sf 36 health questionnaire: ii. responsiveness to health ... · sf 36 questionnaire to changes in...

Quality in Health Care 1994;3:186-192

SF 36 health survey questionnaire: II.Responsiveness to changes in health status in fourcommon clinical conditions

Andrew M Garratt, Danny A Ruta, Mona I Abdalla, Ian T Russell

Department ofPublicHealth, University ofAberdeenAndrew M Garratt,research fellowDanny A Ruta, lecturerHealth ServicesResearch Unit,University ofAberdeenMona I Abdalla, researchfellowIan T Russell, directorCorrespondence to:Mr AM Garratt,Department of PublicHealth, University ofAberdeen, PolwarthBuilding, Foresterhill,Aberdeen AB9 2ZD

Accepted for publication30 September 1994

AbstractObjective-To assess the responsivenessofthe SF 36 health survey questionnaire tochanges in health status over time for fourcommon clinical conditions.Design-Postal questionnaires at baselineand after one year's follow up, with tworeminders at two week intervals ifnecessary.

Setting-Clinics and four training generalpractices in Grampian region in the northeast of Scotland.Patients-More than 1700 patients aged 16to 86 years with one offour conditions: lowback pain, menorrhagia, suspected pepticulcer, and varicose veins; and a randomsample of 900 members of the localgeneral population for comparison.Main measures-A transition questionmeasuring change in health and the eightscales ofthe SF 36 health survey question-naire; standardised response means

(mean change in score for a scale dividedby the standard deviation of the change inscores) used to quantify the instrument'sresponsiveness to changes in perceivedhealth status, and comparison of patientscores at baseline and follow up with thoseofthe general population.Results-The response rate exceeded 75%in a patient population. Changes acrossthe SF 36 questionnaire were associatedwith self reported changes in health, asmeasured by the transition question.The questionnaire showed significant im-provements in health status for all fourclinical conditions, whether in referred or

non-referred patients. For patients withsuspected peptic ulcer and varicose veinsthe SF 36 profiles at one year approximateto the general population.Conclusions-These results provide thefirst evidence of the responsiveness of theSF 36 questionnaire to changes inperceived health status in a patientpopulation in the United Kingdom.(Quality in Health Care 1994;3:186-192)

IntroductionThe need for measures of outcome for use inthe health service is widely recognized.However, to date most of the focus has beendirected towards establishing the validity andreliability of instruments rather than theirresponsiveness or sensitivity to changes inhealth over time.' 2 Since the purpose of an

outcome measure is to quantify the effect ofhealth care, any instrument that is intended foruse as an outcome measure in the healthservice must be shown to be responsive toclinically important changes in patients'perceived health.Although generally agreed methods exist for

assessing validity and reliability, there is lessconsensus in assessing the responsiveness ofoutcome measures. Previous work has relatedchange scores on outcome measures toexternal criteria and processes of care. Thedegree of concordance between change scoresand external criteria has been assessed withtransition questions,3 4 which ask the patient orclinician whether there has been a change inhealth over some given period and in clinicaldata - for example, erythrocyte sedimentationrate in rheumatoid arthritis.5 Several studiesexamined changes in outcome variables afterinterventions ofknown efficacy, including totaljoint arthroplasty6 and total hip arthroplasty.7Different methods have also been proposed forquantifying the responsiveness of outcomemeasures. These include receiver operatorcharacteristic curves,4 responsiveness indices,9effect sizes,'0 and standardized responsemeans.6The SF 36 health survey questionnaire" 12

has been shown to be valid and reliable in theUnited States'2 13 and in populations in theUnited Kingdom."'4-' However, only twopublished studies in the United Statesexamined the ability of the questionnaire todetect changes in health status over time.7 18 Inthe first study7 the 108 item sickness impactprofile,'9 previously shown to be responsive bydemonstrating significant improvements inpatients undergoing joint arthroplasty,20 wasadministered with the SF 36 questionnairepreoperatively and postoperatively to patientsundergoing total hip arthroplasty. The SF 36questionnaire was as responsive as the longersickness impact profile to changes in healthstatus after surgery. In the second study, the SF36 questionnaire was administered to patientsbefore and after heart valve replacementsurgery.'8 Patients' scores were compared withscores for the general population aftercorrecting for sex and age. Before surgery,compared with the general population, thepatients had lower scores across the eight scalesof the SF 36 questionnaire and scoredparticularly low on the scales of physicalfunctioning, role limitations, and energy andfatigue. One month after surgery patients'

186

on April 22, 2021 by guest. P

rotected by copyright.http://qualitysafety.bm

j.com/

Qual H

ealth Care: first published as 10.1136/qshc.3.4.186 on 1 D

ecember 1994. D

ownloaded from

http://qualitysafety.bmj.com/

Responsiveness ofSF 36 health survey questionnaire to changes in health status

scores remained below those of the generalpopulation on all but the scale of general healthperception, and six months after surgery theyapproximated to those of the generalpopulation on all but the scales of rolelimitations.

