Operational Biosurveillance

Public Health: An Inconvenient Truth

"Even if we had a fully credible warning of an imminent influenza pandemic, we wouldn't know what to do with it."  ----  Senior Official with U.S. Health and Human Services

 

Defeats the entire purpose of a "NORAD for infectious disease", doesn't it?  Indeed, the strategy writ large in our country is to build out capability in the area of medical countermeasures (vaccines, antimicrobials, etc), with the hope that we have covered all possible biological agent incursion events of all sources of attribution (natural, accidental, intentional).

During the 2009 H1N1 influenza pandemic, we discovered quickly that delays in countermeasure production and deployment of the H1N1 vaccine resulted in a situation of massive expenditure with precious little mitigation to show for. 

We are often asked by others "isn't the government already doing this?"  Specifically, "isn't CDC forecasting, providing early warning, to promote the use of such information to improve our nation's health security in a manner analogous to our National Weather Service?"

Unfortunately, the answer is no.  Strategic assessments, well-crafted forensic epidemiological studies and clinical guidelines, yes, but the production of data whereby any American is able to type a zip code in and retrieve an infectious disease forecast simply does not exist.  Surely, our public health system is among the best in the world for certain functions, but near-real time operational forecasting, situational awareness, and emergency response is not the forte of our public health professional community.

Take the policy of our National Oceanic and Atmospheric Administration (NOAA), which is the parent organization of our nation's National Weather Service:

NOAA will carry out activities that contribute to its mission, including conducting research; providing environmental assessments; collecting and archiving data; ensuring their quality; issuing forecasts, warnings, and advisories; and providing open and unrestricted access to publicly-funded observations, analyses, model results, forecasts, and related information products in a timely manner and at the lowest possible cost to users.

It is interesting to note that public health has a tendency to withold information until after a forensic investigation has concluded.  This is the byproduct of concern for maximal credibility when speaking to the public as a recognized authority.  Unfortunately, this operational hesitation comes at a cost: an advisory issued often after an event is well-recognized by the involved community.  Granted, an advisory issued regarding one community may be pre-event when viewed by an adjacent, as-yet uninvolved community.  But as has been demonstrated several times in this blog, public health's ability to communicate in an open, unrestricted, and operationally productive fashion that demonstratively contributes to community resilience is inversely proportional to the several features of an infectious disease event:

1. How fast the agent spreads;
2. How many hosts are affected;
3. Ease of clinical or laboratory recognition of the agent;
4. Clinically apparent disruption to the host;
5. Socially apparent disruption to the community; and
6. ... Other factors.

Operational failure is typically observed in public health agencies upon the appearance of a highly transmissible, difficult to diagnose, massively disruptive infectious disease event.  This should be obvious to students of infectious disease crises and disasters. So what is the current state of the art in anticipating and providing near-real time situational awareness for infectious disease events?

Those of a military mindset often look to Pearl Harbor as an example of warning failure that led to grave consequences for the United States (and the world).  A well-known public health analogy is the protracted recognition and warning of HIV-AIDS.  But other major warning failures in US history have occured as a repetitive pattern that has existed for as long as we have had a public health service, including:

• the 1889 influenza pandemic, which showed up in copious media reporting in Asia and Europe, yet did not result in a timely warning to the US

• the 1918 influenza pandemic, which likewise appeared in media reporting overseas long before its appearance in the US

• the 1957 H2N2 influenza pandemic, where lack of attention to publicly available Chinese media reports resulted in tremendous warnings delays:

The first reports to reach WHO were from Singapore early in May to the effect that an extensive outbreak of influenza was occurring and that it appeared to have been introduced from Hong Kong. Later, information was received that the epidemic began in continental China about the third week in February and according to Chu (1958) it originated in Kweichow Province between Kweiyang and Kutsing, which is in Yunan Province. In early March the outbreak had spread to Yunan Province and by the middle of March it had spread all over China. The virus was first isolated in Chanchung by Chu et al. (1957) and in Peking by Tang and Liang (1957) and it is clear that they recognized most of the important features of the virus which have since been described elsewhere. It is unfortunate that this information did not reach the rest of the world until the epidemic was already spreading widely. If it had we should have had two more months in which to prepare. [Society of Medical Officers of Health, 1958]

