Animal & Veterinary
Presentation: Risk Modeling Biological Hazards
by Dr. Gerald L. Rushin
DR. RUSHIN: Great. ‘Evening, everyone. I guess before I get started, I would like to know: Do you want to take a break now or after my presentation. By a show of hands, do we want to take a break now? For “now,” raise your hands. For “later,” leave them down. All right, looks “later” is going to work. Great.
Did you guys get everything that Phares said? I hope you guys did. If not, with my presentation, we are going to do a lot of repeating, so we will get it. I will take it a little bit slower. I have the tendency to talk fast. I guess in me -- I am a D.V.M. J.D., so I will try to keep that under control for you guys.
Once again, my presentation is going to be the risk modeling biological hazards. I like to move around -- I don’t like standing behind a podium. It is not just not my personality type. So I will be all over. But we will get there.
What I am going to do is -- the purpose of my presentation is, one, I am going to describe the model. Actually, I am going to actually demonstrate the model. I am going to rate seven biological hazards for the starter swine, the grower swine and the finisher swine.
Microbials I am going to rank this afternoon for you guys are going to be E. coli, Salmonella typhimurium, Clostridium perfringens Type C, Bacillus anthrax, Staphylococcus aureus, Pseudomonas aeruginosis, and Listeria monocytogenes.
Once again, remember what Phares risk. Our proportional risks are health consequence score times our exposure score divided by a constant. To get our health consequence score, we have our severity of illness score times our likelihood of illness score.
To generate our severity of illness score, we need our 1A, our organ system score times 2B, our severity of illness signs and symptoms score times 1C, duration of signs.
To generate our likelihood of illness score, we need 2A, the median infective dose, 2B, the pathogen virulence, 2C, survivability and recoverability.
And to generate our exposure score, we need 3A, the prevalence of pathogen, 3B, effect of manufacturing process, 3C, the effect of post-processing safety control system, 3D, the proportion of population consuming the product, and 3E, the frequency of consumption.
Once again, severity of illness is 1A, the organ system, 2B, severity of illness sign and symptoms, and 1C, the duration of signs.
Assumptions -- to achieve the objective of the risk model, certain assumptions had to be taken. And as I go through my presentation, we will discuss some of the assumptions that we actually had to take.
For example, one assumption, as Phares discussed, was: Assume that 90 percent of the cases will occur in this organ system for the 1A. If you notice here, we have choice of digestive system, the immune system, the neurological system, musculoskeletal system. Unfortunately, on the slide, I was only able to get four of them up -- there is a total of 7.
If you were to choose digestive system, you are going to get a score of 1.000. For 1B, the severity of illness sign and symptoms, once again it is another assumption. Assume there are degrees of disability caused by disease and there are clearly defined clinical signs in our symptoms. You have minor, which rarely requires veterinary intervention. We have mild -- it sometimes required veterinary intervention. And we have moderate, requires veterinary intervention in most cases. And we have severe, and actually severe, it usually causes death, and this is one that is selected. By choosing severe, we will get a score of 10.
Now we are at 1C, the duration of illness. Once again, we assume that a patient was normal at the time of exposure and reasonable care and treatment was given. We have a choice of 0 to 1 day, 2 to 4 days, 5 to 10 days, and over 10 days. Now, you choose 1, 0 to 1 days, we will get a score of 0.0011.
Now we are able to determine our severity of illness score, which is our 1A, our organ system, score 1, times our 1B, our severity of illness signs and symptoms, which is a score of 10, times our 1C, the duration of signs, which is a score of 0.0011. You multiply these factors together, this is going to give us a --- severity of illness score of 0.011.
Now we just figured out severity of illness score. Let us go ahead and figure out how we are going to figure out likelihood of illness score.
Once again, for Number 2, the likelihood of illness score, we start off with 2AS, the median ineffective dose, 2B, the pathogen virulence, 2C, survivability and recoverability. Once again, it is another assumption -- we have to assume the organisms are typical. We have a choice of, moderate and low for a median ineffective dose. If the user chooses high, this is going to give a score of 0.1. For 2B, the pathogen virulence, disregard the notion of unusual virulent factors and also just assume the species is not antibiotic resistant. We have a choice, for our pathogen virulence, of low, moderate and high. The user chooses high, and this is going to give a score of 25.
For 2C, the viability and recoverability, this means that a certain number of organisms survive processing and they recover to a level approximate the original level in the feed are somewhere between survival and original level. This is what Phares explained to you earlier.
