Introduction

Best management practices during calf rearing are essential to ensure the availability of future replacement stock (Lanyasunya et al. 2006) and thereby the total herd sustainability (Chang’a et al. 2011; Odero-Waitituh 2017). Appropriate housing for good health and growth of the dairy calf is of vital importance for its future production capacity (Bolt et al. 2017).

Dairy heifer growth rates and body weight at first calving are regarded as important benchmarks in farm management because lower body weight heifers are at a greater risk of dystocia at first calving (Mee 2008), produce less milk, and have a shorter lifespan in the herd (Archbold et al. 2012; McNaughton and Lopdell 2013). It is, therefore, important to understand reasons for growth impairment in calves, especially if they are associated with housing factors that can easily be fixed by farmers themselves.

Methods of rearing dairy calves in Kenya vary considerably from farm to farm. In smallholder resource-limited farms, dairy calves are commonly kept in pens but sometimes they may be allowed to graze or be tethered on natural tropical pastures without protein supplementation. Calves are also often housed poorly, in small pens together with other livestock, particularly small ruminants. As a result of inadequate feeding and poor housing management, stunted growth and high mortality have been reported (Lanyasunya et al. 2006).

Due to decreasing land space and tick infestation of pastures, complete zero-grazing is becoming a common alternative; the animals are housed throughout the year, and feed is provided to them. In such situations, farmers are often unable to produce sufficient fodder because of their smallholdings, and so they have to supplement with forage harvested from public areas or purchased fodder (Rademaker et al. 2016). With complete housing, there is often an excessive accumulation of slurry, presence of uneven and/or hard floors in calf pens and leaking roofs (Nguhiu-Mwangi et al. 2013; Kimeli et al. 2015).

Wide-ranging growth rates of calves have been reported in Kenya. Lanyasunya et al. (2006), in their study on the effect of supplementation on performance of calves in the Nakuru area of Kenya, reported an overall average daily gain (ADG) of 0.338 kg/d, while Gitau et al. (2001) reported a mean ADG of 0.26 kg/d in Central Kenya. Both figures were below 0.4–0.5 kg/d which is the recommended target for smallholder dairy farms in central highlands of Kenya (Lukuyu et al. 2012) and 0.5–0.75, targets for developed countries. In two recent cross-sectional studies, Kathambi et al. (2018) observed an ADG of 0.50 kg/d in calves aged between 1 week and 12 months, while Makau et al. (2018) reported 0.443 kg/d in heifers aged less than 36 months. The latter two studies may have suffered survivor bias; thus, a possible over-estimation is possible. No study has systematically evaluated the smallholder farm impact of calf housing enhancement on ADG while controlling for other related factors of ADG.

Focused prospective studies on calf growth, the housing environment and individual calf parameters are limited in Kenya. The few research studies on this topic (Gitau et al. 2001; Lanyasunya et al. 2006; Thumbi et al. 2013) were largely explorative or based on extensive management systems that are dominated by indigenous breed crosses. Analysis of growth patterns following physical environmental enhancement would be useful for determining the relative importance of housing factors that may affect production efficiency parameters such as ADG.

The objective of our study was to evaluate the effects of practical and affordable calf housing improvement on ADG during pre- and post-weaning periods on smallholder dairy farms.

Material and methods

Description of the study area

The study was carried out in Kiambu and Murang’a Counties of Kenya (Fig. 1). These counties experience an average annual environmental temperature of about 18–20 °C and an average rainfall of 962–1195 mm/year Murang’a County is on − 0.7957° S and 37.1322° E, and has a human population of 1,056,640 (KNBS 2019) and area of 2558.8 km2, with an altitude of between 914 and 3353 m. Kiambu County is adjacent to the northern border of Nairobi and is on − 1.1462° S and 36.9665° E, and has a human population of 2,417,735 (KNBS 2019) with an area of 2543.5 km2 and an altitude of 1200 to 2550 m. Kiambu County is about 40% rural and 60% urban, owing to Nairobi’s consistent growth northwards, and it has experienced tremendous growth in dairying due to its ready access to a large urban market and the existence of local milk processing facilities.

