Introduction
Animal welfare is a critical component of the sustainability and productivity of dairy farms (Rivero & Lee Reference Rivero and Lee2022). Improved dairy cattle welfare strongly correlates with positive economic and environmental outcomes (Brennan et al. Reference Brennan, Hennessy and Dillon2021), and is an important element in maintaining dairy farming’s social licence and its social sustainability (Hampton et al. Reference Hampton, Jones and McGreevy2020). Indeed, systematic assessment of animal welfare was originally introduced to provide assurance around animal welfare to consumers, in order to maintain the social sustainability of livestock farming.
Originally introduced to provide ‘farm assurance’ (Blokhuis et al. Reference Blokhuis, Jones, Geers, Miele and Veissier2003; Main et al. Reference Main, Mullan, Atkinson, Cooper, Wrathall and Blokhuis2014), systematic welfare assessments can provide more than just assurance. In particular, they can identify specific welfare issues at both the industry and farm level (Des Roches et al. Reference Des Roches, Veissier, Coignard, Bareille, Guatteo, Capdeville, Gilot-Fromont and Mounier2014; Schenkenfelder & Winckler Reference Schenkenfelder and Winckler2021; Poulopoulou et al. Reference Poulopoulou, Zanon, Alrhmoun, Katzenberger, Holighaus and Gauly2023). At the industry level, systematic welfare assessment can help identify problems that dairy producers may often overlook, such as welfare problems related to behaviour rather than production (Sadiq et al. Reference Sadiq, Song-Lin, Ramanoon, Syed-Hussain, Shaik Mossadeq, Salisi and Mansor2021). As such, a well-planned, systematic welfare assessment can be an effective tool to improve farm conditions and control welfare problems by detecting problems early (Sadiq et al. Reference Sadiq, Ramanoon, Mansor, Syed-Hussain and Mossadeq2024). At the individual farm level, systematic assessment can identify farm-specific issues and provide, through feedback, a tool to support farmers to improve welfare (DuBois et al. Reference DuBois, DeVries, Haley, Lawlis and Merkies2019), as well as exposing farmers to scientific information and advice from veterinarians and other experts (Platto et al. Reference Platto, Zhu, Guo, He, Hu, Valros and Irwin2020).
As welfare issues can vary markedly between dairy farming systems, there is a clear need to design a system-specific welfare assessment that is relevant and achievable within a specific production context (Sapkota et al. Reference Sapkota, Laven, Müller and Kells2020). Most current widely available dairy cattle welfare assessment frameworks (e.g. Welfare Quality® in Europe, Farmers Assuring Responsible Management [FARM] in the US) have been primarily designed for commercial, intensive dairy farms (Krueger et al. Reference Krueger, Cruickshank, Trevisi and Bionaz2020). However, welfare assessment should be targeted to the specific needs, constraints and welfare priorities of that system (Hernandez et al. Reference Hernandez, Berg, Eriksson, Edstam, Orihuela, Leon and Galina2017), especially for systems with limited human and farm resources, such as smallholder dairy farms. But even within ‘smallholder dairy farming’, there can be marked differences. Smallholder farms can be extensive farms with cows kept principally at pasture (e.g. Hernandez et al. Reference Hernandez, Berg, Eriksson, Edstam, Orihuela, Leon and Galina2017), but can also, as is common in Indonesia, use tie-stalls with zero grazing (de Vries & Wouters Reference de Vries and Wouters2017; Fadillah et al. Reference Fadillah, van den Borne, Poetri, Hogeveen, Umberger, Hetherington and Schukken2023). Thus, for Indonesia, systematic welfare assessment specifically targeted at smallholders using tie-stalls is required to improve cow welfare in the system.
Simply applying welfare assessment measures from a published framework developed in a different production context may not fully capture the welfare priorities and problems of Indonesian smallholder dairy farms. Several studies in South Asia have assessed welfare in smallholder tie-stall farms, including Islam et al. (Reference Islam, Sharma, Ahsan, Mazumdar, Rudra and Phillips2020) and Rajpoot et al. (Reference Rajpoot, Singh, Kotresh and Kamboj2024), and these have provided useful examples of welfare assessment in broadly comparable systems. However, many assessment frameworks were originally developed for larger and more intensive dairy systems with higher inputs and outputs. A welfare assessment approach specifically developed for the Indonesian smallholder system may therefore improve the relevance and feasibility of identifying problems expected to occur on smallholder farms with limited resources, where welfare problems, farm practices and feasible interventions may differ from those in more intensive dairy production.
