Signs and symptoms for diagnosis of serious infections in children:

ABSTRACTBackgroundSerious infections in children (sepsis meningitispneumonia pyelonephritis osteomyelitis and cellulitis)are associated with considerable mortality andmorbidity In children with an acute illness the primarycare physician uses signs and symptoms to assess theprobability of a serious infection and decide on furthermanagement

AimTo analyse the diagnostic accuracy of signs andsymptoms and to create a multivariable triageinstrument

Design of studyA prospective diagnostic accuracy study

SettingPrimary care in Belgium

MethodChildren aged 0ndash16 years with an acute illness for amaximum of 5 days were included consecutively Signsand symptoms were recorded and compared to thefinal outcome of these children (a serious infection forwhich hospitalisation was necessary) Accuracy wasanalysed bivariably Multivariable triage instrumentswere constructed using classification and regressiontree (CART) analysis

ResultsA total of 3981 children were included in the study ofwhich 31 were admitted to hospital with a seriousinfection (078) Accuracy of signs and symptomswas fairly low Classical textbook signs (meningealirritation impaired peripheral circulation) had highspecificity The primary classification tree consisted offive knots and had sensitivity of 968 (95confidence interval [CI] = 833 to 999) specificity885 (95 CI = 875 to 895) positive predictivevalue 62 (95 CI = 42 to 87) and negativepredictive value 1000 (95 CI = 998 to 1000) bywhich a serious infection can be excluded in childrentesting negative on the tree The sign paramount in alltrees was the physicianrsquos statement lsquosomething iswrongrsquo

ConclusionSome individual signs have high specificity A seriousinfection can be excluded based on a limited numberof signs and symptoms

Keywordschild sensitivity and specificity serious infectionssigns and symptoms triage

INTRODUCTIONSerious infections in children are usually defined assepsis meningitis pneumonia pyelonephritisbacterial gastroenteritis osteomyelitis and cellulitis1

Their consequences can be severe the mortality ofmeningococcal disease can be as high as 252 andapproximately 7 of children who survive bacterialmeningitis suffer from hearing loss3 In Flandersinfectious diseases are responsible for 80 of alldeaths in children under the age of 1 year and for136 of deaths in children aged between 1 and14 years4 comparable to death rates previouslyreported in the UK5

Incidence rates for serious infections in primarycare have been reported to be around 1 per yearin children between 0 and 14 years old Thisrelatively low incidence contrasts with the highannual incidence of lsquonormalrsquo acute infectionschildren aged 0ndash14 years present an average of 11infections per year to primary care with higher ratesin children under the age of 4 years6 An important

Ann Van den Bruel MD PhD research fellow in general

practice B Aertgeerts MD PhD associate professor in general

practice Department of General Practice Katholieke

Universiteit and CEBAM Belgian Centre for Evidence Based

Medicine Leuven R Bruyninckx MD research fellow in

general practice Department of General Practice Katholieke

Universiteit Leuven M Aerts MSc PhD professor in general

practice and clinical epidemiology Center for Statistics

Universiteit Hasselt Diepenbeek Belgium F Buntinx MD

PhD professor in general practice and clinical epidemiology

Department of General Practice Katholieke Universiteit and

CEBAM Belgian Centre for Evidence Based Medicine Leuven

Belgium and Department of General Practice Universiteit

Maastricht the Netherlands

Address for correspondenceDr Ann Van den Bruel Academic Centre for General

Practice Katholieke Universiteit Leuven

Kapucijnenvoer 33 blok J 3000 Leuven Belgium

Email annvandenbruelmedkuleuvenbe

Submitted 7 September 2006 Editorrsquos response23 October 2006 final acceptance 20 November 2006

copyBritish Journal of General Practice 2007 57 538ndash546

British Journal of General Practice July 2007

A Van den Bruel B Aertgeerts R Bruyninckx et al

538

Signs and symptoms for diagnosisof serious infections in children

a prospective study in primary careAnn Van den Bruel Bert Aertgeerts Rudi Bruyninckx Marc Aerts and Frank Buntinx


task for the primary care physician is to triagechildren with an acute illness into either a very-low-risk group in which a serious infection can be safelyexcluded or a higher-risk group in which furtheraction is warranted Although textbooks accuratelydescribe the signs and symptoms of a specifiedillness or even of a specified bacterial infection thisdoes not reflect a clinical situation in which thephysician has to decide on further managementbased on the childrsquos signs and symptoms Inaddition in primary care children presentthemselves at an early stage of the disease whensigns and symptoms of serious and non-seriousinfections appear similar In a recent paper byThompson et al on the course of meningococcaldisease signs in the first 4 hours of the illness werenon-specific such as coryza or sore throat typicalsigns such as meningeal irritation or haemorrhagicrash appeared only at a median time of13ndash22 hours7 The Dutch College of GeneralPractitioners identified the accuracy of presentingsigns and symptoms for the diagnosis of seriousinfections in children as a gap in the scientific baseof general practice8 In fact evidence directlyanswering diagnostic questions from clinicalpractice remains scarce especially that related tothe value of history taking observation or clinicalexamination910