If the SF 36 questionnaire is to be adoptedas a measure of outcome for use in the NHSits ability to detect changes in health status forpatient populations in the United Kingdomneeds to be assessed. We report the results ofa study in which the questionnaire wasadministered to a large sample of patients inthe United Kingdom with four common con-ditions - low back pain, menorrhagia, sus-pected peptic ulcer, and varicose veins - whowere followed up for a year. For the eight scalesof the questionnaire change scores were calcu-lated and compared with patient responses ona transition question. A standardized measurewas used to quantify responsiveness, andpatient scores at baseline and follow up werecompared with those of the generalpopulation.

MethodsSAMPLING AND DATA COLLECTIONThe methods of identifying and recruitingpatients were as described in the previouspaper (p 180)17 in relation to study 1. In thispaper we assessed the responsiveness of the SF36 questionnaire by sending a follow up ques-tionnaire to patients taking part in study 1 atone year. A random sample of 900 members ofthe general population, selected from theelectoral register for Aberdeen, served as acomparison group; they were sent a similarquestionnaire.

STATISTICAL ANALYSIS

Several methods have been proposed forquantifying the responsiveness of outcomemeasures.4 6 8 9 The standardised responsemean - that is, the mean change in score fora scale, divided by the standard deviation of thechange in scores - allows statisticallymeaningful comparisons to be made betweeninstruments. Higher standardised responsemeans indicate a greater effect or clinicallyimportant change, with standardised responsemeans of 0.2, 0.5, and 0.8 or aboverepresenting small, moderate, and large clinicalchanges respectively.6The SF 36 questionnaire contains a

transition question which is not used to scoreany of the eight scales. This question("Compared to one year ago, how would yourate your health in general now: much better,somewhat better, about the same, somewhatworse, much worse?") was used in this studyas a criterion by which to judge the responsive-ness of the questionnaire. Such questions area valid way of measuring changes in perceivedhealth2' and were used to assess the responsive-ness of instruments designed to measure out-comes.3 4 22 23 For the SF 36 questionnaire tobe a valid measure of outcome which reflectsperceived changes in health status a significantrelation would be expected between changeson the eight scales over the year and the

responses to the transition question. Patientsindicating an improvement in health on thetransition question would be expected to havehigher standardized response means across theeight scales than patients who stated that theirhealth remained the same.Although standardized response means

represent a measure of responsiveness of aninstrument, a standard is required thatfacilitates interpretation of scale scores andchanges in those scores. To fulfil this purposenormative data can be collected for general ordisease free populations."5 16 24 We comparedpatient scores on the SF 36 questionnaire atbaseline and follow up with data for the generalpopulation of Aberdeen after correcting forage, sex, and socioeconomic status. Ordinaryleast squares regression25 was used to estimatethe effect on each of the eight scales for eachof the four conditions, age, sex, and socio-economic status (housing tenure and age onleaving full time education) 16 24

Referred patients have a lower perceivedhealth status than those being managed solelyin general practice. 16 In this study wecompared referred and non-referred patientsfor changes in their perceived health status overone year. Regression analysis was used toestimate the effect of referral on health statusat baseline and follow up for each of the fourconditions after correcting for age, sex andsocioeconomic status.The results are presented as absolute and

standard scores at baseline for a member of thegeneral population with average characteristicsand at baseline and after one year's follow upfor members of the four patient groups withthe same characteristics. Absolute scores,presented as mean deviations from the scoresof the general population, allow comparisonsbetween the patient groups and the generalpopulation for each individual SF 36 scale.Standard scores are calculated by dividing thedifferences between the scores for each patientgroup with a specific condition and for thegeneral population by the standard deviation ofthe score for general population. Presented asline graphs, these standard scores allowcomparisons between the patient groups andthe general population at baseline and after oneyear's follow up across the entire SF 36 healthquestionnaire profile.

