A physical symptom-based measure of quality of care for children affected by HIV/AIDS

Abstract

Based on cross-sectional data of 1,625 children aged 6–18 years who were affected by parental HIV/AIDS in central rural China, we evaluated a physical symptom based measure of quality of care (QOC_PS) for these children. The QOC_PS has acceptable reliability estimates for children of both genders, at different age, and by orphanhood status. Known-group validation and construct validity analysis demonstrated a good validity of QOC_PS indicator. The QOC_PS score was significantly associated with psychosocial well-being measures among children. The physical symptom based measure provides a psychometrically appropriate indicator of quality of care for children affected by HIV/AIDS in China. It could be used as an alternative measure to assess quality of care in resource-poor settings where other objective measurements are not available or feasible. Future research is needed to further validate the scale among children in different living environments across various cultural settings.

Introduction

Globally, an estimated 36.7 million people were living with HIV, and one million people died of AIDS in 2016 (UNAIDS, 2017). As a result of the AIDS epidemic, 13.4 million children had lost one or two parents to AIDS (i.e., AIDS orphans) by 2015, with many more children living with HIV-positive parents or caregivers (i.e., vulnerable children) (UNICEF, 2016). Compared with other orphans, children affected by parental HIV/AIDS (AIDS orphans or vulnerable children) could be more disadvantaged due to HIV-related stigma and discrimination and many other challenges associated with parental HIV/AIDS (UNICEF, 2012; J. Zhao et al., 2010). One of such challenges is the lack of adequate care (Li et al., 2008). In resource-limited settings, particularly in settings with high HIV prevalence among the adult population, the burden of care for AIDS orphans is largely falling on the very young and the very old (Drew, Makufa, & Foster, 1998; Safman, 2004). The caregivers are usually experiencing financial hardship, bereavement, high rates of morbidity and mortality, and migration which may inevitably compromise the stability and quality of their caregiving (Govender, Penning, George, & Quinlan, 2012; Heymann, Earle, Rajaraman, Miller, & Bogen, 2007; Li et al., 2008; Safman, 2004). Parents living with HIV/AIDS who struggle with problems of their own physical and psychosocial health may have difficulties in providing their children with adequate care (Lv et al., 2010).

Quality of care (QOC) has profound consequences for children in terms of their adjustment and long-term well-being (Bryant & Ramey, 1986). Lower quality care threatens children’s cognitive, language, and social adjustment (Bornstein, 2012). Previous studies have mostly focused on the effect of various care arrangements (“type of care”) on the children’s outcomes (Hong et al., 2011; G. Zhao et al., 2010), but studies on QOC were limited (Kuo & Operatio, 2009). A few studies on QOC for children affected by HIV/AIDS relied on either qualitative interviews or self-reports by children or caregivers (Q. Zhao et al., 2009) and these measures might be subject to potential biases such as recall errors or socially desirable reporting.

Physical symptoms have been used as alternative estimates of QOC in a variety of health care and child care settings. For example, physical symptoms (e.g., falling and fractures, urinary tract infections, weight loss, pain, occurrence of pressure ulcers, use of antipsychotics medicines among residents without severe mental illness, and any hospitalization within the past 180 days) were used by the U.S. Center for Medicare and Medicaid Services (CMS) and other agencies as quality indicators for nursing homes (Cefalu, 2011). Global literature suggests that adequate care might be associated with fewer respiratory illnesses and other infectious diseases among children (Niffenegger, 1997). Research has demonstrated that children can reliably report their physical symptoms as young as 6 years of age (Riley, 2004). One study based on 8591 children 5 to 16 years of age suggested that children as young as 5 years old could reliably and validly report their health-related quality of life given the opportunity to do so with an age-appropriate instrument (Varni, Limbers, & Burwinkle, 2007). However, such approaches have not been used or validated among children affected by parental HIV/AIDS in resource-poor settings including rural China.

