comparing disability questions for censuses and surveys in asia...

SD/Discussion Paper/1.2007

Comparing disability questions for censuses and surveys in Asia and the Pacific

Jan Smit Wei Liu*

Statistics Division

United Nations Economic and Social Commission for Asia and the Pacific

February 2007

Abstract

This paper compares the construct an predictive validity of a set of disability

questions tested on a sample of respondent in five Asia-Pacific countries (Fiji, Indonesia, India, Mongolia and the Philippines). It finds that the construct validity of the Washington Group questions for the seeing, hearing, mobility and self care domains is good when WHO questions for the corresponding domains are used as a benchmark; this does not, however, apply to the questions for the cognition and communication domains. The Washington Group questions perform similar to corresponding WHO questions in terms of predictive validity. For the four models examined — explaining difficulty with household responsibilities, work and school, and joining community activities, as well as employment status — the different question sets perform similar in terms of significance and magnitude of the odds ratios. Keywords: disability measurement, construct validity, predictive validity

* Corresponding author: [email protected].

Contents

1 Introduction 4

2 Sample selection and data 5

3 Methodology 7

4 Results 9

4.1 Prevalence comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

4.2 Association . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

4.3 Domains and factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

4.4 Disability outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

5 Discussion and conclusions 25

6 Further study 26

List of Tables

1 Descriptive Statistics by Country . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 WG to WHO population ratio at different cut-offs . . . . . . . . . . . . . . . . . 11

3 Association between Washington Group and WHO questions . . . . . . . . . . 14

4 Loading of disability questions on first 10 factors . . . . . . . . . . . . . . . . . 15

5 Activity limitations at home: Comparison of WG and WHO impairment ques-tions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

6 Activity limitations outside: Comparison of WG and WHO impairment ques-tions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

7 Participation restrictions: Comparison of WG and WHO impairment ques-tions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

8 Employment status: Comparison of WG and WHO impairment questions . . 23

9 Effect of replacing WG question with single WHO question on AIC . . . . . . . 24

A.1 Seeing: b1.6 vs. w1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

A.2 Seeing: b1.7 vs. w1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

A.3 Seeing: b1.6 and b1.7 vs. w1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2

A.4 Hearing: b1.8 vs. w2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

A.5 Hearing: b1.9 vs. w2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

A.6 Hearing: b1.8 and b1.9 vs. w2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

A.7 Mobility: b1.13 vs. w3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

A.8 Mobility: d2.5 vs. w3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

A.9 Mobility: b1.13 and d2.5 vs. w3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

A.10 Cognition: d1.1 vs. w4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

A.11 Cognition: d1.2 vs. w4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

A.12 Cognition: d1.1 and d1.2 vs. w4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

A.13 Self care: d3.1 vs. w5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

A.14 Self care: d3.2 vs. w5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

A.15 Self care: d3.1 and d3.2 vs. w5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

A.16 Communication: d1.5 vs. w6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

A.17 Communication: d1.6 vs. w6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

A.18 Communication: d1.5 and d1.6 vs. w6 . . . . . . . . . . . . . . . . . . . . . . . . 35

List of Figures

1 Comparision of prevalence at different cut-offs . . . . . . . . . . . . . . . . . . 10

3

1 Introduction

Despite recent progress, many countries in Asia and the Pacific still do not collect informa-tion on disability as part of their regular census and survey programmes. Data collectionson disability information that have occurred do not reflect the full extent of disability preva-lence or effectively capture the true needs of the target population. Moreover, the interna-tional comparability of disability statistics suffers tremendously from the wide variance ofstandards, methodologies, and definitions used.

The upcoming 2010 census round, in which the majority of the countries in the region willconduct population census activities within the next five years, presents a crucial chanceto address the deficiencies of information on persons with disabilities. A population cen-sus potentially provides a unique source of information on the prevalence of disability ina country. It can also form the basis for developing disability surveys for collecting moredetailed statistics for policy design and evaluation, such as basic care needs and the envi-ronmental constraints to the level of participation in society of persons with disabilities.

The World Health Organization (WHO) International Classification of Functioning, Dis-ability and Health (ICF) provides — since 2001 — a unified and standard concept and ap-proach for disability measurement, and has been accepted globally as part of the UnitedNations family of social and economic classifications. However, proper use of the ICF fordisability measurement has only been instituted in a limited number of countries in Asiaand the Pacific. The United Nations Economic and Social Commission for Asia and thePacific (ESCAP)/WHO 2004-2006 project on ‘Improvement of Disability Statistics and Mea-surement in Support of the Biwako Millennium Framework (BMF)’ attempted to addressthe need to introduce and implement the ICF in the region.

As part of the this project, pilot testing on ICF-based questionnaires was performed in Fiji,Indonesia, India, Mongolia and the Philippines, with the aim of formulating recommenda-tions for a standard set of disability questions for inclusion in censuses and surveys. Thedesign of the pilot test exercise included studies on the sensitivity and specificity, reliabilityand cognitive qualities of a comprehensive range of question phrasings and domains.

The ICF sees disability as a complex of difficulties, such as impairment, activity limita-tions and participation restrictions, associated with a health condition. It is impossible tomeasure all aspects of disability, and its severity, through a few questions. The number ofdisability questions that can be included in a census and survey not dedicated to disabil-ity, however, is limited. For this reason, the Washington Group on Disability Statistics, a‘City Group’ operating under the aegis of the United Nations Statistical Commission, hasformulated six questions intended to identify the population at the most basic end of thespectrum of functional activity (seeing, hearing, walking and climbing, remembering andconcentrating, self care, and communication) limitation. These six questions were there-fore included in the questionnaire (annex II, section 3, question set 1).

The questionnaire also includes disability-related questions from the WHO World HealthSurvey and questions based on the WHO Disability Assessment Scale (DAS) II (annex II,

4

section 3, question set 2, parts 2 and 3). These 53 questions encompass all disability do-mains defined in the ICF. A final group of four disability questions (annex II, section 3,question set 2, part 4) are those recommended by the Australian Bureau of Statistics (ABS)on the need for assistance domain.

This paper presents some of the results of the analysis of the pilot test data. In particular,it addresses the question how the Washington Group questions perform in comparison tothe extended set of disability questions. It does so by addressing the sensitivity and speci-ficity issue, but places it in a broader framework of construct and predictive validity. Thereliability and cognitive qualities of the questions cannot be assessed because the relevantdata have not been made available to the authors.

The outline of the remainder of this paper is as follows. Section 2 discusses the design of thesample and the collected data. Section 3 discusses some methodological issues. Section 4presents the results of the analysis, which are discussed in section 5. Section 6, to conclude,suggests avenues for further study.

2 Sample selection and data

The data was collected using translated versions of a questionnaire (attached as annex A)during May-July 2005 by the national statistical offices (NSOs) of Fiji (FJI), Indonesia (IDN),India (IND), Mongolia (MNG) and the Philippines (PHL). Each country utilized two ver-sions of the questionnaire, with the second version containing identical questions as thefirst but in reverse order to control for potential question order bias.

Each NSO was requested to administer the questionnaire to a sample of at least 1,000 re-spondents of 18 years and older representative of the national population in terms of age,sex and educational attainment. This request was interpreted by the collaborating NSOs indifferent ways. The sampling procedures followed were:

• Fiji: Since the pilot testing was conducted during the 2004-2005 Labour Force Survey(LFS), the sample was selected from the same preselected areas as the LFS. From thehousehold listing compiled for the LFS 2004-2005, a sample of 1,100 possible respon-dents was identified. The sample consisted of two rural and eight urban enumera-tion areas (although it is unclear whether these areas were completely enumerated,or that respondents were sampled from them);

• India: respondents were randomly selected from one urban enumeration block andone village in the district of Meerut in the state of Uttar Pradesh it is unclear whetherthese areas were completely enumerated, or that respondents were sampled fromthem);

• Indonesia: respondents were randomly selected in two stages from urban CentralJakarta. In the first stage, a (unknown) number of enumeration blocks were selected

5

Table 1: Descriptive Statistics by Country

N FJI IDN IND MNG PHL CombinedN = 995 N = 800 N = 590 N = 1062 N = 1056 N = 4503

Sex 4481Male 47% ( 467) 44% ( 351) 58% ( 345) 44% ( 466) 42% ( 433) 46% (2062)

Age 4503(17,34] 48% ( 482) 40% ( 324) 45% ( 267) 43% ( 459) 46% ( 485) 45% (2017)(34,49] 30% ( 299) 33% ( 266) 30% ( 178) 33% ( 351) 33% ( 346) 32% (1440)(49,64] 16% ( 162) 20% ( 158) 17% ( 100) 16% ( 165) 13% ( 140) 16% ( 725)(64,95] 5% ( 52) 6% ( 52) 8% ( 45) 8% ( 87) 8% ( 85) 7% ( 321)

Education 4264(0,6] 10% ( 93) 38% ( 292) 26% ( 117) 4% ( 43) 23% ( 244) 19% ( 789)(6,12] 53% ( 511) 52% ( 394) 43% ( 191) 53% ( 556) 61% ( 638) 54% (2290)(12,34] 37% ( 355) 10% ( 79) 31% ( 137) 43% ( 443) 16% ( 171) 28% (1185)

Marital status 4494Never married 30% ( 293) 13% ( 105) 15% ( 89) 24% ( 256) 26% ( 269) 23% (1012)Married, cohabiting 60% ( 597) 78% ( 627) 78% ( 459) 64% ( 680) 64% ( 671) 68% (3034)Sep., div., widowed 10% ( 100) 8% ( 68) 7% ( 42) 12% ( 126) 11% ( 112) 10% ( 448)

Employment status 4495Employed 50% ( 497) 47% ( 377) 51% ( 298) 49% ( 523) 57% ( 603) 51% (2298)Unemployed 3% ( 30) 4% ( 33) 3% ( 15) 13% ( 139) 9% ( 92) 7% ( 309)NILF 47% ( 460) 49% ( 390) 47% ( 277) 38% ( 400) 34% ( 361) 42% (1888)

N is the number of non–missing values. Numbers after percents are frequencies.

with probability proportional to the number of households in them. In the secondstage, 40 households were randomly selected from the selected enumeration blocks(it is unclear how respondents were seleced from the selected households);

• Mongolia: an equal number (30) of households were selected from randomly selecteddistricts in Ulaanbaatar (it is unclear how households were selected from within dis-tricts and how respondents were selected from within households);

• Philippines: three urban and two rural districts were selected on basis of high ex-pected incidence of disability, diversity in occupations, activities an income levels,as well as cost and geographical considerations. Households were then randomlyselected from ‘clusters’ (it is unclear how the clusters relate to the districts and howrespondents were selected from within households).

