Use in human health. Mortality rates are used extensively in human health literature to monitor progress in human development. However, mortality rates are usually qualified by factors such as age, life stage, or cause. To illustrate this, Table 1 shows the definitions of selected Sustainable Development Goal (SDG) targets and indicators measuring mortality rates from SDG 3: “Ensure healthy lives and promote well-being for all at all ages” ( UNSD, 2020a; UNSD, 2020b). The specificity of each mortality indicator allows comparison across time and space. Changes in trend can also be traced back to a relatively discrete number of factors for each indicator, and therefore drive targeted research, programs and interventions.

Although only encompassing narrow age groups, neonatal, infant and under-five mortality rates are often used as barometers for overall population health ( UNSD, 2020a; WHO, 2021). Reidpath & Allotey (2003) examined the ability of infant mortality rate (IMR), defined as the number of deaths in children under 1 year of age per 1000 live births in the same year, to represent whole population health compared to the more comprehensive measure of disability-adjusted life expectancy (DALE), which accounts for mortality as well as non-fatal morbidity. The analysis found a strong, linear correlation between IMR and DALE, showing that IMR is a feasible and useful indicator of whole population health. Reidpath & Allotey (2003) also discussed the difference in resources required to collect data for each indicator, highlighting that the simpler IMR was more feasible to monitor in resource-poor countries.

This is important considering the resources that are required to collect a broad set of data to construct complex indicators, compared to the resource limitations often faced in LMICs.

Use in livestock health. Livestock fulfil multiple roles in various parts of the world; however, their primary roles are generally for income generation, food, and employment ( Herrero et al., 2013; Salmon et al., 2020). This focus on production marks a significant difference in perspective between human and animal health and means that indicators that measure progress in humans may not be directly transferrable to livestock. In addition, the mixture of public and private interests in livestock production complicates and often limits the availability of public resources for animal health. Practically, this necessitates efficiency in resource allocation for data collection, analysis, and action.

In animal health, attempts to devise a system analogous to disability-adjusted life years (DALYs) or DALEs for humans has not, as yet, gained general acceptance in the livestock development community ( Shaw et al., 2017). Therefore, at the present time, SEBI is attempting to define an indicator of livestock mortality that is a stable reflection of the animal health status of a country.

Livestock mortality has always been an important issue, as loss of stock represents a loss of wealth, livelihood, nutrition, genetic material, and a waste of investment, whether financial or through labour. The magnitude of lost value experienced by livestock keepers may be comparatively greater for those whose livestock fulfil several purposes, as is the case in many LMICs.

A suite of mortality indicators available for use in animals, as described by Thrusfield & Christley (2018), are presented in Table 2. Issues with the current definitions are immediately evident: the authors do not define the age for which calf/lamb/kid or neonatal mortality rates apply, as “there is not a universal agreement on the age at which animals cease to be neonates in veterinary medicine”. With this range of indicators and a lack of consensus on age groupings, it is important to gain a better understanding of what the term “mortality rate” actually means in a practical sense, particularly when referencing specific age groups. To this end, a literature search was performed with the aim to collate and review the terms that are most frequently used within the livestock farmer, professional and scientific research communities, and to review the age categories which have the highest incidence of mortality with a view to refining the current definition of “livestock mortality rate”. This review is reported in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) ( Wong, 2021).

A preliminary search was performed in Google to collect commonly used terminology from a mixture of academic and professional publications. This search showed that indicators used to monitor mortality rates are often specific for age groups (especially young animals), defined time periods (e.g., annual, a study duration), or specific diseases (i.e., case fatality rates). Common terms encountered during this preliminary search contributed to the development of a search strategy (outlined in Table 3) to retrieve articles reporting on mortality rates and age at mortality in cattle (both dairy and beef production systems), sheep, and goats.

Inclusion criteria. Articles from all countries published between 1900 and 2020 were considered for inclusion. A wide date range was used due to the scoping nature of the indicator definition review, and to maximise the inclusion of data from LMICs. The search included peer-reviewed journal articles (original research, secondary data analysis and reviews for both mortality indicator definitions and age distribution of mortality, and method articles also for indicator definitions). Data from published research reports related to the selected journal articles were included if they expanded on or clarified definitions used in the article. For age distribution of mortality, articles studying interventions were only included if baseline data were reported, and only baseline data was considered for inclusion. To ensure all nuances in indicator definitions were captured, articles were only included in the indicator definition section if they were in English.

