A Pictorial Sleepiness Scale Based on Cartoon Faces

INTRODUCTION

SLEEPINESS IS THE CRAVING OR DESIRE FOR SLEEP AND ISTHE PHYSIOLOGICAL DRIVE THAT PROMOTES THE ONSET OFSLEEP1 AND REFLECTS THE BALANCE BETWEEN SLEEPDRIVE AND AROUSAL.2 The degree of sleepiness may be estimatedobjectively using the multiple sleep latency test (MSLT), which mea-sures the time to sleep onset evident in the electroencephalogram.3 TheMSLT has been used extensively to measure the extent of daytimesleepiness in patients with sleep disorders, particularly narcolepsy, aswell as sleep propensity in research settings.4,5 The MSLT is sensitive topartial sleep deprivation and the cumulative effects of sleep restriction.6A modification of the MSLT, the Maintenance of Wakefulness Test, isused to quantify the ability to remain awake and has been used to eval-uate treatment efficacy in patients with excessive somnolence.7Although these electroencephalographic techniques provide the mostaccurate measurements of sleepiness, they are inappropriate for large-scale epidemiologic studies and some fieldwork.

Subjective perceptions of sleepiness are reasonable indicators ofphysiological sleepiness and may even precede performance decre-ments.8 Generally, we are able to recognize how sleepy we are, thoughmisjudgment of sleepiness may occur with prolonged sleep restriction.6Subjective measures of instantaneous perceived sleepiness, such asthose reflected in the Karolinska Sleepiness Scale (KSS)9 and theStanford Sleepiness Scale (SSS),10 are quick and simple to complete,provided that the subjects have some level of literacy and comprehen-sion of what is required. Visual analog scales (VAS),11-15 which are sim-

ple, cheap, and easy to execute, also have been used to assess sleepi-ness.8,9,16,17 However, indicating a response on a VAS requires someunderstanding of geometry and the ability to map an abstract phe-nomenon onto a VAS. The concept of a scale, even one as simple as theVAS, may be foreign to children at a prelogical cognitive stage of devel-opment. Also, preliminary studies in our laboratory have indicated thatpoorly educated and, especially, rural people fail to understand the con-cept of a VAS.

Conceptually, the measurement of perceived sleepiness is a task sim-ilar to the measurement of perceived pain. Pictorial scales, using cartoonfaces, are well established as instruments for measuring perceived painin subjects and patients for whom semantic or geometric scales are inap-propriate and, particularly, for children.18 We therefore designed a picto-rial scale, using cartoon faces, to measure perceived sleepiness and test-ed it using, as a subject base, a diverse South African adult populationthat included many subjects whose home language was not English andSouth African preschool and junior-school children. We also validatedthe scale against established scales and assessed its utility in practicalapplications.

MATERIALS AND METHODS

Subjects and Ethical Considerations

A cohort of 835 subjects participated (Table 1). Separate groups ofsubjects were used for each phase of development and testing. Subjectsfor the ranking procedure were recruited from patients attending a diag-nostic sleep laboratory and their guests, healthy adults employed at ouruniversity, and healthy children from local English-medium preschoolcare centers and junior schools. Subjects for the validation were recruit-ed from patients attending the same diagnostic laboratory for daytimestudies and healthy university students. The application phase consistedof subjects recruited from 2 English-medium junior schools and adultshift workers employed in the mining industry.

Each component of this project was approved by the Committee forResearch on Human Subjects of the University of the Witwatersrand.Informed verbal consent was obtained from all subjects of consenting

A Pictorial Sleepiness Scale Based on Cartoon Faces—Maldonado et alSLEEP, Vol. 27, No. 3, 2004 541

A Pictorial Sleepiness Scale Based on Cartoon FacesClaudia C. Maldonado,BSc(Hons); Alison J. Bentley,MBBCh; Duncan Mitchell, PhD

Wits Dial·A·Bed Sleep Laboratory, Brain Function Research Unit, School of Physiology, University of the Witwatersrand, Johannesburg, South Africa

Study Objectives: To develop a sleepiness scale devoid of semantic orgeometric elements. Design: Subjects were asked to rank in order 7 cartoon faces represent-ing degrees of sleepiness. We used Thurstone’s scaling procedure totransform these rankings into an interval scale, which allowed us to elim-inate 2 of the faces. The remaining 5 faces were ranked again using othersubjects. In a validation study, subjects rated their perceived level ofsleepiness using our scale and other sleepiness scales. Employed shift-workers and school-going children used our scale to assess its practicalapplicability.Settings: Research and diagnostic sleep laboratories, pre-primary to ter-tiary institutions, shift-working industry.Participants: Ethnically diverse healthy and sleep-disordered adults (n =490), and school-going children (n = 345).Measurements and Results: Our faces scale correlated with theKarolinska Sleepiness Scale (P < .05), the Stanford Sleepiness Scale (P< .04), and a visual analog scale measuring sleepiness (P < .0001).

