The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States government.
Hospital volume and surgical outcomes after elective hip/knee arthroplasty: A risk-adjusted analysis of a large regional database†
Version of Record online: 1 AUG 2011
Copyright © 2011 by the American College of Rheumatology
Arthritis & Rheumatism
Volume 63, Issue 8, pages 2531–2539, August 2011
How to Cite
Singh, J. A., Kwoh, C. K., Boudreau, R. M., Lee, G.-C. and Ibrahim, S. A. (2011), Hospital volume and surgical outcomes after elective hip/knee arthroplasty: A risk-adjusted analysis of a large regional database. Arthritis & Rheumatism, 63: 2531–2539. doi: 10.1002/art.30390
- Issue online: 1 AUG 2011
- Version of Record online: 1 AUG 2011
- Accepted manuscript online: 7 JUN 2011 10:59AM EST
- Manuscript Accepted: 31 MAR 2011
- Manuscript Received: 5 OCT 2010
- Western Pennsylvania Chapter of the Arthritis Foundation
- NIH (Clinical Translational Science Award to the Mayo Clinic Center for Clinical and Translational Research). Grant Number: 1-KL2-RR-024151-01
- NIH (National Institute of Arthritis and Musculoskeletal and Skin Diseases). Grant Number: K24-AR-055259
To examine the relationship between hospital procedure volume and surgical outcomes following elective primary total hip arthroplasty/total knee arthroplasty (THA/TKA).
Using the Pennsylvania Health Care Cost Containment Council database, we identified all patients who underwent elective primary THA/TKA in Pennsylvania. Hospitals were categorized according to the annual volume of THA/TKA procedures, as follows: ≤25, 26–100, 101–200, and >200. The 30-day complication rate and 30-day and 1-year mortality rates were assessed by logistic regression models, adjusted for age, sex, race, insurance type, hospital region, 3M All Patient Refined Diagnosis Related Group risk of mortality score, hospital teaching status, and bed count.
In the THA and TKA cohorts, the mean age of the patients was 69 years, and 42.8% and 35%, respectively, were men. Compared with patients whose surgeries were performed at very-high-volume hospitals (>200 procedures/year), patients who underwent elective primary THA procedures at hospitals with a very low volume (≤25 procedures/year), a low volume (26–100 procedures/year), or a high volume (101–200 procedures/year) had higher multivariable-adjusted odds ratios (ORs) for venous thromboembolism (OR 2.0, 95% confidence interval [95% CI] 0.2–16.0), OR 3.4 [95% CI 1.4–8.0], and OR 1.1 [95% CI 0.3–3.7], respectively) and 1-year mortality (OR 2.1 [95% CI 1.2–3.6], OR 2.0 [95% CI 1.4–2.9], and OR 1.0 [95% CI 0.7–1.5], respectively). Among patients ages ≥65 years who underwent elective primary TKA at very-low-volume, low-volume, and high-volume hospitals, the ORs for 1-year mortality were significantly higher (OR 0.6 [95% CI 0.2–2.1], OR 1.6 [95% CI 1.0–2.4], and OR 0.9 [95% CI 0.6–1.3], respectively), compared with very-high-volume hospitals.
Performance of elective primary THA and TKA surgeries in low-volume hospitals was associated with a higher risk of venous thromboembolism and mortality. Confounding due to unmeasured variables is possible. Modifiable system-based factors/processes should be targeted to reduce the number of complications associated with THA/TKA procedures.
Elective total hip arthroplasty (THA) and total knee arthroplasty (TKA) are highly successful surgical treatment options for patients with treatment-refractory, end-stage hip or knee arthritis. Both procedures are associated with significant improvement in pain, function, and health-related quality of life (1, 2). Perioperative and postoperative medical complications, including cardiac and thromboembolic events, can lead to significant morbidity and mortality after THA/TKA. Furthermore, implant-related complications, including infection, loosening, and periprosthetic fractures, can lead to early implant failure, necessitating revision surgery (3). Thus, these complications lead to higher patient morbidity, which can lead to higher health care utilization and costs (4).
