The Diffusion of Back Surgery in the U.S., 1992-2014
Frakt A, Skinner J. The puzzling popularity of back surgery in certain regions. NY Times. 2017 Feb 13. https://www.nytimes.com/2017/02/13/upshot/medical-mystery-why-is-back-surgery-so-popular-in-casper-wyo.html
Research Methods for Back Surgery Series
Methods developed by John E. Wennberg and the Dartmouth Atlas working group (1992-2000) and Brook Martin (2001-14).
Implemented by Sandra Sharp
Notes compiled by Jonathan Skinner
Maps created by Kristen Bronner
A. All rates are rates are based on the Medicare fee-for-service population age 65-99 enrolled in Parts A and B. Exclusions include enrollees who are in Medicare Managed Care for any month during the calendar year, and those eligible because of End-Stage Renal Disease.
B. Regional measures are adjusted for differences in age (5-year intervals), sex, and race (black or non-black), by Hospital Referral Region (HRR) as described in the Dartmouth Atlas Appendix
Note that the Atlas measures are adjusted to the age-sex-race composition of the elderly Medicare fee-for-service population for a given calendar year. However, there may also be changes over time in (e.g.) average age that will not be adjusted for in these estimates. However, generally back surgery declines with age, so an older age composition is not necessarily predictive of higher surgical rates.
C. There are two distinct series of data. The first set of annual rates, from 1992-2000, are based on inpatient back surgery only and come directly from the Dartmouth Atlas database. Table 1 below presents the ICD-9 procedure and diagnosis codes used to define back surgery admissions. The second set, from 2001-2014, includes both inpatient and outpatient rates, and were developed as part of an National Institute on Aging project to create databases focused on diffusion; the entire database can be accessed here: dartmouthdiffusion.org
The “seam” is therefore between 2000 and 2001. However, in comparing the new (inpatient plus outpatient) with the old (inpatient only) 2001 data, the new data is on average just 3% higher, and the correlation coefficient between the new 2001 data and the old 2001 data, by HRR, is 0.99.
D. For the 2001-14 data, “Surgery” includes the procedure variables from Brook Martin’s algorithm that defines discectomy (SPTXDECO), fusion (SPTXFUS), artificial disc (SPTXADR), or spinal distraction devices (SPTXSPACER); the defining ICD-9 and CPT codes are presented in Table 2 below (ICD-9 for MedPAR, CPT for part B and outpatient data). The incident record of spine surgery for each beneficiary was identified on an annual basis using MedPAR and outpatient claims, which was then linked on BENE_ID with Part B data to provide additional corroborating information. (We did not retain the Part-B-only sample since for early years we have only a 40% sample.)