O R I G I N A L A R T I C L E

Measurement of Health Statusin Diabetic PatientsDiabetes Impact Measurement Scales

G. STEVEN HAMMOND, MD, PHDTHOMAS T. AOKI, MD

OBJECTIVE — To develop an instrument to measure health status in adult insulin-dependent (type I) and non-insulin-dependent (type II) diabetic patients.

RESEARCH DESIGN AND METHODS— Correlative study to examine psy-chometric properties of the questionnaire. Test-retest reliability, item-scale correla-tions, principal-components analysis, correlations with global clinical ratings, andcorrelations with clinical data extracted from medical records were examined at thediabetes clinics at the University of California, Davis, Medical Center. Patients werevolunteer clinic patients able to complete the questionnaire. One hundred thirtypatients completed a first administration of the questionnaire, and 52 completed asecond administration.

RESULTS — Test-retest reliability was satisfactory. Item-scale correlations showedthat 40 of 44 questionnaire items were highly correlated with subscale and total scalescores. Principal-components analysis identified one major factor measured by thequestionnaire. Cronbach's a, a measure of the scales' internal consistency, was ofsatisfactory magnitude. Global ratings of clinical status by patients and clinicians werehighly correlated with scale scores. Correlations of scale scores with clinical data weregenerally of low magnitude but, where significant, were consistently in the directionhypothesized if the scale truly measures health status or disease impact.

CONCLUSIONS— The Diabetes Impact Management Scales (DIMS) is an easilyadministered questionnaire with internal consistency and test-retest reliability. Pre-liminary correlative analyses support the validity of the instrument as a measure ofhealth status in adult type I and type II diabetic patients. Further work will benecessary to firmly establish the validity of the DIMS and its usefulness in clinicaloutcomes research.

FROM THE UNIVERSITY OF CALIFORNIA, DAVIS, MEDICAL CENTER, SACRAMENTO, CALIFORNIA.

ADDRESS CORRESPONDENCE AND REPRINT REQUESTS TO THOMAS T. AOKI, MD, UNIVERSITY OF CALI-

FORNIA, DAVIS, DIVISION OF ENDOCRINOLOGY, 4301X STREET, BUILDING FOLBII-C, SACRAMENTO, CA

95816.

RECEIVED FOR PUBLICATION 12 SEPTEMBER 1990 AND ACCEPTED IN REVISED FORM 3 JULY 1991.

Over the past 20 yr, considerableeffort has been devoted to the de-velopment of health-status mea-

surement instruments to study outcomesin public-health and clinical research.These surveys provide valuable informa-tion complementing more traditionalmeasures of clinical outcome such asmorbidity and various physiological andpathological parameters (1,2). Some ofthese instruments have also been usedeffectively to study quality of life anddisease impact in specific patient popu-lations including cancer (3), arthritic (4 -6), and hypertensive patients (7). Aninstrument used to measure health statusin insulin-dependent (type I) diabeticpatients in the Diabetes Control andComplications Trial has been described(8). This study describes another instru-ment developed to measure health statusor disease impact in diabetic patients.

We aimed to develop a measureof clinical outcome in diabetes researchthat would complement traditional mea-sures, e.g., blood glucose and glycosylatedhemoglobin levels and presence and sever-ity of diabetic complications. Our mainpurpose was to develop what has beentermed an evaluative index (9), specificallydesigned to measure longitudinal changein diabetic patients to quantitate treatmentbenefit in clinical trials.

We drew heavily from previousresearch efforts to measure health statusin general populations, most notably thehealth-status measures used in the RandHealth Insurance Study (2,10,11), theSickness Impact Profile (12), and a scaledesigned to measure health status in ar-thritic patients, the Arthritis Impact Mea-surement Scales (4-6). The consensus ofthese and other groups has been that ahealth-status measure should reflect thepresence of important symptoms, func-tional capacity or impairment, and gen-eral well being. These factors contributeto quality of life, yet are only indirectly, ifat all, reflected in traditional measures ofclinical outcome.

DIABETES CARE, VOLUME 15, NUMBER 4, APRIL 1992 469

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/15/4/469/441272/15-4-469.pdf by guest on 15 January 2022

Diabetes impact measurement scales

RESEARCH DESIGN ANDMETHODS — Questionnaire items werederived from discussions among a group ofexperienced clinicians including physi-cians, a diabetes nurse specialist, and aregistered dietitian associated with theDiabetes Clinics at the University of Cal-ifornia, Davis, Medical Center. We fo-cused on common and significant symp-toms of diabetes, patients' attitudestoward their disease and its manage-ment, patients' abilities to fulfill their so-cial roles, and patients' general sense ofwell being. Some of the items relating togeneral well being were drawn directlyfrom the Rand Study Scale (11).