ResultsRESPONSE RATE

The results of the recruitment of patients aredescribed in the previous paper (p 180) inrelation to study 1. Of the 1148 patientsremaining in the study at one year, 240 failedto respond to the follow up questionnaire, 101refused to take part, and 32 questionnairescould not be delivered by the post office. Of the775(67-5%) patients who returned acompleted questionnaire at one year, 759attempted the SF 36 questionnaire at both oneyear and baseline. Of the comparison sampleof 900 members of the general population,542(60.2%) returned a questionnaire. Theirmean age has 47 9 years (range 18-91 years)and 292(53-9%) were female.

187



j.com/

Qual H


ecember 1994. D

ownloaded from


Garratt, Ruta, Abdalla, Russell

Table 1 Mean changes in SF 36 scales during one year by self reported health transition over one year*

Reported transition* No of Phvsical Social Role Role Mental Energy and Pain Generalsubjects functioning functioning limitation- limitation- health fatigue health

physical emotional perception

Much better 148 10-50 23-50 44-33 26-57 13-81 21-59 34-11 10-89Standard deviation 22-91 26-02 44-83 48-70 19-99 18-73 27-50 17-9095% Confidence interval 6-76 to 14-23 19-23 to 27-77 36-94 to 51-71 18-52 to 34-62 10-51 to 17-10 17-74 to 25-43 29-59 to 38-63 7-99 to 13-80SRM 0-46 0-90 0-99 0-55 0-69 0 93 1-24 0-61

Somewhat better 173 7-60 12-62 26-84 17-47 7-29 9-31 16-95 3-66Standard deviation 18-77 24 10 42-35 42-75 16-85 15-51 26-84 14-7395% Confidence interval 4-78 to 10-42 8-99 to 16-25 20-38 to 33-28 10-92 to 24-02 4-74 to 9-83 6-48 to 12-14 12-88 to 21-01 1-46 to 5 88SRM 0-40 0-52 0 63 0-41 0-43 0-50 0-63 0-25

About the same 332 2-69 4-82 10-82 6-17 1-39 4-37 10-14 1-33Standard deviation 16-83 20 80 41-71 43 16 15 72 17 51 23-51 13-4795% Confidence interval 0-87 to 4-51 2-57 to 7-08 6-24 to 15-39 1-46 to 10-89 -0-32 to 3.10 2-46 to 6-27 7-57 to 12 71 0- 13 to 2-78SRM 0-16 0-23 0-26 0-14 0(09 0-25 0-43 0-10

Somewhat worse 74 -8-89 -9-60 -10-76 -5-71 -2-01 -6-53 0-70( 9-84Standard deviation 19-77 23-49 36-30 48-14 19-56 19-81 22-18 19-52950/0 Confidence interval -13-47 to -4-31 -15-04 to -4-16 -19-29 to -2-23 -17-19 to 5-77 -6-64 to 2-62 111-15 to -1-91 4-57 to 5-87 14-36 to 5 39SRM -0-45 -0-41 -0-30 -0-12 -0-10 -0-33 0-03 ()-50

Muchworse 22 -16-83 -21-52 -3-57 -7-94 -3-46 -7-14 1-05 6-35Standard deviation 23-56 34-40 24-09 39-31 17-90 15-54 23-28 18-46950/0 Confidence interval -27 28 to -6-39 -37-18 to -5-86 --14 54 to 7-39 -25-83 to 9-96 -11-39 to 4-48 -14 22 to -0 07 -9 54 to 11.66 14 53 to 1 83SRM -0-71 -0-63 -0-15 -0-20 -0-19 -0-46 0-05 -0-34

*Question 2 of the SF 36 questionnaire: "Compared to one year ago how would you rate your health in general now?"SRM = Standardised response mean.Changes in all eight scales are linearly related to transition scores (F test for linear trend25; p < 0-0001).

CHANGES IN HEALTH STATUS

Table 1 shows the mean changes on each SF36 scale by self reported health transition atone year. Of the 749 patients who attemptedthe transition question, 321(42-9%) perceivedtheir health as better, 332(44-3%) stated thattheir health had remained the same and96(12-8%) stated that their health had gotworse over the year compared to baseline.The relation between changes in each of the

eight SF 36 scales and the responses to thetransition question was highly significant(table 1). For example, patients who perceivedtheir health as much better compared with one

year previously showed an average improve-ment of 10 50 on the physical functioning scalewhereas those who perceived their health as

much worse showed an average deteriorationof -16-83 on this scale. For patients whoperceived their health as much betterstandardised response means representinglarge changes in health were seen for four of theeight scales and moderate standardisedresponse means for the remainder, with

