Age and C-reactive protein as modifiers of Kocher criteria in pediatric septic knee monoarthritis

a b s t r a c t

Introduction: Although Kocher criteria can distinguish a septic hip from an aseptic cause, they may not apply to a septic knee. We aimed to identify predictors to discriminate septic and aseptic causes of acute knee monoarthritis in children who underwent arthrocentesis.

Methods: We conducted a retrospective cohort study among children who underwent arthrocentesis for suspected Septic arthritis of the knee. Collected data included demographic, clinical and laboratory characteris- tics. We performed univariate and multivariable analyses to identify predictors of the septic knee. We further investigated accuracy of different predictive models.

Results: A total of 60 patients who underwent arthrocentesis for suspected knee septic arthritis were included in this study. Septic arthritis of the knee was confirmed in 32 (53%) patients. Age <= 5 years (OR 4.237, [95% CI 1.270-14.127], p = 0.019), WBC > 12,000 cells/mm3 (OR 5.059, [95% CI 1.424-17.970], p = 0.012), and CRP >

2 mg/dL (OR 3.180, [0.895-11.298], p = 0.074) were the most important predictors of a septic knee. Three-tier model comprising these three factors (AUC 0.766) and 4-tier model with addition of fever >38.5?C (AUC 0.776) performed better than Kocher criteria (AUC 0.677), modified Kocher criteria (AUC 0.699) and Full Model (adding age <= 5 years and CRP >2 mg/dL to Kocher criteria) (AUC 0.746). Full Model successfully ruled out septic arthritis if all 6 criteria were negative.

Conclusion: Based on these findings, we propose an algorithm to identify low, intermediate and high-risk patients for knee septic arthritis. Our proposed two-step algorithm incorporating major (age, WBC, CRP) and minor (fever, ESR, non-weight bearing) criteria can serve as a simple decision-support tool to justify arthrocentesis in children with suspected knee septic arthritis.

(C) 2022

  1. Introduction

Septic arthritis is the most concerning diagnosis when a patient pre- sents to the emergency department (ED) with acute monoarthritis. It is rare with an incidence of 4-37 cases per 100,000 children annually [1,2]. Prompt diagnosis and appropriate treatment of septic arthritis is very important, since delays may result in significant disability and even death [3,4].

Abbreviations: AUC, Area under curve; ANC, Absolute Neutrophil count; CBC, Complete blood cell count; CI, Confidence interval; CRP, C-reactive protein; ED, Emergency depart- ment; ESR, erythrocyte sedimentation rate; OR, Odds ratio; PMN, Polymorphonuclear leu- kocytes; ROC, Receiver-operator curves; WBC, White blood cell count.

* Corresponding author.

E-mail address: [email protected] (D. Hanalioglu).

The most affected joint in children is the knee. Patients with a septic knee usually present with fever, swelling, erythema, pain with motion, and refusal to bear weight on the affected joint. However, equivocal sep- tic arthritis cases may demonstrate less severe symptoms [3-5]. There- fore, clinical suspicion is very important in diagnosing septic arthritis and as in many cases, differentiating septic arthritis from other causes of acute monoarthritis is challenging due to overlapping clinical and lab- oratory manifestations. Cultures of joint fluid, the gold standard for di- agnosing septic arthritis, yield positive results in about 52-82% of cases with high suspicion for septic arthritis [1,3,6,7]. Hence, clinical Prediction tools and models are critical to estimate the risk of septic ar- thritis, to identify high-risk patients for subsequent arthrocentesis and antibiotic treatment, and to avoid unnecessary interventions and resource utilization in low-risk patients.


0735-6757/(C) 2022

Kocher criteria, a validated clinical prediction rule to identify chil- dren at low risk for septic arthritis of the hip have been widely used in clinical practice [8-11]. However, its applicability and validity in knee monoarthritis are debated. Previous studies have demonstrated that its sensitivity to rule out knee septic arthritis is low, and thus it should be supplemented by other findings such as CRP > 2 mg/dL (i.e. modified Kocher criteria) [12-16]. Therefore, independently validated clinical prediction rules specific for septic knee monoarthritis are needed to re- place, modify, or complement the existing Kocher criteria.

In this study, we aimed at identifying predictors to differentiate sep- tic arthritis from non-septic causes in a cohort of children who under- went arthrocentesis due to high suspicion for septic arthritis of the knee. Our secondary objective was to develop a clinical decision- support algorithm to assist in justifying arthrocentesis decision by com- bining clinical and laboratory data acquired from patients who had synovial fluid sampling.