A welfare assessment framework should be focused on animal-based measures that can directly assess welfare outcomes (Whay et al. Reference Whay, Main, Green and Webster2003; de Vries et al. Reference de Vries, Engel, den Uijl, van Schaik, Dijkstra, de Boer and Bokkers2013; Linstädt et al. Reference Linstädt, Thöne-Reineke and Merle2024). Ideally, a holistic animal welfare assessment should also consider broader dimensions of welfare, including feeding, housing, biological functioning and health, natural or appropriate behaviour, affective states, and positive welfare (Fraser et al. Reference Fraser, Weary, Pajor and Milligan1997; Keeling et al. Reference Keeling, Evans, Forkman, Kjaernes, Blokhuis, Miele, Veissier and Jones2013; Mellor & Beausoleil Reference Mellor and Beausoleil2015), while most importantly, aiming to alleviate animal suffering (Webster Reference Webster2016). However, in the Indonesian smallholder dairy context, where empirical welfare data are still limited, identifying locally relevant priority animal welfare problems and measures through expert consensus is an important starting point for developing future welfare assessment approaches.
The first animal-focused welfare assessment study, by Whay et al. (Reference Whay, Main, Green and Webster2003), used a Delphi process to identify animal-based measures for farm animals, in which the initial findings of an expert survey of appropriate measures were summarised and fed back to the respondents, who then commented on and ranked them, resulting in consensus on the most appropriate measures. The aim of this study was to apply the Delphi process using local experts to reach consensus on animal-based measures to address priority welfare problems on Indonesian smallholder dairy farms.
Materials and methods
Ethical statement
This research was peer-reviewed and judged to be low risk and notified to Massey University’s Human Ethics Committee (4000027616). Informed written consent was collected from every respondent during the pilot surveys and the two rounds of the Delphi surveys. The identities and responses of the panellists were only known to IMJ and were not disclosed to any other panel member.
Study design
Two rounds of expert consultations as per Whay et al. (Reference Whay, Main, Green and Webster2003) were used (Figure 1).
Flow diagram of Delphi process adapted from Whay et al. (Reference Whay, Main, Green and Webster2003) to identify potential animal-based measures to assess dairy cattle welfare on Indonesian smallholder dairy farms.
Figure 1 Long description
The Delphi process began with drafting the Round 1 questionnaire in English and translating it into Indonesian language. The questionnaire was piloted with five veterinarians and refined before participant recruitment. In Delphi survey round 1, the consultation began with gathering expert baseline opinions and knowledge. The responses were then summarised and analysed. A round 2 questionnaire was then developed from round 1 results, piloted again with five veterinarians, and refined based on their feedback. In round 2, the round 1 summary was returned to the panellists, who then rated and ranked the measures. The process concluded with analysis of the responses and identification of consensus.
Delphi survey 1
Recruitment
The respondents in this study were people with understanding and competence in dairy cattle welfare obtained through work, education, research, and/or veterinary practice. They were recruited using networks of the authors combined with recommendations from professional groups, associations and Indonesian livestock veterinarians who worked in livestock industries. In the first recruitment wave, at least one of the officials (chairman, general secretary, or secretary) and senior fellows of the Indonesian Dairy Cattle Veterinary Association (IDHSI), Indonesian Veterinary Association (PDHI), Association of Indonesian Animal Husbandry Scholars (ISPI), Alumni of Productive Livestock Study Association (HSTP), Faculty of Veterinary Medicine, Universitas Gadjah Mada were contacted via electronic messaging systems (WhatsApp or LinkedIn). They were asked to recommend experts and share contact details with the aim of getting ≥ 50 contacts. Once potential experts had been identified, a personal invitation to participate was sent via WhatsApp or LinkedIn. Experts contacted in the first wave were asked to recommend additional experts and provide their contact details after finishing the survey.
Delphi survey 1 questionnaire
The initial form of Delphi survey 1 was an English-language adaptation of Whay et al. (Reference Whay, Main, Green and Webster2003). Once finalised, it was translated into the Indonesian language and sent electronically to five veterinarians and academics who were not going to be part of the expert panel. These experts were asked to comment on the questionnaire’s feasibility and practicability. Based on their input, the survey was modified to improve language clarity and eliminate typographical errors. In addition, as recommended by the five experts, example answers were added to selected questions in order to minimise misinterpretation (see Supplementary material 1). The questionnaire was then sent as a digital link to the experts identified as being suitable for the Delphi panel, alongside a brief explanation of the research.
Survey 1 had four sections (see Figure 2). Section 1 consisted of a brief introduction, and consent to participate and to acknowledge their contribution. Section 2 consisted of questions around panellists’ occupation and contact details. For Section 3, each panellist was asked to identify three welfare problems that can affect the mental and/or physical state of dairy cattle (Whay et al. Reference Whay, Main, Green and Webster2003), and to rate each of these three problems in four areas using a 1 (low) to 5 (high) scale:
-
1) Its importance to individual cow welfare;
-
2) Its importance to the herd or to other cows on the same farm;
-
3) Its importance to the general Indonesian dairy cow population and to the dairy farmer;
-
4) The likelihood of solving the problem by managing farming practice.