• the 1968 H3N2 influenza pandemic, where lack of attention again to publicly available Chinese media reports resulted again in tremendous warning delays:

We are dependent on a single newspaper report that the outbreak in Hong Kong was immediately preceded by an epidemic of acute respiratory disease in southeastern China. There is no information on the etiology of this outbreak in China but its close temporal relationship to subsequent events makes it possible that it was due to the Hong Kong strain. It will have escaped none of the members of the Conference that the 1957 pandemic first came to light in southern China, and the experience in 1968, though very tenuous, adds a little more information to the often-expressed hypothesis that strains of influenza virus which have the capacity to spread widely and rapidly often arise in that part of the world. Unfortunately contact between health authorities in China and other countries is even more difficult than in 1957 and it is impossible to obtain information on the possible origin or behavior of the epidemic prior to its appearance in Hong Kong. [WHO, 1969]

• the translocation and subsequent permanent ecological establishment of West Nile virus in the North America (1999), where initial warning signatures were ignored.

• the emergence of SARS in Asia, 2002, where our team documented on a log scale the number of publicly available Chinese media articles describing a grossly unusual infectious disease disaster in southern China months before any public announcement by WHO or CDC.

• the influenza vaccine mismatch of 2007, where thousands of advisories were issued to CDC that resulted in no warning to the American public.  Public health would rightly point out the parents of these dead children chose not to vaccinate.  However the better question to ask is whether these parents had a full and uninhibited right to the risk information in order to better inform their decision not to vaccine?

As of June 19, 2008, 83 deaths associated with laboratory-confirmed influenza infections have occurred among children aged < 18 years during the 2007--08 influenza season that were reported to CDC. These deaths were reported from 33 states (Alaska, Arizona, Arkansas, California, Colorado, Connecticut, Florida, Georgia, Illinois, Indiana, Iowa, Maine, Maryland, Massachusetts, Michigan, Minnesota, Mississippi, Nevada, New Hampshire, New Jersey, New Mexico, New York, North Carolina, Ohio, Oklahoma, Oregon, Pennsylvania, Tennessee, Texas, Utah, Vermont, Washington, and Wisconsin). Among the 83 cases, the mean and median age was 6.4 years and 5.0 years, respectively; seven children were aged < 6 months, 16 were aged 6--23 months, 18 were aged 2--4 years, and 42 were aged 5--17 years. Of the 79 cases for which the influenza virus type was known, 51 were influenza A viruses, 27 were influenza B viruses, and one had co-infection with influenza A and B viruses. Of the 63 cases aged 6 months and older for whom vaccination status was known, 58 (92%) had not been vaccinated against influenza according to the 2007 Advisory Committee on Immunization Practices recommendations. [CDC]

• the 2009 H1N1 influenza pandemic, where escalating warning information was available for several weeks prior to public announcements issued by public health.

• CDC estimates that between 43 million and 89 million cases of 2009 H1N1 occurred between April 2009 and April 10, 2010. The mid-level in this range is about 61 million people infected with 2009 H1N1. 
• CDC estimates that between about 195,000 and 403,000 H1N1-related hospitalizations occurred between April 2009 and April 10, 2010. The mid-level in this range is about 274,000 2009 H1N1-related hospitalizations.
• CDC estimates that between about 8,870 and 18,300 2009 H1N1-related deaths occurred between April 2009 and April 10, 2010. The mid-level in this range is about 12,470 2009 H1N1-related deaths. [Ref] 
• This is more than six times the fatalities seen during 9-11.

• Here we deliberately ignore historical acts of biological warfare and terrorism, with associated gross delays in event signature recognition, warning, and mitigation.  The point being, Mother Nature has provided us plenty of examples without needing to invoke examples of political sensitivity.

 

So where is the NORAD for infectious disease?  Indeed, where is the demonstration of policy in public health equivalent to that of NOAA?  Where is the demonstrable embracing of forecasting as an invaluable instrument of proactive, cost-effective medicine? 