Exclusion criteria. For both studies, articles were excluded if their abstracts or full text papers were not available. For the indicator definition section, studies were excluded if the definitions were based on events such as ear-tagging or weaning rather than age group. For age distribution of mortality, articles that only reported on cause-specific mortality rates, or articles that presented experimental data in laboratory settings that did not reflect realities in the field were excluded.

An overview of the criteria used to identify literature relating to the two purposes of this review is presented in in Table 4.

Literature searches were then performed in Web of Science (core databases) in June 2020. An initial search for “mortality” in combination with the full series of additional search terms yielded too broad a range of results, many of which were not related to livestock production. The search strings were refined, and the specific search strings used are presented in Table 3. A Google Scholar literature search for “livestock mortality rate” was also performed to ensure as many articles were included, as well as inclusion of relevant articles found within reference lists.

For all terms, article titles were reviewed for relevance to livestock mortality rates. Article abstracts for relevant titles were then reviewed by two authors (J. W. and A. P.) according to the selection criteria ( Table 4) and articles selected for full-text review. Full-text articles were reviewed by J. W. and A. P., and disagreements on inclusion were resolved through discussion. Where multiple papers arose from the same research study, or where research groups used the same definition or presented the same information across multiple papers, only the most detailed publication was included.

Data were extracted into a piloted form in Microsoft Excel version 2102. Data extracted included: author/s, citation, country of origin and income group, species and breed, production system characteristics, whether the aim/objectives of study directly related to livestock mortality, study type, recruitment procedure, sample size, time span covered, mortality indicator name, indicator enumerator, indicator denominator, whether abortions, stillbirths or culling events were included, and age distribution of mortality events recorded.

For the study of mortality indicator definitions, definitions were grouped by species and age range, and studied for patterns or common age ranges to produce a narrative summary. Given the scoping nature of this section of the review, studies were not individually assessed for bias.

For the study on age distribution of mortality, studies were first grouped into age ranges studied. Where studies overlapped in age range coverage, mortality risk by calendar month (365.25 days/12) was calculated, and the cumulative mortality risk by age in months graphed for each study along with the average mortality risk and standard deviation across studies. These studies were assessed for bias using the Risk of Bias Tool for Prevalence Studies ( Hoy et al., 2012). Where studies did not have commonalities in age range, the results were included in a narrative summary. These studies were not assessed for bias. All analysis was done in Excel version 2102.

This review was not registered in PROSPERO and did not require ethical approval.

The literature search for “mortality rate” in cattle, sheep and goats mostly yielded articles focussed on mortality in calves, lambs and kids. Few articles defined specific age groups for older animals – general herd or flock mortality rates tended to be reported instead. For all species and age groups, the literature was divided into those studies reporting a true mortality rate, and those reporting mortality risk.

Mortality incidence rate versus mortality incidence risk. The Centres for Disease Control and Prevention (CDC) define mortality rate as a “measure of the frequency of occurrence of death in a defined population during a specified interval” ( CDC, 2012). For livestock, movement of animals in and out of the herd, from mortality, sale, or slaughter, is common and makes quantifying the “population” as a denominator in the ratio, more complex than in static populations. The most accurate way to define a varying population is to calculate the number of animal-days-at-risk, thereby no longer counting animals as being at risk from the day they leave the population ( Thrusfield & Christley, 2018). Mortality incidence rate is often expressed per 100-animal-months or -years. However, crude mortality rates, more correctly known as mortality risk, are often used. These use estimates of the population, such as the number of animals counted or average herd/flock sizes, as a denominator and are accepted proxies. Mortality risk is commonly reported as “mortality rate” ( Thrusfield & Christley, 2018). Within the 52 included studies, 42 (81%) reported mortality risk, four reported mortality rate (8%), four studies reported both (8%) and in two review articles, it was not always clear whether the included articles reported risk or rate.

Age ranges used to define mortality rate. For cattle, the literature search identified 30 articles that used 20 mortality indicators ( Table 5). Geographically, the 30 articles spanned 18 countries, with two studies having a global coverage. Using World Bank Group classifications ( World Bank Data Team, 2019), nine were from high-income countries (HICs), and nine were from LMICs.