Figure 2.Flowchart of Delphi Survey 1 questionnaire used to identify perceived priority welfare problems and animal-based measures to assess dairy cow welfare on Indonesian smallholder farms.
Figure 2 Long description
Section 1 consists of an introduction and informed-consent form. Section 2 gathers panellist information, including name, address, telephone number, and affiliation. In Section 3, panellists were asked to identify three dairy-cattle welfare problems. Each problem was rated on a five-point Likert scale for its importance to an individual cow, the herd, and the Indonesian dairy-cow population, and for the possibility of addressing it through good farming practices. Panellists identified two animal-based measures for each problem and rated each measure for robustness, practicability, suitability for smallholder farms, and suitability for large-holder farms. Section 4 requested general recommendations, nominations of additional experts, and provided general information about Survey 2.
The panellists were then asked to identify, for each of the three problems they identified, two animal-based measures that could be used to assess the problem. Animal-based measures are used as defined by Whay et al. (Reference Whay, Main, Green and Webster2003): measures that reflect the effect of farming on the animal which can be assessed either through direct observation or through the use of farm records.
The panellists were asked to rate each identified measure in four areas using a 1 (low) to 5 (high) scale:
-
1) Robustness of the measure;
-
2) The practicability of the measure for use during a 1–2-h farm visit;
-
3) The suitability of the measure on smallholder farms;
-
4) The suitability of the measure on larger enterprise-owned dairy farms.
Lastly, in Section 4, the expert panel were given general information about what would happen in the second round of the Delphi process, an opportunity to comment on the process so far and to recommend another expert who could participate in survey 1.
Delphi survey 2
The problems and measures identified by the expert panel in Delphi survey 1 were exported to Microsoft Excel®. They were then coded, categorised, and summarised (see Figure 3).
Process of developing Delphi Survey 2 questionnaire to rate and rank the most important welfare measures for assessing welfare in Indonesian smallholder dairy farms.
Figure 3 Long description
Flowchart illustrating the process of developing Delphi survey 2, starting from exporting problems and measures to Excel, followed by coding similar measures into English words, categorising into welfare principles, criteria, and problems, and summarising results. Final steps include creating a questionnaire for a second Delphi survey where panellists rate importance on a 1-5 Likert scale and rank their top five measures, with boxes connected by arrows showing sequence.
Number of responses for each dairy cattle welfare problem (n = 54) identified by an Indonesian expert panel (n = 47) in Delphi Survey 1. Each expert was asked to identify three welfare problems and two potential animal-based measures suitable for detecting each problem in Indonesian smallholder dairy farms.
Figure 4 Long description
Each of the 47 panellists nominated three welfare problems, resulting in 54 distinct problems after coding and categorisation. Lameness was nominated most frequently, with 18 responses, followed by mastitis with 14 and heat stress with 12. Negative energy balance received eight responses, while the remaining problems received seven or fewer responses each; many were nominated by only one or two panellists. Of the 54 problems, 33 were health-related, 13 related to housing or the environment, seven concerned feeding, and one related to emotional state or behaviour.
Coding process
The identified problems/measures were translated from Indonesian to English and coded into one or two words in English and then listed in a Microsoft Excel® spreadsheet. Problems and measures that were effectively equivalent were merged. During translation from Indonesian to English, the consistency of terms was evaluated, standardised, and coded.
Categorisation process
Coded problems were categorised based on the adaptation of the Welfare Quality® 4 principles and 12 criteria (Keeling et al. Reference Keeling, Evans, Forkman, Kjaernes, Blokhuis, Miele, Veissier and Jones2013) (Table 1).
Adaptation of Welfare Quality® (Keeling et al. Reference Keeling, Evans, Forkman, Kjaernes, Blokhuis, Miele, Veissier and Jones2013) principles and criteria used to categorise welfare problem and assessment measures identified in Delphi survey 1 for Indonesian smallholder dairy farms
Table 1. Long description
The table presents an adapted framework comprising of four welfare principles and 12 criteria. Good feeding includes absence of prolonged hunger and absence of prolonged thirst. Good environment includes comfort around resting, thermal comfort, and ease of movement. Good health includes absence of pain and injury, absence of disease, and absence of painful management procedures. Appropriate behaviour includes expression of social behaviour, expression of other behaviour, positive or negative emotional state, and a good human–animal relationship.
For all measures, two additional descriptors were recorded: the type of measurement (i.e. animal-, management-, or environment-based), and type of observation (i.e. recording, direct, management practice, and environmental condition). After the coding and categorisation process, the frequencies of the problems and measures identified by survey 1 were calculated and summarised. This summary was sent to all experts who responded to survey 1, alongside a link to survey 2.
Delphi survey 2 questionnaire
Panellists were again asked for consent and for brief personal information including name, address, and whether they usually worked on small-scale farms, large-scale farms, or government dairy farms. The main section of the survey listed all the individual animal-based measures identified after the coding and categorisation of the results from Delphi survey 1. These measures were grouped based on the welfare criteria associated with the welfare problem the expert had linked them to. The panellists were asked to indicate, for each animal-based measure, the degree of importance it had as a means of measuring the welfare criterion to which they were linked (1 [low] to 5 [high] scale). After scoring all 83 measures individually, they were asked to write and rank their top five most important measures (irrespective of any linked welfare criterion).
Survey delivery
Before survey 1 began, the research objectives were further explained to each participant via Zoom meetings or text sent via WhatsApp. For survey 1, the experts were given four weeks to complete the survey. Non-responding participants were sent personal reminders after 3 days. Survey 2 was sent to all participants in survey 1 and was available for 3 weeks, with non-respondents being followed up every 3–4 days during that period (maximum of three times).
Data analysis
Delphi survey 1
The practicability, robustness and suitability of all animal-based measures remaining at the end of the coding process were estimated by calculating the mean score for each measure of all respondents who included that measure in their response. Measures were categorised based on mean scores. Scores of ≤ 4 were considered low in practicability, robustness, or suitability, while scores of > 4 were deemed high in these aspects. Cohen’s Kappa (κ) was then used to assess the strength of agreement between a measure’s robustness and practicality, and between a measure’s suitability for smallholder and large farms. Cumulative probabilities for each benchmark (Landis & Koch 1977) were calculated (Gwet Reference Gwet2014). All data were statistically analysed using IBM SPSS® (Version 30, USA, IBM).
Delphi survey 2
For survey 2, the percentage of the maximum possible score (MPS) was calculated for each measure based on the importance score given by each panellist using the following formula.
A threshold of MPS > 85% was applied to identify consensus in the participants around a measure’s importance. Once all measures with an MPS > 85% had been identified, duplicate or similar measures were reviewed and amalgamated. Amalgamation was undertaken in three situations: when duplicate measures were linked to more than one welfare criterion (see Table 1 for the welfare criteria used); when similar measures were linked to different welfare criteria; and when different measures assessed the same welfare aspect. In these cases, only one measure was retained in the final set, irrespective of its link to specific welfare criteria. The decision on which measure to retain in the final list was made by IMJ. To avoid duplication, the final measures were not categorised according to specific welfare criteria.
In addition, the cumulative sum of rank (CSR; Whay et al. Reference Whay, Main, Green and Webster2003) was calculated for each measure listed by an expert as one of their top-five measures, with scores ranging from 5 for a measure ranked as the most important, to 1 for the fifth most important. All measures with at least one rank were then checked for duplication and relevance. Measures that were irrelevant or not listed on the initial list of 83 measures in the Delphi 2 questionnaire were eliminated. If duplication was identified in relevant measures, the measures were amalgamated; the final term used for that measure was the same as that used after the amalgamation of the measures identified as having a MPS > 85%. For all amalgamated measures it was confirmed that there were no panellists who had duplicate amalgamated measures in their list of 5 ranked measures (e.g. including both ‘feeding behaviour’ and ‘rumination behaviour’). Once this was completed, the final CSR was calculated for each remaining measure. If a remaining measure had been amalgamated, its final CSR was the total CSR of all measures amalgamated into that measure (e.g. if the CSR of feeding behaviour was 6 and that of ‘rumination behaviour’ was 5, the CSR for the amalgamated measure would be 11).
Results
Response rate and demographics
The first wave of recruitment identified 52 experts with purposive snowball sampling identifying an additional 11. The Delphi survey 1 questionnaires were thus sent to a total of 63 respondents, of whom 47 (75%) filled in the survey. Of the 47 experts sent Delphi survey 2, 44 (94%) responded. Thus, of the initial 63 experts contacted, 44 (70%) completed the two rounds of the Delphi survey. Of those 44 respondents, 38 were working veterinarians, five were academics and one was a dairy consultant. In terms of experience, 28 of the 44 experts had experience working in smallholder farms, 20 in large farms (with eight having experience in both), and four had experience with government farms. The 44 panellists originated from IDHSI (23), ISPI (3), PDHI (4), and HSTP Alumni (6) with eight being recommendations from other experts.
Delphi survey 1: Problem identification
All 47 panellists identified three welfare problems each. After coding and categorising their responses, 54 different welfare problems were identified (see Figure 4 and S1 [Supplementary material 2] for the detailed categorisation), of which 33 were health-related, 13 were related to housing or environment, 7 were linked to feeding problems, and 1 was classified as an emotional state or behaviour-related problem. The three most frequently highlighted welfare problems were lameness (18), mastitis (14), and heat stress (12).
