Brief Report

The Accuracy of Motherrsquos Touch to Detect Fever in Children

A Systematic Review

by C L Tenga C J Ngb H Nik-Sherinab A H Zailinawatic and S F Tongd

aDepartment of Family Medicine International Medical University Seremban Negeri Sembilan MalaysiabDepartment of Primary Care Medicine University of Malaya Kuala Lumpur MalaysiacKlinik Keluarga Kuala Lumpur MalaysiadDepartment of Family Medicine Universiti Kebangsaan Malaysia Kuala Lumpur Malaysia

Summary

Universally mothers often use touching to detect fever in their children We perform a systematic review

of published diagnostic studies evaluating the ability of mothers to detect fever in their children by

touching We found 10 studies satisfying our inclusion criteria The meta-analysis revealed a summarysensitivity of 892 and summary specificity of 50mdashmaternal touch is perhaps more useful to exclude

fever rather than to lsquorule inrsquo fever However due to significant heterogeneity in the included studies

interpretation of the summary data is difficult

Key words fever mother palpation children systematic review

Introduction

Fever in children is a common problem in primarycare In resource-poor settings where the use ofthermometer is restricted to the healthcare facilitiesthe initial diagnosis of fever in a child is based almostentirely on the assessment by the mothers often bymere touching [1 2] Even in developed countrieswhere thermometers are more readily available athome mothers may not use them correctly or theystill prefer touching to detect fever [3ndash5] Maternaltouch as a method for fever detection is especiallyimportant in regions endemic for malaria anddenguemdashthe carers often need to make quickdecision whether to seek professional help or toadminister treatment (eg antimalarials [2]) Whenthe child has reached a healthcare facility thepresence of fever can be confirmed by thermometerHowever not infrequently the maternal report offever in their children cannot be confirmed eg whenthe prior use of antipyretic has lowered the bodytemperature and the intermittent nature of most

fevers The question that is in our mind is lsquoin a sickchild for whom fever is suspected can maternaltouching accurately detect feverrsquo To answer thisquestion we proceeded to perform a systematicreview of diagnostic studies evaluating this issue

Methods

One investigator (CLT) performed a literaturesearch of Medline covering the period 1956ndash2006for diagnostic studies using the method reported byDeville et al [6] We then looked within the aboveretrieval for citations containing these search terms(fever OR febrile) AND (palpate OR palpation ORtouch) [note lsquoORrsquo lsquoANDrsquo are Boolean operatorsthe asterisk () is a truncation symbol used inPubMed] The initial search identified 27 studies(full list is available from the author) Perusal ofthis list and their references identified 10 studiescomparing maternal touching with thermometer fortemperature measurement [7ndash16] We excluded thefollowing studies (i) one study that did not formallyevaluate touching [2] (ii) one prospective studywhere calculation was based on patient-visit ratherthan individual patients [17] and (iii) three studiesthat evaluated only the accuracy of healthcarepersonnelrsquos palpation for fever [18ndash20] Seven inde-pendent assessors (all primary care doctors) evalu-ated included studies using a checklist [21] Weextracted numerical data of true positive truenegative false positive and false negative from theincluded studies and attempted a meta-analysis(random effect models) using Meta-Disc [22]

Acknowledgements

We wish to thank Dr Ambigga and Dr Chen WS forparticipating in the appraisal of the publisheddiagnostic studies

Correspondence Associate Prof C L Teng Departmentof Family Medicine International Medical UniversityJalan Rasah 70300 Seremban MalaysiaE-mail lttengclgmailcomgt

The Author [2007] Published by Oxford University Press All rights reserved For Permissions please email journalspermissionsoxfordjournalsorg 70doi101093tropejfmm077 Advance Access Published on 25 November 2007

at University of N

ewcastle on Septem

ber 25 2014httptropejoxfordjournalsorg

Dow

nloaded from

The study by Hooker et al [14] was excluded fromthe meta-analysis because numerical data were notavailable

Results

All 10 studies were cross-sectional studies evaluat-ing maternal touching vs an objective method oftemperature measurement Considerable differen-ces in the study design in these studies were noted(Table 1)

(i) Setting five studies were conducted in hospitaloutpatient clinics or paediatric clinics [8 9 1112 15] four in emergency room [7 10 13 14]and one in a hospital ward [16] Five of themwere conducted in the United States [9 1012ndash14] the rest were conducted in four othercountries (Brazil [1] India [11 15] Malawi [8]Zambia [16])

(ii) Subject selection in two studies children withcomplaints of fever were selected [7 11] Onestudy included only children whose mothersused palpation method for fever detection [12]The rest of the studies either included childrenwith various complaints or did not mention thetype of presentation

(iii) Site of maternal touching only four studiesmentioned the body parts touched by mothers[7ndash9 16]

(iv) Type of thermometer three studies usedelectronic thermometer [10 12 14] fourstudies used mercury thermometer [7 11 1516] one study used a mixture of electronic andmercury thermometer [9] and two studiesdid not identify the type of thermometerused [8 13]

(v) Site of temperature measurement two studiesmeasured axillary temperature [7 16] twostudies measured rectal temperature [8 14]one study measured forehead temperature [10]five studies measured temperature at more thanone site [9 11ndash13 15]

