Mapping Strategy

Success Begets Complacency: Germs, Vaccines & the FDA

Wide-ranging and thought-provoking article by Peter Huber in the latest City Journal on vaccines, public health, politics and the economics of drug development for germ-fighting pharmaceuticals:

The FDA... worries that with too few patients tested, statistical anomalies might allow an inferior antibiotic to win its license, and the new one might then become the benchmark for a third even worse one, and so on until the industry slouches its way down to licensed sugar pills. The agency calls this scenario “biocreep.” It’s just the sort of logical but overly theoretical concern that sneaks into government offices where the paramount objective is to avoid mistakes. But nowadays, “no mistakes” means no new drugs licensed until the germs start killing lots of people again...

Enveloped in bureaucracy, the germ-killing segment of the drug industry has lost much of its flexibility, resilience, and reserve capacity—and has become painfully slow in developing what new science makes possible. Short-term economics and federal law have converged to create a systematic bias in favor of the germicidal drug invented and licensed decades ago...

In 1957, five important vaccines were being supplied by 26 companies. By 2004, just four companies were supplying 12. Mergers accounted for some of the attrition, but most of it resulted from companies getting out of the business. More than half of the vaccines routinely given to young children in 2005 came from just one manufacturer; four others had only two suppliers.

Money quotes:

The new infectious diseases... have evolved to be as nimble as we are now institutionally stolid, as flexible as we are rigid...

Germs are terrorists: they let the dead past bury its dead, they are always changing, and the ones you know aren’t the ones that will kill you.

Five-Dimensional Visualization: Lifespan vs. Income

Check out this extremely cool visualization of how life expectancy relates to national income, region and population--all over time (be sure to press the 'play' button in the lower left). Cleverly (a la Tom Barnett's core and gap concept), the tool is called the "Gap Minder".

Everyone's mind works a little differently, but if there's universal truth in the adage that "a picture is worth a thousand words", then such data-rich portrayals of complex interrelationships have got to be worth millions. In a world increasingly dominated by PowerPoint, we can only hope that they will play a much bigger part in our public and private dialogues about dynamic, complex problem spaces.

Infrastructure Priorities in a Pandemic

Glad someone is thinking about this. Unfortunately, it might well fall under the category of "planning" that Hayek rightly criticized decades before it became acceptable to do so.

Pandemic Preparation

Former Health and Human Services Secretary Tommy Thompson (who consults on pandemic planning) makes a coldly harsh but thoroughly rational statement about bird flu in the January/February, 2007 issue of Corporate Board Member Magazine (quick, free registration required):

“I keep telling them this planning is good for the company. If we have a pandemic we’ll build market share, because our competitors aren’t doing the planning.”

What kind of market share, you may ask? Earlier in the article, another consultant provides examples of opportunity in the midst of hypothetical chaos after avian flu has hit:

A big hotel chain, figuring tourism would collapse, is studying how to rent its properties to governments as places to tend the sick. A hospital operator has contracted for refrigeration trucks [because] their in-house mortuaries are too small... A communications corporation has built a self-contained “clean facility” within its headquarters to house critical personnel and key operations.

Sounds like the kind of grim-but-true advice I'd expect to hear on Six Feet Under. Elsewhere in the article, Robert E. Mittelstaedt Jr., dean of Arizona State University’s W.P. Carey School of Business comments on scenario planning, noting (as I have before) that scenarios are often misused as a predictive tool and at a far too detailed level for business continuity planning (or much of anything else, for that matter):

...it’s difficult to engage in detailed scenario planning, because no one knows how, exactly, a pandemic would unfold—“but we want to make sure there’s a plan there, while not micromanaging it.”

Shortly after 9-11, a client showed us a massive spreadsheet of disaster possibilities, deeming them all 'scenarios'. They were not. Every conceivable thing was covered--from half a dozen megatonnages and locations for nuclear detonations to countless permutations of bridges and tunnels being taken out, anthrax being mailed, utilities and emergency services being cut and the simultaneity of all of the above.

Too much. Waaaaay too much. The 'scenarios' were also confined by definition to what the client could conceive of on its own, colored, it's now clear in hindsight, by the headlines of the day. We were brought in in part to help make sense of it all.

Another angle some now see, and for which real scenarios are well suited, is thinking through supply chain risk and contingency--something many have woken up to in the wake of the Taiwan earthquake (some of the resilience implications of which I blogged about last week: here, here and here). The article continues:

For manufacturers, supply-chain resilience is a chief concern. Some have been conducting due diligence on the readiness of suppliers, knowing that if critical components aren’t delivered it will render their own contingency plans useless. Others are establishing relationships with alternative suppliers, just in case... [though] some directors privately concede that little attention is being paid to the specific challenges posed by a pandemic. Having lived through Y2K, 9/11, and Hurricane Katrina, many outfits feel comfortable that their general risk-management and business-continuity plans will suffice during a pandemic. [emphasis added]

Key differences to note: 9-11 was local; Katrina was regional; Y2K was long-anticipated. Yes they all had global effects, but none were even in the same league as a 1918-style pandemic. In no case were people unable or unwilling on a wholesale basis to travel to work for more than a few days.