The aim of this study was to establish the accuracyof presenting signs and symptoms for the diagnosisof a serious infection in children in primary care Inaddition it aimed to create a triage instrument thatclassifies children into a very-low-risk group inwhich a serious infection can be safely excluded ora higher-risk group in which further action iswarranted

METHODIn a prospective diagnostic accuracy study allchildren with an acute illness presenting to primarycare were included consecutively The accuracy ofpresenting signs and symptoms (index test) wasanalysed using hospitalisation for a specified seriousinfection as the reference standard Triageinstruments were created based on multiple signsand symptoms

Data collectionThe study was performed in primary care in FlandersBelgium All children consulting a GP paediatrician orthe emergency department not referred by anotherphysician at the moment of their inclusion in thestudy were considered to be consulting primarycare First two hospitals with a paediatric emergencydepartment were contacted for collaboration in thestudy each in a geographically distinct area

Secondly GPs working close to these hospitals wererecruited for participation in the study on a voluntarybasis Data collection started on 1 January 2004 andended on 30 November 2004 Every physicianparticipated during four separate months equallydistributed over the year 2004 to ensure datacollection in every seasonPatients aged 0ndash16 years with an acute illness for

a maximum of 5 days were consecutively included inthe study Children were excluded if the acuteepisode was caused by a merely traumatic orneurological illness intoxication psychiatric orbehavioural problems without somatic cause or anexacerbation of a chronic condition If children wereentered twice in the study by the same physicianwithin 5 days the second registration wasconsidered a repeated measurement on the samesubject and was subsequently excluded from theanalysis Finally physicians were excluded if theassumption of consecutive inclusion was violated(inclusion of fewer than five children in 1 month)

Index testsPresenting signs and symptoms from history takingand physical examination were recorded on apredefined form The signs and symptoms werechosen based on a systematic review (Van den Bruelet al unpublished data 2007) and on the results of aqualitative study11

lsquoBody temperaturersquo was defined as the highestbody temperature measured by the parents or thephysician Before analysis 05degC was added totemperatures measured under the axilla12 or with atympanic thermometer13

lsquoSomething is wrongrsquo was defined as a subjectivefeeling of the physician that things were not rightSimilar although not identical was the sign lsquodifferentillnessrsquo which was defined as a statement by theparents that this illness was different from previousillnesseslsquoDyspnoearsquo was defined as difficult or laboured

breathing lsquotachypnoearsquo as breathing frequency ofge40 per minute lsquochanged breathingrsquo as any changeas compared to normal breathinglsquoImpaired peripheral circulationrsquo was present when

How this fits inSerious infections are the cause of considerable mortality and morbidity inchildren and primary care physicians need to triage children with an acuteillness for these serious infections The predictive value of classical textbooksigns is sufficient to take action when any of these signs is present and shouldbe evaluated in every acutely ill child Classification trees using a limitednumber of signs and symptoms are able to exclude a serious infection in themajority of children with an acute illness

Original Papers

539



the capillary refill took more than 3 secondslsquoMeningeal irritationrsquo was based on the presence ofneck stiffness Kernigrsquos sign Brudzinskyrsquos sign 1 or2 and a bulging fontanelle or irritability onmanipulation of the head or legs in children agedlt1 year old lsquoPetechiaersquo were present in cases of anon-blanching rash The signs lsquoirritablersquo and lsquodrowsyrsquowere used in the analysis separately and combinedas one variable lsquochanged behaviourrsquo on the basis ofprevious research11

All presenting signs and symptoms were coded aslsquoyesrsquo when present lsquonorsquo when absent and lsquorsquo whenthey could not be evaluated for example headachein a babyParticipating physicians also noted a working

hypothesis for each child at the time of recordingAll procedures were pretested in a small number of

practices

Reference standardSerious infections were defined as admission tohospital with one of the following infectionspneumonia (infiltrate on chest X-ray) sepsis(pathogen in haemoculture) viral or bacterialmeningitis (pleocytosis in cerebrospinal fluid andidentification of bacteria or a virus) pyelonephritis(ge105ml pathogens of a single species and white

blood cells in urine and serum C-reactive proteinelevation) cellulitis (acute suppurative inflammationof the subcutaneous tissues) osteomyelitis(pathogen from bone aspirate) and bacterialgastroenteritis (bacterial pathogen in the stool)Sepsis and meningitis were combined a priori as onediagnostic categoryTwo different and complementary methods were

used to establish the final outcome of the childrenincluded in the study First hospitalisation wasverified for all children by checking hospital recordsfrom the 10 regional hospitals in the areas As aback-up every participating physician completed afollow-up form after every registration period and atthe end of the study on which any known seriousinfection had to be reportedFrom all children thus identified all available

evidence from clinical laboratory radiology andother tests was collected and presented to a panel oftwo professors of paediatrics one paediatrician in aregional hospital and one professor of generalpractice The panel was blinded to the diagnosis ofthe treating physicians decisions were made byconsensus Children were considered as not havingsuffered from a serious infection if no seriousinfection was identified from hospital records orduring follow-up