To address this knowledge gap, the current study aims to evaluate a physical symptom based measure of QOC, utilizing a large sample of children affected by parental HIV/AIDS (both AIDS orphans and vulnerable children) from rural China. We anticipated that a self-reported symptoms-based QOC measure (QOC_PS) may have appropriate psychometric properties to assess the quality of care among these children in rural China. The psychometric properties assessed in this study included reliability and construct validity. The construct validity of the scale was assessed using both known-group validation (DeVellis, 2003) and associations with other measures of psychosocial functioning. We hypothesized that the children affected by AIDS (i.e., AIDS orphans or vulnerable children) will report poorer QOC_PS than their peers from the same community who did not experience HIV-related illness and death in their families (i.e., comparison children). We also hypothesized that QOC_PS measures will demonstrate a good construct validity by showing a positive association with psychosocial well-being (e.g., self-esteem) and negative associations with psychopathological symptoms (e.g., PTSD) among these children.

Method

Sample

Data used in the current study were derived from the baseline survey of a longitudinal study conducted in 2006–2009 examining psychological distress and adjustment in children from two rural counties with the highest prevalence of HIV infection in central China (X. Li et al., 2009). The study sites (with ≥94% of their population being rural residents) suffered a rapid spread of HIV infection mainly through unhygienic blood collection from the late 1980s to the middle 1990s (J. Zhao et al., 2010). Poor farmers who sold their blood and/or blood plasma were infected with HIV through contaminated centrifuges and non-disposable needles (Wang, 1993). A total of 1,625 children 6 to 18 years of age participated in the baseline survey including 755 AIDS orphans, 466 vulnerable children, and 404 comparison children who were from the same communities but did not have HIV-related illness or death in their families. With the support from the anti-epidemic stations of each county we identified the villages with the highest number of HIV-related deaths or confirmed HIV infection. Participating children were recruited from five administrative villages (rural administrative units under the county) with approximately 250,000 residents in their jurisdictions. The recruitment procedure has been described elsewhere (X. Li et al., 2009). Briefly, we recruited children from four government-funded AIDS orphanages, eight community-based group homes, and various family or kinship care settings. With the collaboration of village leaders, we developed lists of families that were either caring for orphans or having parents with confirmed diagnosis of HIV/AIDS. We then visited families and children on the lists and recruited one child per family until we reached the target sample size (about 1,200) for the AIDS orphans and vulnerable children. The recruitment of comparison group followed the similar process. Comparison children (with a target sample size of 400) were recruited from the same villages where the AIDS orphans and vulnerable children were recruited. Children with known HIV infection were not included in the study, with HIV-infection status being confirmed by current caregivers or local community workers. The research protocol, including consenting procedure, was approved by the Institutional Review Boards at both Wayne State University in the United States and Beijing Normal University in China. All children provided appropriate consent/assent before their participation.

Survey procedure

Participating children were asked to complete a confidential survey of demographic characteristics, living condition, and psychosocial outcomes. The questionnaires were self-administered for older children (e.g., ≥9 years of age) and were administered verbally by interviewers for younger children (e.g., ≤8 years of age) or children with low literacy. Necessary clarification or instruction was provided when needed during the survey. The interviewers were faculty and graduate students from a local university and received extensive training prior to data collection in areas of research ethics and skills in working with disadvantaged children and conducting quantitative survey. The time of completing the survey varied by children but generally ranged from 75 to 90 minutes. Younger children were instructed to take a 10–15 minutes break after every 30 minutes during the survey. All the children received an age-appropriate gift worth ~20 Chinese Yuan (~3 US dollars) after finishing the survey as a token of appreciation (X. Li et al., 2009).

Most of the psychological measures used in the survey have been adapted and validated in China (Wang, 1993). Some psychosocial scales were translated into Chinese and pilot-tested prior to the field data collection. Following the pilot-testing, several minor modifications were made to improve the comprehensibility of survey questions for younger children (e.g., those<10 years of age) (Fang et al., 2009).

