Sunday, February 03, 2008

GSK, Avandia and Medical Informatics: More on Why Pharma Fails

"If I have seen further it is by standing on the shoulders of Giants."- Isaac Newton
"We would like to pick the low hanging fruit while standing on the shoulders of Giants, but we got rid of them all." - Felix Fulmer

The GSK diabetes drug Avandia and the Pfizer anti-smoking drug Chantix are new entries in a long "honor roll" of drugs from major pharmaceutical companies to come under suspicion of having an unacceptable degree of adverse drug effects (ADE’s).

The recent Avandia issues reminded me of assertions about need for more medical informatics expertise in pharma to help track drug ADE’s and improve drug safety. These recommendations come from organizations such as Gartner Group and from the prestigious Institute of Medicine of the National Academies (link).

The Avandia issue also reminded me of essays I’ve written about the lack of understanding of the formal specialty of Medical Informatics in the pharma industry e.g., "Why Pharma Fails" and "We don't need medical informatics here Part 2." I am also reminded how this problem, complicated by the conflation of IT and information science, cause the industry to suffer at both ends of the pipeline, that is, discovery and postmarketing surveillance.

The Avandia debate also reminded me of the preposterous, contrary to reason concept that in clinical information/IT initiatives there can be "too much talent" in an organization, and of ill-judged biases and stereotypes regarding Medical Informatics and MD's in the same context, especially regarding leadership roles.

Finally, in those cases and others I have written about the dangers of non-medical personnel holding leadership positions in healthcare that permit them to make decisions best left to medical or cross-disciplinary (e.g., medicine and information science) professionals, or otherwise put them outside the bounds of their competencies.

All this brought me back to GSK and, on searching my files, yet another experience I had with pharma illustrating my concern about their hiring practices. The experience may also shed some light on why pharma seems to have a problem with drug monitoring. I use this case as an example, but have had similar experiences with AstraZeneca, Wyeth, McNeil, Johnson and Johnson, Merck, and others over the past several years.

In July 2004 I received an unsolicited email about a new job opening from a prestigious British recruiting firm (“headhunter”) retained by GSK, Armstrong Craven Ltd. It referred to a cheminformatics position I'd once applied to (and never heard back about), and sought me out for a converstation about a new role: Director of Medical Informatics (text of letter here).

I set up a phone interview with the recruiter, who called me from the UK a short while later. We spoke for over an hour about the Director of Medical Informatics role. As a result of our detailed conversation, the recruiter felt I was highly qualified for the position. As a further check, he emailed me a GSK document that described the position in great detail. It included a description of the Informatics and Knowledge Management unit at GSK, and one concerning item:

Medical Informatics is a discipline that studies, invents and implements the structures and algorithms necessary to improve communication, understanding, value derivation and management of pharmaceutical scientific information. The objective is the coalescing of data, knowledge and tools necessary to apply these data and knowledge in the decision making process, at the time and place that a decisions needs to be made.

I felt this definition of medical informatics a bit dated. It was written by a pioneer who is now semi-retired, Homer Warner, MD, PhD of Univ. of Utah, who I knew from various informatics professional meetings. However, it was clear why the recruiter was interested in me.

With extensive medical informatics training and experience in both hospitals and pharma, having been Director of Scientific Information Resources and of The Merck Index at Merck Research Labs, I must have seemed to him well qualified for the job description. That JD included items such as identifying, prioritizing and partnering on projects where Medical Informatics can deliver business impact, and developing both strategic (long term) collaborations that will have a sustained impact as well as short term efforts that will impact high-priority R&D projects. The full job description can be seen here.

Finally, the job was described as reporting at a high level but not to an M.D. It reported to the Vice President of Informatics & Knowledge Management who apparently authored the description, Dr. Tomas Flores, a Ph.D. The recruiter told me he would speak with the hiring team and present my credentials to them ASAP. He fully expected at least a follow up interview.

Unfortunately, this was not to occur.

I received a call back from the recruiting firm several days later to tell me that GSK had no interest in my background, and Dr. Flores did not wish to speak to me at all. When I inquired why, I was told he was looking for someone with “an extensive CS background to write algorithms to solve business problems.”

Needless to say, I was rather disappointed by this response. The Armstrong Craven response raises a number of important questions:

  • Why do we have computer-oriented VP's lacking medical credentials evaluating resumes of those who do?
  • Did the lack of medical credentials and experience prevent proper understanding of critical line items on my resume?
  • Why was a dated definition of "Medical Informatics" being employed in this job description for a new Director of Medical Informatics, when broader, more current definitions of the field were readily available?
  • Was the Homer Warner definition itself taken far too literally regarding "algorithms?" Could Dr. Warner have meant something more than computer algorithms?
  • How was the prejudicial decision that algorithms needed to be written by a Medical Informatics specialist made? Who set this as a strategic objective? Why was it done by someone with no apparent Medical Informatics credentials?
  • Is the best use of a medical informatics specialist to write algorithms (i.e., do programming), or is it to create and direct the writing of algorithms, separating the high-level biomedical cognitive tasks form the lower ones (i.e., programming)? See example of this issue here.
  • Was there anything in my resume with regard to my training and experience in computing and in healthcare suggesting I could not write such algorithms to solve biomedical business problems?
  • Was it typical at GSK for a hiring manager to read a resume that came from its own retained recruiter and, on the basis of secondhand knowledge from the recruiter and line items on a piece of paper, make a decision not to even talk to the recommended candidate – or even meet in person, since that candidate lived perhaps fifteen miles from one of the sites the hiring manager frequented? Is this above-board, state-of-the-art talent management in an industry as critical and under such societal mistrust as pharma?
  • Finally, was my resume actually read? Was this a preferential job posting? Was there a favored internal candidate? (HR departments, usually annoyed when lawsuits happen, force hiring managers to advertise new positions per company policies on same, rather than allow favoritism. However, HR cannot challenge hiring managers' stated reasons why outside candidates they might not have expected to apply might not be suitable for a role.)
  • Were other potentially valuable candidates turned away on similar reasoning?
  • Do such decision makers also make decisions concerning who gets to bid on their own job, and ultimately, who stays and who leaves during downsizing?
  • Might that be a "double whammy" adversely affecting the talent pool at both ends, creating a “brick-headed” wall at the intake end, and a “leaky membrane” at the exit end?
Some of these questions can be answered by excerpts from my response to Armstrong Craven, cc’d also to the hiring manager (full text here):

Dear (recruiter),

Thanks for presenting my resume to GSK regarding a position in Medical Informatics. Your colleague explained to me yesterday that the GSK Informatics & Knowledge Management group is seeking someone with an "extensive CS background to write algorithms to solve business problems."

... by way of education ... It is my belief that a view of medical informatics professionals as "writers of algorithms to solve business problems" reflects a fundamentally narrow and mechanistic view of the field, or perhaps a mislabeling of the position as being one of Medical Informatics. The lack of a requirement for formal Medical Informatics education and training suggests the latter.

One of Medical Informatics' founders, Dr. Homer Warner, was the author many years ago of the definition adapted for the ArmstrongCraven draft brief. Homer wrote that "medical informatics is the study, invention and implementation of structures and algorithms to improve communication, understanding and management of medical information." See my website on this and other definitions of the field at .

By "algorithms", however, Homer meant not computer algorithms but more broadly the development of processes, procedures, methodologies -- and software systems -- for management, process control, decision making and scientific analysis in healthcare. That is the current understanding of the field that my academic colleagues and I teach our students.

The value of the field stems not so much from writing algorithms (as in bioinformatics or computational biology), but from an integrative approach focusing on data definitional issues, language issues, human factors, organizational issues, information needs and flows, and strategies for indexing, retrieval and information dissemination. It is in the latter areas that I excel (although having done the former -- the writing of software and algorithms -- earlier in my career).

My definition of the field is that Medical Informatics is the science and art of modeling and recording real-world clinical concepts and events into computable data used to derive actionable information, based on expertise in medicine, information science, information technology, and the scholarly study of issues that impact upon the productive use of information systems by clinical personnel.

Perhaps a better definition of Medical Informatics is found in MeSH itself, written by the National Library of Medicine itself: "Medical Informatics is the field of information science concerned with the analysis and dissemination of medical data through the application of computers to various aspects of health care and medicine."

I received a thank you from Armstrong Craven for helping them understand the field better (which helps their competitive position in the headhunting domain), but did not receive a reply from the company hiring manager.

Even worse, this was not my first encounter with GSK / SKB. I had also had fruitless discussions with SKB in the 1996-8 timeframe, while I was Director of Clinical Informatics at Christiana Care. In February 2000 as well, a senior recruiter named Iain McKenzie had contacted me, interviewed me in person after a phone discussion and then enthusiastically tried to sell my background to GSK discovery scientists. He was quite puzzled that they expressed no interest.

In conclusion, one wonders if pharma really "gets it" regarding Medical Informatics. Many of my observations indicate the industry does not. That the IOM needs to call pharma's attention to a specialty field whose professional education is largely NIH-sponsored is quite revealing.

One also should wonder if it is wise to put nonmedical personnel in charge in pharma of what basically amounts to the hiring of medical/medical information professionals.

One also wonders if such hiring practices impair companies’ abilities to monitor the safety of their drugs, and if Avandia and other safety issues would have been detected earlier if, as IOM suggests, more Medical Informatics professionals were involved in roles beyond "writing computer algorithms to solve business problems."

I report, you decide.

Finally, I'm not sure I would have wanted to work for this company anyway. For some of the reasons why, see my post "How To Demoralize Pharma Workers in One Easy Lesson."

-- SS

No comments: