- Email: [email protected]
Impact of hospital and surgeon volumes on outcomes following pelvic reconstructive surgery in the United States
American Journal of Obstetrics and Gynecology (2006) 195, 1778–83
www.ajog.org
Impact of hospital and surgeon volumes on outcomes following pelvic reconstructive surgery in the United States Vivian W. Sung, MD, MPH,a Michelle L. Rogers, PhD,b Deborah L. Myers, MD,a Melissa A. Clark, PhDb Division of Urogynecology and Pelvic Reconstructive Surgery, Department of Obstetrics and Gynecology,a Brown Medical School; Center for Gerontology and Health Care Research, Department of Community Health,b Brown University, Providence, RI Received for publication January 15, 2006; revised June 23, 2006; accepted July 5, 2006
KEY WORDS Pelvic reconstructive surgery Pelvic floor disorders Hospital volume Surgeon volume Mortality
Objective: The purpose of this study was to estimate the effect of hospital and surgeon volumes on outcomes following urogynecologic surgery. Study design: This was a retrospective cohort study of women who underwent urogynecologic procedures between 1998 and 2003 from the Nationwide Inpatient Sample. Hospitals and surgeons were categorized as low, medium, or high volume based on average number of cases per year. Outcomes included in-hospital mortality, complications, and nonroutine discharges. Multivariable analyses were performed using generalized estimation equations to estimate relative risks. Results: There were 310,759 women and 2986 hospitals. Women who had procedures at lowvolume hospitals were 2.75 (95% CI 2.33-3.16) times more likely to die and 1.63 (95% CI 1.44-1.83) times more likely to have a nonroutine discharge, compared to those at high-volume hospitals. Women who had procedures by low-volume surgeons were also more likely to suffer complications and have nonroutine discharges compared to those with high-volume surgeons. Conclusion: Differences in hospital and surgeon volumes of urogynecologic procedures may contribute to variations in mortality and morbidity risks. Ó 2006 Mosby, Inc. All rights reserved.
Recently, the media, lay public, and policy makers have given increased attention to differences in postoperative morbidity and mortality associated with the
Supported in part by grant T32: HD040674-03; WIH/Brown Epidemiology/Clinical Trials Training Program; National Institute of Child Health and Human Development. Presented at the Thirty-Second Annual Meeting of the Society of Gynecologic Surgeons, Tucson, AZ, April 3-5, 2006. Reprints not available from the authors. 0002-9378/$ - see front matter Ó 2006 Mosby, Inc. All rights reserved. doi:10.1016/j.ajog.2006.07.015
volume of surgical procedures performed by a hospital and/or surgeon (eg, hospital and surgeon volume). Several studies have reported that hospitals and/or surgeons that perform higher volumes of specific surgical procedures, such as complex oncologic, cardiac, and vascular procedures, have lower risks of postoperative death and complications.1-5 These findings have led to initiatives aimed at concentrating selected surgical procedures in high-volume hospitals, or with high-volume surgeons.6 To date, the effect of hospital and surgeon
Sung et al volume on mortality and morbidity following urogynecologic procedures has not been well studied. It is estimated that over 135,000 inpatient antiincontinence procedures7 and over 225,000 procedures for pelvic organ prolapse8 are performed each year. It is also predicted that the number of women seeking care and undergoing treatment for pelvic floor disorders will increase.9 Following these procedures, the overall postoperative mortality rate has been reported to be 0.01% to 0.1%, but with complication rates as high as 20%.7,8,10,11 The primary goal of these elective procedures is to improve the quality of life of women; therefore, identifying factors that may reduce surgical mortality and morbidity, as well as maintaining a woman’s function, are critical. The purpose of this study was to determine whether patients undergoing urogynecologic procedures at lowvolume hospitals or with low-volume surgeons had higher risks of in-hospital death, complications, and nonroutine discharge. Our primary objective was to estimate the effect of hospital volume on outcomes using data from the Nationwide Inpatient Sample (NIS). Our secondary objective was to estimate the effect of surgeon volume on outcomes within high-volume hospitals.
Material and methods Data for this study were derived from the NIS for the years 1998 to 2003. The NIS contains all-payer data on inpatient hospital stays from states participating in the Healthcare Cost and Utilization Project and is maintained by the Agency for Health Care Research and Quality (AHRQ). The NIS approximates a 20% stratified sample of US hospitals and includes approximately 1000 community hospitals, public hospitals, and academic medical centers each year. Each sampled hospital is given a hospital identifier and provides data on all discharges for each year it participates and thus, the NIS can be used to obtain annual total volumes of specific procedures at individual hospitals. The principal surgeon is often identified using a synthetic code, although not all states permit release of surgeon identifiers for confidentiality reasons. Patient and hospital characteristics are also included in the NIS. A detailed explanation regarding the coding, sampling, and weighting strategy of the NIS can be found on the AHRQsponsored website.12 This study was approved by the Institutional Review Board of Women and Infants’ Hospital. We identified urogynecologic cases using urogynecologic procedurediagnosis code combinations based on the International Classification of Diseases, Ninth Revision-Clinical Modification (ICD-9). A detailed list of ICD-9 procedure codes included can be found online in supplementary Table A. We excluded patients who did not have an
1779 accompanying urogynecologic diagnosis. For example, women who underwent a hysterectomy for reasons other than a urogynecologic diagnosis were excluded. We restricted the study population to include only women over the age of 20 years who were admitted electively from a routine source. Patient demographic information and hospital characteristics were abstracted from the data set. Comorbidities were identified using ICD-9 secondary diagnosis codes and were compiled into a weighted comorbidity index according to the Dartmouth-Manitoba algorithm. According to this algorithm, ICD-9 diagnoses that could reflect both preexisting comorbidities and postoperative complications are excluded. For example, the ICD-9 code 437.0, ‘‘other and ill-defined cerebrovascular disease,’’ could be interpreted as either a comorbidity or complication and would be excluded from the comorbidity index. These methods are described in detail by Romano et al.13
Hospital and surgeon volumes Hospital and surgeon volumes of urogynecologic procedures performed were derived by counting the number of cases performed by each hospital and each identified surgeon in the database from 1998 to 2003. We calculated the mean number of cases per year for each hospital and surgeon. Hospitals were then ranked in order of increasing mean annual volumes. Three volume groups were defined by selecting whole-number cutoff points for mean annual hospital volume that most closely sorted the cases into 3 groups of equal size (tertiles). The cutoff points were established before outcomes were examined, assuring objective cutoff points. Low-volume hospitals (LVH) performed on average !92 cases per year, medium-volume hospitals (MVH) performed 92 to 185 cases per year, and high-volume hospitals (HVH) performed O185 cases per year. We used an analogous approach to define surgeon volumes. Using this method, low-volume surgeons (LVS) performed on average !8 cases per year, medium-volume surgeon (MVS) performed 8 to 18 cases per year, and high-volume surgeon (HVS) performed O18 cases per year.