Fig. 1
figure 1

Map of Kenya showing households that were visited in Kiambu and Murang’a in a randomized controlled trial on the Kenyan smallholder dairy calves

Full size image

Study design, sampling method and data collection

This study was a randomised controlled trial, in which calf-pen improvement was implemented on smallholder dairy farms in Kiambu County and control farms were located in neighbouring Murang’a County. Randomisation was achieved at the county level (via a coin toss) to minimise contamination of control group farms, which can happen if they are located in the same area as intervention farms, since housing improvement is something fixable by owners if/when control group farmers visit intervention group farmers, a phenomenon that has been documented in Kenya previously (Kathambi et al. 2019; Richards 2017).

The enrolment of farms to the study was enhanced through a presentation at an information meeting with the farmers via their dairy cooperative society. Farm-level inclusion criteria were as follows: (1) being in the target county; (2) practicing zero-grazing; (3) having at least one unweaned heifer calf aged less than 6 weeks; and (4) having less than ten milking cows. A total of 150 farms, seventy-five from each county, were enrolled in the study over a period of 8 months.

On the initial visit, the principal farmers consenting to participate in the study were asked to respond to a questionnaire covering issues related to the calf characteristics, management practices, farmer training and demographics. The calves were also assessed by observation and thorough physical examination. We then administered calf-pen floor improvements on intervention farms which included (1) filling floor holes if on the ground or repairing wooden slats if on an elevated floor; (2) introduction of rubber mats on the lying area; (3) ensuring a floor gradient for drainage; and (4) patching leaking roofs. The control farms received the same housing improvement (as needed by each farm) after data collection was complete at the end of study period, in appreciation for their participation in the trial.

Each study animal was routinely monitored bi-monthly for up to 6 farm visits, where applicable, to obtain up to a year’s worth of data on the study animals. During the follow-up period, July 2018 to October 2019, routine clinical examinations on study calves were conducted. Data on farm management practices, and herd health and veterinary interventions on the herd since the last visit were collected.

Body weight (kg) via girth measurements (cm) were recorded at recruitment and during all monitoring visits. Faecal samples were also collected per rectum on all visits. Faecal analysis for the presence and counts of coccidia oocysts and helminth eggs was undertaken using a centrifugal faecal flotation method with Sheather’s sugar and the McMaster method, as described elsewhere (Kimeli et al. 2020). The initial and bimonthly questionnaires are available upon request.

Data management and analysis

Questionnaire data were collected using a mobile phone-based data collection platform (Epicollect5®; https://five.epicollect.net/), which was then exported to MS Excel 2010 (Microsoft Inc., Sacramento, California, USA) and coded. Laboratory data were added to the Excel spreadsheet, and the combined data were imported to Stata® 15.1 statistical software (StataCorp LLC, College Station, TX, USA) for statistical analyses.

We anticipated that factors associated with the growth of calves pre-weaning were different than for the post-weaning period. Therefore, we modelled factors associated with average daily gain (ADG) in weight for the periods between (1) first and third visits and (2) third and sixth visits. For each animal, therefore, ADG was calculated as the difference between weights measured on (1) the third visit and first visit, divided by the number of days between the visits (model I); and (2) the sixth visit and third visit, divided by the number of days between the visits (model II). When farms had multiple calves, they were monitored and recorded separately in the data.

Descriptive statistics, such as means, standard deviations, median and ranges, were determined for continuous variables, while proportions were calculated for binary and categorical variables. Between-group (counties) differences at baseline (visit 1) were calculated using independent sample t tests and Chi-squared statistics.

Ordinary least-squares linear regression was used in building models using ADG as the outcome of interest. Attempts to use the farm as a random effect (controlling for clustering of calves within farms) did not significantly change the model estimates because there was only one calf per farm for most farms, and therefore it was not utilised. Assessment for normality of ADG was performed using histogram, Q-Q plot and Shapiro-Wilk test. Box-Cox analysis was used to estimate a suitable transformation factor for the ADG outcome in the pre-weaning period. These assessments found that ADG for model I was right-skewed, and a natural logarithm of ADG achieved normality.