Items were grouped into foursubscales: 1) symptoms, subdivided intosymptoms relatively specific for diabetesand less specific for diabetes; 2) diabetes-related morale, pertaining to the patients'attitudes toward managing their disease;3) social-role fulfillment; and 4) well be-ing. Although it was recognized thatthere was considerable overlap betweenthese categories, the subscales were con-structed to explore the possibility thatthere may be identifiable independentfactors contributing to health status indiabetic patients. Items were constructedto allow a graded response (i.e., Likertscales) for the purpose of providing highsensitivity to longitudinal change. Effortwas made to use simple and unambigu-ous wording of items and responsechoices. The language of the items isestimated to require a sixth-grade read-ing ability. The items were designed foruse with adult subjects but, with a fewexceptions (e.g., items dealing with li-bido and sexual functioning), could beused with children and adolescents. Thescale was designed to be applicable toboth type I and non-insulin-dependent(type II) diabetic patients.

The test items are listed in theAPPENDIX. A copy of the complete ques-tionnaire including instructions, re-sponses, and scoring key is availablefrom the authors on request. The ques-tionnaire was self-administered and tookan average of 15-20 min to complete.

Individual items were scored accordingto the response selected. Higher valueswere assigned for less-severe or less-frequent symptoms, greater diabetes-related morale, greater social-role fulfill-ment, and greater well being. Some itemswere reverse keyed on the administeredquestionnaire to avoid response bias.Subscale and total-scale scores werecomputed by adding individual itemscores. The handling of missing values isdiscussed below.

The experimental protocol wasapproved by the University Human Sub-jects Review Committee. Subjects wereselected from patients visiting the Diabe-tes Clinic at the University of California,Davis, Medical Center. Selection criteriawere as follows: 1) diagnosis of diabetesmellitus by conventional criteria (>2fasting blood glucose levels >7.7mM[> 140 mg/dl] and/or random blood glu-cose levels >11.11 mM [>200 mg/dl]);2) English-language reading ability suffi-cient to read and complete the question-naire; and 3) willingness to participate inthe study.

The questionnaire was completedduring a clinic visit usually before thepatient was seen by a clinician. Ratingscales (see below) were also completedby clinicians (endocrinology faculty orfellows) at the time of the clinic visit.

A second administration of thequestionnaire was performed at a subse-quent visit for as many subjects as could beaccomplished. The minimum time betweenfirst and second administrations was 1 mo.

Validation studiesEach questionnaire packet included twoglobal rating scales consisting of a99-mm line. The first scale was precededby the instruction "show by marking anX on the line below how well your dia-betes has been controlled during the pastmonth (in terms of blood sugar, symp-toms, and complications)." The secondscale was preceded by the instruction"show by marking an X on the line belowhow good your general health (physical,mental, and emotional) has been during

the past month." Beneath the leftwardend of each scale was the label "could nothave been worse;" beneath the center ofthe scale was the label "in between;" andbeneath the rightward end of the scalewas the label "could not have been bet-ter." The scale was scored by measuringthe number of millimeters away from theleftward end X was marked.

Analogous scales for each partic-ipating subject were presented to clini-cians for completion.

Patients' charts were abstractedfor clinical data including sex, age, typeof diabetes (I or II, according to chartnotes), age of onset, and duration of di-abetes. The presence of diabetic compli-cations was tabulated for retinopathy (0,no retinopathy; 1, background retinopa-thy; 2, proliferative retinopathy),nephropathy (persistent proteinuria onurinalysis >75 mg protein in 24-h urinesample), neuropathy, gastropathy, andfoot ulcer. Other medical diagnoses werecoded as present or absent: hyperten-sion, obesity, coronary artery disease,cerebrovascular disease, peripheral vas-cular disease, congestive heart failure,hypercholesterolemia (>5.2 mM [>200mg/dl] total cholesterol), hypertriglycer-idemia (> 16.6 mM [>300 mg/dl]), renalinsufficiency, rheumatologic disease orarthritis, other endocrinologic disease,pancreatic insufficiency, chronic obstruc-tive pulmonary disease (COPD), andother miscellaneous medical diagnosis.The presence of these conditions was de-termined from problem lists or notesfrom other clinics found in the patients'medical record.

Clinical parameters includedweight, blood pressure, pulse rate,HbAlc (by high-performance liquidchromatography; 13), serum potassium,serum creatinine, blood urea nitrogen,creatinine clearance, protein on urinaly-sis, 24-h urinary protein excretion, se-rum cholesterol, and serum triglycerides.Laboratory values closest in time to thefirst administration of the questionnairewere used. If available, HbAlc was re-corded for the time of both the first ad-

470 DIABETES CARE, VOLUME 15, NUMBER 4, APRIL 1992

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/15/4/469/441272/15-4-469.pdf by guest on 15 January 2022

Hammond and Aoki

ministration and the second administra-tion of the questionnaire. Blood forserum lipid levels was usually collectedfrom nonfasted patients. Because HbAlc

and self-monitored blood glucose resultswere customarily used in these clinics toevaluate glycemic control, few fastinglaboratory blood glucose levels wereavailable for analyses.