The user has a choice here of high, moderate, low, very low or negligible. If the user chooses moderate, it is going to give a score of 0.1.
And guess what? We are now able to determine our likelihood of illness score. We are going to multiply 2A, our median ineffective dose score of 0.1 times 2B, the virulence score of 25, times 2C, our survivability and recoverability score of 0.1. This is going to give us a likelihood of illness score of 0.25. Everybody with me? Didn’t lose anybody? Great.
Now we will determine our health consequence score, which is comprised of severity of illness score 0.01 times likelihood of illness score of 0.25. We will multiply these two numbers together and that is going to give us a health consequence score of 0.00275.
Once again, remember, or proportional risk is our health consequence score of 0.01 times 0.25, so it is going to give us this number at 0.00275.
Let us determine how we are going to determine our exposure score. Once again, the likelihood of exposure score is determined by multiplying 5 factors. The first is 3A, the prevalence of the pathogens times 3B, the effect of feed manufacturing process, times 3C, the effectiveness of the post-process safety control system, times 3D, the proportion of population consuming the product, times 3E, the frequency of consumption.
For 3A, the prevalence of pathogen, for animals we have to assume that there is contamination of raw material. For here, we have a choice of negligible, which is nearly 0 percent, rare, which is 0.1 to 1 percent, infrequent, which is 1 to 10 percent and sometimes 11 to 50 percent. If we choose rare, this is going to give us a score of 0.0001.
For 3B, this is the effect of manufacturing on the feed pathogen level. This category is going to assess the capacity of the feed manufacturing process to reduce or increase the number of pathogens from the levels prior to manufacturing.
Here we have a choice of eliminates, usually eliminates, slightly reduces, no effect. Here you can notice there is actually more -- we have increased and greatly increased, but unfortunately we couldn’t get that on the slide.
If you use -- choose slightly reduces, you are going to get a score of 0.001.
For 3C, the effectiveness of the post-process safety control system, this category assessed the degree to which the post-processing safety control system maintains the levels of pathogens in the feed following manufacturing.
Here we have a choice of in control, which is the effective system in place. We have partial control, which is a partially effective system in place, and out of control, which is an inadequate system is in place, or severely out of control, which is gross management occurs.
If you choose the in control, this is going to give a score of 1.
For 3D, the proportion of animals consuming the feed formulation, this is going to estimate the percent of animals consuming the product. We have very few which is less than 5 percent. We have some which are between 6 and 25 percent. And most, between 26 and 75 percent. Or very large, 76 to 100 percent.
If the user chooses very large, this is going to give a score of 0.1.
For 3E, this is the frequency of consumption. The feed for animals, we are going to ad lib consumption, which is average consumption. And this is going to give us a score of --- of 10.95.
Now, we are able to determine our likelihood of exposure score. So we have our 3A, which is going to be 0.001, times our 3B, which is going to be 0.001 also, times our 3C, which is going to give us a score of 1, times our 3D, which is going to give us a score of 0.1, times our 3E, which is going to give a score of 10.95. You multiply all these together, it is going to give us a likelihood of exposure score of .00001095.
Now, we will determine our proportional risk. Remember, we have a health consequence score, a severity of illness score, times our likelihood of illness score. So we have a health consequence score times our exposure score, which is .00001095, divided by the constant, which is going to give us a proportional risk of 1.66 times 10 to the negative 18. That is very ---. This number is then scaled for our risk score. It gives a risk score of 73. Everybody catch up with that? We are still together -- great.
Okay, so why don’t we go ahead and actually use the model for E. Coli in starter swine. For the model, our organ system is going to be the digestive system. Severity signs will be mild. Duration of signs will be 5 to 10 days. For likelihood of illness score, 2A for infective dose, will be moderate. For 2B, our pathogen virulence will be moderate. For 2C, our survivability and recoverability will be very low.
For 3, our likelihood of exposure score -- our 3A, prevalence of pathogen, will be sometimes. Our effectiveness of feed manufacturing process will be usually eliminates. For 3C, the effectiveness of post-processing safety control system, will be in control. For 3D, the proportion of the population consuming their product, will be very large. And our frequency of consumption will be ad lib.
What I will do with that -- I will pull it up. Remember, as I said before, it is going to be -- if you notice here, you can see the whole model itself.