Delphi survey 1: Measure identification
All 47 panellists identified six measures, except for one who listed only five. After categorisation and coding, 83 different animal-based measures were identified (see Figure 5). These measures were linked to good feeding (11), good environment (16), good health (55) and appropriate behaviour (1). The behaviour measure, fear behaviour, covered only two criteria of appropriate behaviour, so no measures covered the expression of social or other behaviour (see Table 1). After the coding and categorisation process, eleven non-animal-based measures were identified; eight were environment-based measures, and three were management-based (see Figure S1; Supplementary material 2).
Potential animal-based measures (n = 83) to assess dairy cow welfare in Indonesian farms identified from Delphi survey 1 categorised based on modified welfare principles and criteria adapted from Keeling et al. (Reference Keeling, Evans, Forkman, Kjaernes, Blokhuis, Miele, Veissier and Jones2013).
Figure 5 Long description
The 83 potential animal-based measures identified from Delphi survey 1 were grouped according to welfare principles and criteria from Keeling et al. (2013). Good feeding included 11 measures: eight related to the absence of prolonged hunger and three to the absence of prolonged thirst. Good environment included 16 measures: seven related to comfort around resting, seven to thermal comfort, and two to ease of movement. Good health included 55 measures: 38 related to the absence of disease and 17 to the absence of pain and injury. Appropriate behaviour included only one measure, fear behaviour, classified under negative emotional state and the human–animal relationship.
Delphi survey 1: Suitability of identified measures
The comparison of the suitability of the 83 measures identified during survey 1 for smallholder and large farms is presented in Tables S2, S3 (see Supplementary material 2). Of the 54 measures that were thought to have high suitability for large farms by the panellists, 32 were thought to be also suitable for smallholder farms. Whereas, of the 29 measures thought to have low suitability for large farms, only one was thought to be highly suitable for smallholder farms. Cohen’s Kappa (κ) was 0.48 (95% CI 0.32–0.64), showing moderate agreement. Analysis of cumulative probability (Gwet Reference Gwet2014) indicated that there was only a 5.5% chance that agreement was substantial or better.
Delphi survey 1: Robustness and practicability of the identified measures
Of the 46 measures which were thought to be highly robust by the panellists (mean robustness score > 4), 20 (43%) were thought to have low practicality on smallholder farms (mean practicality score ≤ 4). In contrast to the 33 measures thought to be highly practical for smallholder farms (mean practicality score > 4), only 7 (21%) were thought to be low in robustness. Cohen’s Kappa indicated fair agreement (κ = 0.36; 95% CI: 0.18–0.55) with only a 7.4% chance that agreement was substantial. See Tables S2, S4 (Supplementary material 2) for each measure’s mean robustness and practicality and their categorisation.
Delphi survey 2: The most important measures based on maximum possible score
Table 2 shows the 29 measures which achieved MPS > 85% (consensus threshold), see Table S2 (Supplementary material 2 ) for the full MPS result for each measure. The only suggested measure, which was related to appropriate behaviour, fear behaviour, did not meet the 85% threshold but was included in the list to ensure that there was one behaviour measure included in the final list. After amalgamation, a total of 20 separate measures remained (Table 3). The majority of measures were health-related (12/20), followed by environment/housing measures (5/20), and feeding-related measures (3/20).
Welfare measures to assess dairy cattle welfare in Indonesian dairy cows identified in Delphi survey 2 as having a percentage of the maximum possible score (MPS) for importance above 85%
Table 2. Long description
The table lists 29 measures that exceeded the consensus threshold of 85% of the maximum possible score, together with fear behaviour, which scored 71% but was retained to ensure inclusion of a behavioural measure. Body condition score and lameness received the highest scores, both at 94%, followed by reproduction recording at 93%. Cow comfort and mastitis testing each scored 92%. The remaining measures generally scored between 86% and 91%. Overlapping or similar measures were amalgamated, including milk-production assessment and records; body and rectal temperature; rumination and feeding behaviour; appetite and rumen fill score; ease of movement, cow comfort, and movement from lying to standing; and repeated measures of ectoparasites and hoof condition.
*These measures were amalgamated before inclusion in the final list; **as there was only one measure in the appropriate behaviour welfare principle, it was included in the consensus by default even though its MPS was <85%.
Final list of proposed animal-based measures identified through the Delphi process for addressing perceived priority welfare problems in Indonesian smallholder dairy farms
Table 3. Long description
Body condition is assessed using a 1-10 scoring method and rumen fill using a 1-5 scoring method. Dehydration is assessed using a 0-3 eyeball recession scale, cleanliness categorised as clean, dirty, and very dirty. Ectoparasites assessed by flies and ticks counts. Heat stress using a 0-4.5 panting score. Rumination behaviour using 0-3 scores. Ease of movement assessed from easy movement to inability to stand or lie down. Additional measures assess lameness using adapted indicators from stall lameness score. Limb abnormalities used the 1-3 hock scoring, Hoof condition by presence of overgrown hoof. Mastitis test used the California Mastitis Test. Reproduction recording uses presence of the recording on farm. Udder condition using palpation and visual observation. Downer cow using interview. Rectal temperature using thermometer. Vulvar discharge using 0-3 discharge scores. Milk production recording the reduction of milk production. Pain behaviour and fear behaviour assessed using the presence of those behaviours.