The therapeutic regimen was alsocoded. Use of insulin and the number ofinjections per day were recorded as wasthe use of oral hypoglycemic agents. Alsorecorded was use of diuretics, P-blockers,central a-blockers, angiotensin convertingenzyme (ACE) inhibitors, calcium-channelblockers, tricyclic antidepressants, pheny-toin, and metoclopramide. Also, the totalnumber of prescribed medications and thepresence or absence of reported hypogly-cemic episodes and of polydipsia/polyuriawere recorded.

Statistical analysesDescriptive statistics were performed onall variables. Most of the remaining sta-tistical tests were correlative. Computedvariables included a diabetes complica-tions index (sum of scores for retinopa-thy, nephropathy, neuropathy, gastropa-thy, and foot ulcers) and a medical-diagnoses index (sum of all codeddiagnoses).

Standard item analyses were per-formed, including interitem correlationsand item-scale correlations for the totalscale and the subscales. The item scorewas not included in the computation ofscale scores for item-scale correlations.Cronbach's a, a measure of internal con-sistency or homogeneity of the scale andsubscales, was also performed. A princi-pal-components analysis and a canonicalanalysis between principal componentsand the subscales were performed.

Missing values were handled inseveral ways. When responses wereomitted, the scale scores were adjusted toreflect the proportion of the total possi-ble score when that item was omitted.This essentially assigns the average itemscore on a particular scale when an item

is omitted. As discussed below, there wasnot a great number of missing responsesto test items, but responses to some testitems were more likely to be omittedthan others. In all correlational analyses,pairwise deletions were made for missingvalues.

RESULTS

Sample characteristicsOne hundred thirty patients completed afirst administration of the questionnaire,and 52 completed a second administra-tion. Fifty-five male (42% of sample) and75 female (58% of sample) patients wereincluded. Age range was 18-78 yr of age,with a mean ± SD age of 45 ± 15 yr.Fifty-one patients (39%) were diagnosedas type I and 77 patients (59%) as type II(2 patients could not be classified as typeI or II). The age of onset of diabetesranged from 1 to 74 yr with a mean of 34yr. The number of years of duration ofdiabetes ranged from 0 (<1 yr) to 34 yrwith a mean of 11 yr.

Seventy-eight (60%) patientswere not documented to have retinopa-thy, whereas 30 (23%) were described ashaving background diabetic retinopathyand 21 (16%) as having proliferative ret-inopathy. Thirty-six (28%) patients weredocumented as having nephropathy, i.e.,abnormal proteinuria. Thirty-nine pa-tients (30%) had documented evidenceof neuropathy (sensory loss or unequiv-ocal loss of deep tendon reflexes), and 90patients (70%) had none. Only 4 pa-tients (3%) were documented as havinggastropathy, whereas 125 (97%) showedno documented evidence of this diabeticcomplication. Five patients (4%) weredocumented as having foot ulcers, and124 patients (96%) were not. Data re-garding complications were not availablefor 1 patient. Frequencies of other med-ical diagnoses are shown in Table 1.

Among our sample, HbAlc valuesranged from 4.3 to 14.1% (mean ± SD7.59 ± 1.8%; normal range 4-6%) at ornear the time of the first administrationof the questionnaire.

One hundred patients (77%)were taking insulin and 29 (22%) werenot, with data missing for 1 patient. Ofthe 100 patients taking insulin, 4 tookone injection per day, 25 took two injec-tions, 23 took three injections, and 46took four injections per day. Twenty-nine patients (22%) took an oral hypo-glycemic agent and 101 (78%) did not.

Item analysesEach item elicited responses throughoutthe possible range. Though some itemswere more frequently left unansweredthan others, it appeared that the itemswere easily answered by most patients.No effort was made to insist that subjectsanswer every item. Ninety of 130 sub-jects responded to every item in a scor-able fashion. Some omissions appearedto be due to simple oversight, whichseemed particularly apparent when apage of the questionnaire was left blank(the printing format resulted in 4 itemsbeing presented on each page). Therewere occasional unscorable idiosyncraticresponses (write in responses). One item(no. 15) was not applicable to 4 subjectsbecause they were not taking hypoglyce-mic medications at the time of the ques-tionnaire administration. Item 30 (5missing values) was considered not ap-plicable by several patients who statedthat they had no family. Item 40 (3 miss-ing values) may have been considerednot applicable by patients who were notdoing self-monitoring of blood glucose.

Four items (6, 21, 22, and 32)had relatively high rates of missing val-ues (16, 13, 15, and 10 answers missing,respectively). Some patients appeared tobe inhibited in responding to the items(21 and 32) pertaining to sexual func-tioning and libido. This phenomenonhas been noted previously (4). A fewpatients wrote in that they were sexuallyinactive. The items concerning the oc-currence of hypoglycemia with exercisewere perhaps difficult to understand orperhaps felt to be not applicable in casesof extreme limitation of exercise. Thesefour items were retained for analysis be-

DIABETES CARE, VOLUME 15, NUMBER 4, APRIL 1992 471

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/15/4/469/441272/15-4-469.pdf by guest on 15 January 2022

Diabetes impact measurement scales

Table 1—Frequency of other medical diagnoses and diabetes and related morbidity in studysample

PRESENT ABSENT DATA MISSING

HYPERTENSION

OBESITY

CORONARY ARTERY DISEASE

CEREBROVASCULAR DISEASE

PERIPHERAL VASCULAR DISEASE

CONGESTIVE HEART FAILURE

HYPERCHOLESTEROLEMIA

HYPERTRIGLYCERIDEMIA

RENAL INSUFFICIENCY

ARTHRITIS/RHEUMATOLOGIC DISEASE

OTHER ENDOCRINE DISEASE

PANCREATIC INSUFFICIENCY

COPDOTHER DIAGNOSIS t

RETINOPATHY

BACKGROUND

PROLIFERATIVE

NEPHROPATHY

NEUROPATHY

GASTROPATHY

FOOT ULCER

33 (26)49 (38)14(11)4(3)2(2)4(3)

54 (43)24 (19)17(13)10(8)13 (10)4(3)9(7)

27 (21)

30 (23)21 (16)36 (28)39 (30)4(3)5(4)

96 (74)80 (62)

115(89)125 (97)127 (98)125 (97)71 (57)

100 (81)112(87)119(92)116(90)125 (97)120 (93)103 (79)78 (60)

93 (72)90 (70)

125 (96)124 (96)

:hronic obstructive pulmonary disease.

11*1111561111101

If111

*Not listed as medical problem or diagnosis.tMiscellaneous, e.g., psychiatric disorders, gout, migraine.f Abnormal proteinuria not documented.

cause they were felt to be potentiallysignificant in assessment of disease im-pact of diabetes and because they weresatisfactorily answered by most respond-ents.

The strategy of assigning a valuefor missing responses equal to the aver-age of all other responses on the scale (orsubscale) was felt to be justified for sev-eral reasons. First, if all questionnaireswith a missing value were disqualifiedfrom analysis, a significant proportion ofthe sample (40 of 130) would have beenlost, which could have introduced a biasin the patient sampling. Second, theitems with the highest rate of nonre-sponse deal with potentially importantclinical information. Again, even theitems with the highest rate of nonre-sponse were satisfactorily answered bymost subjects. Finally, with one excep-tion (item 22), these items correlatedhighly with scale scores, which mini-

mized the likelihood of introducing asignificant bias by retaining them.

Item-scale (and interitem) corre-lations are available from the authors onrequest. Four items (16, 22, 39, and 42)performed poorly, not correlating signif-icantly with the total-scale score (P >0.01). All other items showed significant(P < 0.01) correlations with their re-spective subscale totals and the total-scale score. Because of their poor perfor-mances, items 16, 22, 39, and 42 will bedeleted from the scoring of the DiabetesImpact Measurement Scales (DIMS) in itsfuture applications.

Distributions of subscale andtotal-scale scoresDistribution characteristics of each sub-scale and the total scale are presented inTable 2. Scores are presented as T scores.The T scores represent the scale score asthe proportion of the total possible score

on the scale, multiplied by 10, yielding arange of possible values from 0 to 10.Each scale yielded a reasonably widerange of scores. The scales are all nega-tively skewed, reflecting the generallyobserved characteristic of quality-of-lifemeasures to concentrate scores closer tothe upper end of the scale (11). Note thatthe total-scale score here is derived froma simple sum of responses to all items.This gives disproportionate weight to thesubscales with the greater number ofitems. It can be argued that the subscales(symptoms, well being, diabetes-relatedmorale, and social role fulfillment)should be weighted equally in determin-ing the total-scale score. This would beeasily accomplished by simply taking theaverage of the subscale scores as the to-tal-scale score. In future applications ofthe DIMS, we will explore both methodsof obtaining total-scale scores.

Subscale analysesInformation regarding the relationshipsbetween the subscale scores and the to-tal-scale scores is listed in Table 3, acorrelation matrix of these scores. Allsubscale scores were highly correlated(P < 0.001) with all other subscales, butthe correlations are far enough fromunity to suggest that the subscales arenot redundant. (An exception is the highcorrelation of the two subdivisions of thesymptoms subscale—specific and non-specific symptoms of diabetes, with thecombined-symptoms subscale.)

Cronbach's a, a test of the inter-nal consistency of the subscales and totalDIMS scale, was performed (Table 4).The obtained values are within the rangegenerally considered desirable by psy-chometric standards.

Principal-components analysisA principal-components analysis wasdone to explore statistically the factorstructure of the questionnaire. A majorprincipal-component accounting for32% of the variance was identified. Be-yond this, nine components accountingfor <7.5% of the variance each were

472 DIABETES CARE, VOLUME 15, NUMBER 4, APRIL 1992

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/15/4/469/441272/15-4-469.pdf by guest on 15 January 2022

Hammond and Aoki

Table 2—Distribution

administration

SCALE

0/ subscale and total-scale scores on the 1st questionnaire

SKEWNESS

1AIB111111IVTOTAL

7.70 ± 1.21(3-9.7)7.30 ± 1.29 (3.8-10.0)7.40 ± 1.17(4.5-9.4)6.70 ± 1.71 (2.6-10.0)7.70 ± 1.08 (4.2-9.6)7.10 ± 2.08 (2.2-10.0)7.40 ± 1.15(3.8-9.5)

-0.77-0.23-0.37-0.25-0.45-0.36-0.31

Values are means ± SD of proportion of total possible score on the scale x 10. Ranges given in paren-theses. 1A, diabetes-specific symptoms; IB, nonspecific symptoms; I, combined symptoms; II, well being;III, diabetes-related morale; IV, social role fulfillment; Total, total diabetes impact measurement scalesscore.

identified, bringing the cumulative vari-ance explained to 69%.

