial inter-individual variation was observed, the results show that nesting material type had no significant effect on material use. During the first three days postpartum, the odds of crushing were 46 % lower in the straw group compared to the jute bag control (p < 0.05), whereas no significant difference was detected between hay and the control. No significant treatment effects were observed for crushing losses at the end of lactation. Sows provided with straw showed approximately 28 % fewer posture changes than control sows (p < 0.05), and odds of crushing increased by approximately 8 % for every 10 posture changes (p < 0.05). These findings indicate that not only the presence but also the functional properties of nesting material are relevant for piglet protection. The provision of structured materials such as long-stemmed straw represents a practical management measure to improve welfare outcomes under commercially relevant conditions and may contribute to more stable and less restless sow behavior during the critical periparturient period, even in systems with temporary confinement.

In recent years, increasing societal and political attention has been directed toward improving animal welfare in pig production, particularly with regard to housing systems that allow greater freedom of movement for sows (Clark et al., 2016; Vandresen and Hötzel, 2021a, 2021b; Wallenbeck et al., 2024). In several countries, regulatory frameworks have been revised to reduce the duration of confinement or to promote alternative farrowing systems (Wallenbeck et al., 2024). In Germany, recent amendments to the Animal Husbandry Regulations have introduced new requirements for farrowing housing, including restrictions on the use of farrowing crates and transitional implementation periods (Federal Republic of Germany, 2021). These developments reflect a broader international trend toward welfare-oriented housing concepts. At the same time, practical experience indicates that transitioning toward systems with increased sow mobility poses considerable challenges for pig producers (Baxter et al., 2025). A survey among German farmers identified the farrowing area as one of the most critical production stages during this transition (Geisler et al., 2025). Within this context, the provision of organic nesting material has been discussed as a potentially feasible measure to support the welfare of both sows and the piglets while remaining compatible with commercial production systems (Wischner et al., 2009a; Yun and Valros, 2015; Rosvold et al., 2019).

Nest-building behavior in sows has been extensively studied and is considered a fundamental component of the periparturient behavioral repertoire (Monteiro et al., 2023). Under semi-natural conditions, nest-building typically begins within 48 h prior to farrowing, with peak activity occurring during the final 24 h before birth (Jensen, 1986). The behavior unfolds in distinct phases and is characterized by increased activity, including rooting, pawing, and the collection and manipulation of nesting material (Jensen, 1993). The onset and intensity of nest-building behavior are primarily hormonally regulated, whereas its expression and termination are modulated by internal and external cues associated with nest completion (Algers and Uvnäs-Moberg, 2007). This behavioral phase plays an essential role in preparing a suitable farrowing environment and may contribute to the development of maternal behavior (Thodberg et al., 1999). Despite intensive domestication and selection for production traits, domestic sows retain a strong motivation to perform nest-building behavior, underscoring its biological significance (Wischner et al., 2009a).

When sows are hindered from expressing nest-building behavior, behavioral alterations and signs of impaired welfare have been consistently reported (Jensen, 1993; Damm et al., 2003; Yun and Valros, 2015). In environments lacking suitable nesting material or opportunities for manipulation, sows may redirect their motivation toward non-functional or repetitive activities (Wischner et al., 2009a). These behaviors include stereotypies such as bar biting, pawing, or excessive rooting at pen fixtures, which are commonly interpreted as indicators of frustration (Jensen, 1988; Yun and Valros, 2015; Plush et al., 2024). In addition to behavioral changes, restricted nest-building has been associated with elevated stress responses around farrowing (Yun et al., 2013). Increased restlessness, prolonged activity, and altered lying behavior have been observed when sows are unable to perform prepartum nest-building (Thodberg et al., 2002). Physiological indicators, including changes in stress-related hormones, further support the conclusion that the inability to express this highly motivated behavior compromises sow welfare during the periparturient period (Jarvis et al., 2001).

When sows are provided with adequate space and suitable nesting material that allows the expression of nest-building behavior, positive effects have been observed not only for sow welfare but also indirectly for early piglet survival (Yun and Valros, 2015). Expression of nest-building behavior has been associated with improved maternal behavior around farrowing, including enhanced responsiveness toward piglets (Cronin and van Amerongen, 1991). In particular, sows that are able to perform nest-building behavior have been reported to respond more promptly to piglet distress calls, potentially reducing the likelihood of piglets being crushed (Herskin et al., 1998; Yun et al., 2013). Piglet crushing occurs predominantly within the first 72 hours after farrowing and represents a major cause of pre-weaning mortality (Nicolaisen et al., 2019). A frequent underlying mechanism is the change of body position by the sow, particularly during lying down and rolling from one side to the other while already lying (Marchant et al., 2001; Damm et al., 2005). Calmer sows with more coordinated maternal behavior and improved responsiveness to piglet vocalizations may therefore contribute to a reduced crushing risk, especially during the early postpartum period when piglets are most vulnerable.