All of which leads me to the conclusion the esteemed Former Secretary has already reached: surviving bird flu as a business entity may require as much clever seizing of new opportunities as it does determination to keep the old business running. In some cases there is no rational continuity strategy, only one of strategic resilience (building the capability--structurally, culturally and otherwise) to flexibly adapt to radically changed and changing circumstances.

Scenarios for H5N1; Scenarios for Resilience

This month's issue of Strategy and Business features an article on Avian Flu that begins with a fairly mild hypothetical future newspaper headline (a scenario 'event' in our lexicon):

Atlanta, November 24, 2008 — The U.S. Centers for Disease Control and Prevention today attributed two deaths in northern New Jersey and one in Philadelphia to a human-to-human transmissible strain of the avian influenza virus. Although reports of the disease among humans have been accelerating in Asia and Europe, these are the first deaths to occur in the United States. Health officials are calling for the closure of schools in the tri-state region and for citizens to limit all unnecessary travel until further notice... [italics in original]

The authors go on to make an important if somewhat exaggerated observation about resilience:

Pandemics are unique among large-scale disasters; they leave the supporting physical infrastructure of roads and buildings largely intact. Only the people who operate these lifeline infrastructures are affected.

Pandemics are not as unique as the authors would like to make them out (though at scale, we'll admit, they probably are). After the August 2003 New York City blackout, as well as in the immediate aftermath of 9-11, the availability of people for critical functions was compromised by the difficulty they encountered in moving around. In the case of the blackout that had to do with the subways not working and secondarily with the sudden capacity crunch for roads, bridges and even sidewalks as everyone wanted to get home all at once. On 9-11 the crunch came more as a result of security restrictions. 

Ultimately, the reason why people can't get where they need to go (to do their jobs and/or to recover from doing them) is less important in thinking about resilience than is the simple fact that they cannot move around efficiently - if at all.

That's something we've written about before in "Dependence on Fragile Networks".

Then there's this data-rich if still hypothetical statement:

The U.S. Department of Health and Human Services estimates that a pandemic comparable to the 1918 outbreak could cause 1.9 million deaths in the United States. In that projection, there would be 90 million people ill, 1.5 million hospitalized in intensive care, and 742,000 people needing mechanical ventilators to breathe. This situation would put a huge strain on public and private health systems; for example, there are now only 105,000 mechanical ventilators in the United States. In the U.S. alone, economic losses would range between $71 and $167 billion.

It's not clear if the 90 million ill number is all at once or over a period of months (we suspect it's the latter). Either way, that's a scenario that ought to get everyone thinking: What would I do? What would my organization do? What critical services do we take for granted today? (Telephone? Sewer? Water? Payment systems? Electricity? Gasoline? Fresh food? Roads? Police? Fire protection?) Are we really ready for a few months' collective foray into what could look, feel and smell a lot like the Middle Ages? Probably not. Let's hope we don't have to go there.

Then there's this note:

Organizational planners cannot prepare for a pandemic the same way they would prepare for other crises — identifying the most likely scenarios and anticipating ways to deal with them — but only in a more flexible way: building the capacity for continuous sensing and responding, and striving for the resilience that will be needed when disruptions or disasters occur.

That's true if one thinks of scenarios in their more traditional form: as rigidly scripted set-pieces, screened for their apparent likelihood as seen from the standpoint of today. Interactive, modular scenarios by contrast, are far more fluid affairs. They enable (even force) participants to role play, try on possibilities and rehearse together what might be seen as 'unlikely' but which in fact might be completely unpredictable. It is the process of rehearsing together more than the particulars of the scenarios themselves that give participants the awareness and shared frameworks to respond effectively when "the plan" must be thrown out the window.

Side note: The Tradesports contract on Bird Flu being confirmed in the U.S. on or before the end of 2006 is currently trading at 42. That's down from a brief high of 78 shortly after it floated in March. It's still frighteningly credible. That still leaves ample time for a human outbreak to begin in the U.S. - as imagined in the article - before the end of 2008.

Avian Flu - Predicting the Unprecedented?

