Conventional wisdom that provider capacity drives cost questioned in new study


Remember the idea that coordination will improve care? Well, if physicians do not get timely reports from other providers, their patients seem to have lower costs!!!  This from a new study out from the Center for Studying Health System Change. 

More importantly, this paper throws cold water on the idea that providers generate a lot of unnecessary cost to fill up excess capacity in the delivery system.   As you know, conventional wisdom driven by the Dartmouth Atlas and other studies has it that that care utilization and cost can vary sharply across regions without being tied to better outcomes, e.g. when there are a lot of ICU beds in a region, then a lot of patients end up in ICUs; when there are a lot of specialists, a lot of specialists tend to get seen, etc. Some studies suggest as much as a 50% cost gap between top decile and bottom decile geographies.

But this new study found – once you adjust for health status of the patient — no real correlation between cost and a variety of provider system structural features (such as bed capacity).


Even if you can change the delivery system structure, don’t expect any real reduction in cost.  Therefore, the study suggests, a lot of public policy and health plan strategy may be focused on a much smaller cost reduction prize than many people hope.

  • Additional thoughts: It is a pretty dense article — lots of OLS regressions, etc — and the methodology warrants close scrutiny. Two possible cautions:
    1) the analysis explains only about 50% of the variation in cost for the very sick patients and 23% of the variation for healthier patients (see Table 4). Not bad for cross-sectional analysis but there still is a lot not explained in the model
    2) many of the structural variables look like they were measured on a county basis. But obviously medical markets operate across counties, especially for the severely sick who drive the costs. In Boston, for instance, the big AMCs serve at least 5 to 6 counties (in fact, one can argue they serve 3 or more states). So the relevant structural factors related to the cost of a particular patient (e.g. availability of beds in the nearby tertiary care center) may not be measured correctly.

    Putting these points aside, the statistical results (Table 4 in the study) are mostly a sea of statistical insignificance, consistent with the core thesis of the article. Take a look: perhaps your favorite idea about what drives healthcare costs has been tested and found wanting :)
    Provider in a solo practice? doesn’t matter
    Provider with a staff model HMO? No significance
    Provider has a lot of capitated care? Nope
    Lots of specialists in the county? You guessed it. nada…
    There are some variables which are significant. For example, for profit status of the beds (hospital or SNF) appears to add cost for the very ill. But not that much.