Shiftworkers showed a time-on-task effect on the evening shift (P < .0001)and a peak in sleepiness at 4:00 and 5:00 (P < .0001) on the night shift.Eight to 10 year old children appeared sleepier than older childrenthroughout a school day (P ≤ .02) and became sleepier as the day pro-gressed (P < .0001). We confirmed that our scale measures sleepiness,uncontaminated by pain, anger, or happiness. Conclusions: We have devised a sleepiness scale suitable for people tooyoung or insufficiently educated to employ more-conventional scales. Weenvisage the scale being used for diagnostic, therapeutic, and researchpurposes.Key Words: subjective sleepiness, sleepiness scale, children, pictorialscale, faces scaleAbbreviations: MSLT, Multiple Sleep Latency Test; KSS, KarolinskaSleepiness Scale; SSS, Stanford Sleepiness Scale; VAS, visual analogscale; ESS, Epworth Sleepiness Scale; ANOVA, analysis of varianceCitation: Maldonado CC; Bentley AJ; Mitchell D. A pictorial sleepinessscale based on cartoon faces. SLEEP 2004;27(3):541-8.

Disclosure StatementNo significant financial interest/other relationship to disclose.

Submitted for publication October 2003Accepted for publication November 2003Address correspondence to: CC Maldonado, School of Physiology, University ofthe Witwatersrand, Private Bag 3, Wits 2050, South Africa. Tel: +27 11 717 2363;Fax: +27 11 643 2765; E-mail: [email protected]

Dow

nloaded from https://academ

ic.oup.com/sleep/article/27/3/541/2708011 by guest on 01 Septem

ber 2022

age and from school authorities for the children. For underage subjects,written informed consent was obtained from adults legally responsiblefor them.

Experimental Procedures

Ranking and Reduction of 7-Face Scale

Seven cartoon faces depicting degrees of sleepiness weredrawn by a specialist in medical graphic images. Particularattention was paid to the eyes, and the faces were devisedsuch that they were devoid of gender and ethnic features.The cartoons included 4 faces showing various degrees ofeye closure, yawning, a hand rubbing an eye, and a personasleep (Figure 1). The cartoon of each face (approximately55 mm x 60 mm), mounted on a magnet, was attached to aportable metal display board (640 mm x 640 mm). This con-struction allowed us to position the 7 faces randomly in 2dimensions to avoid any implied rank or order.

Subjects (Table 1) were told the purpose of the study,using terms appropriate for their age, language, and level ofcognition. Subjects were asked their age and with what eth-nic group they identified themselves or what their home lan-guage was. The procedure took approximately 5 minutes persubject. They were assured that the task was not a test, quiz,or game and that there was no right or wrong answer.Respondents were asked to rank the faces in order, from theface they considered to depict the person most awake to theperson most sleepy. Each subject’s response was anony-mous, confidential, and free from outside interference. Oncethe faces were ranked, each subject was asked to look at theassembly of faces in their selected order and was given achance to amend their ranking.

Thurstone’s scaling procedure (see below) was used totransform the individual rankings to a consolidated intervalscale for each group of subjects. On the basis of that analy-sis, 2 faces were eliminated from the original set. The


Figure 1—Interval positions of the 7 original faces, calculated by the Thurstone scaling procedure, from the rank order selected by subjects in 4 groups, from the face that subjects consid-ered to depict the person most awake to that which depicted the person most sleepy (total n = 202). A. Patients attending a diagnostic sleep laboratory. B. Patients’ guests who accompaniedthem to the sleep laboratory. C. Healthy schoolchildren aged 6 and 7 years. D.Ostensibly healthy adults employed by our university. The positions shown are relative to the face on the extremeleft.

Table 1—Characteristics of subjects

Phase Task Source Number Age Male, of subjects (mean ± SD (%)

or range)

Ranking Reduction of Diagnostic 109 44 ± 12 857-face scale sleep laboratory

Guests of patients 36 40 ± 14 45University 16 34 ± 10 57 Junior school 41 7 ± 1 100

Ranking of Diagnostic 20 45 ± 14 805-face scale sleep laboratory

University 20 - 30 Preschool care center 20 5 ± 1 35

Validation Comparison with Diagnostic 8 45 ± 11 75KSS & SSS sleep laboratory

University 26 18 - 41 38

Comparison with University 234 22 ± 5 36VAS andcontamination study

Application Drivers Mining industry 21 - 100 Grade-3 school Junior school 137 8 ± 1 48children Grade-7 school Junior school 147 13 ± 1 47children

KSS, Karolinska Sleepiness Scale; SSS, Stanford Sleepiness Scale; VAS, visual analog scaleSubjects identified themselves as South African Caucasian (English-speaking and Afrikaans-speaking), SouthAfrican Black (Tswana, Ndebele, Xhosa, Zulu, and Sotho), British, Portuguese, Chinese, Israeli, Indian,Polish, Italian, Turkish, Ethiopian, and Greek.

