Ninety-Day Follow-up Is Inadequate for Diagnosis of Fracture-related Infections in Patients with Open Fractures

Abstract

Background

Fracture-related infection (FRI) is a challenging complication in musculoskeletal trauma surgery and often complicates the management of open fractures. The CDC currently advocates a surveillance period of 90 days after fracture fixation, but it is unclear what duration of follow-up constitutes adequate surveillance for FRI. Inadequate follow-up will underestimate infections and, in clinical research, will make any interventions studied appear better than they really are, thereby resulting in misleading conclusions.

Questions/purposes

(1) What is the timing of FRI onset in patients with open fractures? (2) What is the proportion of FRIs captured when follow-up is limited to 90 days postoperatively versus when follow-up is extended to 1 year?

Methods

This is a secondary analysis of patient data from a previous retrospective cohort study that investigated whether the duration of perioperative antibiotic prophylaxis was independently associated with FRI in patients with open fractures. Of the 530 eligible patients in the source study, 3% (14) died. Of the remaining 516 patients, 97% (502) patients with 559 long-bone open fractures had 2 years of follow-up constituted the base cohort. Forty-seven fractures in 46 patients were complicated by FRI and were the focus of this secondary analysis. Medical records were reviewed in detail specifically for the current study. Seventy-eight percent (36 of 46) of patients were male, and the mean ± SD age was 42 ± 16 years. The most common mechanism of injury was a motor vehicle accident (63% [29 of 46] of patients), and the tibia was the most involved site (53% [25 of 47] of fractures). The median (interquartile range) time to debridement was 3.0 hours (IQR 2.0 to 4.0). FRIs developed in 3% (7 of 247) of Type I open fractures, 7% (11 of 164) of Type II, 17% (18 of 107) of Type IIIA, 29% (9 of 31) of Type IIIB, and 20% (2 of 10) of Type IIIC open fractures. Each clinic visit of each patient was reviewed, and data about the time of onset of any symptoms and signs suggesting or confirming an FRI, as reported by patients and/or determined by treating surgeons, were recorded. The proportions of FRIs with onset by specific time periods were determined. A Kaplan-Meier survival analysis was performed, and the FRI event rates with 95% confidence intervals were calculated.

Results

The median (IQR) time to the onset of FRI was 52 days (IQR 15 to 153). Follow-up of 90 days captured only 64% (30 of 47) of FRIs, whereas follow-up of 1 year captured 89% (42 of 47) of FRIs. The proportion of FRIs with onset within 1 year increased to 95% (42 of 44) in the presence of an already healed fracture.

Conclusion

Follow-up of 90 days after the management of an open long-bone fracture is inadequate for postoperative surveillance, especially for research purposes. Clinical research on interventions would report results appearing to be much better than they really are, potentially resulting in misleading conclusions. Follow-up of 1 year is preferable because most FRIs will develop before that time, especially when fracture union has occurred. A small percentage of patients may still develop infections beyond the first year after the management of an open fracture. The risk of missing these infections by not extending follow-up beyond 1 year must be balanced against the additional logistical burden. Future prospective multicenter studies and registries with long-term patient follow-up would help clarify this issue.

Level of Evidence Level III, diagnostic study.

Introduction

Fracture-related infection (FRI) can result in nonunion, functional loss, decreased quality of life, and even amputation of the injured extremity [18, 23]. FRI is particularly concerning after open fractures, with the infection risk progressively increasing as injury severity increases. For example, FRI develops after 1% to 5% of Type I and II open injuries and in up to 25% to 30% of Type III open injuries [11, 12, 15, 26, 30]. Considerable research has focused on delineating strategies to prevent this complication, including systemic and local antibiotics, debridement and irrigation, and management of the soft tissue envelope [11, 12, 16, 21, 25, 26, 30, 31]. Evaluation of these strategies requires adequate postoperative surveillance, or follow-up, to capture as many FRIs as possible and to accurately determine the success or failure of these strategies. Previously, the CDC recommended a surveillance period (follow-up) of 1 year for procedures involving the use of implants [13, 14, 17], but this period was subsequently reduced to 90 days in 2013 [2, 3] and remains at 90 days [4].