Outcomes Our primary outcome was in-hospital mortality which was coded directly in the NIS. Secondary outcomes included perioperative complications and nonroutine discharge. Perioperative complications were determined using secondary ICD-9 diagnosis codes and are available online in supplementary Table B. Because our study population was restricted to elective admissions admitted from home, we defined nonroutine discharge as patients who were transferred to a skilled nursing facility or other intermediate or short-term facility. This was also coded directly in the NIS.
1780 Table I volume*
Sung et al Patient and hospital characteristics, by hospital Hospital volume
Characteristic Mean age (years G standard error) Mean comorbidity score Number procedures per patient (mean G standard error) Non-white race (%) Payer status (%) Medicare Private Medicaid Other Median area income (%) %$24,999 $25,000-$34,999 $35,000-$44,999 R$45,000 Large hospital bedsize (%) Hospital in urban setting (%) Teaching hospital (%) Hospital region (%) Northeast Midwest South West
Low (!92 cases/y)
Medium (92-185 cases/y)
High (O185 cases/y)
that statistical correction for clustering is critical for volume-outcome studies to prevent overestimation of associations.14,15 All statistical analyses were performed using SAS 8.2 (SAS Institute, Inc, Cary, NC) and SUDAAN 9.0 (Research Triangle Institute). All analyses were 2-tailed and P values ! .05 were considered significant in final analyses.
56.0 (.11) 55.4 (.15) 54.5 (.21) 0.14 2.7 (.01)
0.13 2.9 (.02)
0.13 3.0 (.03)
19.6
17.6
14.5
30.9 56.1 7.2 5.8
29.0 62.0 4.6 4.4
26.0 66.6 4.0 3.4
8.7 31.9 28.1 31.3 44.5
6.3 26.5 29.5 37.7 64.8
5.9 21.7 27.5 44.9 77.1
69.1
86.6
96.3
22.8
40.2
57.9
21.9 26.3 31.8 20.0
16.4 23.0 36.5 24.1
9.9 21.4 42.9 25.8
* Data are presented as weighted percentages unless otherwise indicated.
Statistical analysis Univariate and bivariate analyses were performed using Chi-square and simple linear regression analysis as appropriate. Potential confounders found to be significant on univariate analysis, interaction effects, and variables based on clinical relevance were added to our models. Generalized estimation equations (Taylor series linearization) were obtained to estimate relative risks for the effects of hospital and surgeon volumes on outcomes, using the patient as the unit of analysis, with volume measured at the hospital or surgeon level. Our final multivariable models adjusted for patient age, race, comorbidities based on the DartmouthManitoba index, year, median area income, number of procedures per admission, hospital teaching status, location, and region. We also adjusted for hospital and surgeon clustering, which may occur when patients treated by the same hospital or by the same surgeon are likely to experience similar outcomes as a result of other processes of care. Previous studies have shown
Results There were 310,759 women who underwent urogynecologic procedures at 2986 hospitals between 1998 and 2003. After weighting, this represents over 767,000 women over the age of 20 who underwent urogynecologic procedures each year in the US. Of the participating hospitals, 2390 (80%) were classified as LVHs, 404 (13.5%) as MVHs, and 192 (6.5%) as HVHs. These hospital volume cutoffs separated patients into 3 equal groups, with LVHs performing 33.9% of all urogynecologic procedures, MVHs performing 34.7%, and HVHs performing 31.4%. Table I presents the relative frequencies of patient and hospital characteristics and their relationships with hospital volume strata. Of note, non-white women and women with government insurance were more likely to undergo surgery at a lower volume hospital. The risk of death was inversely proportional to hospital volume, with women at low-volume hospitals having a higher risk of death (LVH 0.05%; MVH 0.04%; HVH 0.03%, P = .03) and nonroutine discharge (LVH 1.12%; MVH 0.71%; HVH 0.58%, P ! .01). There was no difference in risks of perioperative complications between LVH, MVH, and HVH. As shown in Table II, after adjusting for confounders, women who had urogynecologic surgery at LVHs were 2.75 (95% CI 2.33-3.16) times more likely to die and 1.63 (95% CI 1.44-1.83) times more likely to have a nonroutine discharge compared to those at HVHs.