Although numerous studies have demonstrated beneficial effects of nesting material provision, most investigations have focused on the presence versus absence of enrichment rather than comparing functionally different materials under practical housing conditions. In addition, evidence regarding the use of grass hay as nesting material remains limited, despite its potential relevance for commercial systems. Furthermore, indicators related to sow activity and crushing risk have traditionally been assessed through manual observation, which restricts scalability and standardization (Chen et al., 2021). Advances in computer vision offer opportunities to automate the quantification of posture-related activity measures, and previous studies have successfully applied such approaches to detect sow postures, posture changes, and other activity-related behaviors (Zheng et al., 2020; Wang et al., 2021; Yang et al., 2021).

Despite the growing body of evidence supporting nesting material provision, the implementation of natural organic substrates in commercial farrowing systems remains limited. Practical guidance on suitable material types and quantities under commercial conditions is still scarce. Generating applied evidence that reflects actual housing environments is therefore essential to support informed management decisions.

Therefore, the overarching aim of the present study was to generate more practical evidence on the provision of organic nesting material under commercially relevant housing conditions that comply with current regulatory requirements. Specifically, the study evaluated (1) whether pre-partum material use differed between sows provided with grass hay or long-stemmed straw, (2) whether the provision of grass hay or long-stemmed straw, compared to a jute bag control, affected piglet crushing losses during the first three days postpartum and until weaning, and (3) whether the provision of nesting materials altered the frequency of sow postural changes during the first 72 h postpartum. For the analysis of sow posture changes, an automated quantification was implemented to enable more standardized and reproducible data generation.

Farm description and housing conditions

The study was conducted on a commercial sow farm in north-western Germany. The herd consisted of approximately 150 sows (Suisag genetics) managed in a five-week batch-farrowing system with a four-week lactation period. Farrowing took place in a mechanically ventilated barn containing 64 pens arranged in a single open room without subdivision into compartments. Sows were housed in Schauer BeFree movement pens (Schauer Agrotronic GmbH, Austria) that allow temporary confinement (Figure 1). Each pen provided 7 m² of total floor area, including approximately 5 m² accessible to the sow when the crate was open. Pens were equipped with a closed and heated creep area for piglets. The flooring was partially slatted, and manure was removed by an underfloor scraper system.

According to routine farm procedures, sows were confined in the crate from two days before the expected farrowing date until about one week postpartum, resulting in an overall confinement period of approximately 9 to 10 days, depending on the actual farrowing date. Although longer than the maximum duration permitted under future German legislation, this confinement period reflected the farm’s standard management. Around farrowing, pens were cleaned manually once daily. During the first postpartum days, sows were fed twice daily, and piglets were temporarily confined in the creep area during feeding. Piglets were released after the sow had completed feeding and settled into a lying position to reduce the risk of crushing. In addition, routine cross-fostering was performed by the farm manager after farrowing to homogenize litters and to adjust the number of piglets per sow based on previous rearing performance and sow-specific characteristics, including previous litter performance, and teat number and quality. This procedure was part of the farm’s standard management and was not influenced by the experimental treatments.

Experimental design and treatments

In this study, two different nesting materials were used and compared with a control group. As nesting materials, hay (Lolium perenne) and straw (barley, long-stemmed) were selected because they are cost-effective, readily available, and compliant with future German regulations. A jute bag was used as a control treatment, as it represents the minimum enrichment standard currently required under German legislation.

Over a one-year data collection period (October 2022 to September 2023), complete data from a total of 221 farrowings of 141 individual sows were recorded in 11 batches. The data collection included both gilts and multiparous sows, with 13 first farrowings by gilts and the remaining farrowings from multiparous sows. Of all farrowings, 76 were assigned to the hay treatment, 71 to the straw treatment, and 74 to the jute bag control group. If sows were participating in the study for the first time, treatments were allocated to them randomly and evenly. In cases where sows participated in the experiment more than once, they were consistently assigned to the same treatment.