If you go with the digestive system, we have severity, it is going to be mild. And we are going to have -- 5 to 10 days. This is going to give us a score of 0.01. We are going to have median infective dose will be moderate. Pathogen virulence will be moderate. --- select. We have very low, which is already selected.
Go down for presence of pathogen in the feed, we have sometimes we have selected. We are going to have usually eliminates for 3B. We are going to have in control and we are going to have very large and we are going to have ad lib consumption. And this is going to give us a score of 1.66, 7 times to the negative 21, for our personal risk score. If you move it over here to our scaled risk score, it will give a score of 69.
Now, let us see if you play with some of these numbers, what is going to happen. For example, if we go to usually effective of manufacturing processing, I would say eliminates, which is nearly 100 percent. This gives a scaled risk score of 55, so you see I actually lowered it.
So this is the model itself.
Now, what we end up doing, we went through all organisms with the actual model itself, and this is what we end up coming up with.
Let me go back to the presentation again, and we will start with the rank for our starter swine, for the starter ---. If you notice, we have the rank -- we have Listeria at 74, Clostridium is 73, Bacillus is 72, Pseudomonas is 66, Staph aureus is 64, E. Coli is 58, and Salmonella typhimurium is uncommon in starter swine due to pathogen immunity.
Look at our grower swine rank. We have Listeria at 78, Bacillus also at 78, --- exact number. We have Staph at 76, Pseudomonas at 75, Clostridium at 73, Salmonella at 66, and E. Coli is uncommon in grower swine due to immunity.
We Bacillus at 78, Staph at 76, and Listeria is 76, so exactly the same number. The Pseudomonas is 75, Clostridium is 73, Salmonella at 67, and E. Coli is uncommon.
These are the references that were used to help design a model and some information.
Thank you, and have a safe trip home. Great. Everybody get all that?
Any questions at all?
You guys are ready for a break? Yes?
MR. LITTLE: Dan Little, Brookings, South Dakota. Two questions. One is, have you been able to validate this in any way? I mean, there are a lot of neat calculations, but have you actually been able to validate it against diseases?
Secondly, as you look at this, there doesn’t seem to be anything that takes into account the resilience of the host, and so when you put this into a production system, will it really hold up because you have varying levels of immunity, environmental conditions? I mean, you know the list.
DR. RUSHIN: Right.
MR. LITTLE: Won’t that just blow this away?
DR. RUSHIN: I guess -- this is a group effort. I guess we can all sign in.
DR. RUSHIN: Phares, I guess some this we discussed to design a model. And can you repeat your first question again?
MR. LITTLE: On the validation. I guess, have you actually taken naturally occurring problems and validated it against the model?
DR. RUSHIN: Not yet. We have not done that. And your second question was?
MR. LITTLE: In terms of the host, the whole issue of the host, because you have stacked up all of the risk factors but you haven’t considered the host resistance, which in real life is the main issue. I mean, even as of intake, the animal may stop eating a very toxic substance and then not take as much in as something that is less toxic that they eat more. How do you control for that?
DR. RUSHIN: You want to answer that one, Phares?
MR. OKELO: You might recall that there were some assumptions. And we are assuming average host resistance. I mean, average healthy animal and average healthy human being, so that takes care of the host resilience.
DR. RUSHIN: As you can see from both these presentations, there are some assumptions we have to make because it is not, like we said before --- (away from microphone).
DR. GRABER: I think the first part of the validations is going to be hugely helpful. I think the validation will be hugely helpful in figuring out whether we have made the right assumptions. And again, you know, we are going to have a hard time getting that data, but we will try.
MR. DZANIS: Yes, I have a question. You went through pretty well how you calculated the proportional risk, but that leap from proportional risk to the scaled risk score, I mean, what is in the spreadsheet there that automatically adjusts to that? Is it compared to everything else that is already in the system?
DR. RUSHIN: Remember, we divided by (away from microphone).
MR. DZANIS: Okay.
MR. OKELO: We can give you details, if you like.
MR. DZANIS: Okay, thank you.
MR. OKELO: Any other questions?
DR. McCURDY: Give you 10 minutes for the break -- okay? No, 10!
(Whereupon, a break was taken from 2:23 to 2:43 p.m.)
DR. McCURDY: I have an announcement to make. Dr. Hooberman’s next two talks -- the hand-outs are not in your package, so Dr. Graber has hand-outs for you on these next two talks if you want a copy, okay? Dr. Barry?