  1. Methods
    1. Study design, setting and ethics approval

We conducted a retrospective analysis of all consecutive patients

<18 years of age who presented to the pediatric ED and underwent arthrocentesis due to clinical suspicion of septic arthritis of the knee at a single institution between January 2014 and January 2021. This study was approved by the Institutional Review Board and complies with the Helsinki Declaration.

    1. Study population, inclusion, and exclusion criteria

We identified patients through the electronic medical records data- base by using a process that consisted of several steps. First, we identi- fied potential subjects among those who initially presented to the ED with a chief complaint of limping, joint pain, joint swelling, erythema on the joint, Musculoskeletal pain, myalgia, cellulitis of the lower ex- tremities, who were then admitted to the hospital and all patients with a discharge diagnosis of septic arthritis, arthritis, or synovitis. In ad- dition, we screened all patients with a synovial fluid culture to avoid missing any potential subjects. Next, two study authors independently reviewed each potential subject’s patient chart to select those who met the inclusion criteria. We included subjects with documented acute knee monoarthritis (symptom onset <2 weeks) [17] on physical examination who had undergone arthrocentesis due to concern for septic arthritis and had complete laboratory work-up. We excluded patients who had joints involved other than the knee or multiple joint involvements, and those who did not undergo synovial fluid aspiration, those with a previous history of arthritis, knee trauma, accompanying osteomyelitis, overlying cellulitis, recent knee surgery, and immunodeficiency.

    1. Management

At our institution, we follow an institutional pathway for the evalu- ation of patients with an acute irritable joint; obtain a detailed medical history and physical exam and pay attention to Red flags, in particular fever >=38.5?C and non-weight bearing status, draw CBC count with dif- ferential, ESR, CRP, and blood culture. Patients with a clinical or radio- logical concern for septic arthritis and have either elevated serum WBC (>12,000 cells/mm3) and/or ESR (>40 mm/h) and/or CRP (>2 mg/dL) undergo diagnostic arthrocentesis for synovial fluid cell count, differential, Gram staining, and culture, are admitted to the hospital, and receive empiric intravenous antibiotic treatment. The antibiotic regimen is tailored based on microbiological data. Patients with low clinical suspicion for septic arthritis and no predictors present are discharged with close follow-up within two days (Supplementary

Fig. 1). The region where this study was conducted is not a Lyme- disease endemic area, therefore Lyme tests are not performed routinely.

    1. Data collection

Two authors independently reviewed medical records of each pa- tient individually to confirm a diagnosis of septic arthritis. In case of dis- agreement, a consensus was reached through discussion. Data obtained from medical records included patient demographics such as age, gen- der, presenting symptoms, symptom duration, associated symptoms such as fever, rash, recent/concurrent infectious symptoms, recent wound, trauma at affected joint, recent antibiotic use, recent joint sur- gery and/or intraarticular injection, past medical history of arthritis, septic arthritis, osteomyelitis and/or joint surgery, general physical ex- amination findings such as body temperature, rash, wound, joint evalu- ation such as swelling, erythema, warmth, tenderness, limited range of motion, pain with micromotion and ambulation features such as non- weight bearing status or limping and laboratory results such as WBC count, absolute neutrophil count, Neutrophil to lymphocyte ratio, eryth- rocyte sedimentation rate, C-reactive protein (CRP), blood culture, sy- novial fluid Bacterial culture, WBC count and differential (percentage polymorphonucleocytes, (% PMN)), and synovial fluid Gram staining. Only initial serum laboratory tests and synovial aspirate analyses at the presentation were included in the analyses.

    1. Definitions and outcome measures

We defined fever as a temperature of 38.5?C or higher according to Kocher criteria [8], weight-bearing status based on the clinical history and physical exam documented in the medical records, recent antibiotic use as receiving antibiotics 72 h prior to the ED visit. We defined culture-positive septic arthritis as the presence of a positive synovial fluid culture. However, cultures that grew Bacillus species, Corynebacte- rium, and Streptococcus viridans were considered contaminants. If two or more organisms were grown in culture, we used the term “polymicrobial”. We defined a culture negative septic arthritis in case bacterial culture of synovial fluid did not grow any microorganisms and either 1) synovial WBC count was >50,000 cells/ mm3 or 2) Gram stain was positive and/or synovial WBC count was >25,000 cells/ mm3 with a predominance of polymorphonuclear cells (>90%) and clinical records demonstrating a consensus between Pediatric Infectious Diseases and Orthopedics consultants for final diagnosis of septic arthritis. These two groups together formed the septic arthritis group. We classified subjects outside these definitions as “non-septic arthritis”.

