relationship between the volume of total hip replacements performed by providers and the rates of...
TRANSCRIPT
The PDF of the article you requested follows this cover page.
This is an enhanced PDF from The Journal of Bone and Joint Surgery
1997;79:485-94. J Bone Joint Surg Am.KREUTER HANS J. KREDER, RICHARD A. DEYO, THOMAS KOEPSELL, MARC F. SWIONTKOWSKI and WILLIAM
of Washingtonby Providers and the Rates of Postoperative Complications in the State Relationship between the Volume of Total Hip Replacements Performed
This information is current as of August 5, 2010
Reprints and Permissions
Permissions] link. and click on the [Reprints andjbjs.orgarticle, or locate the article citation on
to use material from thisorder reprints or request permissionClick here to
Publisher Information
www.jbjs.org20 Pickering Street, Needham, MA 02492-3157The Journal of Bone and Joint Surgery
Copyright 1997 by The Journal of Bone anlf Joint Surgery, Incorporated
Relationship between the Volume of Total Hip Replacements Performed by Providers and the Rates of
Postoperative Complications in the State of Washington* BY HANS J. KREDER, M.D.f, RICHARD A. DEYO, M.D4, THOMAS KOEPSELL, M.D4,
MARC F. SWIONTKOWSKI, M.D.§, AND WILLIAM KREUTER, M.P.A.l, SEATTLE, WASHINGTON
Investigation performed at the University of Washington, Seattle
ABSTRACT: Since the late 1970's, an empirical relationship between the volume of procedures performed by a provider (a hospital or surgeon) and the outcome has been documented for various operations. The present study examines the relationship between the volume of hip replacements performed by surgeons and hospitals and the postoperative rate of complications. A statewide hospital discharge registry was used to identify patients who had had an elective hip replacement between 1988 and 1991. Patients who had had a revision procedure, who had been referred on an emergency basis, or who had had a diagnosis of a fracture or a malignant tumor on admission were excluded. There were 7936 eligible patients who had had 8774 hip replacements. The average annual number of all hip replacements performed from 1987 through 1991 was subsequently determined for each hospital and surgeon who had cared for at least one patient in the study cohort. The rate of operative complications was modeled as a function of the volume of procedures performed by the surgeon or hospital (the surgeon or hospital volume), with adjustment for the age of the patient, gender, co-morbidity, and operative diagnosis.
We noted significant differences in the case mix of low-volume providers compared with that of high-volume providers (p < 0.01). In general, surgeons and hospitals with a volume below the fortieth percentile
*No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. Funds were received in total or partial support of the research or clinical study presented in this article. The funding sources were the Orthopaedic Research and Education Foundation and The American Academy of Orthopaedic Surgeons through the Health Services Fellowship.
tMuseuloskeletal Health Status Working Group, Division of Orthopaedics, University of Toronto, Sunnybrook Health Science Center, Suite A-315, 2075 Bayview Avenue, North York, Ontario M4N 3M5, Canada. E-mail address: [email protected].
^Department of Medicine, Clinical Scholars Program, University of Washington, Varsity Apartments, Box 355300, 3747 15th Avenue, N.E., Seattle, Washington 98105.
§Department of Orthopaedics, University of Washington, Har-borview Medical Center, Box 359798, 325 Ninth Avenue, Seattle, Washington 98104.
^Department of Health Services, University of Washington, Box 354807, Seattle, Washington 98195-4807.
managed patients who had a more adverse risk profile in terms of age, co-morbidity, and diagnosis. Even after adjustment for the case mix, there was a significant relationship between surgeons who averaged fewer than two hip replacements annually (low-volume surgeons) and a worse outcome (p < 0.05). Patients managed by these low-volume surgeons tended to have higher mortality rates, more infections, higher rates of revision operations, and more serious complications during the index hospitalization. The duration of hospitalization was inversely related to surgeon volume and directly associated with hospital volume. Hospital charges were inversely related to hospital volume, even after adjustment for patient-related factors as well as the duration of hospitalization, the year of the operation, and the destination after discharge (p < 0.05).
More detailed information is required to investigate the reason for these observed variations in the rates of complications. If future studies confirm an association between low-volume providers and an adverse outcome, performance of some types of elective total hip replacements at regional centers should be considered.
The expected rate of adverse outcomes after various operative procedures depends in part on factors such as the age of the patient, severity of the disease, and co-morbid conditions482430. It has been suggested that the experience of the surgeon and hospital with the specific procedure is an additional factor that may affect the rate of operative complications1214'1618. With regard to total hip arthroplasty, issues that continue to be debated revolve around the selection of candidates for primary and revision operations, the type of prosthesis that should be used, the best type of fixation, the proper operatingTroom environment, the best way to prevent venous thrombosis and infection, the appropriate operative approach, and the best program for rehabilitation after the operation. The experience of the hospital and surgeon with total hip arthroplasty might affect the outcome in several ways. Experienced providers might make more appropriate decisions regarding the indications for the operation and other operative details as
VOL. 79-A, NO. 4, APRIL 1997 485
486 H. J. KREDER ET AL.
they gain technical expertise and learn which factors result in a better outcome. Moreover, rehabilitation"and other important ancillary services may be more readily available to providers who care for a high volume of patients. With large numbers of similar cases, both hospitals and surgeons may develop routines that minimize the risk of errors in^treatment.
More than 200,000 hip replacements are performed annually in the United States22; thus, even a low rate of preventable complications could affect a large number of people. Excellent results have been reported after total hip arthroplasties performed by high-volume providers in both academic centers31 and community settings21. There is little information in the literature regarding the outcome specifically for patients who are managed by surgeons or hospitals that perform a low volume of procedures7-15,16,23. As far as we know, most volume-related outcome studies reported to date have failed to separate elective primary hip replacements from revision operations and emergency procedures (for acute fractures of the hip). The unequal distribution of these factors across low and high-volume providers could lead to biased conclusions. Moreover, only one of the previous studies23 excluded individuals who had a malignant tumor. Patients who have a reconstructive operation because of neoplastic hip disease have a much worse prognosis than those who do not have cancer. If valid comparisons are to be made across providers, it is essential to consider the varied case mix for the surgeons and hospitals involved. Although a detailed prospective study would shed the most light on this topic, the large number of patients that is required in order to detect variation in the prevalence of low-frequency complications makes such a study prohibitively expensive. Analysis of computerized administrative data sets allows the records of a large number of patients to be reviewed in order to identify trends that can be investigated in more detail later.
