Surgeon and hospital factors associated with the use of differentially-reimbursed hip fracture procedures Mary L. Forte DC, Beth A.Virnig PhD, MPH, Roger Feldman PhD, Sara Durham MS, Marc Swiontkowski MD, Mohit Bhandari MD, MSc, Robert L. Kane MD Research supported in part by a contract from CMS to the Research Data Assistance Center under contract #
2 Background 266,000 + hip fractures annually in U.S. Vast majority are treated surgically Intertrochanteric (IT) hip fractures: ~47% of elderly hip fractures Two devices: –Plate with screws –Intramedullary nail (IMN) New in U.S. ~1988; specific CPT code 1992
3 The devices Plate with screwsIntramedullary nail
4 How the implants compare Outcomes similar for most IT fractures: –Functional outcomes: ~same –Mortality: ~same –Complications: IMN higher –Stay-related: Length of stay: ~same OR time: ~same Blood use: IMN less (smaller incision) IMN better for unstable fxs (3-29%) No outcomes evidence to support the need for IMN for stable IT fractures
5 Background Surgeons: paid by RVUs –Two CPT codes differentiate the procedures (27244, 27245) –Surgeons paid $270 more by Medicare to use IMN than plate/screws (range $ ) Hospitals: DRGs –Both devices in the same two DRGs: 210, 211 –Not reimbursed for device costs IMN costs hospital ~$1000 more per implant than plate/screws
6 Study aim Identify the surgeon and hospital factors that were associated with IMN use among Medicare intertrochanteric hip fracture patients treated with internal fixation Identify the surgeon and hospital factors that were associated with IMN use among Medicare intertrochanteric hip fracture patients treated with internal fixation
7 IMN variation by State was not explained by patient factors in 2002 Forte et al JBJS 2008;90:691-9
8 Methods Patients: MedPAR, Carrier, Denominator files –Age 65+, Parts A & B enrolled, non-HMO –Inpatient surgery with internal fixation for IT hip fracture (MedPAR) –Exclude high-energy trauma, cancer-related, revisions, infection, bilateral fxs –Surgeon claim for specific device (Carrier) Kept first surgery per patient: 3/1/00-12/31/02 Surgeons: MPIER file Hospitals: Provider of Services (POS) file
9 Methods Analysis –Binary outcome: IMN or plate/screws –Surgeon and hospital characteristics used as predictors, adjusted for patient factors –Nonlinear mixed models: SAS Proc NLMIXED
10 Methods Predictors: –Surgeons: age, degree, Orthopaedic Board certification, Medicare IT fx case volume (quartile), # of case hospitals –Hospitals: Medicare IT fx case volume (quartile), ownership, teaching status (3 options) –Patient covariates: age, sex, race, nursing home-Medicaid assistance status Excluded: Charlson (screened in nlmixed: not sig.)Excluded: Charlson (screened in nlmixed: not sig.)
11 Patient sample 192,365 cases 3/1/00 – 12/31/02 Mean age 84 years 77% female 94% white 11% subtrochanteric (unstable) fractures 20.1% admitted from a nursing home IMN: 8% in 2000; 17.1% 2002
12 Surgeons 15,091 surgeons 15,091 surgeons Median age: 46 yrs Median volume: 10 cases (7.8%=1 case) 95% MDs, 5% DOs 65% Orthopaedic Board-certified 64% group practice 56% operated at one hospital 14% operated at 3 or more hospitals
13 Hospitals 3,480 U.S. hospitals –Median IT fx volume: 41 cases (2.3%=1 case) –Teaching status: 30.3% teaching hospitals 4.3% teaching hospitals with resident case(s) 4.3% teaching hospitals with resident case(s) –Type of ownership: 15.0% For-profit 18.5% Government
14 Results: Model selection 3 models considered: –No random effects, hospital random effects, surgeon random effects No random effects model: –Significantly worse fit by likelihood ratio test (p< for both) Surgeon random effects models fit better than hospital random effects models by AIC, BIC Patient and surgeon characteristics contributed substantially to model fit; hospital characteristics less so
15 Predictor (ref. group*) Random surgeon intercepts model Odds ratio (OR) Confidence interval for the OR Surgeon age <35 yrs <35 yrs – yrs yrs – yrs yrs – yrs* yrs* yrs yrs yrs yrs – yrs yrs – yrs 65+ yrs – 0.96 p<0.0001; p<0.05 Results: Surgeon factors
16 Predictor (ref. group*) Random surgeon intercepts model Odds ratio (OR) Confidence interval for the OR Professional degree Dr. of Osteopathy (DO) Dr. of Osteopathy (DO) – 2.81 Medical doctor (MD)* Medical doctor (MD)*1.00- Ortho. Board Certification Not Ortho. Board certified Not Ortho. Board certified – 1.31 Ortho. Board certified * Ortho. Board certified *1.00- Practice structure Group * Group *1.00- Other Other – 1.14 p<0.0001; p<0.05 Results: Surgeon factors
17 p<0.0001; p<0.05 Predictor (ref. group* ) Random surgeon intercepts model Odds ratio (OR) Confidence interval for the OR Number of IT fractures – – – * 18 + *1.00- Case hospitals one * one *1.00- two two – 1.41 three three – 1.84 four or more four or more – 3.38 Results: Surgeon factors
18 Predictor (ref. group * ) Random surgeon intercepts model Odds ratio (OR) Confidence interval for the OR Number of IT fractures – – – * 79+ *1.00- Type of ownership Non-profit * Non-profit *1.00- For profit For profit – 1.23 Government Government – 1.19 Results: Hospital factors p<0.0001; p<0.05
19 Predictor (ref. group*) Random surgeon intercepts model Odds ratio (OR) Confidence interval for the OR Teaching status Non-teaching * Non-teaching * Teaching Teaching – 1.22 Teaching-resident on case Teaching-resident on case – 2.00 Year 2000 * 2000 * – – 4.97 Results: Hospital factors, year p<0.0001; p<0.05
20 Overall findings Best model fit: patient, surgeon and hospital predictors with surgeon random intercepts; the addition of hospital predictors only minimally improved model fit after inclusion of surgeon random effects
21 Conclusions 1.Surgeon effects are stronger than hospital effects in the use of IMN for Medicare IT hip fracture patients 2.Surgeon factors, resident case involvement and teaching hospital status were strong predictors of IMN use 3.Surgeons under age 45, those operating at more than one hospital and DOs were significantly more likely to use IMN 4.The effects of higher IMN use on patient outcomes warrants further investigation
Limitations NLMIXED: one random effect Orthopaedic Board certification status, group practice: under-identified Hospital IT fx volume: IT fx-specific; may not parallel overall hospital case volume Claims data
23 Policy implications 1.Higher Medicare reimbursement to surgeons for IMN may contribute to higher IMN use when a less- expensive procedure would give similar outcomes in the majority of cases 2.IMN use can be expected to increase as long as the RVU payment incentive remains and IMN procedures are not harder to perform 3.No objective evidence exists that IMN procedures require more surgeon work 4.The process of assigning RVUs to procedures using physician/surgeon surveys may be contributing to the propagation of RVU-related financial incentives
Acknowledgement: Lynn Eberly, PhD Dept. of Biostatistics University of Minnesota