The Haiti Epidemic Advisory System (HEAS)

HEAS Operational Infectious Disease Forecasting Has Been Reactivated

Through an incredibly generous donation from Direct Relief International, the HEAS' forecasting functions for Haiti have been reactivated.  Stay tuned...

Village Level Cholera Forecast Validation

Below are the results of our assessment of forecast validation for Artibonite Province, Haiti.  We are quite pleased with the results from an operational perspective.

This is the area:



Basically, the forecasts vs. the actual outcomes for:

 

1)    Anse-Rouge was outstanding with three peaks in both, and with both the timing and the amplitudes of the 2^nd and 3^rd peaks right on.

 

2)    Govaives was solid with three peaks in both but only okay as the timing and amplitude of the 1^st and 3^rd peaks are off while that of the 2^nd peak was right on.

 

3)    Marmelade was solid with three peaks in both and okay as the timing and amplitude of the 1^st and 3^rd peaks are right on while that of the very broad 2^nd peak was off.

 

4)    TerrNeuve was off in the timing of the peaks, but the amplitudes of the peaks were solid and the number of peaks, with six predicted and six occurring, was outstanding.

 

5)    Ennery was very good as both the timing and amplitudes of the actual and predicted 5^th peaks were right on, however while the forecast was for four peaks and there were five actual, the predicted 3^rd  peak fell between the 3^rd and 4th actual peaks, the agreement was not too bad overall. Additionally, considering the relatively high and low numbers of actual and predicted cases, the comparison is very good.

 

6)    Grosmorne was okay considering the low number of cases.

 

7)    St. Michel shows a mixed outcome as the only one peak was predicted and the timing and amplitude of that peak vs. the actual realization was outstanding, however the forecast missed the 2^nd and 3^rd peaks entirely.

 

So why did the model do well in some cases and not so well in others?

 

We believe that there are several fundamental reasons.

 

On the positive side, there is good structure and momentum in the individual and collective time series. As such the memory in the systems is capable of being carried forward in a probabilistic sense. So, one can conclude that there are individual and distinct signals which can be and have been separated from the noise in each of the time series.

 

Albeit, one must recognize that the forecasts are totally probabilistic and not deterministic. Moreover. and this point cannot be overstated, the time series are very short (~2 years) in all cases, so there are very few data points. Also, there have not been any correlative analyses done with other likely relevant parameters and variables, other then the cross correlations that we conducted between the stations. 

 

The challenge created by the short time series can be briefly summarized with a simple example. To wit, if Dave left his office to walk across campus and the first two students he saw were wearing NC State tee shirts, then what would he guess, probabilistically speaking, that the next student would be wearing; an NC State tee shirt or not; 50/50?

 

Again, probabilistically speaking, if Bob found that using 117 years of hourly temperature data that the coldest two weeks of the year occur during the 2^nd and 3^rd weeks of January, then what about the forecast for 2013? What are the odds that it would necessarily be those two weeks or could it end up being the 4th week of January and the 1^st week of February, et cetera? As such, probabilistic forecasts are not deterministic.

 

In terms of the timing and amplitude, while we have several observations, they cover only a bit over 1 year so we have had only one look at a pattern that is not necessarily going to repeat the next year. The 117 year example gets at the idea of probabilistic forecasting but our case is much worse, having seen not 117 instances of a yearly pattern but only a bit over 1. 

In our forecasts we did not account for any of the external variables that the different regional facilities are located in, such as: relationships to weather and climate factors, with the former and latter including such conditions as warm and humid conditions, altitude, arid versus no-arid conditions, thus low or high low precipitation, transportation availability, water contamination, the ability of the stricken to get to the hospitals, the exposure of the populace to disease carrying hosts, such as outsiders carrying the disease into the locale(s), polluted water, and so on.

 

So the solution to the improvement of forecasts would be to have longer time series at individual and multiple locations which document incidence and also to be able to conduct correlative analyses with other factors that may be conditional and or causal in nature. We actually showed the power of such studies and analyses with our cross-correlations between station to station cholera data and also our cross-correlation study of the other diseases and medical conditions. These powerful relationships, or lack therein, were in the report as well. 

 

All of that said, the results are very encouraging, especially given the paucity of data, suggesting that highly reliable relationships can be created given appropriate data sets, and sufficient time and effort.   

 

 

Towards Validation of Forecast For Cholera in Haiti

On July 27, 2012, the Praecipio International infectious disease forecast team, using technology from Ascel Bio LLC, issued a forecast for cholera activity in Haiti.  Most relevant was the comment,

It appears that in Artibonite, the peak weeks for transmission are Epidemiological Weeks 27 (first week of July) and 46 (second week of November), respectively.

