Moneyball in Healthcare

Its tempting to draw parallels when a new paradigm emerges. Normally, one would not think of many similarities between the glamorous world of baseball and the stodgy, sterile world of healthcare. However, when it comes to the central theme of "Moneyball" - the power of data to outpredict the "instincts" of the experts - there may be many things in common. Medicine has traditionally operated much more in the "artisan" mode than the "scientist" - there is often a strong belief in the experience and instinct of the practitioner, and the practitioner often resents the intrusion of "data" or "evidence-based (cookbook) medicine" into his/her practice.

Unfortunately it has been shown time and again that experience and intuition alone are not sufficient at producing the best outcomes for a specific patient or even for a population. Intuition is often led astray by the "recall effect" - the most recent cases that a practitioner has seen, and particularly negative outcomes, can often be over-emphasized in a practitioner's mind.

Furthermore, the physician's intuition - much like the scouts in Moneyball - may be led astray by subtle biases - the way someone looks, talks, walks, color of skin, religion, etc. It happens much more often than many physicians would like to acknowledge - patients are attributed as "drug seeking" thereby leading to unnecessary struggles between the care providers about their course of treatment, emotive labels of "unfortunate" or "pleasant" are applied in the medical workup that actually have no role or meaning in the care of the patient, but may paradoxically taint the medical encounter.

In this recent posting on KevinMD, Drs. Jerome Groopman and Pamela Hartzband discuss common types of cognitive error that can contribute to undermining of logic, intuition, and experience:

http://t.co/GmntggS4

(1) "Anchoring" - fixation on a particular bit of information, which can then lead us down a particular path, closing down prematurely the possibility of other paths.

Example: A patient was admitted with abdominal pain and elevated WBCs. Abdominal ultrasound revealed some thickening of the gallbladder. The diagnosis was made of "acute cholecystitis" and patient was taken for cholecystectomy. However, the patient also had gram positive cocci in a blood culture; this finding was not congruent with the diagnosis, but was ignored because of the diagnostic path that the clinicians had taken. The patient turned out to have sepsis, resulting in a poor outcome.

(2) "Availability" - similar to "recall" bias, powerful memories tend to hold greater sway in our decision making. Patient deaths tend to have this impact, and although certainly should hold a place in a clinician's memory, they can be detrimental when they exert an undue influence.

Example: A practitioner prescribes long-term anticoagulation for a number of patients after discharge because one of his recent patients died of a pulmonary embolism after hospitalization.

(3) "Attribution" - excessive influence of stereotypes; this is a bias that all of us have to be vigilant for in our day-to-day lives, as such it should be no surprise that it also plays an undue role in medical decision-making.

Example: Countless patients are "labeled" as "chronic pain" or "frequent flyers" - in part because of their socio-economic background - only leading the providers astray from their goal of helping each patient.

(4) "Confirmation" - contradictory data is discarded in order to make a "neat" diagnosis. This error was also present in the example of gram positive bacteremia given above.

(5) "Satisfaction of Search" - stopping the search for an underlying diagnosis because we have found one potential diagnosis. This error was also found in the gram positive bacteremia example above.

Drs. Groopman and Hartzband propose a few remedies to combat these errors, asking clinicians to pose these questions:

(a) What else could it be?

(b) Does anything else fit?

(c) Could there be more than one process at work?

While data is not something that the doctors discuss directly in this post, synthesizing data (for example from clinical trials, evidence-based medicine, systematic reviews, patient registries) into the medical decision-making process should provide a further grounding for sound decision-making.