In light of last Friday's Wyoming Pandemic Flu Summit and controversial comments by Secretary of Health and Human Services Michael Leavitt there ("When you go to the store and buy three cans of tuna fish, buy a fourth and put it under the bed. When you go to the store to buy milk, buy powdered milk and put that under the bed.") we were motivated to research the Avian Flu (aka 'A/H5N1') phenomenon, concluding that anyone's ability to anticipate the course of such an unprecedented potential pandemic goes beyond complex and uncertain, entering into the realm of semi-informed speculation.

Sophisticated models are not terribly useful when the only meaningful reference point (the 1918 Spanish Flu outbreak that killed tens of millions) dates from a time when disease vectors were far different and the speed and sophistication with which knowledge could be disseminated in combating them were far less.

Nor in our opinion, are prediction markets especially useful here - something that hasn't stopped Tradesports from trying. As of this writing, Tradesports bets on a confirmed report of an A/H5/N1 case in the U.S  by June 30th were going for 36 out of a possible 100, while contracts for September 30th and December 31st recently changed hands at 58 and 71 respectively. March contracts were trading at 7 - a low but not entirely speculative value. All of which may mean a lot... or very little.

Predicting two weeks in advance the geographic coverage of the typical annual flu (something we opined upon last March) is a far more tightly bounded problem than predicting the almost unprecedented - something that the IEM results prove out. We suspect that the prediction market values will swing wildly on A/H5/N1. As with weather and terrorism markets they'll digest mainstream news faster but not necessarily that much better.

In theory of course, a lone CDC scientist willing to break a news embargo could make herself several grand overnight under the right circumstances, but at the risk of her job and professional reputation. Not a great bet even if she 'won'. Prediction markets reliant on insider information (e.g., trials, the new Pope, etc.) have been amusing but seldom reliable.

Human behavior is notoriously hard to predict, as the vastly different evacuations of New Orleans and Houston last fall amply illustrate. And human behavior (of officials as well as ordinary citizens) is just one element in an impossibly complex set of interdependent variables whose interactions are likely to produce chaotic effects in the system. Computing power can only illustrate the uncertainties and generate an unmanageable raft of 'scenario'. At least with weather forecasting we have deep, rich historical data on which to draw. 

While the possibility of a pandemic must be taken seriously (infected birds from Asia are likely to make it to Alaska in a few weeks as a result of annual migration patterns, mixing with North American populations and spreading south to the Lower 48), close-in analysis of other recent catastrophes (e.g., 9-11's impact on securities industry communication grids) has led us to observe that systems in which people are positive actors for recovery are far more resilient than media hand-wringing might lead us to believe.

Cancer Research - Limits and Hopes

I didn’t intend to be away from blogging for this long (three weeks). The day of my last post, I was with my brother at the hospital when he received the definitive version of the news we’d been anticipating:

You don’t have long. Days, probably. Maybe a week or two. Go home. Rest. Pray. Enjoy your family. Finalize your affairs.

He had a few good days there, breaking dietary restrictions he’d endured these past seven months of chemotherapy and immunosuppression, telling stories, living life. Then he began to slip. His wife, their daughter and I sat vigil through hospice care: four days and three long nights. He passed peacefully in his own bed, shortly after sunrise, Friday the 14th of October, surrounded by family: as it should be. Thirty-nine years young: as it should not.

To say that business has taken a backseat this month (and frankly, this year) would be the understatement of all time. But the mental gears grind on. This whole ordeal has caused me to reflect on the many ways in which medical science is amazing... but still limited.

In February, my brother was a healthy guy with a great life just coming into bloom. That changed in less than a week. In early March, he came home from a business trip, feeling a little tired. (Who doesn't?) The next day he felt like he had a cold. (It's March, so what?) The day after that he felt he had the flu but made it in to work anyway to give an important presentation. (Wouldn't we all?.) He told me it went well. Nobody remembers what it was about.

The day after that he felt like it might be a bad case of the flu. He went out to clear his driveway anyway. (How bad could it be?) That night his wife got worried and took him to the emergency room, thinking it might be meningitis. Good thing she did. Within hours he was in chemo. They told us he wouldn’t have lasted another 48 hours without it.

Because of medical science and my sister-in-law’s vigilance, my brother lived seven months after that awful diagnosis of Acute Lymphocytic Leukemia. A young teacher at my daughter’s school missed that diagnostic window, earlier this year. She died within days.

My brother received the best care available – of that I am absolutely certain. Yet it was not enough. The beast that slew him remains thoroughly untamed. Nevertheless, it’s exciting (albeit a little sad for us), to realize that new treatments are coming down the pipeline all the time. My brother missed a new, promising clinical trial by a little less than a month.