    1. Statistical analyses

Statistical analyses were performed with IBM SPSS v23.0 program. Data were presented as mean +- standard deviation, median [inter- quartile range] or number (percentages), whenever appropriate. Student’s t-test and chi-square were performed to detect the differ- ences between two independent groups for continuous and categor- ical variables, respectively. Logistic regression models were used for identifying predictor variables. Different cut-off values for age were tested as in univariate analyses. Variables with p < 0.1 in the univar- iate analysis and those deemed clinically important were introduced into the multivariable model. The model coefficients were presented as odds ratios (OR) and 95% CI. The performance characteristics of the independent predictors were analyzed using receiver-operator curves (ROC) and areas under curve (AUCs) were calculated. A p value <0.05 was considered statistically significant, whereas a p value ranging 0.05-0.1 was considered to represent a trend towards significance.

  1. laboratory characteristics“>Results

Table 1

Demographic and clinical characteristics of the study cohort.?

    1. Demographic and clinical characteristics

Among a total of 237 subjects identified through screening, there were 89 subjects with acute knee monoarthritis who underwent arthrocentesis. Twenty-nine patients with possible confounding condi- tions were excluded. A total of 60 subjects were included in the study. Thirty-two (53%) subjects had septic arthritis of the knee (joint fluid culture positive, n = 17; culture negative, n = 15) and 28 (47%) had non-septic arthritis. A flow diagram is presented in Fig. 1.

Patients in the septic arthritis group were younger than the patients

in the non-septic arthritis group (median age 5 [2-9] vs. 7 [5-14], p = 0.017). The percentage of patients aged <=5 years was significantly higher in the septic arthritis group than non-septic arthritis group (n = 19 (59%) vs. n = 7 (25%), p = 0.007). History of fever (n = 13

(41%) vs. n = 8 (29%), p = 0.329), symptom duration (4 [1-11] days vs. 2 [1-6.5] days, p = 0.319), and proportion of patients with recent ill- ness (n = 12, (38%) vs. n = 10, (36%), p = 0.886) were similar in both groups. Medication use for fever and pain, either acetaminophen or ibu- profen, and antibiotic pretreatment prevalence were higher in the sep- tic arthritis group however these differences were not statistically significant.

Joint pain/tenderness was present in almost all patients (n = 31,

Characteristic Total

(n = 60)


Age, years

6.3 [4-10]

5 [2-9]

7 [5-14]


Age <= 5 years

26 (43)

19 (59)

7 (25)


<1 year

3 (5)

2 (6)

1 (4)

1-2 years

7 (12)

5 (16)

1 (4)

3-5 years

16 (27)

11 (34)

5 (18)


38 (63)

19 (59)

19 (68)


Medical history

History of fever

21 (35)

13 (41)

8 (29)


Symptom duration, days

2.5 [1-7.5]

4 [1-11]

2 [1-6.5]


Recent illness

22 (37)

12 (38)

10 (36)


Recent anti-pyretic or

19 (32)

12 (38)

7 (25)


analgesic use

Recent antibiotic use

15 (25)

11 (34)

4 (14)


Physical examination

Triage temperature, ?C

37.3 +- 1.1

37.5 +- 1.2

37.1 +- 0.9


Left knee

29 (48)

15 (47)

14 (50)


Joint pain/tenderness

58 (97)

31 (97)

27 (96)


Joint warmth

36 (60)

20 (63)

16 (57)


Joint redness

8 (13)

3 (9)

5 (18)


Limited range of motion

55 (92)

30 (94)

25 (89)


Non-weight bearing

45 (75)

24 (75)

21 (75)


Septic (n = 32)

Non-septic (n = 28)



(97%) vs. n = 27, (96%), p = 0.923). Limited range of motion on the af- fected knee (n = 30, (94%) vs. n = 25, (89%), p = 0.476), non-weight

bearing status (n = 24, (75%) vs. n = 21, (75%), p = 1) and joint

warmth (n = 20, (63%) vs. n = 16, (57%), p = 0.673) were most com- mon physical examination findings and were similar among groups. De- mographic and clinical characteristics of the patients are presented in Table 1.

    1. Laboratory characteristics

Laboratory evaluation of patients with septic arthritis demonstrated significantly higher WBC count than patients with other inflammatory arthritis (12.6 +- 3.9 x 103 cells/mL vs. 10.1 +- 3.5 x 103 cells/mL, p = 0.011). The proportion of patients with WBC >12,000 cells/mL was also significantly higher in the septic arthritis group as compared to the non-septic arthritis group (n = 18, (56%) vs. n = 7 (25%), p = 0.007). There was a trend towards significance in comparison of per- centages of patients with CRP > 2 mg/dL which was higher in the septic

Image of Fig. 1

Fig. 1. Study flow diagram.