Dow



ber 2022

remaining 5 faces (Figure 2A) were modified slightly according to com-ments made by some subjects.

Ranking of 5-Face Scale

The 5 faces were ranked again in order of perceived sleepiness, usingdifferent subjects, from the same sources. During this secondary rank-ing, in order for us to be sure that the younger children understood theconcept of sleepiness, each child (aged 4 to 7 years) was asked to mimicsomebody feeling sleepy or, if they were older, whether they knew whatfeeling sleepy meant. When presented with all 5 faces together, theyoungest children seemed overwhelmed and often failed to respond tothe instructions. We then presented them with 2 faces and asked them toplace them on the board in order of sleepiness and added the faces 1 ata time until all the faces were ranked. The children then were permittedto revise the order if they so wished. Thurstone scaling again was usedto transform the individual rankings into an interval scale for each sub-ject group.

Validation

Validation Against Other Sleepiness Scales

Eight patients suffering from excessive daytime sleepiness who hadbeen referred to a diagnostic sleep laboratory and 26 healthy universitystudents paid honoraria for their participation participated in a compari-son of our faces scale with the SSS and KSS. Patients and students wereasked to rate their personal level of sleepiness or alertness using ourfaces scale (see Figure 2B), the SSS, and the KSS, with the scales pre-sented in a random order on 4 occasions during 1 day at 10:00 AM,12:00 noon, 2:00 PM, and 4:00 PM. All the subjects were fluent inEnglish and were given the opportunity to familiarize themselves withthe subjective scales before the testing procedure began.

For validation against a geometric scale, a large population of univer-sity students (n = 234) was asked to mark the 1 face that best describedhow sleepy or alert they felt at that time and to complete a 100-mm VAS,with anchors as alert as ever and as sleepy as ever.

Contamination with Other Affective States

The same university students participated in a protocol designed toassess to what extent the faces scale might be contaminated by the intru-sion of affective states other than sleepiness in perceptions of what thefaces depicted. On the basis of a recent recommendation made in respectof faces scales employed to assess pain,19 we first modified our scale to

eliminate the smile on the first 3 faces (neutral anchor, Figure 2C).Students were asked to rank our faces both with the smile anchor (Figure2B) as well as the neutral anchor (Figure 2C) in order of intensity in var-ious affective states, from most happy to most sad, from least pain tomost pain, from least angry to most angry, and from as alert as ever to assleepy as ever, with the possible contaminating states presented in ran-dom order. Due to the tedious nature of this task, not all the rankingswere performed by every subject. Thurstone scaling procedure wasapplied to each affective state.

Application

Subjects with Relatively Low Literacy Levels

Three groups of subjects were used for this phase. Heavy-vehicledrivers engaged in shift work in the mining industry used our scale(Figure 2B) to gauge their level of sleepiness during their routine work.The faces were presented in a linear horizontal array. Every hour, for theduration of a shift, each subject marked the face that best described howsleepy or alert he felt at that time. There were two 8-hour shift cyclesover a 24-hour day, the evening shift beginning at 4:00 PM and the nightshift beginning at midnight.

Children with Differing Ages

Two groups of school-going children used our scale to rank their levelof sleepiness on 4 occasions during a school day. The faces were pre-sented in a random 2-dimensional array, at 8:00 AM, 9:30 AM, 11:30AM and 12:30 PM. The children chose the face that best described howsleepy or alert they felt at that time. The assessments were administeredby the teacher in charge of that class.

Statistical Analyses

Thurstone Case V scaling procedure20 converts rankings into a zero-based interval scale, in which the linear distance between adjacent itemsrelates directly to the frequency of selection of the items in that order,rather than the reverse order. If items cluster on the scale, there is dis-agreement among subjects about their rank order, so they may be con-sidered interchangeable. Using the interval value derived by theThurstone procedure allows for the use of interval, rather than categori-cal, methods of statistical analyses. VAS values were normalized, for sta-tistical purposes, with the arcsin transformation. Pearson correlationswere used to compare the Thurstone scores from the faces scale to thescores from the VAS, and repeated-measures Pearson correlations are


Figure 2—A. The 5 faces remaining from the initial ranking. B. The same faces after the smile had been reduced on the face representing greatest wakefulness. C. The faces after the smilehad been eliminated from all faces, following similar adjustments to a scale used to assess pain.19

Dow



ber 2022

reported for the validation phase of the study. Most of the statistical testsemployed the statistical package GraphPad InStat (Graphpad Software,Inc., San Diego, Calif). Student t tests were used to compare regressionlines using the neutral and smile anchors, and χ2 statistics were used tocompare the measurement of various affective states using a neutral orsmile anchor on the faces scale. Probability values less that .05 wereconsidered significant. Sample sizes are shown in each case.

RESULTS

Ranking

Characteristics of the subjects who participated are shown in Table 1.Subjects ranged in age from 4 to 73 years and came from a wide varietyof backgrounds. For many, English was not their home language, but allcould communicate adequately in English.

Figure 1 shows the Thurstone scores generated from the ranking of thefaces by each of the 4 groups of subjects, when the original 7 faces werepresented to 202 subjects, asked to rank them in order of increasingdepicted sleepiness. Four of the faces, 3 of which represented stages ofapparent wakefulness rather than sleep, were ranked in the same orderby the 4 groups of subjects and were positioned in distinctly separatepositions on the Thurstone scale by all groups except university staff.The remaining 3 faces, representing apparent states of drowsiness, wereclustered on the Thurstone scale and were not ranked in the same orderby the 4 groups of subjects. We therefore eliminated 2 of the 3 faces and,on the basis of spontaneous comments made by subjects, retained theface representing drowsiness by a hand rubbing the eye (Figure 2A).Figure 2B and 2C shows the same faces after the smile first was reducedand then was eliminated to reduce confusion between alertness and hap-piness (or other affective states).

Figure 3 shows the Thurstone scores generated by the responses of all60 subjects (sleep-laboratory patients and healthy adults and children)ranking the remaining 5 faces in the improved format of Figure 2B.These 5 faces occupied distinct positions on the scale. Only 5% of thesubjects ranked the faces in an order different from that of Figure 3.

Validation

Patients and university students assessed their personal level of sleepi-ness using our faces scale and both the SSS and KSS at 4 times spreadover 1 day. However, no statistically significant circadian variation insleepiness was evident in the SSS and was evident only in the studentcohort for both our faces scale and the KSS (repeated measures analysisof variance [ANOVA] n = 25, P < .05, F3,72 = 2.817 for faces and P < .05Fr (nonparametric) = 11.794 for KSS. We therefore compared theThurstone scores of faces chosen averaged over the day for each subject,with the median score obtained from both the KSS and the SSS (Figure4A and 4B). In both cases, there was a very highly significant linear cor-relation between our faces scale and the 2 well-established sleepinessscales. More specifically, our faces scale showed a better correlationwith both the KSS and the SSS in the symptomatic patient population(KSS: r = 0.729, P = .0458; SSS: r= 0.946, P = .0011) than in thehealthy control group (KSS: r = 0.475, P = .0143; SSS: r = 0.412, P =.0364).

Also shown in Figure 4 is the linear correlation between our facesscale and a VAS, when a large cohort of university students assessedtheir instantaneous sleepiness on the scale concurrently. They employedour scale in the versions with a reduced smile (Figure 2B) and the smileeliminated (Figure 2C). Neither the slopes nor the intercepts of the lin-ear regression lines relating VAS to the Thurstone scores of the facesscales with the reduced and absent smiles differed significantly, imply-ing that, for numeric purposes, the 2 versions of our scales are inter-changeable. The Thurstone scores of the 2 versions of our scale corre-lated significantly with each other (r = 0.992, P = .0009). We have the-oretic reasons to prefer the scale without smiles though, and Figure 5shows a more-detailed analysis of the instantaneous sleepiness assessedconcurrently on the VAS and on the faces scale without smiles. Thisanalysis reveals that our faces scale covers virtually the full range ofsleepiness incorporated in a VAS anchored at as alert as ever and assleepy as ever and correlates very highly significantly, but in a slightlynonlinear way, with scores on the VAS.

Though the high correlation between our faces scale and the standardscales of sleepiness makes it unlikely that our scale measured anythingother than sleepiness, we considered the possibility that subjects mayhave confused the cartoon faces with affective states other than sleepi-ness. We therefore asked subjects to arrange the faces as if they relatedto other states. The Thurstone scores resulting from attempts by univer-sity students to rank the faces in order of other affective states, in addi-tion to sleepiness, are shown in Figure 6. The scale was nondiscrimina-tory except when used to assess sleepiness. Significantly more subjectswere able to rank all 5 faces according to sleepiness than according tohappiness, anger, or pain (Table 2).

Application

The sleepiness of heavy-vehicle drivers employed in the miningindustry, during different shift cycles, is shown in Figure 7. The driverssimply indicated which face best represented their instantaneous sleepi-ness at every hour of the shift, and we then calculated sleepiness scoresfrom the Thurstone interval value for those faces. Perceived sleepinessincreased progressively during the evening shift toward midnight andpeaked at 4:00 AM and 5:00 AM on the night shift. Maximum perceivedsleepiness was not significantly different between the evening and nightshift. The drivers were able to implement the procedure of selecting aface without disrupting their work schedule and delivered a full data set,our analysis of which delivers a plausible record of the sleepiness trendin the drivers. The data also illustrate the value of Thurstone scaling inconverting categorical to interval data, amenable to further analysis.

