1 Introduction

Pediatric septic arthritis is a relatively uncommon disease that occurs in children, with an estimated incidence of 1–10 per 100,000 children.^1–7 The most common sites of pediatric septic arthritis are large joints in the lower extremities, especially of the knee and the hip.^2,8 In comparison to adult septic arthritis, this disease process is more common and can have much more devastating consequences, leading to sepsis, more serious functional consequences, and even death.^9,10 Additionally, infections can be made more severe by the presence of adjacent infection loci, including periarticular infections (PAI).¹¹ This can include a range of pathologies including osteomyelitis with or without subperiosteal abscess and infectious myositis.

The criteria set forth by Kocher et al. are used frequently for decision making about further treatment options for pediatric hip patients.¹² However, this cannot be applied to all joints uniformly as Obey et al. found that the Kocher criteria predictors, which specifically target septic arthritis of the hip, did not apply to septic arthritis of the knee and would result in up to 52% risk of missing septic knee if applied as a screening tool.¹³ More recently, studies have observed the prevalence and criteria for diagnosing septic arthritis with adjacent infection in children as this can impact both the diagnostic and treatment pathways, specifically including magnetic resonance imaging (MRI) prior to intervention.^14–16 Elucidating new criteria may be beneficial to decide which children should obtain an MRI prior to surgical intervention in order to ensure adjacent infections are not missed during the initial surgery. The ability to discern which patients are most likely to benefit from a preoperative MRI could prevent unnecessarily costly studies as well as decrease the need for additional surgical interventions.

Rosenfeld et al. found that the following criteria were predictive of septic arthritis with adjacent infection in their patient population: age older than 4 years, C-reactive protein (CRP) level greater than 8.9 mg/L, duration of symptoms (including fever, effusion, or impaired weightbearing) more than 3 days, platelet count less than 310 × 10³ cells/μL, and absolute neutrophil count greater than 7.2 × 10³ cells/μL.¹⁷ For patients with ≥3 of these factors, the study found that the sensitivity and specificity to be 90% (95% CI, 0.78, 0.96) and 67% (95% CI, 0.49, 0.81), respectively. The positive predictive value (PPV) was 80% (95% CI, 0.66, 0.88), and negative predictive value (NPV) was 83% (95% CI, 0.64, 0.93). These criteria are intended to help physicians decide whether to obtain an MRI to identify possible additional foci prior to septic joint irrigation and debridement. However, the external validity of these criteria remains to be seen, as this study was performed in Texas in the Southwestern United States, and the prevalence of different bacteria or different antibiotic susceptibilities varies based upon geography.^5–7

One study by Refakis et al. applied the Rosenfeld criteria to their own patient population retrospectively in a geographically distinct location, the Northeastern United States in Pennsylvania.¹⁸ This study found that the Rosenfeld criteria had lower sensitivity (86%) and specificity (54%) as well as an elevated false positive rate (50%) when compared to the findings in Texas of 90%, 67%, and 33%, respectively. The results demonstrated that the criteria were not as reliable for their geographical area. As a response, Welling et al. applied the Rosenfeld criteria to a new population from the same geographical area as the original paper in Texas and found similar results to their first study, with a sensitivity of 86%, specificity of 85%, PPV of 91%, and NPV of 77%.¹⁹ This evidence seemingly suggests that the original criteria are valid for the Southwestern United States but not as helpful for the Northeastern United States.

The purpose of this study is to apply the Rosenfeld criteria to a Southeastern United States pediatric population to verify the external validity of these criteria in a geographically distinct area from those that have previously been studied.

2 Materials and methods

After receiving IRB approval, a retrospective chart review of 185 patients between August 2009 and December 2018 was performed using International Classification of Diseases (ICD)-9 and ICD-10 codes of children treated at a tertiary referral center located in the Southeastern United States. The ICD-9 codes were 711.0–711.9, and the ICD-10 codes were M00-02. Inclusion criteria included septic arthritis in any appendicular joint in a child between the ages of one and 18.

Patients' medical records that met inclusion criteria were queried for demographic information, including sex, current age, age at diagnosis, birth date, race, height, weight, and body mass index (BMI). Additionally, the following information about the presentation was recorded: date of diagnosis, affected joint, infection type, symptom duration, weight bearing status, concomitant diagnoses, history of diabetes, history of immunosuppression, and receipt of antibiotics prior to intervention. Prior to definitive intervention, imaging studies (x-ray and MRI), lab values (leukocytes, platelets, hematocrit, absolute neutrophil count, percent segmented neutrophils, erythrocyte sedimentation rate (ESR), and CRP), vital signs (temperature, heart rate, and blood pressure), and pathologic specimens (blood culture and joint aspiration cell count and cultures) were collected. The CRP value is recorded on a milligram per deciliter (mg/dL) scale at this institution. Information about intervention and clinical progression, including operative treatment, associated complications, repeat surgeries, antibiotic use, antibiotic complications, follow-up timeline, and clinical status at follow-up were also recorded. All MRIs were performed at the authors' institution, often under the same anesthesia as the operative case.

