Associations Between Surgical Site Infection Risk and Hospital Operation Volume and Surgeon Operation Volume Among Hospitals in the Dutch Nosocomial Infection Surveillance Network
Objective. To examine the association between hospital operation volume and surgeon operation volume and the risk of surgical site infection (SSI).
Design. Prospective, multicenter cohort study based on surveillance data.
Methods. Data were obtained from the Dutch surveillance network for nosocomial infections (Preventie Ziekenhuisinfecties door Surveillance [PREZIES]) on 9 different types of orthopedic surgery, general surgery, and gynecology procedures performed during 1996‐2003. Multilevel logistic regression analysis was performed to assess the independent effect of hospital volume and surgeon volume on SSI risk.
Results. Hospital volume was not significantly associated with SSI risk for any of the selected procedures. Low surgeon volume was associated with an increased risk for an infection for 7 of 9 types of procedures, although this effect was statistically significant only for knee arthroplasty. For 4 procedures, the odds of exceeding the 75th percentile for duration of surgery were greater when the surgeon volume was low than when the surgeon volume was moderate or high.
Conclusions. Patients operated on by surgeons with a low operation volume seem to have a higher risk of developing an SSI with some procedures, particularly knee arthroplasty. The higher SSI risk for surgeons with a low operation volume is possibly partly mediated by the longer duration of surgery, a well‐known risk factor for development of SSI.
Received April 11, 2006; accepted July 24, 2006; electronically published April 13, 2007.
Recently, a number of articles have been published about the relation between the volume of procedures at hospitals (hereafter, “hospital volume”) and different outcome measures, such as mortality, surgical site infection (SSI), and deep vein thrombosis. Most articles have focused on the relation between hospital volume and the operation‐related mortality among patients who had high‐risk surgery performed.1 In a systematic review, an inverse relation between hospital volume and mortality was confirmed, but the methodological shortcomings of a number of studies complicated the conclusion.1 Fewer publications focused on the effect that the volume of operative procedures performed by surgeons (hereafter, “surgeon volume”) has on mortality. A study concerning the joint effects of hospital volume and surgeon volume for 8 surgical procedures concluded that the observed association between hospital volume and mortality was largely mediated by surgeon volume.2
In a multicenter study performed in the United Kingdom, a higher risk for SSI after orthopedic surgery in centers with a high hospital volume was reported.3 However, in the US Medicare population, an inverse relation between SSI risk and both hospital volume and surgeon volume was observed for total hip arthroplasty and knee arthroplasty.4,5 In the Preventie Ziekenhuisinfecties door Surveillance (PREZIES) network, the Dutch surveillance network for nosocomial infections, an inverse relation between hospital volume and SSI risk after total hip arthroplasty was also observed.6 Surgeon volume was not considered in that study. In this study, we analyzed both hospital volume and surgeon volume as independent risk factors for SSIs for 9 different types of orthopedic surgeries, general surgeries, and gynecologic surgeries in the PREZIES network, taking into account the multicenter origin of the data.
Methods
Patient Population and Data Collection
The PREZIES network was started in 1996.7 Hospital participation in the PREZIES network is voluntary and confidential. According to the PREZIES SSI protocol, hospitals are free to choose the type of surgery they will include in the surveillance. Hospitals can also choose how long they want to participate, but a minimum of 3 months is recommended. Patients aged less than 1 year and patients with an infection at the time of hospital admission are excluded. Data collection for each patient is performed according to the PREZIES protocol.8 The following information is routinely collected on all patients, procedures, hospitals, and infections: sex, date of birth, date of admission, date of surgery, date of discharge, type of procedure, duration of procedure, American Society of Anesthesiologists physical status classification (hereafter, “ASA classification”), wound contamination class (hereafter, “wound class”), use of antibiotic prophylaxis (yes or no), elective or emergency procedure, university or peripheral hospital, date of infection, and type of infection (superficial incisional, deep incisional, or organ space).