It has often been observed it takes the death of many in a tight geotemporal focus for a society to recognize the inadequacy of its cultural protections vis a vie 9-11.

How many more infectious disease crises or disasters must we observe before we overcome our affliction aka the Cassandra Syndrome?

The Use of Forecasting to Rapidly Identify Non-Routine Infectious Disease Signatures: Part 2

Our team has been doing a full review of outpatient / inpatient / transfer stats from our department's perspective for Jan-Mar 2012, which was a time period where we saw an epidemic of human metapneumovirus (hMPV) that overlapped an unusually severe season of RSV. We have issued several reports of our experience here in this blog.  In Part 2, we explore further the "so what" of operational infectious disease forecasting coupled to early warning. 

On the 25th of every month the Ascel Bio Black Canyon Infectious Disease Forecasting Station released forecasts for infectious disease activity specific to our county.  RSV and other respiratory pathogens were therefore expected to appear and cause a given level of strain to our medical infrastructure. 

From Jan 4th to Feb 10th, we admitted a low number of patients exhibiting typical RSV infection, where we followed American Academy of Pediatrics (AAP) guidelines for the management of RSV: no antibiotics, steriods, with the rare use of inhalers.  In the midst of these admissions, the Station issued an advisory for unusual human metapneumovirus activity which at the time was primarily disruptive to the outpatient component of our medical infrastructure.

Then, from Feb 12-16th, multiple back to back admissions presented as typical RSV infections but abruptly decompensated on the floor, precipitating transfers for all of them to the Front Range by fixed wing aircraft.  In the midst of this, we were told the Front Range's PICU grid capacity was close to diversion status.  It was during this time we had report from the PICU that received our patients they were coinfected with Haemophilus influenzae.  One of these patients would later be triply infected with Streptococcus pneumoniae while intubated.

On Feb 18th, a warning was issued by the Station, indicating an unusual epidemiological situation now existed for epidemic RSV of high clinical severity.  Our department issued guidance to abandon standard AAP protocol for the treatment of RSV and consider the use of antibiotics prophylactically given the apparent high coinfection rate being seen.

Results:

Before the warning was issued to all providers with pediatric admitting privledges, our hospital exhibited a 33% transfer rate, all of whom were over the course of four days.  Of those transferred, none of the patients were given antibiotics prophylactically upon admission.  Forty-six percent of all admissions pre-warning were provided antibiotics for comorbid conditions such as otitis media.

After the warning was issued, the transfer rate dropped to 8%, as now 88% of RSV admissions were provided antibiotics prophylactically.  Of total admissions to the hospital during the entire epidemic of RSV, 71% of the admissions were seen during the post-warning period.  The 8% who were transferred were placed on antibiotics prophylatically, however it is not clear they were truly infected with RSV.

Adherence to the Station's warning heeded at least a 76% reduction in transfer rates due to RSV during the epidemic. 

This is one of the rare examples of operational validity and proof of performance using near real time surveillance coupled to operational forecasting that altered the outcome of an epidemic.

 

 

 

 

 

 

 

Successful Operational Forecast: Influenza Vaccine Drift

Approximately one year ago, our team produced a series of forecasts for influenza vaccine drift anticipated to occur in the 2011-2012 northern hemisphere influenza season.  Today's ProMED post by the influenza group at the British Columbia CDC has provided affirmation of these forecasts, which were available to us months prior to the appearance of influenza in our patient catchment area. As noted here, we have been tracking clinical reports of patients infected with influenza who were previously vaccinated, some of whom died.

This again highlights the value of forecasting and early warning that has been reduced to practice in the clinical setting to rapidly recognize non-routine infectious disease signatures.

ADVISORY (Updated): Influenza in Previously Vaccinated Patients (Pacific NW and Colorado, USA)

Per our prior Advisory, we have report of another metropolitan hospital in the Pacific NW seeing influenza in previously vaccinated children. 

Here in western Colorado, we saw the death of an elderly woman who had a prior medical history of chronic pulmonary disease.  Both she and her husband were vaccinated for influenza, but the husband was the first to be infected, then the wife who later succumbed.

A notable phenomenon given we forecasted this outcome months ago for our hospital, prior to the advent of influenza season.

Several comments from readers indicate a healthy debate related to a chicken-egg problem: is what we are seeing true vaccine drift or simply a poor vaccine match in the first place?  A tricky question to answer...

ADVISORY: Influenza in Previously Vaccinated Children (Pacific NW, USA)

We have report of a significant bump in influenza-positive pediatric patients presenting to a major metropolitan hospital in the Pacific NW, USA over the last 2-3 weeks.  Of note, many of these patients were previously vaccinated for influenza at the start of the school year (~September). 

 

Differential diagnosis (in order of probability):

1. Waning immunity as the predominant type A influenza circulating is H3N2 in the Pacific NW.  Waning immunity over the course of a routine influenza season is an expected phenomenon.  Our team forecasted a moderate vaccine drift probability for A/H3N2 this season several months ago.

2.  Waning immunity as there are still significant levels of circulating A/H1N1 (2009) in the Pacific NW.  We believe this is highly unlikely and forecasted very low probabilities of vaccine drift for A/H1N1.

3.  Possible emergence of other type A influenza viruses such as A/H3N2-variant.  While we believe this is unlikely, we cannot exclude it.  One of the key indicators we have been monitoring is the appearance of influenza in patients who were previously vaccinated.

The Use of Forecasting to Rapidly Identify Non-Routine Infectious Disease Signatures: Part 1

Our team has been doing a full review of outpatient / inpatient / transfer stats from our department's perspective for Jan-Mar 2012, which was a time period where we saw an epidemic of human metapneumovirus (hMPV) that overlapped an unusually severe season of RSV.  We have issued several reports of our experience here in this blog.

The forecast, initially released by the Ascel Bio Black Canyon Infectious Disease Forecasting Station in Nov (i.e. 60 days before the onset of RSV and non-RSV activity), indicated we would see RSV activity precede non-RSV (i.e. human metapneumovirus, hMPV).  We should have expected higher RSV activity than MPV in Jan. 

What actually happened:

More MPV activity in Jan than RSV, with RSV exhibiting nearly zero activity (grossly unusual based on prior baselines, as indicated by the forecast).  This tipped us off immediately we were seeing something not observed at least in the last 5 years.  We were the first in the nation to publicly declare the possibility of a "bad" / unusual season of RSV/MPV; this was confirmed by multiple pediatric intensive care units in many other states.

Our department saw +180%, +213%, and +180% hospital admissions for the months of Jan, Feb, and Mar 2012, respectively, based on prior baselines.  From the perspective of total pediatric admissions for Jan-Mar inclusive, we saw a difference of +238%.

From the perspective of transfers from inpatient care to an outside facility (i.e. due to level of acuity), we transferred +100%, +746%, and +150% for the months of Jan, Feb, and Mar 2012, respectively, based on prior baselines.  The expected average transfer rate for Jan-Mar, inclusive, was 7%.  What we saw in Jan-Mar 2012 was 13% (an increase in clinical acuity of 186%). 

The unusually high clinical acuity triggered a change in clinical protocol based on report of bacterial coinfection (e.g. Haemophilus influenzae, among other)- we are still analyzing the data to see if there is an inference of transfer avoidance after implementing this protocol.  It is notable that the number of transfers for Jan-Mar 2012 was +160% over the combined transfers for Jan-Mar 2009, 2010, and 2011, and +480% over the expected average annual Jan-Mar period.

Conclusions:

1.  The 2012 RSV / MPV season broke all available hospital records in terms of admission volume and clinical severity. 

2.  The forecast was accurate, and highlighted a subtle difference in the 2012 disease activity signature that empowered us to rapidly recognize unusual disease activity early in the season involving RSV and MPV.  This then empowered us to alert colleagues nationwide about a change in RSV/MPV activity.