In the first regression step, univariable linear regression models for all the predictor variables were fitted into separate models, with those having P ≤ 0.25 being eligible for the multivariable analyses. Multicollinearity was then evaluated among the univariable association variables, and one of the correlated predictors was dropped, based on statistical significance unless there was a substantial difference in the biological plausibility. A multivariable model was then built through backward removal, while evaluating for potential confounding. Additionally, a curvilinear relationship was explored for continuous predictors. Two-way interactions between the biologically plausible variables in the final model were tested, and an interaction term was retained if the P value of interaction terms was found to be significant at P ≤ 0.05.

Residual analysis was performed to evaluate model assumptions and diagnostics. The normality of residuals was evaluated using histograms, Q-Q plots and normality tests. Homoscedasticity was evaluated by plotting standardised residuals against predicted values, and by using the Breusch-Pagan test. Identification of extreme and influential observations was done by sorting and graphing the standardised residuals and comparing changes in coefficient estimates and their significance when modelling with and without influential observations. Model fitness was assessed using PRESS statistics (Holiday et al. 1995).

Results

Descriptive statistics and univariable analysis

A total of 187 heifer calves from 150 smallholder farms were recruited at different times between July 2018 and February 2019 and followed for up to 477 days, with up to 6 on-farm assessment visits done bimonthly. The number of study farms and candidate calves decreased in subsequent visits due to loss to follow up and late enrolment. One hundred sixty-five animals from 132 farms were evaluated three times, while only 100 animals from 91 farms were assessed and sampled six times (Table 1).

Table 1 Total number of farms and calves (with median weight) at each visit of the study
Full size table

Tables 2 and 3 report on baseline differences between the intervention and control group farms. Participating farms from Murang’a County had significantly higher owned land acreage but leased less land and earned lower gross income from milk as compared to Kiambu County farms. The differences in bodyweight at enrolment and farmer education level were also statistically significant. However, there was no baseline difference in the number of adults per household, age/gender/marital status of the principal farmer, duration involved in dairy farming, previous attendance of training, floor area of the calf pen, average weaning age, calf breed or dam parity number. Average age at enrolment was 3 ½ weeks, and average weaning age was slightly over 13 weeks, regardless of the group. This weaning age reflects the common weaning age of 3 months in Kenya, and demonstrates that the first two farm visits were recording animal-level information on pre-weaned calves. Therefore, our plan to model pre-weaning ADG for visits 1 to 3 and post-weaning ADG for visits 3 to 6 was appropriate.

Table 2 Summary of animal and farm characteristics (continuous variables) pre-intervention for 187 calves on 150 Kenyan smallholder dairy farms
Full size table
Table 3 Summary of animal and farm characteristics (categorical variables) pre-intervention for 187 calves on 150 Kenyan smallholder dairy farms
Full size table

At least two pre-weaning and one post-weaning housing improvement(s) were implemented on all of the 75 intervention farms (Table 4). Rubber mats were provided to all intervention farms to improve softness of the lying area. Fixing the floor gradient to improve liquid drainage (adding dirt to the pen floor more on one side than the other), and fixing holes or slats in the floor were other common fixes applied. Over half of the intervention farms with pre-weaned calves and with post-weaned calves had at least two fixes applied.

Table 4 Proportions of 75 and 61 intervention farmers with pre-weaned and post-weaned calves, respectively, that received each of the housing improvements on their Kenyan smallholder farms
Full size table

The overall ADG of the animals that completed the study was 0.429 kg/d (SD=0.297), with a median of 0.365 kg/d. The calves examined in the pre-weaning period (1st to 3rd visits) had a mean ADG of 0.324 kg/d (SD=0.133; median= 0.307), while the animals examined in the post-weaning period (3rd to 6th visits) had a mean ADG of 0.510 kg/d (SD=0.220; median= 0.487). The mean body weight at recruitment was 46.5 kg/d (SD=8.5) at a mean age of 26.5 days (SD= 17.8), while the mean weight on the sixth (last) visit was 188 kg (SD=58) at a mean age of 356 days (SD=37.8). Figure 2 shows the large variability of bodyweight gains with age among individual calves.