Hay and straw were provided to the sows in racks that were installed in all farrowing pens at the beginning of the study (Figure 2). The racks were positioned in a way that allowed access to the material even when the crate was closed. When selecting the rack design, particular attention was paid to large openings to ensure easy accessibility of the material for the sows. Each morning, the racks were filled with 120 g of material (fresh weight). If more than two thirds of the material had been removed by the afternoon, an additional 120 g were provided. The standardized amount of 120 g per provision was determined based on the capacity of the racks, allowing the material to remain sufficiently loose for easy manipulation and withdrawal by the sows.

In the control group, jute bags were attached to the lower bar of the farrowing crate to prevent them from being displaced outside the sow’s reach (Figure 2). Once a bag was no longer manipulable due to extensive wear and tearing, it was replaced with a new one. All materials were provided from day 112 of gestation onwards to ensure that nesting material was available at the onset of nest-building behavior. Sows had continuous ad libitum access to the provided material. After parturition, no further replenishment of the materials was carried out.

Data collection

Data collection comprised the documentation of nesting material use, farrowing performance, pre-weaning piglet mortality, and video recordings of individual sows. Sows showing signs of lameness were excluded from the study. The use of nesting material under the applied provision regime was recorded daily until the onset of farrowing. Each morning, remaining material was removed from the racks and weighed before refilling. Based on the amount of material provided and the remaining quantity, the daily material use per sow for both treatments was calculated. No corresponding data was recorded for the jute bag control.

After farrowing, the number of live-born and stillborn piglets was documented for each litter. Piglet losses attributed to crushing were recorded daily during the first three days postpartum and at the end of the lactation period. The diagnosis of crushing losses was based on the position and external appearance of the carcass, including signs of traumatic injury such as bruising or visible impressions of the slatted floor on the piglet’s body (Nicolaisen et al., 2019). Piglets that did not show typical signs of crushing were classified as having died from other causes. As routine cross-fostering was occasionally performed as part of standard farm management and could also occur days postpartum, minor deviations in litter size during the early postpartum period cannot be fully excluded. Because this procedure was not fully documented on the farm, crushing losses were calculated relative to the number of live-born piglets per litter.

From the 221 farrowings, 133 were video recorded, starting two days before the expected birth date and continuing until three days postpartum. Here, cameras of type Reolink RLC-520A network cameras (Reolink Innovation Limited, Hong Kong) were installed at a height of approximately 3.2 m above the pens, positioned centrally to obtain a top-down perspective. Recordings were conducted at a resolution of 2560 x 1920 pixels with a frame rate of 30 frames per second. Infrared illumination was used for night-time recordings to allow continuous monitoring of sow behavior.

Video-based analysis of posture changes

The proposed approach in Wahmhoff et al. (2025) was applied to determine the number of posture changes as an indicator of restlessness and general activity. Following previous studies, which demonstrated the beneficial use of combining independent predictors to improve overall prediction performance (Marathe et al. 2021; Walambe et al., 2021), we further extended the method and applied a consensus-based approach using two independent YOLOv11 models (Jocher et al., 2023) to quantify sow-specific posture changes. In detail, two well-known computer vision tasks in the form of object detection (OD) and image classification (IC) have been applied.

In the first branch, an OD model was used to locate the sow using a bounding box annotation style and to determine the respective posture directly within the video frame. Since previous studies showed that the performance of a single OD approach can be influenced by the respective surrounding and the existence of non-relevant features (Tian et al., 2021), a second, independent extraction pipeline was used to first crop the detected bounding box and separately classify the reduced representation into a target posture class. Here, both branches have been trained to recognize four distinct body posture classes (standing, sitting, lying, and lateral lying; Figure 3). To reduce complexity, for this study these four classes have been aggregated into two functional states. Standing and sitting were combined into the state active, whereas lying and lateral lying were combined into the state resting. A posture change was defined as a transition between these two states.

The training was carried out on a workstation equipped with an AMD Ryzen Threadripper PRO 5945 WX CPU, 512 GB RAM, and an NVIDIA RTX A6000 GPU. All models were trained with a batch size of 16 images per iteration for 250 epochs. The implementation was carried out in Python (version 3.9) using the PyTorch deep learning framework (version 2.0.1), running on an Ubuntu 22.04.3 environment. During training, data augmentation techniques, including geometric transformation and color-space operations, were applied using the default augmentation parameters provided in the Ultralytics documentation (Jocher et al., 2023).

Both the OD and IC models were trained on 4,800 annotated images each using independent datasets. Images for the OD dataset were manually extracted from the video recordings. For the training of the IC model, all images from environment A of the publicly available dataset SowPostureDS (Wahmhoff et al., 2026) were used. Each dataset was balanced across the four posture classes. The data were split into training, validation, and test sets using a ratio of 7:2:1.

To improve computational efficiency, one frame per second was evaluated. Only frames for which both models produced the same posture prediction were retained for further processing. Prior to posture change determination, short-term fluctuations were filtered by retaining only posture states that persisted for at least 5 s, ensuring that transient misclassifications did not influence the analysis. Subsequently, transitions between posture states were identified and quantified to determine the total number of posture changes per sow.

The performance of this automated pipeline was evaluated using 500 randomly selected and manually annotated reference videos. For posture prediction, precision, recall, and F1 scores for both branches ranged between 0.89 and 0.91. The consensus-based evaluation resulted in improved prediction performance scores between 0.92 and 0.93. The accuracy of the posture change estimation was further assessed by calculating the mean absolute error (MAE) between manually counted and automatically detected posture changes for all 500 videos, which was 0.44 for the consensus-based approach. In contrast, MAE for the individual branches was 5.19 or higher, underlining the considerable improvement achieved through the consensus-based approach.

Video analysis for activity assessment was restricted to the straw and jute bag treatments, as a significant effect on crushing losses was only observed within these groups. From the 85 recorded farrowings belonging to the straw (n = 44) and jute bag (n = 41) treatments, preselection criteria were applied to obtain comparable subsamples. These criteria included a litter size of 10–20 live-born piglets and a parity between 2 and 5. After applying these, 30 farrowings remained in the straw group and 33 in the jute bag group. From these, 25 farrowings per treatment were randomly selected for further analysis due to technical and practical constraints. Posture changes were quantified over the first 72 hours following the birth of the last piglet.

Statistical analysis

Statistical analyses were conducted using mixed modeling approaches appropriate to the distribution of each response variable in R (version 4.4.2; R Core Team, 2024). Pre-parturition material use (continuous) was analyzed using a linear mixed-effects model assuming a Gaussian error distribution. Normality and homoscedasticity of residuals were assessed using residual diagnostics and Q–Q plots.

Crushing losses were modeled using a binomial generalized linear mixed model (GLMM) with a logit link function to appropriately model proportional event probabilities. The number of crushed piglets was modeled relative to the number of live-born piglets per litter. Postural changes were analyzed as count data describing the occurrence of discrete behavioral events within a defined observation period. Therefore, Poisson GLMM with a log link function was applied to model event rates while accounting for the mean-variance relationship inherent to count data. To explore a potential behavioral mechanism underlying differences in crushing losses, an additional model was fitted in which crushing losses within 72 h postpartum were analyzed using the same binomial GLMM structure described above but including the number of postural changes as an additional fixed effect alongside material and the other covariates. The number of postural changes was scaled (per 10 events) to facilitate interpretation of the estimated effect sizes. This model allowed assessment of whether the association between nesting material and crushing losses persisted after accounting for sow movement behavior.

Comparisons were defined according to the study hypotheses: Material use was compared between hay and straw; crushing losses for the hay and straw treatment were analyzed relative to the jute bag control; and postural changes were evaluated only between straw and the jute bag treatment. The hay group was not included in the postural change analysis, as the primary behavioral hypothesis focused on the material associated with reduced crushing losses. Because postural changes were quantified only for a subset of sows, analyses involving postural changes were conducted on this subsample. Consequently, the model assessing the association between postural changes and crushing losses was also restricted to this subset and therefore included only straw treatment and jute bag control.

For all models, nesting material was included as the primary fixed effect of interest. Parity class (1, 2, 3, 4, 5+) and experimental batch (2, 3, 4, 5, 6, 7, 8, 9, 10, 11) were incorporated as additional fixed effects to control for biological and management-related variation. In the models assessing crushing losses, the number of stillborn piglets was additionally included as a fixed effect to account for potential litter-related influences on piglet mortality. Sow identity was fitted as a random intercept to account for repeated observations within animals, and pen was included as an additional random effect where applicable to account for environmental variability.

Model assumptions were evaluated using residual diagnostics, and dispersion tests were applied to generalized models to confirm appropriate model fit. Potential interaction effects between fixed factors were tested but were not retained in the final models if they did not reach statistical significance and did not improve AIC and BIC substantially. Optimizer settings were adjusted where necessary to ensure numerical convergence. Statistical significance was declared at p < 0.05.