physical functioning on the borderline. Forpatients who perceived their health as

somewhat better standardised response means

of a moderate level were seen for four of thescales, the remainder representing smallchanges. For patients who perceived theirhealth as about the same standardized response

means representing a small change were seen

for four scales. For patients who perceivedtheir health as being somewhat worse

standardized response means representing a

small negative change were seen for five of thescales. Finally, for patients who perceived theirhealth as much worse at one year, standardisedresponse means representing a moderatenegative change were seen for two of the scales,the remainder representing a small negativechange or less (table 1).To facilitate interpretation of baseline and

follow up scores they were compared with datafor the general population of Aberdeen. Themean baseline scores for the generalpopulation and the mean score deviations forthe condition specific groups after correcting

Table 2 Mean SF 36 questionnaire scores for general population and mean score deviations for condition specific groups (mean changes at one year'sfollow up) after correctingfor score age, sex, and socioeconomic status

Group No of Physical Social Role Role Mental Energy and Pain Generalsubjects functioning functioning limitation- limitation- health fatigue health

physical emotional perception

General population 542 79-2 78-6 76-5 75-0 73-7 61-2 76-9 68-7Patients with:Low back pain 322 -27.4*(5.4*) -23.3*(9.5*) -57 7*(23-8*) -31-1*(10-6*) -13-7*(4-4*) -20-2*(5.6*) 43-8*(19 0*) -,13-2*(1 0)Menorrhagia 168 -1 6(0-00) -12.3*(8-0*) -31-3*(14-6*) -27.9*(15.8*) -12-3*(6.3*) -18.5*(11.8*) -20.9*(9.9*) -10.9*(3.41)Suspectedpeptic ulcer 122 +2-7(1-7) -6.3¶(10.7*) -15 1*(18.5*) -6-8(10-7t) -7.5*(5.5t) -11.3*(10-1*) -23.0*(19.5*) 7.6*(3.9¶)Varicose veins 157 -4.1¶(5.0*) +1-3(2-9) -14.3*(1I.3t) -8-6*(9.O¶ 0-0(2-7¶) -2-9(3-9t) -7.8*(9-0t) -I 3(3-it)

Referred patients with:Low back pain 147 -34.5*(5.9t) -28-0*(8 7*) -60-7*(18-1*) -39-7*(14-4*) -18-3*(6-3t) -25-6*(6-4t) -46.2*(15-0*) -17.5*(0.5)Menorrhagia 125 -1-2(-0-0) -13-8*(8.4*) -35-4*(16-9*) -31.2*(16-3*) -13-7*(7-2*) -20-9*(13-0*) -24-3*(12-0*) -13-0*(5.2t)Suspectedpepticulcer 58 +1-5(1-4) -10-3t(13-2*) -21*5*(17-3t) -9-2(12-91) -10-7*(8.2t) -18-1*(15-4*) -27-9*(24.8*) 10.2*(9-0t)Varicoseveins 116 -5.0¶(5.7*) +0-8(3-5) 15.5*(13.3*) -9-3T(10-8t) -0-7(3-5¶) -4-5T(5-0t) -8-8*(10.1*) -1 4(3-5t)

Non-referred patients with:Low back pain 175 -21-4*(4-9t) -19-2*( 0-3*) -55.6*(28-4*) -23.2*(7.511) -9-8t(2-81) -1 5-7t(5-0t) -41-9t(22-2*) -9-6t(1 -4)Menorrhagia 43 -3-7(0-0) -9-6¶(6-6) -21-5*(7-5) -15-7¶(14-5¶) 7-7t(3-9) -1-9t(8-3¶) -11-5t(3-6) -6-811(-2-3)Suspected peptic ulcer 64 +3-8(2-0) -2.4(8.3*) -9-71(19-5t) -4-3(8-7) -4-6(3-0) -5-2¶(5-21D -18-4t(14-4*) -4-9(-0-6)Varicose veins 41 -1-2(2-5) +2-5(0-8) -10-7(4-0) -4-5(2-2) 2-8(-0-1) -2-5(-0-2) -4-4(4-9) -1-6(1-7)

Mean score for a given patient group can be calculated by adding the mean deviation to the mean score for the general population. Adding the change scorein parentheses gives the mean score at 12 months' follow up.Significant score differences from the general population and significant change scores: Ip < 0-05; t# < 0-01; tp < 0-005; *p < 0-001.Underlined change scores refer to comparisons of change scores for referred and non-referred patients with the same condition (single underlining denotes p < 0-05and bold underlining p < 0-01).