We are still in the process of collecting all of the data for Artibonite, but we note a remarkable correlation in an area just north of the original forecast site in the Cap Haitien area:

At the time, we had used the Artibonite data as a proxy for the north.  We are surprised to see how "tight" the forecast was temporally.  Obviously, this is not a final assessment.  But it is compelling nonetheless.

Progress in Operational Infectious Disease Forecasting in Haiti (Updated)

We have an update from the forecast team.  While we only have two years of baseline available to produce a tentative forecast, here are the results:

1.  There is a 3% decline in cholera cases across all of the sites surveyed.  This is likely due to herd immunity, which is a good thing.  This phenomenon will mitigate the explosive degree to which cholera will transmit in Haiti.  In other words, we do NOT expect cholera to be as bad as it was during the initial period of introduction (i.e. the first year of contact).  We DO believe cholera is established now in Haiti's ecosystem, so it will be an ongoing problem.

2.  The overall seasonality summarized across multiple sites is peaks at Weeks 27 (early July) and 46 (mid/late November).

3.  Specific forecasts, by locale in Artibonite.  Note the subtle differences when you examine the local data:

St. Michel: in Week (W) 27 the forecast is for 2 cases rising by 1 case/wk to 10 at W 34 and to 15 by W 37 the receding to 11 in W 39 and down to 2 by Ws 46-47 and minimal cases after that.

Gros-Morne, by W 35 there could be 1 case increasing to 4 cases by W 38 dropping to 1 case by W 41 and staying at zero until 1 case occurs in W 4, dropping back to minimal cases thereafter.

Terr-Neuve, 1-2 cases occur by Ws 38-40, increasing to 3 by W 50  and decreasing to minimal cases thereafter;

Ennery, cases begin with 1 case increasing to 10 cases by W 35, and then rising to 24 by W 39, dropping to 4 cases by Ws 44-46, rising to 6 cases by W 48 dropping to minimal cases by W 52 and thereafter;

Anse-Rouge there are no cases until W 36 with 1, then 2-3 by Ws 38-40, dropping to minimal cases then rising from 1 in W 46 to 7 by W 49, dropping to 1 to 2 cases by W 11 and then dropping to minimal cases thereafter;

Govaives there is 1 case in W 33, rising to 4-5 by Ws 37-38 dropping to 1 by W 42 and then minimal cases until Ws 4-5, with 2 cases by W 7 and then minimal cases thereafter;

Marmalade, there could be 1 case during period from Ws 27-33, then 2 in W 44, rising to 9 in W 47 dropping to 1 by W 52, rising to 7 by W 8, dropping to 1 by W 16, and minimal cases thereafter.  

To our knowledge this has never been done before- site-specific forecasting of infectious disease that includes a tie-in with response.  We thank Direct Relief International, who supports this important work.

Operational Forecasting of Cholera in Artibonite, Haiti

The Praecipio International infectious disease forecasting team, sponsored by Direct Relief International and thanks to HEAS partner efforts in sharing data, has just completed the first phase of an extensive statistical review of epidemiological data for cholera in Artibonite.  We have had the honor and benefit of working with some of the world's experts in meteorological forecasting, including three National Academy of Science members.  All of these individuals were excited to have the opportunity to support the People of Haiti. 

We went through this exercise to get our hands around the question of "what seasonality is cholera settling into"?  While we have often seen media reports of association with rainfall, which makes sense based on prior experience in countries like Bangladesh, we needed to make sure of this in Haiti. 
 
The bottom line is we have identified a target signature pattern for Artibonite that is now testable in the humanitarian theater.  Obviously, we would be more sure of our conclusions if we had 10-20 years of reliable data, but with 2 years of baseline it is the best we have at the moment.  It appears that in Artibonite, the peak weeks for transmission are Epidemiological Weeks 27 (first week of July) and 46 (second week of November), respectively.  This means we would issue forecasts for Artibonite at 60 and 30 day intervals prior to all of the Haitians and NGOs operating in this area.  We are in the process of determining whether variation in these signatures exist from village to village.
 
We will now be executing live forecasts from this point forward.  The crucial next step is validation, of course, of these forecasts.  Validation contributes to refinement of the forecast libraries.
 
One of the other observations made was a statistically significant decrease in overall rates of cholera transmission, which of course makes sense as the disease is settling into an endemic state.  We are planning to publish this as an academic paper for the world's benefit.
 