Effect of surgeon volume Approximately 30% (n = 88,410) of our study population had masked surgeon identifiers available. Hospitals that reported surgeon information were similar to those that did not in hospital bed size, location, teaching status, and patient characteristics but were more likely to be in the Northeast and Southern regions of the US (P ! .01). The distribution of this subset of women by hospital and surgeon volumes is shown in Table III. Low-volume hospitals had a higher proportion of patients who underwent surgery by low-volume surgeons compared to high-volume hospitals. We examined the effect of surgeon volume on complications and nonroutine discharges within the highvolume hospital stratum. After adjusting for confounders, patients who had procedures by LVS and MVS were
Sung et al Table II
1781 Effect of hospital volume on outcomes following pelvic reconstructive procedures Hospital volume
Outcome Death Nonroutine discharge
Low
Medium
High
RR* (95% CI) 2.75 (2.33-3.16) 1.63 (1.44-1.83)
RR* (95% CI) 1.48 (1.03-1.93) 1.15 (0.95-1.37)
RR* (95% CI) Reference Reference
* Model adjusts for age, race, comorbidity, income, number of procedures, year, hospital teaching status, hospital location, hospital region, and hospital clustering.
Table III
Distribution of patients based on hospital and surgeon volumes* Hospital volume
Surgeon volume Low (n = 31,015) Medium (n = 29,329) High (n = 28,066)
Low (n = 28,326)
Medium (n = 30,214)
High (n = 29,870)
12,305 (43.4) 9644 (34.1) 6377 (22.5)
10,074 (33.4) 10,049 (33.2) 10,091 (33.4)
8636 (28.9) 9636 (32.3) 11,598 (38.8)
* Data are presented as unweighted counts with weighted column percentages (n [%]).
Table IV
Effect of surgeon volume on outcomes, within high-volume hospital stratum Surgeon volume
Outcome Complication Nonroutine discharge
Low
Medium
High
RR* (95% CI) 1.39 (1.21-1.57) 1.95 (1.43-2.48)
RR* (95% CI) 1.22 (1.03-1.42) 1.58 (1.11-2.05)
RR* (95% CI) Reference Reference
* Model adjusts for patient age, race, comorbidity index, income, number of procedures, year, hospital teaching status, hospital location, hospital region, and surgeon clustering.
more likely to suffer complications and have nonroutine discharges compared to women with HVS (see Table IV). This analysis was not performed for the outcome of death, given the overall small number of deaths in this stratum.
Comment This current study provides population-based outcomes from a nationally representative sample of almost 3000 hospitals that performed urogynecologic procedures from 1998 to 2003. Our findings suggest that the volume of urogynecologic procedures performed by hospitals as well as surgeons has a significant impact on postoperative outcomes and disposition. Women having surgery at LVH had almost a 3-fold higher risk of death. Our findings are consistent with studies in other fields.16,17 Birkmeyer et al evaluated cardiovascular procedures and major cancer resections using the national Medicare claims database and found that higher-volume hospitals had lower mortality rates. The authors also found that the absolute magnitude of the relation between volume and outcomes varied among different
types of procedures. This indicates that the effects of hospital and surgeon volumes on outcomes may vary depending on the type of procedure, and thus these findings cannot be generalized to all types of surgeries without further evaluation. The mechanisms related to the effect of hospital volume on outcomes are not fully characterized. Some have speculated that high-volume hospitals may have improved processes for postoperative care, better-staffed intensive care or nursing units, greater resources to provide complex perioperative care, and more subspecialty surgeons.1,18 Several studies have reported that subspecialty surgeons, or surgeons with higher caseloads, have improved outcomes and have suggested that the hospital-volume effect is at least in part mediated by improved outcomes of high-volume surgeons.19-21 For example, in a second study by Birkmeyer et al, the authors found that for many cardiovascular and oncologic procedures, the association between hospital volume and operative mortality was largely mediated by surgeon volume, suggesting that patients may improve their chances of survival, even at high-volume hospitals, by selecting surgeons with higher caseloads. Our results are also consistent with these findings.
1782 There are several limitations in utilizing administrative data. Only data from the index hospitalization were available and thus, women could not be followed after discharge from the hospital. Therefore, our outcomes are limited to in-hospital death, complications, and nonroutine discharge. Although some women may have a preoperative plan to be transferred to another facility, this would only be problematic as a confounder in our study if women in low-volume hospitals or who had low-volume surgeons were more likely to make these plans. Currently, there is no evidence to suggest this is the case. Furthermore, transfers to facilities, particularly skilled-nursing facilities, may reflect prediction of future functional status, as this implies a judgment that the patient’s return to independent living is unlikely. Therefore, nonroutine discharge is an important outcome to consider, especially after elective surgery. Determination of postoperative complications and comorbidities is based on secondary diagnosis codes, which are subject to human error. Whereas the outcome of death is an objective outcome and directly coded in the NIS, complications are subject to a coder’s interpretation. This may result in nondifferential misclassification and most likely explains why we did not find a hospital volume effect for the outcome of complications. To minimize misinterpretation, we also excluded ICD-9 codes that may represent either comorbidities or complications. This may also result in nondifferential misclassification. Using the NIS also precludes adjusting for confounders not included in the database, such as number of previous procedures, length of surgery, or patient functional status. Because not all patients had surgeon identifiers available, we were only able to perform surgeon-volume analyses on a subset of women, which may limit the external validity of these findings. We were also unable to identify surgeons who operate at multiple hospitals and thus, these surgeons may have been misclassified as low-volume surgeons. However, we would expect that this nondifferential misclassification would have biased our results towards the null. Despite these limitations, utilizing the NIS has many advantages. The sampling of hospitals is comprehensive and includes all geographic regions, hospital, and patient characteristics. There is a complete representation of all discharges from a random sample of US hospitals, and thus, there is no potential for selection bias on an individual or institutional level. Finally, the inclusion of all discharges from each institution allows accurate accounts of hospital volume, unlike other national databases, which may only capture a fraction of discharges per hospital. In conclusion, although urogynecologic procedures are associated with low overall mortality risks, differences in hospital and surgeon volumes may contribute to variations in mortality and morbidity. Reduction in
Sung et al these variations is likely to lead to improved care of women undergoing surgery for pelvic floor disorders.