The impact of this change on the adequacy of surveillance after fracture fixation has rarely been evaluated [10, 27], and it is unclear what duration of follow-up constitutes adequate surveillance for FRI. Greater clarity on these points will inform both clinical practice and clinical research. This is important because inadequate follow-up will have adverse effects in both areas, especially the latter. Specifically, inadequate follow-up in clinical research will underestimate infections and will make any interventions studied appear much better than they really are, thereby resulting in misleading conclusions. In an earlier study on patients with open fractures, we compared patients with postoperative FRIs to patients without FRIs to assess whether the duration of perioperative antibiotic prophylaxis, among other factors, was associated with FRI [8]. We subsequently realized that the almost complete 2-year follow-up of these patients would allow us to assess the time to onset of FRI more accurately than any of the very few studies that have attempted to address this question [10, 27].

Therefore, we performed a secondary analysis of data from the patients of our previous study to answer the following questions, which are separate from the questions of our previous study: (1) What is the timing of FRI onset in patients with open fractures? (2) What is the proportion of FRIs captured when follow-up is limited to 90 days postoperatively versus when follow-up is extended to 1 year?

Patients and Methods

Study Design and Setting

This is a secondary analysis of data from the patients of a previous retrospective study that investigated whether the duration of perioperative antibiotic prophylaxis was independently associated with FRI in open fractures treated at the University Hospitals Leuven, which is an urban, academic, teaching, 1995-bed hospital and designated Level I trauma referral center in Belgium [8].

Patients treated for open fractures from January 2003 to January 2017 were identified by searching the institution’s digital patient file system, and these patients were screened for eligibility. Exclusion criteria were skeletally immature patients (younger than 16 years); pathologic fractures; fractures of the hand, foot, or axial skeleton; and definitive treatment elsewhere.

Baseline Cohort

Five hundred thirty eligible patients with long-bone open fractures were treated during the 14-year study period. A total of 3% of patients died (14; eight from trauma-related causes within 30 days of injury and six within 6 months from cardiovascular disease or cancer). None of the deceased patients had an FRI. Of the remaining 516 patients, 3% (14) were lost to follow-up immediately after treatment for their open fracture; no patient had an FRI at that time. The remaining 97% (502) of patients with 559 long-bone open fractures underwent follow-up with planned visits at 6 weeks, 3 months, and every 3 months thereafter until fracture healing, and then at 12 and 24 months. All 502 patients had 2 years of follow-up and constituted the base cohort.

FRI Cohort

FRIs developed in 46 patients and in 8% (47 of 559) of the long-bone open fractures. Seventy-eight percent (36 of 46) of patients were male, and the mean ± SD age was 42 ± 16 years. The most common mechanism of injury among patients was a motor vehicle accident (63% [29 of 46]). Forty-eight percent (22 of 46) of patients sustained polytrauma (Injury Severity Score ≥ 16), and 35% (16 of 46) of patients were admitted to the intensive care unit. The tibia was the most involved site (53% [25 of 47] of fractures), followed by the femur (23% [11 of 47] of fractures). The median (interquartile range) time to debridement was 3.0 hours (IQR 2.0 to 4.0). FRIs developed in 3% (7 of 247) of Type I open fractures, 7% (11 of 164) of Type II, 17% (18 of 107) of Type IIIA, 29% (9 of 31) of Type IIIB, and 20% (2 of 10) of Type IIIC open fractures.

Current Study Versus the Previous One

The above descriptive data on the 502 base cohort patients and the 46 patients with FRIs were published in our previous study [8] and are the only data that are reported again here because the current study is based on the same patients. We believe that this secondary analysis of data of the same patients was justified by the fact that the time of FRI onset was not assessed in the previous study and the 2-year follow-up of all patients in the base cohort would allow us to address this important issue.

Therefore, the medical records of all patients with FRIs were independently reviewed again by two surgeons (LA, WJM), specifically for the current study. The records of each clinic visit of each patient were reviewed, and detailed data about the time of onset of any symptoms and signs suggesting or confirming an FRI [19], as reported by the patients and/or determined by the treating surgeons, were recorded. Any disagreements between the two surgeons were resolved by evaluation of the data by a third investigator (PD). The data consistency and completeness of all variables were reviewed independently by an external reviewer (CGZ), and the statistical analysis was performed by a biostatistician (AB).

Definitions and Outcome Measures

The diagnosis of FRI was determined based on the confirmatory criteria of the FRI consensus definition [19] , and specifically on the presence of (1) phenotypically indistinguishable pathogens identified by culture from at least two separate deep tissue/implant (including sonication fluid) specimens taken during an operative intervention (in 42 FRIs); (2) fistula, sinus, or wound breakdown with communication to the bone or the implant (in three FRIs); and (3) purulent drainage from the wound or presence of pus during surgery (in two FRIs).