The purpose of this study was to evaluate the relationship between the volume of all total hip arthroplasties performed by a provider (a hospital or surgeon) and the rate of complications after elective total hip arthroplasty in the State of Washington. Specifically, our aims were to describe the case mix as a function of provider volume and to analyze case-mix-adjusted rates of operative morbidity and mortality as a function of provider volume.
Materials and Methods
Selection of the Study Cohort
The Comprehensive Hospital Abstract Reporting System (CHARS) computerized data set of the Washington State Department of Health was used for this Study. This data set is restricted to inpatient admissions and includes the surgeon, the hospital, and encrypted patient identifiers as well as the procedure and diagnostic codes of the International Classification of
Diseases, Ninth Revision (ICD-9). Some information regarding the hospital admission and patient demographics is also available. We initially identified all records with a procedure code of 8151 or 8159 during the years 1988 through 1991 (Appendix I). These dates were chosen so that at least one year of follow-up data were available for all patients, as well as one year of information preceding the earliest study date in order to calculate co-morbidity scores for all patients. Admissions labeled as elective with diagnostic codes for osteoarthrosis, inflammatory disease, avascular necrosis, or late post-traumatic osteoarthrosis for patients who were more than eighteen years old at the time of admission were selected for additional consideration. All records that identified the source of admission as the emergency department or the diagnosis as fracture or malignant tumor (in any anatomical region) were excluded. Finally, any hospitalization that met the criteria for a revision (a procedure code for revision or for removal of the hip prosthesis in conjunction with a diagnostic code suggesting a complication related to a prosthetic device; Appendix II) was excluded from consideration.
Definitions of Variables
The hospital discharge data were linked to the Washington State Death Index, an abstraction of all death certificates filed in Washingtorl. The dates of all deaths were recorded. Patients who were not identified as dead in the match were assumed to be living at the end of the study period (December 31,1992).
All admissions to the hospital subsequent to an elective index admission for total hip arthroplasty were identified. A revision or infection was deemed to have occurred if specific diagnostic and procedure codes pertaining to a subsequent admission were recorded in the hospital discharge database (Appendix II).
Serious complications during the index hospitalization were documented as a binary variable (no major complications compared with one major complication or more). Only codes for diagnoses related to operative mishaps, myocardial infarction, stroke, and the like were included in the definition (Appendix II). In order to minimize the effect of coding inaccuracies for conditions that are subject to wide interpretation, no attempt was made to evaluate less serious events such as postoperative anemia".
Surgeon and Hospital Volumes
The term provider is used to refer to both hospitals and surgeons. For each surgeon and each hospital, the annual number of hip replacements, including hemiarthroplasties and revision total hip procedures, performed from 1987 through 1992 was determined, and this number was designated the surgeon or hospital volume. No exclusion criteria were applied as it was reasoned that experience is gained with any hip replace-
THE JOURNAL OF BONE AND JOINT SURGERY
RELATIONSHIP BETWEEN THE VOLUME OF TOTAL HIP REPLACEMENTS PERFORMED BY PROVIDERS 487
TABLE ]
CHARACTERISTICS OF SURGEON AND HOSPITAL-VOLUME GROUPS
Surgeon Volume Hospital Volume
Variable Low Medium High Low Medium High
Volume percentile <40th 40th to 80th >80th <40th 40th to 80th >80th
No. of providers 170 246 78 27 25 15
No. of patients Per year per provider <2 2 to 10 >10 <16 16 to 65 >65 Total 280 3125 4531 476 2667 4793
Average hospital volume per year 77.1 87.5 113.1
Average surgeon volume per year 9.4 42.7 143.8
ment regardless of the surrounding circumstances. The year 1987 was included in order to gain information regarding the volume of replacements performed by the provider (the provider volume) for at least one year before the earliest hip replacement performed in the cohort. The average annual number of hip arthroplasties performed during the study period was then determined for each provider. Descriptive statistics were calculated to divide providers into five equal groups on the basis of the twentieth, fortieth, sixtieth, and eightieth percentiles. Although this method ensures an equal number of providers in each group, the number of patients managed by each group is markedly skewed as a given number of providers in the lowest-volume group will have managed far fewer patients than the same number of providers in the highest-volume group. In our preliminary analyses, we noted that the small number of patients in the lowest-volume group resulted in poor statistical power and parameter estimates with wide confidence intervals. In order to increase the number of patients in each volume group, the two lowest-volume groups and the two middle-volume groups were combined. Thus, three groups of surgeons and hospitals were studied: low-volume (a volume below the fortieth percentile), medium-volume (a volume between the fortieth and eightieth percentiles), and high-volume (a volume above the eightieth percentile).
Covariates
All reported statistics were adjusted for age, comorbidity, gender, and diagnosis. The operative diagnosis was based on an algorithm depending on the presence of certain codes (Appendix II). With this algorithm, a dichotomous diagnosis variable was modeled as osteoarthrosis compared with other forms of arthropathy, such as inflammatory disease, avascular necrosis, or late post-traumatic osteoarthrosis.
The co-morbidity score was calculated on the basis of hospital discharge records at the time of the index admission and during the previous year according to the method developed by one of us (R. A. D.) and colleagues5. This score, which is based on work by Charlson et al.3, is calculated as the weighted sum of values assigned for various serious medical conditions.
Analysis of Data
Data were analyzed with SPSS UNIX version-5.0 software (SPSS, Chicago, Illinois) on the University of Washington IBM RS-6000 computer. For each patient, the occurrence or absence of death, infection, and revision within three months and within one year after the time of the index hospitalization was recorded. The duration of hospitalization and the hospital charges were evaluated as continuous outcome variables. Ordinary linear or logistic regression requires that all observations be statistically independent of each other. This assumption was violated in our data as multiple patients had received care from the same hospital or surgeon. For this reason, generalized estimating equations suitable for correlated data were applied with an SAS macro32 running under SAS UNIX version-6.09 software (SAS Institute, Cary, North Carolina) on the University of Washington IBM RS-6000 computer. Interactions between covariates, surgeon volume, and hospital volume were individually tested for every model. The interaction term between hospital and surgeon volume was also evaluated by entering it into the model after all main effects had been included. None of the interaction terms was found to reach significance. For clarity, and in order to maximize the degrees of freedom, only the main-effects model data are presented.