Ethics statement

The animal study was approved by the Animal Welfare Officer of the Georg-August-University Göttingen, Germany. The study was conducted in accordance with the local legislation and institutional requirements.

Descriptive statistics of the study population

Descriptive statistics of the study population are summarized in Table 1. The values for parity, gestation length, number of live-born piglets, and number of stillborn piglets were comparable across the three treatment groups. The distribution across parity classes was uneven, with 13 observations in parity 1 and 50 observations in parity 2, while parities 3 to 6 comprised 30 to 48 observations each, parity 7 included 11 observations, and parity 8 only one observation.

Nesting material use

The total usage of nesting material from two days before farrowing until the onset of farrowing is presented in Figure 4. On average 417 g hay and 354 g straw have been applied. In both treatments, a wide variation in individual use can be seen, ranging from sows that hardly used the provided material to sows that almost completely removed the offered amount from the rack. Across the 441 daily observations from two days before farrowing until farrowing, the full daily amount of 240 g was removed in 68 observations (15.42%).

Although Figure 4 shows a tendency toward lower use in the straw group, statistical analysis indicated that material type had no significant effect on material use (p > 0.05). Parity class showed a weak influence on material use, with sows in parity class 2 consuming significantly less material compared to parity class 1 (–128.4 g ± 57.2, p < 0.05). No significant differences were detected for higher parity classes. The experimental batch had no detectable effect on material consumption (p > 0.05). Differences between individual sows accounted for the largest share of the total variation in material use.

Crushing losses

Crushing losses are summarized in Figure 5. The percentage of crushed piglets differed between treatment groups both during the first three days postpartum and at the end of lactation. During the early postpartum period (0–3 days), the lowest proportion of crushing losses was observed in the straw group (3.51%), followed by the hay group (4.68%), whereas the highest proportion occurred in the jute bag group (6.52%). A similar pattern was observed for crushing losses at the end of the lactation (Straw: 6.09%, Hay: 6.85%, Jute bag: 8.23%). While 58% of crushing losses in the straw group occurred within the first three days postpartum, higher proportions were observed in the hay group (68%) and the jute bag group (79%).

Mixed model analysis for the first three days postpartum confirmed a significantly lower crushing risk in the straw group compared to the jute bag control (estimate = – 0.621 ± 0.279, p < 0.05), corresponding to an odds ratio of 0.54. Thus, the odds of crushing were 46% lower in the straw group compared to the jute bag control. No significant difference was detected between the hay and jute bag groups (p > 0.05). Parity class and batch also had no significant effects. The number of stillborn piglets was positively associated with crushing losses (estimate = 0.134 ± 0.063, p < 0.05), indicating that each additional stillborn piglet increased the odds of crushing by approximately 14%. At the end of lactation, no significant effect of material type on overall crushing was detected. Neither hay nor straw differed significantly from the jute bag group (p > 0.05). However, the number of stillborn piglets remained significantly associated with total crushing losses (estimate = 0.117 ± 0.054, p < 0.05), corresponding to an increase in crushing odds of approximately 12% per additional stillborn piglet. No significant effects of parity class or batch were observed in the final model (p > 0.05).

Posture changes and their relationship to crushing

The number of posture changes during the first 72 hours after birth of the last piglet differed between treatments (Figure 6). Sows in the jute bag group showed a higher number of posture changes compared to sows in the straw group. Median posture changes were 160 in the jute bag group and 128 in the straw group, respectively.

Results of the Poisson GLMM for posture changes confirmed a significant effect of nesting material on postural activity. Compared to the jute bag control, sows provided with straw showed a significantly lower number of posture changes (estimate = – 0.323 ± 0.099, p < 0.05). This corresponds to an incidence rate ratio of 0.72 for the 72-h observation period, indicating that straw provision was associated with approximately 28% fewer posture changes per sow during the first 72 h postpartum. Parity class did not significantly affect the number of postural changes (p > 0.05). The experimental batch showed a small but significant negative effect (estimate = –0.035 ± 0.014, p < 0.05), corresponding to a reduction of approximately 3.5% in posture changes per batch increment. The random effect of sow identity indicated additional inter-animal variation in postural activity.