Acknowledgments This work was completed as part of a Master’s Degree thesis in Public Health, Brown University. We thank Vincent Mor, PhD, for discussions during the development phase of this study.
References 1. Birkmeyer JD, Siewers AE, Finlayson EV, Stukel TA, Lucas FL, Batista I, et al. Hospital volume and surgical mortality in the United States. N Engl J Med 2002;346:1128-37. 2. Birkmeyer JD, Stukel TA, Siewers AE, Goodney PP, Wennberg DE, Lucas FL. Surgeon volume and operative mortality in the United States. N Engl J Med 2003;349:2117-27. 3. Halm EA, Lee C, Chassin MR. Is volume related to outcome in health care? A systematic review and methodologic critique of the literature. Ann Intern Med 2002;137:511-20. 4. Schrag D, Cramer LD, Bach PB, Cohen AM, Warren JL, Begg CB. Influence of hospital procedure volume on outcomes following surgery for colon cancer. JAMA 2000;284:3028-35. 5. Bianco FJ Jr, Riedel ER, Begg CB, Kattan MW, Scardino PT. Variations among high volume surgeons in the rate of complications after radical prostatectomy: further evidence that technique matters. J Urol 2005;173:2099-103. 6. Leapfrog Group for Patient Safety Standards. Available at: http:// Leapfroggroup.org. Accessed January, 2006. 7. Waetjen LE, Subak LL, Shen H, Lin F, Wang TH, Vittinghoff E, et al. Stress urinary incontinence surgery in the United States. Obstet Gynecol 2003;101:671-6. 8. Brown JS, Waetjen LE, Subak LL, Thom DH, Van den Eeden S, Vittinghoff E. Pelvic organ prolapse surgery in the United States, 1997. Am J Obstet Gynecol 2002;186:712-6. 9. Luber KM, Boero S, Choe JY. The demographics of pelvic floor disorders: current observations and future projections. Am J Obstet Gynecol 2001;184:1496-501; discussion 1501-3. 10. Sultana CJ, Campbell JW, Pisanelli WS, Sivinski L, Rimm AA. Morbidity and mortality of incontinence surgery in elderly women: an analysis of Medicare data. Am J Obstet Gynecol 1997;176: 344-8. 11. Sung VW, Weitzen S, Sokol ER, Rardin CR, Myers DL. Effect of patient age on increasing morbidity and mortality following urogynecologic surgery. Am J Obstet Gynecol 2006;194:1411-7. 12. Barker FG 2nd. Craniotomy for the resection of metastatic brain tumors in the U.S., 1988-2000: decreasing mortality and the effect of provider caseload. Cancer 2004;100:999-1007. 13. Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol 1993;46:1075-9; discussion 1081-90. 14. Panageas KS, Schrag D, Riedel E, Bach PB, Begg CB. The effect of clustering of outcomes on the association of procedure volume and surgical outcomes. Ann Intern Med 2003;139:658-65. 15. Greenfield S, Kaplan SH, Kahn R, Ninomiya J, Griffith JL. Profiling care provided by different groups of physicians: effects of patient case-mix (bias) and physician-level clustering on quality assessment results. Ann Intern Med 2002;136:111-21. 16. Hannan EL, Wu C, Ryan TJ, Bennett E, Culliford AT, Gold JP, et al. Do hospitals and surgeons with higher coronary artery bypass graft surgery volumes still have lower risk-adjusted mortality rates? Circulation 2003;108:795-801. 17. Dimick JB, Pronovost PJ, Cowan JA Jr, Lipsett PA, Stanley JC, Upchurch GR Jr. Variation in postoperative complication rates
Sung et al after high-risk surgery in the United States. Surgery 2003;134: 534-40; discussion 540-1. 18. Taub DA, Miller DC, Cowan JA, Dimick JB, Montie JE, Wei JT. Impact of surgical volume on mortality and length of stay after nephrectomy. Urology 2004;63:862-7. 19. Schrag D, Panageas KS, Riedel E, Hsien L, Bach PB, Guillem JG, et al. Surgeon volume compared to hospital volume as a predictor of outcome following primary colon cancer resection. J Surg Oncol 2003;83:68-78; discussion 78-9.