PREZIES uses the definitions for infections from the Centers for Disease Control and Prevention, translated by the Dutch Working Party on Infection Prevention.9 Double and illogical data entries were omitted from our analysis.
Selection of Data and Procedures
Data on patients who had a surgical procedure performed during 1996‐2003 and underwent active postdischarge surveillance for 30 days or for 1 year (if the patient received a nonhuman implant) were selected for analysis.10 In The Netherlands, patients routinely return to the surgeon in the hospital where the procedure was performed for postdischarge examination and, if necessary, further treatment, which makes outpatient surveillance feasible. In addition, only records without missing values for the evaluated risk factors and with a valid surgeon code were included in the analysis. Next, we selected all procedures with more than 1,000 records and for which a sufficient number of infections (ie, more than 35) were recorded. The following 9 procedures satisfied these criteria: radical mastectomy, resection of colon, appendectomy, total abdominal hysterectomy, cesarean section, partial hip arthroplasty, total hip arthroplasty, revision of hip arthroplasty, and knee arthroplasty.
Data Analysis
Each procedure was analyzed separately for the effect of hospital volume and the effect of surgeon volume. The hospital volume for each procedure was calculated by dividing the number of surgical procedures performed at each participating hospital by the number of years of the study period. Next, hospitals were divided into 3 classes based on the tertiles of their hospital volume.
The surgeon volume for each procedure was calculated in the same way. However, in the rare event that a surgeon had performed only 1 specific operation during a surveillance period of more than 1 year, the surgeon volume was set to 1 per year. Next, surgeons were divided into 3 classes based on the tertiles of their surgeon volume.
To determine the associations between the hospital volume and surgeon volume and the risk of SSI adjusted for the case mix, the hierarchical structure of the data had to be taken into account. In the analysis of hospital volume, a 2‐level model was used, with operations as the first level and hospitals as the second level. In the analysis of surgeon volume, a 3‐level model was used, with operations as the first level, surgeons as the second level, and hospitals as the third level. In all analyses, the lowest volume stratum was used as the reference category, and random intercept models were used. Multilevel data analysis was performed with Proc Glimmix, version 9.1 for Windows (SAS Institute).
All possible risk factors that are routinely recorded were considered as potential confounders. The ASA classification was divided in 2 classes: nonsevere systemic disease (ASA classification of 1‐2) and severe systemic underlying disease (ASA classification of 3‐5). Wound class was also divided into 2 classes: clean and clean‐contaminated surgery (wound class of 1‐2) and contaminated or dirty or infected surgery (wound class of 3‐4). Procedure‐specific age categories were based on age tertiles.
Potential confounders associated with an SSI, defined as variables with a P value of less than .20 in a univariate analysis, were considered for multivariate analysis. Next, the best model was created by forward selection of these potential confounders to the model. Because the Glimmix procedure gives only pseudofit statistics, we used the significance of an added confounder for model selection. A confounder was added to the multivariate model if its estimate had a P value of less than .05. The association between surgeon volume and the 75th percentile duration of surgery in our database, adjusted for potential confounders, was examined separately with a 3‐level model, using the Glimmix procedure in SAS.
Results
Patient Characteristics
Table 1 provides the characteristics of the patient population, according to surgical procedure. As expected, the wound class for operations on the digestive system was high. With respect to the orthopedic procedures, patients who underwent partial hip arthroplasty more often had an ASA classification greater than 2 and an emergency operation than patients who underwent a different arthroplastic procedure. The operation volume tertiles for both hospitals and surgeons are described in Table 2. Partial hip arthroplasty was performed by a relative high number of surgeons.