In the binomial GLMM for crushing losses, including posture changes as an additional predictor, postural activity showed a significant positive association with crushing losses during the first 72 h postpartum (estimate = 0.080 ± 0.037, p < 0.05). When scaled to increments of 10 posture changes, this corresponds to an odds ratio of 1.08, indicating that every additional 10 posture changes were associated with an approximately 8% increase in odds of crushing. In this model, nesting material was no longer a significant predictor of crushing losses (p > 0.05), while the remaining covariates did not show significant effects.

Nesting material

In this study, hay and straw were selected as nesting materials. Previous research has explored a variety of substrates, including sand, peat, sawdust, branches or newspaper sheets as nesting or enrichment materials (Herskin et al., 1998; Damm et al., 2005; Yun et al., 2013; Rosvold et al., 2018; Hukkinen et al., 2024). The selection of hay and straw in this study was motivated by their high likelihood to be adopted under commercial conditions due to their wide availability, comparatively low costs, and practical suitability for use in systems with perforated or slatted floors, particularly when compared to substrates such as sawdust or peat. This selection is further supported by the European Food Safety Authority (EFSA), which identifies straw, hay and haylage as appropriate nesting materials based on their functional characteristics (Nielsen et al., 2022).

No significant differences in material use were observed between hay and straw in the two days before farrowing, although the hay group showed a numerical trend toward greater use. This tendency may be explained by the measurement approach, as material use was defined as the quantity removed from the rack. Due to its finer structure, hay may have dispersed more easily during manipulation, potentially leading to slightly higher recorded values without reflecting a biologically meaningful difference in behavioral engagement.

Compared with previous studies, nesting material use in the present study was substantially lower than the amounts provided in other experimental studies, where several kilograms of material were placed in the racks or directly into the pen (Jensen, 1993; Thodberg et al., 2002; Andersen et al., 2014; Illmann et al., 2015; Westin et al., 2015a). Moreover, the data for material use indicate that the smaller quantities of nesting materials provided under the conditions of this study were sufficient, as the offered amounts were not completely removed from the racks by the majority of sows. Given that the use of large amounts of nesting material in systems with fully or partly slatted floors may interfere with manure systems (Nielsen et al., 2022), the provision of comparatively small quantities can be regarded as a practical and feasible approach under commercial conditions.

Parity class affected material use only between first-parity sows and animals in their second parity. Previous studies conducted under semi-natural conditions have reported greater material collection with advancing parity (Jensen, 1989). Similarly, increased nest-building intensity in older sows has been observed in free farrowing systems (Rosvold et al., 2018) and has been associated with endocrine changes, including elevated prolactin concentrations prior to parturition (Yun et al., 2014). These patterns were not consistently supported by the present data. Compared with findings reported in earlier studies, first-parity sows in this study showed higher material use than animals in their second parity. However, this observed pattern may partly be influenced by the relatively small number of first-parity sows included in the dataset.

Furthermore, differences between individual sows accounted for the largest share of total variation in material use, indicating that sow-specific characteristics substantially influenced interaction with the provided nesting material. A comparable degree of inter-individual variability in prepartum nest-building behavior has also been reported by Westin et al. (2015b). Although parity effects were not consistently statistically supported in the present dataset, individual variation may partly reflect differences in parity or previous experience, which have previously been identified as important determinants of nest-building behavior intensity (Damm et al., 2003).

Crushing losses

Relative to earlier studies, overall piglet mortality on the study farm was relatively low. Even prior to the implementation of the experimental treatments, pre-weaning mortality amounted to 12.6% of live-born piglets, which is very comparable with values reported for farrowing systems with fully crated sows [13.0% (Marchant et al., 2000), 11.8% (Hales et al., 2014), and 12.3% (Nicolaisen et al., 2019)]. This indicates that the management measures routinely applied on the farm were already well established and effectively coordinated.

Despite the generally low level of piglet losses, significantly lower odds for crushing were observed in the straw treatment compared with jute bag control during the first three days postpartum. This finding is consistent with previous studies demonstrating that the provision of nesting material around parturition can improve piglet survival (Herskin et al., 1998; Bolhuis et al., 2018; Plush et al., 2021). However, an important difference between the present study and previous studies can be found in the experimental design. Earlier studies primarily compared the presence versus absence of nesting material, whereas in the current experiment the control group received a minimal enrichment in the form of a jute bag. The persistence of a treatment effect despite the presence of this baseline enrichment suggests that the mere provision of manipulable material is not sufficient. Rather, the functional characteristics of the material appear to determine whether the sow’s behavioral and physiological needs around parturition are adequately addressed. In line with the observation of Plush et al. (2021), who reported that although hessian stimulated nest-building behavior in crated sows, it did not yield the same improvements in farrowing outcomes as straw. Straw provision was associated with reduced indicators of sow distress and improved piglet survival, whereas hessian failed to elicit comparable effects. Together with the present results, this supports the notion that not all enrichment materials exert equivalent biological effects.

