Administrative Data Systems

7. Illustrations

Figure 2

Figure displaying variation in the rates of surgery among Medicare beneficiaries across 300+ hospital referral regions for 4 types of surgery; hip fracture, back surgery, knee and hip replacement, at varying levels of discretion.

"Preference-Sensitive Care: A Dartmouth Atlas Project Brief." The Dartmouth Atlas Project, January 2007.

Figure 2 presents these data, clearly revealing the increasing degree of variation as a function of the level of discretion in the surgical procedures in question. 

For example, hip fracture repair rates per 1000 Medicare beneficiaries is relatively similar across the 306 HRRs.  On the other hand, back surgery is extremely variable, with some areas manifesting rates that are almost 8 times as great as areas with the lowest rates (Weinstein, Lurie et al., 2006).  Earlier studies clearly documented the relationship between the number of medical specialists (or orthopedic surgeons) and the average level of Medicare health care spending per capita in an area, adding further weight to the argument that supply stimulates demand for care (Welch, Miller et al., 1993; Fisher and Welch, 1999).

The creation of variables that characterize the style of medical treatment in one part of the country or another has worked to define an entire field of research and to continuously push the boundaries of epidemiological, health services, and social science research.

In particular, the DAP website is a great source of data for researchers interested in understanding the intra and inter-regional differences in various health care consumption measures that might shed light on a whole new phenomenon (Wennberg, Fisher et al., 2005; Adamson, 2008).  The possibilities, particularly once the data are linked to other geographically rooted data, are limitless. Here are some examples of research data structures that DAP created:

  • Hospital Referral Regions (HRRs). One of the first major methodological advances made by the group, these determined the boundaries of a region based upon the proportion of Medicare beneficiaries who received cardio-vascular surgeries, then drew them to maximize that proportion across all regions of the country, all of which was done with Medicare hospital claims data, aggregated to region based upon the zip code of the beneficiary (McPherson, Wennberg et al., 1982).  
  • Hospital Service Areas (HSAs). Using the same approach to conducting a “patient origin” study, these are considerably smaller than tertiary referral regions (HRRs) (Fisher, Welch et al., 1992).  HSAs are smaller groups of hospitals which have the advantage of being a more coherent market for general hospital care but result in considerably more overlap among patients and physicians. 
  • Primary care service areas. DAP researchers constructed even smaller units of measure by undertaking another patient origin study that assigned zip code clusters to those beneficiaries who received a preponderance of primary care (based upon claims data containing information on type and location of service) from physicians located in those areas (Goodman, Mick et al., 2003).  In all these market definitions, the patient origin studies revealed that between 65% and 85% of the relevant type of service was provided by physicians or hospitals located in the area.