Differences in Proxy-Reported and Self-Reported Disability in the Demographic and Health Surveys

Abstract

This article examines the presence, direction, and magnitude of differences between proxy reports and self-reports of disability among adults aged fifteen to fifty-nine years and elderly individuals aged sixty years and older in Demographic and Health Surveys conducted in Uganda, South Africa, and Mali. We use the propensity score weighted multivariate logistic regression to balance the weighted distributions of the covariates between self-reports and proxy reports. Disabilities that have an immediate effect on the communication with others or that require one-to-one help are likely to be over-reported by proxies or under-reported if proxies are not used, especially among the elderly aged sixty years and older. Disabilities that are not observable might be under-reported by proxies.

1. BACKGROUND

The Demographic and Health Surveys (DHS) are the main source for national and sub-national estimates of demographic and health indicators in many countries, including the prevalence of disabilities. The DHS Program disability module includes a series of questions based on the Washington Group on Disability Statistics Short Set (WG-SS) of questions that address six core functional domains: seeing, hearing, communicating, remembering or concentrating, walking or climbing steps, and washing all over or dressing. As an extension for the Child Functioning Module developed by UNICEF and the Washington Group, the WG-SS was developed to measure functional disabilities among general population. These modules focus on the functional difficulties rather than impairments, diagnoses, or conditions, which would require medical assessments to be identified (Cappa, Mont, Loeb, Misunas, Madans, et al. 2018; Loeb, Mont, Cappa, De Palma, Madans, et al. 2018; Massey 2018).

In DHS surveys that include the disability module, questions on disability are added to the household questionnaire about all household members and visitors aged five years and older. Similar to others in the household questionnaire, the disability questions are to be answered by the respondent of the household questionnaire on behalf of all reference individuals. Answers to the disability questions are classified into two groups: (1) self-reports: the answers of the respondent about him/herself and (2) proxy reports: the answers of the respondent about other household members or visitors. Some proxy reports might be classified as proxy-assisted reports when the respondent was assisted by the reference individual who was present during the interview. Since the respondent of each household questionnaire must be at least age fifteen years, all answers to disability questions for children aged five to fourteen are based on proxy reports from a parent or other family member.

Although this interviewing strategy assumes an alignment between the information provided by the respondent and the condition of the reference individual, it is more efficient to collect the disability data from the respondent of the household questionnaire. This might be true for questions about background characteristics such as gender, age, education, and marital status, but it is not necessarily true for questions about disability. Differences between self-reports and proxy reports of disability have been examined in previous studies. Systematic differences stemmed from factors such as the age of the reference individual, the nature of the disability, and the relationship between the reference individual and the respondent (Todorov and Kirchner 2000; Ellis, Bannister, Cox, Fowler, Shannon, et al. 2003; Shields 2004; Roydhouse, Gutman, Keating, Mor, and Wilson 2018; Lapin, Thompson, Schuster, and Katzan 2019). There was some evidence that proxies under-report disabilities for adults aged eighteen to sixty-four years, but over-report disabilities in the elderly (Epstein, Hall, Tognetti, Son, and Conant 1989; Rothman, Hedrick, Bulcroft, Hickam, and Rubenstein 1991; Todorov and Kirchner 2000; Li, Harris, and Lu 2015). There was also some evidence that proxies under-report symptoms and minor disabilities but not obvious and chronic problems that require assistance from others (Tennant, Badley, and Sullivan 1991).

Two types of errors are associated with the discrepancies between proxy reports and self-reports: (1) the proxy effect, which is the under-reporting of events from lack of knowledge, and (2) the saliency principle, or the effect of proxy respondents reporting only the most salient features (Tennant et al. 1991; Rajmil, Fernández, Gispert, Rué, Glutting, et al. 1999). Such errors might lead to systematic differences between self-reports and proxy reports that can affect the national estimates of disabilities. Unfortunately, it is not common to find reliable benchmarks of disability prevalence to assess the potential bias that arises from using the proxy reports. However, self-reports are justified as the standard for evaluating proxy reports of disabilities among adults who are capable of providing information. Among adults who are not, proxy reports are acceptable as a standard for the evaluation (Beckett, Weinstein, Goldman, and Yu-Hsuan 2000; Edwards, Anderson, and Deokar 2015; Li et al. 2015).