Delphi survey 2: Cumulative score rank
Forty-seven measures were nominated and ranked by at least one panellist in their top five measures. Following relevance screening, a total of 12 measures were eliminated. Five were identified as a welfare problem rather than a measure, three were not listed on the initial list of 83 measures, three were not animal- or record-based measures and one was not a routinely assessed welfare measure (see Table S5; Supplementary material 2 for the overview of this process). The remaining 35 measures were amalgamated into 24 measures for which CSR was calculated. Of the 24 measures, 15 were listed in the final 20 list from MPS, and nine (cow body morphology, skin turgor, examine stress hormones, cow vocalisation, itchy/discomfort behaviour, standing duration, faecal condition, calculating weight and standing behaviour) were not. Five measures (udder condition, eyeball recession, limb abnormalities, downer cow and vulvar discharge) received a CSR of zero as they were not ranked in the top five by any respondents. The top five measures based on CSR were BCS (184), RFS (76), mastitis (71), hoof condition (67), and fear behaviour (44).
Discussion
This Delphi process identified a list of potential animal welfare measures to address welfare problems perceived as a priority on Indonesian smallholder farms by a group of Indonesian experts. This set of measures can be used as a starting point for the future development of a systematic animal welfare assessment protocol for Indonesian smallholder farms and other similar production systems. To our knowledge, no similar Delphi process has previously been used to identify priority animal-based welfare assessment measures for smallholder dairy farms. Kamboj et al. (Reference Kamboj, Kumar and Mahla2022) consulted experts when developing an assessment protocol for dairy farms, including smallholder dairy farms, in Northern India. However, their study was a one-stage consultation process to obtain expert opinion on pre-selected indicators and was not focused on animal-based measures.
This study aimed to recruit Indonesian experts who were familiar with both dairy farming and animal welfare and could identify welfare problems and measures that were relevant, effective, and applicable in the Indonesian context. The study was successful in recruiting 70% of those approached, resulting in 44 Indonesian experts providing responses to both surveys. This study’s response rate (70%) was higher than Whay et al. (Reference Whay, Main, Green and Webster2003) (22%) and other e-Delphi studies, e.g. 8.2% (n = 39/477) (Truelove et al. Reference Truelove, Martin, Langford and Leach2020), 48% (n = 98/206) (Campos-Luna et al. Reference Campos-Luna, Miller, Beard and Leach2019). A low drop-out rate (6%) was observed between our first and second survey, perhaps due to pre-survey explanations, personalised reminders, a simple survey interface and a relatively long time-frame to finish the survey.
The experts’ opinions suggest that welfare assessment measures may not be equally suitable for smallholder and large-holder dairy farms. As agreement on the suitability of measures for large and small farms was only moderate (κ = 0.48; 95% CI 0.32–0.64), this supported the need for a smallholder-specific welfare assessment approach. Only 32/54 measures identified as highly suitable for large farms were thought highly suitable for small farms. Fair agreement (κ = 0.36; 95% CI: 0.18–0.55) between our experts’ opinions on the robustness and the practicality of the 83 measures was also identified. This suggests that while there was some level of agreement, the robustness and practicality of measures are not always equivalent. In any large-scale assessment programme, the welfare assessment measures need to be robust. Ideally, each measure should be validated as an indicator of the welfare outcome it is intended to assess, be repeatable within- and between-observers, and correlate well with other indicators of the intended welfare problems (Browning Reference Browning2023). However, particularly on dispersed, often remote Indonesian dairy farms, practicality is important, as there is a clear need to develop a time-limited assessment that does not require repeated measurements or significant infrastructure. Literature on the robustness of the final list of suggested measures is common (de Vries et al. Reference de Vries, Engel, den Uijl, van Schaik, Dijkstra, de Boer and Bokkers2013; de Graaf et al. Reference de Graaf, Ampe, Buijs, Andreasen, Des Roches, van Eerdenburg, Haskell, Kirchner, Mounier, Radeski, Winckler, Bijttebier, Lauwers, Verbeke and Tuyttens2018; van Eerdenburg et al. Reference van Eerdenburg, Di Giacinto, Hulsen, Snel and Stegeman2021); however, practicality can only be tested in the field. Therefore, our future study will focus on testing the feasibility and practicality of this set of measures in a real farm setting.