1783 20. Barker FG 2nd, Amin-Hanjani S, Butler WE, Ogilvy CS, Carter BS. In-hospital mortality and morbidity after surgical treatment of unruptured intracranial aneurysms in the United States, 1996-2000: the effect of hospital and surgeon volume. Neurosurgery 2003;52:995-1007; discussion 1007-9. 21. Konety BR, Dhawan V, Allareddy V, Joslyn SA. Impact of hospital and surgeon volume on in-hospital mortality from radical cystectomy: data from the health care utilization project. J Urol 2005; 173:1695-700.
www.ajog.org
Impact of hospital and surgeon volumes on outcomes following pelvic reconstructive surgery in the United States Vivian W. Sung, MD, MPH,a Michelle L. Rogers, PhD,b Deborah L. Myers, MD,a Melissa A. Clark, PhDb Division of Urogynecology and Pelvic Reconstructive Surgery, Department of Obstetrics and Gynecology,a Brown Medical School; Center for Gerontology and Health Care Research, Department of Community Health,b Brown University, Providence, RI Received for publication January 15, 2006; revised June 23, 2006; accepted July 5, 2006
KEY WORDS Pelvic reconstructive surgery Pelvic floor disorders Hospital volume Surgeon volume Mortality
Objective: The purpose of this study was to estimate the effect of hospital and surgeon volumes on outcomes following urogynecologic surgery. Study design: This was a retrospective cohort study of women who underwent urogynecologic procedures between 1998 and 2003 from the Nationwide Inpatient Sample. Hospitals and surgeons were categorized as low, medium, or high volume based on average number of cases per year. Outcomes included in-hospital mortality, complications, and nonroutine discharges. Multivariable analyses were performed using generalized estimation equations to estimate relative risks. Results: There were 310,759 women and 2986 hospitals. Women who had procedures at lowvolume hospitals were 2.75 (95% CI 2.33-3.16) times more likely to die and 1.63 (95% CI 1.44-1.83) times more likely to have a nonroutine discharge, compared to those at high-volume hospitals. Women who had procedures by low-volume surgeons were also more likely to suffer complications and have nonroutine discharges compared to those with high-volume surgeons. Conclusion: Differences in hospital and surgeon volumes of urogynecologic procedures may contribute to variations in mortality and morbidity risks. Ó 2006 Mosby, Inc. All rights reserved.
Recently, the media, lay public, and policy makers have given increased attention to differences in postoperative morbidity and mortality associated with the
Supported in part by grant T32: HD040674-03; WIH/Brown Epidemiology/Clinical Trials Training Program; National Institute of Child Health and Human Development. Presented at the Thirty-Second Annual Meeting of the Society of Gynecologic Surgeons, Tucson, AZ, April 3-5, 2006. Reprints not available from the authors. 0002-9378/$ - see front matter Ó 2006 Mosby, Inc. All rights reserved. doi:10.1016/j.ajog.2006.07.015
volume of surgical procedures performed by a hospital and/or surgeon (eg, hospital and surgeon volume). Several studies have reported that hospitals and/or surgeons that perform higher volumes of specific surgical procedures, such as complex oncologic, cardiac, and vascular procedures, have lower risks of postoperative death and complications.1-5 These findings have led to initiatives aimed at concentrating selected surgical procedures in high-volume hospitals, or with high-volume surgeons.6 To date, the effect of hospital and surgeon
Sung et al volume on mortality and morbidity following urogynecologic procedures has not been well studied. It is estimated that over 135,000 inpatient antiincontinence procedures7 and over 225,000 procedures for pelvic organ prolapse8 are performed each year. It is also predicted that the number of women seeking care and undergoing treatment for pelvic floor disorders will increase.9 Following these procedures, the overall postoperative mortality rate has been reported to be 0.01% to 0.1%, but with complication rates as high as 20%.7,8,10,11 The primary goal of these elective procedures is to improve the quality of life of women; therefore, identifying factors that may reduce surgical mortality and morbidity, as well as maintaining a woman’s function, are critical. The purpose of this study was to determine whether patients undergoing urogynecologic procedures at lowvolume hospitals or with low-volume surgeons had higher risks of in-hospital death, complications, and nonroutine discharge. Our primary objective was to estimate the effect of hospital volume on outcomes using data from the Nationwide Inpatient Sample (NIS). Our secondary objective was to estimate the effect of surgeon volume on outcomes within high-volume hospitals.
Material and methods Data for this study were derived from the NIS for the years 1998 to 2003. The NIS contains all-payer data on inpatient hospital stays from states participating in the Healthcare Cost and Utilization Project and is maintained by the Agency for Health Care Research and Quality (AHRQ). The NIS approximates a 20% stratified sample of US hospitals and includes approximately 1000 community hospitals, public hospitals, and academic medical centers each year. Each sampled hospital is given a hospital identifier and provides data on all discharges for each year it participates and thus, the NIS can be used to obtain annual total volumes of specific procedures at individual hospitals. The principal surgeon is often identified using a synthetic code, although not all states permit release of surgeon identifiers for confidentiality reasons. Patient and hospital characteristics are also included in the NIS. A detailed explanation regarding the coding, sampling, and weighting strategy of the NIS can be found on the AHRQsponsored website.12 This study was approved by the Institutional Review Board of Women and Infants’ Hospital. We identified urogynecologic cases using urogynecologic procedurediagnosis code combinations based on the International Classification of Diseases, Ninth Revision-Clinical Modification (ICD-9). A detailed list of ICD-9 procedure codes included can be found online in supplementary Table A. We excluded patients who did not have an
1779 accompanying urogynecologic diagnosis. For example, women who underwent a hysterectomy for reasons other than a urogynecologic diagnosis were excluded. We restricted the study population to include only women over the age of 20 years who were admitted electively from a routine source. Patient demographic information and hospital characteristics were abstracted from the data set. Comorbidities were identified using ICD-9 secondary diagnosis codes and were compiled into a weighted comorbidity index according to the Dartmouth-Manitoba algorithm. According to this algorithm, ICD-9 diagnoses that could reflect both preexisting comorbidities and postoperative complications are excluded. For example, the ICD-9 code 437.0, ‘‘other and ill-defined cerebrovascular disease,’’ could be interpreted as either a comorbidity or complication and would be excluded from the comorbidity index. These methods are described in detail by Romano et al.13
Hospital and surgeon volumes Hospital and surgeon volumes of urogynecologic procedures performed were derived by counting the number of cases performed by each hospital and each identified surgeon in the database from 1998 to 2003. We calculated the mean number of cases per year for each hospital and surgeon. Hospitals were then ranked in order of increasing mean annual volumes. Three volume groups were defined by selecting whole-number cutoff points for mean annual hospital volume that most closely sorted the cases into 3 groups of equal size (tertiles). The cutoff points were established before outcomes were examined, assuring objective cutoff points. Low-volume hospitals (LVH) performed on average !92 cases per year, medium-volume hospitals (MVH) performed 92 to 185 cases per year, and high-volume hospitals (HVH) performed O185 cases per year. We used an analogous approach to define surgeon volumes. Using this method, low-volume surgeons (LVS) performed on average !8 cases per year, medium-volume surgeon (MVS) performed 8 to 18 cases per year, and high-volume surgeon (HVS) performed O18 cases per year.
