Retail Meat Reporting
by Dr. Elvira Hall-Robinson
DR. ROBINSON: Good afternoon. For those of you who do not know me, my name is Elvira Hall-Robinson, and I am the manager and coordinator for the retail meat database activities. Today I will be talking to you about the retail meat database, the current reporting of all the NARMS Arms, and also some future plans for a more collaborative report.
First, I would like to start with the retail meat log sheet. After the sample is actually collected, as Terry mentioned, each meat type has a separate log sheet that is submitted from the FoodNet sites. This logsheet is sent to the Office of Research monthly or quarterly electronically. In addition, the original log sheet is Fed Ex with the isolate to the Office of Research.
Part one of the log sheet is the demographic information on the meat sample, and you have the sample I.D., the store name and city, the brand name, the lot number, whether the product was cut or ground in the store, the sell-by date, the process date and the lab process date.
Part two of the log sheet is where the isolate information is recorded from the meat sample, and it is recorded as a growth of yes or no. Once it is recorded as "yes," the sites assigned isolate I.D. number that is identical to the sample ID except a C for Campylobacter, “S” for Salmonlla, “EC” for E. coli and “E” for Enterococcus is added the sample ID number to create the isolate ID number.
We have developed protocols for the sites on guidance for completion of this log sheet, for shipping of the isolates, and a methods protocol, so that the information from site to site is consistent.
That was the log sheet used from 2002 to 2004. In May of 2005 a new log sheet was created so that the brand name and plant name information would remain at the site. We also added on the log sheet an organic column, and the reason we added this field which is not part of the sampling plan, but we noticed that some stores were organic that we were not a where of and they ended up on the grocery store chain guide list that Terry mention. This is column was added to distinguish this type of meat.
We also assigned a code list to 173 brand names so that we would not have misinformation. This is information is actually going to be at the site. So, prior to the site sending us the log sheet they remove brand name and establishment number, and we do not have that information anymore.
Now, the database that we used to mange the data for the retail meat is Access. The reason that we chose Access is it’s component of the Microsoft office suite that is loaded on all the CVM computers, It will allows easier exchange of information between Word, excel, adobe, and SAS, Access is user-friendly, we can get timely reports as the data is entered. Disadvantage is the security and the number of users using the DB at one time.
This is an example of the USDA’s front page. They use Access as their front page and also use Access as their back end.
This is an example of CDC’s front page, and CDC uses Access as their front page, but they use SQL Server as their back end. With the retail meat, we use Access as our front end and back end, but looking at using Oracle or SQL as our backend and Access as the front end. I think both USDA and FDA are looking at using another relational database as their back end because of the power.
This is actually an example of a model that we could use, as to how the data for the three could be brought together. One thing about relational databases is that you can bring other partners’ tables in. Okay? As you see this model, this actually was developed by John Stelling, which Paula mentioned in her talk when she talked about reporting. We have a data analysis for USDA, a data analysis for CDC and a data analysis for FDA.
There are tables in the back end of this that actually links to those particular tables from the three arms. You don’t have to integrate the actual data, but you can have their tables in it to create reports that are very similar.
Now, I would like to talk about how the data for the retail meat is actually managed. I am going to through our database and then talk about reporting, as far as what is the current reporting with all three arms, and then go into what is in the future. Actually, that Paula and Tom has already mentioned.
Again, this is our front page and we have three buttons. We have the data entry button, search data and the reports button. As far as the data entry, the fields are very similar to the monthly log sheet that I went over prior to this.
The information from the log sheet is just extracted as it comes in electronically. So, once it comes in, it actually gets entered into the database, which means we can get prevalence data out very quickly.
Again the brand names are going to be changed from a name to a code. We will not be recording any information on the lot number field.
Now, the second part of the log sheet, as I mentioned, was part two, which is the isolate information, and that is here. What happens is the isolate I.D. is actually extracted down here into the database, and once the site recovers the bacterium, then they actually put in the yes or no and an Isolate ID is entered.