Surgeon Volume and Duration of Surgery
In Table 3, the odds of exceeding the 75th percentile of duration of surgery for surgeons with moderate and high surgeon volumes is compared with that for surgeons with low surgeon volumes. With total abdominal hysterectomy, cesarean section, total hip arthroplasty, and knee arthroplasty, the odds of exceeding the 75th percentile for duration of surgery were greater when the surgeon volume was low than when the surgeon volume was moderate or high. For the other procedures, no clear effect of surgeon volume on the 75th percentile for duration of surgery was observed (
).
Hospital Volume and SSI Risk
Table 4 gives the number of operations and the number of subsequent infections for each hospital volume and surgeon volume class. Because the volume classes are based on the tertiles of the number of procedures per hospital and per surgeon, high‐volume classes contain a higher absolute number of operations.
There was no clear effect of hospital volume on crude infection rates (Table 4). Centers with a low hospital volume had a lower rate of infection after radical mastectomy and partial hip arthroplasty but a greater rate of infection after colon resection and total hip arthroplasty. Also, after adjustment for confounders, none of the hospital volumes for the procedures were associated with the infection rate (Table 5). For most procedures, the confidence intervals were broad. Small confidence intervals were revealed only for colon resection, total hip arthroplasty, and knee arthroplasty, for which the infection rates were slightly lower for both high hospital volumes and high surgeon volumes. The SSI risk was somewhat lower after total hip arthroplasty and knee arthroplasty at centers with a moderate hospital volume, compared with centers with a high hospital volume, but this difference was not statistically significant.
Surgeon Volume and SSI Risk
The association between crude infection rates and surgeon volume is given in Table 4. For resection of the colon, total abdominal hysterectomy, partial and total hip arthroplasties, and knee arthroplasty, infection rates seemed to decrease as the surgeon volume increased. After adjustment for relevant confounders, surgeon volume seemed inversely related to the infection rate for all procedures except appendectomy, cesarean section, and hip revision. For all 6 procedures in which an effect was observed, differences between low surgeon volumes and high and moderate surgeon volumes were small. Although the SSI risk was lower after these 6 procedures for surgeons with a high surgeon volume than for surgeons with a low surgeon volume, the risk was statistically significantly lower only after knee arthroplasty. For colon resection, the SSI risk was substantially lower for surgeons with a high surgeon volume, compared with surgeons with a low surgeon volume, although the difference was not statistically significant.
Discussion
For some operative procedures, we observed a trend toward a higher SSI risk for surgeons with a low surgeon volume. In orthopedic surgery, low surgeon volume is a known risk factor for overall adverse outcomes,4 dislocation,11 prolonged length of stay,12 and mortality.11,13
Because of the relatively low numbers of some procedures, the statistical power of our study was insufficient for detecting relatively small effects of operation volume. Also, several possibilities exist for bias toward the null in our analyses. First, it is possible that some surgeons worked in different hospitals simultaneously or sequentially. Because names of surgeons are not registered, we could not combine data on the operations performed by such surgeons. However, most surgeons in our database were probably restricted to one hospital; therefore, bias is small and, if present, is directed to an underestimation of the effect of surgeon volume. Second, we analyzed the surgeon volume according to surgical procedure. Generally, surgeons regularly perform several different procedures (eg, different types of orthopedic arthroplasties). More general‐surgery experience through increased total operation volume can possibly also affect the SSI risk for specific procedures, thus taking into account the fact that operation‐specific volume alone may lead to an underestimation of the effect of surgeon volume on specific procedures. Third, from one hospital in our network, we know that all assistant surgeons are registered under one surgeon‐identifying number. If this practice is common enough, bias toward the null will result.
Multilevel models are an appropriate and efficient method to analyze multicenter data, because they take into account variation between subjects within one level (eg, hospitals) that is not explained by the registered risk factors. However, as we reported elsewhere,14 the estimates of effect are smaller and the confidence intervals are broader than those in the commonly used regular 1‐level regression models that do not take into account this variation.