Perinatal and neonatal mortality indicators were the most commonly reported. For perinatal mortality, though there was variation between authors on definitions, most included stillbirths and measured mortality between birth and up to 24 or 48 hours, while only a small number included abortions. There was similar variation in definitions for neonatal mortality, although from birth or 2–3 days through to one month of age was common. For older calves, there was a complete lack of consensus as to what age range indicators included.

For sheep, 20 studies were identified, reporting on 16 indicators of mortality ( Table 6). These articles originated from 10 countries, including two HICs and eight LMICs and one article having a global scope. Perinatal, neonatal and pre-weaning mortality rates were most commonly reported. The definition of perinatal mortality varied greatly between authors, with indicators covering death between birth to 24 hours, 48 hours, seven days, and 14 days. There was a greater level of consensus for definitions of neonatal mortality, with most authors including mortalities from birth to 28 or 39 days of age. Pre-weaning mortality was most reported to be between birth to 90 days.

There was a dearth of articles reporting on mortality rates for goats. Only 11 articles were identified, from eight LMICs. These articles reported on 17 indicators ( Table 7). Perinatal, neonatal and pre-weaning mortality rate definitions were reported by three authors each. Two of the three authors defined perinatal mortality as death within the first 48 hours of birth, while two authors also defined perinatal mortality as that occurring between 48 hours and one month of age. All three authors reporting on pre-weaning mortality defined this as mortality occurring up to three months of age, although two authors measured this from birth, while one author measured this from one month of age.

A small number of studies arbitrarily divided the pre-weaning period into early and late stages of weaning, which differed greatly between authors. Papers that monitored mortality to weaning but did not specify when weaning typically occurred were excluded from the study. In the review article by Peeler & Wanyangu (1998), the authors collated the weaning ages of lambs and kids reported in grey literature from Kenya. Across 12 studies, weaning age ranged between 120 – 224 days, with an average of 152 days and a standard deviation of 30 days, showing significant variation within weaning ages of production systems used in one country.

Few studies reported mortality rates specific to adult animals - most studies reporting adult mortality rates used crude flock or herd mortality risk, rather than an age-based indicator. The exception to this is in dairy cattle, where a small number of studies reported mortality by parity in dairy cattle ( Traore & Wilson, 1988; Upadhyay et al., 2014).

Across all indicators for cattle, there was no more consistency in definitions used within HIC/LMIC groupings than between the two income categories. The small number or absence of studies from HICs for sheep and goats precluded comparisons between different production income groups. For all species, age-defined indicators concentrated on the first few months of life.

While many articles discussing livestock mortality state that mortality rates are highest in younger age categories, few articles reported the distribution of mortality by age, particularly for animals older than one year. In cattle, 15 articles ( Table 8) presented detailed tables of mortality by age, allowing crude comparison between studies. However, for sheep and goats, this information is scarce and spread across varying timeliness, making direct comparison difficult.

Mortality in adult cattle. Only four articles detailed age at mortality to at least three years of age. Data from these articles are presented in Figure 2. It is evident that mortality rises steeply in the first few months of life and begins to taper off by 12 months of age.

In addition to these studies reporting exact numbers, several papers presented cumulative mortality (or proportion of survival) over time graphically. In Mali, Traore & Wilson (1988) showed that the proportion of cattle surviving dropped most steeply within the first three months of life, then declined at a fairly steady rate between 1–3 years, before becoming negligible between 3–4 years. Raboisson et al., 2013 showed that in three different dairy cattle breeds in France, heifer survival rates dropped most rapidly within the first 200 days of age, continuing at moderate rates between 200–400 days, before reaching a lower, stable rate between 400–1400 days. Zhang et al. (2019) reported on mortalities and involuntary culling rates in dairy calves and replacement heifers in China to 60 months of age. Frequency of mortality was highest in the <3-month age group, dropped dramatically between 3–6 months, then continued to drop until 60 months of age.

Mortality in young cattle. For cattle in the first 12–15 months of life, a total of 11 articles reported detailed mortality incidence risk over time. Study sizes, age groups for which data are presented and proportions of total mortality are included in Table 9. As demonstrated in the table, there is inconsistency in the age groupings used. Three studies reported mortality risk for each month, however, one study reported mortality for each 28-day period. Other studies reported mortalities for arbitrarily-determined age groups. To be able to compare and present this data graphically, figures were either averaged or consolidated to give monthly values, and this cumulative mortality risk for the first year of life is presented in Figure 3. For studies reporting mortality risk monthly, cumulative mortality rises sharply in the first 2–3 months. In all but one study, 80% of mortalities that occur in the first year have occurred by six months of age.