Excludedbull 181 records not meeting inclusion criteriabull 70 records second inclusion during one episode

Final diagnosis 3950 children without

a serious infection

Excludedbull 23 records not meeting inclusion criteriabull 11 records second inclusion during one episodebull 9 records referred by GP or referral unknown

3658 children included in the study in general practice

3981 children included in the analysis

Final diagnosis 31 children with an acute

serious infection

481 children included in the study in ambulatory paediatric care

256 children included in the study at the emergency department

Excludedbull 7 records not meeting inclusion criteriabull 2 records second inclusion during one episodebull 111 records referred by GP or paediatrician or referral unknown

Figure 1 Patient flowchart

Original Papers

AnalysesFirst the accuracy of the presenting signs andsymptoms was analysed for any serious infection andfor each diagnostic category separately In case of anempty cell in the 2 times 2 table 05 was added to everycell Analyses were performed with STATA (version 8)Secondly a classification and regression tree

analysis (CART)14 was performed to create a triageinstrument using lsquorpartrsquo of the R package (wwwr-projectorg) CART is a form of binary recursivepartitioning The term lsquobinaryrsquo implies that eachgroup of patients represented by a lsquonodersquo in adecision tree can only be split into two groups Thuseach node can be split into two child nodes in whichcase the original node is called a parent node Theterm lsquorecursiversquo refers to the fact that the binarypartitioning process can be applied over and overagain Thus each parent node can give rise to twochild nodes and in turn each of these child nodesmay themselves be split forming additionallsquochildrenrsquo The term lsquopartitioningrsquo refers to the factthat the dataset is split into sections or partitionedCART analysis has a number of advantages over

other classification methods including multivariablelogistic regression First it is inherently non-parametric In other words no assumptions aremade regarding the underlying distribution of valuesof the predictor variables Secondly theinterpretation of results summarised in a tree is verysimple It is much simpler to interpret than themultivariable logistic regression model making itmore practical in a clinical setting The tree producespositive and negative predictive measures and othermeasures such as sensitivity specificity andlikelihood ratios can easily be derived Additionallythe inherent lsquologicrsquo in the tree is easily apparent andmakes clinical senseAn important feature of the analysis in this study

was that the signs and symptoms could be used toeither include or exclude the possibility of a seriousinfection thus exploiting the asymmetry of testsAlso indeterminate test results that is signs thatwere scored as lsquorsquo were considered during theanalysisSensitivity and negative predictive value of the

trees were maximised by introducing a weighingfactor of 75 to the misclassification of a seriousinfection The minimum number of observations in anode in order for a split to be attempted and in anyterminal lsquoleafrsquo was set at 100 in both to obtainsensible and robust splits and accurate predictionsThis method deals effectively with missing datathrough surrogate splits The selection of the finaltree was based on a 50-fold cross-validationprocedure thereby validating the classification treesinternally

RESULTSDescription of the populationIn total 121 physicians participated in the study ofwhich 113 were GPs and eight paediatricians 66were male with an average of 17 years of clinicalpractice experience (range 2ndash35 years) The eightpaediatricians also recruited patients at two differentemergency departmentsThe analyses were based on 3981 patients the

patient flow is illustrated in Figure 1 Children were onaverage 50 years old (range = 002ndash169 years) and2131 were boys (535) Hospital records wereretrieved for 196 children of which 48 were admittedfor reasons other than an acute infection (forexample scheduled surgery) and 117 for an acutebut non-serious infection (predominantlygastroenteritis) A serious infection was diagnosed in31 children (prevalence 078 95 confidenceinterval [CI] = 053 to 111) with 16 cases ofpneumonia five cases of pyelonephritis nine casesof sepsis or meningitis and one case of cellulitisThere were no cases of bacterial gastroenteritis forwhich hospital admission was required No patientdied during the study period The average age of

British Journal of General Practice July 2007 541

Children without Children withserious infection serious infection

(n = 3950) (n = 31)

Age in years (SD range) 50 (42 002ndash169) 22 (27 006ndash141)

Sex male () 2108 (537) 24 (742)

Illness duration hours (SD range) 442 (286 0ndash120) 457 (350 3ndash120)

Included by GP (n = 3407) 3394 13

Paediatrician in ambulatory care (n = 438) 433 5

Paediatrician at emergency department (n = 136) 123 13

Chronic condition present n () 269 (77) 6 (194)

Body temperature ge38degC n () 1761 (542) 24 (774)

Working hypothesisUpper respiratory infection 2076 7Viral infection 876 4Viral gastroenteritis 629 3Other 209 2Pneumonia 46 7Pyelonephritis 47 1Bronchiolitis 15 0Bacterial gastroenteritis 6 0Sepsismeningitis 3 5Cellulitis 1 2No illness present 2 0

Definite diagnosisPneumonia 0 16Sepsismeningitis 0 9Pyelonephritis 0 5Cellulitis 0 1Non-serious infection 3950 0