Fig. 2
figure 2

Growth patterns of 187 calves on 150 Kenyan smallholder farms. The blue dots are the individual weights recorded while the grey lines connect the repeated measurements for each calf

Full size image

Descriptive statistics and univariable analyses for categorical and continuous variables eligible for multivariable analyses of factors associated with pre-weaning and post-weaning ln ADG and ADG are presented in Tables 5, 6, 7 and 8, respectively. There were 27 and 30 variables meeting the p < 0.25 eligibility cut-off for the pre-weaning and post-weaning multivariable models for ln ADG and ADG, respectively.

Table 5 Descriptive statistics and p values (<0.25) for Model I (pre-weaning visits) univariable linear regressions for categorical variables associated with ln average daily gain for 165 calves on 132 Kenyan smallholder farms
Full size table
Table 6 Descriptive statistics and p values (<0.25) for Model II (post-weaning visits) univariable linear regressions for categorical variables associated with average daily gain for 100 calves on 91 Kenyan smallholder farms
Full size table
Table 7 Descriptive statistics and p values (<0.25) for Model I (pre-weaning visits) univariable linear regressions for continuous variables associated with ln average daily gain for 165 calves on 132 Kenyan smallholder farms
Full size table
Table 8 Descriptive statistics and p values (<0.25) for Model II (post-weaning visits) univariable linear regressions for continuous variables associated with average daily gain for 100 calves on 91 Kenyan smallholder farms
Full size table

Multivariable linear regression models

Model I

In the final pre-weaning model, the following explanatory variables were found to be significantly associated with ln ADG: Education level and Gender of the principal farmers, Dam parity number, Calf frequency of milk feeding, Calf age at first acaricide application, Total number of calf pens, Calf mortality risk over the last 5 years and Calf age at first ad lib access to water (Table 9).

Table 9 Final linear regression model for pre-weaning ln average daily gain (model I) for 165 calves on 132 Kenyan smallholder farms
Full size table

In an interaction term, the relationship between the principal farmer’s level of education and ln ADG depended on the gender of the principal farmer. ADG was significantly lower by 31.6% for female than male principal farmers but only when primary education was the highest education level achieved; ADG was similar by principal farmer gender at higher levels of education (Fig. 3).

Fig. 3
figure 3

Interaction plot of the predicted pre-weaning ln average daily gain and 95% CI for gender and education level of the principal farmer, based on the final model of 165 calves in 132 Kenyan smallholder farms

Full size image

In a second interaction term, the relationship between dam parity number and ln ADG depended on the frequency of milk feeding. For calves from parity 3+ dams, ADG was lower by 28.1% when milk was fed twice/day than thrice/day; however, there was no ADG difference in calves of lower parity dams (Fig. 4).

Fig. 4
figure 4

Interaction plot of the predicted pre-weaning ln average daily gain and 95% CI for dam parity number and frequency of milk feeding, based on the final model of 165 calves in 132 Kenyan smallholder farms

Full size image

The coefficients of variables not involved in interactions can be interpreted directly through exponentiation of the coefficients due to the transformed outcome variable of ln ADG. Based on the exponentiated coefficient, ADG increased by 0.6% for every 1-week delay in acaricide application. For number of calf pens, the exponentiated coefficient showed that ADG increased by 5.3% for every additional calf pen. There was a reduction in ADG by 0.5% for every increased percentage point in mortality risk over the last 5 years. Delayed age at first ad lib access to water was associated with a reduced ADG, by 0.7% for every 1-week delay.

Model II

In the final model, the following explanatory variables were found to be significantly associated with post-weaning ADG: housing improvement, age and gender of the principal farmers, total number of calf pens, fed maize bran, fed wheat bran, fed calf pellets, fed hay, age at first introduction of concentrate, dam average daily milk yield in the first 6 weeks, mortality risk over the last 5 years, presence of faecal coccidia oocyst and presence of dam-calf connection by way of having adjacent housing, allowing the calf physical or visual contact with the dam (Table 10).