All subscale scores were highlycorrelated with the first principal com-ponent (r = 0.75-0.95) and variablycorrelated with the remaining compo-nents.

A canonical correlational analysiswas performed to examine relationshipsbetween the principal components andsubscales. Aside from again showing thestrong positive correlation between allthe subscales and the first principal com-ponent, a contrast between scale I(symptoms) and scale III (diabetes-related morale) in relation to the secondprincipal component was suggested(scale I correlated negatively and scale IIIcorrelated positively with the secondprincipal component).

The principal-components analy-sis suggests that there is a major factorbeing measured by all subscales and thetotal-DIMS score and several minor fac-tors. Although we did not see statisticalsupport for unique significance of thesubscales, we will continue to study therelationship of subscale scores to clinicalvariables to determine whether the sub-scales provide information beyond thatof the total-DIMS score. The subscalesare expected to perform differently whennondiabetic patient populations arestudied and compared with diabetic pa-tients.

Validation studiesA traditional test of test-retest reliabilitywas not possible within the design of thisstudy because the interval between re-peated administrations of the scale variedfrom subject to subject. However, scalescores on the first and second adminis-trations (Table 4) were all highly signif-icantly correlated.

Construct validation of the DIMSwas sought by exploring correlations ofthe DIMS scores with the global ratingscales and the clinical variables (Table5). For the most part, the DIMS subscaleand total-scale scores were highly corre-lated with both patients' and clinicians'global ratings of overall diabetic controland health status.

We hypothesize that higherDIMS scores would generally be nega-

tively correlated with clinical variablesindicating presence of disease. Correla-tions between DIMS scores and the nu-merous clinical variables recorded wereof variable but generally low magnitude.Clinical variables showing no significantcorrelation with DIMS scores includeddiabetes type, duration of disease, pres-ence of retinopathy, nephropathy, gastr-opathy, foot ulcer, diagnosis of hyperten-sion, coronary artery disease, peripheralvascular disease, congestive heart failure,renal insufficiency, hypercholesterol-emia, hypertriglyceridemia, systolic anddiastolic blood pressures, presence of in-sulin reactions, use of insulin, number ofdaily insulin injections, use of oral hypo-glycemic agents, and most other medica-tions recorded (diuretics, (3-blockers,central a-blockers, ACE inhibitors, andcalcium-channel blockers). The diabe-tes-complications index (sum of allcoded complications) was not signifi-cantly correlated with DIMS scores. Forsome of these variables, the frequencydistribution within the sample was notconducive to generating a correlation,e.g., less-common diabetic complica-tions (gastropathy and foot ulcers).Other variables would not have been ex-pected to relate to the scale scores (e.g.,blood pressure, which was generally rea-sonably well controlled and is typicallyasymptomatic). Other variables might beexpected to be differently related to dia-betes impact in different situations, e.g.,

Table 3—Correlations of subscale and total-scale scores

1AIBI11

111

IV

TOTAL

1A

1.0000.7000.8590.5510.4850.4590.735

IB

0.7001.0000.9670.7040.5560.6340.875

I

0.8590.9671.0000.7030.5710.6170.890

11

0.5520.7040.7031.0000.6220.7590.890

III

0.4850.5560.5710.6221.0000.5180.797

IV

0.4590.6340.6170.7590.5181.0000.811

TOTAL

0.7350.8750.8900.8900.7970.8111.000

IA, diabetes-specific symptoms; IB, nonspecific symptoms; I, combined symptoms; II, well being; III,diabetes-related morale; IV, social role fulfillment; Total, total diabetes impact measurement scales score.P < 0.001 for all comparisons.

DIABETES CARE, VOLUME 15, NUMBER 4, APRIL 1992 473

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/15/4/469/441272/15-4-469.pdf by guest on 15 January 2022

Diabetes impact measurement scales

Table 4—Cronbach's a. scores for subscales and total Diabetes Impact MeasurementScales (DIMS) scale and test-retest correlations

SCALE

SPECIFIC: SYMPTOMS

NONSPECIFIC SYMPTOMS

COMBINED SYMPTOMS

WELL-BEING

DIABETES-REIATED MORALE

SOCIAL ROLE FULFILLMENT

TOTAL DIMS

a

0.60450.78610.85320.85220.76660.85450.9394

N

10911294

11811112387

TEST-RETEST CORRELATION

(N = 52)

0.7500.6890.7240.6090.7780.6510.770

n varies because casewise deletion was used for missing values to calculate Cronbach's a.P < 0.001 for all comparisons.

use of insulin, which is generally re-quired in more-severe diabetes, butwhich clearly produces therapeutic ben-efits, or presence of hypoglycemic epi-sodes, which can be seen in both poorlycontrolled, "brittle", diabetes and tightlycontrolled diabetes.