The absence of a significant effect for hay further emphasizes this point. Although both hay and straw are natural substrates, they differ in structural properties. Straw, due to its coarser and more persistent texture, may provide more pronounced tactile feedback during manipulation and remain accessible on slatted floors for a longer period. Such properties may enhance the effectiveness of the material in satisfying nest-building motivation, thereby indirectly supporting maternal behavior and reducing early crushing risk, as observed in other studies (Cronin and van Amerongen, 1991; Herskin et al., 1998; Yun and Valros, 2015). Evidence on hay as nesting material remains limited. Although positive effects have been reported for lucern hay (Edwards et al., 2019), its structural characteristics differ markedly from the grass hay used in the present study. Studies investigating the influence of material structure similarly suggest that coarser and more stable materials tend to exert stronger effects on maternal behavior and piglet survival (Damm et al., 2000; Rosvold and Andersen, 2019).

Taken together, these findings indicate a two-step differentiation: first, the provision of structured organic material appears superior to minimal enrichments; second, even among natural substrates, material-specific properties influence biological outcomes. Straw therefore seems more suitable than both jute bags and grass hay for supporting piglet survival under the present housing conditions.

The temporal distribution of crushing losses provides additional insights. In the jute bag group, the majority of crushing events occurred within the first three days postpartum, consistent with previous reports identifying this phase as the period of highest vulnerability (Nicolaisen et al., 2019; Rosvold and Andersen, 2019). In contrast, the proportion of early crushing losses was markedly lower in the straw group. Explanations occasionally discussed in the literature, such as insulating straw layers leading to piglets spending more time near the sow, resulting in higher crushing losses (Westin et al., 2015a), are unlikely to apply here, as nesting material was provided only until farrowing and pens were cleaned daily postpartum. A more plausible explanation may be the opening of the farrowing crate approximately one week after birth, which has been associated with increased crushing risk due to expanded sow movement (Chidgey et al., 2015; Lohmeier et al., 2020). Piglets particularly susceptible to crushing may therefore have survived the immediate postpartum phase in the straw group but were affected later after crate opening. This temporal shift could explain why no significant differences in overall crushing probability were detected at the end of lactation despite the early advantage observed with straw.

The positive association between the number of stillborn piglets and crushing probability has also been described in previous studies (Westin et al., 2015a; Rosvold and Andersen, 2019). Higher numbers of stillborn piglets are often linked to prolonged or more complicated farrowings, which may compromise piglet vitality and increase vulnerability during the early postpartum period. Reduced vitality may in turn elevate the likelihood of crushing, particularly when piglets respond more slowly to postural changes of the sow.

Routine cross-fostering was performed as part of standard farm management and may have influenced litter composition and size. As the exact timing of transfers was not possible to document in detail, crushing losses were calculated relative to the number of live-born piglets. This should be considered when interpreting treatment effects and represents a limitation of this study.

Postural changes

Behavioral indicators of sow activity around farrowing, such as posture changes, have traditionally been assessed through manual video evaluation. In recent years, computer vision-based approaches have increasingly been explored to automate the analysis of animal behavior. These methods offer clear advantages in terms of scalability and consistency, as they allow large volumes of video data to be processed in a standardized and objective manner. In the present study, an automated pipeline was implemented to automate and objectify the quantification of posture changes during the early postpartum period of sows confined in farrowing crates. By combining two independent prediction branches in a consensus-based approach, overall prediction performances could be improved in contrast to the individual model performances. This strategy increased the robustness of posture classification and enabled a more reliable and consistent identification of posture changes across animals and recordings, which is in line with other studies (Rosa et al., 2024).

A limitation of this analysis concerns the use of a subsample of farrowings. Due to restricted availability of suitable video recordings as well as practical constraints related to time and computational resources, posture changes could not be quantified for all experimental animals. Despite this constraint, the automated analysis provided valuable explanatory insights into differences observed in crushing risk between treatment groups and demonstrates the feasibility of efficiently generating standardized behavioral data under practical farm conditions.