3 Results

A total of 185 patient charts were reviewed. Fifty-eight patients had suspected or confirmed acute hematogenous musculoskeletal infection. A total of 24 patients were excluded, leaving 23 patients with isolated septic arthritis (ISA) and 11 patients with septic arthritis with adjacent periarticular infection (PAI) for a total of 34 patients. Eight patients were excluded for being less than one year of age, as this population has different levels of normal accepted lab values and vital signs. Other patients were excluded for: sickle cell disease (1), metastatic rhabdomyosarcoma (1), osteomyelitis and/or septic arthritis of a hand (2), septic arthritis of the sacroiliac joint (2), or lack of diagnostic criteria for septic arthritis as specified by Rosenfeld et al. (10), meaning either an aspiration with cell count >50,000, positive Gram stain, or positive culture (Fig. 1).

Close

Fig. 1 Breakdown of included and excluded patients in the study. ICD indicates International Classification of Diseases; ISA, isolated septic arthritis; PAI, periarticular infection; SCD, sickle cell disease; SI, sacroiliac.

Export to PPT

The patients' ages ranged from 22 months to 11 years old (mean of 6.31 ± 2.97 years) (Table 1). The joints involved were: 14 knees, 8 hips, 6 ankles, 4 elbows, 1 shoulder, and 1 wrist (Table 2). The average follow-up in weeks was 12.45 (range 2–80 weeks). One patient had a very complicated course, including recurrent infection of the hip with pathologic slipped capital femoral epiphysis (SCFE) requiring pinning procedure with Spica cast, which then recurred after pin removal. The rate of patients with adjacent PAI was 32.3% in our population. Of the PAI group, osteomyelitis was present in all 11 patients. A subperiosteal abscess was present in 4 patients, and pyomyositis was present in 5 patients. Of all cases studied, intraoperative cultures were positive in 29 of the cases (85.3%), with the other 5 patients having aspirate cell count >50,000 despite negative cultures (Table 3). Blood cultures were only positive in 18 patients (52.9%) but were more likely to be positive in patients with PAI (10 out of 11, p < 0.001). All patients underwent at least one operative intervention. Five of the 11 patients with PAI required a second surgical intervention (45.5%) while 7 of the 23 patients (30.4%) with ISA underwent secondary procedure (p = 0.39). MRIs were performed in 27 of the 34 patients (76.4%), 19 of which (70.3%) were completed preoperatively. Patients with ISA completed preoperative MRIs in 11 out of 15 cases (73.3%), whereas only 7 out of 11 MRIs (63.6%) were performed preoperatively in patients with septic arthritis in addition to adjacent infection. Of the 18 patients who received preoperative MRIs, 5 patients required more than 1 surgical intervention, including 3 (27.3%) ISA patients and 2 (28.6%) PAI patients, respectively.

The sensitivity, specificity, PPV, and NPV of the Rosenfeld criteria when applied to our patient population were: 91.7%, 22.7%, 39.3%, and 83.3%, respectively (Table 4). Accordingly, if these criteria were used to make the decision to obtain a preoperative MRI, a beneficial preoperative MRI would not have been obtained in 8.3% of our patients using the calculation 1-sensitivity. Likewise, an unnecessary preoperative MRI would have been obtained in 60.7% of our patients using the calculation 1-PPV. Our patient population demonstrated very few patients that met all four of Rosenfeld criteria (Table 5).

4 Discussion

Periarticular adjacent infections in pediatric patients presenting with septic arthritis were identified in 32.3% of patients treated in the Southeastern United States. These patients were more likely to present with positive blood cultures. This could be due to the multiple foci of infection generally due to a more virulent organism.²⁰ This information, however, does not assist with timely diagnostic decision-making given blood cultures may not result for up to 48–72 h after presentation. The retrospective application of the Rosenfeld criteria demonstrated a sensitivity of 91.7% for identifying a PAI, however, the specificity was only 22.7%, which would have resulted in 8.3% of patients with PAI being undiagnosed. Additionally, most of the MRIs would be unnecessary and increase cost to the healthcare system without apparent benefit.

As with any expensive intervention in medicine, the cost versus benefit ratio must be considered. MRIs can add significant financial costs and can delay treatment if done preoperatively, which can result in further irreversible damage to joint cartilage.^21–23 At some institutions, this may also mean a second intubation for the patient in a short timespan if the institution is unable to guarantee the MRI under the same anesthesia as the operation. One potential benefit of a preoperative MRI is that it may also identify additional foci of infection, preventing a second surgery. If foci are missed, it can increase hospitalization time, overall treatment duration, and likelihood of persistent sepsis.^24,25 Griswold et al. demonstrated that implementing a preoperative MRI protocol may reduce the consequences of missed foci, by decreasing the rate of additional operative intervention in pediatric infections.²⁶ However, the patients observed in this study required reoperation in nearly one quarter of patients who underwent preoperative MRI. Additional procedures were required at similar rates in both ISA and PAI despite an attempt to identify adjacent disease early.