Diagnosis of other endocrine dis-ease (primarily thyroid disease) was neg-atively correlated with the symptomssubscales (P = 0.005, 0.01, and 0.005,respectively, for specific symptoms, non-specific symptoms, and combined-symp-toms subscales) and the total-DIMS score(P = 0.03). Presence of COPD was neg-atively correlated with the symptomssubscales (P < 0.001 for specific, non-specific, and combined-symptoms sub-scales), well being (P = 0.01), social-rolefulfillment (P = 0.02), and the total-DIMS score (P = 0.001). Presence ofmiscellaneous diagnoses was negativelycorrelated with the nonspecific-sump-toms subscale (P = 0.03) and combined-symptoms subscale (P = 0.03), well-being subscale (P = .001), social-role-fulfillment subscale (P = .008) and total-DIMS score (P = 0.005). The medical-diagnoses index (sum of medicaldiagnoses) was significantly negativelycorrelated with the nonspecific-symp-toms (P = 0.001) and combined-symp-toms subscales (P = 0.003), the socialrole-fulfillment subscale (P = 0.03), andthe total-DIMS score (P = 0.03).

Presence of neuropathy was neg-atively correlated with the nonspecificsymptoms subscale (P = 0.009) and thecombined-symptoms subscale (P =0.02), although not with the specific-symptoms subscale (P = 0.14), despitethat the latter includes an item pertainingto uncomfortable paresthesias. Presenceof obesity was negatively correlated withthe specific-symptoms subscale (P =0.03). Pulse rate was negatively corre-lated with the specific-symptoms sub-scale (P = 0.03), the nonspecific-symp-toms subscale (P = 0.02), and the total-DIMS score (P = 0.03).

The use of certain medicationswas negatively correlated to DIMS

scores. Tricyclic antidepressant use wasnegatively correlated to the specific-symptoms subscale (P = 0.004), thenonspecific-symptoms subscale (P =0.001), the combined-symptoms sub-scale (P < 0.001), and the social role-fulfillment subscale (P = 0.01) and thetotal-DIMS score (P = 0.004). Phenytoinuse was negatively correlated with thespecific-symptoms subscale (P = 0.005),nonspecific-symptoms subscale (P <0.001), and the combined-symptomssubscale (P < 0.001), and the well-beingsubscale (P = 0.01), the diabetes-related-morale subscale (P < 0.001), social-role-fulfillment subscale (P = 0.01), and thetotal-DIMS score (P < 0.001). Use ofmetoclopramide was negatively correlatedwith the nonspecific-symptoms subscale(P = 0.002), the combined-symptomssubscale (P = 0.007), the well-beingsubscale (P = 0.02), the social role ful-fillment subscale (P = 0.04), and the to-tal-DIMS score (P = 0.007). Finally, thetotal number of prescribed medicationswas negatively correlated with the non-specific-symptoms subscale (P < 0.001),the combined-symptoms subscale (P =0.003), the well-being subscale (P =0.04), and the total-DIMS score (P =0.02).

The report of polydipsia and/orpolyuria was negatively correlated withthe specific-symptoms subscale (P <

Table 5—Correlations of rating scales with subscale and total-scale scores

SCALE

IAIBIIIIIIIV

TOTAL

DIABETES

PATIENT

0.3490.249*0.3050.3670.5460.222t0.432

RATING SCALE

CLINICIAN

0.3400.294*0.3340.3470.3380.243t0.383

WELLNESS

PATIENT

0.269*0.3170.3250.4700.3610.3080.429

SCALE

CLINICIAN

0.3680.3300.3680.4520.321*0.288*0.432

P < 0.001, unless noted.*P<0.01.tP > 0.01.

474 DIABETES CARE, VOLUME 15, NUMBER 4, APRIL 1992

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/15/4/469/441272/15-4-469.pdf by guest on 15 January 2022

Hammond and Aoki

0.001; this subscale contains an itemspecifically addressing this symptom),the nonspecific-symptoms subscale (P <0.001), and the combined-symptomssubscale (P < 0.001), and the well-beingsubscale (P = 0.008), the social-role-fulfillment subscale (P = 0.01), and thetotal-DIMS score (P = 0.001). TheHbAlc levels were negatively correlatedwith the nonspecific-symptoms subscale(P = 0.04), the combined-symptomssubscale (P = 0.04), the well-being sub-scale (P = 0.05), the diabetes-related-morale subscale (P = 0.01), and the to-tal-DIMS score (P = 0.005).

The hazards of relying on tests ofstatistical significance when multipletests are conducted are well known.However, virtually all of the above-listedsignificant correlations are in the direc-tion that would be expected if the DIMSis a valid measure of health status ordisease impact. Of the 266 correlationsexamined above (7 subscale and totalscores with 38 clinical variables), 57(21%) had P < 0.05, well above thenumber that would be expected to occurrandomly.