There is growing evidence linking surgical outcomes with the surgery volume for a variety of procedures (5–7). With respect to hip and knee replacements, studies have demonstrated an association between hospital volume and a decreased risk of some complications but not others. Katz et al studied 90-day complication rates in 80,904 Medicare patients who underwent primary TKA, adjusting the analyses for age, sex, Medicaid eligibility, comorbidities, and underlying diagnosis (8), and reported that the risk of pneumonia was significantly higher among patients for whom joint replacements were performed in “low volume” hospitals (8). However, no significant differences were noted in the rate of mortality, acute myocardial infarction, or pulmonary embolism (8). Similarly, in another study of 76,627 Medicare patients who underwent primary or revision THA, there was a significant association between higher hospital volume and lower 90-day mortality (9).
In contrast, a population-based study of 14,352 patients who underwent TKA in Canada between 1993 and 1996 showed no association between low hospital volume and the rates of in-hospital major complications, 90-day mortality, or knee infection at 1 year or 3 years (10). Similarly, Kreder et al observed no association between hospital volume and the rate of complications or mortality following hip arthroplasty (11). Consequently, it is unclear whether these conflicting findings regarding the relationship between hospital volume and the rate of postoperative complications are attributable to differences in the study setting (US versus Canada), cohort characteristics (Medicare patients ages ≥65 years versus population-based), or the volume thresholds considered. Furthermore, previously reported estimates were not adjusted for the overall risk of surgical mortality, which can lead to residual confounding.
Therefore, the purpose of this study is to examine the relationship between hospital surgical volume and postoperative complications, including 30-day and 90-day mortality, in a group of 29,000 patients undergoing elective THA/TKA, using a large regional database, adjusting for the overall risk of surgical mortality.
PATIENTS AND METHODS
Study sample and data collection.
We used the Pennsylvania Health Care Cost Containment Council (PHC4) database to identify all elective primary THA and TKA surgeries performed in Pennsylvania during fiscal year 2002. Cases were identified using the International Classification of Diseases, Ninth Revision (ICD-9) codes of 81.54 (THA) and 81.51 (TKA). Patients who had previously undergone hip or knee replacement were excluded from the analysis. The data set includes information on demographics for all patients who underwent TKA or THA at 169 acute-care nongovernmental hospitals in Pennsylvania between July 1, 2001 and June 30, 2002. For THA, patients were excluded if they had a hip fracture (ICD-9 code 820) as the cause of arthroplasty or if they had undergone hemiarthroplasty (ICD-9 code 81.52), a procedure commonly performed in the management of hip fractures. This study was approved by the Institutional Review Board at the Veterans Affairs Pittsburgh Healthcare System.
Main predictor and covariates/confounders.
The primary predictor of interest was hospital volume, defined as the annual number of joint arthroplasties performed in each hospital. The hospital volume categories were as follows: <25 procedures/year, 26–100 procedures/year, 101–200 procedures/year, and >200 procedures/year. The reference group was very-high-volume hospitals (>200 procedures/year). The covariates for the study included sex, race, age, region, hospital teaching status (teaching or nonteaching), and insurance status (categorized as none or unknown, Medicaid, Medicare/government, or private). For surgical risk adjustment, we used the 3M Health Information Systems All Patient Refined Diagnosis Related Group (APR-DRG) risk of mortality score. This risk-adjustment tool provides a categorical risk assessment based on interactions of age, type of surgical procedure, comorbidity, and the principal diagnosis and has been previously validated (12–15). The 3M APR-DRG score for the risk of mortality assigns a risk of death to each surgical procedure as minor, moderate, major, or extreme.
The study outcomes of interest were overall mortality at 30 days and at 1 year, and the 30-day complication rate. To assess complications, we used ICD-9 codes to identify 5 major patient-centered complications. These complications comprise the most common major complications after total joint replacement, including acute myocardial infarction (ICD-9 codes 410.00, 410.01, 410.10, 410.11, 410.20, 410.21, 410.30, 410.31, 410.40, 410.41, 410.50, 410.51, 410.60, 410.61, 410.70, 410.71, 410.80, 410.81, 410.90, and 410.91), venous thromboembolism (pulmonary embolism/deep venous thromboses; ICD-9 codes 415.1, 415.11, 415.19, 451.11, 451.19, 451.2, 451.81, 451.9, 453.40, 453.41, 453.42, 453.8, and 453.9), catheter-associated urinary tract infection (ICD-9 codes 996.54, with an additional ICD-9 code of 595.xx or 599.0), prosthetic device malfunction (ICD-9 codes 996.40, 996.41, 996.42, 996.43, 996.46, 996.47, and 996.49), and/or surgical wound infection (ICD-9 codes 682.5, 682.6, 682.8, and 682.9). To assess mortality, the cohort was linked to the National Death Index.