Table 10 Final linear regression model for post-weaning average daily gain (model II) for 100 calves on 91 Kenyan smallholder farms
Full size table

Calf housing improvement was associated with increased ADG by 0.054 kg/d among the animals on the farms that received the intervention. An increase in the age of the principal farmer was associated with a decrease of ADG by 0.004 kg/d for every additional year. The calves feeding on maize bran and wheat bran had 0.086 kg/d and 0.079 kg/d better ADG than those not fed maize bran and wheat bran, respectively. Also, calves feeding on calf pellets and on hay had a 0.204 kg/d and 0.077 kg/d increase in ADG, respectively, compared to those not fed calf pellets or hay. Delay in the introduction of concentrate to calves was associated with decreased ADG, by 0.005 kg/d for every 1-week delay. For each additional litre of the dam’s average daily milk yield in the first 6 weeks, there was an increase in ADG of 0.008 kg/d. There was a reduction of ADG by 0.002 kg/d for every increased percentage point in mortality risk over the last 5 years.

There was an interaction between dam-calf connection and the faecal shedding of coccidia oocysts. Compared to when faecal coccidia oocysts were absent, the ADG was reduced by only 0.009 kg/d when there were faecal coccidia oocysts and the calves were allowed to visually or physically connect with the dam; however, ADG was reduced by 0.089 kg/d when faecal coccidia oocysts were present but the dam-calf connection was absent (Fig. 5).

Fig. 5
figure 5

Interaction plot of the predicted post-weaning average daily gain with 95% CI for the presence of faecal coccidia and dam-calf connection, based on the final model of 100 calves in 91 Kenyan smallholder farms

Full size image

In a second interaction term, the relationship between principal farmer gender and the ADG depended on the total number of calf pens (Fig. 6). ADG was similar (slightly above or below 0.50 kg/d) by principal farmer gender when there were less than four calf pens; however, ADG increased with more calf pens, to over 0.8 kg/d with more than 6 calf pens, but only when the principal farmer gender was female.

Fig. 6
figure 6

Interaction plot of the predicted post-weaning average daily gain and 95% CI for the gender of the principal farmer and a total number of calf pens, based on the final model of 100 calves in 91 Kenyan smallholder farms

Full size image

Model evaluations

In both cases, the model assumptions of normality and homoscedasticity were met. Scatter plots of fitted values and standardised residuals did not depict distinct patterns in the distribution of residuals, with only 3 (model I) and 5 (model II) observations that were outside 2 standard deviations, and these observations were not outliers. The standardised residuals had a good fit on the Q-Q plot. A model fit without these observations had no difference from a model with all observations. Therefore, all observations were retained in the final model. The predicted residual error sum of squares (PRESS) statistic was 0.04 for both (Model I) and (model II) which is less than 0.1, suggesting that our final models were good for prediction.

Discussion

This is the first controlled field trial focussed on calf housing improvement within the zero-grazing system with smallholder dairy farms. While controlling for other factors associated with ADG in the final post-weaning model, housing improvement was associated with increased ADG by 0.054 kg/d among the calves in the intervention group, which would translate into a 5.4 kg improvement over 100 days or 19.7 kg if applicable over the first year of life. This increased weight is substantial, representing a 10% improvement, assuming calves reach 197 kg by a year of age (average weight in our study calves on their sixth visit was 188 kg). The reasons for these improvements could be because a better roof provides a drier area to lie down, and better flooring provides a softer drier area to lie down. Both of these improvements make the lying area more comfortable, leading to a longer lying down time, as found in a companion paper (Kimeli 2020), and therefore higher ADG.

We observed an overall ADG of 0.429 kg/d with a median of 0.365 kg/d during the entire study period. These ADGs are higher than 0.135 kg/d (Thumbi et al. 2013), 0.338 kg/d (Lanyasunya et al. 2006) and 0.260 kg/d (Gitau et al. 2001) that were reported from Western, Nakuru and Central Kenya, respectively. It is however slightly lower compared to 0.50 kg/d (Kathambi et al. 2018) and 0.443 kg/d (Makau et al. 2018) reported from studies conducted in the Meru region of Kenya. Generally, the observed ADG was within 0.4–0.5 kg/d which is the recommended target for smallholder dairy farms in Kenya (Lukuyu et al. 2012) but lower than 0.5–0.75 kg/d, the target for developed countries (Radostits and Blood 1985).