Table 1 Baseline characteristics for the whole group andfor those with a serious infection



542

children with a serious infection was 22 years (range= 006ndash141 years) 24 (742) were male Otherbaseline characteristics are listed in Table 1

Working hypothesisPhysicians labelled six of the 16 children correctly as

having pneumonia the other 10 were diagnosed ashaving a non-serious infection Five children out ofnine were correctly identified as having sepsis ormeningitis three children were diagnosed withanother serious infection and one was diagnosedwith a non-serious infection Pyelonephritis was

Description of the tree Priorknots in order probability Sensitivity Specificity PPV NPV LR+ LRndash OR

Diagnostic category of appearance () (95 CI) (95 CI) (95 CI) (95 CI) (95 CI) (95 CI) (95 CI)

Tree 1Any serious infection 1 Something is wrong 08 968 885 62 1000 84 004 2310

2 Dyspnoea (833 to 999) (875 to 895) (42 to 87) (998 to 1000) (76 to 94) (001 to 02) (314 to 16981)3 Temperature ge3995degC4 Diarrhoea5 Age ge242 years6 Age le118 years

Tree 2aAny serious infection 1 Something is wrong 08 968 869 59 1000 79 004 2174

2 Dyspnoea (833 to 999) (868 to 888) (40 to 83) (998 to 1000) (71 to 88 ) (001 to 02) (296 to 15978)3 Temperature ge3995degC4 Diarrhoea5 Age ge25 years6 Age le10 years

Tree 3Any serious infection 1 Different illness 08 936 855 48 999 64 008 853

2 Dyspnoea (786 to 992) (843 to 866) (32 to 68) (998 to 1000) (57 to 72) (002 to 03) (203 to 3584)3 Age le324 years4 Temperature ge3795degC5 Diarrhoea6 Age ge064 years

Tree 4aAny serious infection 1 Different illness 08 936 821 39 999 52 008 665


Tree 5Pneumonia 1 Dyspnoea 04 938 932 53 1000 139 007 2069

2 Something is wrong (698 to 998) (924 to 940) (30 to 86) (999 to 1000) (117 to 165) (001 to 05)(272 to 15725)

Tree 6Pneumonia limited 1 Dyspnoea 04 938 921 84 1000 119 007 1747to children lt4 years 2 Something is wrong (698 to 998) (908 to 932) (48 to 135) (997 to 1000) (98 to 144) (001 to 05) (229 to 13307)


2 Different illness (698 to 998) (908 to 925) (25 to 71) (999 to 1000) (96 to 133) (001 to 05) (218 to 12548)

Tree 8Pneumonia limited 1 Dyspnoea 04 938 899 67 1000 92 007 1329to children lt4 years 2 Different illness (698 to 998) (885 to 911) (38 to 108) (997 to 1000) (77 to 111) (001 to 05) (175 to 10114)

Tree 9Sepsismeningitis 1 Something is wrong 02 889 971 65 1000 307 011 2683

(518 to 997) (965 to 976) (29 to 124) (999 to 1000) (229 to 412) (002 to 07) (333 to 21631)

Tree 10Sepsismeningitis 1 Different illness 02 778 953 36 1000 164 023 705

(400 to 972) (946 to 959) (15 to 73) (998 to 1000) (113 to 239) (007 to 08) (145 to 3414)

aUsing easier to remember cut-offs for age LRndash = negative likelihood ratio LR+ = positive likelihood ratio NPV = negative predictive value OR = odds ratio PPV =positive predictive value

Table 2 Test characteristics of all classification trees


diagnosed correctly in only one child the other fourchildren were missed as was the one child withcellulitis Overall physicians diagnosed 12 of 31children correctly at the time of registration (387)Apart from the working hypothesis physicians

found 310 children to be seriously ill of which 17 hada serious infection

Bivariable analysesOverall the diagnostic accuracy of presenting signsand symptoms is limited (Supplementary Table 1)Sensitivities are low only body temperature ge38degChas sensitivity over 80 Specificities are higherwith maximum specificity of 999 for the symptomslsquocyanosisrsquo and lsquomeningeal irritationrsquo Odds ratios(ORs) range from 62 for the sign lsquosomething is wrongrsquoto 019 for the sign lsquoheadachersquoThe probability of a serious infection increases

with increasing body temperature But two childrenpresented with a normal body temperature lowerthan 375degC one child with pneumonia and one withcellulitisSigns of an upper respiratory tract infection do not

exclude a serious infection 21 children of the 31 with

a serious infection showed signs of upper respiratoryinfection Coughing was present in 14 children of the16 with pneumonia however coughing was alsopresent in 47 of the children with a non-seriousinfection Crepitations and tachypnoea two classicalsigns for the diagnosis of pneumonia were presentin eight children with pneumonia dyspnoea in 11children and decreased breathing sounds ordullness on percussion in five Only two children didnot have any sign suggesting pneumoniacrepitations tachypnoea dyspnoea or dullness onpercussionMeningeal irritation was present in one child and

impaired peripheral circulation in two of the ninechildren with sepsis or meningitis leading to lowsensitivity In contrast specificity was very high andpositive predictive value sufficient to take furtheraction when present

