7/30/2019 Jb Ja 0811216620

1/9

Differentiating Between Septic Arthritisand Transient Synovitis of the Hip in Children:

An Evidence-Based Clinical Prediction Algorithm*

BY M INI NDER S. KOCHER , M.D., DA VID ZURAK OWSKI , PH.D., AND JAMES R. KASSER, M.D., BOSTON, MA SSACHUSETTS

Investigation performed at Childrens Hospital, Harvard Medical School, Boston

AbstractBackground: A child who has an acutely irritable

hip can pose a diagnostic challenge. The purposes ofthis study were to determine the diagnostic value ofpresenting variables for differentiating between septicarthritis and transient synovitis of the hip in childrenand to develop an evidence-based clinical predictionalgorithm for this differentiation.

Methods:We retrospectively reviewed the cases of

children who were evaluated at a major tertiary-carechildrens hospital between 1979 and 1996 because ofan acutely irritable hip. Diagnoses of true septic arthri-tis, presumed septic arthritis, and transient synovitiswere explicitly defined on the basis of the white blood-cell count in the joint fluid, the results of cultures of

joint fluid and blood, and the clinical course. Univari-ate analysis and multiple logistic regression analysiswere used to compare groups. A probability algorithmfor differentiation between septic arthritis and tran-sient synovitis on the basis of independent multivariatepredictors was constructed and tested.

Results: Patients who had septic arthritis differedsignificantly (p < 0.05) from those who had transientsynovitis with regard to the erythrocyte sedimentationrate, serum white blood-cell count and differential,weight-bearing status, history of fever, temperature, ev-idence of effusion on radiographs, history of chills,history of recent antibiotic use, hematocrit, and gender.Patients who had true septic arthritis differed signifi-cantly (p < 0.05) from those who had presumed septicarthritis with regard to history of recent antibiotic use,

history of chills, temperature, erythrocyte sedimenta-tion rate, history of fever, gender, and serum whiteblood-cell differential. Four independent multivariateclinical predictors were identified to differentiate be-tween septic arthritis and transient synovitis: history offever, non-weight-bearing, erythrocyte sedimentationrate of at least forty millimeters per hour, and serumwhite blood-cell count of more than 12,000 cells percubic millimeter (12.0 x 109 cells per liter). The pre-

dicted probability of septic arthritis was determined forall sixteen combinations of these four predictors and issummarized as less than 0.2 percent for zero predictors,3.0 percent for one predictor, 40.0 percent for twopredictors, 93.1 percent for three predictors, and 99.6percent for four predictors. The chi-square test fortrend and the area under the receiver operating char-acteristic curve indicated excellent diagnostic per-formance of this group of multivariate predictors inidentifying septic arthritis.

Conclusions: Although several variables differedsignificantly between the group that had septic arthritisand the group that had transient synovitis, substantialoverlap in the intermediate ranges made differentia-tion difficult on the basis of individual variables alone.However, by combining variables, we were able to con-struct a set of independent multivariate predictors that,together, had excellent diagnostic performance in dif-ferentiating between septic arthritis and transient sy-novitis of the hip in children.

The initial presentation of an acutely irritable hip ina child can pose a diagnostic challenge to the orthopae-dic surgeon, pediatrician, emergency-room physician, orprimary-care physician. A fter the more apparent radio-

graphic abnormalities of Legg-Perthes disease, slippedcapital femoral epiphysis, and fracture have been ruledout, the differential diagnosis commonly involves septicarthritis and transient synovitis.

The differentiation between septic arthritis andtransient synovitis of the hip in children is essential,since the two clinical entities have different treatmentsand different potential for negative sequelae. Septic ar-thritis is treated with operative drainage and antibiotics,whereas transient synovitis is usually self-limited and istreated symptomatically1,5,8,12,25,26,30. Complications of sep-tic arthritis include osteonecrosis, growth arrest, andsepsis, whereas transient synovitis usually has a benign

*No benefits in any form have been received or will be receivedfrom a commercial party related directly or indirectly to the subject of

this article. Funds were received in total or partial support of theresearch or clinical study presented in this article. The funding sourcewas the American Academy of Orthopaedic Surgeons/OrthopaedicResearch and Education Foundations Health Services Research Fel-lowship Grant (M. S. K.).

Winner of the 1999 A merican Orthopaedic A ssociation-Zimmer A nnual Travel Award for Orthopaedic Residents, the 1999Richard K ilfoyle Award of the New England Orthopaedic Society,and the Best Paper A ward at the Annual A merican OrthopaedicA ssociation Residents Conference, Chapel Hill, North Carolina,March 12, 1999 (M . S. K.).

Departments of Biostatistics (D. Z.) and Orthopaedic Surgery(M. S. K. and J. R. K.), Childrens Hospital, Harvard Medical School,300 Longwood Avenue, Boston, Massachusetts 02115. E-mail addressfor Dr. Kocher: kocher_m@al.tch.harvard.edu.

Copyright 1999 byThe Journal of Bone and Joint Surgery, Incorporated

THE JOURNA L OF BONE AND JOINT SURGERY1662

7/30/2019 Jb Ja 0811216620

2/9

clinical course3,7,13,16,18,24,27,31. In addition, early, accurate di-agnosis of septic arthritis of the hip is essential, as pooroutcomes have been associated with a delay in the di-agnosis of this condition in children9,10,17,21,23.

The differentiation between septic arthritis and tran-sient synovitis of the hip in children can be difficult, since

they often have a similar presentation: an atraumatic,acutely irritable hip in a child who has progressive symp-toms and signs of fever, has a limp or refuses to bearweight, and has limited motion, joint effusion, and ab-normal laboratory findings. Empirically, clinicians haveused various clinical, laboratory, and radiographic vari-ables to distinguish between septic arthritis and tran-sient synovitis.

The purpose of this study was to determine the di-agnostic value of presenting variables for differentiatingbetween septic arthritis and transient synovitis of thehip in children and to develop an evidence-based clini-cal prediction algorithm for this differentiation.

Materials and Methods

The cases of all 282 patients who were evaluated ata major tertiary-care childrens hospital between 1979and 1996 because of an acutely irritable hip for which thedifferential diagnosis involved transient synovitis andseptic arthritis were reviewed. The patients were identi-fied with use of the hospital admissions database forinpatients and the hospital emergency-room databasefor outpatients. The records of identified patients werecross-tabulated with the clinical laboratory databasefor joint-fluid analysis and the radiology database for

fluoroscopic or ultrasonographic aspiration of the hip.The diagnosis of true septic arthritis (thirty-eightpatients) was explicitly assigned when a patient had apositive finding on culture of joint fluid or a whiteblood-cell count in the joint fluid of at least 50,000 cellsper cubic millimeter (50.0 109 cells per liter) with pos-itive findings on blood culture. The diagnosis of pre-sumed septic arthritis (forty-four patients) was explicitlyassigned when a patient had a white blood-cell countin the joint fluid of at least 50,000 cells per cubic milli-meter with negative findings on culture of joint aspirateand blood. The group with septic arthritis (eighty-twopatients) included both the group with true and the

group with presumed septic arthritis. The diagnosisof transient synovitis (eighty-six patients) was explicitlyassigned when the patient had a white blood-cell countin the joint fluid of less than 50,000 cells per cubic mil-limeter with negative findings on culture, resolution ofsymptoms without antimicrobial therapy, and no fur-ther development of a disease process as documentedin the medical record.

Excluded patients (114) included those in atypicalgroups, such as those with immunocompromise (four-teen patients), renal failure (six patients), neonatal sep-sis (six patients), postoperative infection of the hip (fourpatients), later development of rheumatological dis-

ease (four patients), later development of Legg-Perthesdisease (one patient), or associated proximal femoralosteomyelitis (six patients) confirmed with bone aspira-tion in patients who had either radiographic changesin the proximal aspect of the femur or failure to appro-priately respond to arthrotomy and antibiotics given

intravenously. To avoid information bias associated withincomplete data analysis and to avoid selection biasassociated with inclusion of patients who had presump-tive and inconsistent diagnoses, a patient was excluded if

joint-fluid aspirate had not been obtained for a cell count,gram stain, or culture (fifty-seven patients); if peripheralblood had not been obtained for culture or a cell count(eight); if the white blood-cell count in the joint fluid wasless than 50,000 cells per cubic millimeter with negativecultures but the patient was managed with an arthrotomyand intravenous administration of antibiotics (six); or ifthe white blood-cell count in the joint fluid was less than50,000 cells per cubic millimeter with negative cultures

but the patient was managed with intravenous adminis-tration of antibiotics alone on the medical service (two).

Data obtained for all of the patients included age,gender, date of presentation, duration of symptoms, his-tory of fever, history of chills, weight-bearing status, his-tory of trauma, history of concurrent or recent infection,history of recent antibiotic use, temperature, erythrocytesedimentation rate, serum white blood-cell count anddifferential, platelet count, hematocrit, results of bloodculture, evidence of effusion on radiographs, and resultsof gram-staining, cell count, differential, and culture of

joint fluid. A history of fever was operationally defined

as an oral temperature of more than 38.5 degrees Cel-sius during the week before the initial evaluation. Ahistory of chills was operationally defined as positivedocumentation of chills; a negative history of chills wascoded either for documentation of no chills or for nodocumentation of chills. Weight-bearing status was de-termined on the basis of the clinical history.