Results
Surgeon and Hospital Volume Four hundred and ninety-four surgeons performed
at least one elective primary total hip arthroplasty from 1988 through 1991; these procedures were done in sixty-seven different hospitals. The provider-volume distribution was skewed, with a large number of surgeons and hospitals caring for a small number of patients (Table I). Although only the providers who performed at least one elective primary hip replacement are considered, the volume calculations include all of the hip replacements that were performed (without any exclusion criteria).
Study Cohort
Of the 18,081 patient records that contained the procedure codes 8151 or 8159 from 1988 through 1991,
VOL. 79-A, NO. 4, APRIL 1997
488 H. J. KREDER ET AL.
TABLE II
UNADJUSTED PATIENT CHARACTERISTICS BY PROVIDER GROUP*
Surgeon Groups Low Medium High Low
Hospital Groups
Medium High
Average age (yrs.)
Co-morbidity (per cent) >0 >2
Male gender (per cent)
Diagnosis other than osteoarthrosis (per cent)
Bilateral procedure (per cent)
Average hospital charge (U.S. dollars) Total Inflation-adjusted
Discharged to home (per cent)
Average duration of hospitalization (days)
Died (per cent) During index hospitalization <3 mos. after <1 yr. after
Infection (per cent) <3 mos. after <1 yr. after
Revision (per cent) <3 mos. after <1 yr. after
Complications during index hospitalization (per cent)
Deep venous thrombosis <3 mos. after (per cent)
Urinary tract infection during index hospitalization (per cent)
68.0 67.6 66.5f 69.2 67.7 66.3t
34.3 3.6
40.0
23.9
10.7
i,538 ,449
85
8.96
1.8 2.1 3.2
1.1 1.1
1.8 3.2
12.9
1.1
3.9
26.4 3.0
40.9
15.6
10.8
12,329 10,436
87.4
7.72
0.4 0.9 2.0
0.7 1.2
0.7 1.9
7.9
1.1
4.5
24.lt 3.4
40.9
14.lt 10.4
12,355t 10,365t
91.lt
7.73t
O.lt 0.5t 1.3t
0.3+. 0.6+.
0.5t 1.6
8.8+
1.1
3.7
24.6 2.3
43.1
12.4
9.5
13,294 11,252
86.1
6.99
0.2 0.6 1.7
0.6 0.8
1.7 2.5
7.4
0.8
2.5
27.4 3.2
39.4
15.9
10.5
13,103 11,120
87
7.76
0.3 0.9 2.2
0.6 1.0
0.6 1.8
8.3
1.0
4.2
24.3 3.4
41.5
14.8
10.7
ll,729t 9788t
91.lt
7.85t
0.2 0.6 1.3*
0.4 0.8
0.5* 1.7
9.1
1.2
4.0
*The statistical comparisons were made with either the Kruskal-Wallis one-way on whether the variable of interest represented a continuous quantity (comparison (comparison of the percentage of positive cases across groups).
+P<0.01. +.0.01 < p < 0.05.
analysis of variance or the Pearson chi-square test depending of average values across groups) or a dichotomous variable
7936 patients met all of the entry criteria during 8774 hospitalizations. Unfortunately, the hospital discharge data set used does not indicate the anatomical side to which a particular procedure or diagnostic code refers. The 838 patients who met the inclusion criteria twice during the study period were considered to have had bilateral total hip arthroplasty. No patient met our criteria for elective primary total hip replacement more than twice. The proportion of patients who had bilateral hip replacement did not differ significantly across surgeon or hospital-volume categories (Table II). The same patterns of complications and statistical trends were demonstrated when the analysis was repeated after exclusion of the patients who had bilateral elective total hip arthroplasty.
Low-volume surgeons tended to manage patients with more adverse risk profiles in terms of age, comorbidity, and diagnosis (p < 0.01) (Table II). The average age of the patients who were managed in the lowest-volume hospitals (69.2 years) was approximately three years older than that of the patients who were managed in the highest-volume institutions (66.3 years)
(p < 0.01). However, with the numbers available, there was no significant difference in patient co-morbidity or diagnosis across hospital-volume groups (p > 0.05). The comparison of the rates of complications across provider-volume groups may be biased because no adjustment was made to account for differences in case mix (Table II).
Mortality
Twenty patients (0.3 per cent) died during the initial elective hospitalization. This number was too small to establish adequate levels of significance for comparison of provider-volume groups. Fifty-seven patients (0.7 per cent) died in the first three months after the index procedure. The age of the patient, co-morbidity, and male gender were found to be significantly related to the probability of dying within three months after elective total hip arthroplasty (p < 0.05) (Table III). Patients in the lowest-volume surgeon group had three times the risk of dying within three months after elective total hip replacement than those in the highest-volume surgeon group. With the numbers available, no significant differ-
THE JOURNAL OF BONE AND JOINT SURGERY
RELATIONSHIP BETWEEN THE VOLUME OF TOTAL HIP REPLACEMENTS PERFORMED BY PROVIDERS 489
TABLE III
ADJUSTED RATES OF COMPLICATIONS BY PROVIDER (HOSPITAL AND SURGEON) VOLUME*
Death Infection Revision
<3 Mos. after Index
Hospitalization
<lYr . after Index
Hospitalization
<3 Mos. after Index
Hospitalization
<lYr . after Index
Hospitalization
<3 Mos. after Index
Hospitalization
<1 Yr. after Index
Hospitalization
Complications during Index
Hospitalization
Surgeon volume Low vs. high Medium vs. high
Hospital volume Low vs. high Medium vs. high
Age of patient per 10 yrs.