Again, these observations are supported by many of the other studies that either did not report exact numbers over time or that presented data over a shorter period of time. Findings reported by authors included those where the largest proportion of deaths in cattle occurred intrapartum ( Mock et al., 2020), in the first 48 hours ( Busato et al., 1997; Raboisson et al., 2013), the first week ( Gardner et al., 1990), first one- ( Menzies et al., 1996), two- ( Santman-Berends et al., 2019), or three- ( Hyde et al., 2020) months of life.

For both extensively-managed beef cow-calf herds and large-scale, intensive dairy cattle production in Estonia, Motus et al. (2017) and Reimus et al. (2017) found that mortality rates were highest in female and male calves 0–3 months of age. For beef cattle, mortality rates dropped markedly after 3 months until 18 months of age in females, when they began to rise again, while mortality rates in male calves was more variable. In dairy cattle, for both sexes, mortality rates remained moderately high between 3–5 months, before dropping at 6 months of age. Data available for older age groups in female cattle show that mortality rates start to rise again at 24 months. Similarly, Selvan et al. (2019) reported mortality risk for two Zebu breeds and Zebu crossbreed calves under 6 months of age in India from longitudinal data from a research station. For the two Zebu breeds (Sahiwal and Tharparkar), mortality risk was highest in calves aged 0–1 month, whereas mortality risk was comparatively high for calves 0–1 month and 1–3 months of age for the crossbred calves. Norberg et al. (2013) studied mortality in Danish Jersey heifer calves aged 1–180 days and found that the risk of mortality was highest between days 1–14, moderate between days 15–30, and continued to decrease until 180 days of age. Slavik et al. (2009) found that in beef cattle herds in the Czech Republic, 62% of mortalities within the first six months occurred in the first week, with a further 28% occurring between one and four weeks of age.

In smallholder mixed crop-livestock systems in Kenya, Thumbi et al. (2013) found three periods where risk of cattle mortality was higher: the neonatal period, immediately after birth; between 150–190 days, when maternal antibody levels are waning; and towards one year of age, when calves are weaned. Knopf et al. (2014) reported that all deaths occurred in the first 210 days, while Pannwitz (2015) recorded the highest mortality rates in calves less than six months of age, then a decreased mortality rate between 6–24 months, and an uptick after 2 years of age.

In contrast to these findings, two authors, Debnath et al. (1990) ( Figure 3) and Gitau et al. (1994) reported that no strong age patterns were seen in calf mortalities, while Seppa-Lassila et al. (2016) found that mean mortality risk in large-scale Finnish dairy cattle was 5.2 ± 2.3% in calves <7 days, while the mortality risk in calves 7–180 days was 5.7 ± 6.2%.

Risk of bias of included studies. For the 15 studies that published data sets showing age distribution of mortality in cattle, risk of bias was assessed using a tool refined by Hoy et al. (2012) ( Table 10). All seven studies from HICs and five of the studies from LMICs utilised data from national registries, or research/breeding institutes, therefore random participant selection and non-response bias was unable to be assessed. It is possible that research or breeding station data may not be a true representation of realities in the field. Although some herds were managed as per the local normal, some stations had much more intensive management than would be found in surrounding areas. For the remaining three LMIC studies, these were conducted in the field using convenience sampling, based on farms being in accessible locations, being the site of previous studies, and willingness to participate by farmers. Convenience sampling may select participants who have a greater interest in the health and welfare of their cattle and therefore may employ better management practices, or who have better access to veterinary advice or treatments. Overall, the risk of bias in studies originating from HICs was assessed to be low, and moderate - though difficult to avoid due to constraints on how research can be undertaken - in studies originating from LMICs. Other types of bias, such as reporting bias, were not reported by studies and could not be assessed.