All patients with an FRI underwent surgical management with exposure of the fracture site, and no patient was treated with antibiotics only. Therefore, these infections were not superficial and could be classified as organ or space infections according to the current CDC definition [4].

The time of FRI onset was defined as the interval from the date of the first debridement and fixation of the open fracture to the date of appearance of any of the FRI-suggestive or confirmatory criteria. Presence of a confirmatory criterion (such as fistula) was needed to establish the diagnosis. In the presence of a suggestive criterion only (such as wound erythema or elevated inflammatory markers), the date of appearance of this criterion was considered the date of FRI onset, provided that the diagnosis was established using a subsequent confirmatory criterion (such as a pathogen identified in at least two deep tissue cultures).

In two patients with subclinical FRIs, who underwent surgery to treat a presumably aseptic nonunion and had positive intraoperative culture results, the date the cultures were obtained was used as the date of FRI onset.

Ethical Approval

Ethical approval for this study was obtained from the University Hospitals Leuven, Leuven, Belgium (number S58022).

Statistical Analysis

Continuous data are expressed as the mean ± SD or median (IQR), depending on adherence to a normal distribution. Categorical data are presented as percentages with counts. We performed a Kaplan-Meier survival analysis, and the FRI event rates with 95% confidence intervals were calculated for specific time points using Kaplan-Meier methods. Patients with more than one fracture were included in our analysis; therefore, the data are clustered. Since this paper focuses solely on the estimation of FRI event rates and these are not affected by clustering of the data, we deemed it appropriate not to adjust our analyses for clustering. The proportions of FRIs with onset within specific time periods were calculated for the entire cohort. We did not analyze our results by sex (or by any patient, injury, or treatment factors) because the number of FRIs in our study would not allow for a meaningful multivariate analysis to investigate whether such factors are independently associated with time to onset of infection.

Results

Timing of FRI Onset in Patients with Open Fractures

Among the patients in our study who had FRIs, the median (IQR) time to onset was 52 days (15 to 153) (Fig. 1).

Fig. 1.

This Kaplan-Meier curve represents the time to FRI after open fracture fixation.

Proportion of FRIs at 90 Days Versus 1 Year

Sixty-four percent (30 of 47) of FRIs were captured by 90 days of follow-up, compared with 89% (42 of 47) of FRIs captured by 1 year of follow-up (Fig. 2). A large proportion (36% [17 of 47]) of postoperative FRIs would be missed with a shortened surveillance period limited to 90 days, as advocated by the CDC after the 2013 guideline revision [2, 3].

Fig. 2.

This bar graph shows the proportions of FRIs captured, divided into specific follow-up periods.

A surveillance period of 90 days would result in an FRI event rate of 5.4% (95% CI 3.8% to 7.6%), whereas surveillance for 1 year postoperatively would yield an FRI event rate of 7.5% (95% CI 5.6% to 10.0%). Overall, 11% (5 of 47) of infections developed more than 1 year postoperatively. Surveillance for 2 years postoperatively demonstrated an FRI event rate of 8.2% (95% CI 6.2% to 10.8%).

Interestingly, 3 of 5 infections with onset beyond 1 year were associated with nonunion. Only two infections developed beyond 1 year in fractures healed by that time; therefore, follow-up of 1 year captured 95% (42 of 44) of infections in the presence of an already healed fracture compared with 89% (42 of 47) overall. Follow-up of 9 months captured 93% (40 of 43) of infections in the presence of an already healed fracture, compared with 85% (40 of 47) overall.

Discussion

The appropriate postoperative follow-up time to assess the development of surgical site infection remains controversial. A minimum follow-up time of 2 years has been proposed for evaluating arthroplasty to treat periprosthetic joint infection [9]. A 1-year follow-up period for infection surveillance after fracture fixation was recommended by experts [20] and Clinical Orthopaedics and Related Research® [5], and it was advocated by the CDC in the pre-2013 surveillance guidelines [13, 14, 17]. The revised CDC guidelines limited the surveillance time to 90 days, even when an implant is used, as is the case in fracture fixation and arthroplasty [2-4]. As a result, studies reporting on postoperative infection with a surveillance time of only 90 days are being widely published. Clarifying which duration of follow-up constitutes adequate infection surveillance would have important implications regarding prognosis, patient care, and reporting of research findings. We found that a surveillance period of 90 days after the management of an open long-bone fracture captured only 64% (30 of 47) of infections in this study’s analysis. In contrast, extending follow-up to 1 year would help diagnose 89% (42 of 47) of infections overall and 95% (42 of 44) of infections if the fracture has already united by that time.