The drivers constituted a group of subjects precluded from usingsemantic or geometric scales by virtue of their limited education. Schoolchildren, some of whose ages precluded them from using more-sophisti-cated scales, also were able to relate to our faces scale and to indicatetheir sleepiness during the school day by picking a face (Figure 8). Onall 4 occasions during the school day, the younger children chose facesthat indicated that they were sleepier than the older children (t test, P ≤.02). The younger children reported progressively increasing sleepinessas the day progressed, whereas the older children were sleepier only at12:30 PM. Even so, both groups of children reported relatively low lev-els of sleepiness and never reached the levels of sleepiness reported by


Figure 3—Thurstone scaling of the rankings of 5 faces (Figure 2B), in order of perceived sleepiness based on the responses of 60 subjects selected from patients attending a diagnostic sleeplaboratory, ostensibly healthy adults employed by our university, and preschool children aged 4 to 7 years.

Dow



ber 2022

the vehicle drivers. Again, the faces scale delivered plausible assess-ments of sleepiness in a population too young to employ other scales.

DISCUSSION

Subjects differing widely in age and ethnicity were able to relate car-toon faces to perceptions of sleepiness. They could rank 5 faces consis-tently in order, from the face representing the greatest wakefulness tothat representing sleep. The cognitive complexity required for this rank-ing task was low. There was no semantic element requiring sophisticat-ed comprehension of English or any other language. Indeed, the subjectsrecruited to develop our scale spoke at least 16 different home lan-guages. The geometric skills required were confined to an ability to rankin order, a skill considerably simpler than the mapping required whenemploying a VAS. Children in the age range of 4 to 7 years, with littleor no reading or writing skills and limited vocabulary, were able to carryout the task successfully. Presenting the cartoon faces in a random 2-dimensional array and requiring the subjects then to arrange them in asingle dimension eliminated possible observer-imported bias on theorder subsequently chosen.

Our faces scale showed good agreement with established subjectivesleepiness scales, such as the SSS and the KSS, more so in somnolentthan healthy populations, and has the added advantage that no translationor explanation is required. Our scale also correlated well with positions


Figure 4—Correlation (linear regression with 95% confidence limits) between other sleepi-ness scale and the Thurstone scores of our faces scale, based on concurrent assessments ofprevailing sleepiness. A. Concurrent assessment of personal sleepiness over a day bypatients attending a diagnostic sleep laboratory (n = 8) and university students (n = 26)using our faces scale (Figure 2B) and the Stanford Sleepiness Scale. r = 0.68, P < .0001. B.Concurrent assessment, as in A, but with the Karolinska Sleepiness Scale. r = 0.58, P =.0003. C. Concurrent assessment of instantaneous perceived sleepiness by university stu-dents (n = 234) using our faces scale in the version of Figure 2B (smile present) and 2C(smile absent) and a visual analog scale (0 = “as alert as ever,” 100 = “as sleepy as ever”).95% confidence limits shown for the neutral anchor regression; slopes of the 2 regressionlines were not significantly different (t test, P = .59). For neutral anchor regression, r = 0.76,P < .0001. For smile anchor regression, r = 0.73, P < .0001.

Figure 5—Concurrent assessment of instantaneous personal sleepiness by 234 universitystudents on our faces scale with smile absent and a visual analog scale (mean ± SEM, 0 =as alert as ever, 100 = as sleepy as ever). Least nonlinear regression y = 60.8x0.5 – 52.2, r= 0.98, P = .0014, shown together with 95% prediction limits.

Dow



ber 2022

on a VAS that covered nearly three quarters of the spectrum between asalert as ever and as sleepy as ever. Our scale, though, does not requirethe abstract task of mapping a perception onto a geometric scale as isrequired with the VAS.

Thurstone scaling was employed to convert individual rankings by agroup of subjects into an interval scale representing both the group’sconsolidated rank order and the degree of agreement about that order

among the members of the group. That analytic approach initiallyallowed elimination of the 2 redundant faces from the assembly, reduc-ing the assembly to 5 elements. Thurstone scaling has another advan-tage: the position of each face on the scale is identified by a numeric dis-tance from the anchor point of the scale (which we chose to be the facedepicting a state of greatest alertness).

Applying our faces scale in an occupational setting showed its poten-tial industrial usefulness. The well-documented increased sleepinessobserved on the night shift,9,21-25 particularly from 4:00 AM to 6:00 AM,is reflected in our scale, and the increased sleepiness associated with atime-on-task effect24,26 also is reflected in drivers’ responses on our scalein the evening shift. Children, aged 7 to 14 years, could relate to our car-toon faces describing various states of sleepiness, and analysis of thefaces that they chose delivered plausible assessments of sleepiness,known to be low in preadolescents,27 particularly the maximum levels ofsleepiness noted at 12:30 PM,27 and of how sleepiness progresses overthe school day. Even though the younger children in our study indicatedgreater perceived levels of sleepiness than did the older children, thisphenomenon may very well reverse as the children mature into adoles-


Figure 6—Thurstone scaling of the rankings of the five faces from the set with smile absent (Figure 2C) (n = 232) in order of perception of sleepiness (D) and other affective states: A. Rankedfrom least pain to most pain. B. Ranked from least happy to most happy. C. Ranked from least angry to most angry. Faces tended to rank in the same order for all affective states, but theirpositions were distinct only for perceptions of sleepiness.

Table 2—Percentage of subjects who were able to rank all 5 facesaccording to sleepiness and other affective states, in the versions with-out a smile (Figure 2C)

State Sample Size Able to rank (%)

Sleepiness 232 99 Happiness 157 48 Anger 111 34 Pain 118 52

Significantly more subjects could rank the faces according to sleepiness than according tothe other affective states χ2(3) = 195.6, P <.0001

Dow



ber 2022

cence, and the rise in physiological sleepiness associated with pubertaldevelopment17,28 may be reflected in our faces scale. Using our facesscale in an occupational setting and in a school setting caused minimaldisruptions to work and allowed an excellent response rate.

Attempts by educated subjects to rank our faces according to otheraffective states, namely pain, happiness, and anger, showed that it is apoor measure of these affective states. Almost all (99%) of the subjectswere able to rank the faces according to sleepiness, whereas significant-ly fewer subjects were able to rank our faces scale according to happi-ness, anger, and pain, irrespective of the anchor used. Thus our scaledoes not appear to be contaminated by the other affective states; the statereflected by choice of a face indeed is sleepiness.

We envisage the faces scale being used for diagnostic, therapeutic,and research purposes. Patients or subjects can be presented with the 5cartoon faces in a random arrangement in a 2-dimensional or linear arrayand asked to point out the single face that best represents their currentperceived state of sleepiness. An equivalent procedure, employing a dif-ferent set of cartoon faces, has been used successfully in both therapeu-tic and research settings for the measurement of perceived pain in chil-dren, even as young as 3 years.18,19,29 VAS, with pictorial anchors at eachpole, have been used in sleep research (M. A. Carskadon, PhD, person-al communication)17 but still require subjects to master the geometricmapping. We have taken the functionality of our scale beyond thatafforded by simple choice of a face by employing the Thurstone proce-dure to convert the ranking of the faces, in order of sleepiness, to aninterval scale, thereby generating a number attached to each face, with-out assuming linearity. Other subject or patient groups could rank thefaces and so establish their own Thurstone scores, or the scores generat-ed by our subject groups could be employed. Thurstone scaling gener-ates values that depend on sample size as well as degree of agreement,but the responses of different subject groups each with its own Thurstonescale still may be compared, provided the scales correlate with eachother.

Our scale cannot have the sensitivity of more-sophisticated measuresof sleepiness, and we do not expect it to replace these measures.Electroencephalographic measures of sleep onset (for example, theMSLT or nighttime sleep recordings) are likely to remain the most reli-able measures of sleep propensity. Also, the MSLT continues to showdecreased sleep latencies with accumulating sleep loss, whereas subjec-tive perceptions of sleepiness (such as the SSS and VAS) plateau withaccumulating sleep loss,6 and our scale may well do so, too. Althoughthe MSLT is a highly reliable measure of daytime sleepiness in healthynormal subjects, provided there are more than 3 nap sessions,30 it andother electroencephalographic techniques9 are time consuming andexpensive to perform. Also, patients with insomnia and those with mildobstructive sleep apnea have variable sleep latencies determined largelyby the quality of sleep on the preceding night.30

The KSS,9 a useful subjective assessment of sleepiness, correlateswith the VAS9 and has been used successfully in research settings.9,25

The SSS is sensitive to sleep deprivation and correlates with perfor-mance measures8, 10 and also is sensitive to circadian dips in perfor-mance.8 The ESS,30 which records recent sleep-related history, is reliableover time and internally consistent and distinguishes normal adults frompatients suffering from excessive sleepiness.31 It cannot be used, howev-er, to detect instantaneous variations in sleepiness. The ESS correlateswell with nighttime polygraphic measures of sleep.32 Although someauthors have shown ESS scores to be modestly to highly reliably corre-lated with MSLT scores,33,34 other authors have found no correlationbetween the ESS and MSLT scores in patients suffering from sleep-dis-ordered breathing or periodic limb movement disorder.35 The ESS hasbeen used extensively in clinical and research adult populations and dis-tinguishes sleepy patients from healthy controls. 30,31,33 The ESS has beenused successfully in Italian and Portuguese subjects, (M Casagrande, CViolani, P Testa and G Curcio; A Pedroso, CG Rose, P Kadre, et al,reported at the meeting of the Associated Professional Sleep Societies inSan Francisco, Calif., June 1997) although translating and back-translat-ing the scale, which is the minimum requirement for translating seman-tic scales, is time consuming and demanding.

