• Decrease font size
  • Return font size to normal
  • Increase font size
U.S. Department of Health and Human Services

Animal & Veterinary

  • Print
  • Share
  • E-mail

Section Contents Menu

Safety & Health

Presentation: Risk-Ranking Model

by Phares O. Okelo

MR. OKELO: Good afternoon. I am going to be talking for a few minutes, describing the risk-ranking model that we are developing for ranking the risk posed by hazards found in feed.

(Slide)

Just a quick definition. The model -- in the model, the risk score is composed of two parts, the health consequence score and the exposure score. And these two are multiplied together to get the risk score.

(Slide)

To give you a little background as to how we are going to be setting up the scales and the weight for the variables in the model, we observed that trends in biological systems are exponential in nature, and so we are going to select increments that are in multiples of 10 for the variables. And as we may notice in this graph, in the vertical scale, we have the probability of infection, and in the horizontal scale, we have the dose of the pathogen.

If you look at the different pathogens represented by the solid line and the other lines, you notice that these are an exponential portion of the curve, and this is what we are talking about when we say that biological phenomena are exponential in nature. And so we use that knowledge to set our scales and the levels, the scores, that we assign to the variables in increments of 10, and that would give us the exponential portion of the responses.

The reason we want to set it up like this is so that we could use log transformation to convert the exponential trends into linear trends, and that is a result that we would like to have in modeling.

We also know that log transformation condenses sparse data points and also disperses dense data points, and that is also a result that we would like to have in modeling.

(Slide)

Now, we break down the health consequence portion of the model into the severity of illness score and the likelihood of illness score. And in the next few slides, I will describe further what severity of illness score and likelihood of illness score refers to.

(Slide)

By severity of illness score, we have selected three different variables. That is, the organ system score, and the severity of signs and symptoms score, and then the duration of symptoms score.

By organ system score, we are talking about if it is the respiratory organ system that the disease occurs in, rather, if the respiratory system is the one that is mostly affected by the hazard, or is it the digestive system that is mostly affected, we are saying that the organs, the different organs, would carry different weights. And we start with the least important organ system and we assign those scores that are in multiples of 10 till we get to the organ system that would carry the highest score.

(Slide)

Next, we go to describe the severity of signs and symptoms score, and by that we are referring to how severe the sign or symptom is when you see it. In this case, we see pigs that have diarrhea. That is the digestive system. The digestive system organ system would be the one that would be affected. Now, the digestive system would be the sign that appears in the digestive system. And again we assign the severity levels -- we assign scores to severity levels ranging from the lowest to the highest, and those would be in increments of -- that are in multiples of 10.

(Slide)

Next, I want to describe a little bit the duration of signs and symptoms. By that, we are simply referring to how long the sign or symptom lasts under general, reasonable care to the animal or human.

Just to summarize what severity of illness score is referring to, we have three variables. The organ system score would range from, say, the renal system being assigned the lowest score of 0.001. The next organ system would be assigned a score of .01. And so they would be increment of -- that are in multiples of 10, until we get to the highest score in the range, which in this case would be the cardiovascular system.

The same thing is done for the severity of signs and symptoms score. You have the minor severity being assigned a score of .01 in increments of -- multiples of 10, again, until we get to the most severe sign or symptom score, which would have a score of 10.

Duration -- if you have the least, which is the sign or symptom lasting less than 1 day, would be assigned a score of 0.0011, and the highest score would be assigned a score of 1.1, and that would be if the sign or symptom lasts right around 10 days under normal, reasonable care to the animal or human.

(Slide)

Now, I need to mention here now that we break into the chemical hazards, and then later on I will come back and talk about the biological hazards because we -- the score goes slightly differently.

I am going to be talking about the likelihood of illness score in -- for chemical hazards. We have 2 variables, the acceptable daily intake score as a metric for estimating the likelihood of illness score, and by acceptable daily intake, if we look in this graph, for example, you have 2 chemicals, A and B, and they show different responses when you change the dose, you have the response in the vertical scale.