Duration of Surgery
Adjustment of odds ratios is often done with the National Nosocomial Infections Surveillance system risk index, which consists of the 75th percentile duration of surgery, wound class, and ASA classification. However, we observed that, even after adjustment for the case mix, surgeons with a low surgeon volume exceeded the 75th percentile for duration of surgery significantly more frequently than surgeons with moderate or high surgeon volumes for total abdominal hysterectomy, cesarean section, total hip arthroplasty, and knee arthroplasty. This finding indicates that the duration of surgery may be an intermediate variable in the association between surgeon volume and SSI risk for these procedures: a low surgeon volume leads to a longer duration of surgery, which is a well‐known risk factor for the development of SSI.15 This correlation is in line with the “practice makes perfect” hypothesis16 and with findings from a 4‐year survey on the development of a surgeon’s expertise.17 Because of the strong correlation between duration of surgery and surgeon volume in some procedures, we did not use the 75th percentile as a confounder in our models.
Hospital Volume
Despite the vast number of articles that have discussed the relation between hospital volume and mortality, only a few articles, mainly those concerning orthopedic surgery, discuss the relationship between hospital volume and SSI risk.3‐6,18 The results of some studies on orthopedic surgery were not comparable with our findings, because the volume categories were different6 or because other procedures were included.5 Two studies were more comparable, but they did not find evidence that hospital volume has a relevant effect on infection rates.3,4 Our data revealed that the SSI risk slightly decreased as the hospital volume increased, especially in centers with a moderate hospital volume, for colon resection, total hip arthroplasty, and knee arthroplasty. However, the broad confidence intervals suggested no important relation between hospital volume and SSI risk.
Surgeon Volume
Compared with the results of our analysis of hospital volume, confidence intervals for surgeon volume were smaller, and for most procedures an inverse relation between surgeon volume and SSI risk was found. These results indicate that, for most procedures, surgeon volume may be more important than hospital volume for determining the SSI risk. Differences between low surgeon volume and the other 2 surgeon volume classes might be partly mediated by differences in operation duration, because exceeding the 75th percentile duration of surgery is a well‐known risk factor for SSI.15 The P value for surgeon volume class in the multilevel analysis was less than .10 for all orthopedic procedures except revision of hip arthroplasty and radical mastectomy, suggesting clustering of infections according to surgeon volume. For these procedures in particular, a model that would ignore clustering of SSI among specific surgeons would lead to biased results.
Our results are in line with those in the literature, although most articles that addressed the relationship between surgeon volume and SSI risk were concerned only with orthopedic procedures (eg, knee and total hip arthroplasties).3,4 We also tried to analyze deep SSI separately, but for most procedures the number of deep SSIs was too low for statistical analysis.
Although many surgeons perform only a small number of specific procedures per year, only a small proportion of patients have operative procedures performed by these surgeons. The procedure with the lowest surgeon volume involved only approximately 6% of all operations, because of the classification system we used. Therefore, the overall effect of low surgeon volume on the absolute number of SSIs was rather small. In addition, our analyses are not fit to determine the required minimum number of specific operations to be performed in a hospital or by a surgeon that would minimize the risk of SSI, because we merely looked at trends on cutoff points, which we arbitrarily based on tertiles, for operation volume.
In conclusion, we observed a longer duration of surgery and a trend for more infections among surgeons with a low surgeon volume. Lack of routine performance of specific operative procedures, as indicated by surgeries that exceed the 75th percentile duration of surgery, may be the main reason for these observed trends in the inverse relation between surgeon volume and the risk of an SSI. Surgeons who rarely perform a certain procedure should be aware that they may impose a greater risk of SSI on their patients.
Acknowledgments
We gratefully acknowledge all participants in the PREZIES network for their contribution to the collection of data. We thank Dr. H. C. Boshuizen and Ir. G. Doornbos for statistical advice, E. Harding for editing the text, and the members of the PREZIES team, in particular Ir. J. Manniën, Ing. C. N. Lau, and Ir. T. van der Kooi for helpful discussions.
Potential conflicts of interest. All authors report no conflicts of interest relevant to this article.