Outcomes Our primary outcome was in-hospital mortality which was coded directly in the NIS. Secondary outcomes included perioperative complications and nonroutine discharge. Perioperative complications were determined using secondary ICD-9 diagnosis codes and are available online in supplementary Table B. Because our study population was restricted to elective admissions admitted from home, we defined nonroutine discharge as patients who were transferred to a skilled nursing facility or other intermediate or short-term facility. This was also coded directly in the NIS.
1780 Table I volume*
Sung et al Patient and hospital characteristics, by hospital Hospital volume
Characteristic Mean age (years G standard error) Mean comorbidity score Number procedures per patient (mean G standard error) Non-white race (%) Payer status (%) Medicare Private Medicaid Other Median area income (%) %$24,999 $25,000-$34,999 $35,000-$44,999 R$45,000 Large hospital bedsize (%) Hospital in urban setting (%) Teaching hospital (%) Hospital region (%) Northeast Midwest South West
Low (!92 cases/y)
Medium (92-185 cases/y)
High (O185 cases/y)
that statistical correction for clustering is critical for volume-outcome studies to prevent overestimation of associations.14,15 All statistical analyses were performed using SAS 8.2 (SAS Institute, Inc, Cary, NC) and SUDAAN 9.0 (Research Triangle Institute). All analyses were 2-tailed and P values ! .05 were considered significant in final analyses.
56.0 (.11) 55.4 (.15) 54.5 (.21) 0.14 2.7 (.01)
0.13 2.9 (.02)
0.13 3.0 (.03)
19.6
17.6
14.5
30.9 56.1 7.2 5.8
29.0 62.0 4.6 4.4
26.0 66.6 4.0 3.4
8.7 31.9 28.1 31.3 44.5
6.3 26.5 29.5 37.7 64.8
5.9 21.7 27.5 44.9 77.1
69.1
86.6
96.3
22.8
40.2
57.9
21.9 26.3 31.8 20.0
16.4 23.0 36.5 24.1
9.9 21.4 42.9 25.8
* Data are presented as weighted percentages unless otherwise indicated.
Statistical analysis Univariate and bivariate analyses were performed using Chi-square and simple linear regression analysis as appropriate. Potential confounders found to be significant on univariate analysis, interaction effects, and variables based on clinical relevance were added to our models. Generalized estimation equations (Taylor series linearization) were obtained to estimate relative risks for the effects of hospital and surgeon volumes on outcomes, using the patient as the unit of analysis, with volume measured at the hospital or surgeon level. Our final multivariable models adjusted for patient age, race, comorbidities based on the DartmouthManitoba index, year, median area income, number of procedures per admission, hospital teaching status, location, and region. We also adjusted for hospital and surgeon clustering, which may occur when patients treated by the same hospital or by the same surgeon are likely to experience similar outcomes as a result of other processes of care. Previous studies have shown
Results There were 310,759 women who underwent urogynecologic procedures at 2986 hospitals between 1998 and 2003. After weighting, this represents over 767,000 women over the age of 20 who underwent urogynecologic procedures each year in the US. Of the participating hospitals, 2390 (80%) were classified as LVHs, 404 (13.5%) as MVHs, and 192 (6.5%) as HVHs. These hospital volume cutoffs separated patients into 3 equal groups, with LVHs performing 33.9% of all urogynecologic procedures, MVHs performing 34.7%, and HVHs performing 31.4%. Table I presents the relative frequencies of patient and hospital characteristics and their relationships with hospital volume strata. Of note, non-white women and women with government insurance were more likely to undergo surgery at a lower volume hospital. The risk of death was inversely proportional to hospital volume, with women at low-volume hospitals having a higher risk of death (LVH 0.05%; MVH 0.04%; HVH 0.03%, P = .03) and nonroutine discharge (LVH 1.12%; MVH 0.71%; HVH 0.58%, P ! .01). There was no difference in risks of perioperative complications between LVH, MVH, and HVH. As shown in Table II, after adjusting for confounders, women who had urogynecologic surgery at LVHs were 2.75 (95% CI 2.33-3.16) times more likely to die and 1.63 (95% CI 1.44-1.83) times more likely to have a nonroutine discharge compared to those at HVHs.