(vi) Cut-off value of fever was variable(375ndash383C) depending on the type ofthermometers

Two studies excluded children who had receivedantipyretics [11 13] Four studies [7 9 11 15]enquired whether the children had temperature takenat home with a thermometer but only one of themexcluded these children [11] Maternal touchingappeared to precede objective temperature measure-ment in all studies but only one study [8] mentionedblinding of the assessment by mother and healthcarepersonnelThe sensitivities of maternal touching for fever

were 706ndash973 but the ranges for specificities weremuch wider (192ndash906) (Table 1) Eight out of tenstudies showed higher sensitivities than specificities

for maternal touching The meta-analysis showeda summary sensitivity of 892 and a summaryspecificity of 50 with a diagnostic odds ratio of 15(Table 2) However analysis showed statisticallysignificant heterogeneity (Plt 001) in the includedstudies

Discussion

We used only MEDLINE for our literature search(plus tracking of cited references in the full papers)and may have missed potentially useful studiescaptured only in other databases eg CINAHL orEMBASE (but we think this is unlikely) Thissystematic review suggests that maternal touch forfever detection in children is probably moresensitive than specific The 95 confidence levelfor diagnostic odds ratio is above one this showsthat maternal touching has some diagnostic valueAs the summary sensitivity is higher than thesummary specificity maternal touching is probablymore useful to exclude fever rather than to confirmfever However the significant heterogeneity ofthe included studies (hence the wide confidenceinterval of the diagnostic odds ratio) also makes theresults difficult to interpret Among the includedstudies only two of them specifically includedchildren complaining of fever [7 11] This groupof children is closest to the intended study group(lsquosick child for whom fever is suspectedrsquo) In thestudy by Alves et al [7] maternal report of fever wasalmost invariably correct [positive predictive value(PPV) 972] on the other hand in the studyby Chaturvedi et al [11] the PPV was only381 It appears that the PPV in these twostudies were influenced a great deal by theprevalence of fever in the study samples (Table 1)Of the 10 studies included in our systematic reviewonly three of them showed a PPV exceeding 80[7 14 15]Based on the aforementioned perhaps it is unwise

to rely on maternal report of fever unless it iscollaborated by more objective measurementHowever Dunyo et al [17] argued otherwise Ina prospective study where 130 children wereexamined weekly for 3 months by carers (usingtouching) and trained interviewers (using electronicthermometer) they showed that lsquohot bodyrsquo turnedout to be fever in 79 of instances while theproportion of afebrile children not having lsquohotbodyrsquo was 993 This study suggested that carerscould be taught to recognize fever accurately giventhe chance to verify their touching with objectivemeasurement We conclude that notwithstandingour systematic review the utility of maternaltouching as a method of fever detection is stillunresolved and deserved further study Clearlybetter studies should be done by avoiding themethodological problems identified the main issues

BRIEF REPORT

Journal of Tropical Pediatrics Vol 54 No 1 71

at University of N

ewcastle on Septem

ber 25 2014httptropejoxfordjournalsorg

Dow

nloaded from

TABLE 1Characteristics of included diagnostic studies

Study Subjects and setting Methods Prevalenceof fever ()

Sensitivity Specificity PPV NPV LRthorn LR

1 Alves [7] 169 children reporting feverage 2 months to 13yearsER Brazil

Touching of neck axillarytemp (38C mercurythermometer 2min)

811 759104137

9062932

972104107

4682956

807 027

2 Nwanyanwu [8] 1120 children with variouscomplaints age lt5 yearshospital OPD Malawi

Touching of foreheadrectal temp (38C type ofthermometer NA)

367 973399410

192136710

410399973

925136147

120 014

3 Banco [9] 261 children with acute(non-trauma) illness age5 days to 15 years hos-pital paediatric clinicUSA

Touching various parts ofbody (544 forehead)

Oralrectal temp (378C383C electronic ormercury time NA)

176 7393446

856184215

5233465

939184196

506 030

4 Callanan [10] 179 infants with variouscomplaints agelt3 months ER USA

Touching site not men-tioned forehead temp(38C electronic)

128 9132123

788123156

3892154

984123125

431 011

5 Chaturvedi [11] 200 children reporting feverage 0ndash12 years hospitalpaediatric clinic India

Touching site not men-tioned oralrectal temp(375C380C mer-cury thermometer 2min)

340 7064868

40954132

38148126

7305474

119 072

6 Ernst [12] 100 children whose mothersused palpation to mea-sure temperature age1 month to 18 yearshospital outpatient clinicUSA

Touching site not men-tioned oralrectal temp(377C383Celectronic)

400 9003640

7334460

6923652

9174448

337 037

7 Graneto [13] 322 children age 10 yearsER USA

Touching site not men-tioned rectalTM temp(38C type of thermo-meter NA)

385 839104124

758150198

684104152

882150170

347 021

8 Hooker [14] 180 children age 0ndash4 yearsER USA

Touching site not men-tioned rectal temp(38C electronic)