Limitations

This retrospective study has several limitations. First, precisely determining the exact time at which FRI clinically occurs depends on the patient’s accurate recollection and a detailed history-taking by the provider. It is possible that some patients did not come back promptly after symptoms developed or did not remember the exact date of symptom appearance; this may have resulted in imprecise documentation of symptom onset and perhaps some delay relative to the actual onset. However, we would not expect this delay to be large because our follow-up period was strict, and it was based on direct communication with general practitioners. Prospective studies with close monitoring of all patients, even between the traditionally scheduled clinic visits, would improve the accuracy of determining the onset of symptoms, but such studies are not available now and will not be easy to perform.

Second, patients with subclinical FRIs (those with nonunions presumed to be aseptic with positive intraoperative culture results) present a unique challenge [1, 22]. The diagnosis of FRI can only be determined intraoperatively in these patients, and the exact onset of infection is often unclear. Moreover, the timing of surgical management of a presumably aseptic nonunion (and therefore the time of establishing the diagnosis) may vary among surgeons and institutions, leading to a delay in diagnosis. However, only 4% (2 of 47) of the infections in our series were subclinical; therefore, we are confident they had a minor, if any, effect on our findings. Furthermore, we believe these subclinical infections should still be included in the analysis despite the aforementioned limitation. Otherwise, a relatively short follow-up period, especially before the fracture has healed, would miss these infections; this would lead to underestimating the infection rate and overestimating the proportion of infections identified within this period and the adequacy of follow-up.

Third, our study is limited by the small number of FRIs, which does not allow us to perform a meaningful multivariate analysis to investigate whether time to infection onset might be associated with certain factors; such factors might involve the patient (such as sex, age, or comorbidities), the injury (anatomic location and severity of open fracture, and the presence or absence of polytrauma), the treatment (type and duration of antibiotic prophylaxis, time to debridement, or fixation method), or the pathogen type. Also, the small number of FRIs does not allow us to explore the potential association of the progression of fracture healing with the onset of infection. However, exploring these associations was beyond the scope of our study. We aimed to assess the proportion of FRIs captured when follow-up is 90 days versus 1 year, and we were able to demonstrate an important difference between these time points.

Lastly, our reported time to onset of infection after the treatment of open fractures may not apply to other clinical scenarios, such as the fixation of closed fractures or total joint arthroplasty. Further studies are needed to clarify the time to development of infection following these procedures.

Timing of FRI Onset in Patients with Open Fractures

Extremely limited information is available on the time to the onset of infections in patients with open fractures. A study of 214 FRIs (47% of which developed after the management of open fractures) reported that 50% of all infections presented within 35 days [29]. An important limitation of that study is that the authors did not provide data on the completeness and length of follow-up of the entire cohort of patients. A recently published study reported a median time to diagnosis of deep infection after the treatment of open fractures of 53 days [27], which is similar to our median time of 52 days. However, the upper bound of their IQR was much shorter than ours (119 days versus 153 days). The shorter 75th percentile reported in that study might be attributed to the fact that they followed 90% of their cohort for only 1 year, even though FRIs might still develop beyond that period. This difference suggests that data on the timing of infections and the adequacy of surveillance depend on the completeness and length of follow-up of the underlying study; these are important considerations when assessing the data and recommendations of a study dealing with this topic.

In contrast to the previous two studies [27, 29], our study had an extended follow-up of 2 years for almost all (97%) patients. This increases confidence in our findings compared with the two previous studies.

Proportion of FRIs at 90 Days Versus 1 Year

In 2013, the CDC shortened the minimum necessary surveillance time for operative fixation of fractures and any procedures involving implant insertion from 1 year to 90 days [2, 3]. This decision aimed to reduce the burden of collecting additional data because 90 days of follow-up was deemed adequate. However, the impact of this decision has been assessed infrequently. Our study convincingly demonstrated that 90 days of follow-up is inadequate for surveillance after fixation of open fractures, with 36% of infections developing after this period. We want to emphasize that our findings apply only to fixation of open fractures and not to other orthopaedic procedures. One study found that 21% of FRIs after open reduction and internal fixation of fractures would be missed during the shortened 90-day period [10], and another study reported that 32% of deep infections after the treatment of open fractures were diagnosed between 90 days and 1 year [27]. We reported a higher proportion of missed infections when follow-up was limited to 90 days than the previous two studies. This may be attributed to the longer and more complete follow-up in our study. Dicks et al. [10] based their analysis on follow-up of 1 year, but they did not report the proportion of patients who did not achieve 1 year of follow-up. Prada et al. [27] described a follow-up period of 1 year in 90% of their patients. In contrast, our study had extended follow-up of 2 years for almost all patients.