188



j.com/

Qual H


ecember 1994. D

ownloaded from



05 for age, sex, and socioeconomic status areLow back pain shown in table 2. The absolute change scores60% in brackets showed that significant improve-0.0 50% ments occurred for all four patient groups40% across seven of the eight SF 36 scales. The

a-05 wow Z 30% scores for patients with back pain improved0 * significantly across seven scales, but they did0~~~~~~~~~~~~~~~0CO 210 a/ \\//°%@ not show any change on the scale of general

\a) health perception. Patients with menorrhagia- 10% 0 and peptic ulcer showed significant improve-

°n-1.5-/-

/ ments in scores on all but the physicalo-9 Baseline A! 1 5% functioning scale, a scale on which they did not

-2 0 0-0 One year's follow up differ significantly from the general population- General population at baseline. Patients with varicose veins showed

1% significant improvements in scores for seven-2.5 I I I I Iscales but not for social functioning, on which05 Menorrhagia they did not differ significantly from the

_r60% general population at baseline.0.0 50% Figure 1 shows the standard scores for anv.0 50%

average member of the general population and40% the same patient with each condition in turn at

a,-0.5 _ B O XJD30%baseline and one year. The mean for the0 * general population was set at zero on each0~~~~~~~~~~~~~~~0 *r-_. health scale, allowing comparisons to be made

a,( in terms of standard deviations of the generalC_ 10% 0 population. For example, patients with lowW -1 5 _-_ 5% back pain with a mean standard score of about

-2 for pain at baseline corresponded to the

-2 0 lowest scoring 2.5% of the general population.The figure serves to highlight the scales on

L1 s|A% which the patient groups showed the greatest-2.5 improvement, represented by those areas of0.5 Suspected peptic ulcer health that get closer to the standard score of

Suspected 60% zero for the general population. The largest0.0 50% areas of improvement included the scales of

40%pain and role limitations attributable to

|l40% physical problems for low back pain; energyw --0 5 r ~ Ax30% and fatigue and pain for menorrhagia; pain ando * role limitations attributable to physical

2-10%\ */.problems for suspected peptic ulcer; and painCD' and role limitations attributable to physical

C _ 10% I' problems for varicose veins. After one year theX -1.5

- 5% health profiles for patients with low back painand menorrhagia still deviated considerably

-2-0 from the health profile of general populationwhereas the profiles for patients with suspected

L| l1% peptic ulcer and varicose veins more closelyapproximated to the profile of the general

0.5 Varicose veins population (fig 1)._ 60% Table 2 shows the mean score deviations and

° °0\-== 50% change scores for the referred and non-referredpatient groups. The same data are presented as

40~Ad-°% standard scores in figure 2. Although signifi-a-0 5 _ _ 30% cant improvements in SF 36 scores were seen

0 o _20% * for referred and non-referred patients withV -10 _ @ back pain and menorrhagia, the health profiles

of these two groups of patients at one year wasCo _10% still appreciably lower than the health profile

C) -1 5 5% for the general population. However, the healthprofiles of referred and non-referred patients

-2.0 with suspected peptic ulcer and varicose veinsat one year were much closer to the health

-25 1% profileof the general population.Physical Social Role Role Mental Energy Pain General Compared with non-referred patients,

functioning functioning limitations limitations health and health patients referred with menorrhagia, suspectedphysical emotional fatigue perception peptic ulcer, and varicose veins had a

Questionnaire scale significantly higher level of improvementFig1 MeanSF36healthprofilesatbaselineandoneyear'sfollowupforgeneral across the SF 36 scales at one year, forpopulation and patients with one offour study conditions (*centiles ofgeneral population:for example, X% ofgeneral population falls below Xth centile and remaining members fall menorrhagia on the scale of general healthabove that point) perception and for suspected peptic ulcer on

189

w-w rev -a -->i

i



j.com/

Qual H


ecember 1994. D

ownloaded from



_-* Baselineo-o One years follow up- General population

0.5

u ua)

c8 -0.5

cn-2

1.0

CU 1.5U)

-2.0

2.5 --

0.5

n r , L

-0.5

-1 .0

-1.5

-2.0

-2.5

0.5

0.0

-0.5

Referred patients with lowback pain 60%

50%40%30%20%

10%5%

*Cf)

cCD(a)