Univariate analysis was performed with use of thetwo-sample Student t test for continuous variables andFishers exact test for categorical variables. Comparisonwas made between the group with septic arthritis andthe group with transient synovitis and between thegroup with true septic arthritis and the group with pre-

sumed septic arthritis. Stepwise multiple logistic regres-sion with backward selection was performed to identifyindependent clinical predictors, with the same groupcomparisons as were performed for the univariate anal-ysis. Variables with a p value of less than 0.20 in theunivariate analysis were chosen as candidates for themultivariate model, with significance determined withuse of the likelihood ratio chi-square test14. Regressionmodel fit was estimated with the Hosmer-Lemeshowgoodness-of-fit test15. Adjusted odds ratios and 95 per-cent confidence intervals were derived with the methodof maximum likelihood, and the probability of septicarthritis of the hip was estimated for each combina-

DIFFERENTIATING BETWEEN SEPTIC ARTHRITIS AND TRANSIENT SYNOVITIS OF THE HIP IN CHILDREN 1663

VOL. 81-A, NO. 12, DE CEMBER 1999

7/30/2019 Jb Ja 0811216620

3/9

tion of predictors. The Pearson chi-square test for trendwas used to evaluate the relationship between the in-creasing number of predictors and the proportion ofpatients who had septic arthritis. In addition, a receiveroperating characteristic curve was constructed to assessthe diagnostic performance of the group of multivariatepredictors in identifying septic arthritis11. Sensitivity andspecificity were calculated with standard formulae29.Statistical analysis was performed with SPSS (version8.0; SPSS, Chicago, Illinois) and SAS (version 6.12; SASInstitute, Cary, North Carolina) software packages.

Results

Descriptive Data

Arthrocentesis was performed in the operating

room or the fluoroscopy suite in fifty-two (31 percent)of the 168 patients and in the ultrasound suite in theother 116 (69 percent). Of the eighty-two patients withseptic arthritis, thirty-eight (46 percent) had positive re-sults on culture and the remaining forty-four (54 per-cent) had negative results on culture. Of the thirty-eightpatients who had positive results (true septic arthritis),twenty-six (68 percent) had positive joint-fluid andblood cultures, nine (24 percent) had a positive joint-fluid culture and a negative blood culture, and three (8percent) had a negative joint-fluid culture and a positiveblood culture. Organisms isolated on culture includedStaphylococcus aureus (twenty-two patients; 58 per-

cent), Streptococcus pneumoniae (six patients; 16 per-cent),Hemophilus influenzae(five patients; 13 percent),Neisseria meningitidis (three patients; 8 percent), andgroup-A Streptococcus (two patients; 5 percent). A llinfections with Hemophilus influenzae developed be-fore 1988. Of the thirty-five patients who had a positiveculture of joint fluid, twenty-seven (77 percent) had apositive gram stain of joint fluid. There were no positivegram stains of joint fluid from the patients who hadnegative cultures.

Univariate Analysis

Septic Arthritis Comparedwith Transient Synovitis

The eighty-two patients who had septic arthritis dif-

fered significantly (p < 0.05) from the eighty-six whohad transient synovitis with regard to erythrocyte sed-imentation rate (51.6 23.5 compared with 21.3 12.5millimeters per hour), serum white blood-cell count(15,000 5700 compared with 9900 3100 cells percubic millimeter [15.0 5.7 compared with 9.9 3.1 109 cells per liter]) and differential (neutrophils [66.812.7 compared with 60.1 13.8 percent], lymphocytes[19.9 10.4 compared with 29.4 12.6 percent], andbasophils [3.4 2.8 compared with 1.2 0.6 percent]),temperature (38.7 1.0 compared with 37.4 0.6degrees Celsius), non-weight-bearing status (seventy-eight patients [95 percent] compared with thirty pa-

TABLE I

UNIVARIATE A NALYSIS: SEPTIC A RTHRITIS COMPARED WITH TRANSIENT SYNOVITIS

VariableSeptic Arthritis

(N =82)*Transient Synovitis

(N =86)* P Value

Age (yrs.) 6.0 4.2 5.3 2.3 0.17

Male gender (no.) 41 (50%) 57 (66%) 0.03

Duration of symptoms(days) 2.3 1.7 2.1 2.7 0.77History of fever (no.) 67 (82%) 7 (8%)

7/30/2019 Jb Ja 0811216620

4/9

tients [35 percent]), history of fever (sixty-seven pa-tients [82 percent] compared with seven patients [8

percent]), evidence of effusion on radiographs (sixty-three patients [77 percent] compared with thirty-threepatients [38 percent]), history of chills (nine patients[11 percent] compared with zero patients [0 percent]),history of recent antibiotic use (twenty patients [24percent] compared with nine patients [10 percent]), he-matocrit (34.5 3.5 compared with 36.0 2.6 percent),

and male gender (forty-one patients [50 percent] com-pared with fifty-seven patients [66 percent]) (Table I).