* Data are presented as median [interquartile range], mean +- standard deviation or

n (%).

arthritis group than non-septic arthritis group (n = 23, (72%) vs. n = 14, (50%), p = 0.082).

There was no statistically significant difference in other laboratory parameters between the two groups. Laboratory characteristics of the patients are presented in Table 2.

Of the 32 children with septic arthritis, 17 (53%) had positive syno- vial fluid culture result. common pathogens grown included Staphylo- coccus aureus (n = 7), Staphylococcus epidermidis (n = 7), Klebsilella pneumonia (n = 2), Escherichia coli (n = 1), Propionibacterium acnes (n = 1), Pantonea agglomerans (n = 1), Burkholderia cepacia (n = 1) and Kocuria maria (n = 1). Four patients had polymicrobial culture results.

    1. Univariate and multivariable analyses of predictors

Univariate and multivariable analyses results are shown in Table 3. Among all tested cut-off values for age (4, 5 and 6 years), <=5 years showed the highest discriminating power. We tested six parameters

Table 2

Laboratory characteristics of the study cohort.

Laboratory work-up





(n = 60)

(n = 32)

(n = 28)


WBC, cells, x 103 per mL

11.4 +- 3.9

12.6 +- 3.9

10.1 +- 3.5


WBC > 12 x 103 per mL

25 (43)

18 (56)

7 (25)


ANC, cells, x 103 per mL

6.9 +- 3.0

7.5 +- 2.5

6.4 +- 3.3



3.2 +- 2.2

3.4 +- 2.4

3.0 +- 2.1


NLR > 2.4

31 (52)

18 (56)

13 (46)


NLR >4

17 (28)

9 (28)

8 (29)


ESR, mm/h

31.0 +- 29.0

36.5 +- 32.4

25.6 +- 24.5


ESR > 40 mm/h

22 (37)

14 (44)

8 (29)


CRP, mg/dL

6.7 +- 8.5

7.6 +- 8.3

5.7 +- 8.8


CRP > 2 mg/dL

37 (62)

23 (72)

14 (50)


Synovial WBC count x 103

44.4 +- 11.8

49.9 +- 14.6

31.9 +- 19.0


per mL

Synovial ANC count x 103

39.2 +- 11.7

45.2 +- 13.7

24.5 +- 19.8


per mL


75.9 +- 7.6

81.6 +- 7.0

61.5 +- 19.5


Abbreviations: WBC; white blood cell, ANC; absolute neutrophil count, NLR; neutrophil to lymphocyte ratio, ESR; erythrocyte sedimentation rate, CRP; C-reactive protein.

Table 3

Univariate and multivariable analyses of septic arthritis predictors.

Univariate Multivariable


OR (95% CI)

p value

OR (95% CI)

p value

Modified Kocher Criteria

Kocher Criteria

Non-weight bearing

1.421 (0.461-4.384)


0.248 (0.034-1.804)


Temperature > 38.5?C

4.333 (0.836-22.469)


1.494 (0.224-9.953)


ESR > 40 mm/h

1.944 (0.662-5.709)


1.665 (0.456-6.085)


WBC > 12,000 cells/mm3

3.857 (1.278-11.638)


5.059 (1.424-17.970)


CRP > 2 mg/dl

2.556 (0.878-7.443)


3.180 (0.895-11.298)


Age <= 5 years

4.385 (1.447-13.288)


4.237 (1.270-14.127)


in the multivariable model. Age <= 5 years (OR: 4.237, [1.270-14.127],

p = 0.019) and WBC > 12,000 cells/mm3 (OR: 5.059, [1.424-17.970],

p = 0.012) were found to be the independent predictors of septic arthri- tis diagnosis. CRP > 2 mg/dL (OR: 3.180, [0.895-11.298], p = 0.074) was only marginally significant.

Based on multivariable analysis results and previous literature, we tested six different predictive models (Fig. 2). Kocher criteria showed poor discrimination for knee septic arthritis (AUC = 0.677). When CRP was added to the model (i.e., modified Kocher criteria), discrimina- tion power increased to AUC = 0.699. Our results show that Kocher criteria missed 22% and 69% of septic arthritis patients if cut-off was chosen >=2 and >=3 respectively. Modified Kocher criteria missed 6%, 41% and 72% of septic arthritis patients if cut-off was chosen >=2, >=3 and >=4 respectively.