For baseline comparisons, we performed chi-square testing for categorical variables and used the Kruskal-Wallis equality-of-populations rank test for continuous variables. For all analyses, hip replacement cases and knee replacement cases were assessed separately. A total of 10,187 patients underwent hip replacement surgery, and 19,418 patients underwent knee replacement surgery. We excluded 6 patients who underwent hip replacement surgery and 1 patient who underwent knee replacement surgery, for whom the APR-DRG classes for risk of mortality could not be calculated.
For the analysis for 30-day and 1-year mortality, we used logistic regression models, clustered on hospital, to take into account the correlation of outcomes in patients admitted to the same hospital. The fitted models were adjusted for age, sex, race, APR-DRG class for risk of mortality, insurance type, hospital geographic region within Pennsylvania, hospital teaching status, and hospital bed count. We performed logistic regression analyses separately for the overall 30-day complication rate and the rate of each individual complication clustered on hospital. The covariates were the same as those used in the models for mortality analysis. Last, to examine whether volume–outcome associations are specific to older patients, as shown in 2 studies of Medicare data (8, 9), we conducted additional analyses restricted to patients age ≥65 years.
Hospital and study cohort characteristics.
The distribution of hospitals according to region, number of beds, and teaching status for THA and TKA are shown in Tables 1 and 2, respectively. For THA, there were significant differences in hospital volume by region, with a larger proportion of high-volume hospitals in the more urban regions around Philadelphia, Pittsburgh, and northwest Pennsylvania (i.e., Erie). For TKA, this difference did not reach statistical significance.
|Very low||Low||High||Very high|
|Total no. of hospitals||169||69||75||17||8|
|Pittsburgh and surrounding area||33 (19.5)||9 (13.0)||17 (22.7)||5 (29.4)||2 (25.0)|
|Northwest Pennsylvania||24 (14.2)||13 (18.8)||8 (10.7)||3 (17.7)||0 (0)|
|Southern Laurel Highlands||11 (6.5)||6 (8.7)||5 (6.7)||0 (0)||0 (0)|
|North central Pennsylvania||12 (7.1)||8 (11.6)||1 (1.3)||3 (17.7)||0 (0)|
|South central Pennsylvania||16 (9.5)||3 (4.4)||9 (12.0)||2 (11.8)||2 (25.0)|
|Northeast Pennsylvania||14 (8.3)||5 (7.3)||8 (10.7)||1 (5.9)||0 (0)|
|East Pennsylvania||13 (7.7)||5 (7.3)||5 (6.7)||1 (5.9)||2 (25.0)|
|Surrounding Philadelphia||25 (14.8)||8 (11.6)||15 (20)||2 (11.8)||0 (0)|
|Philadelphia||21 (21.4)||12 (17.4)||7 (9.3)||0 (0)||2 (25.0)|
|No. of hospital beds, median (IQR)||204 (111–314)||104 (74–171)||239 (171–314)||366 (319–504)||550 (475–689)||<0.01|
|Teaching hospital||24 (14.2)||3 (4.4)||9 (12.0)||9 (52.9)||3 (37.5)||<0.01|
|Very low||Low||High||Very high|
|Total no. of hospitals||169||69||75||17||8|
|Pittsburgh and surrounding area||33 (19.5)||5 (13.5)||11 (17.2)||8 (19.1)||9 (34.6)|
|Northwest Pennsylvania||24 (14.2)||8 (21.6)||10 (15.6)||4 (9.5)||2 (7.7)|
|Southern Laurel Highlands||10 (5.9)||4 (10.8)||2 (3.1)||4 (9.5)||0 (0)|
|North central Pennsylvania||12 (7.1)||3 (8.1)||5 (7.8)||1 (2.4)||3 (11.5)|
|South central Pennsylvania||17 (10.1)||1 (2.7)||6 (9.4)||6 (14.3)||4 (15.4)|
|Northeast Pennsylvania||14 (8.3)||3 (8.1)||5 (7.8)||6 (14.3)||0 (0)|