Differences in disability prevalence based on proxy reports and self-reports might be attributed to the varied sociodemographic characteristics of the inference population. For example, individuals with disabilities who have immediate effect on communication with others are less likely to be respondents for the household questionnaire, while individuals with functional disabilities who are able to communicate with others are more likely to be respondents because they stay at home most of the time. The differences between proxy reports and self-reports might also be attributed to “true” differences in the reports (Epstein et al. 1989; Rothman et al. 1991; Todorov and Kirchner 2000; Li et al. 2015).

Since the reporting status, either proxy reports or self-reports, is not typically randomly assigned, assessments must be done with caution. When traditional regression models are used to examine the differences between proxy reports and self-reports, equal weights are given to all cases, including those with almost no chance of membership in the comparison group. Several studies used propensity score weighting (PSW) to balance the weighted distributions of the sociodemographic characteristics of reference population between proxy reports and self-reports and to assure that both groups are similar (Hirano and Imbens 2001; Ellis et al. 2003; Elliott, Beckett, Chong, Hambarsoomians, and Hays 2008).

In the DHS surveys, responses from proxies constitute 40–60 percent of all responses to disability questions among adults aged fifteen years or older. The differences between proxy-reported and self-reported disabilities in the DHS survey have never been examined. In this article, we examined the presence, direction, and magnitude of differences between proxy reports and self-reports on disability among adults aged fifteen years and older in three DHS surveys. To detect true differences between the self-reports and proxy reports, we control for variations in the sociodemographic characteristics of the inference population using PSW multivariate logistic regression. We use the DHS surveys conducted in Uganda, South Africa, and Mali. The differences are assessed for the two age cohorts separately, adults aged fifteen to fifty-nine years and the elderly aged sixty years and older.

2. METHODS

2.1 Data

We used data from DHS surveys conducted in Uganda (DHS 2016; 19,588 households and 46,247 adults aged fifteen years and older), South Africa (DHS 2016; 11,083 households and 26,372 adults aged fifteen years and older), and Mali (DHS 2018; 9,510 households and 27,009 adults aged fifteen years and older). In these surveys, the disability questions are administered to the respondent of the household questionnaire, who is any adult member of the household capable of providing the required information. In all surveys, respondents provided information for all household members and visitors aged five years and older (including him/herself) about whether they had no difficulty, some difficulty, a lot of difficulty, or did not have the ability at all in each of the following domains: seeing, hearing, communicating, remembering or concentrating, walking or climbing steps, and washing all over or dressing. In all disability questions, the word “disability” is avoided, and all questions are phrased around “difficulties” instead. (See table 1 for the question wording in the three surveys.) In addition to the disability questions, information was collected on sociodemographic characteristics of all household members such as gender, age, education, and marital status.

Table 1.

Question Wording for Disability Module in DHS Surveys

Domain	South Africa DHS	Mali DHSa	Uganda DHS
Seeing	Does (NAME) have difficulty seeing, even if wearing glasses? IF NO, CIRCLE “0.” IF YES, PROBE: With some difficulty, with a lot of difficulty, or cannot see at all?	Does (NAME) wear glasses or contact lenses to help them see? Yes/No I would like to know if (NAME) has difficulty seeing even when wearing glasses or contact lenses. Would you say that (NAME) has no difficulty seeing, some difficulty, a lot of difficulty, or cannot see at all?	Does (NAME) wear glasses or contact lenses to help them see? Yes/No I would like to know if (NAME) has difficulty seeing even when wearing glasses or contact lenses. Would you say that (NAME) has no difficulty seeing, some difficulty, a lot of difficulty, or cannot see at all?
Hearing	Does (NAME) have difficulty hearing, even if wearing a hearing aid? IF NO, CIRCLE “0.” IF YES, PROBE: With some difficulty, with a lot of difficulty, or cannot hear at all?	Does (NAME) wear a hearing aid? Yes/No I would like to know if (NAME) has difficulty hearing even when using a hearing aid? Would you say that (NAME) has no difficulty hearing, some difficulty, a lot of difficulty, or cannot hear at all?	Does (NAME) wear a hearing aid? Yes/No I would like to know if (NAME) has difficulty hearing even when using a hearing aid? Would you say that (NAME) has no difficulty hearing, some difficulty, a lot of difficulty, or cannot hear at all?
Communicating	Does (NAME) have difficulty communicating in (his/her) usual language? For example, understanding others or others understanding (him/her)? IF NO, CIRCLE “0.” IF YES, PROBE: With some difficulty, with a lot of difficulty, or cannot communicate at all?	I would like to know if (NAME) has difficulty communicating when using his/her usual language. Would you say that (NAME) has no difficulty understanding or being understood, some difficulty, a lot of difficulty, or cannot communicate at all?	I would like to know if (NAME) has difficulty communicating when using his/her usual language. Would you say that (NAME) has no difficulty understanding or being understood, some difficulty, a lot of difficulty, or cannot communicate at all?
Remembering or concentrating	Does (NAME) have difficulty remembering or concentrating? IF NO, CIRCLE “0.” IF YES, PROBE: With some difficulty, with a lot of difficulty, or cannot remember or concentrate at all?	I would like to know if (NAME) has difficulty remembering or concentrating. Would you say that (NAME) has no difficulty remembering or concentrating, some difficulty, a lot of difficulty, or cannot remember or concentrate at all?	I would like to know if (NAME) has difficulty remembering or concentrating. Would you say that (NAME) has no difficulty remembering or concentrating, some difficulty, a lot of difficulty, or cannot remember or concentrate at all?
Walking or climbing steps	Does (NAME) have difficulty walking a kilometer or climbing a flight of steps? IF NO, CIRCLE “0.” IF YES, PROBE: With some difficulty, with a lot of difficulty, or cannot walk or climb steps at all?	I would like to know if (NAME) has difficulty walking or climbing steps. Would you say that (NAME) has no difficulty walking or climbing steps, some difficulty, a lot of difficulty or cannot not walk or climb steps at all?	I would like to know if (NAME) has difficulty walking or climbing steps. Would you say that (NAME) has no difficulty walking or climbing steps, some difficulty, a lot of difficulty, or cannot walk or climb steps at all?
Washing all over or dressing	Does (NAME) have difficulty with self-care such as washing all over or dressing? IF NO, CIRCLE “0.” IF YES, PROBE: With some difficulty, with a lot of difficulty, or cannot do at all?	I would like to know if (NAME) has difficulty washing all over or dressing. Would you say that (NAME) has no difficulty washing all over or dressing, some difficulty, a lot of difficulty or cannot wash or dress at all?	I would like to know if (NAME) has difficulty washing all over or dressing. Would you say that (NAME) has no difficulty washing all over or dressing, some difficulty, a lot of difficulty, or cannot wash all over or dress at all?

a

English translation for the French questionnaire used in the Mali DHS.

Samples of the three surveys utilized two-stage sampling designs, in which census enumeration areas (EAs) were selected in the first stage; fixed numbers of households were selected in the second stage (Uganda DHS 2016: 697 EAs and 20,910 households; South Africa DHS 2016: 750 EAs and 15,000 households; Mali DHS 2018: 379 EAs and 10,574 households). Survey weights were calculated to account for selection probabilities in different selection stages, and for nonresponse in different levels. In this study, weighted analysis was produced using the household weight (HV005) that accounts for the survey base weight and the household level nonresponse. Outcomes from the Uganda DHS are presented in section 3, and tables for results from the South Africa DHS and Mali DHS are shown in the Appendix. Since children aged five to fourteen years were not among the target population of this article, they were excluded from the analysis.

2.2 Measures

We analyzed six disability domains reported in six variables (seeing, hearing, communicating, remembering or concentrating, walking or climbing steps, and washing all over or dressing). In each variable, difficulties were reported in four categories (no difficulties, some difficulties, a lot of difficulties, cannot function at all). The disability prevalence estimates were calculated as the proportions of individuals who have a lot of difficulty or cannot function at all.

Five variables were considered for the PSW adjustment and regression models: gender (male, female), age (fifteen to twenty-nine, thirty–forty-four, forty-five to fifty-nine, sixty to seventy-four, seventy-five to eighty-nine, ninety+ years), education (no education, primary, secondary or higher), marital status (never married, married, widowed/divorced), and proxy status (self-reported, proxy-reported).