Multivariable analysis for any seriousinfectionThe variable lsquosomething is wrongrsquo was the first knotin the primary tree (tree 1 in Table 2 Figure 2) andcorrectly classified 20 children with a serious

543

Original Papers

2994 patients negative1 with a serious infection



gt242 years lt118 years

No or lt242 years gt118 years100 patients positive

3 with a serious infection

101 patients positive2 with a serious infection


137 patients positive 20 with a serious infection

lt3995degC gt3995degC

No or

Yes or No

3981 patients

Yes

Diarrhoea

Temperature

Dyspnoea

Somethingis wrong

Age Age

Figure 2 Classificationtree for any seriousinfection Priorprobability of seriousinfection is 08(n = 31)



544

infection The following five steps using fourvariables (age is used twice) added 10 correctlyclassified children with a serious infection One childwith a serious infection (pyelonephritis) was missedby the tree and 454 children had a false-positiveresult This corresponds to a sensitivity of 968(95 CI = 833 to 999) specificity 885 (95CI= 875 to 895) positive predictive value 620(95 CI = 42 to 87) and negative predictive value1000 (95 CI = 998 to 1000)Analyses were repeated excluding lsquosomething is

wrongrsquo This second tree mainly used the samevariables as the first tree but the sign lsquoillness isdifferentrsquo as stated by the parents has replaced thesign lsquosomething is wrongrsquo (Supplementary Figure 1tree 3 in Table 2)

Multivariable analysis for pneumoniaFor the diagnosis of pneumonia the classificationtree used only two presenting signs and symptomslsquodyspnoearsquo and lsquosomething is wrongrsquo (tree 5 in Table2) classifying 15 of 16 cases of pneumonia correctlyand only 268 children testing false positiveExcluding the sign lsquosomething is wrongrsquo it was

replaced by lsquoillness is differentrsquo as shown inSupplementary Figures 2 and 3 (tree 7 in Table 2)With this tree only one child was missed as in theprevious tree but the number of children testing falsepositive was higher Applying these trees to childrenunder the age of 4 years the positive predictive valueincreased and the specificity was slightly lower (trees6 and 8 in Table 2)

Multivariable analysis for sepsis or meningitisThe tree for sepsis or meningitis used only one signlsquosomething is wrongrsquo eight cases were identified andone case was missed a false-positive resultoccurred in 115 children (tree 9 in Table 2)

When the sign lsquosomething is wrongrsquo was excludedthe tree used lsquoillness is differentrsquo as shown insupplementary figures 4 and 5 (tree 10 in Table 2)two cases of sepsis or meningitis were missed thuslowering sensitivity and 188 children tested falsepositive

DISCUSSIONSummary of main findingsThe prevalence of serious infections for whichhospitalisation was required was low (078)Depending on the practice population a primary carephysician will encounter a serious infection in a childfor which hospitalisation is required two or threetimes a year In contrast the need for triage is high asacute illnesses in children are extremely commonAll signs and symptoms had sensitivity below

90 Specificities were better even over 99 incases of the classic textbook signs such asperipheral circulation cyanosis convulsionsmeningeal irritation and petechiaeContrary to individual signs the classification trees

had high sensitivity and were all superior to thephysiciansrsquo working hypotheses The testcharacteristics of the trees compare favourably tothose of other triage instruments of which theOttawa ankle rules is one of the best-knownexamples15

The sign lsquosomething is wrongrsquo was paramount inevery classification tree In this statement thephysician synthesises results from various sources ofinformation and finds something is not right he orshe has a gut feeling about it However it is notknown which signs the physician based thisconclusion on Possibly some signs and symptomsare counted twice the physician finds the child hasdyspnoea as such and concludes that something iswrong partly based on the same dyspnoea

COMMENTARYIdentifying the really sick child is vital Alarm signals like high fever poor circulation and neck stiffness are well recognised Thisstudy provides a thorough prospective analysis of the presenting features and outcomes of nearly 4000 acutely ill children 31 ofwhom had an acute serious infection Thirty-two findings were considered as of potential prognostic value and their individualcontributions assessed A little used but well-established computer technique CART mdash classification and regression tree analysismdash was employed to tease out those questions that taken together best distinguish between the seriously and only moderately illat an early stage The idea is simple Yesno answers in sequence that are best at assigning severemild cases to one category orthe other are derived Diagrams of the decision tree are clear and demonstrate predictive values at once The lsquoblack boxrsquo mathshave to be taken on trustThe general conclusions conform to current beliefs and support them Negative findings (ruling out) are more reliable than positive

findings (ruling in) What was reassuring was that clinical intuition mdash lsquosomething is wrongrsquo proved the best predictor What generatedthat feeling wasnrsquot discovered Work here for psychomethodologists

