Salmonella Saint Paul outbreak

[Robert A. LaBudde]

CDC has given the outbreak curve, which now allows some conclusions
to be reached:

http://www.cdc.gov/salmonella/saintpaul/epidemic_curve.html

1. Cause of the outbreak stopped feeding into the distribution system
around the end of May. The growth was exponential from April through
May, which suggests much more than simple contamination of a few
lots. Some type of feed back may have been present.

2. Cases since then are due to capacitance of the distribution
system, and discharging this capacitance is still underway, but is nearly over.

Because the source and amplification of the outbreak was disconnected
at the end of May, any investigations (e.g., sample and test) by CDC
and FDA after that point would be fruitless.

However, there would have been a useful purpose in examining product
already in distribution (e.g., warehouses, storage at restaurants, etc.)

It would be interesting to know what happened at the end of May to
terminate the propagation mechanism. Presumably it had something to
do with the CDC investigation and fixation on tomatoes.

So CDC appears to have done the job, even if it did so in partial ignorance.

Of course, we may never know how the outbreak came about in the first
place, as the trail is now cold.
================================================================
Robert A. LaBudde, PhD, PAS, Dpl. ACAFS e-mail: r...@lcfltd.com
Least Cost Formulations, Ltd. URL: http://lcfltd.com/
824 Timberlake Drive Tel: 757-467-0954
Virginia Beach, VA 23464-3239 Fax: 757-467-2947

"Vere scire est per causas scire"
================================================================

[CarlHansenMD]

"There is an inherent delay between the date that an illness starts,
and the date that the case is reported to public health authorities.
It typically takes 2-3 weeks for Salmonella infections."
See link below (http://www.cdc.gov/salmonella/saintpaul/
epidemic_curve.html)

Lyons and colleagues reported in 1980 an animal to human spread that
led to a hospital nursery nosocomial outbreak of Salmonella:
"An epidemic of resistant Salmonella in a nursery. Animal-to-human
spread." JAMA. 1980 Feb 8;243(6):546-7.

The index patient lived on a farm where the animals received
antibiotic feed mixture.

The curve presented in the CDC page follows roughly a normal
distribution.

Time series analysis woul provide more useful information when coupled
with data relevant to specific geographic locations.

Within a given geographic location it would be useful to cross
correlate data with the entry of the produce into the market and then
also the complete removal of produce from the market.
The occurence of a subsequent spike at a later time would suggest th
signal has moved to a vector generating a new cluster of events
(clinical infections.)

Given empiricallty available data it should be possible to generate
statistical mathematical models and then conduct clinical and public
health work in the communities to ascertain what exactly happened and
why there was such a long time course. Local professionals and
facilities could then take appropriate educational steps with family
and patient care and hygiene.

It would also help to establish who is a carrier and the attendant
risk of spread in health facilities, food facilites, schools and child
care programs.

[Robert A LaBudde]

At 08:49 AM 7/19/2008, Carl wrote:
><snip>

>The curve presented in the CDC page follows roughly a normal
>distribution.
>
>Time series analysis woul provide more useful information when coupled
>with data relevant to specific geographic locations.
>
>Within a given geographic location it would be useful to cross
>correlate data with the entry of the produce into the market and then
>also the complete removal of produce from the market.
>The occurence of a subsequent spike at a later time would suggest th
>signal has moved to a vector generating a new cluster of events
>(clinical infections.)
>
>Given empiricallty available data it should be possible to generate
>statistical mathematical models and then conduct clinical and public
>health work in the communities to ascertain what exactly happened and
>why there was such a long time course. Local professionals and
>facilities could then take appropriate educational steps with family
>and patient care and hygiene.
>
>It would also help to establish who is a carrier and the attendant
>risk of spread in health facilities, food facilites, schools and child
>care programs.

To the casual observer, the outbreak case curve vs. date of onset
might appear as a Gaussian (bell-shaped) curve.

However, this is not the case. It is a splice of two exponential
curves, one increasing due to amplification during the initial phase
of the outbreak, a cusp where the outbreak is turned off, and an
exponential decline due to discharging of the capacitance of the
distribution system plus the incubation period lag. (These latter
lags smooth out the cusp into a blurred peak.)

The entire curve can be modeled accurately with 4 parameters for the
listed effects. See my chapter in Microbial Food Contamination, 2nd
ed., for a description of the model, meaning of the effects, and the
fitting process.

To a first approximation, the back of the outbreak was broken about
one incubation period (2 weeks?) before the visible peak, or in the
last half of May. Whatever started the outbreak ended then. The rest
of the outbreak curve is like floodwater moving downstream. The rains
are over, but the flooding has just started.

[CarlHansenMD]

Dear Bob,

Thanks for the reference.

As you described, the process lends itself to additional time series
analysis such as post-interval and interval event based studies They
would provide some very useful data that could give insights into what
happened and different factors that led to the number of individuals
who became ill.

The exponential build up and decay as well transmission process have
parallels with information or neuronal transmission systems.

The introduction of human to human spread or surface to human spread
would create a superimposition of different events that are discrete.
Based on a common endpoint of an iacutely ill person, it would give
the appearance of the graphic data presented.

A secondary analysis of the data would give more powerful information
for tracking out how well the food, medical, and public health systems
are managing an outbreak in a given geographical region.

[Roy Costa]

Please see below.

Roy E Costa, R.S., M.S./M.B.A.
Public Health Sanitarian Consultant
Environ Health Associates, Inc
1.386.734.5187
www.haccptraining.org
www.safefoods.tv
rco...@cfl.rr.com

> Date: Sat, 19 Jul 2008 08:52:59 -0700
> Subject: [Foodsafe] Re: Salmonella Saint Paul outbreak
> From: crha...@mend.net
> To: Foodsa...@googlegroups.com

>
>
> Dear Bob,
>
> Thanks for the reference.
>
> As you described, the process lends itself to additional time series
> analysis such as post-interval and interval event based studies They
> would provide some very useful data that could give insights into what
> happened and different factors that led to the number of individuals
> who became ill.
 

Its always a relationship between the host, the agent and the environment.

 
 
> The exponential build up and decay as well transmission process have
> parallels with information or neuronal transmission systems.
 

Then we look at the reservoir and pathways of transmission.

 
 
> The introduction of human to human spread or surface to human spread
> would create a superimposition of different events that are discrete.
> Based on a common endpoint of an iacutely ill person, it would give
> the appearance of the graphic data presented.
 

They are not superimposed, meaning coming from without the process, they are actually causualy linked in a chain of events over time.

>
> A secondary analysis of the data would give more powerful information
> for tracking out how well the food, medical, and public health systems
> are managing an outbreak in a given geographical region.
 

The problem is we sometimes do not get enough data, and maybe sometimes like in the tomato outbreak we get too much data, too much background noise and distortion to be able to adequately test the effect.