This study assessed the burden of post-surgical SAI among orthopedic surgery patients in Germany. We identified a similar near term SAI infection risk as a prior study in Germany [5]. In our analysis 0.77% of patients were infected by S. aureus within 90 days of index endoprosthetic surgery (0.46% for knee surgeries; 1.01% for hip surgeries; 0.66% spine surgeries). In previous German studies, 0.98% of patients who underwent orthopedic surgeries acquired SSI, with approximately one-third of SSI caused by S. aureus [5, 9]. As late SSIs (e.g. 180 days, 365-days) were not assessed in these German studies, we cannot directly compare our SSI results [24]. However, our 90-day post-surgery SAI incidence of 0.77% is similar to results from a United States study assessing post elective orthopedic surgery SAI of 0.8% [6]. Our study assessed a longer follow-up period after the index surgery (up to 365 days) in order to capture late deep infections, which has not previously been assessed. To ascertain whether SAIs are related to surgeries other than the index orthopedic surgery, we performed a sensitivity analysis censoring at time of a follow-up surgery not conducted on the same body part and found SAI incidence reduced by 50%. We did not censor for surgeries that were done on the same body part, given that some revision surgeries could have occurred due to a SAI. We recognize using retrospective database did not allow to further determine whether the SAI came from the index surgery or as a result of a follow-up surgery. Therefore, we did not censor for any follow-up surgeries in the risk factor or outcomes analyses. When evaluating the impact of infection risk reduction or prevention methods that could have implications across multiple surgeries within a given time period (e.g. smoking cessation and vaccines) understanding the SAI impact regardless of causative surgery is helpful to assess the potential overall impact of the infection prevention intervention. To further inform this, we assessed the incidence of SAI among those with recent fractures at the location of the index surgery and those without, assuming recent fractures to be most indicative of urgent or emergent surgeries and therefore less likely to allow for implementation of infection prevention activities that require advance time (e.g., smoking cessation, vaccinations).
We identified risk factors for SAI consistent with those reported in the literature for orthopedic SSIs [25]. Older patients and male patients were found to have a higher risk for infection after orthopedic surgery, consistent with previous literature [2, 25, 26]. In addition, the IR of SAI was higher in patients who underwent a hip surgery compared with those patients undergoing a knee surgery, in line with previous studies [2, 26]. Lai et el. found that the presence of more than 2 comorbidities can increase the risk of SSI [27]; our study further supports this by reporting that a higher number of comorbidities (measured as CCI) had a significant impact and increases the SAI risk. Other bacterial infections and longer index hospitalization stay were also found to increase the risk for infections, both in our study and in existing literature [25]. We identified some additional risk factors such as complications with devices/implants during index hospitalization, recent fractures at the location of the index surgery, and number of previous antibiotic prescriptions. Nevertheless, we did not assess the SAI risk associated with other known risk factors, such as obesity, body mass index multiple follow-up hospitalizations or additional surgeries.
In our study, the length of index hospitalization was higher in patients with SAI compared with those without SAI (20.05 vs. 13.95 days). Also, hospitalization days during 365-days follow-up period were substantially higher (76.04 vs. 21.39 days). This is in line with previous observations in Germany, reporting considerable differences in the length of stay between infected and non-infected patients [28, 29]. Compared with previous studies in the United States [3, 30, 31], our hospitalization day results are higher. This reflects differences in the health care practice and reimbursement systems of both states which also account for a longer length of stays in the hospitals in Germany compared to the United States [32]. Our results showed overall unadjusted cost ratio within 365 days after index orthopedic surgery of 3.1 among SAI compared with non-SAI patients (p < 0.001). A slightly higher cost ratio (3.7) was seen in patients who underwent knee arthroplasty in a German hospital setting and experience/did not experience any SSI [29, 30].
The results for 90-day post-surgery mortality (2.79% in the non-S. aureus and 4.90% in the S. aureus group) are comparable to previous studies conducted in the United States, reporting death rates of 1.5–3% for non-SAI patients, 6.7–20.7% for SAI-patients during 90-day follow-up after any surgery [6, 33, 34]. Razavi et al. reported death rates of 6.6%/16.8% within 180 days of orthopedic surgery, which is slightly higher than the reported mortality of 3.9%/11.6% in our study [2]. However, our reported mortality risk ratio within 180 days after any orthopedic surgery (3.05) is slightly higher than that publication (2.56) [2].
This study has some limitations. First, generalizability could be affected by the fact that the health insurance fund only covers patients in two regions of Germany (Saxony and Thuringia). However, since health reimbursement rules are identical across Germany, considerable differences in the treatment of patients are unlikely and therefore results are expected to be generalizable within Germany, but not outside of Germany, by the authors. Given the lack of laboratory data, identification of SAI was limited to documented outpatient and inpatient diagnosis codes (ICD-10), which may have introduced misclassification bias and underestimated the true SAI in the real world. Moreover, underestimation might have happened since patients could have died of SAI after the index hospitalization, without being diagnosed as such.
The index surgery date was not available, therefore the hospital admission date was used as the index date, which could have caused overestimation of the time until infection, and hospitalization days attributable to SAI. Since the specific date when the SAI was identified was not available either, we may have the approximate time to infection in this study, particularly for outpatients who only had data on the quarterly basis on the timing of SAI diagnosis. Although some risk factors for SAI could also be risk factors for dying, death was not included as a covariate in our Cox regressions as a competing risk factor. Ultimately this might have caused overestimation of SAI risk.
Moreover, the correlation between length of hospitalization stay and SAI risk as identified in the analysis might be biased in that SAIs might themselves lead to increased hospitalization time for the index stay, even though they might not have been diagnosed at that time. Finally, when examining the association between index surgery with SAI, any actual association between the index surgery and following SAI is presumed, especially since no specific ICD-10 code classifying an infection as a SSI is available in our dataset. Although an association can be assumed with some degree of certainty for early infections. Infections observed during longer follow-up periods might be attributable to other factors. To account for this, sensitivity analyses were performed for incidence with censoring for follow-up surgeries performed on a different body part from the index surgery. Surgeries on the same part of the body were not censored, as these might have been revision surgeries caused by SAI.