We notice here that to get the same level of response, you require different doses of the two chemicals, and if that is the level that is acceptable, then we would consider Chemical A to be more lethal than Chemical B because you require only a small dose to cause the same response.

So we would say in general that the chemical that is close to the same level with only a little -- with a lower dose, has a higher likelihood of causing illness.

(Slide)

Then, the second variable of the likelihood of illness score is the agent stability score. Now, that simply refers to the hazard’s ability to maintain its toxicity, and you recognize here we are talking about chemical hazards, so its ability to remain toxic under unfavorable conditions.

(Slide)

Again, to summarize, we would score them on a scale going from high to low of .01, .1 to the low dose, the low acceptable daily intake would take the highest score of 10. Now, the reason for this is if you need a great amount to cause illness, we consider that chemical hazard would have a lower potency, and that is why we would assign it a lower score. And if you require only a little bit of it to cause the same level of illness, we consider it more potent and it has a higher likelihood to cause illness and therefore we would assign it the greatest score.

(Slide)

For agent stability, if the agent has a low stability, we would assign it the lowest score of .25 and the highest score would be a 25. Now, the reason for the numbers being different there, in increments that are multiples of 10, but the numbers are selected so that the weight -- and I will show that a little later, but they are selected such that the weights for the acceptable daily intake variable and the agent stability variable sort of balance each other. So we select them such that they carry different weights, or about equal weight for the likelihood of illness score.

(Slide)

Next, we move onto the biological hazards, the likelihood of illness scoring for biological hazards. And if we look at this graph on the right-hand side, we have on the horizontal scale the dose of pathogen, and in the vertical axis we have the response. In this particular case, it is the probability of infection.

Now, we notice that for the first pathogen that a lower dose is required to cause the same level of illness, about 50 percent, and a greater dose is required of the second organism that is represented by the axis. And so we consider the organism that requires the lowers median infective dose to cause illness as being more potent and therefore having a greater likelihood to cause illness.

(Slide)

For pathogen virulence, we are talking about the power of the pathogen to cause illness, and, again, we go from the lowest virulence being assigned the lowest score. In increments of 10, we move to the highest virulence, which would have the highest score.

By pathogen survivability and recovery, which is a factor for likelihood of illness score, we are talking about the ability of the organism to survive unfavorable conditions, to survive during processing or storage, depending on what the conditions are in the feed.

In summary, the way that we assign scores to those factors is you have a high dose. The highest dose would be assigned the lowest score of .1 and the low dose would be assigned the highest score of 10. The reason, again, is because you require a low dose to cause illness compared to the high dose in a different organism to cause illness. We consider the one that requires a low dose to cause illness as being more potent and therefore having a greater ability to cause illness.

For pathogen virulence, we go from low to high, low being assigned the lowest score. And you notice we have the same numbers for the chemicals as well as for the biological hazards.

The pathogen survivability is scored the same way, but you notice, again, that negligible is assigned a very small number. We just would be avoiding assigning it a zero because the model is multiplicative -- that is, we multiply the numbers across to get the final risk score, and if we had a zero in there, it would knock out everything. But negligible survivability would be assigned the lowest score and a high survivability and recovery would be assigned the highest score of 1.

(Slide)

Then, there is the last aspect of our risk score, which is exposure. We have identified five factors to use as predictors of the exposure score. By exposure, we are talking about the pathways by which the hazards get into the feed and then to the animal.

In this particular diagram, we see different activities that are involved in making feed. So if you have feed coming into the mixer, we are saying: How does mixing affect the level of hazard, whether chemical or biological? How does it affect the amount of hazard in the feed?

We have other processes -- extrusion -- and then you have some drying and cooling. How do those processes affect the levels of the hazards in the feed? And that is what we assess in trying to score the exposure score.