2.3 Analytic Approach

Data analysis was done separately for two age groups: adults aged fifteen to fifty-nine and elderly adults aged sixty years or older. Two propensity score models were estimated as multivariate logistic regressions, one for each age group. Each model estimated the probability of proxy reports, p(α), as opposed to self-reports, (1 − p(α)). To adjust the survey weight, HV005, it was multiplied by weights of 1/p(α) for proxy reports and 1/(1 − p(α)) for self-reports. The adjusted survey weight was used in multivariate logistic models (PSW-adjusted models) to examine the association between the disability domains and the proxy status. In building models to estimate the adjustment propensities or the PSW-adjusted models, weighted main effects were estimated taking sampling strata and sampling clusters into account. In all models, the dependent variable was the reporter (0: self-reports, 1: proxy reports); in addition to the main effect of the six disability variables, the main effects of four sociodemographic characteristics were retained in the models: gender, age, education, and marital status. Before and after PSW adjustments, we calculated the odds ratio (OR) and 95 percent confidence interval (CI) for each disability domain. The data analysis used the PROC PSMATCH of SAS/STAT version 9.4 (SAS Institute, Cary, NC, USA) for the PSW and the svy: logit of Stata-16 (StataCorp, College Station, TX) for multivariate logistic regressions.

3. RESULTS

In Uganda DHS, the data on the disability questions are reported for 46,247 adults. Among adults aged fifteen to fifty-nine years, there are 17,060 self-reports and 24,665 proxy reports, and among the elderly aged sixty years and older, 2,522 self-reports and 2,000 proxy reports. As shown in table 2, regardless of the age cohort, the sociodemographic characteristics of the reference population differ significantly between those who self-reported and those who were proxy-reported. Among adults aged fifteen to fifty-nine years, survey-reference individuals who are proxy-reported were more likely to be male, younger than age thirty years, and single. Among elderly individuals aged sixty years or older, the proxy-reported are more likely to be male, married, and not educated.

Table 2.

Characteristics of Self-Reports and Proxy Reports among Adults and the Elderly, Uganda DHS 2016

	Adults aged fifteen to fifty-nine years			Elderly aged sixty years or older
	Self-report (%) (n = 17,060)	Proxy report (%) (n = 24,665)	p-Value	Self-report (%) (n = 2,522) (%)	Proxy report (%) (n = 2,000) (%)	p-Value
Gender			<.00001			.0005
Male	38.46	53.64		41.54	48.20
Female	61.54	46.36		58.46	51.80
Age			<.00001			<.00001
15–29/60–74	41.39	62.64		74.30	70.32
30–44/75–89	38.55	24.62		23.36	23.71
45–59/90+	20.06	12.74		2.34	5.97
Education			<.00001			.0002
No education	11.73	8.79		42.65	48.66
Primary	53.61	55.88		45.97	39.15
Secondary or higher	34.67	35.33		11.39	12.19
Marital status			<.00001			<.00001
Never married	12.52	43.09		1.13	1.79
Married	70.05	49.70		44.27	60.12
Widowed/divorced	17.43	7.21		54.60	38.08

Proxy-reported disabilities also differed significantly from self-reported disabilities. The direction and magnitude of differences between two types of responses varied by disability type. As indicated in table 3, proxies are more likely to report difficulties in hearing, communicating, and washing all over or dressing among adults and the elderly. The over-reporting was significant in all comparisons, except for washing all over or dressing among adults (difference: 0.06, 95 percent CI: −0.07 to 0.2). Proxies are less likely to report difficulties in seeing, remembering or concentrating, and walking or climbing steps. The under-reporting was significant in all comparisons, except for remembering or concentrating (difference: −0.57, 95 percent CI: −2.58 to 1.45) and walking or climbing steps among the elderly (difference: −1.66, 95 percent CI: −4.58 to 1.26).

Table 3.