Harry Hall


However especially in primary care wherephysicians need to triage patients and refer them tosecondary care if necessary this gut feeling couldprove very useful No difference was found when thediagnostic value of lsquosomething is wrongrsquo wasstratified according to the physicianrsquos experience ofmore or less than 10 years but confidence intervalswere wide In essence the study is not sufficientlypowered for these secondary analysesExcluding the sign lsquosomething is wrongrsquo it was

replaced by a similar although not identical sign thestatement of the parents that this illness is differentfrom previous illnesses It is fair to assume that thissign is a synthesis of information and may equallydepend on the parentsrsquo experience This is illustratedby the fact that only one parent of the three infantsunder the age of 3 months with a serious infectionstated that this illness was different Although theclassification trees did not miss a serious infection inany of these infants using the sign lsquodifferent illnessrsquoin this of population very young children should bedone with caution

Strengths and limitations of the studyThe most important strength of the present study isthe prospective design including all eligible childrenconsecutively which is considered the optimaldesign for diagnostic accuracy studies1617 Inaddition several serious infections were consideredas outcome as triage would be done for a variety ofserious infectionsHowever verification of the outcome had to rely on

information obtained from hospital records andduring follow-up as most patients are seen onlyonce for an acute illness and additional testing israre in primary care Although it is possible that notevery child with a serious infection was identified itis reasonable to assume that this probability wasmade as low as possible by the measures taken Inaddition presenting signs and symptoms could havebeen the reason for additional testing andsubsequently led to a diagnosis of a seriousinfection This may increase sensitivity andspecificity18

Comparison with existing literatureThe present study included unselected children fromall ages as the authors believed a primary carephysician would need to triage every child regardlessof its age In contrast previous studies have focusedon infants or young children and selected patientpopulations For example Pantell et al evaluated theaccuracy of signs and symptoms in infants of aged3 months or younger for the diagnosis of bacteraemiaor bacterial meningitis and found age and lsquovery illappearancersquo the best clinical predictors19

A similar population of children was used to derivethe Baby Check Cards infants under the age of6 months Here a serious illness was defined ifinfants had a positive body fluid bacterial culture apositive chest X-ray or if significant treatment wasrequired in hospital The combination of eitherdrowsiness on history or examination pallor onhistory or examination chest wall recessiontemperature gt38degC and a lump being presentidentified 825 of all babies deemed subsequentlyto be seriously ill20 Bleeker et al created a predictionrule for serious infections in children aged1ndash36 months referred by the GP for fever without acause and included duration of fever poormicturition vomiting age temperature lt367degC orge40degC at examination chest-wall retractions andpoor peripheral circulation in the model1

Implications for future research or clinicalpracticeThe classification trees were validated internally tocorrect for optimism However external validation isnecessary before implementation in clinical practiceespecially in studies with a limited number of casesas in the present one2122 as results from validationstudies can be worse or better than the originalresults23 Future studies should also assess whetherphysician characteristics influence diagnosticaccuracy Considering the low prevalencemulticentre studies may be necessaryIn clinical practice finding any of the classic

textbook signs should be a reason for further actionas their positive predictive value is higher than theprior probability of disease But the absence ofthese signs is no argument for ruling out a seriousinfection In contrast the classification trees missonly one or two cases in other words a seriousinfection can be excluded in the vast majority ofchildren with an acute illness on the basis of a fewsimple clinical tests But it is less clear what theprimary care physician should to do with every childthat tests positive to any of the decision trees In factthe positive predictive value is low leading to asubstantial number of false positives It is not theintention to promote immediate referral for all ofthose children except for those suspected of sepsisor meningitis For the others a review a few hourslater or additional testing such as a chest X-ray orblood samples would be a reasonable option Futurestudies will need to provide the evidenceunderpinning these choicesThe diagnostic value of the individual signs and

symptoms is limited although some signs have highspecificity Combining a limited number of signs andsymptoms in classification trees very few cases aremissed The sign lsquosomething is wrongrsquo as stated by

545

Original Papers

the physician is the strongest predictor for a seriousinfection

Supplementary informationAdditional information accompanies this article athttpwwwrcgporgukbjgp-suppinfo

Funding bodyThe study was financed by an official grant of the FondsWetenschappelijk Onderzoek Vlaanderen (FWO) and by anunconditional grant of Eurogenerics (reference numberG023204N)

Ethics committeeThe study was approved by the ethical committee of theKatholieke Universiteit Leuven (reference number ML2193)

Competing interestsThe authors have stated that there are none



546

AcknowledgementsFirst of all we would like to thank all participating physiciansand patients Secondly we would like to thank thoseinvolved in the expert panel the collection of data andadvice on the analyses and manuscript

REFERENCES1 Bleeker SE Moons KG Derksen-Lubsen G et al Predicting serious

bacterial infection in young children with fever without apparentsource Acta Paediatr 2001 90 1226ndash1232

2 Strang JR Pugh EJ Meningococcal infections reducing the casefatality rate by giving penicillin before admission to hospital BMJ1992 305 141ndash143