Addition of age <= 5 years to modified Kocher criteria (i.e., full model) further improved the discrimination power to AUC = 0.746. Moreover, this model perfectly ruled out septic arthritis if none of the six criteria was present (Table 4). This model was able to identify one patient <5 years who had Kocher = 0 and Modified Kocher = 0 but indeed was diagnosed with septic arthritis. This patient had mild symptomatology but had an elevation in CRP two days after initial test results and underwent synovial fluid sampling and eventually was diagnosed with septic arthritis. Next, we investigated if backward elimination of these criteria would yield a simpler yet effective prediction model. Removal of non- weight bearing criterion improved AUC to 0.768 (5-tier model). Next, the elimination of ESR > 40 mm/h criterion improved AUC even further to 0.776 (4-tier model). Finally, reduction of the model to 3 major criteria by eliminating temperature of 38.5?C slightly decreased AUC to 0.766 (3-tier model). However, when the septic arthritis frequencies in respective scores in each of these models were analyzed, the final 3- tier model missed only 1 patient (3% of 32 septic arthritis patients) if all 3 criteria were absent, similar to 4-tier and 5-tier models. Based on these findings, we proposed a two-step algorithm (major and minor criteria) for risk stratification and decision-making for arthrocentesis

in patients with clinical suspicion of knee septic arthritis (Fig. 3).

This missed patient was a 4-year-old male who had ongoing symp- toms for 5 days. He presented with non-weight bearing on the right side. He had been given acetaminophen and ibuprofen routinely to con- trol the pain. On his initial evaluation, his body temperature was 36?C, WBC 8700 cells/mm3, ANS 5700 cells/mm3, CRP 0.7 mg/dL and ESR 16 mm/h. On follow-up (24 h later) his blood work was repeated which showed a trend of increase: WBC 11,300 cells/mm3, ANS 4800 cells/mm3, CRP 8.5 mg/dL and ESR 35 mm/h. His blood culture was negative. The joint fluid culture was negative but on microscopical examination, gram- positive coccus and 90% PM were detected. This patient was treated as septic arthritis.

  1. Discussion

In this study, we examined Predictive powers of various models in patients with acute knee monoarthritis and showed that age 5 years

or younger, an elevated CRP >2 mg/dL, leukocytosis >12,000/mm3, and a fever >38.5?C were the most powerful predictors of a septic knee. Patients with all six criteria (Kocher criteria plus age <= 5 years and CRP > 2 mg/dL) negative were at the lowest risk for septic arthritis of the knee. Based on our findings, we propose a decision-support algo- rithm for identifying low-risk patients to rule out septic arthritis and thus avoid arthrocentesis.

    1. Timely and accurate diagnosis of septic arthritis

A timely and accurate diagnosis of septic arthritis is very important since Delays in diagnosis may cause a more complicated disease course and potentially result in death and disability [3]. The gold standard of septic arthritis diagnosis is joint fluid culture which is obtained by an in- vasive procedure [3]. Arthrocentesis itself is a trauma to the joint, causes significant pain and requires sedation and analgesia, which adds further potential undesired complications or adverse event risks. Moreover, the benefit of a synovial fluid culture in diagnosing septic arthritis is limited by its low sensitivity [18]. Cultures yield positive results in about 52-82% of symptomatic cases [1,3,6]. In the current study, this propor- tion was at the lower margin. This low rate of culture positivity might be partly related to high frequency of pre-emptive antibiotic use in our cohort.

    1. Kocher criteria in septic arthritis

The original Kocher study included 282 patients with an irritable hip. There were 82 patients with septic arthritis vs. 86 with transient syno- vitis. In this study, fever >38.5?C, WBC > 12,000/mm3, ESR >40 mm/h and non-weight bearing status were identified as independent predic- tors of septic arthritis of the hip. The probability for septic arthritis based on the presence of these predictors was as follows: none <0.2% risk, 1 criterion 3%, 2 criteria 40%, 3 criteria 93.1% and 4 criteria 99.6%. Even though the subsequent validation studies demonstrated lower probabilities for these criteria (none 2% risk, 1 criterion 9.5%, 2 criteria

35%, 3 criteria 72.8% and 4 criteria 93.0% vs. 59.1-59.9% in presence of all four criteria), this prediction rule has been widely accepted to iden- tify children at low risk for septic arthritis of the hip [8-11].

    1. C-reactive protein as an addition to Kocher criteria

Caird et al. [14] modified the Kocher criteria by adding CRP >2 mg/dL as the fifth criteria and found that predicted probability of septic arthritis, among children who had hip synovial fluid aspiration due to suspected septic arthritis, was as follows: none 16.9% risk, 1 crite- rion 36.7%, 2 criteria 62.4%, 3 criteria 82.6%, 4 criteria 93.1% and 5 criteria 97.5%. C-reactive protein >2 mg/dL was identified as a strong independent predictor of septic arthritis and the addition of CRP improved diagnostic sensitivity of the Kocher criteria [14,15]. How- ever, both Kocher and modified Kocher criteria were developed and validated only for distinguishing septic arthritis from transient synovitis of hip, but not knee.