A lthough the mean erythrocyte sedimentation rateand white blood-cell count differed significantly (p 12,000 cells per mm3(12.0 109 cells per L)

PredictedProbability of

Septic Arthritis

(percent)

Y es Yes Y es Yes 99.8

Y es Yes Y es No 97.3

Y es Yes No Yes 95.2

Y es Yes No No 57.8

Y es No Y es Yes 95.5

Y es No Y es No 62.2

Y es No No Yes 44.8

Y es No No No 5.3

No Yes Y es Yes 93.0

No Yes Y es No 48.0

No Yes No Yes 33.8

No Yes No No 3.4

No No Y es Y es 35.3

No No Y es No 3.7

No No No Y es 2.1

No No No No 0.1

Receiver operating characteristic curve of the simplified clinical prediction algorithm for septic arthritis.

FIG . 3

DIFFERENTIATING BETWEEN SEPTIC ARTHRITIS AND TRANSIENT SYNOVITIS OF THE HIP IN CHILDREN 1667

VOL. 81-A, NO. 12, DE CEMBER 1999

7/30/2019 Jb Ja 0811216620

7/9

probability of having septic arthritis of the hip. ThePearson chi-square test for trend indicated a verystrong relationship between an increasing number ofpredictors and the proportion of patients who had sep-tic arthritis (chi square = 121.96 on 4 degrees of free-dom) (p < 0.0001). This relationship was also examined

with receiver operating characteristic analysis (Fig. 3).The area under the receiver operating characteristiccurve was 0.96, indicating excellent diagnostic perfor-mance of this group of four multivariate predictors inidentifying septic arthritis.

True Septic Arthritis Comparedwith Presumed Septic Arthritis

Two independent multivariate predictors were iden-tified to differentiate between true septic arthritis andpresumed septic arthritis: history of chills (regressioncoefficient = 3.1, likelihood ratio = 12.21, p < 0.001,adjusted odds ratio = 22.7, and 95 percent confidence

interval = 4.0 to 132.5) and male gender (regressioncoefficient = 1.5, likelihood ratio = 7.46, p = 0.006, ad-

justed odds ratio = 4.3, and 95 percent confidence inter-val = 1.7 to 11.5). For example, the odds that the accuratediagnosis is true septic arthritis as opposed to presumedseptic arthritis are approximately twenty-three timesgreater for a patient who has a history of chills than fora patient who does not have a history of chills, with thelower and upper confidence limits of that estimation at4.0 and 132.5 times greater odds, respectively.

Discussion

When a child has an acutely irritable hip, potentialetiologies include Legg-Perthes disease, juvenile rheu-matoid arthritis, osteomyelitis, psoas abscess, pyogenicsacroiliitis, and Lyme arthritis. However, the differentialdiagnosis commonly involves septic arthritis and tran-sient synovitis. Differentiating between septic arthritisand transient synovitis in an accurate and timely manneris cardinal, as the two diagnoses have vastly differentimplications in terms of treatment and morbidity. Fur-thermore, a delay in the diagnosis of septic arthritis ofthe hip is associated with a poor outcome9,10,17,21,23. None-theless, the differentiation can often be challenging,since patients have similar demographic characteristics,

symptoms, signs, and findings.Clinicians have empirically used different variables

to help distinguish between septic arthritis and tran-sient synovitis. Bennett and Namnyak emphasized thewhite blood-cell count in children who are more thanthree years old2. Morrey et al.23and Klein et al.19 focusedon an elevated erythrocyte sedimentation rate. Chen etal. found both leukocytosis and an elevated erythrocytesedimentation rate to be common4. However, becausethese case series of children with septic arthritis did notinclude a comparison group of children with transientsynovitis, the authors were unable to establish differ-ences between the two groups or to determine the diag-

nostic performance of assorted variables.Various presenting clinical, laboratory, and radio-

graphic data for distinguishing between children whohave septic arthritis and those who have transient sy-

