Surgical site infection (SSI) after hip replacement is a major complicationReference Le Manach, Collins and Bhandari ¹ targeted for epidemiologic surveillance.Reference Le Manach, Collins and Bhandari ^{1 –} Reference Zimmerli ⁴ Measuring and reducing the risk of SSI through surveillance improvement is a key target of the French program for prevention of healthcare-associated infection.Reference Carlet, Astagneau and Brun-Buisson ^{5 ,} Reference Astagneau, L’Hériteau and Daniel ⁶

Hip arthroplasty after fracture or trauma is associated with an increased SSI incidence, especially when practicing partial hip replacement (PHR). Moreover, this surgery has been reported to have higher crude in-hospital mortality compared with patients undergoing a total hip replacement (THR).Reference Le Manach, Collins and Bhandari ¹ However, case fatality could vary according to several patient underlying conditions often associated with such fracture. In the scope of epidemiologic surveillance aiming for hospital benchmarks, it is essential to account for differentiating patient characteristics that might affect the probability of infection. Recent studies worldwide reported numerous patient risk factors for SSI, including obesity, malnutrition, diabetes mellitus, and blood transfusion.Reference Zhu, Zhang, Chen, Liu, Zhang and Zhang ⁷ Recent studies have demonstrated the usefulness of a hospital information system for SSI surveillance, especially improvement of postdischarge surveillance.Reference Grammatico-Guillon, Baron and Gettner ^{2 ,} Reference Grammatico-Guillon, Baron and Rosset ^{3 ,} Reference Grammatico-Guillon, Baron, Gaborit, Rusch and Astagneau ⁸ Using the French hospital discharge database, this study aimed to determine SSI outcome and mortality after primary THR compared with PHR after adjusting for confounding factors.

The national hospital discharge database provides computerized SSI detection in an affordable data-reporting system allowing robust analyses.Reference Grammatico-Guillon, Baron and Gettner ^{2 ,} Reference Grammatico-Guillon, Baron and Rosset ^{3 ,} Reference Grammatico-Guillon, Baron, Gaborit, Rusch and Astagneau ⁸ A previous study demonstrated the potential of hospital discharge algorithms for SSI detection after hip arthroplasty with an acceptable performance (positive predictive value, 87%, without difference between PHR or THR).Reference Grammatico-Guillon, Baron, Gaborit, Rusch and Astagneau ⁸ Data on French public and private hospitals were extracted from the hospital discharge database during the 2008–2012 period (Appendix). Overall 476,778 patients undergoing hip replacement were selected and followed up, including 371,889 (78%) receiving THR. Comorbidities, used as confounding factors, were extracted using International Classification of Disease, Tenth Revision, coding algorithms (ie, the specifications to identify each condition) from significant associated diagnoses (Appendix). A time-dependent model was calculated using the Kaplan-Meier method and Cox regression to determine the effects of different confounding factors on the SSI and mortality risk comparing THR and PHR, first evaluated in a bivariate analysis and included in the multivariate model if P<.2. We checked proportionality of hazards and log-rank test by SAS, version 9.1 (SAS Institute).

During the study period, primary hip replacement was performed mainly in private hospitals (60%). Patients were mainly female (74.9%), aged more than 65 years. The median (interquartile range) age at replacement was 84 (10) years. The main reasons for hip replacement were degenerative osteoarthritis (85.9%) and fracture (8.8%). THR was mostly performed for osteoarthritis (89%), whereas 90% of PHR occurred after hip fracture. PHR was mostly performed in older women after hip fracture. Over the period, the number of arthroplasties per year was stable. Most patients were readmitted to the hospital after the arthroplasty stay, 24.4% once and 40.8% at least twice (mean follow-up, 381 days). A total of 30,846 (8%) of the 371,889 patients had another primary arthroplasty during their follow-up, corresponding to the contralateral hip replacement in more than 80% of these cases. In-hospital case fatality during follow-up after hip replacement was 3.7%.

One-year SSI incidence was 1.2% after THR vs 2.2% after PHR. The median age at infection was significantly higher in PHR than THR (82 vs 71 years, P<.001).

The Kaplan-Meier curves for SSI occurrence showed a higher and early risk of SSI after PHR over time, and the same result was found for mortality after hip replacement (Figure 1).

FIGURE 1 Annual risk of surgical site infection and death after hip arthroplasty in French 5-year cohort of hip prosthesis.

In multivariate analysis, PHR was associated with a higher risk of SSI than THR (hazard ratio, 1.29 [95% CI, 1.22–1.36]) after adjusting for risk factors of age greater than 65 years (1.17 [1.09–1.21]), body mass index (calculated as weight in kilograms divided by height in meters squared) greater than 40 (1.20 [1.08–1.32]), urinary disorders (1.30 [1.23–1.39]), renal failure (1.17 [1.09–1.25]), malnutrition (1.17 [1.09–1.29]), and decubitus ulcer (1.11 [1.03–1.19]), whereas gender, diabetes mellitus, and heart diseases were not associated with SSI.

Overall crude mortality rate was 7% and was significantly lower in THR than PHR (4% vs 7%, P<.01). In the multivariate analysis, PHR was associated with a higher risk of mortality (hazard ratio, 2.13 [95% CI, 2.08–2.19]) as well as SSI (1.22 [1.15–1.29]).

As recently checked in Canada, medical chart review for cases identified through administrative data is an efficient supplemental SSI surveillance strategy.Reference Rusk, Bush and Brandt ⁹ It improves case-finding by increasing SSI identification, and it identifies SSIs presenting at nonprocedure facilities.Reference Rusk, Bush and Brandt ⁹

At first, on the basis of a nationwide hospital information system of more than 400,000 procedures, our study confirmed the significant increase of crude SSI incidence previously reported in a regional subset during the same study period.Reference Grammatico-Guillon, Baron and Rosset ³ An association of SSI coding with new hospital financial incentives has been suspected since the labeling of French medical centers in 2009 for rates of complicated bone and joint infection,Reference Grammatico-Guillon, Baron and Gettner ² but this trend does exist in other countries, making this argument insufficient to explain such an increase.Reference Dale, Fenstad and Hallan ¹⁰ Hence, the hospital discharge database model has limitations, particularly regarding the quality of the data coded. However, the routine definition of SSI cases after hip arthroplasty has been assessed with reliable performance parameters. The strength of this work is the use of a national database, avoiding underestimation of SSI incidence by limiting postdischarge surveillance to the operating hospital.

We demonstrated that infection and mortality risk were higher after PHR compared with THR and after adjusting for confounders. In-hospital mortality associated with PHR was also higher in SSI than non-SSI patients. As recently exposed,Reference Le Manach, Collins and Bhandari ¹ patients needing hip fracture surgery mainly underwent PHR, favoring SSI and death. However, even if patient conditions represent important independent factors of SSI and death, PHR remains an important independent factor of SSI and mortality.