KOLAHUN

The Perversion of Oncology Economics

In an insidiuously chilling editorial in the NEJM, the intersection of micro- and macro-economics of oncolgy is explained.  Oncology through the 20th century consisted of using available poisons to hurt cancer cells (bad guys) quicker than other body cells (good guys).  Given the obvious potential to seriously harm patients, only few physicians did so.  This anachronistic process led to these few physicians 'selling' chemotherapy medicines out of their office.  

Towards the end of the 20th century, this vending became incredibly profitable as cancer medications became more expensive (and as people stopped dying of communicable diseases with better public health, vaccines and antibiotics).  But as the rate of medical costs soar, medical payers (CMS and private insurers) have begun focusing on decreasing oncology care costs.  And so we now end up with a non-sustainable system.

But what's the actual problem?  Greedy pharmaceuticals - not really, as they're not humanitarians, they're capitalists.  The governmentn - well, for anyone to the left of Rick Perry...perhaps.  But what exactly is the government supposed to do - pay more for medicines when it doesn't need to?  

No, I'd say the fault lies in the fundamental pay scheme of modern American medicine which emphasizes pay-for-service rather than pay-for-outcome.  Here's a model of how it would work, per the NEJM article:

One solution is adopting clinical pathways for which practices are paid disease-management fees that are not based on chemotherapy sales. For instance, one large oncology group has developed care pathways specifying preferred drug combinations and sequences — for example, allowing only a few first-line, mostly generic regimens for patients with non–small-cell lung cancer, as compared with the 16 possible drugs and many more combinations included in National Comprehensive Cancer Network pathways. This approach has been shown to result in equal or better survival, less use of chemotherapy near the end of life, and 35% lower costs than usual care.

"The Plural of Anecdote is not Data"

Nancy Pelosi channeled the above phrase from an unknown source in the political sphere.  But in the sphere of medicine, sometimes clinicians are confronted with the case that, in fact, no data exist and all that's left to help guide decisions are the plural of anecdote, whatever that may be called.

Here's a NEJM article in which pediatricians Stanford confront the problem.  Of note, among the various fields in medicine, few are as poorly researched as pediatrics:

Without clear evidence to guide us and needing to make a decision swiftly, we turned to a new approach, using the data captured in our institution's electronic medical record (EMR) and an innovative research data warehouse.

They simply did a quick case-control study withing their own institution's database, found some apparently north-star-ish advice and made a clinical treatment decision.  Luckily for the patient and them, the situation worked out for the best.  They then conclude grandiosely:

Our case is but one example of a situation in which the existing literature is insufficient to guide the clinical care of a patient. But it illustrates a novel process that is likely to become much more standard with the widespread adoption of EMRs and more sophisticated informatics tools. Although many other groups have highlighted the secondary use of EMR data for clinical research,2,3 we have now seen how the same approach can be used to guide real-time clinical decisions. The rapid electronic chart review and analysis were not only feasible, but also more helpful and accurate than physician recollection and pooled colleague opinion. Such real-time availability of data to guide decision making has already transformed other industries,4 and the growing prevalence of EMRs along with the development of sophisticated tools for real-time analysis of identified data sets will no doubt advance the use of this data-driven approach to health care delivery. We look forward to a future in which health information systems help physicians learn from every patient at every visit and close the feedback loop for clinical decision making in real time. (Emphasis added)

These authors have identified a fundamental problem in clinical analysis - the volume of perplexing clinical questions supersedes the available volume of decent studies answering them.  In this case, most clinicians revert to, what worked yesterday, or the day before.  And clinicians with longer memories can go back even further.  And those special clinicians with higher computing power can even figure out mental matrices in which they subdivide patients based on comparative characteristics.  

But, of course, what these Stanford pediatricians have done is outsourced the mental computing to a local computer which does the hard work.  And good for them and their patients.  Perhaps the epidemiologists can figure out how to synthesize these datum to produce meaningful statements, similar to meta-analyses.  

But clinicians shouldn't fool themselves or their patients into accepting that this is 'optimal' medicine.  It's not.  It's second best.  Nor is this a problem if clinicians de-throne themselves from the alter of deities and begin to explain to patients the uncertainty in diagnoses and treatments.  

Take Two Twitters and Call Me in the Morning

Chorbev et. al describe in the Journal of Telemedicine and Applications a plausible model for improved diabetes through group share.  They set up a platform where:

We have developed a healthcare social network that provides methods for distance learning; opportunities for creation of virtual self-help groups where patients can get information and establish interactions among each other in order to exchange important healthcare-related information; discussion forums; patient-to-healthcare specialist communication. The mission of our virtual community is to increase the independence of people with diabetes, self-management, empower them to take care of themselves, make their everyday activities easier, enrich their medical knowledge, and improve their health condition, make them more productive, and improve their communication with other patients with similar diagnoses.

On one hand, one of the flaws with this is that this network captures those who already are invested in improving their care - the self-selection bias.  If you're online and have the choice between TMZ (Humpdashian!) versus your HbA1C, even with marginal costs, the uninterested patient will choose Humpdashian.  But on the other hand, the Gladwell understanding of the social network model, would argue that given the low cost of interaction (send some emails, perhaps check your sugars more often or take your insulin more regularly), even with low motivation, the participation of these patients should be high.  So while each patient doesn't have to do so much, the fact that many are doing a bit improves substantially the median health outcome.

More systematically, the Gladwell argument discounts the power of networks to find answers.  The beginning examples of chaos theory generally use the patterns of birds or fish as the analogy to describe the complex patterns developed without hierarchy:

Chaos theory looks at how very simple things can generate very complex outcomes that could not be predicted by just looking at the parts by themselves.
 
You may have noticed the wonderful swirling patterns of birds flying together in the sky or fish schooling in the oceans. At first glance we would think the birds would have to be rather intelligent to work out how to fly in formation like that. We would probably also assume there must be a 'bird in charge' giving  instructions to all the others. 

Research into swarms has shown, however, that all that is required is for each bird to maintain the distance between itself and its neighbours and fly in the average direction of its neighbours. From this alone the wonderful, swirling, complex patterns the birds or fish make are seen. Simple rules can generate complex behaviours that just seem to emerge out of nowhere. While each individual agent does not need much intelligence, the is swarm intelligence which resides with the collective.

And this is the argument put forth by the social network as improved health model.  It's actually the feature (rather than the bug) of this model that it's non-hierarchical.  In this way, simple modifications to the program or to the rules set forth to the patients can have large, cascading effects.  And done well, these can still be inexpensive in cost to the patients (effort) and to society ($$).