In hopes that it not be so for others, I will be running the Boston Marathon next April in support of the Claudia Adams Barr Program in Innovative Basic Cancer Research at Dana-Farber Cancer Institute. I’ve met some of these folks. They are amazing: doing cutting edge stuff that (it’s not too presumptuous to say), has and continues to have a fundamental global impact on the understanding and treatment of all kinds of cancers.

All of the proceeds from my run go directly to the research program. (The marathon part is staffed by volunteers.) You can donate on-line here. Please think about giving generously and passing this along to friends and colleagues in your network. As my brother’s case illustrates all too well, no one is immune...

Flu Prediction Update II

As I reported last month, the University of Iowa Influenza Prediction Market seems to be doing a creditable job of forecasting the nation’s health status a week ahead of formal CDC reports, including representing the uncertainty of what turned out to be a mixed picture during the last week of January. The market will be less interesting for awhile, as we’re now into a few weeks when the answer is patently obvious. April may be more interesting. We’ll see if the market can maintain itself as outbreaks (presumably) decline over the course of the Spring.

(Side note: I had to chuckle at the rough visual correspondence between the CDC’s red/blue flu map for the last week of February and red/blue political maps from the last election. Are red-staters less healthy? Conspiracy theories anyone?)

Ridiculous Precision

A report by biotech consulting firm NanoMarkets predicts that by 2012, nano-enabled drug discovery solutions will be a $2.5 billion market. The topic is fascinating in its own right, but that’s not what struck me. I was caught instead by the astounding precision being conferred – seven years out – on a nascent market that today is just a fraction of that size.

Unfortunately this particular forecast is far from unique. Such pseudo-precise forecasts help to sell reports for the analysts that write them. Sometimes they make good headlines. What they don’t do is enable firms to recognize quickly when the forecasts may be going off track, or deal with completely alternative scenarios that may emerge.

Let’s face it: humans crave certainty. We’d like to believe that the future is as concrete and well documented as the past. It isn’t. We’re vulnerable as a species to claims that the world will be a certain way on a certain date. The real world is more complicated.

In my work with clients trying to anticipate the future and build strategic plans around a range of contingencies, we will often take a point forecast like this (i.e., “such and such a market will be of a certain size by a particular date”) and do two things with it:

1) Contextualize it by writing several orthogonal scenarios (industry visions that include chains of assumptions and events) and seeing how it ‘plays’ in each one. I.e., assess whether the point forecast will be more or less likely to be true under different sets of assumptions about overall industry conditions and drivers.

2) Establish a prediction market around it, where the payoff is tied to the actual size of the market on that date. The resulting set of moving market prices can help establish a probability curve for how large the market may grow (albeit with some qualifications.)

By doing these two things, it’s possible to turn an overly simplistic point forecast into a more dynamic and realistic framework for ongoing planning and realignment of investments. That doesn’t always have the sizzling sound-byte quality of a point forecast seven years into the future, but it’s a lot closer to the complexity and nuance of the real world.

Update: Flu Prediction

As previously noted here and here, the 2004/2005 Iowa influenza markets appear to be in full swing. Their success however, is hard to gauge. The CDC national outbreak maps on the U of Iowa site (titles of which are impossibly hard to read) seem to show forecasts that point accurately to the definitive CDC post-assessment most weeks. But the exceptions are notable: overly optimistic forecasts (i.e., less widespread flu than actual) for the last week of 2004 followed by slightly pessimistic forecasts (i.e., more widespread flu than actual) for the second week of January, 2005. (Scroll down the long page to get to more recent markets.)

"Contracts to be traded in the markets are based on the extent of influenza activity in Iowa during the 2004-2005 influenza season. Contracts will be introduced in sets of 5, with one set for each of thirteen selected weeks. The five contracts in a set cover all possible color coded activity levels as reported to the CDC by the Iowa State Epidemiologist. The color coded activity levels are those used by the CDC in its weekly report."

Several things make it difficult to assess the midstream performance of these markets. It's not clear, for example, why they cover only one week periods with one week gaps in between. Nailing the prevalence of flu in any given week is a lot harder than pinning it to a two week period. The national flu maps also beg the question of how such markets would do on a broader scale, (i.e., beyond Iowa, which seems to be having a fairly unremarkable flu season so far.) In addition, total market volumes (dollars and/or contracts), and numbers of participants in each market are hard to find on the site (e.g., compared with NewsFutures which displays volumes prominently.)

From my quick and dirty calculations drawn from this page, it would seem for instance, that trading volume during December on the 'Week 52' (Dec. 26-Jan.1) Iowa flu contract, amounted to just 534 units, worth a grand total of $169. Some Wall Street traders I know spend this much on drinks and cigars when the market's hot. That may sound like a slam on the flu markets, but it's not. For such a thin market, that kind of 'pretty close' predictive performance comes obscenely cheap compared to any prospective alternative I can think of.