Furthermore, our study used time to infection onset as the main outcome variable instead of time to confirmation of diagnosis. Our approach accounts for delayed patient presentation or delayed diagnosis of an FRI when findings are initially attributed to a superficial infection instead and treated with antibiotics only. Indeed, differentiating infections based on the depth of involvement has inherent problems [19, 28]. The depth of bacterial colonization can only be assessed by opening the wound and taking samples for culture from different tissue layers. When we assume that the infection involves the incision only, we may miss the diagnosis of FRI and delay surgical treatment, especially when the fracture is subcutaneously located, as in ankle fractures.

The inadequacy of the 90-day surveillance period for open fractures is underscored by the fact that most open tibia fractures will not have healed by this timepoint [7] and subclinical infections may present as nonunions [1, 22]. The median time to union of open tibia fractures, which comprise most open long-bone fractures [6], ranges from 18 weeks for Type I open fractures to 41 weeks for Type IIIB injuries [7]. Therefore, a low-grade infection resulting in nonunion is unlikely to be captured during the 90-day surveillance period proposed by the CDC. For the above-mentioned reasons, the first FRI consensus definition did not dictate a timepoint after which follow-up would no longer be necessary [19].

Our study findings have important implications for clinicians, clinician scientists, and policymakers. For the clinical care of patients with open fractures, achieving long-term follow-up may not be required, especially if the fracture has healed and the patient does not have symptoms. A patient who develops symptoms but does not return to us for follow-up would likely get care elsewhere. Nevertheless, we should counsel the patient that an infection may present even after 1 year, explain the symptoms and signs of this complication, and instruct the patient to seek care should this occur.

For clinical research, however, adequate follow-up is important. Without it, a considerable proportion of complications would be missed, their prevalence would be underestimated, and clinical research on interventions would report results appearing to be much better than they really are, potentially resulting in misleading conclusions. This underscores the importance of having a minimum follow-up requirement when reporting data on adverse outcomes such as FRI. Adequate follow-up becomes even more important given the fragility of published orthopaedic trauma trials, with only a few event reversals required to alter trial significance. In studies reporting dichotomous categorical outcomes, if only 4 of 100 patients experienced the alternative outcome, a statistically significant result would become nonsignificant and vice versa [24]. On the other hand, obtaining long follow-up is labor-intensive, and requiring such a follow-up period would exclude a considerable number of patients, thereby limiting the sample size available for analysis. Therefore, the minimum required follow-up should be able to capture most events of interest, while also avoiding any low-yield logistical burden. To make this determination, it is essential to know the proportion of FRIs that develop during distinct time periods postoperatively, and our study offers valuable information in this regard. We recommend follow-up of 1 year because most FRIs will develop before that time, especially when fracture union has occurred. A small percentage of patients may still develop infections beyond that time, but the risk of missing these infections is the tradeoff for avoiding the additional logistical burden of longer follow-up. Nevertheless, confirmation of our findings in larger multicenter studies would be necessary.

For policy, we believe that the current 90-day follow-up does not provide adequate postoperative surveillance for infection in patients with open fractures. We understand that the CDC has the difficult task of creating surveillance guidelines that apply across all surgical disciplines and all types of procedures. We believe that there is a need for dedicated guidelines for FRIs that consider the specific characteristics of these infections. This has already been done by the CDC for periprosthetic joint infections [4], and we believe it would be very useful to also have specific definition and surveillance guidelines for FRIs.

Conclusion

Follow-up of 90 days after the management of an open long-bone fracture is inadequate for postoperative infection surveillance, especially for research purposes. Follow-up of 1 year allows the diagnosis of 89% of infections overall and 95% of infections if the fracture has already united by that time. A small percentage of patients may still develop infections later, but the risk of missing these infections by not extending follow-up beyond 1 year must be balanced against the additional logistical burden. Future prospective multicenter studies and registries with long-term patient follow-up would be required to clarify this issue. Until then, we recommend a 1-year follow-up period for open long-bone fractures.