Effect of surgeon volume Approximately 30% (n = 88,410) of our study population had masked surgeon identifiers available. Hospitals that reported surgeon information were similar to those that did not in hospital bed size, location, teaching status, and patient characteristics but were more likely to be in the Northeast and Southern regions of the US (P ! .01). The distribution of this subset of women by hospital and surgeon volumes is shown in Table III. Low-volume hospitals had a higher proportion of patients who underwent surgery by low-volume surgeons compared to high-volume hospitals. We examined the effect of surgeon volume on complications and nonroutine discharges within the highvolume hospital stratum. After adjusting for confounders, patients who had procedures by LVS and MVS were
Sung et al Table II
1781 Effect of hospital volume on outcomes following pelvic reconstructive procedures Hospital volume
Outcome Death Nonroutine discharge
Low
Medium
High
RR* (95% CI) 2.75 (2.33-3.16) 1.63 (1.44-1.83)
RR* (95% CI) 1.48 (1.03-1.93) 1.15 (0.95-1.37)
RR* (95% CI) Reference Reference
* Model adjusts for age, race, comorbidity, income, number of procedures, year, hospital teaching status, hospital location, hospital region, and hospital clustering.
Table III
Distribution of patients based on hospital and surgeon volumes* Hospital volume
Surgeon volume Low (n = 31,015) Medium (n = 29,329) High (n = 28,066)
Low (n = 28,326)
Medium (n = 30,214)
High (n = 29,870)
12,305 (43.4) 9644 (34.1) 6377 (22.5)
10,074 (33.4) 10,049 (33.2) 10,091 (33.4)
8636 (28.9) 9636 (32.3) 11,598 (38.8)
* Data are presented as unweighted counts with weighted column percentages (n [%]).
Table IV
Effect of surgeon volume on outcomes, within high-volume hospital stratum Surgeon volume
Outcome Complication Nonroutine discharge
Low
Medium
High
RR* (95% CI) 1.39 (1.21-1.57) 1.95 (1.43-2.48)
RR* (95% CI) 1.22 (1.03-1.42) 1.58 (1.11-2.05)
RR* (95% CI) Reference Reference
* Model adjusts for patient age, race, comorbidity index, income, number of procedures, year, hospital teaching status, hospital location, hospital region, and surgeon clustering.
more likely to suffer complications and have nonroutine discharges compared to women with HVS (see Table IV). This analysis was not performed for the outcome of death, given the overall small number of deaths in this stratum.
Comment This current study provides population-based outcomes from a nationally representative sample of almost 3000 hospitals that performed urogynecologic procedures from 1998 to 2003. Our findings suggest that the volume of urogynecologic procedures performed by hospitals as well as surgeons has a significant impact on postoperative outcomes and disposition. Women having surgery at LVH had almost a 3-fold higher risk of death. Our findings are consistent with studies in other fields.16,17 Birkmeyer et al evaluated cardiovascular procedures and major cancer resections using the national Medicare claims database and found that higher-volume hospitals had lower mortality rates. The authors also found that the absolute magnitude of the relation between volume and outcomes varied among different
types of procedures. This indicates that the effects of hospital and surgeon volumes on outcomes may vary depending on the type of procedure, and thus these findings cannot be generalized to all types of surgeries without further evaluation. The mechanisms related to the effect of hospital volume on outcomes are not fully characterized. Some have speculated that high-volume hospitals may have improved processes for postoperative care, better-staffed intensive care or nursing units, greater resources to provide complex perioperative care, and more subspecialty surgeons.1,18 Several studies have reported that subspecialty surgeons, or surgeons with higher caseloads, have improved outcomes and have suggested that the hospital-volume effect is at least in part mediated by improved outcomes of high-volume surgeons.19-21 For example, in a second study by Birkmeyer et al, the authors found that for many cardiovascular and oncologic procedures, the association between hospital volume and operative mortality was largely mediated by surgeon volume, suggesting that patients may improve their chances of survival, even at high-volume hospitals, by selecting surgeons with higher caseloads. Our results are also consistent with these findings.
1782 There are several limitations in utilizing administrative data. Only data from the index hospitalization were available and thus, women could not be followed after discharge from the hospital. Therefore, our outcomes are limited to in-hospital death, complications, and nonroutine discharge. Although some women may have a preoperative plan to be transferred to another facility, this would only be problematic as a confounder in our study if women in low-volume hospitals or who had low-volume surgeons were more likely to make these plans. Currently, there is no evidence to suggest this is the case. Furthermore, transfers to facilities, particularly skilled-nursing facilities, may reflect prediction of future functional status, as this implies a judgment that the patient’s return to independent living is unlikely. Therefore, nonroutine discharge is an important outcome to consider, especially after elective surgery. Determination of postoperative complications and comorbidities is based on secondary diagnosis codes, which are subject to human error. Whereas the outcome of death is an objective outcome and directly coded in the NIS, complications are subject to a coder’s interpretation. This may result in nondifferential misclassification and most likely explains why we did not find a hospital volume effect for the outcome of complications. To minimize misinterpretation, we also excluded ICD-9 codes that may represent either comorbidities or complications. This may also result in nondifferential misclassification. Using the NIS also precludes adjusting for confounders not included in the database, such as number of previous procedures, length of surgery, or patient functional status. Because not all patients had surgeon identifiers available, we were only able to perform surgeon-volume analyses on a subset of women, which may limit the external validity of these findings. We were also unable to identify surgeons who operate at multiple hospitals and thus, these surgeons may have been misclassified as low-volume surgeons. However, we would expect that this nondifferential misclassification would have biased our results towards the null. Despite these limitations, utilizing the NIS has many advantages. The sampling of hospitals is comprehensive and includes all geographic regions, hospital, and patient characteristics. There is a complete representation of all discharges from a random sample of US hospitals, and thus, there is no potential for selection bias on an individual or institutional level. Finally, the inclusion of all discharges from each institution allows accurate accounts of hospital volume, unlike other national databases, which may only capture a fraction of discharges per hospital. In conclusion, although urogynecologic procedures are associated with low overall mortality risks, differences in hospital and surgeon volumes may contribute to variations in mortality and morbidity. Reduction in
Sung et al these variations is likely to lead to improved care of women undergoing surgery for pelvic floor disorders.