The demographic characteristics of the overall sample of 4,503 respondents is summarizedin table 1 (two respondents are dropped from the sample because their age is less than 18

6

years). Fiji, Mongolia and the Philippines contribute close to the requested 1,000 respon-dents each; Indonesia and especially India substantially less. The majority of respondentsis female (54%); this also holds for every country except India, where 58% of the respon-dents is male. There are no major differences in the age distribution across countries.

The majority of respondents, overall and in every country, has received between six andtwelve years education. Indonesia and the Philippines have relatively large shares of re-spondents with less than six years education and relatively small shares of respondentswith more than 12 years education; the reverse is true for Fiji and Mongolia. India has arelatively large proportion of respondents in the low and high educational categories.

Over two-thirds of all respondents are married or cohabiting, although there are differ-ences between the countries, with relatively more respondents having this marital statusin Indonesia and India than in Fiji. The proportion of respondents who have never beenmarried in the first two countries (13% and 15% respectively) is substantially lower than inthe three other countries.

A slight majority (51%) of all respondents is employed. However, this is not the case forIndonesia and Mongolia; in the former country, the number of respondents not in thelabour force (NILF) even exceeds the number employed, while in the latter unemploymentamongst respondents is high (13%). An issue here is the non-standard response categoriesattached to the ‘main work status’ question. There is, for example, no response option fornon-paid family workers. Students who spend close to half-time studying are likely to in-dicate that they are students, even though they are part of the labour force. People whodo not work because of a disabling condition are likely to indicate that they are ‘unem-ployed (health reasons)’, even though they are outside the labour force. For the purposesof this study, respondents who have selected the response opitons ‘paid work’ and ‘self em-ployed, such as own your business or farming’ are considered to be employed, and respon-dents who have selected ‘unemployed (health reasons)’ and ‘unemployed (other reasons)’are considered to be unemployed; all other respondents are considered to be outside thelabour force. This classification no doubt seriously deviates from the grouping that wouldhave been obtained had the question been asked in a standard manner.

The second column of table 1 also gives an early indication of the extent of item non-response. The age indicator is complete, but the other demographic variables (including,surprisingly, sex) are not; for 5.3% of the respondents information on the years of educationreceived is missing.

3 Methodology

The objective of this paper, as stated in section 1, is to compare the performance of Wash-ington Group questions with WHO questions for the same domains. This performance isdefined in terms of instrument validity, in particular construct validity and predictive va-lidity.

7

Within six domains, the following questions are compared:

• Seeing: w1 with b1.6 and b1.7;

• Hearing: w2 with b1.8 an b1.9;

• Mobility: w3 with b1.13 and d2.5;

• Cognition: w4 with d1.1 and d1.2;

• Self care: w5 with d3.1 and d3.2;

• Communication: w6 with d1.5 and d1.6.

The WHO questions are compared with Washington Group questions individually and com-bined. The value for respondent i for the combined WHO questions for domain j are de-rived as:

xci j = max(x1i j , x2i j ), (1)

where xc is the combined question and x1 and x2 are the individual questions within do-main j . A respondent is classified as, for example, ‘extreme/cannot do’ for the combinedb1.6 and b1.7 question (b1.6_b1.7) if s/he has answered ‘extreme/cannot do’ to either b1.6or b1.7.

One approach to assess the construct validity of disability census questions with respect tosurvey questions is to treat the latter as ‘true’ (or as the ‘gold standard’) and then to identifythe proportion of ‘true positives’ (or sensitivity) and ‘true negatives’ (or specificity) in theformer (ABS, 2003, 2006; WHO and ESCAP, 2006). This approach, however, involves makingat least two arbitrary decisions. First, there is no a priori reason to assume that answers tosurvey questions are ‘more true’ than answers to census questions. The decision to use sur-vey questions as the base of comparison, and not census questions, is therefore arbitrary.Second, the questions studied in this paper do not have binary response options. In orderto be able to calculate sensitivity and specificity a — arbitrary — decision has to be madeat what response options questions are ‘cut’ into true-false variables. As an unwelcomecorollary, information contained in the scaling of the response options is discarded.

Sensitivity and specificity are, furthermore, highly sensitive to the relative number of timesthe denominator is ‘true’ and ‘false’ (Agresti, 2002, p. 229). Finally, when the content andphrasing of census and survey questions differ, as they obviously do in the present context,potentially incomparable underlying concepts are being measured.

This paper follows an alternative approach in assessing the construct validity of Washing-ton Group variables by asking three questions:

• To what extent do Washington Group and WHO questions agree — for each of the sixdomains — in terms of estimated disability prevalence?

8

• To what extent do each respondent’s answers to Washington Group questions corre-spond to her/his answer to WHO questions for each of the six domains?

• Do the Washington Group and corresponding WHO questions belong to the samelatent factors in the dataset?

The predictive validity of disability variables has in several studies been assessed throughstepwise linear regression. Cieza et al. (2006) regress ICF-based questions against a self-assessed measure of overall health. McMenamin et al. (2006) regress candidate disabilityquestions against a disability index constructed by experts but apparently using the samequestions. There are, however, at least three problems with such an approach in the presentcontext. First, the dataset under study does not contain an overall health outcome variable;the construction of a disability index as a dependent variable and then regressing it againstvariables that have been utilized in its construction is theoretically unappealing. Second,there are many theoretical and conceptual problems associated with stepwise procedureswhich make them less suitable for purposes beyond initial model screening.1 Third, a linearregression approach is less suitable when the dependent variable can only take on a limitednumber of discrete values.

This paper tests the predictive validity of the Washington Group questions in comparison toWHO questions by examining the power of both in explaining some key variables in otherdomains: ‘household responsibilities’ (d5.1-d5.4) and ‘work and school tasks’ (d5.5-d5.8) inthe life activities domain, and ‘joining community activities’ (d6.1) in the participation do-main, all using ordinal logistic regression models (in particular, proportional odds models),and ‘employment status’ (a5) using multinomial logistic regressions models.

4 Results

4.1 Prevalence comparison

The prevalence of disability in the six domains generated by the different Washington Groupand WHO questions is compared in figure 1. Each of the panels in this figure compares thesample disability prevalence in each of the six domains on the basis of the WashingtonGroup question, to the two ‘corresponding’ WHO questions, as well as the latter two ques-tions combined. The horizontal axis of each panel measures disability prevalence, startingwith the most extreme cases. The first dot for question w1, for example, represents thepercentage of respondents who say they are ‘unable’ to see, even if wearing glasses. Thesecond dot adds to the respondents who are unable to see those who have ‘a lot’ of diffi-culty seeing; the dot represents, therefore, the respondents in the ‘top-two boxes’. The thethird and final dot for question w1 represents the percentage of respondents who have atleast ‘some’ difficulty seeing.

1These problems are summarized at http://www.stata.com/support/faqs/stat/stepwise.html.

9

http://www.stata.com/support/faqs/stat/stepwise.html

10

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40

d1_1&2

d1_2

d1_1

w4

d1_5&6

d1_5

d1_6

w6

b1_8&9

b1_8

b1_9

w2

b1_13&d_5

b1_13

d2_5

w3

b1_6&7

b1_6

w1

b1_7

d3_1&2

d3_1

d3_2

w5

Cognition Communication Hearing

Mobility Seeing Self care

unable severesome

Source: ESCAP/WHO pilot test.

Cumulative prevalence (%)

0 2 4 6 8 0 2 4 6 8 0 2 4 6 8

w4

d1_1&2

d1_2

d1_1

w6

d1_5&6

d1_6

d1_5

b1_8&9

b1_8

w2

b1_9

b1_13&d_5

w3

d2_5

b1_13

w1

b1_6&7

b1_6

b1_7

d3_1&2

d3_1

w5

d3_2

Cognition Communication Hearing

Mobility Seeing Self care

unable severeSource: ESCAP/WHO pilot test.

Cumulative prevalence (%)

Figure 1: Comparison of prevalence at different cut-offs

Two main conclusions emerge from figure 1. First, there is reasonable correspondence upto the second most severe response option (‘a lot’ and ‘severe’ difficulty) between preva-lence as measured through the Washington Group questions on the one hand and the com-bined WHO questions on the other for four domains: seeing (4.2% and 4.3%), cognition(3.3% and 3.2%), self care (1.5% and 1.7%) and communication (1.9% and 2.1%). The ratiobetween the respondents being picked up by the Washington Group and combined WHOquestions for these domains varies, as shown in table 2 between 0.95 and 1.02. However, forthe self care domain, the ratio between respondents in the top-two boxes of an individualWHO question (d3.1, ‘washing’) and the same boxes of the correspondending WashingtonGroup question (w5) is exactly 1.0. The ratio between respondents who say that have atleast ‘mild’ difficulty seeing (w1) and who have ‘some’ short-vision difficulty (b1.6) is alsoclose to one.

Table 2: WG to WHO population ratio at different cut-offs

Domain WHO question Response Extreme Severe Moderate MildSeeing b1.6 Unable 0.62 0.15 0.06 0.03(WG w1) A lot 5.53 1.38 0.50 0.23

Some 24.00 6.00 2.19 1.01b1.7 Unable 0.68 0.17 0.07 0.03

A lot 6.06 1.49 0.60 0.26Some 26.29 6.47 2.61 1.14

b1.6 and b1.7 Unable 0.44 0.11 0.04 0.02A lot 3.92 0.97 0.38 0.18Some 17.00 4.23 1.64 0.77

Hearing b1.8 Unable 0.68 0.19 0.06 0.02(WG w2) A lot 2.68 0.73 0.24 0.09

Some 15.58 4.23 1.40 0.52b1.9 Unable 0.68 0.25 0.08 0.03

A lot 2.68 0.96 0.31 0.11Some 15.53 5.57 1.82 0.64

b1.8 and b1.9 Unable 0.50 0.15 0.05 0.02A lot 1.96 0.61 0.20 0.07Some 11.38 3.52 1.15 0.43

Mobility b1.13 Unable 1.22 0.27 0.10 0.04(WG w3) A lot 5.30 1.17 0.44 0.17

Some 18.28 4.04 1.51 0.59b2.5 Unable 0.60 0.22 0.11 0.05

A lot 2.60 0.95 0.46 0.21Some 8.94 3.28 1.58 0.73

b1.13 and d2.5 Unable 0.52 0.17 0.07 0.03A lot 2.24 0.72 0.31 0.14Some 7.71 2.46 1.06 0.48

Cognition d1.1 Unable 1.14 0.30 0.08 0.03

11

(WG w4) A lot 6.95 1.80 0.48 0.17Some 38.90 10.09 2.66 0.93

d1.2 Unable 1.41 0.20 0.07 0.02A lot 8.65 1.24 0.40 0.12Some 48.12 6.87 2.22 0.69

d1.1 and d1.2 Unable 0.89 0.17 0.05 0.02A lot 5.44 1.02 0.29 0.10Some 30.30 5.68 1.63 0.57

Self care d3.1 Unable 1.20 0.43 0.23 0.13(WG w5) A lot 2.76 1.00 0.54 0.30

Some 7.40 2.68 1.45 0.79d3.2 Unable 1.88 0.71 0.34 0.18

A lot 4.31 1.64 0.78 0.42Some 11.56 4.40 2.08 1.13


Communication d1.5 Unable 1.25 0.43 0.11 0.04(WG w6) A lot 4.40 1.52 0.39 0.13

Some 17.90 6.17 1.60 0.55d1.6 Unable 0.83 0.33 0.10 0.04

A lot 2.93 1.16 0.37 0.13Some 11.97 4.72 1.49 0.55


For the hearing (1.1% and 1.9%) and mobility (5.6% and 8.3%) domains the combined WHOquestions pick up substantively more respondents under the aforementioned response op-tions (i.e., the top-two boxes) than the Washington Group questions. The number of re-spondents who said that they had at least ‘severe’ difficulty hearing what was said in aconversation (b1.9) alone is, on the other hand, close to the number of people who saidthat they had ‘a lot’ of difficulty hearing (w2). For the mobility domain, the ratio betweenrespondents in the top-three boxes of both the Washington Group and combined WHOquestions is closest to one (1.06).

It should be noted that a ratio close to one at any cut-off point on the scales of the questions(in the case of WHO versions, individually or combined) compared does not necessarilymean that the same respondents are being identified. Examination of the contingency ta-bles in annex I shows that the overlap between respondents in the top-two boxes for theWashington Group questions and combined WHO questions, as a percentage of the for-mer, is just 47% for the seeing domain, 46% for the cognition domain, and 47% for thecommunication domain. For the remaining domains, the overlap is over 50%: namely, 51%for the hearing domain, 70% for the mobility domain, and 56% for the self care domain.

12

The overlaps at various cut-off points is reflected in the measures of association discussedin the next subsection.

The second main conclusion from figure 1 is that the number of subjects reporting anydifficulty in response to the combined WHO questions is larger than in response to theWashington Group questions for every domain. This is especially the case for the hearing,mobility and communication domains, where the combined WHO questions pick up morethan twice the number of people with at least some difficulty than the Washington Groupquestions. Moreover, each individual WHO question for these domains, as well as the cog-nition domain, generates higher disability prevalence than the corresponding WashingtonGroup question.

4.2 Association

Table 3 presents the association between the Washington Group questions for the six do-mains on the one hand and the individual and combined corresponding WHO questionson the other, using three common measures for comparing ordinal variables: Kendall’s tau-b, Kendall-Stuart’s tau-c and Goodman-Kruskal’s gamma. For the sake of completeness,the table also has columns for sensitivity and specificity (with positive cases representingrespondents reporting any difficulty).

Given two ordinal variables x and y , Kendall’s tau-b is defined as the excess of concordant(C ) over discordant (D) pairs, divided by a term representing the geometric mean betweenthe number of pairs tied on x by not on y (Tx) and the number of pairs tied on y but not onx (Ty ):

τb =C −D√

(C +D +Tx)(C +D +Ty ). (2)

Kendall-Stuart’s tau-c is defined as the excess of concordant over discordant pairs, adjust-ing for the number of rows and columns of the contingency table of x and y :

τc = (C −D)2mn2(m −1) , (3)

where n is the sample size and m is the minimum of the number of rows and columns ofthe contingency table.

Goodman-Kruskal’s gamma, finally, is defined simply as the proportion of the excess ofconcordant over discordant pairs over the total number of pairs, ignoring ties:

γ= C −DC +D . (4)

The three measures of association have in common that they are 0 in case of independencebetween x and y , and range in value from -1 to +1, which represent negative and positive

13

Table 3: Association between Washington Group and WHO questions

Domain WHO question Tau-b Tau-c Gamma Sensitivity Specificity

Seeing b1.6 0.50 0.21 0.83 0.60 0.91(WG w1) b1.7 0.51 0.20 0.84 0.63 0.90

b1.6 and b1.7 0.56 0.26 0.85 0.58 0.94Hearing b1.8 0.49 0.11 0.91 0.39 0.98(WG w2) b1.9 0.43 0.09 0.88 0.38 0.97

b1.8 and b1.9 0.47 0.12 0.90 0.34 0.98Mobility b1.13 0.50 0.28 0.80 0.47 0.93(WG w3) d2.5 0.49 0.26 0.79 0.51 0.91

b1.13 and d2.5 0.52 0.31 0.82 0.43 0.96Cognition d1.1 0.36 0.15 0.70 0.46 0.88(WG w4) d1.2 0.47 0.22 0.79 0.48 0.92

d1.1 and d1.2 0.46 0.24 0.79 0.43 0.94Self care d3.1 0.56 0.07 0.96 0.51 0.98(WG w5) d3.2 0.53 0.05 0.96 0.58 0.98

d3.1 and d3.2 0.57 0.07 0.96 0.49 0.99Communication d1.5 0.36 0.10 0.80 0.31 0.96(WG w6) d1.6 0.36 0.10 0.80 0.31 0.96

d1.5 and d1.6 0.37 0.11 0.80 0.28 0.97

respectively perfect monotonicity (weak in the case of gamma, strong in the case of thetaus). They give the same values in the absence of ties (τb = τc = γ). The advantage ofgamma is its ease of interpretation; however, it inflates the level of association because itignores ties. The tau measures do not have straightforward intuitive interpretations, buthave the advantage that they do not discard information on ties. Tau-b penalizes ties lessthan tau-c. In the presence of ties, τc < τb < γ.Table 3 shows that there is reasonably close association between the Washington Groupand WHO questions — individually and combined — for the self care, seeing and mobilitydomains in terms of tau-b (0.5 or higher) and gamma (0.8 or higher). For each of thesethree domains, the association is highest when the WHO questions are combined. In termsof tau-c the association is much lower for the self care domain. This is caused by the lowprevalence of (any) self care difficulties, as discussed in foregoing subsection and illustratedin figure 1, resulting in many ties on the variables within this domain.

The association between the Washington Group and WHO questions for the hearing do-main, again in terms of tau-b and gamma, is somewhat lower than for the the three afore-mentioned domains. Interestingly, the association between question b1.8 (‘hearing some-one talking’) and question w2 is higher than between the latter and questions b1.8 and b1.9(‘hearing what is said in a conversation’) combined. A similar issue applies to the cognition

14

domain: the association between question d1.2 (‘remembering’) and question w4 is higherthan between the latter and d1.2 and d1.1 (‘concentrating’) combined. The association be-tween the Washington Group and WHO questions for the communication domain is ratherlow, particularly in terms of tau-b.

4.3 Domains and factors

Table 4 presents the results of exploratory factor analysis conducted on all the disabilityvariables in the dataset using the Kendall’s tau-b correlation matrix as input. The primaryobjective of this exercise is to check whether the Washington Group questions and WHOquestions for the respective domains are part of the same underlying latent constructs (fac-tors). The analysis allows, however, two further questions to be addressed. First, as the fac-tor loadings are the correlation coefficients between the disability variables and the factors,the exercise also indicates how the number of questions asked might be reduced (or mini-mized). Second, the factors present the ‘statistical’ dimensions of the dataset which can becross-checked against the ‘stated’ domains in the questionnaire (and the ICF).