|East Pennsylvania||13 (7.7)||3 (8.1)||4 (6.3)||2 (4.8)||4 (15.4)|
|Surrounding Philadelphia||25 (14.8)||3 (8.1)||12 (18.8)||8 (19.1)||2 (7.7)|
|Philadelphia||21 (21.4)||7 (18.9)||9 (14.1)||3 (7.1)||2 (7.7)|
|No. of hospital beds, median (IQR)||204 (111–314)||98 (66–163)||163 (107–239)||256 (222–362)||450 (283–536)||<0.01|
|Teaching hospital||24 (14.2)||3 (8.1)||6 (9.4)||4 (9.5)||11 (42.3)||<0.01|
The mean age of the patients in the THA cohort was 69 years, and 43% were men. All demographic and clinical characteristics differed significantly according to hospital volume. Patients who underwent surgery in very-high-volume hospitals were younger, more likely to be male, less likely to be white, less likely to have government insurance, and had a lower risk of mortality according to the APR-DRG, compared with patients who underwent surgery at all other hospitals (Table 3). The mean age of the patients in the TKA cohort was 69 years, and 35% were men. Similar to patients in the THA cohort, those in the TKA cohort who underwent surgery in very-high-volume hospitals were younger, more likely to be male, less likely to be white, less likely to have government insurance, and had a lower APR-DRG–defined risk of mortality, compared with patients who underwent surgery at all other hospitals (Table 4). Overall, 61% of patients in the primary THA cohort and 64% of patients in the primary TKA cohort were age 65 years or older.
|Characteristic||Overall (n = 10,187)||Hospital volume||P|
|Very low (n = 814)||Low (n = 4,163)||High (n = 2,246)||Very high (n = 2,964)|
|Age, mean (IQR) years||69 (58–76)||72 (62–78)||70 (60–78)||69 (58–76)||65 (54–74)||<0.01|
|Male sex||4,363 (42.8)||304 (37.4)||1,680 (40.4)||995 (44.3)||1,384 (46.7)||<0.01|
|Age ≥65 years||6,256 (61.4)||566 (69.5)||2,757 (66.2)||1,397 (62.2)||153 (51.8)||<0.01|
|White||8,436 (82.8)||716 (88)||3,698 (88.8)||2,107 (93.8)||1,915 (64.6)|
|Black||483 (4.7)||79 (9.7)||194 (4.7)||68 (3)||142 (4.8)|
|Other or unknown||1,268 (12.5)||19 (2.3)||271 (6.5)||71 (3.2)||907 (30.6)|
|Government||6,076 (59.6)||564 (69.3)||2,686 (64.5)||1,353 (60.2)||1,473 (49.7)|
|Medicaid||289 (2.8)||44 (5.4)||141 (3.4)||38 (1.7)||66 (2.2)|
|Private||3,778 (37.1)||195 (24)||1,322 (31.8)||841 (37.4)||1,420 (47.9)|
|None or unknown||44 (0.4)||11 (1.4)||14 (0.3)||14 (0.6)||5 (0.2)|
|APR-DRG risk of mortality†||<0.01|
|Minor likelihood of dying||7,897 (77.5)||579 (71.1)||3,131 (75.2)||1,741 (77.5)||2,446 (82.5)|
|Moderate likelihood of dying||1,718 (16.9)||166 (20.4)||776 (18.6)||368 (16.4)||408 (13.8)|
|Major likelihood of dying||488 (4.8)||59 (7.3)||219 (5.3)||115 (5.1)||95 (3.2)|
|Extreme likelihood of dying||78 (0.8)||9 (1.1)||35 (0.8)||19 (0.9)||15 (0.5)|
|Overall (n = 19,418)||Hospital volume||P|
|Very low (n = 475)||Low (n = 3,681)||High (n = 6,096)||Very high (n = 9,166)|
|Age, mean (IQR) years||69 (60–75)||69 (60–76)||69 (61–76)||69 (61–76)||68 (60–75)||<0.01|
|Male sex||6,797 (35)||165 (34.7)||1,245 (33.8)||2,067 (33.9)||3,320 (36.2)||<0.01|
|Age ≥65 years||12,487 (64.3)||309 (65.1)||2,462 (66.9)||3,966 (65.1)||5,750 (62.7)||<0.01|
|White||16,529 (85.1)||414 (87.2)||3,242 (88.1)||5,494 (90.1)||7,379 (80.5)|
|Black||964 (5)||51 (10.7)||271 (7.4)||271 (4.5)||271 (4.1)|
|Other or unknown||1,925 (9.9)||10 (2.1)||168 (4.6)||331 (5.4)||1,416 (15.5)|
|Government||12,013 (61.9)||328 (69.1)||2,353 (63.9)||3,854 (63.2)||5,478 (59.8)|