This is just the beginning….

Scattered Reports of Fatalities and Patient Surge in Central Haiti

We have been tracking reports over the last 10 days of patient surge and the now-classic and typical reporting of patients dying along the roads in the mountains of central Haiti.  Questions are being raised about the operational status of the St Marc hospital to receive some of this patient flow.  This is the same facility where the original first cases of cholera were reported.

Public health again nowhere to be found...

 

The Passing of a Hero: "Big Dave" Bompart

Today the HEAS is silenced in respect for the passing of "Big Dave" Bompart.  Our prayers are with his family.

 

James M. Wilson V, MD
Haiti Epidemic Advisory System (HEAS)
Executive Director
Praecipio International
Washington-Houston-Port au Prince


Praecipio International is a charitable non-profit organization devoted to the promotion of operational biosurveillance worldwide.

Please Lift "Big Dave" Bompart Up In Prayer

It is with the greatest sadness the HEAS lifts up David Bompart "Big Dave" in worldwide prayer.  Big Dave was shot while transporting funds to an orphanage in Haiti.  He is a Christian, a hero, and an inspiration to us all for his tireless sacrifice for the People of Haiti.  Please take a moment to pray for Big Dave.

James M. Wilson V, MD
Haiti Epidemic Advisory System (HEAS)
Executive Director
Praecipio International
Washington-Houston-Port au Prince


Praecipio International is a charitable non-profit organization devoted to the promotion of operational biosurveillance worldwide.

The Current and Future Status of Cholera in Haiti

Within the HEAS, a  serious discussion has ensued regarding the current and future status of cholera in Haiti, which began with a question from an outside agency:

 

 

• [From a major NGO seeking to implement major prevention efforts in Haiti] We know these predictions can be somewhat uncertain, but we would be grateful to hear your opinion as a medical actor on the ground.  Your first-hand knowledge about the infection patterns and the current pace of health and water sanitation interventions makes you well poised to help us see a few years ahead into the cholera epidemic.

 

• [From me to HEAS partners internationally recognized as cholera experts]  Guys, I could use your input and advice regarding the [above question] and would like to form a consensus opinion, if possible.

 

• The usual cholera epidemic attack rate (AR) among immunological susceptible populations is 2%.  This estimated attack rate considers only severe and some moderate cholera cases requiring hospitalization. Most of the total cases (95%)  are mild or asymptomatic and may not be detected by the surveillance system

 

• [My reply] Is this AR what you guys observed in Peru in '91?  Reason why I'm asking is we are dealing with a totally naive environment (albeit we are in a time period subsequent to first introduction [i.e. first wave]).

 

• Yes, this is what it was observed in the 1919 Latin America epidemic as well as  previous descriptions of cholera epidemics in Asia and Africa. I think that once cholera is established in the environment , all population is affected sooner or later if the water and sanitation conditions are not adequate. That is the case of Southeast Asia, over there the only susceptible population is small children with no antibodies against cholera.

 

• [My reply] This is a very important observation- we are still seeing a high percentage of adults affected versus kids- at least in the case of Pestel if I'm not mistaken.  So, the situation may be endemic in Port au Prince but still epidemic in the rural areas?

 

• There may be a few pockets of susceptible populations. Nevertheless, after cholera strikes, we have to be very careful with the diagnosis. During the last waves of the 1991epidemic, I went to investigate diarrheal diseases outbreaks that were wrongly diagnosed as cholera. Many of these outbreaks were disenteric amebiasis, E coli, salmonella, cryptosporidium, etc. It is time to start to invest some resources in confirming the outbreak agents.

 

• [My response] Right, copy.  That seems to have been the problem with the Pestel situation [i.e. public health officials delaying 72 hrs to respond because the situation in Pestel was not laboratory-confirmed cholera]- a bias that it was not due to cholera until lab confirmation proves otherwise.  If we had not intervened, people would have needlessly died.

 

... Later,

• I just checked the number of cases accumulated in Haiti until January 2012:
  
  The Ministry of Health has reported 524,861 cumulative cholera cases, 282,989 hospitalizations and 7,018 deaths, as of 8th january 2012. The overall fatality rate is 1.3 percent nationwide. The general trend is towards a decrease in the number of new cases and fatality
  *Source: Last report of MSPP as of 8 january 2012
  
  If we have a total population of 10 million inhabitants it looks like the 282,989 hospitalized cases (~2.8%) shows that the population is exhausted and the cholera epidemic may have  entered into a endemic patter of transmission.  If  somebody comes to deliver the vaccine or establishes any control strategy it will look as a very effective intervention,  just by being temporally associated with the development of immunity among the population.