Round 2: The Black Canyon Forecasting Station Versus Google's FluTrends, ProMED, And SickWeather Combined

Round 2. 

The first such comparison generated quite a bit of Internet traffic.  Record-breaking for this blog, in fact, from multiple countries around the world.

Here again we perform a simple comparison between information produced and made available to the public by the Black Canyon Infectious Disease Forecasting Station (#1) and the combined output of Google's Flu Trends, ProMED, and SickWeather.  Screenshots from these websites were taken at approximately 2130 on 19 Feb 2012.

 

Black Canyon Infectious Disease Forecasting Station.

As previously noted, we forecasted in December a surge in patient volume and clinical severity due to a variety of viral respiratory pathogens, notably RSV, for our specific location in the US: Delta County, Colorado, which is near Grand Junction (geographic coordinate Lat/Long: 38° 44' 43.4394", -108° 2' 49.1994").

We indeed did see this, along with the forecast-facilitated detection of unusually severe, high volume RSV disease related to coinfection with other pathogens.  Here in Colorado, this has produce pediatric infrastructure strain at a grid level, meaning near-diversion status for several PICUs along the Front Range (our referral loci).  We were warned that beds may not be available the next time we needed to execute a transfer...

The entire staff complement of our office was infected, yet again, but no one took time off.  Much misery among staff working very long hours, with physicians pulling 48 hr call that afforded little sleep.

 

Google Flu Trends.

Again tells us we are at "moderate" levels of influenza. 

We have not documented a single positive specimen to-date in our clinic.  We are aware of a clinic 20 miles east of us that documented two positive cases in their kids about ten days ago- no activity since.

 

ProMED. 

Absolutely nothing on Colorado.

 

SickWeather.

Thirteen (13) different symptoms selected, as shown below.  One single report of "pneumonia" submitted.  Nothing else.  We were unable to obtain evidence of true pre-event operational forecasting in the SickWeather website.

Summary.

Again, the limitations of using open source information and social media is revealed.  We just experienced an unusual, non-routine infectious disease signature pattern in our county that resulted in significant medical infrastructure strain and near-diversion status at the grid level for the pediatric critical care infrastructure in our state.  There is no indication of this significant level of infectious disease activity in Google's Flu Trends, ProMED, and SickWeather combined.

The Black Canyon Forecasting Station in "War Time Operations"

This is a local healthcare community WARNING posted today by the Black Canyon Infectious Disease Forecasting Station.  Of note, this is a warning related to mostly non-notifiable infectious disease- thus reflective of the socialization process of routine infectious disease forecasting analogous to meteorological forecasting as a vital component to community health.  Public health is certainly a partner and involved, but until a notifiable disease is involved, they do not engage.  Meanwhile, we are dealing with near-maximum patient volumes, clinical severity, and awareness of an entire medical grid strained (which effects our decisions to transport to critical care facilities).

 

A new epidemiological situation in Delta County has been detected by the Black Canyon Infectious Disease Forecasting Station over the course of the last week.

The forecast for February (first issued in December) called for heightened activity levels of multiple pathogens, including RSV, human metapneumovirus, adenoviruses, coronaviruses, influenza A and B, and Moraxella catarrhalis, among other pathogens.  The forecast also indicated we would reach maximum medical infrastructure strain this month, and indeed we and the entire state have been experiencing this- multiple PICUs on the front range are at near-diversion status.  What is remarkable is this is NOT due to influenza virus, as we have had a remarkably mild season (again, as forecasted). 

The current situation represents expected high RSV transmission in the community.  What is not expected is the older age of children involved and the clinical severity- we are seeing an abrupt and unusually high volume of admissions, clinical severity, and age of our patients (2 to 2.5 year olds).  Of significant concern is the atypical pattern of these patients exhibiting apparent clinical recovery, then abruptly decompensating and requiring intubation and transport to facilities with more appropriate critical care facilities.  We have discovered there is an (apparent) high and unusual rate of RSV coinfection with other pathogens such as human metapneumovirus, adenovirus, Moraxella, and notably for Delta County, Haemophilus influenzae.  PICUs in our transport referral area are indicating a rare rate of coinfection with influenza A.  RSV coinfection, in general, is often associated with more severe and life-threatening complications.  Consultation with pediatric emergency departments and PICUs nationwide resulted in reports from several areas of the country indicating similar patterns.  We continue to receive updates...