The commonly used subjective sleepiness scales were devised for themore-literate adult, but children may experience difficulty with theabstract task of describing the exactness of subjective experiences usingverbal cues.12 Children do better matching internal states with pictorialrepresentations of subjective states.19 A simple, quick, and inexpensivemeasure of subjective sleepiness also has potential application forassessing degree of sleepiness in industry, with little disruption of work.A pictorial scale will therefore be particularly beneficial for use in illit-erate or non-English-speaking workers and in children.

Our proposed pictorial scale is simple, quick to use, and requires onlylimited instruction. We predict the pictorial scale could be used to mea-


Figure 7—Sleepiness (mean + SD) of heavy-vehicle drivers (n = 9 evening shift, n = 12night shift) at hourly intervals over two 8-hour shift cycles. The sleepiness scores were cal-culated by assigning the Thurstone interval value of each face in Figure 3 to the face cho-sen by the driver to best represent his sleepiness at the time, with low scores indicatinggreater wakefulness. The drivers on the evening shift showed significantly increasingsleepiness towards midnight (repeated measures analysis of variance, F8,64 = 13.49, P <.0001). A peak in sleepiness occurred at 4:00 AM and 5:00 AM for the drivers on the nightshift (repeated measures analysis of variance, F8,88 = 5.61, P < .0001). Maximum sleepi-ness on this scale = 5.58.

Figure 8—Sleepiness (mean ± SD) of Grade 3 (age 8.4 ± 0.5 years, clear squares) andGrade 7 (age 12.5 ± 0.7 years, solid circles) school children showing degree of sleepinessover a school day. Sleepiness scores were calculated from the faces that the children choseto represent instantaneous sleepiness, as in Figure 7, with low scores indicating greaterwakefulness. The Grade 3 children reported sleepiness that increased progressively as theday progressed (repeated measures analysis of variance, F3,544 = 7.56, P < .0001), where-as the older children were significantly more sleepy only at 12:30 PM (repeated measuresanalysis of variance , F3,438 = 4.25, P < .0001).

Dow



ber 2022

sure subjective sleepiness in children as young as 4 years. This tool alsocould be used for epidemiologic studies or fieldwork and wherever ver-bal or cognitive limitations hamper communications. The sensitivity ofthe pictorial scale to sleep disruption and other interventions known toincrease sleepiness, and its utility in tracking improvements in patientstreated for sleep disorders causing excessive daytime sleepiness, are stillto be explored. The ultimate usefulness of the scale will depend on thediversity of situations in which it can be applied successfully. Weencourage other interested researchers and sleep medicine professionalsto test the scale in their particular fields of interest; copies of our cartoonfaces are available on request.

ACKNOWLEDGEMENTS

We thank Terry Borain of the Photo Illustration Unit of the Faculty ofHealth Sciences, University of the Witwatersrand, for the design of thefaces; Willie De Bruin for constructing the test array; Yvonne Glazer forassisting with data collection in the patient populations; LeanneKatzenellenbogen for collecting some of the data of Figure 4; AlisonSaunders, Brendon Billings, Praneshri Pillay, and Stacey Ramsey forcollecting some of the data of Figure 4, 5, and 6; Schu Schutte forarranging the data collection for Figure 7; the teachers involved in thevarious phases of the study; and Helen Driver for comments on themanuscript. Our research was funded by the University of theWitwatersrand.

REFERENCES

1. Pivik RT. The several qualities of sleepiness: psychological considerations. In: MonkTH, ed. Sleep, Sleepiness and Performance. Chichester, UK: John Wiley and Sons Ltd,1991:3-37.

2. Bonnet MH, Arand DL. Sleepiness as measured by modified multiple sleep latency test-ing varies as a function of preceding activity. Sleep 1998;21:477-83.

3. Carskadon MA. Measuring daytime sleepiness. In: Kryger MH, Roth T, Dement WC,eds. Principles and Practice of Sleep Medicine, 2nd ed. Philadelphia: WB SaundersCompany; 1994:961-6.

4. Carskadon MA, Wolfson AR, Acebo C, Tzischinsky O, Seifer R. Adolescent sleep pat-terns, circadian timing, and sleepiness at a transition to early school days. Sleep1998;21:871-81.

5. Randazzo AC, Muehlbach MJ, Schweitzer PK, Walsh JK. Cognitive function followingacute sleep restriction in children ages 10-14. Sleep 1998;21:861-8.

6. Carskadon MA, Dement WC. Cumulative effects of sleep restriction on daytime sleepi-ness. Psychophysiology 1981;18:107-13.

7. Mitler MM, Gujavarty KAS, Browman CP. Maintenance of wakefulness test: apolysomnographic technique for evaluating treatment efficacy in patients with excessivesomnolence. Electroencephalogr Clin Neurophysiol 1982;53:658-61.