|Medicaid||503 (2.6)||27 (5.7)||120 (3.3)||158 (2.6)||198 (2.2)|
|Private||6,840 (35.2)||118 (24.8)||1,192 (32.4)||2,061 (33.8)||3,469 (37.9)|
|None or unknown||62 (0.3)||2 (0.4)||16 (0.4)||23 (0.4)||21 (0.2)|
|APR-DRG risk of mortality†||<0.01|
|Minor likelihood of dying||15,530 (80.0)||368 (77.5)||2,896 (78.7)||4,959 (81.4)||7307 (79.7)|
|Moderate likelihood of dying||3,100 (16.0)||90 (19.0)||639 (17.4)||915 (15.0)||1,456 (15.9)|
|Major likelihood of dying||666 (3.4)||14 (3.0)||124 (3.4)||184 (3.0)||344 (3.8)|
|Extreme likelihood of dying||121 (0.6)||3 (0.6)||22 (0.6)||38 (0.6)||58 (0.6)|
Surgical outcomes in patients undergoing THA.
The 30-day and 1-year mortality rates following primary THA, respectively, were 0.52% (53 of 10,187 patients) and 2.74% (279 of 10,187 patients). Within 30 days, incident venous thromboembolism was observed in 0.42% of patients (43 of 10,187), myocardial infarction was observed in 0.40% (41 of 10,187), and infection was observed in 0.25% (25 of 10,187). Thirty-day mortality did not differ by hospital volume in the entire cohort or in those age ≥65 years. However, there was a statistically significant association between low hospital volume and higher 1-year mortality (Table 5). This finding was also observed when the analyses were restricted to patients undergoing THA who were age ≥65 years. Low hospital volume was also associated with a higher risk of venous thromboembolism in patients undergoing THA (Table 5). However, this association was not observed when the analyses were restricted to patients age ≥65 years. Thirty-day complication rates did not differ by hospital volume.
|Very low volume||Low volume||High volume||Very high volume||P†|
|n/N||OR (95% CI)||n/N||OR (95% CI)||n/N||OR (95% CI)||n/N||OR (95% CI)|
|30-day mortality||6/814||0.9 (0.2–4.2)||29/4,163||1.6 (0.6–4.1)||9/2,246||1.3 (0.4–4.5)||9/2,964||Ref.||0.53|
|1-year mortality||32/814||2.1 (1.2–3.6)||147/4,163||2.0 (1.4–2.9)||50/2,246||1.0 (0.7–1.5)||25/2,964||Ref.||<0.01|
|Overall complications||25/814||1.3 (0.6–2.5)||129/4,163||1.5 (0.9–2.4)||57/2,246||1.3 (0.7–2.3)||67/2,964||Ref.||0.40|
|Venous thromboembolism||3/814||2.0 (0.2–16.0)||24/4,163||3.4 (1.4–8.0)||7/2,246||1.1 (0.3–3.7)||9/2,964||Ref.||0.02|
|Myocardial infarction||3/814||0.3 (0.1–1.5)||16/4,163||0.7 (0.2–1.9)||10/2,246||1.2 (0.3–4.4)||12/2,964||Ref.||0.37|
|Infection||2/814||0.6 (0.1–3.3)||12/4,163||1.1 (0.4–3.4)||4/2,246||0.3 (0.1–1.7)||7/2,964||Ref.||0.44|
|Only patients age ≥65 years|
|30-day mortality||5/566||1.0 (0.2-4.6)||27/2,757||1.9 (0.7–4.9)||8/1,397||1.1 (0.3–4.5)||7/1,536||Ref.||0.32|
|1-year mortality||26/566||2.2 (1.2-4.3)||127/2,757||2.2 (1.4–3.4)||45/1,397||1.0 (0.6–1.5)||42/1,536||Ref.||<0.01|
|Overall complications||16/566||1.0 (0.4-2.3)||94/2,757||1.4 (0.8–2.6)||39/1,397||1.1 (0.6–2.1)||47/1,536||Ref.||0.45|
|Venous thromboembolism||3/566||2.0 (0.2-18.9)||17/2,757||2.2 (0.7–6.8)||2/1,397||0.3 (0.1–1.2)||8/1,536||Ref.||0.06|
|Myocardial infarction||3/566||0.2 (0.04-1.2)||14/2,757||0.5 (0.2–1.4)||10/1,397||1.2 (0.3–4.6)||11/1,536||Ref.||0.16|
|Infection||2/566||0.9 (0.1-6.9)||8/2,757||1.6 (0.2–12.0)||1/1,397||0.2 (0.02–1.8)||4/1,536||Ref.||0.29|
Surgical outcomes in patients undergoing TKA.