1%

Referred patients withmenorrhagia

-1.0

-1.5

-2.0

-2.5

Referred patients withsuspected peptic ulcer

60%50%40%30%20%

10%5%

*

Q)a,

a)0

0.5- Non-referred patients withlow back pain

a, °-054 /8 -0.5

-1.0CUCU1.5

-2.0

-2-5

0 5 r Non-referred patients withmenorrhagia

0.0a,° -0.5

-1-.0_CD) 1.0

C:C15U)

1%

60%50%40%30%20%

10%5%

*

a)0

a)0co0

E

Cl)

0.5

0.0

-0.5

-1.0 F

-1 .5

-2.0

d 1%

Non-referred patients withsuspected peptic ulcer

-2 -5

Referred patients with varicose veins

P 0> \O

"le1~ 1\> e A

9~OO ,,

60%50%40%30%20%

10%5%

*cna)4-

CDa,

a)

oC/)

VU)

1%

0.5

0.0

-0-5 F-1 0

-1 5

600o5000

, 40%

-~~~~~~~~30%20%

Non-referred patients withvaricose veins oo0%

_ 5%-2.0 H

-2 5 L

9Q

Questionnaire scale

Fig 2 Mean SF 36 health profiles at baseline and one year's follow up for general population atd referred and i1o7-referred patients with one offour study conditions (*centiles ofgeneral population: for example, X% ofgeneral populationfalls below Xth centile and remaining members fall above that point)

the scales of energy and fatigue, pain, andgeneral health perception. For low back painnon-referred patients compared with referredpatients improved significantly more accordingto the scales of role limitations attributable tophysical problems and pain.

DiscussionThe SF 36 health survey questionnaire hasbeen shown to possess a high level of validityand reliability in populations in the UnitedStates.'2 13 26 In the United Kingdom it is areliable and valid measure of health status in

general populations'4 1 and in patient popu-lations.'6 Although some concern has beenexpressed over the reliability of the SF 36questionnaire over repeated administrations,27we showed in the accompanying article that thequestionnaire demonstrates good reliabilityacross all scales when used to monitor healthstatus in groups of patients over time. 17 It is lessreliable when used in managing individualpatients'7 but is still capable of reflectingclinically important deviations from the norm.If the SF 36 questionnaire is to function as anoutcome measure for use in evaluating patient

60%50%40%30%20%

10%5%

*(n

C/)a,

aC

1%

a)o0cnc0El

U)

a)o0

_0cJn

4-(n

60%50%40%30%20%

1000

5%

*

a)

a,0

1%

0.5

60%--5000

40%30%20%

*(na)

ca,)0

10%5%

0.0

-0.5

-1 *0

-1-55

I.a,

00C/)

V

CUEV

U)-2.0

-2.5

*Cl)a,4-

a)0

1%

Questionnaire scale

nrn

{J-n

190



j.com/

Qual H


ecember 1994. D

ownloaded from



care and monitoring patients it must also beresponsive to changes in health status overtime. Results from the United States haveproved encouraging, with the SF 36 question-naire performing as well as the longer sicknessimpact profile.7

In this study changes in all eight SF 36 scaleswere significantly related to changes in selfreported health as measured by the transitionquestion in the questionnaire. Standardisedresponse means, a measure of an instrument'sresponsiveness to change, were calculated foreach SF 36 scale. The group of patientsreporting their health as much better on thetransition question had the largest standardizedresponse means, followed by those reportingtheir health to be somewhat better and, lastly,by those reporting their health to be about thesame.To estimate the burden of each particular

condition on health status and to assess theextent of any improvement, patients' scores onthe eight scales of the questionnaire at baselineand follow up were compared with the scoresfor the local general population. Across alleight questionnaire scales mean improvementsin health status were seen for all fourconditions, which were in line with theresponses to the transition question. For twoconditions - namely, varicose veins andsuspected peptic ulcer - the SF 36 healthprofiles at follow up were close to the healthprofile for the general population. This findingsuggests that for the average patient withsuspected peptic ulcer or varicose veins thehealth outcome at one year was to achieve astandard of health that closely approximates tothat experienced by the average member of thegeneral population. For patients with low backpain and menorrhagia changes in scoressuggest the outcome is relatively lessfavourable. Although significant improvementsin questionnaire scores were seen for thesepatients, particularly for pain in patients withlow back pain and energy and fatigue in womenwith menorrhagia, the average patients' scorestill deviated considerably from that for theaverage member of the general population, onall eight health scales in the case of low backpain and on all but the physical and socialfunctioning scales in the case of menorrhagia.