The first factor of the five is the level of chemical hazard in the feed, and that would be just the amount in nanograms per kilogram or whatever unit is used. Or, we are looking at the microbial hazard and its prevalence in the feed, and that would be assigned a score ranging from “negligible” to when all the samples are contaminated. And for microbial hazard, we are talking about just whether or not it is positive in the feed or negative in the feed. So the prevalence is just percent positive in the feed, and we assign them -- we will show that later on in some other slide.

Then the 2nd factor is talking the effect of feed manufacturing, and those would be assigned scores ranging from the lowest, which again you may notice is a small number but not zero, but we make it so small that if the manufacturing process eliminates the hazard in the feed, then it reduces the risk, or the exposure score, significantly. And then the greatest score would be if it greatly increases the level -- if the process, the manufacturing process, greatly increases the level of the hazard in the feed, we would assign it a score of 6.

Effect of post-process control system is just what is done after the feed is manufactured before it is given to the animals or before -- if it is food, before people eat it. And, again, if the system is in control, we assign it the lowest score of 1, and if it is severely out of control, we assign it the highest score of 100. And the scores in between would be in increments that are multiples of 10.

The 4th factor is the proportion of the population consuming the feed, and so exposure would be proportional to that proportion. That proportion would be just a percentage. And, again, very few -- say, between zero and 5 percent -- of the total population would be assigned a score of 0.001. And a very large proportion consuming the feed would be assigned .1. And in between, there would be increments that are in multiples of 10. And this would be the case for both chemical as well as biological hazards.

Then the last factor would be the frequency of consumption, and this is purely based on -- if it is monthly, that is assigned a score of 12. That is, you are eating once a month. And the whole factor is on the basis of annual, so how much is consumed per year. If the animal consumes the feed just once a month, whatever the product is only given to the animals once a month, that would have a score of 12, but -- and then --- would have a score of 1095. That 1095 is based on the 3 times a day, and if you calculate that for 365 days a year, you get 1095.

(Slide)

All of those scores are multiplied. There are 3 factors for severity of illness and 2 factors in the case of chemical hazard, but 3 factors in the case of biological hazards, and then 5 factors for exposure scoring in both chemical as well as biological hazards. Those are multiplied together across the board to get a final proportional risk score, but we -- the proportional risk score would just be a number, and for each situation, because we are doing risk ranking relative to each other, the number would be compared to -- for 1 pathogen or chemical hazard would be compared -- the numbers would be compared to each other to get the --- risk rank.

Because the proportional risk score would be a number that may not have any physical meaning, we propose to scale that between a minimum number of 1, which would represent the least risk, least relative risk, and a maximum of 100, which would represent the greatest risk. The reason for scaling it between 1 and 100 is that that is a familiar scale and it may be possible when the model is relegated well to, for example, interpret 50 as being halfway between 0 and 100, or 75 being about ¾, but we want to emphasize here that this is all relative. So if one scaled risk score is greater than another, the rank would be -- if you are comparing 2 organisms, then the rank would be just 1 and 2, as long as one number was greater than another. So we are not looking at magnitude so much. But that is the reason for scaling it between 1 and 100.

(Slide)

Then, once we develop the model, we present it in an Excel -- on an Excel platform where the user would be able to go in for chemical hazards and make a selection. For severity of illness, they would make a selection: What organ system is involved? And under the many different parts, all the variables that we have talked about, the user can go in and make a selection based on the scores that have been assigned and finally get a score, once they have made all their selections.

In some cases, they would be prompted to enter some specific numbers. In the case of chemical hazards exposure, you might remember that you could measure the level of the hazard in the feed in nanograms per kilogram. Then the model would at some point prompt you to enter that number, and in the end you get the final scaled risk score using that user-friendly platform. And we will have some example later on.

(Slide)

The same thing would be done for biological hazards. You have the user-friendly platform to make your selections or enter information as prompted by the model.

Thank you for your attention. Are there any questions?

(No response)