Once it becomes a yes, the sites actually assign an isolate I.D. number. This isolate I.D. number is very similar identical to the sample ID except if Salmonella is isolated we put an "s" on the end of it, for Campylobacter we put a "c" and for E. coli we put an "ec", and for Enterococcus we put an "e" on the end of the isolate ID number.
Once the isolate comes into the Office of Research, it is assigned a CVM number, which is going to be changed to what we call a NARMS number. We are calling it a NARMS number instead of a CVM number. What that number is will link the sample data and isolate data to the MIC data.
Here is the actual form for the MIC data. Currently the data that we are collecting is the susceptibility method, the drug, the MIC sign and also the MIC number goes here and then we determine whether the isolate is Susceptible, Intermediate or Resistant.
Now, the second part of the database is the search data, and we can search the data by state, sample I.D. and we can also search by CVM number. If someone called up and said can you actually find this isolate in the database, what is going on with the resistance, we can actually find it very quickly.
The last part of the database, the last button, is the reports button, this button is linked to our reports page, and there are different types of reports on this reports page. We have formal reports, there are bar graphs for MIC distributions, there is a percent positive sample tables, pivot tables and then there is percent resistant pivot tables that we can convert to a pivot chart.
The main buttons that we really use a lot is the percent positive samples that we provide on the quarterly conference call with the FoodNet sites that go through CDC. The other button we use a lot is the percent resistant button.
Here is an example of the percent positive samples by state and meat product for Campylobacter and Enterococcus for 2002 and 2003. This is an example of that, and that is in your book.
The next button that we use I mentioned before is the percent resistance, and this is the pivot table for the percent resistance. We can convert the pivot table to the pivot chart from that pivot table, and we can just hit a button and it actually we can get that chart.
So, the blue bars are your percent susceptible, the burgundy bar is your intermediate susceptible and the yellow bar is your percent resistance.
As far as the current reporting published on the web, both CDC and USDA published reports in 1997, as they mentioned. USDA has published a different variety of formats for their reports and they have some data for 2003. CDC published similar reports from year to year, from 1997 to 2002, and they were published, as Tom mentioned, this summer, their 2003 report.
As far as retail meat, we have published the 2002 annual report last year, and it looks like we are going to be publishing the 2003 report this summer and then hopefully, at the end of the year publishing the 2004 data.
As far as all of the reports, most of the reports cover all or most of this; the interpretive criteria, the frequency positive samples, the frequency of the percent resistance for antimicrobials, also the MIC bar graphs, MIC distribution, multi drug resistant tables and multi drug resistant patterns. In addition, the retail meat for the 2002 report we published PFGE patterns for Salmonella and Campylobacter.
As far as future reporting, as Paula and Tom mentioned, FDA, USDA, CDC and CIPARS met in March of this year to discuss comparing reports of all the three U.S. agencies. During the meeting, we discussed an overview of our databases. We actually share databases; so that we could actually share queries, we can share coding and all of that.
We also discussed developing a new report and, as Tom mentioned, the level of reporting. Should we be looking at drug versus drug class or both? How do we define multi drug resistance, dichotomization? As Tom mentioned, looking at susceptible and non-susceptible resistance.
Also, putting confidence intervals into the percent resistance tables and modeling the Danish and the CIPARS reports. In the meeting we discussed strategies for presenting and coordinating the reports. We talked about executive summaries, web based tables, graphs and reports, as Tom mentioned, and paper-based reports.
One deliverable we hope to have is a collaborative report that is going to be published in the MMWR in 2006 for the 2004 data.
After the meeting we the USDA and CDC programmers John Stelling, Terrell Miller, Tom Chiller and Lauren Stancik came up to the Office of Research and we actually decided to put together a report layout of what a collaborative report will look like. The NARMS working group meeting that occurred in May we decided to present this and this is the percent positive resistance for Campylobacter with the three agencies: CDC, FDA and USDA.
I only actually put this out to Tetracycline because the chart was too large, it would not fit.
And that is all I have. Are there any questions?
DR. KOTARSKI: I have a question as a point of clarification. You had mentioned early on, in terms of the data entry, there was a column or an entry regarding whether or not the meat was cut in the store or not or cut or ground in the store, and they would answer yes or no. What is the significance of that entry?