Despite evidence of variations in referralpatterns28 referred patients have a lower per-ceived health status than those being managedwithin general practice." We showed thatimprovements in perceived health were greaterfor referred patients than non-referred patientsfor three of the conditions under study, theexception being low back pain. The relativelack of improvement in the questionnairescores for referred compared with non-referredpatients with low back pain may have severalpossible explanations. Referred patients mayrepresent a selected group of patients with backpain who are more likely to have chronic orsevere underlying disease and a consequentlypoorer prognosis or who were referred becausetheir back pain was resistant to treatment. Inaddition, few treatments for low back pain havebeen shown to be of clear value.29

We emphasise that it has not been possiblein a study of this kind to attribute with certainlythe health outcomes detected by the SF 36questionnaire to the medical care received. Wewere concerned with evaluating the responsive-ness of the SF 36 questionnaire and not theeffectiveness of treatments. To carry out suchan evaluation would require us to relate moreprecisely the processes of care for individualpatients to the outcome achieved, controllingfor other variables that may influence health.The alternative would be to conduct arandomised controlled trial of treatment usingthe SF 36 questionnaire as part of a packageof outcome measures. Such a package wouldalso include more traditional clinical indicatorsand a condition specific measure.30 For certainconditions, condition specific measures havebeen shown to be more responsive than the SF36 questionnaire to small, but clinicallysignificant, changes over time, with greaterpower to discriminate between patients withvery severe or very mild disease.3' They alsoprovide additional information on sympto-matology which may be relevant to clinicians.For a measure of health status and health

outcome to be suitable for routine use withinthe NHS in a wide variety of clinical settingsit must provide information that is valid,reliable, responsive to change, and quick andeasy to collect. Our findings here and in theaccompanying article suggest that the SF 36questionnaire is responsive to clinical changeand sufficiently reliable'7 for monitoringgroups of patients and, for at least four of itsscales, individual patients.

In conclusion, the SF 36 questionnaire isresponsive to changes in health status overtime, even when used to assess health improve-ments in patients with relatively minor clinicalconditions. Taken together with previouslypublished data on validity and reliability, ourresults provide further evidence for thepotential of the SF 36 questionnaire as aroutine tool in monitoring and assessment ofhealth outcome in the NHS and as anevaluative instrument in clinical research for awide range of conditions.

We thank the staff at Inverurie, Portlethen, Rubislaw Place, andWesthill practices for recruiting patients; Jeremy Grimshaw,Jenny Duncan, and Alison De Ville for help with data collection;Elizabeth Russell for helpful comments; and John Ware and hiscolleagues at the Health Institute of the New England MedicalCenter for permission to use the SF 36 health survey question-naire. This research and the Health Services Research Unit areboth funded by the Chief Scientist Office of the Scottish OfficeHome and Health Department; however, the opinionsexpressed are those of the authors, not of the SOHDD.

1 Tuley MR, Mulrow CD, McMahan CA. Estimating andtesting an index of responsiveness and the relationship ofthe index to power. Y Clin Epidemiol 1991 ;44:417-21.

2 Fitzpatrick R, Ziebland S, Jenkinson C, Mowat A, MowatA. A comparison of the sensitivity to change of severalhealth status instruments in rheumatoid arthritis. JRheumatol 1993;20:417-21.

3 Deyo RA, Inui TS. Toward clinical applications of healthstatus measures: sensitivity of scales to clinicallyimportant changes. Health Serv Res 1984;19:275-89.

4 Mackenzie CR, Charlson ME, DiGioia D, Kelley K. Canthe sickness impact profile measure change? An exampleof scale assessment. Journal of Chronic Diseases1986;39:429-38.

5 Jenkinson C, Ziebland S, Fitzpatrick R, Mowat A, MowatA. Sensitivity to change of weighted and unweightedversions of two health status measures. InternationalJournal of Health Sciences 1991;2:189-94.

191



j.com/

Qual H


ecember 1994. D

ownloaded from



6 Liang MH, Fossel AH, Larson MG. Comparisons of fivehealth status instruments for orthopedic evaluation. MedCare 1990;28:632-42.

7 Katz JN, Larson MG, Phillips CB, Fossel AH, Liang MH.Comparative measurement sensitivity of short and longerform health status instruments. Med Care 1992;30:917-25.

8 MacKenzie CR, Charlson ME, DiGioia D, Kelley K. Canthe sickness impact profile measure change? An exampleof scale assessment. Journal of Chronic Diseases 1986;39:429-38.