Disability Prevalence Comparison between Self-Reports and Proxy Reports among Adults and the Elderly, Uganda DHS 2016

Difficulties	Adults aged fifteen to fifty-nine years			Elderly people aged sixty years or older
	Self-report (%) (n = 17,060)	Proxy report (%) (n = 24,665)	Difference (proxy–self) (95 % CI)	Self-report (%) (n = 17,060)	Proxy report (%) (n = 24,665)	Difference (proxy–self) (95 % CI)
Seeing	2.76	1.21	−1.54 (−1.87, −1.22)***	23.53	19.52	−4.01 (−6.71, −1.31)**
Hearing	0.49	0.79	0.3 (0.14, 0.47)***	5.76	7.85	2.09 (0.43, 3.75)*
Communicating	0.06	0.60	0.54 (0.43, 0.64)***	0.18	1.11	0.93 (0.42, 1.44)***
Remembering or concentrating	2.11	1.65	−0.46 (−0.75, −0.17)**	11.53	10.96	−0.57 (−2.58, 1.45)
Walking or climbing steps	2.00	1.37	−0.63 (−0.92, −0.34)***	21.62	19.96	−1.66 (−4.58, 1.26)
Washing all over or dressing	0.33	0.39	0.06 (−0.07, 0.2)	5.39	8.13	2.74 (0.64, 4.83)*

*

p-Value < .05.

**

p-Value < .01.

***

p-Value < .0001.

To control for potential selection bias, we used PSW regression to adjust the survey weights for adults and the elderly separately. As shown in table 2 and figure 1, there is some imbalance between the groups for adults aged fifteen to fifty-nine years, where the standardized mean differences (the mean difference divided by an estimate of its standard deviation) are greater than 0.25 for females, singles, and adults aged fifteen to twenty-nine years and less than −0.25 for married, widowed, and adults aged thirty to forty-four years. Similarly, for the elderly aged sixty years or older, there is some imbalance between the groups, especially for married and widowed elderly individuals, where the standardized mean differences are greater than 0.25 for married and less than −0.25 for widowed elderly individuals. These differences were balanced by the PSW. The balanced standardized mean differences are around the zero. As indicated in table 4, after using the adjusted survey weight to compare the characteristics of individuals who are self-reported and those who are proxy-reported, almost no significant differences can be found between the two samples.

Table 4.

Adjusted Characteristics of Self-Reports and Proxy Reports among Adults and the Elderly, Uganda DHS 2016

	Adults aged fifteen to fifty-nine years			Elderly aged sixty years or older
	Self-report (%) (n = 17,060)	Proxy report (%) (n = 24,665)	p-Value	Self-report (%) (n = 2,522)	Proxy report (%) (n = 2,000)	p-Value
Gender			.3313			.9478
Male	49.02	48.17		44.74	44.61
Female	50.98	51.83		55.26	55.39
Age			.7981			.5715
15–29/60–74	53.97	53.60		73.11	71.81
30–44/75–89	30.30	30.68		23.39	24.05
45–59/90+	15.73	15.72		3.50	4.15
Education			.0007			.6772
No education	9.69	9.89		45.03	45.80
Primary	51.86	54.19		43.39	42.00
Secondary or higher	38.54	35.92		11.58	12.19
Marital status			.6990			.7552
Never married	30.87	30.30		1.32	1.35
Married	57.78	58.20		52.26	50.90
Widowed/divorced	11.35	11.51		46.42	47.75

Figure 1.

Standardized Mean Differences by Sociodemographic Characteristics between Self-Reports and Proxy Reports among Adults Aged Fifteen to Fifty-Nine Years and the Elderly Aged Sixty Years and Older, Uganda DHS 2016.

As shown in table 5, among adults aged fifteen to fifty-nine years, after controlling for the sociodemographic characteristics (gender, age, education, and marital status), the proxies under-reporting of remembering or concentrating (OR: 0.92, 95 percent CI: 0.76–1.10) and walking or climbing steps (OR: 0.87, 95 percent CI: 0.73–1.04) among adults became insignificant. The over-reporting of difficulties in hearing and communicating and the under-reporting of difficulties in seeing remain significant even after PSW regression adjustment. Among adults, there are more difficulties reported by proxies in hearing (OR: 1.98, 95 percent CI: 1.50–2.63) and communicating (OR: 7.95, 95 percent CI: 3.88–16.28) and more self-reported difficulties in seeing (OR: 0.55, 95 percent CI: 0.46–0.66).

Table 5.