When influenza is exhibiting its typical level of activity, transmission activity produced by these other pathogens are thought to be suppressed in the community by mechanisms still as-yet unclear.  While the forecast of a mild influenza season initially appeared to be "good news", in this case it revealed a disruptive phenomenon involving multiple pathogens (with RSV leading the way).

James M. Wilson V, M.D.
Chief of Station
Black Canyon Infectious Disease Forecasting Station
-and-
Department of Pediatrics
Delta County Memorial Hospital
Delta, Colorado

 

The other important phenomenon is the social network that has evolved around operational forecasting.  Physicians simply come forward quicker with information vital to confirm signals that would otherwise be missed.  This, of course, is a crucial sociological process when contemplating rapid recognition of non-routine or unusual infectious disease activity signatures.

 

 

Peaking Activity of hMPV Confirmed in Multiple US Cities

Cross-checking with other pediatric colleagues nationwide has revealed that yes indeed, from Rhode Island to Chicago to Dallas pediatricians are noting a substantial level of hMPV activity.  Technology is surely a factor here, where we are actually able to 'see' this pathogen via molecular diagnostics. 

Let it be said, we are seeing a sociological shift in our medical colleagues in even discussing this pathogen in the first place, which has largely been ignored because such recognition did not change clinical management.  Important because of a social shift in understanding our country's baselines of biological activity.  Such finer understanding and discourse about this pathogen feeds into the process of developing our core of "astute clinicians" so important to rapidly recognizing unusual biological activity signatures in our country.

 

 

"Frequency-Modulated Shielding" of a Hospital Using Operational Forecasts For Antibiotic Resistance

One of the greatest challenges we face in medicine is the judicious management of our armamentarium of antibiotics as we confront resistant microbes in our community.  There continues to be debate in the literature regarding the ultimate best strategy for deciding which antibiotic, by pathogen, one should use now versus a year from now when resistance profiles may have migrated. 

The Black Canyon Forecasting Station recently operationalized over 250 pathogen-antibiotic forecast lines for one of our client hospitals.  The major challenge of the hospital is how to manage its antibioitic armamentarium with an eye towards the future.  In other words, "when will we lose the ability to use a given antibiotic for a given pathogen, and based on this knowledge, should we continue to use it?"  Sometimes it's best to holster that "weapon" until a later time when it might be the only option left to you as a physician.

This is a key question from a variety of perspectives- pharmaceutical supply & demand, temporal and financial cost to develop new antibiotics, and patient treatment costs, among many others.  What was interesting to us was the disparity between guidance from The Sanford Guide To Antimicrobial Therapy 2011, Epocrates, and current reality in this particular hospital.  With antibiotic resistance forecasting now a part of the analysis, these disparities become even more poignant.  An example is Pseudomonas aeroginosa, which is horribly resistant to multiple antibiotics, and the forecasts are dismal.

Wiesch et al recently published a nice review of "Population Biological Principles of Drug-Resistance Evolution in Infectious Diseases" (see attached), which reminds us of

The emergence of resistant pathogens in response to selection pressure by drugs and their possible disappearance when drug use is discontinued are evolutionary processes common to many pathogens.

Constant forecasting and real-time monitoring of drug resistance may be our best defense given the research, development, and distribution of new antibiotics is not keeping up with the global problem of resistance. 

For those willing to take a leap from reality, look to the Star Trek-themed "frequency-modulated shielding" of a starship (i.e. hospital).  A bit of silliness, but a highly relevant analogy when considering the very real threat of encroaching drug-resistant bacteria and the reality of a too-slowly evolving antibiotic armamentarium. 

As the Black Canyon Forecasting Station brings this new forecasting capability on-line at the hospital, we note this may be the first time this has ever been attempted...

Download Population biological principles of drug-resistance evolution in infectious diseases