Simple Math Can Miss Opportunities
DETERMINING FACTORS TO CONSIDER WHEN LOOKING AT TRUCK DRIVERS AND DEAD ON ARRIVAL
By Stephanie Rutten-Ramos, DVM,
For the many farms that finish pigs, the return for all their efforts is not realizable until the pig walks into the slaughter plant. Since we have come to recognize the role that people have in the success of pig production, it seems only natural to ask about the association between truck drivers and dead on arrival [DOAs] at the plant. However, when working in a biological system, few relationships are as simple as they may seem.
Much of the value of detailed analysis comes in learning the aspects of a system that drive outcomes independent of the people in the barns. Pig populations are dynamic. The parity distribution of the sow base for any group of nursery and finisher pigs is variable. So, too, is the pig age distribution within any group, the health challenges and residual damage from pathogen exposure across lifetimes. As well, within systems, there may also be differences in genetic base and presence of certain diseases. The list goes on.
Consider the following system. In an effort to reduce DOAs at the slaughter plant, they would like to identify and eliminate the services of the driver(s) with the highest rates of DOAs. The logic is fair, but may be too simple. What if there are some factors or circumstances that predispose certain loads to DOAs? In other words, if you put a competent driver into a bad situation, is it reasonable to expect average or above average results?
Of the data that we routinely track, what factors could be associated with DOAs? Is it possible that the health challenges pigs experience across their life affect their ability to handle the stress of transport to market? Does the frequency of DOAs change across seasons? And in the absence of measuring trailer loading times, is there a proxy variable that can describe how long pigs are waiting before the vehicle starts moving?
The following data analysis looks at the associations between DOAs, truck drivers and other known factors from one year’s worth of data in a given system. The system was comprised of several sow farms, with pigs derived from six different nursery sites. Finisher groups came from twenty-five different sites and were able to be linked to nursery site, but not nursery group. Finisher groups were either single or mixed sex. For each finisher group represented, in-barn mortality rate was ranked. Group mortality rate was evaluated as a risk factor for DOAs if it fell in the system’s top third (>5.1%). Truck loads came from either a single finisher group or more than one [split load]. Ship date was assigned to quarter of the year [1, 2, 3 or 4] and being occurring in June-August [summer]. Since truck drivers were the primary variable of interest, only those with 10 or more loads were included. Sixteen truckers were represented. All pigs were hauled to the same packer. More than 180,000 pigs and 1,300 shipped ticket events were represented in the data set.
Overall, this system averaged 2.62 DOAs per thousand head shipped (SD=6.24/1000, minimum=0/1000, maximum=91/1000). DOAs occurred in nearly 24% of shipped ticket events. Loads derived from finisher groups of one nursery source were more likely to have one or more DOAs than all other nursery sources (odds of any DOAs from that nursery=1.74, 95% Wald Confidence Limits: 1.15, 2.63). One or more DOAs were more likely to occur in loads that were not split between two or more barns (odds of any DOAs in an un-split load=2.86, 95% Wald Confidence Limits: 2.15, 3.76). Any DOAs were less likely to occur in loads derived from groups in the highest third of system mortality compared to those with mortality in the lower two thirds (odds of any DOA from an upper mortality group=0.66, 95% Wald Confidence Limits: 0.49, 0.88). There was no association between time of the year and DOAs in this dataset.
Besides the using the odds ratio as a tool to evaluate associations between risk factors and the occurrence of DOA events, we can also use analysis of variance to estimate the long-run expectation for DOAs given a particular group attribute or driver. The merits of using a more formal analysis instead of simply crunching numbers hinge on the ability to simultaneously account both for sample size and variation (repeatability) and multiple sources of variation. In other words, one bad haul by an otherwise good driver will not a bad record make. Just as importantly, however, we can estimate the effects of different factors on the rate of DOAs.
In this analysis, we are able to account for nearly 25% of the variation in DOAs per thousand head shipped with variables describing if the load was from more than one barn [split load], if source group mortality ranked in the system’s top third, if finisher groups originated from a specific nursery, and the truck driver, including the driver’s performance within particular finishing sites.
Adjusting for the known sources of variation in the dataset, we would estimate that split loads average 1.03 (SE=0.25) fewer DOAs per thousand shipped than un-split loads. Groups with the highest third of in-barn mortality average 0.89 (SE=0.3) fewer DOAs per thousand shipped than those with mortality in the lower two thirds. Pigs derived from one nursery were estimated to experience 3.5 (SE=1.0) more DOAs per thousand shipped than from all other nursery sources. As well, three truck drivers stood out as having a predictably higher rate of DOAs compared to the rest of the drivers, with rates of 8.6 (SE=1.5), 5.1 (SE=0.9), and 5.0 (SE=1.0) DOAs per thousand shipped.
While a simple evaluation of DOAs likely would have identified the same drivers as having extraordinarily high rates of DOAs, it most likely would have missed identifying those other areas for system or site-level interventions. Sadly, when those factors bigger than the individual person are overlooked, any opportunity for system-level improvement is lost.
Dr. Stephanie Rutten-Ramos received her DVM and PhD from the University of Minnesota and is an independent consultant. To contact her, email email@example.com.