DR. HALL-ROBINSON: That is a good question. Before I came the log sheet was actually created. But I guess the importance is to know how a product is handled as far as the process of the product itself, if there a more contamination rate based on how the product is handled and processed. Your pork chop is cut in the store, and usually the hamburger is ground in the store. That is routinely done.
DR. KOTARSKI: Okay. And that is reported out?
DR. HALL-ROBINSON: We have not actually published any data on that at this time.
DR. KOTARSKI: Okay. Is there any reason why you have decided not to report whether it was cut in the store or not?
DR. HALL-ROBINSON: We haven’t even talked about it. There is no reason why we don’t publish it or not publish it. We haven’t discussed it.
DR. KOTARSKI: Thank you.
DR. McEWEN: Maybe Terry is the one for this question or maybe you can take a stab at it, Elvira. When it comes to integrating the information that you get from the human side of things and retail and also slaughter I guess, if you -- the tendency will be, I guess, to see if we have something pop in an area. In people, let’s look at the retail and see if we have a comparable sort of phenomenon there.
The tendency would be to treat them as kind of catchments areas I guess and the assumption being perhaps that what is in the retail stores will reflect human exposure, peoples’ purchasing practices in those areas and exposure. If you sort of go further I guess, another assumption might be that what goes in through the slaughter houses through that region will turn up in retail.
I wondered to what extent you have given the thought to investigating the sort of food distribution system to see how it reflects that sort of catchment area?
DR. HALL-ROBINSON: I don’t think we have actually looked at the food distribution, but I can say that as far as when we were collecting the information on the plant that some product for example is process in Indiana may be distributed in Michigan, from the same processing plant. So we haven’t, looked at that information at this time.
DR. WALKER: I think an example of that is when we had that first BSE positive cow in, I think, it was Washington or Oregon. How many states had the meat from that animal been distributed to?
It is a very good question, but I think it would be extremely difficult in this country because of the marketing practices.
DR. HALL-ROBINSON: Products can get to California very quickly, within hours or days.
DR. KOTARSKI: Another question or point of clarification. So the sampling strategy that you have currently for 10 states now, that reflects human populations in terms of exposure? Or does it reflect the number of stores? What is the population base you are representing by the selection of the stores?
DR. HALL-ROBINSON: Terry, I don’t know if you want to explain that as far as when we came up with the -- how we can up with the randomized sampling, because there were some limitations with the sites.
DR. PROESCHOLDT: Are you asking why the 10 FoodNet states or why the areas within each state?
DR. MILLER: What does it represent I would like to know?
DR. KOTARSKI: Yes.
DR. PROESCHOLDT: What does FoodNet represent?
DR. MILLER: No. What does your sampling --- represent?
DR. PROESCHOLDT: It represents Tom; correct me if I put my foot in my mouth. It represents as big an area as we could force the FoodNet labs to cover.
DR. CHILLER: It basically we have 10 FoodNet sites. Some of those FoodNet sites are actually statewide. Other FoodNet sites are very localized. For example, in California it is the three bay area counties. And again, we are not here to talk about FoodNet. That is a whole two and a half day lecture in itself.
But essentially, FoodNet is active surveillance for defined regions within each of these states. So FoodNet goes out. They know every single clinical lab in those regions and they ascertain every month the number of lab confirmed cases of these pathogens, which is why -- when I showed you the model data and E. coli O157 going down and Campylobacter going down, that is all based on FoodNet active surveillance so they can make statements about incidents, unlike when we are just doing lab based surveillance or passive surveillance.
It is harder to make statements about that. They can then estimate burdens because they know the populations living in these counties and they know every single lab based organism coming in.
So what we do in the retail food study is we tried to go after those exact areas, because that would be nice. Again, you guys are getting at the exact point. It would be nice to relate that back to a population where we are actually studying disease, and ultimately, that is what we would like to be able to do.
Some sites we are not willing to drive all over the same catchment areas that they are calling clinical labs to get. In most states we actually are in that same catchment area. So we have taken that chain store list from that catchment area, randomized all the stores within that catchment area based on the chain store list and they are buying meat essentially from stores where they are also collecting active surveillance data on disease.
DR. MILLER: And what proportion of the population does that represent?
DR. CHILLER: FoodNet represents about 15 percent of the U.S. population, and so, it is going to vary in each state, depending how big the catchment area is in each site.
DR. MILLER: Have you done any work in comparing the outcome from the retail meats and looking at human disease in that same area?
DR. CHILLER: Not really. We have begun to look at some interesting issues. I think Dave mentioned one where the prevalence of Campylobacter seems very geographical distinct. So they might isolate a ton in Maryland little in California or vice-versa, and we wondered if that related to prevalence in humans.
And we looked at that, and it didn’t seem to relate to prevalence in human disease. So then we wondered -- I mean, it is -- yes. It is never that simple. You would like it to be a simple association. Of course, it never is.
So then we wondered well, maybe they happen to be buying more organic or maybe they are grinding differently. I think as Bob mentioned and as Scott eluded to, it would be nice to sort of trace, to trace distributors, but distribution in the United States ends up being a 15-legged spider very quickly and things just tend to go out.
So we have -- but we are interested in these regional differences. We are trying to understand why can one state have half as much Campylobacter prevalence on chicken than another and why do some states have a lot more Campylobacter illness than others. So we wondered is there a relationship to the food that we eat or is there something else.
These questions have been generated by this data, but unfortunately, we are not answering them yet. But we are trying to methodologically work out answers to those things. The nice thing is having this retail food study within FoodNet and having the FoodNet investigators who are very committed to understanding these issues and questions provides a wonderful platform for pretty intensive research questions into these various issues.
DR. WHITE: If I could add too? I was invited down to the FoodNet meeting this year to talk about NARMS retail for the first time. So, we were able to talk to them about the data we had and to start coordinating using that data with their active disease surveillance. So, we are starting to go ahead with that.
DR. YOUNGMAN: One thing that I was going to add though is -- I mean, it is not just that we are doing this work and we don’t see any correlations that make sense. For example, the most prevalent serotypes that you find in retail meats aren’t the same most prevalent serotypes that you find in humans and so forth. So, there are associations that we are seeing that make sense.
DR. CHILLER: Yes. I definitely didn’t want to give the impression -- you’re absolutely right. It has been very interesting. As Dave mentioned, we just now are coming along with three years of data from a convenience sampling.
In any surveillance system, as you all know, it is hard to really know how to evaluate the first year. You are working out a lot of methods and testing, especially in complex lab based surveillance systems where you are getting labs and sites involved. Now that we have three years of data and now that we are more comfortable with that, we are now making a lot of effort to look at things together.
I think, as Linda mentioned, one of the first things you can see right away is there are some top serotypes that we are finding in human food. And in Paula’s. I mean, we look at Paula’s as well there. So there are going to be some commonalities and there is going to be some major differences.
Kentucky is one serotype that comes to mind that they are finding a lot of and we don’t find much human illness. So, there are a lot of questions being asked essentially by the data and by comparing data, which is great.
DR. VOGEL: I think this question is either for Dave or Terry. I don’t know that I wasn’t listening. What is happening with the sample size on the retail over the three years and in the future?
In 2005 I understand you are going to randomized. But is the sample size, the number of meats selected, changing or increasing?
DR. WHITE: Yes. I can answer that. The numbers have increased over the number of states that have been involved in the program. We started off with six states in ‘02, and remember, they are doing 40 meats per month. So that is 240 meats per month in ‘02. Now we have 10 states doing 40. So, we are 400 per month. So, we are up to 4,800 meats.
That is plateaued. We have no more states to add. So for the current time that is what we have.
DR. VOGEL: So, what are you getting for serotype numbers in 2004? For example, all you have is available is the 2002. You had eight isolates of Newport that year. Well, you can’t tell a whole lot about resistance trends or prevalence if you have only got eight isolates.
DR. WHITE: I agree completely. But in a way that is good that we are not seeing that much Salmonella in the retail foods. 2004 serotyping is undergoing right now. So we don’t know yet. But hopefully, we will have three years of data to compare.
Unfortunately, as we talked about earlier, the majority of our Salmonella is coming from poultry and we rarely have any Salmonella from a ground beef product or a pork product. So those numbers, from year to year, are only in the single digits.
DR. McEWEN: (Microphone not turned on.) -- according to classification of drugs --- human importance may have come up?
DR. HALL-ROBINSON: Yes, it did.
DR. McEWEN: I just wondered what you folks thought about that, whether you considered doing that at all.
DR. HALL-ROBINSON: Well, I know that FDA -- and, Dave, you can step in at any time -- actually had important drugs with the Guidance 152. So the question is do we use that or do we use another classification of important drugs that CDC used.
I think Tom mentioned at the meeting we talked about that. I think it was WHO. We were getting together and coming up with a list of drug classes.
DR. CHILLER: Yes. It is a great question, Scott. CIPARS, as you know, does do that and we like that. We started talking about that concept at this joint meeting when all of us were together. Lucille and others were there from CIPARS.
We were less -- we thought that maybe the best way to approach that, if we were to present drugs, is in classes of critical importance. So what this is, for those of you who don’t know, is that there are -- in the Guidance 152, for example, FDA has ranked drugs as critically important, highly important and important.
So the way to report on resistance would be then to start with the highly important ones and show resistance and then go to critically important and then go to important.
WHO is about to come out -- they drafted a similar list that will be a WHO drafted list where they have got representatives from all over the world, all the different regions, to come together and draft that same list. When we talked there, we thought that maybe the best thing to do was for CIPARS and the U.S. to use the WHO list.
It wasn’t seen as something we developed in our own countries and then sort of imposed that. The idea is maybe we could all be using that critically important list to rank our drugs and then we would have comparative rankings and comparative lists, including even European colleagues or other people that moved into this sort of reporting. So, that was discussed as an idea.
DR. WHITE: The good thing too is I think our two most critically important drugs in surveillance for Salmonella would be ciprofloxacin and ceftriaxone. If you look at those numbers over the years, they are extremely low. I think resistance is less than one percent to both of those drugs.
DR. McEWEN: (Microphone not turned on.) to comment --- talking about integration --- the surveillance on policy. I think the CIPARS group in Canada --- an example for us to --- this past year --- Lucy talked about it. When they rank them, the resistance results according to human importance, --- and there was one province where there was a lot of Salmonella resistance --- so that begged the question why I guess.
And also, there was a -- there appeared to be a relationship between -- I think it was in the retail isolates. It was a high prevalence and samples from poultry from this one province --- it raised a lot of flags and questions --- and public health ministry got interested and it appears that the poultry industry had made a switch in the --- use of --- may explain it and then decided to switch back and make the corrective action, which I understand has been reflected in this year’s data. So, a good example of integration.
DR. YOUNGMAN: Are there any other questions? Maybe not for Elvira, but questions that you wanted to ask of some of the other speakers from this morning?
DR. YOUNGMAN: Okay. Well, that takes us about to 4:20, close to 4:30 when we were planning to close for today. I just wanted to remind you, those of you who have parked in the garage, if you punch in #1470 that will allow you to leave the garage without paying. Your cost for parking today has been covered by this conference.
DR. McEWEN: Is there any chance we could do the other talk today? If we are going to have --- my calculations each person takes 15 minutes ---
DR. YOUNGMAN: That is fine with me. Are there objections from people in the audience to that, people with kids to pick up or whatever? Because our speaker is here.
DR. YOUNGMAN: An alternative that has been proposed is that we could started earlier tomorrow. We are slated to start at 8:30 tomorrow. Are there objections if we started -- sorry?
DR. ZHAO: I am willing to do it however you decide.
DR. YOUNGMAN: Would you like to give the talk today so we can carry on and then have more time tomorrow for discussion?
DR. YOUNGMAN: If everybody is okay with that, that sounds like a great suggestion.
DR. YOUNGMAN: Well, then our next speaker is
Dr. Shaohua Zhao, who is going to talk about molecular characterization of isolates from the retail meat arm.
Okay. Thank you.