The lower number of posture changes observed in the straw group compared with the jute bag group suggests that straw more effectively supported the expression of nest-building-related behaviors and contributed to a calmer behavioral state of the sow around parturition. Similar observations have been reported in previous studies, in which the provision of suitable nesting material was associated with reduced restlessness and fewer posture transitions during and shortly after farrowing (Martins et al., 2021; Monteiro et al., 2023). These effects have been linked to reduced stress and frustration when sows are able to express nest-building behavior (Plush et al., 2021), which may in turn promote more stable lying behavior during the periparturient period (Bolhuis et al., 2018).

The reduced frequency of posture changes in the straw group is likely to represent a key mechanism underlying the lower crushing risk observed during the early postpartum phase. Postural transitions, particularly lying-down movements and rolling, are considered the primary proximate cause of piglets crushing (Wischner et al., 2009b). A lower rate of such movements may therefore directly decrease the likelihood of piglets being trapped beneath the sow (Yun et al., 2019). In line with this interpretation, the statistical model including posture changes as an additional predictor revealed a significant positive association between sow movement activity and crushing losses, indicating that increased postural activity was associated with higher odds of piglets crushing. When posture changes were accounted for in the model, nesting material itself was no longer a significant predictor of crushing losses. This pattern suggests that the beneficial effect of straw on piglet survival during the early postpartum period may be mediated through its influence on sow movement behavior. However, crushing losses are multifactorial, and factors beyond sow movement behavior, including sow-related characteristics and environmental conditions such as temperature, have previously been associated with crushing risk (Damm et al., 2005; Nielsen et al., 2022). These factors should therefore also be considered when interpreting the observed treatment effects and may represent additional targets for reducing piglet crushing.

When drawing broader conclusions regarding sow welfare, it should also be considered that the behavioral analysis was restricted to the first 72 h postpartum and focused on posture changes. While this approach was appropriate for assessing sow postural activity in relation to piglet crushing risk, it does not capture broader aspects of maternal behavior and sow welfare or their development during later lactation. Future studies should therefore extend the observation period and include a more detailed classification of sow behavior to allow a more comprehensive assessment of sow welfare and maternal care.

Practical applicability

The findings presented in this study indicate that the provision of structurally suitable nesting material, particularly long-stemmed straw, can reduce crushing risk during the early postpartum period even under commercial housing conditions with temporary crate confinement. Thus, meaningful improvements in sow behavior and piglet survival appear achievable without fundamental structural changes to existing systems.

From a practical perspective, the use of racks proved compatible with pens equipped with slatted floors, and the moderate material quantities applied are unlikely to interfere with manure management. Ensuring continuous access to the rack, including during confinement, is essential. Refilling can be integrated into routine management procedures, resulting in only limited additional labor input. Given the relatively low cost of straw compared to the economic value of piglets, even small improvements in survival may be economically relevant.

The timing of crate confinement remains a key factor. While fully loose systems are increasingly promoted, short-term confinement around farrowing has been described as a pragmatic transitional approach, particularly in large litters (Yun et al., 2019; Baxter et al., 2025). The present results suggest that combining carefully timed confinement with structured nesting material may represent a feasible compromise that supports both piglet protection and sow welfare. However, temporary confinement still restricts the sow’s freedom of movement, and further improvements in housing and management should therefore be considered to address broader aspects of sow and piglet welfare. In the future, automated monitoring technologies may further optimize the timing of management interventions.

It should be acknowledged that successful implementation of modified farrowing management often requires an adaptation phase. Both animals and stockpersons may need time to adjust to new procedures (Baxter et al., 2025). Survey data from German producers indicate that certain measures are perceived as less problematic after implementation than initially expected (Geisler et al., 2025), suggesting that practical experience plays an important role in optimization of welfare-oriented systems.

Overall, the targeted provision of structured nesting material represents a realistic and implementable management measure to support welfare outcomes in commercial farrowing systems.

The provision of structured organic nesting material influenced sow behavior and piglet survival under commercially relevant farrowing conditions. Long-stemmed straw reduced crushing risk during the early postpartum period and was associated with fewer posture changes compared with minimal enrichment. The results indicate that not only the presence but also the functional properties of nesting material are relevant for sow welfare and piglet protection.

Furthermore, the automated quantification of posture changes proved reliable, providing efficient and objective activity measures that complemented the analysis of crushing losses. Together, these findings support the use of structured nesting material as a practical management tool in commercial and transitional farrowing systems. Future research should further investigate management strategies beyond the immediate postpartum phase, including the timing of crate opening and husbandry practices during later lactation, to optimize welfare outcomes for both sows and piglets.