We also found some significant(P < 0.05) correlations between clinicalvariables and scale scores that were notpredicted by our general hypothesis butare interesting. Age was positively corre-lated with the well-being subscale (P =0.026) and the diabetes-related-moralesubscale (P = 0.003). Sex was stronglycorrelated to the symptoms subscale, forspecific symptoms of diabetes (P < 0.001),nonspecific symptoms (P < 0.001), andcombined-symptoms subscales (P <0.001) (female patients reported moresymptoms). Sex was also similarly corre-lated with well-being scores (P = 0.002)and diabetes-related morale (P = 0.002),with female patients scoring lower. Therelationship of sex to social-role fulfill-ment was in the same direction but notas strong (P = 0.06). There was a highnegative correlation between sex and thetotal-DIMS score (r = -0.332, P <0.001). The age of onset of diabetes was

positively correlated with diabetes-related morale (P = 0.01).

CONCLUSIONS— In this article, wediscuss the development, psychometriccharacteristics, and preliminary valida-tion studies of a new instrument de-signed to measure health status in dia-betic patients. The DIMS is designed foruse in adult patients with either type I ortype II diabetes mellitus. It was specifi-cally designed as an evaluative index tomeasure longitudinal changes in healthstatus or disease impact in clinical trialsof therapeutic interventions in diabetesbut may also serve for other purposessuch as comparisons across groups ofdiabetic patients.

The questionnaire is simple andstraightforward, consisting of easily un-derstood items covering a broad range ofcontent relevant to diabetes impact.

The psychometric properties ofthe DIMS reflected in our analyses ap-pear satisfactory. Four items that did notperform well in terms of item-scale cor-relations (items 16, 22, 39, and 42) willnot be scored. The additional informa-tion provided by the subscales of theDIMS is of uncertain value, but it is ex-pected that unique properties of the sub-scales may appear when diabetic andnondiabetic patient populations are com-pared with the DIMS. Future studieswith the DIMS will examine the potentialusefulness of the subscales. For now, wesuggest that each subscale of the DIMSbe considered a separate indicator ofclinical outcome and that the averagescore of the major subscales (symptoms,well being, diabetes-related morale, andsocial-role fulfillment) be used as anoverall index of diabetes impact. Furtherrefinements of the scoring may evolvewith future use of the DIMS. In futurestudies, we will compare responses to theDIMS in diabetic and nondiabetic pa-tients and study the change of DIMSscores in response to therapeutic inter-ventions.

Acknowledgments—Francine Cisneros re-cruited subjects and administered question-naires. Howard Schutz, PhD, gave advice re-garding statistical validation of questionnaire.Neill Willitts, PhD, gave advice on principalcomponents analysis.

APPENDIX

IA 1. How often were you bothered byexcessive thirst and urinationduring the past month?

II

III

2. During the past month, have youbeen anxious or worried?

3. During the past month, have youfelt optimistic about your diabe-tes?

IB 4. How good has your muscularstrength and endurance beenduring the past month?

IB 5. Over the past month, have youbeen bothered by blurring of vi-sion?

IA 6. Over the past month, how muchexercise could you do withoutdeveloping low blood sugar?

II 7. During the past month, have youfelt that you were good at doingthe most important things youdo (for example, your work,school, homemaking, parenting,handling personal affairs)?

III 8. Over the past month, how muchhave you felt personally in chargeof managing your diabetes?

2. Over the past month, how muchenergy have you had?

IB 10. During the past month how wellhave you slept?

Ill 11. During the past month, how wor-ried have you been about havingan insulin reaction or a danger-ously low blood sugar?

IB

DIABETES CARE, VOLUME 15, NUMBER 4, APRIL 1992 475

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/15/4/469/441272/15-4-469.pdf by guest on 15 January 2022

Diabetes impact measurement scales

IV 12. Have you met the obligationsand responsibilities you feeltoward your family during thepast month?

IB 13. During the past month, have youbeen bothered by constipation?

II 14. Have you felt depressed duringthe past month?

II 15. During the past month, was it aninconvenience or bother to you totake your diabetes medicine (pillsor insulin)?

III 16. During the past month, have youeaten too much?

IA 17. During the past month, were youbothered by burning, tingling,pain, or numbness in your feet orhands?

II 18. During the past month, howworried or fearful have you beenabout your future?

II 19. Have you eaten what you wantedto during the past month?

III 20. During the past month, have youfelt it was worth the effort to takecare of your diabetes?

IB 21. During the past month, how of-ten were you able to functionsexually as well as you wantedto?

IA 22. Over the past month did you de-velop low blood sugar with exer-cise?

IV 23. Have you functioned well, notlimited by your health, in yourusual occupation (homemaking,school, work, etc.) during the pastmonth?

IA 24. How often did you vomit aftereating during the past month?

Ill 25. During the past month, mywhole schedule of activities wasrestricted by my diabetes.