3 Koomen I Grobbee DE Roord JJ et al Hearing loss at school agein survivors of bacterial meningitis assessment incidence andprediction Pediatrics 2003 112 1049ndash1053

4 Care-and-health The Flemish agency for care and health (agencyof the Flemish ministry for health and family) httpwwwzorg-en-gezondheidbe (accessed 7 Jun 2007)

5 Wilson D Bhopal R Impact of infection on mortality andhospitalization in the North East of England J Public Health Med1998 20 386ndash395

6 Van den Bruel A Bartholomeeusen S Aertgeerts B et al Seriousinfections in children an incidence study in family practice BMCFam Pract 2006 7 23

7 Thompson MJ Ninis N Perera R et al Clinical recognition ofmeningococcal disease in children and adolescents Lancet 2006367 397ndash403

8 Tasche M Oosterberg E Kolnaar B Rosmalen K Inventory of gapsin the evidence base of general practice Huisarts Wet 2001 4491ndash94

9 Straus SE Bridging the gaps in evidence based diagnosis BMJ2006 333 405ndash406

10 Frank C Evidence based checklists for objective structured clinicalexaminations BMJ 2006 333 546ndash548

11 Van den Bruel A Bruyninckx R Vermeire E et al Signs andsymptoms in children with a serious infection a qualitative studyBMC Fam Pract 2005 6 36

12 Craig JV Lancaster GAWilliamson PR Smyth RL Temperaturemeasured at the axilla compared with rectum in children andyoung people systematic review BMJ 2000 320 1174ndash1178

13 Craig JV Lancaster GA Taylor S et al Infrared ear thermometrycompared with rectal thermometry in children a systematicreview Lancet 2002 360 603ndash609

14 Breiman L Friedman J Olshen R Stone C Classification andregression trees Belmont CA Wadsworth International Group1984

15 Bachmann LM Kolb E Koller MT et al Accuracy of Ottawa anklerules to exclude fractures of the ankle and mid-foot systematicreview BMJ 2003 326 417

16 Sackett DL Haynes RB The architecture of diagnostic researchBMJ 2002 324 539ndash541

17 Rutjes AW Reitsma JB Di Nisio M et al Evidence of bias andvariation in diagnostic accuracy studies CMAJ 2006 174469ndash476

18 Whiting P Rutjes AW Reitsma JB et al Sources of variation andbias in studies of diagnostic accuracy a systematic review AnnIntern Med 2004 140 189ndash202

19 Pantell RH Newman TB Bernzweig J et al Management andoutcomes of care of fever in early infancy JAMA 2004 2911203ndash1212

20 Hewson P Poulakis Z Jarman F et al Clinical markers of seriousillness in young infants a multicentre follow-up study J PaediatrChild Health 2000 36 221ndash225

21 Bleeker SE Moll HA Steyerberg EW et al External validation isnecessary in prediction research a clinical example J ClinEpidemiol 2003 56 826ndash832

22 Steyerberg EW Bleeker SE Moll HA et al Internal and externalvalidation of predictive models a simulation study of bias andprecision in small samples J Clin Epidemiol 2003 56 441ndash447

23 Van den Bruel A Aertgeerts B Buntinx F Results of diagnosticaccuracy studies are not always validated J Clin Epidemiol 200659 559ndash566


task for the primary care physician is to triagechildren with an acute illness into either a very-low-risk group in which a serious infection can be safelyexcluded or a higher-risk group in which furtheraction is warranted Although textbooks accuratelydescribe the signs and symptoms of a specifiedillness or even of a specified bacterial infection thisdoes not reflect a clinical situation in which thephysician has to decide on further managementbased on the childrsquos signs and symptoms Inaddition in primary care children presentthemselves at an early stage of the disease whensigns and symptoms of serious and non-seriousinfections appear similar In a recent paper byThompson et al on the course of meningococcaldisease signs in the first 4 hours of the illness werenon-specific such as coryza or sore throat typicalsigns such as meningeal irritation or haemorrhagicrash appeared only at a median time of13ndash22 hours7 The Dutch College of GeneralPractitioners identified the accuracy of presentingsigns and symptoms for the diagnosis of seriousinfections in children as a gap in the scientific baseof general practice8 In fact evidence directlyanswering diagnostic questions from clinicalpractice remains scarce especially that related tothe value of history taking observation or clinicalexamination910

The aim of this study was to establish the accuracyof presenting signs and symptoms for the diagnosisof a serious infection in children in primary care Inaddition it aimed to create a triage instrument thatclassifies children into a very-low-risk group inwhich a serious infection can be safely excluded ora higher-risk group in which further action iswarranted

METHODIn a prospective diagnostic accuracy study allchildren with an acute illness presenting to primarycare were included consecutively The accuracy ofpresenting signs and symptoms (index test) wasanalysed using hospitalisation for a specified seriousinfection as the reference standard Triageinstruments were created based on multiple signsand symptoms

Data collectionThe study was performed in primary care in FlandersBelgium All children consulting a GP paediatrician orthe emergency department not referred by anotherphysician at the moment of their inclusion in thestudy were considered to be consulting primarycare First two hospitals with a paediatric emergencydepartment were contacted for collaboration in thestudy each in a geographically distinct area