Among patients who underwent primary TKA, the 30-day and 1-year mortality rates were 0.27% (52 of 19,418 patients) and 1.27% (246 of 19,418 patients). The incidence of venous thromboembolism was 0.98% (190 of 19,418), the incidence of myocardial infarction was 0.30% (59 of 19,418), and the incidence of infection was 0.33% (64 of 19,418). Thirty-day mortality did not differ significantly by hospital volume across the entire cohort (Table 6). The data suggested that 1-year mortality rates were higher in hospitals performing 26–100 TKA procedures annually, but this trend did not achieve statistical significance after adjusting for multiple comparisons. In patients age ≥65 years, however, performance of TKA in hospitals performing 25–100 TKA procedures annually was associated with a significantly higher risk of 1-year mortality compared with the risk for patients whose procedures were performed in very-high-volume hospitals. There were no significant associations between hospital volume and 30-day complications and 30-day mortality in the overall TKA cohort or in those who were 65 years and older.
|Very low volume||Low volume||High volume||Very high volume||P†|
|n/N||OR (95% CI)||n/N||OR (95% CI)||n/N||OR (95% CI)||n/N||OR (95% CI)|
|30-day mortality||0/475||Not estimable||10/3,681||0.7 (0.3–1.5)||13/6,096||0.5 (0.3–1.0)||29/9,166||Ref.||0.18|
|1-year mortality||5/475||1.0 (0.4–2.6)||64/3,681||1.7 (1.1–2.7)||79/6,096||1.2 (0.8–1.8)||98/9,166||Ref.||0.07|
|Overall complications||12/475||1.6 (0.7–3.6)||65/3,681||1.1 (0.6–1.9)||105/6,096||1.1 (0.8–1.6)||182/9,166||Ref.||0.72|
|Venous thromboembolism||8/475||2.4 (0.9–6.5)||27/3,681||1.0 (0.5–2.3)||63/6,096||1.4 (0.9–2.4)||92/9,166||Ref.||0.21|
|Myocardial Infarction||0/475||Not estimable||13/3,681||1.2 (0.5–3.1)||15/6,096||0.8 (0.4–1.5)||31/9,166||Ref.||0.51|
|Infection||4/475||3.4 (0.7–16.3)||20/3,681||2.2 (0.7–7.0)||13/6,096||0.8 (0.3–2.1)||27/9,166||Ref.||0.17|
|Only patients age ≥65 years|
|30-day mortality||0/309||Not estimable||9/2,462||0.5 (0.2–1.5)||10/3,966||0.5 (0.2–1.0)||24/5,750||Ref.||0.16|
|1-year mortality||3/309||0.6 (0.2–2.1)||58/2,462||1.6 (1.0–2.4)||59/3,966||0.9 (0.6–1.3)||83/5,750||Ref.||0.02|
|Overall complications||4/309||1.0 (0.3–2.9)||48/2,462||1.2 (0.7–2.0)||68/3,966||1.1 (0.7–1.6)||123/5,750||Ref.||0.95|
|Venous thromboembolism||3/309||1.7 (0.5–5.9)||20/2,462||1.1 (0.5–2.6)||39/3,966||1.3 (0.7–2.3)||62/5,750||Ref.||0.78|
|Myocardial Infarction||0/309||Not estimable||11/2,462||0.9 (0.3–2.8)||13/3,966||0.7 (0.3–1.6)||27/5,750||Ref.||0.72|
|Infection||1/309||2.6 (0.3–22.3)||13/2,462||2.7 (0.9–8.0)||8/3,966||1.0 (0.4–3.0)||15/5,750||Ref.||0.23|
Total joint replacements are successful for relieving pain and improving function in patients with end-stage arthrosis of the hip and knee joints (1, 2). Although both THA and TKA have proven long-term clinical success, complications that occur either in the perioperative period (e.g., acute myocardial infarction, venous thromboembolism) or in the postoperative period (e.g., infection, loosening, and or fractures) can cause significant morbidity to the patient and increase healthcare costs. Several recent studies showed an association between the surgical volume of a hospital and the risk of certain postoperative complications. Katz et al (8, 9) reported that lower hospital volume was associated with a significantly higher risk of pneumonia in patients undergoing elective TKA and higher 90-day mortality in Medicare patients undergoing elective THA (8, 9). However, 2 Canadian population-based studies failed to prove a correlation between low surgical volume and increased rates of postoperative complications (10, 11). Studies of the Medicare population provide estimates only in patients age ≥65 years, leading to selection bias, because one-third of all knee and hip arthroplasties in the US are performed in adults younger than age 65 years, as reported in a study of trends in California (16).
