Discussion

The results of our study showed that patients treated in departments with 50 or fewer annual arthroscopies or knee replacements had significantly higher rates of SSI than in departments with >50 and ≤100. Such an effect was not seen for hip replacement, but fewer SSIs were found in departments performing more than 100 hip replacements per year.

We did not expect these clear-cut results. The most important decision-making body of the self-government in Germany, consisting of the national associations of doctors and dentists, the German Hospital Federation and the federal associations of health insurance funds, is the Federal Joint Committee (G-BA). Since 2004, the G-BA set an obligatory minimum quantity of procedures for four invasive procedures (liver-, kidney- and stem cell transplantation and complex oesophagus surgery).20 In 2006, this was expanded to include knee replacement. Hospitals are allowed to perform knee replacement only if they have at least 50 procedures per year. In G-BA studies, the risk-adjusted odds for suffering a SSI or secondary bleeding were highest in the group of hospitals with 50–100 procedures per year and a lower risk of complications only in hospitals which performed more than 200 procedures annually.21 The decision to set the threshold at 50 procedures per year was therefore not based on their study results but was probably based on health-policy considerations. Therefore, we were surprised that our results provided evidence for setting a threshold of at least 50 procedures per year.

Numerous studies reported a relationship between adverse events and hospital volume or size—a finding that is often explained by ‘practice makes perfect.’4 5 10 22 23 Interestingly, this relationship was not linear, but differed in our study by type of procedure, and it was most prominent for arthroscopies. The risk for suffering an SSI in departments with <50 arthroscopies per year was almost nine times higher than in medium-volume departments according to the univariate analysis; it was more than twice as high for knee replacements but was not significant for hip replacements. This possible difference between arthroscopy, knee replacement and hip replacement may be explained by several factors. Arthroscopies are performed increasingly in the outpatient setting, as reflected by the relatively small amount of hospital departments (n=25) providing data. Small samples tend to overweight outliers. More importantly, antibiotic prophylaxis is standard care in joint (hip and knee) implantation but not in arthroscopic knee surgery. In response to a questionnaire of perioperative management on outpatient knee arthroscopy, only 62% of German surgeons reported the routine use of an antibiotic prophylaxis.24 However, antibiotic prophylaxis as an important and not routinely applied confounder was not included in our analysis, and so we urge caution with these results (although significant) for arthroscopy.

With respect to knee replacement, the results of our study might contribute to the existing evidence that in primary knee replacement, a higher hospital volume is indeed associated with better patient outcome with respect to SSI. Moreover, we did not find any indicators that there are small ‘centres of excellence,’ that is, low volume centres with very low SSI rates. A recently published study of German hospitals before and after the implementation of the minimum volume regulation demonstrated a risk reduction of 22.5% for postoperative wound infections.25 Volume–outcome relationships can be explained by more effective routine performance and by the implementation of policies and pathways in medium- or high-volume departments, although surgical practice can vary considerably.26 However, an explanation for the difference by type of procedure (≤50 procedures for knee endoprosthesis and >100 procedures protective for hip replacement) cannot be obtained from these data.

Prior studies of the relationship between hospital surgical volume and adverse events such as SSI have also provided conflicting results.27 Anderson et al called their study ‘Size matters.’22 After adjusting for differences between hospital category (low, medium and high volume of 48 different surgical procedures) and confounders (NNIS risk index), they found that the risk for SSI in low-volume hospitals was 1.5 times higher than medium-volume hospitals. However, it was also 1.3 times higher in large-volume hospitals, leading the authors to conclude that there is a U-shaped relationship between SSI and hospital surgical volume. In a recently published Canadian study, the authors did not see any association between provider volume of total hip or knee replacement and surgical complications (including SSI) during index admission.28 Findings for other outcomes were also mixed. For example, surgeon procedure volume was related to rates of revision of hip replacement but not to rates of revision of knee replacement. Data from the Dutch Network for Prevention of Nosocomial Infections through Surveillance (PREZIES) from 1996 to 2000 revealed that a high annual volume of hip arthroplasty was associated with a reduced risk of SSI.29 However, again, data from PREZIES from 1996 to 2003 did not show that hospital volume was associated with SSI risk (total hip arthroplasty and knee arthroplasty were analysed among others).19

A systematic review of MEDLINE from 1980 to 2000 reported strong volume–outcome relationships for surgeries for pancreatic or oesophageal cancer, and of a much smaller magnitude—although statistically significant—for orthopaedic surgery.9 A systematic review focusing on orthopaedic procedures and reviewing studies from 1966 to 2005 found an association between higher hospital or surgeon volumes and lower rates of hip dislocation.30 Recently, Marlow et al published a systematic review on the relationship between volume and outcome in knee arthroplasty. Their results supported the theory that improved outcomes were associated with increased hospital and surgeon volume. However, the authors stated a North American bias (which limits generalisability) and that the association between outcome and volume appeared tenuous, when results were separated by arthroplasty type (primary, revision).

Generally, volume is only a proxy measure of many parameters because it cannot directly produce favourable or unfavourable results, but as long as specific differences in clinical management or procedural techniques between high-volume providers with good outcomes and low-volume providers with poor ones are not exactly identified, it might be justified from a healthcare-policy perspective to implement minimum regulation volumes.

Our study has several strengths. First, the data result from a large network of hospitals (110 surgical departments provided SSI rates after hip replacement). Surveillance data from Krankenhaus-Infektions-Surveillance-System are representative and validated.31–33 Second, standard Centers for Disease Control-definitions for SSI surveillance were applied across all surgical departments. Third, these data are based on a comparatively long study period of 5.5 years.

Our study also has several limitations. We could not evaluate the volume of individual surgeons or the influence of the surgeon experience. Neither could we analyse other parameters such as antimicrobial prophylaxis, postoperative care or other complications such as bleeding, diabetes, immunosuppression, postoperative pneumonia or thrombosis. Furthermore, we did not perform a systematic postdischarge surveillance, which might explain why SSI rates from the Dutch surveillance system were higher (patients underwent active postdischarge surveillance for 30 days and 1 year).

We conclude that our findings offer some support for recommendations to concentrate on arthroscopy and knee replacement in surgical departments with more than 50 procedures and hip replacement in departments with more than 100 procedures per year in order to reduce SSI.