novitis have been found to have poor diagnostic utilitybecause of the substantial overlap between the twogroups. Molteni retrospectively reviewed the cases ofninety-seven patients who had so-called nonspecific ar-thritis and thirty-seven patients who had septic arthritisof various joints, including the hip, knee, ankle, elbow,and wrist22. Diagnoses were established presumptively,with blood-cell analysis of specimens from twelve ofthe ninety-seven patients who had nonspecific arthritisand from thirty-two of the thirty-seven patients whohad septic arthritis. Approximately 50 percent (nine-teen)of the thirty-seven patients with septic arthritishad negative cultures, and the white blood-cell countranged from 5000 to 385,000 cells per cubic millimeter(5.0 to 385.0 109 cells per liter). Without subjectinghis findings to statistical analysis, Molteni compared thegroup with septic arthritis and the group with nonspe-cific arthritis with regard to age (5.8 compared with 6.0years), male gender (twenty-two [59 percent] of thirty-seven patients compared with forty-nine [51 percent]of ninety-seven patients), temperature of more than38.3 degrees Celsius (thirty-one [84 percent] of thirty-seven patients compared with thirty-seven [38 percent]of ninety-seven patients), serum white blood-cell count(7230 compared with 6300 cells per cubic millimeter

[7.23 compared with 6.3

109

cells per liter]), and eryth-rocyte sedimentation rate (fifty compared with thirty-five millimeters per hour). Because of the similaritiesin values and the extent of overlap, he concluded thatthere were no reliable data with which to distinguishthe two different joint processes. Del Beccaro et al. alsofound that overlap between presenting variables im-peded the differentiation of septic arthritis from tran-sient synovitis of the hip in children6. They comparedninety-four children who had transient synovitis (de-fined retrospectively on the basis of the clinical course)with thirty-eight patient who had septic arthritis (de-fined as pyarthrosis). Twenty-one of the ninety-four pa-

TABLE V

DISTRIBUTION OF NUMBER OF MULTIVARIATEPREDICTORS AND SIMPLIFIED A LGORITHM FOR THE

PREDICTED PROBABILITY OF SEPTIC A RTHRITIS

No. ofPredictors*

TransientSynovitis(N =86)

SepticArthritis(N =82)

PredictedProbability of

Septic Arthritis(no. of patients) (no. of patients) (percent)

0 19 (22.1%) 0 (0%)

7/30/2019 Jb Ja 0811216620

8/9

tients with transient synovitis had joint-fluid analysis,and twenty-eight of the thirty-eight patients with septicarthritis had joint-fluid cell counts. On univariate anal-ysis, the authors found significant (p < 0.05) differencesbetween the two groups with respect to temperature(38.1 compared with 37.2 degrees Celsius), erythrocyte

sedimentation rate (forty-four compared with nineteenmillimeters per hour), polymorphonuclear leukocytes(7800 compared with 6300 per cubic millimeter [7.8compared with 6.3 109 per liter]), and band forms(903 compared with 295 per cubic millimeter [0.903compared with 0.295 109 per liter]). Of note, with thenumbers available for study, the serum white blood-cellcount could not be shown to differ significantly (p >0.05) (13,200 compared with 11,200 cells per cubic mil-limeter [13.2 compared with 11.2 109 cells per liter]).A lthough the erythrocyte sedimentation rate was sig-nificantly (p < 0.05) different between the two groups,a substantial amount of overlap made it a poor dis-

criminator. The range of erythrocyte sedimentationrates for the group with septic arthritis of the hip wassix to ninety millimeters per hour (24 percent of thepatients had an erythrocyte sedimentation rate of lessthan twenty millimeters per hour), and the range forthe group with transient synovitis was one to 125 milli-meters per hour (28 percent of the patients had anerythrocyte sedimentation rate of more than twentymillimeters per hour). By combining an erythrocytesedimentation rate of more than twenty millimetersper hour with a temperature of more than 37.5 degreesCelsius, Del Beccaro et al. were able to identify 97

percent of the cases of septic arthritis. However, thoseparameters were also found in 47 percent of the pa-tients who had transient synovitis. Finally, Kunnamoet al. reviewed the cases of 278 children who wereyounger than sixteen years of age and had various ar-thritides, including transient synovitis, juvenile arthri-tis, septic arthritis, serum sickness, enteroarthritis, andSchnlein-Henoch purpura20. They found that the eigh-teen patients who had septic arthritis differed signifi-cantly (p < 0.05) from the other patients with respectto C-reactive protein level, a temperature of more than38.5 degrees Celsius, and a serum white blood-cellcount of more than 12,000 cells per cubic millimeter

(12.0 109

cells per liter).In the present study, we found several significant (p