IB 26. Over the past month, have youbeen bothered by feeling faint ordizzy on sitting up or standingup?

II 27. How much of the time, duringthe past month, has your dailylife been full of things that wereinteresting to you?

III 28. Overall, during the past month,how do you think your diabeteshas been doing?

IB 29. Has your appetite been goodduring the last month?

IV 30. During the past month, have youparticipated in and enjoyed fam-ily life?

IV 31. During the past month, how oftenhave you been able to functionwellinyourusualoccupation(home-making, school, work, etc.)?

II 32. How high has your interest in sexbeen over the past month?

IA 33. How often did you have abdomi-nal discomfort after eating duringthe past month?

III 34. How often during the pastmonth have you been uncertainabout how much to eat and/orhow much insulin to take?

IV 35. Have you enjoyed social and rec-reational activities during thepast month?

II 36. During the past month, have youfelt useful?

IB 37. During the past month, howmuch of the time were you lack-ing enough energy?

IB 38. How often did you have diarrheaduring the past month?

III 39. During the past month, have youbeen able to follow medical rec-ommendations concerning yourdiabetes?

Ill 40. During the past month, was yourdiabetes monitoring an inconve-nience or bother to you?

II 41. During the past month, howmuch of the time did you feelthat things were going well foryou?

II 42. Have you eaten when youwanted to during the pastmonth?

IB 43. During the past month how oftendid you feel nauseated after eat-ing?

III 44. Over the past month, how welldo you feel you have understoodyour diabetes?

IA, diabetes-specific symptoms subscale;IB, nonspecific symptoms subscale; II,well-being subscale; III, diabetes-relatedmorale subscale; IV, social role fulfill-ment subscale.

References

1. Bergner M: Measurement of health sta-tus. Med Care 23:696-704, 1985

2. Brook RH, Davies-Avery A, Greenfield S,Harris LJ, Llah T, Solomon NE, Ware JE:Assessing the quality of medical care usingoutcome measures: an overview of themethod. Med Care 15 (Suppl. 15): 1-165,1977

3. Spitzer WO, Dobson AJ, HallJ, ChestermanE, Levi J, Shephard R, Battista R, CatchloveBR: Measuring the quality of life in cancerpatients: a concise QL-index for use by phy-sicians. J Chronic Dis 34:585-97, 1981

4. Meenan RF, Gertman P, Mason JH: Mea-suring health status in arthritis: the ar-thritis impact measurement scales. Ar-thritis Rheum 23:146-52, 1980

5. Meenan RF, Gertman PM, Mason JH,Dunaif R: The arthritis impact measure-ment scales: further investigations of ahealth status measure. Arthritis Rheum25:1048-53, 1982

6. Meenan RF, Anderson JJ, Kazis LE, EggerM, Altz-Smith M, Samuelson CO, Willk-ens RF, Solsky MA, Hayes SP, Blocka KL,

476 DIABETES CARE, VOLUME 15, NUMBER 4, APRIL 1992

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/15/4/469/441272/15-4-469.pdf by guest on 15 January 2022

Hammond and Aoki

Weinstein A, Guttadauria M, Kaplan SB,Klippel J: Outcome assessment in clinicaltrials: evidence for the sensitivity of ahealth status measure. Arthritis Rheum27:1344-52, 1984

7. Croog SH, Levine S, Testa M, Brown B,Bulpitt CJ, Jenkins CD, Klerman GL,Williams GH: The effects of antihyper-tensive therapy on the quality of life. NEnglJ Med 314:1657-64, 1986

8. DCCT Research Group: Reliability andvalidity of a diabetes quality- of-life mea-sure for the Diabetes Control and Com-plications Trial (DCCT). Diabetes Care

11:725-32, 19889. Kirshner B, Guyatt G: A methodological

framework for assessing health indices. JChronic Dis 38:27-36, 1985

10. Brook RH, Ware RE Jr, Davies-Avery A,Stewart AL, Donald CA, Rogers WH,Williams KN, Johnston SA: Overview ofadult health status measures fielded inRand's Health Insurance Study. Med Care17 (Suppl. 7):1-131, 1979

11. Ware JE Jr, Johnston SA, Davies-Avery A,Brook RH: Conceptualization and Mea-surement of Health for Adults in the HealthInsurance Study: Mental Health. Vol. 3.

Santa Monica, CA, Rand Corporation,

1979

12. Bergner M, Bobbitt RA, Carter WB, Gil-

son BS: The Sickness Impact Profile: De-

velopment and final revision of a health

status measure. Med Care 19:787-805,

1981

13. Wacks MT, Starkman HS, Soeldner JS:

Hemoglobin Au. by high pressure liquid

chromatograph (HPLC). In Clinical Liq-

uid Chromatography. Vol. 2. Kobra PM,

Marton LJ, Eds. Boca Raton, HL, CRC,

1984, p. 131-38

DIABETES CARE, VOLUME 15, NUMBER 4, APRIL 1992 477

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/15/4/469/441272/15-4-469.pdf by guest on 15 January 2022