Choice of experts is fundamental to the value of the Delphi process (Vernon Reference Vernon2009; Shang Reference Shang2023); the predominance of veterinarians among the expert panel who completed both surveys (86%) should be acknowledged. This composition may have influenced the prioritisation bias of welfare problems and measures towards health-related outcomes, with more than half of the final list of measures being health-related (Table 3). Whay et al. (Reference Whay, Main, Green and Webster2003) involved a more diverse expert panel distribution (40% researchers, 33% veterinarians, and 27% farmers) and produced a broader range of welfare measures including those related to social behaviours. Therefore, to create future holistic welfare assessment frameworks, a broader range of stakeholders is needed to capture a wider perspective on welfare priorities in Indonesian smallholder farms. As reflected in the current study, it is difficult to find non-veterinarian experts who understand both dairy cattle farming and animal welfare in an Indonesian context. This highlights the need to expand the knowledge and awareness regarding animal welfare among non-veterinarian stakeholders, such as dairy farmers, cooperative staff, industry representatives, and policymakers, so that the future welfare assessment frameworks can reflect the wider realities of smallholder dairy farming.
In Delphi survey 1, lameness (18), mastitis (14), and heat stress (12) were the most frequently listed welfare problems among 54 different problems nominated by the panellists (n = 47). The majority of the welfare problems were health-related (33/54). Measures that were related to important welfare criteria as described in Keeling et al. (Reference Keeling, Evans, Forkman, Kjaernes, Blokhuis, Miele, Veissier and Jones2013), such as expressing social behaviour, expressing other behaviour, painful management procedures and positive emotional state were not identified by the Indonesian expert panel. Again, this may be due to the veterinarians’ point of view as they formed the majority of panellists. Although Indonesian veterinarians receive formal training related to animal welfare, veterinary training is focused primarily on animal health rather than animal welfare science. This pattern is similar to that reported by Bertocchi et al. (Reference Bertocchi, Fusi, Angelucci, Bolzoni, Pongolini, Strano, Ginestreti, Riuzzi, Moroni and Lorenzi2018), who used a modified Delphi process to characterise hazards and welfare promoters, and to weight the impact of animal-based measures, aimed at developing the first Italian dairy cattle welfare assessment protocol.
In Delphi survey 2, the survey focused on the panel’s perception of the importance of the welfare measures identified in survey 1 with the measures included in the final list being based on MPS and cumulative score rank (CSR) as previously described by Whay et al. (Reference Whay, Main, Green and Webster2003). The MPS reflects the overall agreement of the panellists regarding the importance of each welfare measure, while CSR was used to indicate panellists’ individual opinion on the most important welfare measures. The top 5 most important measures (based on CSR) were: BCS (184); RFS (76); mastitis (71); hoof condition (67); and fear behaviour (44), reflecting the priorities of the panellists in regard to welfare improvement, particularly in the feeding and health areas.
A total of 19 measures reached the pre-defined consensus threshold (MPS > 85%) and were checked for duplication and underwent an amalgamation process. These measures covered all the welfare principles and most of the welfare criteria except for the appropriate behaviour category. Therefore, one measure, fear behaviour (MPS = 71%) which did not reach the pre-defined threshold was included in the final list of measures to ensure the inclusion of at least one measure related to appropriate behaviour and the human-animal relationship. Interestingly, although fear behaviour had the lowest MPS of any of the measures in the final list of measures, it had the fifth highest CSR, indicating that although some of our panellists recognised the critical importance of behaviour measures in welfare assessment, many did not (again demonstrating the bias of our panellists towards health-related measures).
The expert panel’s inherent bias towards health-based measures may reflect the welfare issues currently recognised by Indonesian veterinarians as a priority in Indonesian smallholder dairy farms, where failure to address basic needs (e.g. food and water) and acute negative animal welfare problems (e.g. lameness and mastitis) may be the principal welfare issues on such farms. In this context, the list of measures identified in this study should be treated as an initial set of animal-based measures for addressing priority welfare problems, rather than a state-of-the-art welfare assessment. A complete welfare assessment protocol should include a broader dimension of animal welfare which explicitly incorporates positive animal welfare, behavioural interactions and mental state (Mellor & Beausoleil Reference Mellor and Beausoleil2015) and also “freedom of choice” (Webster Reference Webster2016). Therefore, further refinement is needed to develop a more holistic welfare assessment protocol. However, identifying and addressing priority welfare problems, such as negative welfare experiences, is a practical and feasible first step that can help improve welfare for the animals within a system where the welfare of farmers is often compromised (Toiba et al. Reference Toiba, Rahman, Hartono and Retnoningsih2024).
Another limitation of this study is that the tie-stall system makes it difficult to include measures initially described as ‘ease of movement’. The current list did not capture cows’ full opportunities for locomotion or freedom of movement. Movement restriction is itself an important welfare concern in Indonesian smallholder systems, particularly where cows are tied or tethered, as this may prevent natural behaviour and cause discomfort. Future studies should investigate suitable measures to assess movement restriction, tethering methods, and associated pain or discomfort in the tie-stall system, thereby creating a more holistic welfare assessment framework.