Our study observed a higher ADG over time with 0.324 kg/d pre-weaning and 0.510 kg/d post-weaning. This is contrary to observations by Makau et al. (2018) who reported a general decrease in predicted ADG from about 0.750 g/d at 2 weeks of age to 0.500 g/d and 0.350 g/d for heifers up to 10 and 30 months of age, respectively. Our results may be a typical case of compensatory growth in which the animals accelerate growth after a period of restricted development, mainly due to disease (coccidia infections) and limited feeding during the pre-weaning period (Hornick et al. 2000).

Our study showed that there was an interaction between the highest education level and the gender of the principal farmer (Table 9). The pre-weaning ADG was significantly lower for female than male principal farmers but only when the principal farmer only had primary education. These results are consistent with the findings of Makau et al. (2018) who reported that ADG was lower by 20.7% when the principal farmer was female, compared to male principal farmers or when the male and female were both listed as the principal farmer. Generally, in the Kenyan setup, males are considered the head of the household and can be closely involved in farm activities, if available. However, in many situations where females are the principal farm managers, it means that the male counterpart is either missing (deceased, separated or divorced) or occupied off the farm (by employment or other activities). Women with only primary education are more likely to have more children (Ouma 2006), requiring more time to manage the household, leaving less time to oversee animal husbandry. Additionally, females with lower education and more children likely have a lower socioeconomic status and resource base in order to ensure good housing and nutrition of their calves. It could also be that the women with only primary education have less means to acquire better animals. Overall, ADG improved with increasing education levels, indicating that educated farmers were likely adopting better management approaches, such as better nutrition, improved housing and disease control.

We also observed an interaction between gender of the principal farmer and total number of calf pens in the post-weaning model (Table 10), where ADG increased with an increasing number of calf pens for women while remaining the same for men. We speculate that when women were the principal farmer and farming was their primary source of income, they ensured that the calves were in individual pens for better control over nutrition, hygiene, and bullying, even though it meant more work to keep them clean and well fed because they were in multiple pens rather than all in one pen. Conversely, we observed that men would typically leave multiple animals in one pen because it was quicker to feed the animals and clean only one pen, or not put in the extra effort for good nutrition and hygiene if multiple pens were used. Group housing of calves, especially during the first 1 month of life in suboptimal hygienic conditions, has been shown to have a significantly increased risk of diarrhoeal disease, likely due to increased transmission of enteric pathogens (Curtis et al. 2016). Furthermore, on smallholder dairy farms in developing countries, there is usually a large difference in size between calves on a farm due to the small number of cows on a farm (< 10 cows per farm in our study). Bullying of young calves in group housed pens on such farms is inevitable, leading to lower ADG; therefore, individually housed calves is preferred for better control over nutrition, bullying, and hygiene. Therefore, recent research promoting the benefits of group housing of calves on large farms with cohorts of calves that are approximately the same size (Khan et al. 2016) are unlikely to apply to smallholder dairy farms in Kenya.

The influence of dam parity number on ADG depended on the frequency of milk feeding in our study (Table 9). When the parity number was higher (≥3), the ADG was higher if the frequency of milk feeding was three times or more. All farmers in this study practiced artificial insemination, but many were using local genetics on younger cows (mainly first and second lactation), which is cheaper at $5–10 USD per breeding. Conversely, imported genetics were more often used on older cows once the cow fertility is proven to warrant more expensive semen at $40–150 USD. As a result, the latter calves are likely of high potential, selected for superior performance, including growth rate, which heavily depends on milk quantity offered pre-weaning. It might also be that the higher parity cows were better producers than lower parity cows, and therefore more milk was available to feed the calf and for sale. In their study, Hoseyni et al. (2016) reported that multiparous cows had higher milk production, which was positively correlated with pre-weaning ADG.