Image of Fig. 2

Fig. 2. Receiver operating characteristic curves for different predictive models.

    1. Reliability of Kocher criteria in knee septic arthritis

In 2019, Obey et al. [12] investigated the sensitivity of Kocher criteria plus CRP to rule out the diagnosis of septic arthritis of the knee. This study found that the sensitivity for distinguishing septic arthritis of

Table 4

Frequencies of patients with knee septic arthritis according to different prediction models.


N (%)

M. Kocher

N (%)

Full model

N (%)


1 (33)


1 (50)


0 (0)


6 (33)


1 (11)


2 (22)


15 (58)


11 (55)


4 (36)


10 (77)


10 (53)


12 (55)



9 (90)


8 (80)



6 (86)

p value


p value



p value


5-tier model

N (%)

4-tier model

N (%)

3-tier model

N (%)


1 (13)


1 (11)


1 (11)


5 (39)


8 (40)


9 (43)


10 (50)


11 (65)


15 (71)


7 (78)


10 (83)


7 (100)


9 (90)


2 (100)


p value


p value


p value


the knee in presence of 3 or more criteria was 48.5% for Kocher criteria and 71.6% for modified Kocher criteria (i.e., CRP included) which was lower than that for the hip (84% vs. 79%). Means of the parameters used in clinical practice for the hip had a poor performance in predicting septic arthritis of the knee. In line with these results, our study did not find a significant difference in the percentage of patients with ESR

>40 mm/h between septic and non-septic arthritis groups. This might

be due to relatively smaller sample size or other inflammatory etiologies in patients with non-septic arthritis in the current study, relatively low specificity of ESR for septic arthritis in general, or knee SA in particular. In our study, Kocher criteria missed 22% and 69% of septic arthritis patients if cut-off was chosen >=2 and >=3 respectively. Modified Kocher criteria missed 6%, 41% and 72% of septic arthritis patients if cut-off

was chosen >=2, >=3 and >=4 respectively.

    1. Proposed predictive model and algorithm for knee septic arthritis

To date, only a few studies specifically focused on predictors of the septic knee in children presenting with an acutely swollen/irritable knee [12,13,19,20]. Deanehan et al. [19] developed and validated a clin- ical prediction rule which identified patients at low risk for septic arthri- tis in a Lyme-disease endemic area. They identified ANC >=10,000/mm3 and ESR >=40 mm/h as predictors of septic arthritis. Children with zero criteria were at low risk for septic arthritis (sensitivity 100%, specificity 54%). Interestingly, Kocher criteria identified all patients with septic

Image of Fig. 3

Fig. 3. Proposed two-step algorithm for risk stratification and decision-making for arthrocentesis in patients with suspected acute knee septic arthritis. Three major criteria are selected based on the 3-tier model, while 3 minor criteria are other components of Ko- cher criteria. The presence of 2 or more major criteria regardless of minor criteria or 1 major plus at least 2 minor criteria place a patient in the high-risk group and warrants arthrocentesis. The absence of any major or minor criteria is considered low-risk and can be used to avoid arthrocentesis. The presence of 1 major criterion or no major but at least 1 minor criterion comprise the intermediate risk group. In this subgroup, decision to perform arthrocentesis should be individualized. For instance, a patient with 1 major plus 1 minor criterion or 3 minor criteria with no major criteria can tip the balance to- wards arthrocentesis (high-risk) while the presence of only 1 minor criterion in the ab- sence of other supportive clinical factors can favor the avoidance of synovial fluid sampling (low-risk).

arthritis (sensitivity 100%, specificity 12%) but there was a small num- ber of children with septic arthritis in this study. In a recent study, this prediction rule was externally validated for identification at low risk of septic arthritis in a multicenter prospective cohort of children evalu- ated for Lyme disease [20]. Since these studies took place in Lyme dis- ease endemic areas, their applicability to populations living in areas where Lyme disease is not prevalent is limited.

In a recent study, Thomas et al. [13] found that in children with a painful swollen knee, aged under 5 years and an elevated CRP level had a 95% predicted probability for septic arthritis of the knee. It is note- worthy that probability of septic arthritis was 15% for those children of age >= 5 years and CRP <= 2 mg/dL. Addition of synovial %PMN < 85% decreased the probability of septic arthritis to 11%.