Measures

Physical Symptom based measure of Quality of Care (QOC_PS)

The initial item pool for QOC_PS measure was generated based on global literature on child care (Niffenegger, 1997) and input from pediatricians at a large US-based children’s hospital who have extensive experiences working with children in resource-limited settings including China. The appropriateness of items in measuring quality of care was evaluated and discussed among the pediatricians and the researchers. We finally retained 7 items representing common illnesses or discomforts (i.e., runny nose, fever, persistent cough, ear pain, purulent in ears, diarrhea, and vomiting) among children in China. The prolonged duration or multiple concurrences of these symptoms in children were considered by the pediatricians as being associated with deficient or suboptimal care of children. These items were pilot-tested (along with other scales in our assessment inventory) prior to the data collection. The pilot-test sample included children of both gender aged from 7 to 16 years. The results of pilot-test suggested an adequate understanding of these physical symptoms among Chinese children. During the survey, children were asked to report whether they had any of the 7 physical symptoms in the past month (Yes/No). With appropriate coding, a sum score of the 7 items was used as a measure of QOC, with a higher score reflecting a better QOC.

Individual and Family Characteristics

Children were asked about individual and family demographic information including gender, age, ethnicity, parental educational level (i.e., no school, elementary school, middle school, high school and above) and parental occupation (i.e., farmer, migrant workers, others) in which their parents were engaged currently or before their death. We created a family socioeconomic status (SES) score using parental education and occupation for both father and mother. Specifically, parental education was recoded as 1 for more than elementary school education and 0 otherwise; and parental occupation was recoded as 1 for non-farming occupation or migrant workers and 0 for farming occupation. The SES score thus ranged from 0 to 4 with a higher score suggesting a better family SES. For the purpose of data analysis in the current study, children were divided into three age groups based on their developmental stages: pre-adolescents (younger than 12 years of age), early adolescents (12 through 14 years of age), and middle adolescents (older than 14 years of age).

Other Measurements

Children were also asked to report their perceived health status (poor, fair, good, very good), and satisfaction with the current living conditions (very unsatisfied, unsatisfied, neutral, satisfied, very satisfied). Psychosocial well-being of these children was assessed with several standard and validated psychosocial scales, including depression, loneliness, post-traumatic stress (PTSD), and self-esteem. A brief description of these psychosocial scales is presented in Table 1.

Table 1.

Summary of Psychological Scales

Psychological scale	#items	α	Sample item	Response options
Depression Center for Epidemiological Studies Depression Scale for Children (CES-DC)(Fendrich, Weissman, & Warner, 1990)	20	.81	I was bothered by things that usually don’t bother me.	4-point response option (“not at all” to “a lot”)
Loneliness Children’s Loneliness Scale (CLS)(Asher, Hymel, & Renshaw, 1984; Wang, 1993)	16	.82	It’s easy for me to make new friends at school.	5-point response option (“not true at all” to “always true”)
PTSDa
Post-Traumatic Stress (Briere, 1996; Xiaoming Li et al., 2009)	10	.81	I have nightmares.	4-point response option (“never” to “almost all of time”)
Self-esteem Self-Esteem Scale (Qiao, Li, Zhao, Zhao, & Stanton, 2014)	10	.64	I feel that I have a number of good qualities.	4-point response option (“strongly disagree” to “strongly agree”)
Enacted Stigmaa (G. Zhao, Li, Zhao, Zhang, & Stanton, 2012)	14	.88	Relatives stopped visiting when parents got sick or died.	5-point response option (“never happened” to “always happened”)

Note:

^a.Data were available from only children affected by HIV/AIDS (AIDS orphans and vulnerable children).

Statistical Analysis

Descriptive analysis was used to display the basic demographic characteristics of the participants and their family. Internal consistency of the QOC_PS measure was estimated using Cronbach alpha coefficients for the entire sample, as well as subgroups of children by gender, age and orphanhood status. We conducted inter-item correlations and exploratory factor analysis to examine the dimensionality of the items (Tabachnick & Fidell, 2001). The QOC_PS scores were compared for group differences by gender, age and family SES. To assess construct validity of the QOC_PS measure, we employed the known-group validation procedure (Varni et al., 2007) and examined the associations of QOC_PS score with other available psychosocial scales.