8. Dinges DF, Pack F, Williams K, et al. Cumulative sleepiness, mood disturbance, andpsychomotor vigilance performance decrements during a week of sleep restricted to 4-5hours per night. Sleep 1997;20:267-77.

9. Åkerstedt T, Gillberg M. Subjective and objective sleepiness in the active individual. IntJ Neurosci 1990;52:29-37.

10. Hoddes E, Zarcone V, Smythe H, Phillips R, Dement WC. Quantification of sleepiness:A new approach. Psychophysiology 1973;10:431-6.

11. Aitken RCB. Measurement of feelings using visual analogue scales. Proc Royal SocMed 1969;62:989-93.

12. Herbert M, Johns MW, Dore C. Factor analysis of analogue scales measuring subjectivefeelings before and after sleep. Br J Med Psychol 1976;49:373-9.

13. Bond A, Lader M. The use of analogue scales in rating subjective feelings. Br J MedPsychol 1974;47:211-8.

14. Lee KA, Hicks G, Nino-Murcia G. Validity and reliability of a scale to assess fatigue.Psychiatry Res 1991;36:291-8.

15. Cronin AJ, Keifer JC, Davies MF, King TS, Bixler EO. Postoperative sleep disturbance:influence of opioids and pain in humans. Sleep 2001;24:39-44.

16. De Gennaro L, Ferrara M, Curcio G, Bertini M. Visual search performance across 40 hof continuous wakefulness: measures of speed and accuracy and relation with oculomo-tor performance. Physiol Behav 2001;74:197-204.

17. Fallone G, Owens JA, Deane J. Sleepiness in children and adolescents: clinical implica-tions. Sleep Med Rev 2002;6:287-306.

18. Bieri D, Reeve RA, Champion GD, Addicoat L, Ziegler JB. The faces pain scale for theself-assessment of the severity of pain experienced by children: development, initial val-idation, and preliminary investigation for ratio scale properties. Pain 1990;41:139-50.

19. Chambers CT, Craig KD. An intrusive impact of anchors in children’s faces pain scales.Pain 1998;78:27-37.

20. Thurstone LL, Chave EJ. The Measurement of Attitudes. Chicago: University ofChicago Press; 1929.

21. Torsvall L, Åkerstedt T. Sleepiness on the job: continuously measured EEG changes intrain drivers. Electroencephalogr Clin Neurophysiol 1987; 66:502-11.

22. Kecklund G, Åkerstedt T. Sleepiness in long distance truck driving: an ambulatory EEG

study of night driving. Ergonomics 1993; 36:1007-17.23. Horne JA, Reyner LA. Driver sleepiness. J Sleep Res 1995;4(supp 2):23-9.24. Folkard S. Black times: temporal determinants of transport safety. Accid Anal Prev

1997;29:417-30.25. Åkerstedt T. Work hours, sleepiness and the underlying mechanisms. J Sleep Res

1995;4(supp 2):15-22. 26. Dawson D, Reid K. Fatigue, alcohol and performance impairment. Nature

1997;388:235.27. Palm L, Persson E, Elmqvist D, Blennow G. Sleep and wakefulness in normal preado-

lescent children. Sleep 1989;12:299-308.28. Carskadon MA, Harvey K, Duke P, Anders TF, Litt IF, Dement WC. Pubertal changes in

daytime sleepiness. Sleep 1980;2:453-60.29. McGrath PJ, Unruh AM. Measurement and assessment of paediatric pain In: Wall PD,

Melzack R, eds. Textbook of Pain 3rd ed. Edinburgh: Churchill Livingstone; 1994.30. Johns MW. A new method for measuring daytime sleepiness: the Epworth Sleepiness

Scale. Sleep 1991;14:540-5.31. Johns MW. Reliability and factor analysis of the Epworth Sleepiness Scale. Sleep

1992;15:376-81.32. Zwyghuizen-Doorenbos A, Roehrs T, Schaefer M, Roth T. Test-retest reliability of the

multiple sleep latency test. Sleep 1988;11:562-5.33. Johns MW. Sleepiness in different situations measured by the Epworth Sleepiness Scale.

Sleep 1994;17:703-10.34. Chervin RD, Aldrich MS, Pickett R, Guilleminault C. Comparison of the results of the

Epworth sleepiness scale and the multiple sleep latency test. J Psychosom Res1997;42:145-55.

35. Benbadis SR, Mascha E, Perry MC, Wolgamuth BR, Smolley LA, Dinner DS.Association between the Epworth sleepiness scale and multiple sleep latency test in aclinical population. Ann Intern Med 1999;130:289-92.


Dow



ber 2022

A Pictorial Sleepiness Scale Based on Cartoon Faces

Documents

Transcript of A Pictorial Sleepiness Scale Based on Cartoon Faces