Age seems to be an important criterion in predicting septic arthritis, which is clearly demonstrated in our study. Previously, a preliminary study compared characteristics of septic arthritis of the hip and knee in children and reported that children with septic arthritis of the knee presented at a younger age (mean 2.5 y) compared with those with sep- tic arthritis of the hip (mean 4.8 y) [16]. Similarly, Thomas et al. [13] re- ported mean age of patients with septic arthritis of the knee to be 3.25 y and proportion of patients younger than 5 years of age to be 65%. Our findings are consistent with these previous reports. Interestingly, we also found that a cut-off for age <= 5 years showed the highest discrimi- nating power among all tested cut-off values of age. The reason why septic arthritis has a predilection towards Younger children is not clear. It might be due to anatomic, physiologic, immunological, or devel- opmental factors that make younger children more vulnerable to septic

arthritis. It has been postulated that sluggish blood flow in the metaphyseal capillaries makes growing bones susceptible to infection from hematogenous seeding from any trauma or infection [5]. In addi- tion, unnoticed minor traumas to the knee might also be a predisposing factor in this age group.

In this study, we investigated predictive powers of various models comparatively. The most powerful predictor model of septic arthritis was the 4-tier model which included age younger than 5 years old, an el- evated CRP >2 mg/dL, leukocytosis >12,000/mm3, and fever >38.5?C. When fever was removed from this model (i.e., 3-tier model) we found a slight decrease in AUC but it was still more powerful than Kocher, mod- ified Kocher and even full model. Our findings echo the results of the study by Thomas et al. [13]. Based on the findings, our proposed algorithm is presented in Fig. 3. Integrating age and CRP to the Kocher criteria and assigning different significance to each of four Kocher criteria (major vs minor criteria) can reliably distinguish risk categories.

In our study, there was not a single patient having all 6 criteria neg- ative and diagnosed with septic arthritis of the knee. When non-weight- bearing was removed from the model, 5-, 4- and 3-tier models missed only 1 patient. Prior use of Non-steroidal anti-inflammatory drugs might be the reason for absence of other criteria and thus delay in diag- nosis in this patient. Hence, the proposed algorithm which blends our 3- tier model with other Kocher criteria in a stepwise manner (major and minor criteria) and stratifies patients into three risk groups (low, inter- mediate, high) can serve as a decision-support tool for arthrocentesis in suspected knee septic arthritis in children (Fig. 3).

    1. Strengths and limitations

This study is conducted on a cohort of patients who had undergone arthrocentesis due to clinical suspicion for septic arthritis of the knee in response to the need for an accurate and feasible clinical prediction rule in such patients. Our study is one of the few studies aimed at identifying predictors of a septic knee specifically. Our proposed al- gorithm needs validation since there is a possibility that this model would perform differently in other patient populations. We only in- vestigated those children who underwent synovial fluid sampling. Thus, our findings might not cover other patients managed conser- vatively. Given the rarity of septic arthritis, we had a relatively small sample of patients. Therefore, our study might be underpow- ered to detect statistical differences.

  1. Conclusion

A three-tier model comprising three major criteria (age <= 5 years, CRP >2 mg/dL, WBC > 12,000/mm3) performed better than Kocher criteria, modified Kocher criteria and Full Model. However, the Full Model was the most successful to rule out septic arthritis if all criteria were negative. Based on these findings, we propose a simple two-step algorithm to identify low, intermediate and high-risk patients for knee septic arthritis. Our proposed algorithm incorporating major (age, WBC, CRP) and minor (fever, ESR, non-weight bearing) criteria can serve as a practical decision-support tool to justify arthrocentesis in children with suspected knee septic arthritis.


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

Supplementary data to this article can be found online at https://doi. org/10.1016/j.ajem.2022.08.002.

CRediT authorship contribution statement Damla Hanalioglu: Writing – original draft, Visualization, Methodol-

ogy, Investigation, Formal analysis, Data curation, Conceptualization.

Erdinc Turker: Investigation, Data curation. Saygin Kamaci: Writing – review & editing, Supervision, Investigation. Yasemin Ozsurekci: Writing – review & editing, Visualization, Supervision, Investigation. Ali Bulent Cengiz: Writing – review & editing, Supervision, Investigation. Mehmet Ceyhan: Writing – review & editing, Visualization, Supervision, Investiga- tion. Ozlem Teksam: Writing – review & editing, Visualization, Supervi- sion, Methodology, Investigation, Conceptualization.

Declaration of Competing Interest




The following are the supplementary data related to this article.