Therefore, we examined the relationship between hospital surgical volume and postoperative complications in more than 29,000 patients undergoing elective THA/TKA, using a large, regional database adjusted for overall risk of surgical mortality. Major advances associated with our study compared with previous studies using Medicare and other databases included our ability to adjust for overall surgical mortality risk and the use of a population-based regional database approach that avoided selection bias and allowed inclusion of patients of all age groups undergoing arthroplasty, not just patients age ≥65 years.
In this large study of elective primary THA and TKA, which was not limited to Medicare beneficiaries and was performed in a single fiscal year in Pennsylvania, we observed that lower hospital volume was associated with a higher 30-day risk of venous thromboembolism and 1-year mortality after primary THA. In addition, among the subset of patients age ≥65 years who underwent TKA, lower hospital volume was also associated with a higher risk of 1-year mortality. This result confirms some, but not all, of the findings previously published on this subject. For instance, we showed that low hospital volume is associated with a higher risk of venous thromboembolism following THA. In particular, patients who underwent THA in low-volume hospitals had 2.0–3.4–fold higher odds of developing venous thromboembolism, compared with patients receiving hip replacements at very-high-volume hospitals. This is in contrast to the previous study by Katz et al, which showed no association between hospital volume and venous thromboembolism rates in an analysis limited to the Medicare population (9). Differences in patient population (all patients undergoing TKA versus those age ≥65 years; Pennsylvania versus entire US) and confounders adjusted in analyses may account for the discordant findings. We also were able to adjust for the risk of overall surgical mortality using the APR-DRG score for the risk of mortality, whereas the previous study did not. However, the overall incidence of venous thromboembolism in our cohort (∼1%) is consistent with that previously reported (17, 18).
Among patients who underwent TKA, we did not observe a significantly higher risk of venous thromboembolism in those whose surgeries were performed at low-volume hospitals. This finding is consistent with 3 previous studies by Katz et al (8), Hervey et al (19), and Kreder et al (10), which showed no relationship between hospital volume and venous thromboembolism rates among patients who underwent TKA. There was a suggestion in our study that procedures performed in very-low-volume hospitals (≤25 cases/year) were associated with a higher risk of venous thromboembolism (OR 2.4, P = 0.10). This is consistent with reports of patients who underwent TKA in California (20).