For known-group validation, we conducted analysis of variance (ANOVA) to compare QOC_PS scores by children orphanhood status (i.e., AIDS orphans, vulnerable children, and comparison children), as the existing literature has suggested a better QOC among comparison children than children affected by HIV/AIDS. Similarly, we compared QOC_PS scores by perceived health status (i.e., poor, fair, good, very good) and living condition (i.e., very unsatisfied, unsatisfied, neutral, satisfied, very satisfied). We anticipated that children who perceived a better health status or reported more satisfaction with current living condition would score higher on QOC_PS indicator.

We employed Pearson Correlation Coefficients (Pearson r) to assess the associations between QOC_PS score and psychosocial scales. We anticipated that QOC_PS would be positively associated with psychosocial welling-being (e.g., self-esteem) and negatively associated with psychopathology symptoms (e.g., PTSD). The associations between QOC_PS and psychosocial scales were further examined in multivariate analysis using General Linear Model (GLM) adjusting for key demographic variables including gender, age, orphanhood status, and family SES. We categorized QOC_PS into three groups using the quartiles of the scores as cut-offs: low (bottom 25%), medium (middle 50%), and high (top 25%). The categorized QOC_PS scores, gender and orphanhood status were used as between-subject factors in the GLM analyses; and age and family SES (as continuous variables) were included in the model as covariates. Because two of the psychosocial measures (enacted stigma and PTSD) were only available from children affected by HIV/AIDS (AIDS orphans and vulnerable children), two GLM analyses were conducted with one for children affected by HIV/AIDS with all 5 psychosocial measures (Model 1) and the other for entire sample by removing enacted stigma and PTSD (Model 2). All analyses were conducted using SPSS for Windows V.22.

Results

Sample characteristics

Nearly half of the participants were girls (n=799, 49%). The average age was 12.9 years (SD=2.21) and did not significantly differ between boys and girls. Over 70% children reported that their parents (either father or mother) had no more than middle school education attainment. In addition, 88% mothers and 84% fathers worked as farmers or migrant workers in urban areas. The average family SES score was 2.10 (SD=1.06) in a range of 0 to 4. Two thirds of the participants reported “good” or “very good” health status. About 36% children expressed that they were “satisfied” or “very satisfied” with their current living condition.

Reliability Estimates of QOC_PS Items

Generally, the QOC_PS measure had adequate internal consistency with a Cronbach alpha of .67 for the entire sample. Internal consistency estimates were similar between boys and girls (Cronbach alpha=.65, and .68, respectively), among pre-, early, and middle adolescents (Cronbach alpha=.70, .67, and .61, respectively), and among AIDS orphans, vulnerable children, and comparison children (Cronbach alpha=.66, .73, and .62, respectively).

Dimensionality of QOC_PS Items

Table 2 illustrated that the inter-items correlation coefficients ranged from .04 (between item 1 and item 5) to .45 (between item 6 and item 7). All inter-items correlation coefficients were statistically significant (p<.001) except the one between items 1 and 5 (correlation coefficient=.04, p=.221). The Anti-image correlation matrix demonstrated low correlations across items (all<.25), and the Kaiser-Meyer-Olkin (KMO) measure was .75, suggesting no violations of assumptions to conduct exploratory factor analysis. Using principal axis factoring approach, we identified two initial eigenvalues over 1.00 (2.37 and 1.17) that explained 51% of the total variance (33.87 and 16.65%, respectively). The scree plot of eigenvalues (not shown), however, suggested a single factor solution.

Table 2.

Inter-item correlation of QOC_PS

	Item 1	Item 2	Item 3	Item 4	Item 5	Item 6	Item 7
Item 1	1
Item 2	.38	1
Item 3	.18	.35	1
Item 4	.17	.24	.25	1
Item 5	.04	.06	.10	.31	1
Item 6	.33	.40	.37	.22	.08	1
Item 7	.25	.32	.33	.24	.11	.45	1

Note: All correlation coefficients are significant at p<.001, except the one between Item 1 and Item 5 (p=.221).