Ideally, certain criteria, such as those set forth by Rosenfeld et al., would prove to be universal and aid in selecting the appropriate patient for preoperative MRI. This would, in turn, decrease the need for additional operative intervention, possibly expedite infection eradication, and improve long term outcomes as cartilage can start to see significant detrimental impact as early as 6–8 h from symptom onset.²⁷ In addition, although MRI adds cost to the initial work-up, prolonged hospital stay and secondary procedure also increase costs significantly, so the benefits of both options must be weighed. The consequence of a missed PAI may not necessarily cause significant harm to the patient, as delayed treatment of osteomyelitis and myositis may not cause lasting long-term damage to the patient's limb function like septic arthritis. As long as septic arthritis is diagnosed and treated in a timely manner and the patient is not overtly septic, the main long-term joint and limb sequela may be avoided, even if there is a delay in treatment of the adjacent foci.

The Rosenfeld criteria, when applied to their own study population, demonstrated a similar sensitivity and NPV, but a higher specificity and PPV compared to this population. The original Rosenfeld paper included all appendicular joints and excluded those younger than one year of age. Likewise, we excluded patients less than one year of age and included all appendicular joints. They had higher number of patients studied at 36 with ISA and 51 with adjacent PAI. Interestingly, Rosenfeld et al. demonstrated a higher prevalence of adjacent PAI (59%) compared to the current study (32.3%), which may be explained by geographical differences in pathogens.^20,28 Increased prevalence also increases the PPV, which may help to explain some of the difference in PPV between our studies. However, this should also decrease the NPV, but our NPVs were roughly the same. Another major difference is that all of their patients underwent MRI, either preoperatively or postoperatively. Only 27 of the 34 of our included patients (76.4%) underwent MRI at any point during their care, and of those patients, 19 (70.3%) completed the MRI preoperatively. Patients with isolated septic arthritis completed preoperative MRIs in 11 out of 15 cases (73.3%), whereas only 8 out of 12 MRIs (66.7%) were performed preoperatively in patients with septic arthritis in addition to adjacent infection. Likewise, Welling et al. found either similar or even improved statistical values to Rosenfeld with sensitivity, specificity, PPV, and NPV of 86%, 85%, 91%, and 77% respectively, validating these criteria for the Southwestern United States population in a separate patient population from the original study.¹⁹ Similarly, Welling excluded patients with confounding concurrent illnesses, age less than one, and those without MRI. All appendicular joints were included.

In contrast, Refakis et al. found reduced sensitivity (86%), specificity (54%), and a higher false positive rate (50%) compared to Rosenfeld et al. (90%, 67%, 33%, respectively). Additionally, Refakis et al.’s rate of PAI (35%) was lower than that of Rosenfeld et al. (59%) and more similar to the current study (32.3%). The work by Refakis et al. differs from the initial study in that the patient population was located in the Northeastern United States and only evaluated septic hip arthritis. The inclusion of only septic hips may account for some of the differences found between the two studies. Geographic differences in infectious pathogens is not a new concept as it has been well established that bacterial epidemiology varies from region to region and can change with time.^29–31 The data presented in this study may be applicable to current authors' institution but not to another in a different region.

Similar to the current study, Refakis et al. did not have MRI on all patients and instead decided to obtain an MRI on an individual basis. They only performed an MRI on 43% of their patients, with the majority being obtained postoperatively compared to the 76.4% of patients in the current study. They noted that though they had no proof that the population without MRI did not have PAI, but that it was a relatively safe assumption they had ISA only given that those patients improved with surgical treatment and did not decompensate and require further surgical treatment or MRI.

Gottschalk et al. reviewed retrospective data of their own institution to determine which patients with septic hip arthritis underwent preoperative or pre-aspiration advanced imaging (MRI and/or bone scan) vs. no preprocedural advanced imaging and determined they were more biased toward getting advanced imaging in younger patients and those that were unable to bear weight. After adjusting for age, they found the reoperation rate to be 2.8 times more likely in those who did not undergo preprocedural advanced imaging (30% vs. 16.8%, p = 0.03), though they did not comment on how this may have prolonged their length of stay or any cost comparison analysis. Interestingly, advanced imaging saved 9 out of 77 patients evaluated (12%) from undergoing an unnecessary surgery as they ruled out septic arthritis of the hip.³² Perhaps more institutional reviews of their own practices could help calculate the cost-to-benefit ratio of certain diagnostic and treatment practices.

5 Conclusion

The use of protocol-based algorithms and predictive criteria for determining the necessity of advanced imaging in pediatric septic arthritis has been highly sought, although it has proven difficult to develop a generic approach that extrapolates to various geographic regions. The application of the Rosenfeld criteria was found to be non-specific in determining the presence of an adjacent infection for pediatric patients with septic arthritis treated in the Southeastern United States. Caution should be used with the blanket application of these criteria without augmentation of local factors and clinical judgment to aide in treatment decision-making. The authors recommend using a protocol specific to each institution or region and consider the specific local antibiogram and the implication of more virulent pathogens on the need to obtain advanced imaging in pediatric septic arthritis.

Declaration of interest

None.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.