  1. Cohen E, Katz T, Rahamim E, Bulkowstein S, Weisel Y, Leibovitz R, et al. Septic arthri- tis in children: updated epidemiologic, microbiologic, clinical and therapeutic corre- lations. Pediatr Neonatol. 2020;61(3):325-30. Jun.
  2. Krogstad P. Osteomyelitis and septic arthritis. Textbook of Pediatric Infectious Dis- eases. 6th ed. Saunders; 2009. p. 725-48.
  3. Swarup I, Meza BC, Weltsch D, Jina AA, Lawrence JT, Baldwin KD. Septic Arthritis of the Knee in Children: A Critical Analysis Review. JBJS Rev. 2020;8(1):e0069. https:// doi.org/10.2106/JBJS.RVW.19.00069. PMID: 32105243.
  4. Robinette E, Shah S. Septic arthritis. Nelson Text Book of Pediatrics. 21st ed. 3676-3678.e1.
  5. Paakkonen M. Septic arthritis in children: diagnosis and treatment. PHMT. 2017;8: 65-8. May.
  6. Lyon RM, Evanich JD. Culture-negative septic arthritis in children. J Pediatr Orthop. 1999;19(5):655. Sep.
  7. Yagupsky P. Microbiological diagnosis of skeletal system infections in children. CPR. 2019;15(3):154-63. Dec 9.
  8. Kocher MS, Zurakowski D, Kasser JR. Differentiating between septic arthritis and transient synovitis of the hip in children: an evidence-based clinical prediction algo- rithm*+. J Bone Joint Surg. 1999;81(12):1662-70. Dec.
  9. Kocher MS, Mandiga R, Zurakowski D, Barnewolt C, Kasser JR. Validation of a clinical prediction rule for the differentiation between septic arthritis and transient synovitis of the hip in children. J Bone Joint Surg Am Vol. 2004;86(8):1629-35. Aug.
  10. Luhmann SJ, Jones A, Schootman M, Gordon JE, Schoenecker PL, Luhmann JD. Differ- entiation between septic arthritis and transient synovitis of the hip in children with clinical prediction algorithms. J Bone Joint Surg. 2004;86(5):956-62. May.
  11. Sultan J, Hughes PJ. Septic arthritis or transient synovitis of the hip in children: THE VALUE OF CLINICAL PREDICTION ALGORITHMS. J Bone Joint Surg. 2010;92-B(9): 1289-93. Sep.
  12. Obey MR, Minaie A, Schipper JA, Hosseinzadeh P. Pediatric septic arthritis of the knee: predictors of septic hip do not apply. J Pediatr Orthop. 2019;39(10): e769-72. Nov.
  13. Thomas CS, Schiffman CJ, Faino A, Bompadre V, Schmale GA. Diagnostic criteria for the painful swollen pediatric knee: distinguishing septic arthritis from aseptic effu- sion in a non-Lyme endemic area. Front Surg. 2021;8:740285.
  14. Caird MS, Flynn JM, Leung YL, Millman JE, D’italia JG, Dormans JP.. Factors distin- guishing septic arthritis from transient synovitis of the hip in children: a prospective study. J Bone Joint Surg Am Vol. 2006;88(6):1251-7. Jun.
  15. Singhal R, Perry DC, Khan FN, Cohen D, Stevenson HL, James LA, et al. The use of CRP within a clinical prediction algorithm for the differentiation of septic arthritis and transient synovitis in children. J Bone Joint Surg. 2011;93-B(11):1556-61. Nov.
  16. Joshy S, Choudry Q, Akbar N, Crawford L, Zenios M. Comparison of bacteriologically proven septic arthritis of the hip and knee in children, a preliminary study. J Pediatr Orthop. 2010;30(2):208-11. Mar.
  17. Freed JF, Nies KM, Boyer RS, Louie JS. Acute monoarticular arthritis. A diagnostic approach. JAMA. 1980;243(22):2314-6. Jun 13.
  18. Levine MJ, McGuire KJ, McGowan KL, Flynn JM. Assessment of the test characteristics of C-reactive protein for septic arthritis in children. J Pediatr Orthop. 2003;23(3): 373-7. Jun.
  19. Deanehan JK, Kimia AA, Tan Tanny SP, Milewski MD, Talusan PG, Smith BG, et al. Dis- tinguishing Lyme from septic knee monoarthritis in lyme disease-endemic areas. Pediatrics. 2013.;131(3) Mar 1. e695-701.
  20. Grant DS, Neville DN, Levas M, Balamuth F, Garro AC, Bennett JE, et al. Validation of septic knee monoarthritis prediction rule in a Lyme disease endemic area. Pediatr Emer Care. 2022.;38(2) Feb. e881-5.