Item 1: Did you have runny nose in the past month? Yes/No

Item 2: Did you have a fever in the past month? Yes/No

Item 3: Did you have a persistent cough in the past month? Yes/No

Item 4: Did you have ear pain in the past month? Yes/No

Item 5: Did you have purulent in ears in the past month? Yes/No

Item 6: Did you have diarrhea in the past month? Yes/No

Item 7: Did you have vomiting in the past month? Yes/No

Gender, Age and Family SES Difference of QOC_PS

The QOC_PS measure had a mean score of 3.85 (SD=2.35) for the entire sample. The QOC_PS score was not significantly different between boys (3.90, SD=2.32) and girls (3.81, SD=2.38). Early-adolescents showed a lower QOC_PS score (3.78, SD=2.41) than pre-adolescents (3.81, SD=2.34) or middle adolescents (4.03, SD=2.28), but the overall difference was not statistically significant (F[2,1531]=1.408, p=.245). Pair-wise comparisons suggested that early-adolescents reported a significantly lower QOC_PS compared to middle adolescents (p<.05). Children with the highest SES score (i.e., score =4) reported the lowest QOC_PS score (3.79, SD=2.44), but children with medium SES score (score=2 or 3) reported higher QOC_PS score (3.89, SD=2.33) than children (3.80, SD=2.37) with low SES score (score=0 or 1). However, neither the overall differences across family SES nor the pair-wise comparisons reached statistical significance.

Known-Group Validation

As we anticipated, a significant group difference in QOC_PS was detected by children orphanhood status (F[2,1542]= 15.94, p< .0001). Both AIDS orphans (3.97, SD =2 .34) and vulnerable children (3.34, SD = 2.52) had a lower score than comparison children (4.20, SD = 2.17). Pair-wise comparisons indicated that vulnerable children had a significantly lower QOC_PS compared to both AIDS orphans (p<.001) and comparison children (p<.001). But the difference between AIDS orphans and comparison did not reach statistical significance (p=.087). In addition, ANOVA results suggested that QOC_PS scores were significantly higher among children with better perceived health status (F[3, 1478]=19.68, p<.0001) or with higher satisfaction with current living condition (F[4, 1518]=9.36, p<.0001) than their counterparts who perceived worse health status or satisfaction with their living condition.

Correlations with Psychosocial Measures

We examined the correlations between QOC_PS score and psychosocial scales among children by orphanhood status. Table 3 presented the correlation coefficients across the three groups of children. The QOC_PS was negatively associated with enacted stigma among children affected by HIV/AIDS (r=−.14 for AIDS orphans and r=−.19 for vulnerable children, p<.0001 for both coefficients) and PTSD (r=−.18 for AIDS orphans, p<.0001; and r=−.13 for vulnerable children, p<.05). The QOC_PS score was also negatively associated with depression and loneliness across the three groups of children, although the correlation coefficients between QOC_PS and depression among vulnerable children (r=−.08) did not reach statistical significance. The QOC_PS was positively associated with self-esteem among AIDS orphans (r=.09, p<.05) and vulnerable children (r=.11, p<.05), but its association with self-esteem was not statistically significant among comparison children (r=.09, p=.05).

Table 3.

Correlation coefficients among main study variables

	1	2	3	4	5	6
AIDS Orphans
1. QOC_PS	1
2.Depression	−.168***	1
3.Loneliness	−.104**	.339***	1
4.Enacted stigma	−.143***	.523***	.332***	1
5.PTSD	−.175***	.595***	.222***	.506***	1
6.Self-esteem	.091*	−.176***	−.471***	−.131***	−.124**	1
Vulnerable children
1. QOC_PS	1
2.Depression	−.082	1
3.Loneliness	−.106*	.327***	1
4.Enacted stigma	−.187***	.562***	.312***	1
5.PTSD	−.125*	.628***	.214***	.515***	1
6.Self-esteem	.105*	−.172***	−.491***	−.255***	−.202***	1
Comparison children
1. QOC_PS	1
2.Depression	−.169**	1
3.Loneliness	−.116*	.504***	1
4.Enacted stigma	n/a	n/a	n/a	1
5.PTSD	n/a	n/a	n/a	n/a	1
6.Self-esteem	.091	−.273***	−.357***	n/a	n/a	1

Note:

^*p<.05,

^**p<.01,

^***p<.001.