Table 4: Loading of disability questions on first 10 factors

Var. F1 F2 F3 F4 F5 F6 F7 F8 F9 F10w1 0.12 0.01 0.03 0.11 0.12 0.06 0.02 0.57 0.11 0.09w2 0.10 0.06 0.04 0.03 0.10 0.49 -0.02 0.09 0.10 0.18w3 0.19 -0.05 0.22 0.09 0.43 0.05 0.13 0.20 0.23 0.19w4 0.29 0.14 0.10 0.07 0.11 0.16 0.23 0.16 0.15 0.28w5 0.12 0.06 0.50 0.10 0.08 0.04 0.04 0.02 0.10 0.22w6 0.16 0.21 0.24 0.12 0.05 0.18 0.14 0.10 0.03 0.29b1.1 0.22 0.04 0.08 0.20 0.16 0.05 0.11 0.10 0.66 0.03b1.2 0.22 0.07 0.08 0.20 0.14 0.04 0.10 0.09 0.66 -0.02b1.3 0.16 0.12 0.12 0.22 0.02 0.11 0.09 0.15 0.12 -0.04b1.4 0.15 0.09 0.32 0.13 0.06 0.14 0.10 0.17 0.08 0.00b1.5 0.16 0.09 0.29 0.01 0.18 0.17 0.14 0.15 0.12 0.03b1.6 0.17 0.13 0.08 0.05 0.17 0.16 0.09 0.62 0.08 -0.03b1.7 0.18 0.06 0.10 0.06 0.12 0.14 0.07 0.62 0.07 0.01b1.8 0.17 0.12 0.07 0.11 0.12 0.68 0.08 0.10 0.03 -0.02b1.9 0.14 0.15 0.09 0.09 0.07 0.69 0.06 0.11 0.03 0.01b1.10 0.22 0.11 0.23 0.01 0.06 0.24 0.15 0.15 0.15 -0.13b1.11 0.19 0.06 0.25 0.07 0.05 0.15 0.18 0.14 0.17 -0.16b1.12 0.16 0.03 0.13 0.15 0.31 0.22 0.34 0.11 0.19 -0.13b1.13 0.22 0.12 0.04 0.17 0.47 0.09 0.31 0.15 0.21 -0.01b1.14 0.16 0.12 0.07 0.15 0.21 0.20 0.27 0.13 0.12 -0.13b1.15 0.22 0.08 0.05 0.21 0.19 0.15 0.33 0.11 0.15 -0.02b1.16 0.32 0.13 0.06 0.26 0.13 0.01 0.48 0.06 0.22 0.01

Continued on next page...

15

... table 4 continued

Var. F1 F2 F3 F4 F5 F6 F7 F8 F9 F10b1.17 0.33 0.11 0.08 0.27 0.11 0.07 0.50 0.04 0.24 0.02d1.1 0.34 0.34 0.08 0.08 0.15 0.15 0.31 0.10 0.08 0.05d1.2 0.39 0.27 0.05 0.05 0.13 0.12 0.36 0.09 0.11 0.16d1.3 0.48 0.34 0.01 0.17 0.10 0.12 0.32 0.12 0.08 0.05d1.4 0.39 0.37 0.07 0.13 0.12 0.10 0.32 0.12 0.03 0.12d1.5 0.32 0.44 0.12 0.14 0.10 0.22 0.30 0.11 -0.07 0.12d1.6 0.34 0.48 0.13 0.12 0.08 0.23 0.24 0.07 -0.01 0.08d2.1 0.24 0.12 0.15 0.16 0.58 0.12 0.06 0.11 0.15 0.02d2.2 0.29 0.10 0.23 0.13 0.45 0.13 0.06 0.12 0.18 0.07d2.3 0.34 0.16 0.38 0.11 0.38 0.17 0.04 0.15 0.01 -0.11d2.4 0.33 0.14 0.37 0.17 0.38 0.15 0.02 0.17 -0.00 -0.03d2.5 0.32 0.16 0.16 0.17 0.54 0.08 0.08 0.13 0.12 -0.01d3.1 0.14 0.09 0.70 0.08 0.14 0.06 0.06 0.02 0.07 -0.01d3.2 0.18 0.10 0.67 0.03 0.10 0.06 0.04 0.03 0.06 -0.06d3.3 0.15 0.16 0.40 0.03 0.09 0.07 0.08 0.07 0.05 -0.10d3.4 0.28 0.25 0.20 0.20 0.14 0.12 0.18 0.08 0.01 0.01d4.1 0.35 0.54 0.04 0.15 0.10 0.08 0.08 -0.00 0.04 0.03d4.2 0.31 0.59 0.10 0.10 0.11 0.08 0.00 0.05 0.11 -0.02d4.3 0.30 0.52 0.19 0.07 0.09 0.13 0.03 0.08 0.06 -0.01d4.4 0.34 0.58 0.08 0.13 0.10 0.05 0.01 0.01 0.07 0.02d4.5 0.29 0.26 0.23 0.12 0.18 0.16 0.02 0.05 0.07 -0.05d5.1 0.61 0.17 0.16 0.15 0.12 0.07 0.11 0.08 0.03 -0.03d5.2 0.66 0.14 0.18 0.12 0.11 0.08 0.08 0.07 0.08 -0.06d5.3 0.73 0.14 0.14 0.11 0.11 0.07 0.10 0.04 0.07 -0.09d5.4 0.74 0.13 0.08 0.14 0.14 0.04 0.07 0.03 0.07 -0.03d5.5 0.72 0.11 0.11 0.10 0.09 0.08 0.07 0.09 0.08 0.05d5.6 0.76 0.14 0.08 0.11 0.07 0.05 0.03 0.09 0.09 0.10d5.7 0.80 0.13 0.10 0.12 0.07 0.09 0.07 0.09 0.09 0.04d5.8 0.77 0.12 0.05 0.15 0.11 0.07 0.07 0.05 0.11 0.06d6.1 0.36 0.35 0.14 0.24 -0.00 0.03 0.04 0.13 0.04 -0.10d6.2 0.37 0.38 0.13 0.35 -0.01 0.05 0.07 0.11 0.04 -0.07d6.3 0.38 0.41 0.09 0.32 -0.03 0.13 0.09 0.07 0.03 -0.03d6.4 0.32 0.16 0.07 0.48 0.11 0.06 0.06 0.04 0.20 0.04d6.5 0.39 0.11 0.08 0.55 0.17 0.07 0.16 0.02 0.25 0.08d6.6 0.34 0.14 0.09 0.62 0.16 0.07 0.11 0.07 0.18 0.01d6.7 0.38 0.17 0.13 0.60 0.15 0.13 0.10 0.07 0.14 0.02d6.8 0.43 0.21 0.22 0.41 0.09 0.13 0.11 0.10 0.04 -0.05e1 0.19 0.06 0.48 0.24 0.06 0.06 -0.02 0.10 0.03 0.18e2 0.19 0.07 0.45 0.16 0.06 -0.03 -0.06 0.13 0.02 0.20e3 0.15 0.21 0.21 0.21 0.00 0.12 0.13 0.12 -0.07 0.31

16

The nine dimensions (factors) below emerge from the preliminary results, in order of theircontribution to the overall variance in the dataset. A variable with a loading of 0.4 or higheron a factor is — admittedly, rather arbitrarily — considered to ‘belong’ to that factor; fromthe 10th factor onwards, all factor loadings are lower than 0.4.

• F1: Day-to-day life (d1.3, d5.1, d5.2, d5.3, d5.4, d5.5, d5.6, d5.7, d5.8 and d6.8);

• F2: Getting along with people (d1.5, d1.6, d4.1, d4.2, d4.3, d4.4 and d6.3);

• F3: Self care (w5, d3.1, d3.2, d3.3, e1 and e2);

• F4: Resources and emotion (d6.4, d6.5, d6.6, d6.7 and d6.8);

• F5: Mobility (w3, b1.13, b2.1, b2.2 and b2.5);

• F6: Hearing (w2, b1.8 and b1.9);

• F7: Sadness and anxiety (b1.16 and b1.17);

• F8: Seeing (w1, b1.6 and b1.7);

• F9: Pain and discomfort (b1.1 and b1.2).

An important first result is that out of the six domains covered by Washington Group ques-tions four, namely seeing, hearing, mobility and self care, are generated as factors by theanalysis. Not only do the Washington Group questions load relatively high (≥ 0.4) on thesefactors, all of the corresponding WHO questions — and for the self care factor (F3) also theABS questions — do so as well. There is, in another words, substantive correlation betweenall variables for these four domains.

This is not the case for the questions addressing the remaining two domains covered byWashington Group questions, cognition and communication. The loadings of each Wash-ington Group and WHO question for these two domains are lower than 0.4, and the analysisfails to generate cognition and communication factors. There might be various reasons forthe lack of correlation among the variables for these two domains (which is, of course, aconfirmation of the analysis of the previous subsection). The two concepts, ‘remembering’and ‘concentrating’, asked about in Washington Group question w4 are inherently different.The question is, therefore, double-barreled, and as such potentially confusing for respon-dents. Washington Group question w6 for the communication domain is double-barreledas well; having difficulty ‘understanding’ and having difficulty ‘being understood’ are quitedifferent kinds of disabilities. The rather lengthly preamble to the question adds furtherpotential respondent confusion.

Another important issue with regard to the domains not discussed so far is the extent towhich the predefined (in the questionnaire) disability domains coincide with the domains

17

(factors) extracted by the factor analysis. A high level of agreement does not only indicatethat questions are highly valid for their respective domains, but also that asking one ques-tion — rather than all for the same domain — suffices.

Factor F1 (‘day-to-day activities’) does not only comprise all questions for the ‘life activities’domain (d5.1 to d5.8), but also question d1.3 (‘analyzing and finding solutions to problemsin day-to-day life’) and question d6.8 (‘relaxation and pleasure’). Factor F2 (‘getting alongwith people’) comprises all questions in the predefined ‘getting along with people’ domainexcept the last (d4.5, ‘sexual activities’), and also includes d6.3 (‘living with dignity’) fromthe predefined ‘participation in society’ domain. Factor F4 (‘resources and emotion’) ismade up of a subset of the ‘participation in society’ domain. It can be argued that of theother questions in that domain, the first (d6.1, ‘joining in community activities’) is the onlyone that asks directly about participation in society, and that the remaining two (d6.2, ‘bar-riers and hindrances’ and d6.3, ‘living with dignity’) are either environmental factors or askrespondents perception of how well they get along with others (rather than being partici-pation variables per se). Indeed, as mentioned before, d6.3 loads relatively high on factorF2 (‘getting along with people’).

The remaining factors, F7 (‘sadness and anxiety’) and F9 (‘pain and discomfort’), each con-sist of two apparently closely related questions in the ‘understanding and communicating’and body functions domains.