Unadjusted and Adjusted Comparisons between Self-Reports and Proxy Reports in Disability Domains among Adults Aged Fifteen to Fifty-Nine Years and the Elderly Aged Sixty Years and Older, UDHS 2016

Difficulties	Adults aged fifteen to fifty-nine years		Elderly aged sixty years and older
	Odds ratio (95 % CI)		Odds ratio (95 % CI)
	Unadjusted	PSW regression	Unadjusted	PSW regression
Seeing	0.43* (0.37, 0.51)	0.55* (0.46, 0.66)	0.79* (0.67, 0.93)	0.76* (0.64, 0.91)
Hearing	1.63* (1.24, 2.14)	1.98* (1.50, 2.63)	1.3* (1.07, 1.81)	1.40* (1.07, 1.85)
Communicating	9.58* (5.14, 7.87)	7.95* (3.88,16.28)	6.31* (2.18, 18.25)	7.44* (2.59, 21.35)
Remembering or concentrating	0.78* (0.67, 0.91)	0.92 (0.76, 1.10)	0.94 (0.77, 1.16)	0.98 (0.79, 1.21)
Walking or climbing steps	0.68* (0.58, 0.8)	0.87 (0.73, 1.04)	0.9 (0.76, 1.08)	0.95 (0.78, 1.15)
Washing all over or dressing	1.18 (0.81, 1.73)	1.34 (0.86, 2.07)	1.55* (1.13, 2.14)	1.58* (1.15, 2.17)

*

p-Value < .05.

Among the elderly, the control for the sociodemographic characteristics did not change the direction or the test significance of the comparisons. There are more difficulties reported by proxies in hearing (OR: 1.40, 95 percent CI: 1.07–1.85), communicating (OR: 7.44, 95 percent CI: 2.59–21.35), and washing all over or dressing (OR: 1.58, 95 percent CI: 1.15–2.17), and more self-reported difficulties in seeing (OR: 0.76, 95 percent CI: 0.64–0.91). Differences in remembering or concentrating and walking or climbing steps remain insignificant.

Significant differences were less common in South Africa and Mali. (Results are presented in tables A.1 and A.2.) Difficulties in hearing and communicating were under-reported among self-reported adults and the elderly in both countries. Similar to Uganda, proxies over-report difficulties in washing all over or dressing among the elderly in both countries.

4. DISCUSSION

These findings provide evidence that there might be some significant differences between proxy reports and self-reports of disabilities in the DHS surveys. This might reflect real differences in the reporting, rather than differences in the inference population’s sociodemographic characteristics. The findings showed that differences between self-reports and proxy reports were not random but were related systematically to the nature of the disability and the age of the reference individual. Similarly, the direction of these differences was related to the nature of disability. Unlike other studies, this study did not find variations in the direction of differences between the adults and the elderly groups. This was consistent across the two age cohorts. In general, there is no strong indication for a strong bias in the proxy reports; even when the differences are significant, those between disability prevalence based on proxy reports and self-reports are very small.

Disabilities that have immediate effect on communication with others are likely to be under-reported if proxies are not used. For example, disabilities such as hearing and communicating were under-reported when proxies are not used, with the largest discrepancy found in communicating. For these disabilities, it is necessary to use proxy reports to avoid selection bias in disability estimates based on self-reports. However, such under-reporting has not been noticed across all countries or age cohorts. Among the elderly, disabilities that require one-to-one help, such as washing all over or dressing, are likely to be under-reported if proxies are not used, or over-reported by proxies. The lack of a benchmark for the comparison makes it hard to decide the direction of the bias in these types of disabilities. For difficulties in remembering or concentrating and walking or climbing steps among adults, there was insignificant under-reporting by proxies. Although proxies under-reported difficulties in seeing among adults and the elderly, this under-reporting has not been identified in all countries.

There are some limitations of our study. Since the selected DHS surveys were not designed specifically for the aims of our study, we could not examine the proxy–self pair discrepancies, where proxy reports and self-reports are compared for the same reference population on the individual level. We also could not explore the relationship between proxy and reference individuals as a factor in the discrepancies because such information is not included in the DHS surveys. This factor might explain some of the patterns we noticed in the analysis outcomes; previous studies showed that immediate family such as spouses, parents, and children tend to provide more accurate information than other relatives or unrelated individuals (Mclaughlin, Dietz, Mehl, and Blot 1987; Gilpin, Pierce, Cavin, Berry, Evans, et al. 1994). Also, it was not possible to identify the proxy-assisted reports. With such information, differences between self-reports, proxy-assisted reports, and proxy reports could have been examined. Finally, it was not possible to assess the role of stigma in the over- or under-reporting of disability among self-reports and proxy reports.