818 765 810 775 348 024

9 Singhi [15] 301 children with variouscomplaints age 3 monthsto 2 years hospital pae-diatric clinic India

Touching site not men-tioned

oral or axillary temp(374C mercury ther-mometer time NA)

389 889104117

886163184

832104125

876163176

780 013

10 Whybrew [16] 862 children age 1month to16 years hospitalZambia

Touching of abdomenforehead and neck

aillary temp (378Cmercury thermometer3min)

245 936221236

436273626

385221574

948273288

166 015

Note ER Emergency room LRthorn likelihood ratio positive LR likelihood ratio negative NA not available NPV negative predictive value OPD Outpatientdepartment TM tympanic membrane temp temperature

BRIE

FREPORT

72

JournalofTropica

lPediatrics

Vol54No1

at University of Newcastle on September 25 2014 httptropejoxfordjournalsorg Downloaded from

being variability in the prevalence of fever lackof standardized methods of maternal touchingdifferent reference tests and variable cut-off levelfor fever

References

1 Schwartz N Guendelman S English P Thermometeruse among Mexican immigrant mothers in CaliforniaSoc Sci Med 1997451315ndash23

2 Einterz EM Bates ME Fever in Africa do patientsknow when they are hot Lancet 1997350781

3 Banco L Jayashekaramurthy S The ability ofmothers to read a thermometer Clin Pediatr199029343ndash5

4 Porter RS Wenger FG Diagnosis and treatmentof pediatric fever by caretakers J Emerg Med2000191ndash4

5 Robinson JL Jou H Spady DW Accuracy of parentsin measuring body temperature with a tympanicthermometer BMC Fam Pract 200563

6 Deville WL Buntinx F Bouter LM et al Conductingsystematic reviews of diagnostic studies didactic guide-lines BMC Med Res Methodol 20022

7 Alves JG Correia Jde B Ability of mothers to assessthe presence of fever in their children without using athermometer Trop Doct 200232145ndash6

8 Nwanyanwu OC Ziba C Redd SC et al Palpation as amethod of fever determination in Malawian children

who are less than 5 years old how reliable is itAnn Trop Med Parasitol 199791359ndash63

9 Banco L Veltri D Ability of mothers to subjectivelyassess the presence of fever in their children Am J DisChild 1984138976ndash8

10 Callanan D Detecting fever in young infants reliabilityof perceived pacifier and temporal artery temperaturesin infants younger than 3 months of age Pediatr EmergCare 200319240ndash3

11 Chaturvedi D Vilhekar KY Chaturvedi P et alReliability of perception of fever by touch Indian JPediatr 200370871ndash3

12 Ernst TN Philp M Temperature assessment byparental palpation Am J Dis Child 1985139546ndash7

13 Graneto JW Soglin DF Maternal screening of child-hood fever by palpation Pediatr Emerg Care199612183ndash4

14 Hooker EA Smith SW Miles T et al Subjectiveassessment of fever by parents comparison withmeasurement by noncontact tympanic thermometerand calibrated rectal glass mercury thermometer AnnEmerg Med 199628313ndash7

15 Singhi S Sood V Reliability of subjective assessment offever by mothers Indian Pediatr 199027811ndash5

16 Whybrew K Murray M Morley C Diagnosing feverby touch observational study BMJ 1998317321

17 Dunyo SK Kuorum KA Nkrumah FK Letter Lancet19973501550

18 Bergeson PS Stienfeld HJ How dependable is palpa-tion as a screening method for fever can touchsubstitute for thermometer readings Clin Pediatr197413350ndash1

19 Singh M Pai M Kalantri SP Accuracy of perceptionand touch for detecting fever in adults a hospital-basedstudy from a rural tertiary hospital in Central IndiaTrop Med Int Health 20038408ndash14

20 Valeiras C Calvo A Roca A et al Value of touch inthe diagnosis of fever and of paleness in the diagnosis ofanemia observational study in adults An Med Int200219120ndash5

21 Methodology Checklist 5 Diagnostic Studies In SIGN50 A Guideline Developersrsquo Handbook httpwwwsignacukguidelinesfulltext50checklist5html(6 December 2006 date last accessed)

22 Zamora J Muriel A Abraira V Meta-DiSc forWindows A Software Package for the Meta-analysisof Diagnostic Tests XI Cochrane ColloquiumBarcelona 2003 httpwwwhrcesinvestigacionmetadisc_enhtm (6 December 2006 date last accessed)

TABLE 2Summary statistics of diagnostic parametersa

Diagnostic parameters Summary statistics( 95CI)

Sensitivity 892 (873ndash909)Specificity 500 (479ndash520)Likelihood ratio positive 30 (20ndash45)Likelihood ratio negative 02 (01ndash03)Diagnostic odds ratiob 153 (74ndash319)

aStudy by Hooker excluded (numerical data cannot beextracted)bDiagnostic odds ratio is the odds of positive test resultsin participants with disease compared with the odds ofpositive test results in those without disease

BRIEF REPORT

Journal of Tropical Pediatrics Vol 54 No 1 73

at University of N

ewcastle on Septem

ber 25 2014httptropejoxfordjournalsorg

Dow

nloaded from