 

• [My response] What if these numbers are an underestimate given the 2/3 rural mountainous / unreached / poorly reported areas of Haiti?  Are we able to consider this?  Thank you for the discourse.

 

• I agree. Indeed, there must be an underestimation of the number of cases. The 2% attack rate is based on similar or even less sensitive surveillance systems. If the number of cholera cases is greater, again , this means that most of the Haitian population is immune. The sadness  of this history is that we did not concentrated all our efforts in saving the lives. To achieve this level of natural immunity with more than 7,000 deaths is a failure. At this point water interventions will be strongly associated with no cholera cases, latrines will be strongly associated with no cholera cases, voodoo bone powders will be strongly associated with no cholera cases. The colloquium of this disaster is the return of UN "experts" to say that their presence has been the key factor in stopping the epidemic.

 

 

 

A Moment of Quiet Introspection Following the Disaster in Pestel

Today we received the final assessment of the cholera disaster (IDIS Cat 5) in Pestel and surrounding communes:

Below is a summary of the cases of cholera logged in the Pestel commune.  

---------------------------------------------------------------------------------------------------------------------------------

Incidence of Cholera in Pestel Commune

October 2011 --> 115
November 2011 --> 118
December 2011 --> 126
January 1-12 --> 86 cases

***Assessment by Miriam, Harvest House, Dr. Phillipe January 9-12***
***Cholera eduction/ Aqua tabs + ORS distribution implemented***
Global DIRT assistance January 12-14

1/13 --> 7
1/14 --> 5
1/15 --> 1
1/16 --> 4
1/17 --> 11  (1 death)
1/18 --> 5
1/19 --> 3
1/20 --> 3
1/21 --> 0
1/22 --> 7
1/23 --> 4  (2 deaths)
1/24 --> 0
1/25 --> 3
1/26 --> 3
1/27 --> 4
1/28 --> 0
1/29 --> 3
1/30 --> 3

Total January = 152

Morbidity:
December 1 - January 12 --> 59 deaths
January 12 - 30 --> 3 deaths

 

This was our response:

Clarification: is this solely for Pestel and NOT the surrounding villages as well?  We had report of as many as several hundred cases and 70 fatalities in the region as opposed to just Pestel.  IN THE LEAST, it shows yet again what happens when emergency response occurs- you blunt the fatality rate.

Their response to our question:

These are the cases of cholera which presented to the 4 CTCs (Tozia, Dereveaux, Pela, Pavillon,  Pestel).  I wish you could have heard the stories and seen the faces of the people in Grand Anse.  Then you see how many lives were saved and how many we're affected by the cholera.

 

 

 

Operational Highlights of the HEAS

Our team was interviewed by a major international press organization this afternoon, and they asked about key achievements of the HEAS.  The following was provided as a list of highlights:

1.  HEAS performance / forecasting functions the first 150 days post quake

2.  Our first public statement regarding reports of severe diarrheal disease in Artibonite (ie cholera)

3.  Our first public report of linking the UN Nepalese base in Mirebalais to the cholera disaster (this was actually an update, posted a couple of hours after the initial report).  This was the first public report describing probable attribution for the cholera disaster in Haiti.

4.  Referral to the Paul Keim (a member of the NSABB, the same panel reviewing the controversial studies on avian influenza) study proving the link to Nepal

5.  First case of acute flaccid paralysis identified.  It remains our belief Vaccine-Derived PolioVirus (VDPV) Type 1 continues to transmit in Haiti.  What is required to disprove this belief is a comprehensive field investigation by a neutral third party.

6.  11th case of acute flaccid paralysis identified

7.  Field ops: description of HEAS warning-response ops in rural mountainous areas of Haiti, note the report on Borgne as an excellent example

8.  Intercept Team field ops reports from Belle Fontaine (Dec 10-12th and Jan 12-13th)

The reporter asked us what we thought about cholera transmission now in Haiti.  Our response was cholera continues to transmit in the rural, mountainous 2/3 of Haiti's landmass difficult to appreciate by the major NGOs and officials based in the 1/3 of Haiti that is urbanized.  There is little question the highest fatality rates and destructive social outcomes are observed in these rural areas.  Case fatality rates still reported by officials represent underreporting and bias towards areas of Haiti with the best access to health care. 

We also emphasized the HEAS did all of this for less than $20,000 US.