AVAILABILITY

The Uganda DHS 2016 datasets can be downloaded from the data page (https://dhsprogram.com/data/dataset/Uganda_Standard-DHS_2016.cfm? flag=0; last accessed October 15, 2020).

The South Africa DHS 2016 datasets can be downloaded from the data page (https://dhsprogram.com/data/dataset/South-Africa_Standard-DHS_2016.cfm? flag=0; last accessed October 15, 2020).

The Mali DHS 2018 datasets can be downloaded from the data page (https://dhsprogram.com/data/dataset/Mali_Standard-DHS_2018.cfm?flag=0; last accessed October 15, 2020).

Acknowledgement

The author would like to thank Tom Pullum, Anjushree Pradhan, and Julia Fleuret for helpful discussions that assisted in writing this article. This research is carried out with support provided by the United States Agency for International Development (USAID) through The DHS Program (#720-OAA-18C-00083). The views expressed are those of the authors and do not necessarily reflect the views of USAID or the United States Government.

APPENDIX

 

Table A.1.

Unadjusted and Adjusted Comparisons between Self-Reports and Proxy Reports in Disability Domains among Adults Aged fifteen to fifty-nine years, South Africa DHS 2016 and Mali DHS 2018

Difficulties	South Africa DHS 2016		Mali DHS 2018
	Odds ratio (95 % CI)		Odds ratio (95 % CI)
	Unadjusted	PSW regression	Unadjusted	PSW regression
Seeing	0.85	1.17	0.56*	0.82
	(0.68, 1.07)	(0.91, 1.5)	(0.41, 0.77)	(0.57, 1.2)
Hearing	1.91*	2.07*	1.06	1.7*
	(1.26, 2.9)	7.00 (1.3, 3.3)	(0.68, 1.66)	(1.06, 2.71)
Communicating	4.38*	4.35*	4.33*	6.71*
	(2.15, 8.92)	(2.02, 9.38)	(1.85, 10.15)	(2.76, 16.29)
Remembering or concentrating	1.6*	1.84*	1.13	1.32
	(1.18, 2.15)	(1.35, 2.52)	(0.6, 2.12)	(0.71, 2.44)
Walking or climbing steps	0.77*	1.01	0.57*	0.8
	(0.59, 0.99)	(0.77, 1.32)	(0.39, 0.82)	(0.52, 1.24)
Washing all over or dressing	2.48*	2.78*	0.77	1.08
	(1.4, 4.42)	(1.45, 5.33)	(0.37, 1.6)	(0.5, 2.34)

*

p-Value < .05.

Table A.2.

Unadjusted and Adjusted Comparisons between Self-Reports and Proxy Reports in Disability Domains among the Elderly Aged Sixty Years and Older, South Africa DHS 2016 and Mali DHS 2018

Difficulties	South Africa DHS 2016		Mali DHS 2018
	Odds ratio (95% CI)		Odds ratio (95% CI)
	Unadjusted	PSW regression	Unadjusted	PSW regression
Seeing	1.11	1.1	1.1	1.07
	(0.87, 1.43)	(0.83, 1.45)	(0.89, 1.37)	(0.85, 1.34)
Hearing	2.26*	1.97*	1.51*	1.46
	(1.58, 3.23)	(1.41, 2.74)	(1.02, 2.24)	(0.98, 2.19)
Communicating	2.91*	2.54*	2.45*	1.59
	(1.4, 6.06)	(1.2, 5.41)	(1.09, 5.52)	(0.64, 3.97)
Remembering or concentrating	1.25	1.32	0.82	0.69
	(0.94, 1.66)	(0.98, 1.77)	(0.47, 1.44)	(0.39, 1.22)
Walking or climbing steps	1.16	1.18	1.39*	1.25
	(0.94, 1.44)	(0.95, 1.47)	(1.05, 1.84)	(0.91, 1.72)
Washing all over or dressing	3.36*	2.68*	3.61*	2.41*
	(2.07, 5.45)	(1.63, 4.41)	(1.82, 7.13)	(1.08, 5.39)

*

p-Value < .05.