The lack of focus on the positive welfare assessment, affective state, social behaviour, and the painful management procedures by many in our expert panel indicates that Indonesia lacks expertise related to the behaviour and emotional state of dairy cattle. A similar lack of expertise in our panellists’ understanding of animal welfare was also reflected in the results of Delphi survey 1 and 2, where a number of non-animal-based measures were nominated, despite an explanation and examples of animal-based measures being given to the panellists prior to and during the Delphi consultation process. Together, these findings suggest that there is a clear gap in animal welfare understanding among the panellists, with more focus on health and productivity measures rather than the mental or behavioural domain. Therefore, it is suggested that future efforts to improve animal welfare should include strengthening veterinarians’ and stakeholders’ understanding of animal welfare, particularly animal-based measures that are related to behaviour and emotional state.
There is also likely to be a lack of understanding around animal welfare science in Indonesian smallholder dairy farmers, as although they may have practical insight into dairy farming, the majority have a limited education (Ainy et al. Reference Ainy, Ria Amelia Febriani, Ahmad Fakhrur and Anisa Aulia2023; Fadillah et al. Reference Fadillah, van den Borne, Poetri, Hogeveen, Umberger, Hetherington and Schukken2023) which will not include animal welfare. Thus, further research to understand the knowledge, attitudes, and practices of dairy cattle stakeholders, including veterinarians and farmers, towards dairy cattle welfare is needed. Understanding their values and approaches is essential to develop strategies, actions, and policies to enhance dairy cattle welfare (Fraser Reference Fraser2009; Degeling & Johnson Reference Degeling and Johnson2015; Weary & Robbins Reference Weary and Robbins2019).
The inclusion of example answers in the questionnaires may have influenced the panellists’ response. However, the results suggest that any such influence was limited. Our questionnaire included three examples of welfare problems: lameness, negative energy balance, and cleanliness (see Supplementary material 1). All were frequently mentioned by the panellists (18, 8 and 7 panellists, respectively). This is not surprising as they are common welfare problems on dairy farms around the world. However, the panellists identified 51 other welfare issues that were not included as examples, with ‘mastitis’ (14 panellists) and ‘heat stress’ (12 panellists) being identified more frequently than negative energy balance and cleanliness. This suggests that although our examples may have highlighted these issues to our panellists, the panellists’ responses were not solely driven by the examples.
The current study has identified animal-based measures that Indonesian experts considered relevant for addressing priority welfare problems in smallholder dairy farms. However, this study only focused on identifying and prioritising potential measures to be included in a welfare assessment framework and did not evaluate its application and effectiveness in a real-world setting. Further research is therefore required to test the practicality and feasibility of this set of measures in a time-limited, on-farm assessment, and to determine how they can be used to improve the welfare of dairy cows in Indonesia.
Animal welfare implications
As the first step towards developing a systematic and system-specific welfare assessment framework for Indonesian smallholder dairy farms, this study identified a set of potentially applicable animal- and record-based measures. This study also captured the Indonesian veterinarians’ point-of-view towards health-related and negative animal welfare measures, and further capacity building or training is needed to improve Indonesian stakeholder understanding around dairy cattle welfare.
Conclusion
This Delphi study is the first systematic study in Indonesia to specifically identify animal-based welfare assessment measures for addressing welfare problems (considered by an expert panel to be priority problems) in smallholder farms. Most measures identified in this study related to dairy cattle health and production, reflecting the issues recognised as priority issues by the Indonesian expert panel. Although these measures focus mainly on negative welfare outcomes and do not yet include positive welfare measures, they provide a starting point for a system-specific welfare assessment approach. Future work should test their feasibility and practicality on farms and refine them by incorporating broader behavioural, affective-state and positive welfare measures.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/awf.2026.10124.
Data availability statement
The raw data supporting the findings of this study are available from the corresponding author upon reasonable request. The data are not publicly available due to participant confidentiality considerations.
Acknowledgements
IMJ was co-financed by a PhD research grant from Phyllis Irene Grey Fellowship for Veterinary Sciences (2023), Sinclair Cummings Veterinary and Animal Science Scholarship (2023), Gustav Rosenberger Memorial Fund (2024) and PhD scholarship by Massey University Alumni Doctoral Scholarship (2024-2027). The authors acknowledge Taufiq Fadhlullah, DVM from Indonesian Cattle Veterinarian Association for his valuable contribution in connecting the author with the experts and five Indonesian veterinarians for their contribution on the pilot study. The authors also thank all the survey respondents who contributed their time to responding to the survey.
Competing interests
None.
Open access