4.4 Disability outcomes

This subsection examines the extent of difference, if any, in the power of Washington Groupand corresponding WHO questions in explaining difficulties with regard to ‘household re-sponsibilities’ (d5.1-d5.4) and ‘work and school tasks’ (d5.5-d5.8) in the life activities do-main, and ‘joining community activities’ (d6.1) in the participation domain, all using ordi-nal logistic regression models (in particular, proportional odds models), and ‘employmentstatus’ (a5) using multinomial logistic regressions models.

The results of this analysis are shown in tables 5, 6, 7 and 8. Each table presents the results ofthree specifications. Specification 1 uses the Washington Group questions on seeing, hear-ing, mobility, cognition, self care and communication as covariates, together with all thedemographic variables in the dataset. Specification 2 substitutes the Washington Groupquestions with the combined (see equation 1) corresponding WHO questions. Specifica-tion 3 adds to specification 1 four covariates suggested by the factor analysis conductedin subsection 4.3 as being important: ‘getting along with people’ (d4.1-d4.4), ‘resources’(d6.6), ‘sadness and anxiety’ (b1.16 and b1.17) and ‘pain and discomfort’ (b1.1 and b1.2).The variable ‘male’ is an indicator variable expanded from the factor ‘sex’ (a1) with ‘female’as the reference level. The reference level of the factor ‘marital status’ (a4 collapsed intothree levels) is ‘never married’. India is the reference level of the factor ‘country’.

The tables show for each model the exponentiated coefficient estimates — or odds ratios— and the p-values. The odds ratios represent the effect of a one unit increase in a covari-

18

Table 5: Activity limitations at home: Comparison of WG and WHO impairment questions

(1) (2) (3)WG WHO WG plus

Household responsibilitiesSeeing 1.205 (0.005) 1.102 (0.033) 0.992 (0.910)Hearing 0.861 (0.174) 0.987 (0.828) 0.677 (0.001)Mobility 2.245 (0.000) 1.985 (0.000) 1.760 (0.000)Cognition 1.930 (0.000) 1.865 (0.000) 1.163 (0.058)Self care 2.807 (0.000) 2.563 (0.000) 2.300 (0.000)Communication 1.802 (0.000) 1.651 (0.000) 1.002 (0.983)Male 0.973 (0.697) 1.005 (0.950) 1.050 (0.516)Age 0.983 (0.182) 0.984 (0.212) 0.966 (0.011)Age squared 1.000 (0.116) 1.000 (0.256) 1.000 (0.001)Education 0.983 (0.060) 0.996 (0.638) 1.002 (0.848)Married, cohabiting 1.265 (0.022) 1.249 (0.037) 1.146 (0.220)Sep., div., widowed 1.227 (0.183) 1.198 (0.256) 1.151 (0.390)Indonesia 2.879 (0.000) 2.978 (0.000) 1.604 (0.002)Philippines 5.003 (0.000) 3.112 (0.000) 3.330 (0.000)Fiji 4.200 (0.000) 3.563 (0.000) 2.919 (0.000)Mongolia 1.506 (0.006) 1.548 (0.006) 1.016 (0.920)Getting along 2.239 (0.000)Resources 1.905 (0.000)Sadness, anxiety 1.391 (0.000)Pain, discomfort 1.330 (0.000)Observations 4357 4404 4317Pseudo R2 0.149 0.235 0.272AIC 7558.6 6845.6 6430.2BIC 7686.2 6973.4 6583.1

Exponentiated coefficients; p-values in parentheses

ate on the dependent variable. A one-step increase on the scale for the Washington Group‘seeing’ variable (table 5, specification 1), for example, increases the odds of respondentsreporting difficulty with any household responsibilities by 21% across the entire range ofthe variable. Similarly, the odds of male respondents not being in the labour force (ta-ble 8, all three specifications) are 88% lower than for female respondents. The scores forthe Washington Group responses {no, some, a lot, unable} = {1, 2, 3, 4} and for the WHOquestions {none, mild, moderate, severe, extreme/cannot do} = {1, 2, 3, 4, 5}. Experimentswith recoding the WHO covariates to the midpoints of the ranges suggested in WHO (2001)made the models and specifications fit less on all three the presented measures, the Mc-Fadden pseudo R2, the Akaike information criterion (AIC) and the Bayes (Schwarz) infor-mation criterion (BIC), as did treating every level of these predictors as qualitative.

There are few differences between specifications 1 (Washington Group questions) and 2

19

Table 6: Activity limitations outside: Comparison of WG and WHO impairment questions


Work, schoolSeeing 1.329 (0.001) 1.227 (0.001) 1.197 (0.043)Hearing 1.309 (0.065) 1.232 (0.005) 0.955 (0.759)Mobility 2.106 (0.000) 1.993 (0.000) 1.669 (0.000)Cognition 2.096 (0.000) 1.700 (0.000) 1.153 (0.143)Self care 1.630 (0.004) 1.693 (0.000) 1.332 (0.091)Communication 2.195 (0.000) 1.720 (0.000) 1.315 (0.049)Male 0.909 (0.254) 0.967 (0.702) 1.068 (0.468)Age 0.995 (0.761) 1.000 (0.998) 0.999 (0.976)Age squared 1.000 (0.603) 1.000 (0.946) 1.000 (0.642)Education 0.991 (0.417) 0.998 (0.881) 1.021 (0.067)Married, cohabiting 0.954 (0.697) 0.898 (0.390) 0.799 (0.086)Sep., div., widowed 0.761 (0.163) 0.621 (0.019) 0.659 (0.044)Indonesia 2.819 (0.000) 2.840 (0.000) 1.395 (0.057)Philippines 3.088 (0.000) 1.772 (0.000) 2.007 (0.000)Fiji 5.351 (0.000) 4.923 (0.000) 3.604 (0.000)Mongolia 0.784 (0.181) 0.868 (0.460) 0.524 (0.001)Getting along 2.414 (0.000)Resources 1.768 (0.000)Sadness, anxiety 1.401 (0.000)Pain, discomfort 1.344 (0.000)Observations 3178 3209 3150Pseudo R2 0.146 0.231 0.270AIC 5195.4 4726.9 4428.3BIC 5316.7 4848.4 4573.6


(WHO questions) for each of the response variables modeled in tables 5, 6 and 7. The oddsratios of the seeing, mobility, cognition, self care and communication variables are all largerthan one — as expected — and significant at the 1% level, except the seeing variable in spec-ification 2 of the ‘household responsibilities’ model (table 5), which is significant at the 5%level. The odds ratios of the hearing variable are less than one for both specifications 1 and2 of the ‘household responsibilities’ model, but insignificant so at the 10% level. Hearingdifficulties do increase the odds of having difficulty with school and work and in joiningcommunity activities for both specifications 1 and 2, but insignificantly so for the latterdependent variable.

A number of other interesting and perhaps surprising observations arise from tables 5, 6and 7, which — again — do not differ much between specifications 1 and 2. First, neithersex nor age appear to have any significant bearing on the odds of respondents reporting

20

Table 7: Participation restrictions: Comparison of WG and WHO impairment questions


Community activitiesSeeing 1.313 (0.001) 1.202 (0.001) 1.146 (0.102)Hearing 1.173 (0.208) 0.972 (0.676) 1.046 (0.727)Mobility 1.810 (0.000) 1.555 (0.000) 1.507 (0.000)Cognition 1.683 (0.000) 1.553 (0.000) 1.169 (0.084)Self care 2.189 (0.000) 2.325 (0.000) 1.940 (0.000)Communication 2.161 (0.000) 1.837 (0.000) 1.386 (0.004)Male 1.092 (0.315) 1.111 (0.244) 1.188 (0.062)Age 0.999 (0.950) 0.999 (0.969) 0.991 (0.570)Age squared 1.000 (0.767) 1.000 (0.678) 1.000 (0.656)Education 1.020 (0.085) 1.034 (0.004) 1.045 (0.000)Married, cohabiting 0.883 (0.324) 0.872 (0.296) 0.789 (0.076)Sep., div., widowed 1.170 (0.399) 1.249 (0.243) 1.195 (0.356)Indonesia 3.067 (0.000) 3.294 (0.000) 1.946 (0.002)Philippines 6.778 (0.000) 4.498 (0.000) 4.979 (0.000)Fiji 3.923 (0.000) 3.293 (0.000) 2.902 (0.000)Mongolia 1.622 (0.022) 1.635 (0.026) 1.317 (0.216)Getting along 1.997 (0.000)Resources 1.519 (0.000)Sadness, anxiety 1.084 (0.117)Pain, discomfort 1.281 (0.000)Observations 4295 4335 4257Pseudo R2 0.144 0.206 0.219AIC 5047.7 4720.8 4590.3BIC 5175.1 4848.3 4742.9


difficulty with household responsibilities, work or school, and in joining community activ-ities. Respondents who are married or cohabiting are 25-26% more likely to report havingmore difficulties with household responsibilities than those who have never been married.Those who are separated, divorced or widowed are more likely to have more difficultiesthan those who were never married with work and school (but insignificantly so underspecification 1). Marital status does not otherwise have a significant impact on activitiesand participation. An additional year of education has some impact (2-3%) on the odds ofreporting difficulty in joining community activties, but the significance of this variable islow for specification 1 (Washington Group questions).

Second, the odds of reporting difficulty with household responsibilities, work and school,and in joining community activities do, in contrast to other demographic variables, in bothspecifications 1 and 2 vary with the nationality of respondents. The odds of Indonesians,

21

Filipinos, Fijians and Mongolians reporting difficulty with household responsibilities are1.5 to five times higher than for Indians (the reference category). The odds of Indonesians,Filipinos and Fijians reporting difficulty with work or school are three to five times higherthan for Indians. The difference between the latter and Mongolians, on the other, is in-significant for this dependent variable. Respondents of all countries (including Mongolia)are significantly (from 60% to almost seven times) more likely to select an higher option onthe response scale for experiencing difficulty joining community activities than Indians.It should be remembered — from section 2 — that the country samples are by no meansnationally representative, so that the magnitude and significance of the country indicatorvariables might more reflect sample bias than ‘real’ country differences.