Co-morbidity 1 vs.0 2vs.O
>3 vs. 0
Non-osteoarthrosis vs. osteoarthrosis
Male vs. female
3.0 (1.4 to 7.3) 1.0 (0.6 to 1.9)
0.9 (0.2 to 3.9) 1.8 (1.0 to 3.3)
1.9 (1.3 to 2.6)
1.6 (0.8 to 3.0) 2.4 (1.0 to 5.8) 3.1 (1.1 to 8.9)
1.7 (0.9 to 3.3)
1.9 (0.9 to 3.8) 1.4 (1.0 to 2.0)
1.1 (0.5 to 2.9) 1.7 (1.1 to 2.4) 1.7 (1.4 to 2.2)
1.9 (1.2 to 2.8) 3.3 (1.8 to 5.9) 4.1 (2.2 to 7.6)
2.0 (1.3 to 3.0)
4.3 (1.5 to 12.2) 1.8 (0.9 to 3.6)
1.0 (0.2 to 4.2) 1.2 (0.7 to 2.4)
1.1 (0.8 to 1.4)
1.9 (1.0 to 3.9) 2.3 (0.8 to 6.6) 3.1 (0.9 to 10.2)
1.0 (0.4 to 2.3)
3.2 (1.3 to 7.7) 1.6 (0.9 to 2.8)
0.9 (0.2 to 4.0)
1.3 (0.8 to 2.1)
1.0 (0.8 to 1.2)
1.5 (0.8 to 2.8) 3.0 (1.3 to 6.6) 2.9 (1.1 to 7.6)
1.2 (0.6 to 2.3)
2.9 (1.2 to 6.8) 1.0 (0.5 to 1.9)
1.8 (0.6 to 5.1)
1.1 (0.6 to 2.1)
1.3 (0.9 to 1.7)
0.7 (0.3 to 1.6) 2.0 (0.8 to 4.8) 2.2 (0.7 to 6.5) 1.5 (0.7 to 3.2)
2.1 (1.1 to 3.9) 1.0 (0.7 to 1.4)
0.9 (0.4 to 2.1) 1.0(0.7 to 1.4)
1.1 (0.9 to 1.2)
0.6 (0.3 to 1.1) 1.3 (0.7 to 2.4) 1.8 (0.9 to 3.8) 1.1 (0.7 to 1.9)
2.5 (1.4 to 4.6) 1.4 (1.0 to 2.0) 1.4 (0.7 to 2.7) 1.1 (0.7 to 1.6) 0.8 (0.5 to 1.5) 1.3(0.910 1.
1.6 (1.1 to 2.3) 1.0 (0.8 to 1.3)
0.8 (0.4 to 1.4) 0.9 (0.7 to 1.1) 1.3 (1.2 to 1.4)
1.1 (0.9 to 1.3) 1.3 (1.0 to 1.8) 1.7 (1.2 to 2.5)
1.2 (1.0 to 1.6)
1.6 (1.3 to 1.8)
given as the odds ratios, with the 95 per cent confidence intervals in parentheses. The bold type indicates values that were significant the confidence interval excludes 1.0. Generalized estimating equations for logistic regression were used. Each variable was adjusted
*The values are at p < 0.05 — that is. for all others.
ence in survival could be demonstrated between the lowest-volume hospital group and the highest-volume hospital group (p > 0.05) (Table III).
One hundred and thirty patients (1.6 per cent) died within one year after the index admission. There was a strong relationship between death within one year and the age of the patient, co-morbidity, and diagnosis, even after adjustment for provider volume and gender (Table III). There was a trend toward higher one-year survival rates with higher surgeon volume, although significance was not reached after adjustment for hospital volume, age of the patient, co-morbidity, diagnosis, and gender. Trends for hospital volume were similar to those in the three-month analysis, with the difference in survival between the middle and highest-volume groups achieving significance (p < 0.05).
Infection
Thirty-nine patients (0.5 per cent) were readmitted for infection during the first three months after the index operation. Surgeon volume was significantly related to the development of infection within three months (p < 0.05). Patients who were managed by a surgeon who averaged fewer than two hip replacements annually were more than four times more likely to be readmitted for infection about the hip after elective total hip arthroplasty than those managed by a surgeon who averaged more than ten hip replacements annually (Table III).
Sixty-seven patients (0.8 per cent) were readmitted at least once for infection during the first year after the index admission. Infection was more than three times more likely to develop within one year after the operation in patients who were managed by a low-volume surgeon than in those managed by a high-volume sur
geon. With the numbers available, hospital volume was not significantly related to the development of infection (p > 0.05).
Revision
Fifty patients (0.6 per cent) were admitted for a revision procedure within three months after the index operation. Patients who were managed by a low-volume surgeon were nearly three times more likely to have a revision within three months than those managed by a high-volume surgeon (Table III). With the numbers available, none of the other variables achieved significance at the p < 0.05 level. One hundred and forty-one patients (1.8 per cent) were admitted for a revision within one year after the index operation. Patients who were managed by a low-volume surgeon were approximately twice as likely to be admitted for revision within one year than those managed by a high-volume surgeon (Table III). With the numbers available, hospital volume, age of the patient, co-morbidity, gender, and diagnosis were not significantly associated with revision at one year (p > 0.05).
Complications during the Index Hospitalization
Six hundred and ninety (8.7 per cent) of the patients sustained a serious complication during the initial index procedure. The age of the patient, co-morbidity, male gender, and a diagnosis other than osteoarthrosis were significantly related to such events (p < 0.05) (Table III). After adjustment for hospital volume, age of the patient, gender, co-morbidity, and diagnosis, the risk of a complication was nearly twice as high for patients who were managed by a low-volume surgeon than for those managed by a high-volume surgeon (Table III).
VOL. 79-A, NO. 4, APRIL 1997
490 H. J. KREDER ET AL.
TABLE IV ADJUSTED COMPARISON OF DURATION OF
HOSPITALIZATION AND HOSPITAL CHARGES*
Surgeon volume Low vs. high Medium vs. high
Hospital volume Low vs. high Medium vs. high
Age of patient per 10 yrs.
Co-morbidity 1 vs. 0 2vs.O
>3 vs. 0
Non-osteoarthrosis vs. osteoarthrosis
Male vs. female
Year of admission 1988 vs. 1991 1989 vs. 1991 1990 vs. 1991
Destination after discharge!