Our study revealed that delayed age at first acaricide application was associated with increased ADG (Table 9). Ectoparasites, especially ticks, have been reported to have a significant economic impact on the dairy industry as they cause deleterious effects in cattle bodyweight through blood loss and the transmission of disease-causing agents (Parizi et al. 2012). In exclusively zero-grazing production systems in Kenya, ticks were observed in less than 5% of the calves, compared with over 80% of calves in extensive productive systems utilizing grazing (Okuthe and Buyu 2006; Gachohi et al. 2012; Omondi et al. 2017). All of our study farmers indicated that they used acaricide spray (mainly synthetic pyrethroids, organophosphate or amitraz) but at a reduced frequency of about once every 2 months, unlike the weekly use in high tick-risk areas. We speculate that delayed acaricide use meant that finances which would have been used for acaricide purchases were utilised elsewhere, including buying of feed and thus improved growth.

We observed that farms that had high mortality risk over the last 5 years had lower ADG (Tables 9 and 10). Previous studies in Kenya showed that tick-borne disease played a minimal role in overall calf mortalities (Gitau et al. 2001); instead diarrhoea was recorded as the most common cause of both morbidity and mortality, at 27% and 22% per year, respectively (Gitau et al. 1994). The diarrhoea was associated with calf care and management factors. In a study on calves in smallholder farms, Chang'a et al. (2010) noticed that about half of the cow-sheds had poor hygiene, with a poor drainage system, and flat floors with potholes. In that study, farmers frequently disposed of the dung and waste materials in the vicinity of the animal housing, which often resulted in unhygienic calf-rearing conditions. These conditions can create situations which impair calf health by overwhelming immunity that, in conjunction with poor nutrition, leads to stunted growth.

The current study revealed that ADG was reduced when calves were delayed in their first ad lib access to drinking water (Table 9). Water provision has been associated with early rumen and microbiome development, feed intake, weight gain and immune response (Senevirathne et al. 2018). Wickramasinghe et al. (2018) observed that the provision of drinking water immediately after birth could improve the growth and development of calves pre- and post-weaning, potentially by stimulating rumen development, thus increasing nutrient availability. Therefore, early water provision is of great benefit to calves and should be encouraged. Unfortunately, it is a common perception among farmers that ad lib water access to calves will lead to calves drinking a lot of water all at once, leading to hemolysis and hemoglobinuria. Young bucket-fed calves can drink excessive quantities of water if offered to them in the way they usually receive their milk because the limit of their consumption in that circumstance is governed not by satiety, but by the physical limit of the capacity of the gut (Njoroge et al. 1999). However, calves do not overdrink water when it is available ad lib from birth.

The age at first introduction of commercial concentrate supplements (e.g. calf starter with at least 18% protein) to calves had a significant impact on weight gain (Table 10). Calves that received at least a handful of concentrate early in life performed better than those in which concentrate administration was delayed. Previous reports have shown that incorporating starter diets for young calves at an early age is the best approach for improving both rumen papillae development and subsequent performance to attain a favourable growth rate (Bach et al. 2007; Khan et al., 2011; Reddy et al. 2017).

Our study showed that calves provided calf meal/pellets, maize bran, and wheat bran had greatly improved ADG (Table 10). Calf meal/pellets are commercial supplements that are high in crude protein (18–22%) but may be too expensive for some smallholder farmers. Both wheat and maize bran are milling by-products and can vary in quality, depending on the proportions of the outer skins, germ (fat-rich) and endosperm that are in the products (Mlay et al. 2005). Chemical analyses of similar cattle feed in Pakistan have shown that maize bran has 89% dry matter, 14.0% crude protein and 12.0% crude fibre, while wheat bran has 89% dry matter, 13.1% crude protein and 11.3% crude fibre (Tahir et al. 2002). These chemical analyses confirm that wheat bran and maize bran can be important alternative sources of supplemental protein for ruminants, although they are lower in crude protein than calf meal or pellets. While no other studies could be found demonstrating effects of bran feeding on ADG in calves, a feeding trial has shown that adding wheat bran to a forage diet increased dry matter intake and milk yield, and improved the energy balance (Erlt et al. 2017).Therefore, bran feeding should be encouraged if farmers are not feeding any grain to preweaned calves because they cannot afford calf pellets. Wheat bran and maize bran are inexpensive and readily available in the Kenyan market. There is, however, a need to investigate further the optimal amount of bran to feed, and what are the benefits accrued when bran is mixed with other available by-products such as maize germ, cotton seed cake and sunflower seeds.

The study showed that calves whose dams produced high milk yield at the time of recruitment had higher predicted ADG (Table 10). This result is consistent with the findings of Addisu et al. (2010) who reported a significant positive association between ADG and milk yield traits. It is likely that farmers were consistently providing more milk to calves whose dams had ample milk for calf consumption, household consumption, and sales to acquire much needed income.

Our study found that improved ADG post-weaning was associated with calves with dam-calf connections—rearing a calf in a pen adjacent to her dam with physical or visual contact with her dam (Table 10). The dam-calf connection system best mimics the natural situation and allows free behavioural expression and thus improved welfare (Johnsen et al. 2016). Although there are controversies surrounding dam-calf physical contact at younger ages due to the potential spread of infectious disease and parasites (Gebeyehu et al. 2018; Kimeli et al. 2020), our study showed increased ADG post-weaning if a dam-calf connection was present compared to calves completely secluded from the dam with no physical or visual contact, especially if coccidia shedding was occurring.

Decreased predicted ADG was observed in calves that were positive for faecal coccidia oocysts (Table 10). Coccidia are known to destroy the lining of the intestine, causing inefficiency of nutrient absorption and decreased appetite, which is detrimental to the growth (Coffin 2018). Farmers should be encouraged to consider coccidia control programs, including starter feed with coccidiostats (which are currently uncommon in Kenya) and improved calf hygiene and daily sanitation to remove faeces, thereby decreasing the environmental load of oocysts in the local environment.

A limitation to our study was that county location and intervention grouping were collinear, leading to a possible confounding effect of the county. Although there was a baseline mean difference between groups in a few of the variables (Tables 2 and 3), we believe that the groups were comparable for the study objectives since most of the important variables in the final models were not statistically significantly different between the groups at enrolment. Furthermore, the variables that were found to be statistically significantly different between the two groups at enrolment were controlled for in the models: (1) ADG controlled for body weight at enrolment; (2) number of calf pens in the model controlled for herd size and/or farm size/income; and (3) farmer education in the model controlled for farmer education level at enrolment. Randomisation at the farm level would have been ideal for minimizing the possible effects of selection bias, information bias and confounding. However, randomisation within the county was impractical in the current study since the interventions involved easily fixable housing improvement, which could be implemented by the control farmers if they visited intervention farms, which would have contaminated the control group farms, as we observed in other cow comfort intervention trials in Kenya (Kathambi et al. 2019; Richards 2017). We chose the small amount of residual confounding from county level randomisation after confounding controls in the models over the large amount of contamination from farm level randomisation observed first-hand in our previous Kenyan smallholder dairy farm trials.

Conclusion and recommendation

The growth of calves from birth to 1 year was generally within the recommended smallholder targets for zero-grazing farms in Kiambu and Murang’a Counties, but below the internationally expected performance targets. The positive effect of calf housing improvements on post-weaning ADG was significant, even after controlling for other factors in the final model, suggesting that housing factors have long-term impacts on the growth of calves. ADG was also found to have associations with management factors such as frequency of milk feeding, age at first acaricide application, age at first access to ad lib water, concentrate supplementation of various types, and presence of a dam-calf connection. Additionally, farmer traits such as education level, gender and age were associated with the growth performance. One animal-level factor was associated with ADG, coccidia oocyst shedding. We recommend that farmers should implement housing improvements, coccidia control, and early water and concentrate provision, and construct more calf pens with at least visual contact between calves and dams. Farmers and animal health professionals should be trained with an emphasis on the identified factors to ensure better calf health and growth. Further study should quantify the long-term effects of improved housing on reproduction indices and lactational performance, such as age at first calving and milk yield during the first lactation.