Mortality by age in small ruminants. For both sheep and goats, the variation in the time periods covered by different studies was too great to allow direct comparisons. In sheep, two papers noted that stillbirth and perinatal mortality rates were higher than at any age ( Holmoy & Waage, 2015); Voigt et al., 2019), while other authors report the largest proportion of mortalities occurred within 24 hours of birth ( Binns et al., 2002), or within first week ( Gokce & Erdogan, 2009; Gokce et al., 2013; Khan et al., 2006). Holmoy et al. (2017) reported that 80% of neonatal lamb mortalities occurred within the first two days of life. Sallam (2019) found that average mortality risk for Barki sheep reared at a research station in Egypt was 4.6% for lambs 0–3 days of age, 5.4% for lambs 4–60 days of age, and 2.2% for lambs 61–90 days of age. Turkson (2003) found that for lambs from birth to 12 months of age, 75.6% of the mortalities occurred between 0–3 months, compared to 24.4% occurring between 4–12 months of age. In India, Mandal et al. (2007) reported that mortalities of Muzaffarnagari sheep reared in a research institute were highest in lambs under six months of age. Mortality risk for lambs from 0–3 months of age was 7.5% (with 64% of deaths occurring within 15 days of birth), decreasing to 2.7% from 3–6 months of age, 1.6% from 6–9 months of age, and 1.2% from 9–12 months of age ( Mandal et al., 2007). In Jordan, Abdelqader et al. (2017) found that 63.5% of pre-weaning (from birth to 60 days) lamb mortalities occurred within the first seven days of life. In semi-migratory and village sheep production systems in Iran, Vatankhah & Talebi (2009) found that mortality risk was 6.14%, 12.76%, 3.36% and 0.69% for lambs under 3 months of age, between 3 and 6 months, 6 and 9 months, and 9 and 12 months of age, respectively.

For goats, El-Hassan El-Abid & Nikhaila (2009) observed that 21.2% of pregnancies resulted in abortion, while de Medeiros et al. (2005) reported the highest proportion of mortalities occurring within the first month. Traore & Wilson (1988) reported that 35% of all kids died before five months of age in Mali. For both sheep and goats, Tifashe et al. (2017) observed that lambs and kids had higher mortality rates than “young stock” or “adults”, however, the specific age range covered by these terms were not defined. Kumar et al. (2016) reported that risk of mortality was highest in kids aged 0–3 months (34.6%), >3–6 months (25.42%), and >6 months (19.78%). Turkson (2003) found that for mortalities occurring in kids between birth and 12 months of age, 80.2% occurred between 0–3 months, with the remaining 19/8% occurring between 4–12 months. In India, Thiruvenkadan & Karunanithi (2007) longitudinal data from research station records showed that mortality risk was higher in kids from birth to 12 months of age compared to adults older than 12 months. In this cohort, mortality risk was highest around the time of weaning at three months of age.

In Myanmar, a study on village sheep and goats production found that mortality rates were much higher in young animals aged less than 12 months compared to older animals above 12 months (3.0 deaths/100 animals/month and 0.28 deaths/100 animals/month respectively) ( Hanks et al., 2018).

Ramachandran et al. (2006) reported on longitudinal data on an experimental crossbred dairy goat herd maintained at a research institute. For goats monitored until 78 months of age (6.5 years), 59% of all mortalities occurred in the first 3 months of age, and 72% of mortalities had occurred by 6 months of age. After 6 months of age, mortality risk dropped markedly, and remained low until the goats were >78 months of age.

The inconsistency between definitions of livestock mortality indicators suggests that mortality indicators are not used consistently enough across the industry to be standardised. This is likely due to the predominance of performance indicators based on productivity in HICs, such as daily weight gain, feed conversion ratios, carcass or milk yield or egg production; the irregularity of monitoring in LMICs; and the difficulties in standardising these indicators so that they are applicable across different geographical regions and production systems. However, livestock mortality indicators may become more widely used globally - in HICs due to increasing consumer concerns about animal welfare and the potential utility of mortality indicators in this domain, and in dairy cattle, due to concern about the rising trend in mortality rates seen over the last few decades ( Compton et al., 2017; de Vries et al., 2014; Thomsen & Houe, 2018), and in LMICs due to the inclusion of “Number of animal deaths” in FAO’s recommended minimum set of core data within the Global Strategy to improve Agriculture and Rural Statistics ( FAO, 2018). In this case, a concerted effort should be made to overcome the difficulties in standardising livestock mortality indicators, as this will facilitate comparisons over time. Several sources of inconsistency that need to be addressed have been identified in this review.