9 Guyatt G, Walter S, Norman G. Measuring change overtime: assessing the usefulness of evaluative instruments.Journal of Chronic Diseases 1987;40: 171-8.

10 Kazis LE, Anderson JJ, Meenan RF. Effect sizes forinterpreting changes in health status. Med Care1989;27(suppl):S 178-89.

11 Ware JE, Sherbourne CD. The MOS 36-item short-formhealth survey (SF-36). I. Conceptual framework and itemselection. Med Care 1992;30:473-83.

12 MeHorney CA, Ware JE, Raczek AE. The MOS 36-itemshort-form health survey (SF 36). II. Psychometric andclinical tests of validity in measuring physical and mentalhealth constructs. Med Care 1993;31:247-63.

13 McHorney CA, Ware JE, Rogers W, Raczek AE, Lu JF. Thevalidity and relative precision of MOS short- and longform health status scales and Dartmouth COOP charts.Results from the Medical Outcomes Studv. Med Care1992;30(suppl):MS253-65.

14 Brazier JE, Harper R, Jones NMB, O'Cathain A, ThomasKJ, Usherwood T, et al. Validating the SF 36 healthsurvey questionnaire: new outcome measure for primarycare. BMJ 1992;305:160-4.

15 Jenkinson C, Coulter A, Wright L. Short form 36 (SF-36)health survey: normative data for adults of working age.BMJ 1993;306:1437-40.

16 Garratt AM, Ruta DA, Abdulla MI, Buckingham KJ,Russell IT. The SF-36 health survey questionnaire: anoutcome measure suitable for routine use within theNHS? BMJ 1993;306:1440-4.

17 Ruta DA, Abdalla MI, Garratt AM, Coutts A, Russell IT.SF 36 health survey questionnaire. I. Reliability in twopatient based studies. Quialitv ill Health Care 1994;3:180-5.

18 Phillips RC, Lansky DJ. Outcomes management in heartvalve replacement surgery: early experience. Journal (ofHeart Valve Disease 1992;1:42-50.

19 Bergner M, Bobbitt RA, Pollard WE, Martin DR, GilsonBS. The sickness impact profile: validation of a healthstatus measure. Med Care 1976;14:57-67.

20 Liang MH, Larson MG, Cullen KE, Schwartz JA.Comparative measurement efficiency and sensitivity offive health status instruments for arthritic research.Arthn tiis Rheuni 1985;28:542-7.

21 Fitzpatrick R, Ziebland S, Jenkinson C, Mowat A, MowatA. Transition questions to assess outcomes in rheumatoidarthritis. BrjRheurnatol 1993;32:807-11.

22 Devo RA, Centor RM. Assessing the responsiveness offunctional scales to clinical change: an analogy ofdiagnostic test performance. Journal of (Chronic Diseases1986;39:897-906.

23 Bindman AB, Keane D, Lurie N. Measuring health changesamong severely ill patients: the floor phenomenon. MedCare 1990;28:1142-52.

24 Stewart AL, Greenfield S, Hays RD, Wells K, Rogers MH,Berry SD, et al. Functional status and well being ofpatients with chronic conditions: results from the Iiiedicaloutcomes study. JAMA 1989;262:907- 13.

25 Armitage P, Berry G. Statistical method in edcn.dal rerh/.i2nd ed. Oxford: Blackwell, 1987.

26 Ware JE, Snow KK, Kosinski M, Gandek B. SI-36 hlalthiurccs manunual and illetprctatiiioi guidc. Boston: NimrrodPress, 1993.

27 Sheldon T. Reliability of SF 36 remains uncertain. 1l3A71993;307:125-6.

28 Roland MO, Coulter AM. Hospital rcfc7rals. Oxford: OxfordUniversity Press, 1992.

29 Quebec Task Force. Scientific approach to the assessmentand management of activity-related spinal disorders. Amonograph for clinicians. Report of the Quebec TaskForce on spinal disorders. Spine 1987;12(suppl): 1-59.

30 Garratt AM, Macdonald LM, Ruta DA, Russell II,Buckingham JK, Krukowski ZH. Towards themeasurement of outcome for patients ssith s aricose sells.Quality ini Health Care 1993;2:5-10.

31 Ruta DA, Garratt AM, Wardlaw D, Russell IT. Developinga valid and reliable measure of health outcome for patientsssith low back pain. Spine 1994;19:1887 96.

192



j.com/

Qual H


ecember 1994. D

ownloaded from


sf 36 health questionnaire: ii. responsiveness to health ... · sf 36 questionnaire to changes in...

Documents