Acknowledgments This work was completed as part of a Master’s Degree thesis in Public Health, Brown University. We thank Vincent Mor, PhD, for discussions during the development phase of this study.
References 1. Birkmeyer JD, Siewers AE, Finlayson EV, Stukel TA, Lucas FL, Batista I, et al. Hospital volume and surgical mortality in the United States. N Engl J Med 2002;346:1128-37. 2. Birkmeyer JD, Stukel TA, Siewers AE, Goodney PP, Wennberg DE, Lucas FL. Surgeon volume and operative mortality in the United States. N Engl J Med 2003;349:2117-27. 3. Halm EA, Lee C, Chassin MR. Is volume related to outcome in health care? A systematic review and methodologic critique of the literature. Ann Intern Med 2002;137:511-20. 4. Schrag D, Cramer LD, Bach PB, Cohen AM, Warren JL, Begg CB. Influence of hospital procedure volume on outcomes following surgery for colon cancer. JAMA 2000;284:3028-35. 5. Bianco FJ Jr, Riedel ER, Begg CB, Kattan MW, Scardino PT. Variations among high volume surgeons in the rate of complications after radical prostatectomy: further evidence that technique matters. J Urol 2005;173:2099-103. 6. Leapfrog Group for Patient Safety Standards. Available at: http:// Leapfroggroup.org. Accessed January, 2006. 7. Waetjen LE, Subak LL, Shen H, Lin F, Wang TH, Vittinghoff E, et al. Stress urinary incontinence surgery in the United States. Obstet Gynecol 2003;101:671-6. 8. Brown JS, Waetjen LE, Subak LL, Thom DH, Van den Eeden S, Vittinghoff E. Pelvic organ prolapse surgery in the United States, 1997. Am J Obstet Gynecol 2002;186:712-6. 9. Luber KM, Boero S, Choe JY. The demographics of pelvic floor disorders: current observations and future projections. Am J Obstet Gynecol 2001;184:1496-501; discussion 1501-3. 10. Sultana CJ, Campbell JW, Pisanelli WS, Sivinski L, Rimm AA. Morbidity and mortality of incontinence surgery in elderly women: an analysis of Medicare data. Am J Obstet Gynecol 1997;176: 344-8. 11. Sung VW, Weitzen S, Sokol ER, Rardin CR, Myers DL. Effect of patient age on increasing morbidity and mortality following urogynecologic surgery. Am J Obstet Gynecol 2006;194:1411-7. 12. Barker FG 2nd. Craniotomy for the resection of metastatic brain tumors in the U.S., 1988-2000: decreasing mortality and the effect of provider caseload. Cancer 2004;100:999-1007. 13. Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol 1993;46:1075-9; discussion 1081-90. 14. Panageas KS, Schrag D, Riedel E, Bach PB, Begg CB. The effect of clustering of outcomes on the association of procedure volume and surgical outcomes. Ann Intern Med 2003;139:658-65. 15. Greenfield S, Kaplan SH, Kahn R, Ninomiya J, Griffith JL. Profiling care provided by different groups of physicians: effects of patient case-mix (bias) and physician-level clustering on quality assessment results. Ann Intern Med 2002;136:111-21. 16. Hannan EL, Wu C, Ryan TJ, Bennett E, Culliford AT, Gold JP, et al. Do hospitals and surgeons with higher coronary artery bypass graft surgery volumes still have lower risk-adjusted mortality rates? Circulation 2003;108:795-801. 17. Dimick JB, Pronovost PJ, Cowan JA Jr, Lipsett PA, Stanley JC, Upchurch GR Jr. Variation in postoperative complication rates
Sung et al after high-risk surgery in the United States. Surgery 2003;134: 534-40; discussion 540-1. 18. Taub DA, Miller DC, Cowan JA, Dimick JB, Montie JE, Wei JT. Impact of surgical volume on mortality and length of stay after nephrectomy. Urology 2004;63:862-7. 19. Schrag D, Panageas KS, Riedel E, Hsien L, Bach PB, Guillem JG, et al. Surgeon volume compared to hospital volume as a predictor of outcome following primary colon cancer resection. J Surg Oncol 2003;83:68-78; discussion 78-9.
1783 20. Barker FG 2nd, Amin-Hanjani S, Butler WE, Ogilvy CS, Carter BS. In-hospital mortality and morbidity after surgical treatment of unruptured intracranial aneurysms in the United States, 1996-2000: the effect of hospital and surgeon volume. Neurosurgery 2003;52:995-1007; discussion 1007-9. 21. Konety BR, Dhawan V, Allareddy V, Joslyn SA. Impact of hospital and surgeon volume on in-hospital mortality from radical cystectomy: data from the health care utilization project. J Urol 2005; 173:1695-700.