Venous thromboembolism is a preventable complication following elective THA and TKA. There is an intense ongoing debate regarding the best choice of medication/devices for venous thromboembolism prophylaxis in patients undergoing THA and TKA (21, 22). The risk of venous thromboembolism is most likely impacted not only by the choice of thromboprophylactic agent/device but also by the time of initiation and cessation of such therapy (23). Studies are needed to examine whether the type and duration of the thromboprophylactic agent/device being used in the low-volume hospitals are associated with this increased risk of venous thromboembolism. If differences in thromboprophylaxis regimens between high-volume and low-volume hospitals are found, interventions targeting the thromboprophylaxis regimen may be needed to improve venous thromboembolism–related outcomes in patients undergoing THA/TKA at low-volume hospitals.
Our results show that low surgical volume is associated with higher 1-year mortality in patients undergoing elective THA. These results confirm the findings by Katz et al, who reported a similar correlation between hospital volume and 90-day mortality following THA in the Medicare population (9). However, we observed no association between hospital volume and 30-day mortality after THA. Previous studies have shown that lower hospital volume was associated with higher in-patient mortality following TKA (primary and revision knee arthroplasty combined) in the US National Inpatient Sample (19) and higher 90-day mortality in patients who underwent elective primary TKA in California (20). Our results also show higher 1-year mortality among patients age ≥65 years who underwent TKA surgery at low-volume hospitals. The reasons for this discrepancy remain unclear, but further studies are needed to understand the causes of postoperative mortality and to determine what proportion of mortality following these THA/TKA surgeries is related to the procedure versus management of preexisting medical comorbidities.
The present findings should be interpreted with consideration of the following important limitations. We used a large administrative database, which has potential inconsistencies in documentation and no information on key variables, such as body mass index and patient-reported outcomes, including preoperative and postoperative pain, functional status, quality of life, and satisfaction. Therefore, our ability to assess these outcomes is limited. Leading organizations involved in improving the quality of healthcare, such as the US Agency for Healthcare Research and Quality, support the use of administrative databases, such as the PHC4 data set, to evaluate patient outcomes and address important questions (17, 24–26). Because our sample comprised only patients who underwent surgery in Pennsylvania and for whom data from the PHC4 database were available, we could have missed arthroscopy complications occurring at hospitals outside Pennsylvania for some patients. Due to regional variation in rates of joint arthroplasty across the US (27), these findings may not be generalizable to other regions. Our database lacked information regarding the utilization of specific types/brands of joint prostheses, limiting us from comparing different types of prostheses.
Our study used data from 2002, because this was available to us for analyses, and we wanted to have data to assess 4.5/5-year revision rates. Although it is possible that the volume–complication relationship may have varied over time, it is unlikely given that there have been no major technologic advances in TKA or THA that are expected to impact the volume–outcome relationship. Studies examining these associations longitudinally are required to investigate the time period effect on the volume–complication association after THA/TKA. Residual confounding due to unmeasured variables is possible, because of the lack of availability of all potential confounding factors. Despite the large number of patients studied, the number of events was low for several outcomes, making our results susceptible to Type II errors (i.e, missing significant outcomes when they actually existed, due to a small number of events).
In conclusion, in this large group of elective primary THA and TKA surgeries performed in Pennsylvania during one fiscal year, we observed that arthroplasties performed at low-volume hospitals (<200 procedures/year) were associated with a significantly higher adjusted risk of pulmonary embolism within 30 days and 1-year mortality in patients who underwent primary THA and a higher risk of pulmonary embolism and 1-year mortality in patients who underwent TKA. Future studies should focus on investigating whether the underlying reasons for suboptimal outcomes at low-volume hospitals are modifiable (i.e., system factors, perioperative and postoperative care algorithms). Interventions targeted at modifiable predictors of poor outcomes are likely to improve postarthroplasty outcomes in low-volume hospitals.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Ibrahim had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Kwoh, Ibrahim.
Acquisition of data. Ibrahim.
Analysis and interpretation of data. Singh, Kwoh, Boudreau, Lee, Ibrahim.
- 12Testing of 3M's APR-DRG risk adjustment for hospital mortality outcomes. Abstr Acad Health Serv Res Health Policy Meet 2002; 19: 11., , .