Secondly GPs working close to these hospitals wererecruited for participation in the study on a voluntarybasis Data collection started on 1 January 2004 andended on 30 November 2004 Every physicianparticipated during four separate months equallydistributed over the year 2004 to ensure datacollection in every seasonPatients aged 0ndash16 years with an acute illness for

a maximum of 5 days were consecutively included inthe study Children were excluded if the acuteepisode was caused by a merely traumatic orneurological illness intoxication psychiatric orbehavioural problems without somatic cause or anexacerbation of a chronic condition If children wereentered twice in the study by the same physicianwithin 5 days the second registration wasconsidered a repeated measurement on the samesubject and was subsequently excluded from theanalysis Finally physicians were excluded if theassumption of consecutive inclusion was violated(inclusion of fewer than five children in 1 month)

Index testsPresenting signs and symptoms from history takingand physical examination were recorded on apredefined form The signs and symptoms werechosen based on a systematic review (Van den Bruelet al unpublished data 2007) and on the results of aqualitative study11

lsquoBody temperaturersquo was defined as the highestbody temperature measured by the parents or thephysician Before analysis 05degC was added totemperatures measured under the axilla12 or with atympanic thermometer13

lsquoSomething is wrongrsquo was defined as a subjectivefeeling of the physician that things were not rightSimilar although not identical was the sign lsquodifferentillnessrsquo which was defined as a statement by theparents that this illness was different from previousillnesseslsquoDyspnoearsquo was defined as difficult or laboured

breathing lsquotachypnoearsquo as breathing frequency ofge40 per minute lsquochanged breathingrsquo as any changeas compared to normal breathinglsquoImpaired peripheral circulationrsquo was present when

How this fits inSerious infections are the cause of considerable mortality and morbidity inchildren and primary care physicians need to triage children with an acuteillness for these serious infections The predictive value of classical textbooksigns is sufficient to take action when any of these signs is present and shouldbe evaluated in every acutely ill child Classification trees using a limitednumber of signs and symptoms are able to exclude a serious infection in themajority of children with an acute illness

Original Papers

539






practices







a serious infection





serious infection





Original Papers










(n = 3950) (n = 31)


Sex male () 2108 (537) 24 (742)


Included by GP (n = 3407) 3394 13










542





















(518 to 997) (965 to 976) (29 to 124) (999 to 1000) (229 to 412) (002 to 07) (333 to 21631)


(400 to 972) (946 to 959) (15 to 73) (998 to 1000) (113 to 239) (007 to 08) (145 to 3414)












543

Original Papers











No or

Yes or No

3981 patients

Yes

Diarrhoea

Temperature

Dyspnoea

Somethingis wrong

Age Age




544













Harry Hall











545

Original Papers








546































practices







a serious infection





serious infection





Original Papers










(n = 3950) (n = 31)


Sex male () 2108 (537) 24 (742)


Included by GP (n = 3407) 3394 13










542





















(518 to 997) (965 to 976) (29 to 124) (999 to 1000) (229 to 412) (002 to 07) (333 to 21631)


(400 to 972) (946 to 959) (15 to 73) (998 to 1000) (113 to 239) (007 to 08) (145 to 3414)












543

Original Papers











No or

Yes or No

3981 patients

Yes

Diarrhoea

Temperature

Dyspnoea

Somethingis wrong

Age Age




544













Harry Hall











545

Original Papers








546


























Original Papers










(n = 3950) (n = 31)


Sex male () 2108 (537) 24 (742)


Included by GP (n = 3407) 3394 13










542





















(518 to 997) (965 to 976) (29 to 124) (999 to 1000) (229 to 412) (002 to 07) (333 to 21631)


(400 to 972) (946 to 959) (15 to 73) (998 to 1000) (113 to 239) (007 to 08) (145 to 3414)












543

Original Papers











No or

Yes or No

3981 patients

Yes

Diarrhoea

Temperature

Dyspnoea

Somethingis wrong

Age Age




544













Harry Hall











545

Original Papers








546




























542





















(518 to 997) (965 to 976) (29 to 124) (999 to 1000) (229 to 412) (002 to 07) (333 to 21631)


(400 to 972) (946 to 959) (15 to 73) (998 to 1000) (113 to 239) (007 to 08) (145 to 3414)












543

Original Papers











No or

Yes or No

3981 patients

Yes

Diarrhoea

Temperature

Dyspnoea

Somethingis wrong

Age Age




544













Harry Hall











545

Original Papers








546



































543

Original Papers











No or

Yes or No

3981 patients

Yes

Diarrhoea

Temperature

Dyspnoea

Somethingis wrong

Age Age




544













Harry Hall











545

Original Papers








546




























544













Harry Hall











545

Original Papers








546




































545

Original Papers








546

































546


























Signs and symptoms for diagnosis of serious infections in children:

Documents

Transcript of Signs and symptoms for diagnosis of serious infections in children: