Different attribution rules lead to different assessments of physician efficiency

Warning – a wonky post.

Ateev Mehrotra and colleagues from RAND have analyzed how different rules about what physician is responsible for a patients’ care can lead to impressively different efficiency scores for physicians.   

They evaluated 2 years of medical and pharmacy claims data from 1.1 million commercially insured members in Massachusetts, representing about 80% of those with employer-sponsored insurance.   Claims were assigned to episodes of care using Episode Treatment Groups, ETGs, an industry standard.    Then, twelve ‘attribution rules’ were used to assign the costs of each ETG to one or more physicians.  These rules ranged in how much professional or evaluation and management cost a physician must be responsible for to be credited with the cost of the episode.  Some of the rules allowed crediting a single case to multiple physicians, and others did not.    Each physician was allocated to a high, average or low cost group (and some physicians were in a fourth group, without 30 evaluable episodes). 

The choice of ‘attribution rule’ mattered a lot.  Compared to the ‘standard’ rule,  different attribution assigned a physician to a different efficiency group between 17 and 61% of the time. 

This is an especially elegant study.  Mehtrotra standardized prices so that this evaluation is really all about utilization, not cost per unit.  In the ‘real world,’ though, cost per unit helps drive overall cost – so it’s possible that this study should be repeated without standardizing cost.   The researchers provide data in an appendix showing that most of the difference among the different attribution rules is movement of efficiency categories – not movement between efficiency category and insufficient sample size.  

This matters.   We recognize there is a large utilization difference between “efficient” and “inefficient” physicians.  If the measurements of physician efficiency are this unreliable, it is far harder to develop tiered networks to encourage patients to use more efficient physicians.    This also helps explain physician unhappiness at being assigned to different tiers in different health plan rating systems.

This might be another reason we should profile groups of physicians, rather than individual physicians.