First is the use of mortality risk versus mortality rate. Santman-Berends et al. (2019) compared practical aspects and suitability of mortality rates, where calf days-at-risk was used as the denominator, to mortality risk, where the total number of animals at a specified point in time is used. These authors found that although mortality rate is more accurate, mortality risk was a preferred method of measuring mortality from the scientific, comprehensibility and utilitarian points of view. This finding was supported by the high proportion of studies that reported mortality risk rather than rate. In many cases, method of data capture may not have given authors the degree of precision required to calculate mortality rate. To improve ease of comparison between studies, it may be helpful for studies with access to more detailed data to report both mortality rate and mortality risk, to facilitate comparison with studies with access to less precise data that report mortality risk.

Secondly, based on the studies reviewed here, stakeholders have a greater interest in mortality in young animals. As perinatal and neonatal mortality rates are commonly reported, species-specific definitions for these indicators using age ranges that are appropriate for use across different production systems should be set. For both cattle and goats, more papers defined perinatal mortality as occurring within the first 48 hours of life than other time periods, although variation exists as to whether abortions or stillbirths are included. Due to potential inconsistencies in the detection of abortion or determination of foetal age at abortion under field conditions, the authors propose that perinatal mortality risk be defined for cattle, sheep and goats as animals that are stillborn or die within 48 hours of birth over the total number of still- and live-born animals. For cattle, sheep and goats, most authors defined the upper age limit of neonatal mortality as one month of age. The authors propose that neonatal mortality risk for all three species include deaths occurring from three to 30 days of age over the total number of animals alive at three days.

The literature search for the term “pre-weaning mortality” showed that this is a more commonly used indicator for pig production systems than for ruminants. Conceptually, pre-weaning mortality would be a useful indicator in ruminants as it encompasses the age groups with highest risks of mortality. However, due to the high variability in age at weaning between production systems and species, an indicator based on age may be more universally appropriate. For cattle, studies that report pre-weaning mortality rates should, at minimum, specify the age at weaning for the setting of the study. For sheep and goats, reviewed papers seem to concur that the pre-weaning period should extend to 90 days of age. The authors suggest that pre-weaning mortality risk for sheep and goats be defined as the number of liveborn animals that die between birth and 90 days of age over the total number of liveborn animals.

Lastly, for animals past the weaning stage, the age groups for which mortality rates are reported could be standardised. At present, study authors arbitrarily decide on age groupings for reporting or further analysis. In a review attempting to compare magnitude of calf loss across cattle stations in Northern Australia, Chang et al. (2020) identified 42 studies that reported mortality over 13 different time periods. This variation in timelines limited the usefulness of the data, precluding meta-analysis and allowing only summary statistics to be generated. Although the length of studies may vary, if all studies reported on mortality rates for a consistent set of age ranges, this would aid comparisons across data sets and meta-analyses for more powerful results.

This review found high agreement between studies on the age groups with the highest incidences of mortality in cattle, sheep and goats, although most evidence was available for cattle. Mortality rates were clearly higher in young animals within the first few months of life, and by six months of age, a large proportion of mortalities in herds and flocks had already occurred. This appears to be a common finding regardless of geographic location or production system and is likely why most studies concentrated on reporting mortality rates for younger animals rather than adults.

While it is commonly reported anecdotally in the literature that the perinatal period or first week of life is the most dangerous period for small ruminants, studies from LMICs also reported higher risk of mortality around the time of weaning, which extends the period where higher mortality risk is observed to 6 months of age. Given the findings in this review, young stock mortality risk, where the number of animals dying within six months of birth over the number of live-born animals could be used as an indicator to cover this vital period. Considering the ability for human infant mortality rates to reflect general population health ( Reidpath & Allotey 2003), the relationship between young stock mortality risk and overall herd or flock health could be an area for further analysis.

There are several limitations to this review. Firstly, the study was limited to one citation database and limited use of Google Scholar. Web of Science was chosen for the breadth of journals indexed within this database and their relevance to livestock science. A large number of search results were generated, however, potentially more articles could have been recovered using a second citation database. Secondly, the study criteria restricting inclusion to peer-reviewed journal articles to ensure a high quality of studies may have limited the amount of data included from LMICs, where there may be language or financial barriers to publishing in peer-reviewed journals, and where operational research from governmental or non-governmental organisations may only be published in the grey literature. Finally, an inadequate number of studies reporting sufficient age distribution of mortality data were identified to conduct a meta-analysis or to disaggregate findings by production system or location.

Availability of data. While close to 50% of the cattle studies included in this review originated from HICs, sheep studies more commonly originated form LMICs, and all included goat studies were from LMICs. In HICs, farmers routinely collect data for the purposes of animal identification, registration, and performance monitoring. Due to the availability of these registries, national-level trends are relatively easy to map. Evidence from LMICs is much scarcer, with studies on mortality usually reporting on small-scale retrospective surveys or using data from institutional research or breeding stations. Lesnoff (2009) found that retrospective survey methods for estimating mortality rates over a 12-month period were fairly reliable for cattle and acceptably reliable for small ruminants, however, care is still required in interpreting survey results due to the large degree of variation within and between years. Season, large shocks, and innovations targeting herd productivity can lead to marked variation, which can affect survey results depending on when they are conducted, and the period(s) covered. Variability was higher for small ruminants than cattle, due to higher reproduction and mortality rates. To limit bias from variability, Lesnoff (2009) recommends that whole herd monitoring over several years, rather than 12-month retrospective surveys, should be employed for data collection and analysis where possible. Currently, this is difficult to achieve in LMICs, where monitoring tools are largely unavailable or unused, national livestock registries do not exist, and the reporting systems and investigation of mortality events are still in nascent form.

Data quality. Countries that maintain national databases for livestock registration appear to have robust and complete records including mortality events. However, the grace period for registration of new animals or the requirement for ear-tagging or other forms of identification to be completed before registration mean that in some cases, mortality within the first few days of life may be underreported ( Motus et al., 2018; Ortiz-Pelaez et al., 2008; Raboisson et al., 2013). A large proportion of young stock mortalities occur within the first seven days of life, particularly in small ruminants, so such unrecorded omissions may be significant.

Data quality may also be an issue in countries that rely on retrospective surveys. Accurate reporting of mortality can be influenced by the skill of the interviewer, and the farmer’s ability to accurately recall mortality events ( Lesnoff, 2009). Of the studies reviewed, Debnath et al. (1990) were the only authors to discuss the recall accuracy of farmers. These authors reported that farmers were able to reliably recall livestock mortality events, however, the exact ages of animals that died were more difficult to pinpoint. Farmers were more confident in reporting the age group of the animal. In addition to problems of recall, interviewers may encounter reluctance to report mortality, particularly in areas where disease surveillance has previously led to uncompensated control measures to stamp out disease ( Gilbert, 2012; Otte et al., 2004).

Cause of mortality. In the reviewed papers, it was rare for mortality rates in themselves to be the sole focus of the study: identifying actual or potential causes of mortality or using modelling to identify risk factors was also an important component. This denotes a major weakness in the use of mortality rates to monitor trends in livestock health; mortality rates only indicate the magnitude of the problem. To be able to make improvements, more information is required for interventions to be able to target to the underlying problems. In areas and/or age groups where the causes for mortality remain consistent over time, trends in mortality rate may be useful to track progress or decline, and to monitor for outbreaks. However, both causes and risk factors may vary widely over geographical and temporal scales. In some regions of the world where livestock mortality rates are consistently high, livestock keepers may be exposed to a variety of shocks year on year, including those related to climate or extreme weather events, political and social stability, and human or animal disease epidemics. In the FAO guidelines for estimating livestock production in LMICs, to monitor herd dynamics and animal health, it is recommended to measure the “number of deaths or disappearance per livestock species and by cause” ¹ rather than just mortality rate ( FAO, 2018).

Several of the papers included in this review which presented data spanning multiple years noted that mortality varied greatly from year to year. It was postulated that this was due to environmental stresses such as feed or water shortage, or from disease outbreaks. For livestock mortality data to be able to be useful in a timely manner, long term data needs to be collected and accessible. Causes of fluctuations in mortality rates need to be noted, and a “baseline” mortality rate for that region established from the years without external events. This is a similar concept to the “excess deaths” indicator currently being monitored for human mortalities during the COVID-19 pandemic. For livestock, where mortalities follow a seasonal pattern, baseline mortality rates should be established for each season so that if mortality rates rise above baseline at any time through the year, rapid investigation and action is achievable.

Harvard Dataverse: “Refining livestock mortality indicators: A systematic review PRISMA checklist”. https://doi.org/10.7910/DVN/JJIHJG ( Wong, 2021).

Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).