References
- 1. 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; 137:511‐520.
- 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; 349:2117‐2127.
- 3. Morgan M, Black J, Bone F, et al. Clinician‐led surgical site infection surveillance of orthopaedic procedures: a UK multicenter pilot study. J Hosp Infect 2005; 60:201‐212.
- 4. Katz JN, Losina E, Barrett J, et al. Association between hospital and surgeon procedure volume and outcomes of total hip replacement in the United States Medicare population. J Bone Joint Surg Am 2001; 83:1622‐1629.
- 5. Katz JN, Barrett J, Mahomed NN, Baron JA, Wright RJ, Losina E. Association between hospital and surgeon procedure volume and the outcomes of total knee replacement. J Bone Joint Surg Am 2004; 86:1909‐1916.
- 6. Geubbels EL, Wille JC, Nagelkerke NJ, Vandenbroucke‐Grauls CM, Grobbee DE, de Boer AS. Hospital‐related determinants for surgical‐site infection following hip arthroplasty. Infect Control Hosp Epidemiol 2005; 26:435‐441.
- 7. Geubbels EL, Mintjes‐de Groot AJ, van den Berg JM, de Boer AS. An operating surveillance system of surgical‐site infections in The Netherlands: results of the PREZIES national surveillance network. Infect Control Hosp Epidemiol 2000; 21:311‐318.
- 8. PREZIES. Prevention of Nosocomial Infections by Surveillance: Handbook 2004 [in Dutch]. Bilthoven/Utrecht, The Netherlands: RIVM/CBO; 2004.
- 9. Dutch Working Party on Infection Prevention. Registration of Nosocomial Infections. Leiden, The Netherlands: Dutch Working Party on Infection Prevention; 2003.
- 10. Manniën J, Wille JC, Snoeren RL, van den Hof S. Impact of postdischarge surveillance on surgical site infection rates for several surgical procedures: results from the nosocomial surveillance network in The Netherlands. Infect Control Hosp Epidemiol 2006; 27:809‐816.
- 11. Solomon DH, Losina E, Baron JA, et al. Contribution of hospital characteristics to the volume‐outcome relationship: dislocation and infection following total hip replacement surgery. Arthritis Rheum 2002; 46:2436‐2444.
- 12. Shah SN, Wainess RM, Karunakar MA. Hemiarthroplasty for femoral neck fracture in the elderly surgeon and hospital volume‐related outcomes. J Arthroplasty 2005; 20:503‐508.
- 13. Hervey SL, Purves HR, Guller U, Toth AP, Vail TP, Pietrobon R. Provider volume of total knee arthroplasties and patient outcomes in the HCUP: nationwide inpatient sample. J Bone Joint Surg Am 2003; 85:1775‐1783.
- 14. Muilwijk J, Walenkamp GH, Voss A, Wille JC, van den Hof S. Random effect modelling of patient‐related risk factors in orthopaedic procedures: results from the Dutch nosocomial infection surveillance network ‘PREZIES’. J Hosp Infect 2006; 62:319‐326.
- 15. Mangram AJ, Horan TC, Pearson ML, Silver LC, Jarvis WR. Guideline for prevention of surgical site infection, 1999: Centers for Disease Control and Prevention (CDC) Hospital Infection Control Practices Advisory Committee. Am J Infect Control 1999; 27:97‐132.
- 16. Ihse I. The volume‐outcome relationship in cancer surgery: a hard sell. Ann Surg 2003; 238:777‐781.
- 17. Dagan O, Birk E, Katz Y, Gelber O, Vidne B. Relationship between caseload and morbidity and mortality in pediatric cardiac surgery: a four year experience. Isr Med Assoc J 2003; 5:471‐474.
- 18. Farber BF, Kaiser DL, Wenzel RP. Relation between surgical volume and incidence of postoperative wound infection. N Engl J Med 1981; 305:200‐204.