Third, the variables ‘getting along with people’, ‘resources’, ‘sadness and anxiety’ and ‘painand difficulty’, all framed in terms of having problems, have a significant impact on respon-dents reporting having difficulty with household responsibilities, school and work, and injoining community activities (with the exception of sadness and anxiety for the latter de-pendent variable). The inclusion of these variables also has affects the direction of the oddsand significance of other covariates. For the ‘household responsibilities’ variable, for exam-ple, the seeing and communication variables become insignificant, while the hearing andage variables become significant (but change in odds for the latter is negligible). Cogni-tion and, to a lesser extent self care, become insignificant in the model for work and schooldifficulties.

The observations with regard to the relationship between Washington Group variables andcomparable WHO variables on the one hand and employment status on the other (table8) are similar. The odds for respondents who report a higher level of difficulty with seeing,mobility and self care of being unemployed are — relative to being employed — signifi-cantly higher under both specifications 1 and 2; the hearing and cognition variables areinsignificant for both specifications. The specifications contradict for communication; thevariable is highly significant under specification 1 but insignificant for specification 2.

Under both specifications the odds of being unemployed increase — at a decreasing rate —with age, and decrease — perhaps unexpectedly — with education. The odds for respon-dents who are married or cohabiting of being unemployed are 44-51% lower than thosewho were never married (the effect of being separated, divorced or widowed is insignifi-cant). Filipinos and Mongolians are between two and almost six times respecitively morelikely to be unemployed, everything else being equal, than Indians. Problems with gettingalong with people and sadness and anxiety have no significant effect on unemployment.Pain and discomfort, on the other hand, increase the odds of being unemployed.

The Washington Group and WHO impairment questions also largely agree with regard tothe extent that hearing, mobility, cognition, self care and communication difficulties in-crease the odds of respondents not being in the labour force (relative to being employed);they are all insignificant. Only the combined WHO seeing questions have a significant (pos-itive) impact. The odds ratios for the male dummy, age, age squared and education areremarkably close to each other under specifications 1 and 2, and are all highly significant.Marital status, on the other hand, does not appear to effect the odds of not being in the

22

Table 8: Employment status: Comparison of WG and WHO impairment questions


UnemployedSeeing 1.326 (0.029) 1.305 (0.002) 1.188 (0.192)Hearing 1.303 (0.168) 1.133 (0.234) 1.278 (0.201)Mobility 1.263 (0.079) 1.188 (0.025) 1.084 (0.558)Cognition 0.817 (0.209) 0.977 (0.818) 0.705 (0.035)Self care 2.041 (0.001) 1.589 (0.000) 1.786 (0.007)Communication 1.825 (0.001) 1.030 (0.785) 1.721 (0.003)Male 0.816 (0.132) 0.831 (0.167) 0.827 (0.163)Age 0.861 (0.000) 0.869 (0.000) 0.852 (0.000)Age squared 1.002 (0.000) 1.002 (0.000) 1.002 (0.000)Education 0.956 (0.015) 0.956 (0.013) 0.959 (0.025)Married, cohabiting 0.561 (0.002) 0.493 (0.000) 0.548 (0.002)Sep., div., widowed 0.708 (0.241) 0.636 (0.121) 0.686 (0.207)Indonesia 1.787 (0.084) 1.758 (0.085) 1.574 (0.181)Philippines 2.640 (0.002) 2.220 (0.008) 2.440 (0.004)Fiji 1.237 (0.541) 1.054 (0.878) 1.081 (0.826)Mongolia 5.659 (0.000) 5.080 (0.000) 5.219 (0.000)Getting along 0.996 (0.964)Resources 1.188 (0.035)Sadness, anxiety 1.070 (0.413)Pain, discomfort 1.241 (0.005)Not in labour forceSeeing 1.103 (0.240) 1.212 (0.001) 1.083 (0.348)Hearing 1.163 (0.286) 1.015 (0.837) 1.176 (0.254)Mobility 1.294 (0.002) 1.178 (0.001) 1.235 (0.013)Cognition 1.051 (0.595) 1.044 (0.481) 1.005 (0.960)Self care 1.274 (0.194) 1.057 (0.592) 1.227 (0.269)Communication 1.068 (0.623) 0.888 (0.092) 1.050 (0.723)Male 0.120 (0.000) 0.120 (0.000) 0.122 (0.000)Age 0.729 (0.000) 0.734 (0.000) 0.729 (0.000)Age squared 1.004 (0.000) 1.004 (0.000) 1.004 (0.000)Education 0.961 (0.000) 0.961 (0.000) 0.960 (0.000)Married, cohabiting 1.136 (0.268) 1.093 (0.435) 1.127 (0.301)Sep., div., widowed 0.968 (0.854) 0.936 (0.711) 0.977 (0.896)Indonesia 0.893 (0.401) 0.897 (0.421) 0.874 (0.328)Philippines 0.357 (0.000) 0.340 (0.000) 0.355 (0.000)Fiji 0.876 (0.330) 0.882 (0.354) 0.843 (0.219)Mongolia 0.593 (0.000) 0.605 (0.000) 0.580 (0.000)Getting along 0.951 (0.394)Resources 1.010 (0.852)Sadness, anxiety 1.114 (0.032)Pain, discomfort 1.020 (0.673)Observations 4384 4428 4338Pseudo R2 0.214 0.214 0.216AIC 6210.9 6268.5 6145.0BIC 6428.0 6486.0 6412.8


23

Table 9: Effect of replacing WG question with single WHO question on AIC

WHOquestion

Householdresponsi-bilities

Work,school

Communityactivities

Employmentstatus

Base model 6430 4421 4590 6145Seeing b1.6 6405 4391 4578 6135(replacing WG w1) b1.7 6405 4396 4561 6135Hearing b1.8 6430 4401 4583 6157(replacing WG w2) b1.9 6424 4403 4579 6159Mobility b1.13 6450 4422 4571 6142(replacing WG w3) d2.5 6265 4339 4565 6116Cognition d1.1 6366 4385 4531 6153(replacing WG w4) d1.2 6403 4399 4577 6154Self care d3.1 6374 4410 4550 6145(replacing WG w5) d3.2 6413 4417 4564 6138Communication d1.5 6431 4437 4575 6160(replacing WG w6) d1.6 6441 4442 4560 6163

labour force. The odds of Filipinos and Mongolians not being in the labour force are muchlower than for Indians. Finally, sadness and anxiety increases the odds of not being in thelabour force by 11%, but none of the other additional variables in specification 3 have asignificant impact on the odds of not being in the labour force.

Using combined WHO questions (specification 2) leads to better overall fits — in terms ofall three measures of fit reported in tables 5, 6 and 7 — of the models for household respon-sibilities, work and school, and joining community activities than using Washington Groupquestions (specification 1). For employment status (table 8, there is very little difference inmodel fit between the two specifications. By adding the additional covariates of specifica-tion 3, the fit of all models improves over both specification 1 and specification 2. Althoughnot shown in the tables, the fit of the models improves further by substituting the Washing-ton Group impairment variables by the corresponding WHO variables (constructed fromboth questions in each domain).

It is thus established that the combined WHO impairment variables perform better in ex-plaining difficulties with household responsibilities, work and school, and joining commu-nity activities than the Washington Group questions for the corresponding domains. Table9 asks, and answers, the question whether a single WHO impairment question performsbetter or worse than the Washington Group questions. It does so by measuring the effecton the AIC of substituting — one-by-one — the Washington Group questions in specifica-tion 3 of each model with a single WHO impairment question.

For the employment status model, the Washington Group questions on hearing, cognitionand communication generate better model fits than a single WHO question for the samedomains. The same applies to the communication domain with respect to the models for

24

difficulties with household responsibilities and work and school. In all other cases, a singleWHO questions generates a better fit than the corresponding Washington Group question.

5 Discussion and conclusions

The picture that emerges from the foregoing section is that the construct validity of theWashington Group questions for the seeing, hearing, mobility and self care domains is goodwhen WHO questions for the corresponding domains are used as a benchmark; this does,however, not apply to the questions for the cognition and communication domains.

The Washington Group and WHO questions for the seeing, hearing, mobility and self caredomains load high on four separate factors recovered from the factor analysis exercise. Forthe seeing, hearing and mobility domains, no other question has a loading higher than 0.4on any of these factors. Two of the ABS questions on self care, however and encouragingly,do also load high on the self care factor. The factor analysis fails, however, to generatefactors for the cognition and communication domains.

As the factor analysis exercise used the tau-b correlation matrix as input, it is not surprisingthat rank correlation patterns in general confirm these conclusions. In terms of both tau-band gamma, the association between the Washington Group and WHO questions is highestfor the seeing, mobility and self care domains; it is somewhat lower for the hearing domain.For the cognition and communication domains, the association between the two groups ofquestions is lowest for the cognition and communication domains (again, both in terms oftau-b and gamma).

The Washington Group questions perform similar to the corresponding (individual andcombined) WHO questions in terms of predictive validity. For the four models examined— explaining difficulty with household responsibilities, work and school, and joining com-munity activities, as well as employment status — the different question sets perform sim-ilar in terms of significance and magnitude of the odds ratios.

The question sets agree, for example, that hearing has an insignificant (at the 10% level)impact on difficulty with household responsibilities and joining community activities, butthat a higher level of difficulty in the other five functioning domains increases the odds ra-tios for those two dependent variables significantly (at the 5% level). They also agree thatall six impairment increase the odds significantly (at the 5% level, with the exception of theWashington Group hearing question, which is significant at the 10% level) that respondentsreport a higher level of difficulty with work and school. And they agree that — once again— hearing, and also cognition, difficulties have no significant impact on the odds of bothbeing unemployed and not being in the labour force (vs. being employed), but that mo-bility difficulties has, and that seeing difficulties significantly increases the odds of beingunemployment.

What clearly emerges from the foregoing paragraph is that hearing difficulties, in con-trast to the other functioning difficulties covered by Washington Group questions, have

25

no significant impact on the activity limitations and participation restrictions examined(although the evidence with regard to employment status is not clear-cut). The evidencewith regard to the possible inclusion of additional covariates is not uniform either. Hav-ing a higher level of difficulty in getting along with people, financial resources, pain anddiscomfort, and sadness and anxiety increases the odds of being limited in activities bothat and away from home; the first three covariates do the same for having trouble joiningcommunity activities. Difficulties with resources and pain and discomfort significantly in-crease the odds of being unemployed; the causality with regard to the former covariatemight, however, run in the opposite direction. The sadness and anxiety variable is the onlycovariate that is significant with respect to not being in the labour force.

6 Further study

This section will first examine how the analysis in this paper might be extended, and thendiscuss some ideas for a possible further pilot test.

The models developed in section 4.4 serve the primary purpose of comparing the predictivevalidity of different question sets (Washington Group vs. WHO), not to find ‘best’ fits. Forthe latter purpose, first and obviously, interactions between covariates might be includedin the models. Second, experiments might be conducted to find better scores for the pre-dictors. As mentioned in section 4.4, attempts to fit alternative scores, as well as treatingthe levels of the predictors as separate qualitative variables, were not successful. That doesnot mean, however, that scores leading to better fitting models do not exists; they merelyhave not yet been found.

The models have also have not been used for prediction and classification. There is lit-tle doubt that such attempts with the present models would be unsuccessful because ofthe ‘rare event’ character of disability; the prediction error for the ‘no’ and ‘none’ answerswould overwhelm predictions for the other responses. Finding models suitable for pre-diction and classification in the present context is, however, very important because theability to predict answers to questions accurately would make the necessity to ask themredundant.

A subsequent pilot test of disability question sets should take a few lessons learned fromthe present analysis on board. First, the employment status question should be asked inthe standard way. Second, different response options for the disability questions shouldbe tried. Asking people to assess level of difficulty in terms of {no, some, a lot, unable}and {none, mild, moderate, severe, extreme/cannot do} generates ordinary type variables,which are much more difficult to analyze to interval variables. The latter type of variablescould be generated by asking people to assess difficulty on a scale of, say, one to ten. Inthis context, it would be useful if all questions (i.e., the Washington Group and WHO sets)would have the same response options. The fact established in section 4.4 that model spec-ifications using WHO questions in general fit better than those using Washington Groupquestions is probably due to the fact that the former have an additional response option,

26

making it easier to detect covariance with the dependent variables. Finally, in order to beable to assess the ‘importance’ of the various disability domains, it is important to ask re-spondents one or more questions about there overall health status or quality of life.

27

References

ABS, 2003. Testing a Disability Question for the Census. Australian Bureau of Statistics,Canberra.URL http://www.abs.gov.au/websitedbs/c311215.nsf/22b99697d1e47ad8ca2568e30008e1bc/29ac3ed8564fe715ca256943002c4e3c/

$FILE/Testing%20a%20Disability%20Question%20for%20the%20Census_1.pdf

ABS, 2006. A Disability Question in the 2006 Census: Development, Testing, Analysis. Aus-tralian Bureau of Statistics, Canberra.

Agresti, A., 2002. Categorical Data Analysis, 2nd Edition. John Wiley & Sons, Inc., New York.

AIHW, 2004. Disability and its relationship to health conditions and other factors, aus-tralian Institute of Health and Welfare, Cat. No. DIS 37, Canberra.

Altman, B., Barnartt, S. (Eds.), 2006. International Views on Disability Measures: Movingtoward Comparative Measurement. Elsevier, Oxford.

Borooah, V. K., 2002. Logit and Probit: Ordered and Multinomial Models. Quantitative Ap-plications in the Social Sciences. Sage Publications, Inc., Thousand Oaks, CA.

Cieza, A., Geyh, S., Chatterji, S., Kostanjsek, N., Üstün, B. T., Stucki, G., 2006. Identificationof candidate categories of the international classification of functioning disability andhealth (ICF) for a generic icf core set based on regression modelling. BMC Medical Re-search Methodology 6 (36).URL http://www.biomedcentral.com/content/pdf/1471-2288-6-36.pdf

Greene, W. H., 2002. Econometric Analysis, 5th Edition. Prentice Hall, New Jersey.

Hamilton, L. C., 2003. Statistics with Stata 8. Duxbury Press, Pacific Grove, CA.

Harrell Jr, F. E., with contributions from many other users, 2006. Hmisc: Harrell Miscella-neous. R package version 3.1-2.URL http://biostat.mc.vanderbilt.edu/s/Hmisc,http://biostat.mc.vanderbilt.edu/twiki/pub/Main/RS/sintro.pdf,http://biostat.mc.

vanderbilt.edu/twiki/pub/Main/StatReport/summary.pdf

Hosmer, D. W., Lemeshow, S., 2000. Applied Logistic Regression, 2nd Edition. John Wiley &Sons Inc., New York.

McMenamin, T., Miller, S. M., Polivka, A. E., August 2006. Discussion and presentation ofthe disability test results from the current population survey, U.S. Bureau of Labor Statis-tics Working Paper 396.URL http:/www.bls.gov/ore/pdf/st050190.pdf

Menard, S. W., 2001. Applied Logistic Regression Analysis, 2nd Edition. Sage Publications,Inc., Thousand Oaks, CA.

28

http://www.abs.gov.au/websitedbs/c311215.nsf/22b99697d1e47ad8ca2568e30008e1bc/29ac3ed8564fe715ca256943002c4e3c/$FILE/Testing%20a%20Disability%20Question%20for%20the%20Census_1.pdfhttp://www.abs.gov.au/websitedbs/c311215.nsf/22b99697d1e47ad8ca2568e30008e1bc/29ac3ed8564fe715ca256943002c4e3c/$FILE/Testing%20a%20Disability%20Question%20for%20the%20Census_1.pdfhttp://www.abs.gov.au/websitedbs/c311215.nsf/22b99697d1e47ad8ca2568e30008e1bc/29ac3ed8564fe715ca256943002c4e3c/$FILE/Testing%20a%20Disability%20Question%20for%20the%20Census_1.pdfhttp://www.biomedcentral.com/content/pdf/1471-2288-6-36.pdfhttp://biostat.mc.vanderbilt.edu/s/Hmisc, http://biostat.mc.vanderbilt.edu/twiki/pub/Main/RS/sintro.pdf, http://biostat.mc.vanderbilt.edu/twiki/pub/Main/StatReport/summary.pdfhttp://biostat.mc.vanderbilt.edu/s/Hmisc, http://biostat.mc.vanderbilt.edu/twiki/pub/Main/RS/sintro.pdf, http://biostat.mc.vanderbilt.edu/twiki/pub/Main/StatReport/summary.pdfhttp://biostat.mc.vanderbilt.edu/s/Hmisc, http://biostat.mc.vanderbilt.edu/twiki/pub/Main/RS/sintro.pdf, http://biostat.mc.vanderbilt.edu/twiki/pub/Main/StatReport/summary.pdfhttp:/www.bls.gov/ore/pdf/st050190.pdf

Sarkar, D., 2006. lattice: Lattice Graphics. R package version 0.14-16.

Takamine, Y., 2003. Disability issues in East Asia: Review and ways forward, World BankWorking Paper.URL http://siteresources.worldbank.org/DISABILITY/Resources/Regions/East-Asia-Pacific/Disability_Issues_in_East_Asia_Takamine.pdf

United Nations, 2001. Guidelines and Principles for the Development of Disability Statis-tics. No. 10 in Statistics on Special Population Groups, Series Y. United Nations, New York.

United Nations, 2006. Principles and Recommendations for Population and Housing Cen-suses. United Nations, New York, rev. 2 (draft).

WHO, 2001. International Classification of Functioning, Disability and Health. WorldHealth Organization, Geneva.

WHO, ESCAP, 2006. Disability statistics training manual, Draft, Bangkok.

Wickham, H., 2005. reshape: Flexibly reshape data. R package version 0.7.1.

29

http://siteresources.worldbank.org/DISABILITY/Resources/Regions/East-Asia-Pacific/Disability_Issues_in_East_Asia_Takamine.pdfhttp://siteresources.worldbank.org/DISABILITY/Resources/Regions/East-Asia-Pacific/Disability_Issues_in_East_Asia_Takamine.pdf

Annex I: Contingency tables

Table A.1: Seeing: b1.6 vs. w1

b1.6 N None Mild Moderate Severe ExtremeN = 3685 N = 436 N = 237 N = 103 N = 34

w1 4478No 91% (3331) 46% ( 199) 38% ( 91) 27% ( 28) 21% ( 7)Some 8% ( 293) 48% ( 208) 41% ( 97) 25% ( 25) 15% ( 5)A lot 1% ( 38) 6% ( 25) 19% ( 45) 44% ( 45) 41% ( 14)Unable 0% ( 4) 0% ( 1) 2% ( 4) 4% ( 4) 24% ( 8)


Table A.2: Seeing: b1.7 vs. w1




Table A.3: Seeing: b1.6 and b1.7 vs. w1

b1.6 and b1.7 N None Mild Moderate Severe ExtremeN = 3438 N = 556 N = 306 N = 147 N = 48



30

Table A.4: Hearing: b1.8 vs. w2




Table A.5: Hearing: b1.9 vs. w2




Table A.6: Hearing: b1.8 and b1.9 vs. w2

b1.8 and b1.9 N None Mild Moderate Severe ExtremeN = 3803 N = 433 N = 174 N = 59 N = 26



31

Table A.7: Mobility: b1.13 vs. w3




Table A.8: Mobility: d2.5 vs. w3

d2.5 N None Mild Moderate Severe ExtremeN = 3227 N = 675 N = 299 N = 176 N = 103



Table A.9: Mobility: b1.13 and d2.5 vs. w3

b1.13 and d2.5 N None Mild Moderate Severe ExtremeN = 2590 N = 1041 N = 496 N = 254 N = 120



32

Table A.10: Cognition: d1.1 vs. w4




Table A.11: Cognition: d1.2 vs. w4




Table A.12: Cognition: d1.1 and d1.2 vs. w4

d1.1 and d1.2 N None Mild Moderate Severe ExtremeN = 3045 N = 947 N = 360 N = 117 N = 29



33

Table A.13: Self care: d3.1 vs. w5




Table A.14: Self care: d3.2 vs. w5




Table A.15: Self care: d3.1 and d3.2 vs. w5




34

Table A.16: Communication: d1.5 vs. w6




Table A.17: Communication: d1.6 vs. w6




Table A.18: Communication: d1.5 and d1.6 vs. w6




35

Annex II: Questionnaire

36

UNESCAP/WHO Project on Health and Disability Statistics

Disability Question Set Testing Study One: Specificity and sensitivity testing

Study Two: Test-Retest Reliability Study Three: Cognitive Interview

Questionnaire Version A

May 2005

2

SECTION 1. Face Sheet

ITEMS F1- F5 ARE TO BE COMPLETED BY INTERVIEWERS PRIOR TO STARTING EACH INTERVIEW F1

RESPONDENT I.D . #

Centre # - Subject # - Interview time point

F2

INTERVIEWER I.D. #

Centre # - Interviewer #

F3

INTERVIEW TIME POINT (1, 2, ETC.)

F4

a) INTERVIEW DATE b) STARTING TIME c) TIME INTERVIEW ENDED d) TOTAL DURATION

___ ___/___ ___/___ ___

month day year

___ ___/___ ___ hrs min ___ ___/___ ___ hrs min ___ ___/___ ___ hrs min

F5

LIVING SITUATION AT TIME OF INTERVIEW (CIRCLE ONLY ONE)

Independent in Community 1

Assisted Living 2

Hospitalized 3

F6

SAMPLE (CIRCLE ONLY ONE)

General population 1

Other (specify) 6

__________________________

3

SECTION 2. DEMOGRAPHIC AND BACKGROUND INFORMATION

PREAMBLE SAY TO RESPONDENT: This interview has been developed by the WHO / UNESCAP Project on Health and Disability Statistics to better understand the difficulties people may have due to their health conditions. The information that you provide in this interview is confidential and will be used only for research. FOR RESPONDENTS FROM THE GENERAL POPULATION SAY: Even if you are healthy and have no difficulties, it is necessary that I ask all of the questions for completeness. I will begin with some background questions. A1

RECORD SEX AS OBSERVED

Female 1 Male 2

A2

How old are you now? ___/___ years

A3

How many years in all did you spend studying in school, college or university?

___/___ years

A4

What is your current marital status? (SELECT THE SINGLE BEST OPTION)

Never married 1

Currently married 2

Separated 3

Divorced 4

Widowed 5

Cohabiting 6 A5

Which describes your main work status best? (SELECT THE SINGLE BEST OPTION)

Paid work 1

Self employed, such as own 2 your business or farming

Non paid work, such as 3 volunteer or charity

Student 4

Keeping house/Homemaker 5

Retired 6

Unemployed (health reasons) 7

Unemployed (other reasons) 8

Other (specify) 9

_________________________

4

SECTION 3. DISABILITY QUESTION SETS Question Set 1

The next questions ask about difficulties you may have doing certain activities because of a HEALTH PROBLEM

No Some A lot Unable W1 Do you have difficulty seeing, even if wearing glasses? 1 2 3 4

W2 Do you have difficulty hearing, even if using a hearing aid? 1 2 3 4

W3 Do you have difficulty walking or climbing steps? 1 2 3 4

W4

Do you have difficulty remembering or concentrating? 1 2 3 4

W5 Do you have difficulty (with self-care such as) washing all over or dressing? 1 2 3 4

W6 Because of a physical, mental, or emotional health condition, do you have difficulty communicating, (for example understanding or being understood by others)?

1 2 3 4

5

Question Set 2 Part 1: Introduction

SAY TO RESPONDENT: The interview is about difficulties people have because of health conditions. (HAND FLASHCARD #1 TO RESPONDENT). By health condition I mean diseases or illnesses, other health problems that may be short or long lasting, injuries, mental or emotional problems and problems with alcohol or drugs. I remind you to keep all of your health problems in mind as you answer the questions. When I ask you about difficulties in doing an activity think about (POINT TO FLASHCARD #1). • Increased effort • Discomfort or pain • Slowness • Changes in the way you do the activity (POINT TO FLASHCARD #1). When answering, I’d like you to think back over the last 30 days. I also would like you to answer these questions thinking about how much difficulty you have, on average over the past 30 days, while doing the activity as you usually do it. (HAND FLASHCARD #2 TO RESPONDENT). Use this scale when responding. (READ SCALE ALOUD): None, mild, moderate, severe, extreme or cannot do. (FLASHCARDS #1 AND #2 SHOULD REMAIN VISIBLE TO THE RESPONDENT THROUGHOUT THE INTERVIEW. )

6

Part 2: Questions on body functions

I am going to ask some questions about your body functions. Please remember that I am asking only about difficulties you experienced in the last 30 days that are due to health problems. POINT TO FLASHCARDS #1 AND #2

None Mild Moderate Severe Extreme /Cannot

Do

B1.1 How much of bodily aches or pains did you have? 1 2 3 4 5

B1.2 How much bodily discomfort did you have? 1 2 3 4 5

B1.3 Have you had a problem with a skin defect of face, body, arms or legs? 1 2 3 4 5

B1.4 Have you had a problem with your appearance due to missing or deformed or paralyzed arms, legs, feet? 1 2 3 4 5

B1.5 How much difficulty did you have in using your hands and fingers, such as picking up small objects or opening or closing containers?

1 2 3 4 5

B1.6

How much difficulty did you have in seeing and recognizing a person you know across the road? (take into account eye glasses, if you wear them) Read the brackets if you see respondent wearing glasses.

1 2 3 4 5

B1.7

How much difficulty did you have in seeing and recognizing an object at arm’s length or in reading? (take into account eye glasses, if you wear them) Read the brackets if you see respondent wearing glasses.

1 2 3 4 5

B1.8

How much difficulty did you have in hearing someone talking on the other side of the room in a normal voice? (take into account hearing aids, if you use them) Read the brackets if you see respondent using hearing aid..

1 2 3 4 5

B1.9

How much difficulty did you have in hearing what is said in a conversation with one other person in a quiet room? (take into account hearing aids, if you use them) Read the brackets if you see respondent using hearing aid..

1 2 3 4 5

B1.10 How much of a problem did you have passing water (urinating) or in controlling urine (incontinence)? 1 2 3 4 5

B1.11 How much of a problem did you have with defecating, including constipation? 1 2 3 4 5

B1.12

How much difficulty did you have with shortness of breath at rest? 1 2 3 4 5

B1.13 How much difficulty did you have with shortness of breath with mild exercise, such as climbing uphill for 20 meters or stairs (such as 12 steps)?

1 2 3 4 5

B1.14 How much difficulty did you have with coughing or wheezing for ten minutes or more at a time? 1 2 3 4 5

B1.15 How much of the time did you have a problem with sleeping, such as: falling asleep, waking up frequently during the night or waking up too early in the morning?

1 2 3 4 5

B1.16 How much of a problem did you have with feeling sad, low or depressed? 1 2 3 4 5

B1.17 How much of a problem did you have with worry or anxiety? 1 2 3 4 5

7

Part 3: Questions on Activities and Participation

DOMAIN 1 Understanding and Communicating

I am going to ask some questions about understanding and communicating. Please remember that I am asking only about difficulties you experienced in the last 30 days that are due to health problems. POINT TO FLASHCARDS #1 AND #2

None Mild Moderate Severe

Extreme /Cannot

Do

D1.1 How much difficulty did you have in concentrating on doing something for ten minutes? 1 2 3 4 5

D1.2 How much difficulty did you have in remembering to do important things? 1 2 3 4 5

D1.3 How much difficulty did you have in analysing and finding solutions to problems in day to day life? 1 2 3 4 5

D1.4 How much difficulty did you have in learning a new task, for example, learning how to get to a new place? 1 2 3 4 5

D1.5 How much difficulty did you have in generally understanding what people say? 1 2 3 4 5

D1.6 How much difficulty did you have in starting and maintaining a conversation? 1 2 3 4 5

DOMAIN 2 Getting Around

I am now going to ask you about difficulties in getting around. Please remember that I am asking only about difficulties you experienced in the last 30 days that are due to health problems. POINT TO FLASHCARDS #1 AND #2


Do

D2.1 How much difficulty did you have in standing for long periods such as 30 minutes? 1 2 3 4 5

D2.2 How much difficulty did you have in standing up from sitting down? 1 2 3 4 5

D2.3 How much difficulty did you have in moving around inside your home? 1 2 3 4 5

D2.4 How much difficulty did you have in getting out of your home? 1 2 3 4 5

D2.5 How much difficulty did you have in walking a long distance such as a kilometre [or equivalent]? 1 2 3 4 5

8

DOMAIN 3 Self Care

I am now going to ask you about difficulties in taking care of yourself. Please remember that I am asking only about difficulties you experienced in the last 30 days that are due to health problems. POINT TO FLASHCARDS #1 AND #2


Do

D3.1 How much difficulty did you have in washing your whole body? 1 2 3 4 5

D3.2 How much difficulty did you have in getting dressed? 1 2 3 4 5

D3.3 How much difficulty did you have in eating? 1 2 3 4 5

D3.4 How much difficulty did you have in staying by yourself for a few days? 1 2 3 4 5

DOMAIN 4 Getting along with people

I am now going to ask you about difficulties in getting along with people. Please remember that I am asking only about difficulties you experienced in the last 30 days that are due to health problems. POINT TO FLASHCARDS #1 AND #2


Do

D4.1 How much difficulty did you have in dealing with people you do not know? 1 2 3 4 5

D4.2 How much difficulty did you have in maintaining a friendship? 1 2 3 4 5

D4.3 How much difficulty did you have in getting along with people who are close to

comparing disability questions for censuses and surveys in asia...

Documents