Home vs. transfer
Home-care vs. transfer
Duration of stay per day
Difference in Duration of
Hospitalization (Days)
0.8 (0.4 to 1.3) -0.1 (-0.3 to 0.2)
-1.0 (-1.4 to -0.6) -0.2 (-0.5 to 0.1)
0.4 (0.3 to 0.5)
0.3 (0.1 to 0.4) 0.6 (0.3 to 0.9) 0.7 (0.2 to 1.1)
0.3 (0.1 to 0.5)
-0.5 (-0.6 to -0.4)
1.4 (1.1 to 1.6) 0.9 (0.7 to 1.1) 0.5 (0.3 to 0.7)
-0.7 (-1.0 to -0.4)
-0.1 (-0.5 to 0.2)
Variable not entered
Difference in Hospital Charges
(Dollars)
126 (-858 to 1110) -303 (-992 to 386)
2630 (1758 to 3501) 1838 (1129 to 2546)
-415 (-537 to -294)
334 (47 to 621) 1082 (580 to 1583) 1256 (678 to 1835)
15 (-255 to 286)
508 (303 to 712)
1787 (1490 to 2084) 3355 (2943 to 3767) 4868 (4421 to 5315)
Variable not entered
Variable not entered
744 (576 to 912)
T h e values are given as the beta coefficients, with the 95 per cent confidence intervals in parentheses. The bold type indicates values that were significant at p < 0.05 — that is, the confidence interval excludes 0.0. Generalized estimating equations for multiple regression techniques were used. Each variable was adjusted for all others.
tHome = the patient was discharged home without home-care support, home-care = the patient was discharged home with home-care support, and transfer = the patient was discharged to a skilled nursing facility or another institution after the hip replacement.
Duration of Hospitalization
Surgeon and hospital volume as well as age of the patient, co-morbidity, gender, and diagnosis were significantly related to the duration of hospitalization (p < 0.05). The patients who were managed by a low-volume surgeon stayed in the hospital an average of approximately 0.8 day longer than those managed by a high-volume surgeon (Table IV). The duration of the stay in the low-volume hospitals was an average of one day shorter than that in the high-volume hospitals (Table IV). A major trend was also noted for a shorter duration of hospitalization with a more recent year of operation. For example, the average duration of hospitalization was 1.4 days longer in 1988 than in 1991 (p < 0.05).
Hospital Charges
The age of the patient, co-morbidity, gender, year of admission, and duration of hospitalization were significantly related to the hospital charges (p < 0.05). Af
ter adjustment for all of these factors, hospital volume remained significantly related to cost, with the low-volume hospitals charging approximately 2630 dollars more than the high-volume hospitals (p < 0.05) (Table IV). A breakdown of the charge data revealed that the low-volume hospitals charged more for the prosthetic implant and for the operation as a whole. This increased fee was only partially offset by lower charges for supplies, accommodations, drugs, and laboratory tests. Surgeon volume was not related to hospital charges. Hospital charges were significantly lower for older patients after adjustment for the year of admission, duration of hospitalization, gender, co-morbidity, and provider volume (p < 0.05) (Table IV). It is possible that this reflects the use of less expensive implants in older patients.
As more recent admissions would be expected to be more expensive because of inflation, the entire analysis was repeated with adjustment of all charges to 1988 dollars with use of yearly inflation data specific to hospital-service inflation28. We found no difference in the levels of significance or statistical trends between the analysis with adjusted hospital charges compared with that with unadjusted hospital charges.
Discussion
As a group, patients managed by a surgeon who averages fewer than two hip replacements a year fared worse with regard to most of the outcomes evaluated, even after adjustment for hospital volume, age of the patient, co-morbidity, gender, and operative diagnosis. Hospital volume was significantly related only to the hospital charges and the duration of hospitalization (p < 0.05). Low-volume hospitals charged more for an elective hip replacement, even after adjustment for patient-related factors, provider volume, year of the operation, and duration of hospitalization. The duration of the stay in the low-volume hospitals tended to be shorter. Thus, outcomes seem closely linked to surgeon volume (and, therefore, experience), whereas the use of resources is more closely related to hospital volume. These differences serve to refine our understanding of the relationship between provider volume and outcome. The age of the patient was negatively associated with hospital charges after adjustment for co-morbidity, gender, diagnosis, year of admission, provider volume, and duration of hospitalization. One might speculate that the lower hospital charges for older patients (after adjustment for co-morbidity, duration of hospitalization, and so on) represents the use of cheaper prosthetic implants inserted with cement in this population. The higher cost for men who have elective hip arthroplasty is difficult to explain.
Eftekhar suggested that the outcome of total hip arthroplasty may be influenced by the experience of the surgeon6. That author pointed out that series of patients managed by a single experienced surgeon re-
THE JOURNAL OF BONE AND JOINT SURGERY
RELATIONSHIP BETWEEN THE VOLUME OF TOTAL HIP REPLACEMENTS PERFORMED BY PROVIDERS 491
portedly had better outcomes with respect to rates of mechanical failure and infection than series managed by multiple surgeons with various degrees of experience. While it may be that experienced surgeons provide superior care, an equally possible explanation for this observed difference in outcome is that institutions at which the single-surgeon series were managed had a better case-mix risk profile. This might be due to the restriction of operative candidates or, conversely, low-risk patients may preferentially select a certain type of provider17. Irrespective of the reason for such potential differences in case mix, comparisons across providers may be biased unless patient-related factors are considered813. Johnsson et al. initially found the lifetime experience of the surgeon to be significantly related to the risk of revision after hip replacement; however, this effect disappeared after adjustment for various patient-related factors13. Greenfield et al. compared outcomes within twelve months after total hip arthroplasty across four hospitals involved in a resource utilization study and found, with use of univariate analysis, that function differed significantly across the different hospitals (p < 0.05)8. However, when those authors adjusted for socio-demographics and co-morbidity, the difference in function was no longer significant (p > 0.05). Callaghan et al. described their so-called learning curve with the porous-coated anatomic (PCA) total hip system (Howmedica, Rutherford, New Jersey) in a series of 100 consecutive patients managed at one institution2. The technical results in the fifty patients managed in the second half of the study were noted to be better than those in the fifty patients managed in the first half. However, it was not possible to demonstrate a significant difference in clinical ratings or pain in the thigh at the two-year follow-up evaluation for these two groups of patients.
The implication of the learning-curve concept is that excellence ultimately will be achieved as sufficient experience is gained over a finite period of time20. Thus, to minimize adverse outcomes, one might propose that surgeons who are learning a new technique be supervised by more experienced surgeons until they have demonstrated sufficient facility with the procedure. However, one must also consider the number of operations performed on an ongoing basis. A favorable learning-curve model may not be applicable to a situation in which a given procedure is performed very infrequently over many years. The benefit of previous encounters may be lost because of the large time-interval between operations, so that an appropriate level of skill is never achieved. Furthermore, optimum hospital support staff and services may not be available for an infrequent procedure or, if they are, the staff may be unfamiliar with or inefficient in handling a rare situation. Complications may therefore be minimized by assigning such procedures to regional centers. A consistent direct relationship between institutional volume and patient survival after open heart operations led to
some of the earliest recommendations for minimum-volume standards16181923. Showstack et al. stated that, although an occasional low-volume institution might have low mortality rates (or an occasional high-volume institution might have high mortality rates), average outcomes would be expected to improve if bypass grafting of the coronary artery were performed in higher-volume settings26. Regionalization of other operative procedures, including total hip replacement, has been recommended to minimize adverse outcomes1619. Some authors have found that regionalization of trauma care and the institution of a trauma system reduced adverse outcomes for multiply injured patients"'25-29. Other authors have been unable to demonstrate a benefit after the implementation of state programs recommending minimum patient volumes for providers of certain elective procedures27.
The use of administrative data sets for the determination of outcome requires caution. Clinically relevant information is often not collected, and the coding of the information may be inaccurate11. Several limitations of the hospital discharge data set were encountered in the present investigation. The anatomical side of a procedure or diagnosis was not documented. However, as there was virtually no difference in the distribution of patients who had a bilateral procedure across provider-volume groups, it is unlikely that our conclusions were seriously biased by this limitation. Furthermore, repeating the analysis without the patients who had a bilateral procedure yielded almost the same results. There is good evidence to support marked differences in survival of the implant on the basis of its design and the method of fixation1013. Nonetheless, it is the provider who determines the type of prosthesis and operative approach as well as other details. Thus, while the design of the prosthesis and other information would have allowed a more detailed exploration of the reasons for various types of complications, the basic conclusions regarding variation in rates of complications across providers were not affected by the absence of these variables. While a mounting body of evidence now demonstrates an association between outcome and provider volume18, more detailed data are needed before firm policy recommendations can be made regarding whether certain patients should be referred to more specialized centers for management.
Appendix I: Study Inclusion-Exclusion Algorithm
Inclusion Criteria (for Selection into Inception Cohort)
Diagnostic 1CD-9 Codes At least one of the following was required for inclu
sion. 1.710xy, where x = 0,1,2,3,4,5,8, or 9 and y = 0,5,
or 9 (connective-tissue disorders). 2. 712xy, where x = 1, 2, 3, 8, or 9 and y = 0, 5, or 9
(crystal arthropathies).
VOL. 79-A, NO. 4, APRIL 1997
492 H. J. KREDER ET AL.
3.713x, where x = 0,1,2,3,4,5,6,7, or 8 (metabolic and other arthropathies).
4. 714x, where x = 0,1, 2, 3, 4, 8, or 9 (rheumatoid arthritis).
5. 715xy, where x = 0,1,2, 3, 8, or 9 and y = 0,5, or 9 (osteoarthrosis).
6. 716xy, where x = 2, 3, 4,5, 6, 8, or 9 and y = 0,5, or 9 (other arthropathies).
7.718xy, where x = 0,1, or 5 and y = 0,5, or 9 (other derangement of joint).
8. 719xy, where x = 2 or 3 and y = 0, 5, or 9 (other disorders of joint).
9. 720x, where x = 0,1, 2, 8, or 9 (ankylosing spondylitis).
10. 725 (polymyalgia). 11. 726x, where x = 5, 8, or 9 (enthesopathies). 12. 731x, where x = 0 or 2 (osteitis deformans). 13. 2740 (gouty arthropathy). 14. 732x, where x = 1, 2,4, 6, or 7 (osteochondropa
thies). 15. 7334y, where y = 0, 2, or 9 (aseptic necrosis of
bone). 16. 7338y, where y = 1 or 2 (malunion). 17. 7363y, where y = 0,1,2, or 9 (acquired deformity
of hip). 18. 905x, where x = 3, 4, 5, or 6 (late effect of mus
culoskeletal and connective-tissue injuries).
Procedure ICD-9 Codes
One of the following was required for inclusion (in addition to one of the diagnostic criteria just listed).
1. Throughout entire study period, 8151. 2. Before November 1,1989, 8159. Note: Both procedure codes were revised in 1989 as
follows. Before November 1, 1989, 8151 = total hip replace
ment with cement and 8159 = other total hip replacement. From November 1, 1989 on, 8151 = total hip re
placement and 8159 = revision of lower-extremity joint replacement.
After a discharge date of November 1,1989, no patient who had a procedure code of 8159 was identified by this algorithm, which was intended to select primary operations as index procedures and to exclude revisions.
CHARS Fields
Type of admission = elective.
Exclusion Criteria
Diagnostic ICD-9 Codes
Any one of the following resulted in exclusion. 1. 800 through 899 (fractures, dislocations, and inju
ries). 2. 996 (mechanical complication of implant). 3.140 through 208 (malignant neoplasms). 4. 235 through 239 (neoplasms of uncertain behav
ior).
5. E800 to E869, E880 to E928, and E950 to E999 (supplemental classification of acute non to medical causes of injury and poisoning).
Procedure ICD-9 Codes
1. 8005 alone (arthrotomy for removal of hip prosthesis).
2. 8153 alone (revision of hip replacement). 3. Any combination of procedure and diagnostic
ICD-9 codes (Appendix II) that met the revision criteria.
CHARS Fields
Either one of the following resulted in exclusion. 1. Source of admission = emergency department. 2. Age of eighteen years or less at time of admission.
Appendix II: Definitions of Variables
Diagnoses
ICD-9 diagnostic codes from time of index admission for elective hip replacement.
Osteoarthrosis
At least one of the following was required. I. 715xy, where x = 0,1, 2, 3, 8, or 9 and y = 0, 5, or
9 (osteoarthrosis). 2.716xy, where x = 5,6,8, or 9 and y = 0,5, or 9 (other
arthropathies). 3.718xy, where x = 0,1, or 5 and y = 0,5, or 9 (other
derangement of joint).
Inflammatory Disease, Avascular Necrosis, and Late Post-Traumatic Osteoarthrosis
These diagnoses were initially considered separately. In the final analysis, they were combined into one category. At least one of the following was required for the combined category.
1. 710xy, where x = 0,1, 2, 3, 4,5, 8, or 9 and y = 0, 5, or 9 (connective-tissue disorders).
2. 712xy, where x = 1,2, 3, 8, or 9 and y = 0, 5, or 9 (crystal arthropathies).
3.713x, where x = 0,1,2,3,4,5,6,7, or 8 (metabolic and other arthropathies).
4. 714x, where x = 0 ,1 , 2, 3, 4, 8, or 9 (rheumatoid arthritis).
5.716xy, where x = 2,3, or 4 and y = 0,5, or 9 (other arthropathies).
6. 719xy, where x = 2 or 3 and y - 0, 5, or 9 (other disorders of joint).
7. 720x, where x = 0,1, 2, 8, or 9 (ankylosing spondylitis).
8. 725 (polymyalgia). 9. 726x, where x = 5, 8, or 9 (enthesopathies).
10. 731x, where x = 0 or 2 (osteitis deformans). I I . 2740 (gouty arthropathy). 12. 732x, where x = 1, 2,4, 6, or 7 (osteochondropa
thies).
THE JOURNAL OF BONE AND JOINT SURGERY
RELATIONSHIP BETWEEN THE VOLUME OF TOTAL HIP REPLACEMENTS PERFORMED BY PROVIDERS 4 9 3
13. 7334y, where y = 0, 2, or 9 (aseptic necrosis of bone).
14. 7338y, where y = 1 or 2 (malunion). 15.7363y, where y = 0,1,2, or 9 (acquired deformity
of hip). 16. 905x, where x = 3, 4, 5, or 6 (late effect of mus
culoskeletal and connective-tissue injuries).
Complications
Revision/Failure
After an index procedure, a subsequent inpatient admission that met the following criteria was considered a revision or failure.
At least one of the following diagnostic codes was required.
1. 730xy, where x = 0,1, 2, 3, 8, or 9 and y = 0, 5, or 9 (osteomyelitis of hip).
2. 8350y, where y = 0,1,2, or 3 (dislocation of hip). 3. 9964 (mechanical complication of internal ortho
paedic device). 4. 99666 (infection and inflammatory reaction due
to internal joint prosthesis). 5. 99677 (other complication due to internal joint
prosthesis). 6. 998x; where x = 5 or 6 (postoperative infections). At least one of the following procedure codes was
required (in addition to one of the diagnostic codes just listed).
1. Throughout entire study period, 8005 (arthrotomy for removal of hip prosthesis).
2. From November 1,1989, on, 8153 (revision of hip replacement).
3. From November 1, 1989, on, 8159 (revision of lower-extremity joint replacement).
4. Before November 1,1989, 816x, where x = 1,2,3, 4, or 9 (partial replacement of femur or hip used to code revisions).
Infection
After an index procedure, a subsequent inpatient admission that met at least one of the following criteria was considered an infection.
1. 730xy, where x = 0,1, 2, 3, 8, or 9 and y = 0, 5, or 9 (osteomyelitis of hip).
2. 99666 (infection and inflammatory reaction due to internal joint prosthesis).
3. 998x, where x = 5 or 6 (postoperative infections).
Complications during Index Admission (Mishaps during Operative and Medical Care)
At least one of the following ICD-9 codes during the index admission was considered a complication.
1. 997x, where x = 0,1,3,4, or 5 (complications of a medical or operative procedure affecting specified parts of body).
2. 998x, where x = 0,1, 2, 3, 4, or 7 (other complications of a medical or operative procedure).
3. 999x, where x = 0, 1, 2, 3, 4, 5, 6, 7, 8, or 9 (complications of medical care, not classified elsewhere).
*4. E87xy, where x = 0,1, or 2 and y = 0,1, 3, 5, 7, 8, or 9.
*5. E873y, where y = 0,1,2, 3, 5, 8, or 9. *6. E874y, where y = 0,1,4, 8, or 9. *7. E875y, where y = 0,1, 2, 8, or 9. *8. E876y, where y = 0,1, 2, 3,4, 5, 8, or 9. *9. E878y, where y = 0,1,2,3, 4, 5, 6, 8, or 9.
*10. E879y, where y = 0,1, 2,3,4,5, 6, 7, 8, or 9. * = Mishap during operative or medical care.
Miscellaneous
Deep Venous Thrombosis
1. 9972 (peripheral vascular complications of a procedure).
2. 451x, where x - 1, 2, 8, or 9 (specific site of deep venous thrombosis).
Urinary Tract Infection
At least one of the following ICD-9 codes during the index admission was considered a complicating urinary tract infection.
1. 5990 (urinary tract infection, site not specified). 2. 5978 (urethritis). 3. 590x, where x = 1, 2, 3, 8, or 9 (kidney infection). 4. 595x, where x = 0,3, or 9 (acute cystitis).
References Cales, R. H., and Trunkey, D. D.: Preventable trauma deaths. A review of trauma care systems development. / Am. Med. Assn., 254: 1059-1063,1.985. Callaghan, J. J.; Heekin, R. D.; Savory, C. G.; Dysart, S. H.; and Hopkinson, W. J.: Evaluation of the learning curve associated with uncemented primary porous-coated anatomic total hip arthroplasty. Clin. Orthop., 282:132-144,1992. Charlson, M. E.; Pompei, P.; Ales, K. L.; and Mackenzie, C. R.: A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J. Chronic Dis., 40: 373-383,1987. Cullen, D. J.; Apolone, G.; Greenfield, S.; Guadagnoli, E.; and Cleary, P.: ASA physical status and age predict morbidity after three surgical procedures. Ann. Surg., 220: 3-9,1994. Deyo, R. A.; Cherkin, D. C; and Ciol, M. A.: Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J. Clin. Epidemiol., 45:613-619,1992. Eftekhar, N. S.: Long-term results of cemented total hip arthroplasty. Clin. Orthop., 225: 207-217,1987. Fowles, J.; Bunker, J. P.; and Schurman, D. J.: Hip surgery data yield quality indicators. Business and Health, 4(8): 44-46,1987. Greenfield, S.; Apolone, G.; McNeil, B. J.; and Cleary, P. D.: The importance of co-existent disease in the occurrence of postoperative complications and one-year recovery in patients undergoing total hip replacement. Comorbidity and outcomes after hip replacement. Med. Care, 31:141-154,1993.
VOL. 79-A, NO. 4, APRIL 1997
494 H. J. KREDER ET AL.
9. Guss, D. A.; Meyer, F. T.; Neuman, T. S.; Baxt, W. G.; Dunford, J. V., Jr.; Griffith, L. D.; and Guber, S. L.: The impact of a regionalized trauma system on trauma care in San Diego County. Ann. Emerg. Med., 18:1141-1145,1989.
10. Havclin, L. I.; Espehaug, B.; Vollset, S. E.; and Engesaeter, L. B.: Early failures among 14,009 cemented and 1,326 uncemented prostheses for primary coxarthrosis. The Norwegian Arthroplasty Register, 1987-1992. Acta Orthop. Scandinavica, 65:1-6,1994.
11. Hsia, D. C; Ahern, C. A.; Ritchie, B. P.; Moscoe, L. M.; and Krushat, W. M.: Medicare reimbursement accuracy under the prospective payment system, 1985 to 1988.7. Am. Med. Assn., 268: 896-899,1992.
12. Hughes, R. G.; Garnick, D. W.; Luff, H. S.; McPhee, S. J.; and Hunt, S. S.: Hospital volume and patient outcomes. The case of hip fracture patients. Med. Care, 26:1057-1067,1988.
13. Johnsson, R.; Franzen, H.; and Nilsson, L. T.: Combined survivorship and multivariate analyses of revisions in 799 hip prostheses. A 10- to 20-year review of mechanical loosening. J. Bone and Joint Surg., 76-B(3): 439-443,1994.
14. Kelly, J. V., and Hellinger, F. J.: Physician and hospital factors associated with mortality of surgical patients. Med. Care, 24:785-800,1986. 15. Lavernia, C. J., and Guzman, J. F.: Relationship of surgical volume to short-term mortality, morbidity, and hospital charges in arthro
plasty. J. Arthroplasty, 10:133-140,1995. 16. Luft, H. S.; Bunker, J. P.; and Enthoven, A. C: Should operations be regionalized? The empirical relationship between surgical volume
and mortality. New England J. Med., 301: 1364-1369,1979. 17. Luft, H. S.; Hunt, S. S.; and Maerki, S. C: The volume-outcome relationship: practice-makes-perfect or selective-referral patterns?
Health Services Res., 22:157-182,1987. 18. Luft, H. S.; Garnick, D. W.; Mark, D. H.; and McPhee, S. J.: Hospital Volume, Physician Volume, and Patient Outcomes: Assessing the
Evidence, pp. 1-406. Ann Arbor, Michigan, Health Administration Press Perspectives, 1990. 19. Maerki, S. C; Luft, H. S.; and Hunt, S. S.: Selecting categories of patients for regionalization. Implications of the relationship between
volume and outcome. Med. Care, 24:148-158,1986. 20. Manu, P.; Lane, T. J.; and Matthews, D. A.: How much practice makes perfect? A quantitative measure of the experience needed to
achieve procedural competence. Med. Teacher, 12: 367-369,1990. 21. Moskal, J. T.; Shaffrey, C. I.; and Ripley, L. P.: Prospective analysis of uncemented and hybrid primary porous coated anatomic total hip
arthroplasties in a community setting. Clin. Orthop., 304:139-147,1994. 22. Praemer, A.; Furner, S.; and Rice, D. P.: Medical implants and major joint procedures. In Musculoskeletal Conditions in the United States,
p. 133. Park Ridge, Illinois, The American Academy of Orthopaedic Surgeons, 1992, 23. Riley, G., and Lubitz, J.: Outcomes of surgery among the medically aged: surgical volume and mortality. Health Care Financ. Rev., 7:
37-47,1985. 24. Seagroatt, V., and Goldacre, M.: Measures of early postoperative mortality: beyond hospital fatality rates. BMJ, 309: 361 -365,1994. 25. Shackford, S. R.; Hollingworth-Fridlund, P.; Cooper, G. F.; and Eastman, A. B.: The effect of regionalization upon the quality of trauma
care as assessed by concurrent audit before and after institution of a trauma system: a preliminary report . / Trauma, 26:812-820,1986. 26. Showstack, J. A.; Rosenfeld, K. E.; Garnick, D. W.; Luft, H. S.; Schaffarzick, R. W.; and Fowles, J.: Association of volume with outcome
of coronary artery bypass graft surgery. Scheduled vs nonscheduled operations. J.Am. Med. Assn., 257: 785-789,1987. 27. Sloan, F. A.; Perrin, J. M.; and Valvona, J.: In-hospital mortality of surgical patients: is there an empiric basis for standard setting?
Surgery, 99:446-454,1986. 28. United States Department of Labor, Bureau of Labor Statistics: Monthly reports. CPI Detailed Rep., 1988-1992. 29. West, J. G.; Cales, R. H.; and Gazzaniga, A. B.: Impact of regionalization. The Orange County experience. Arch. Surg., 118: 740-
744, 1983. 30. Whittle, J.; Steinberg, E. P.; Anderson, G. F.; Herbert, R.; and Hochberg, M. C: Mortality after elective total hip arthroplasty in elderly
Americans. Age, gender, and indication for surgery predict survival. Clin. Orthop., 295:119-126,1993. 31. Wroblewski, B. M.: 15-21-year results of the Charnley low-friction arthroplasty. Clin. Orthop., 211: 30-35,1986. 32. Zeger, S. L.; Liang, K. Y.; and Albert, P. S.: Models for longitudinal data: a generalized estimating equation approach. Biometrics, 44:
1049-1060,1988.
THE JOURNAL OF BONE AND JOINT SURGERY