Multivariate Analysis

The results of GLM analyses were summarized in Table 4. As anticipated, the categorized QOC_PS (low, medium, and high), after adjusting for appropriate covariates, remained as a significant between-subject factor in both multivariate and univariate tests for all psychosocial scales for children affected by HIV/AIDS (Model 1). Gender was not significant in the multivariate test or in any of the univariate tests except the loneliness (p=.02). Likewise, family SES was not a significant covariate in the multivariate test or any of univariate tests. Age was a significant covariate in the multivariate test and all univariate tests for psychosocial variables except depression (p=.94) and PTSD (p=.10). Orphanhood status (AIDS orphans, and vulnerable children) was significant in both multivariate and all univariate tests except the loneliness (p=.17). The two-way and three-way interactions among the factor variables (i.e., QOC_PS, gender, and orphanhood status) were also examined. Only the interaction between QOC_PS and orphanhood status was significant for multivariate test and the univariate tests for depression, enacted stigma, and PTSD.

Table 4.

Results of GLM Analysis

	Main Effect			Interaction	Covariates
	QOC_PS (Q)	Gender	Orphanhood Status (S)	Q × Sa	SES	Age
Model 1 (without comparison children)
Multivariateb	6.65c	2.051	9.199c	3.272	.668	20.531c
Depression	14.961c	.333	26.658c	3.870e	.073	.006
Loneliness	10.374c	5.454e	1.880	2.781	1.131	31.892c
Enacted-stigma	18.367c	.826	16.997c	6.904c	.462	57.269c
PTSD	13.565c	.982	13.171c	8.374c	.433	2.713
Self-esteem	10.599c	.552	5.499c	.084	2.777	10.480c
Model 2 (with comparison children)
Multivariateb	9.58c	.97	18.37c	1.39	1.27	15.36c
Depression	23.21c	.08	28.73c	2.03	.17	.02
Loneliness	11.52c	2.67	26.76c	1.20	2.14	37.10c
Self-esteem	8.99c	.16	7.97c	1.21	3.21	12.86c

Notes:

^a.Other possible interactions terms were statistically significant for neither multivariate tests nor univariate tests and were not presented in the table.

^b.Pillai’s Trace F statistics were presented in the table for multivariate test and conventional F statistics (based on Type III Sum of Square) were presented for univariate tests.

^c.p <0.001.

^d.p <0.01.

^e.p <0.05.

Similar results were observed among entire sample with three available measures of psychosocial outcomes (Model 2). Both QOC_PS and orphanhood status (AIDS orphans, vulnerable children, and comparison children) demonstrated a multivariate significance and univariate significance for all three psychosocial measures. Age was a significant covariate for multivariate test and all univariate tests except the one for depression. The two-way and three-way interactions among the factor variables were significant for neither multivariate test nor any of the univariate tests.

Discussion

The current study aimed to evaluate the psychometric property of a brief, physical symptom based QOC measure for use with children in resource-poor settings. Our findings demonstrated both adequate reliability and validity of this measure in a sample of children affected by parental HIV/AIDS in rural China. The QOC_PS is reliable for children of both genders and at different age groups. The known-group validation analysis confirmed our anticipations that children with better perceived health status or better satisfaction with living conditions reported higher QOC_PS score, and children affected by HIV/AIDS reported lower QOC_PS than comparison children. These results suggested that the QOC_PS measure is largely sensitive to the condition of care, especially the condition of care that can be possibly influenced by parental HIV/AIDS. The associations between QOC_PS and psychosocial scales provided preliminary evidence for the validity of the QOC_PS measure. Moreover, the associations between QOC_PS and psychosocial scales remained significant after adjusting for possible individual and contextual confounders such as orphanhood status, gender, age and family SES.

Our data demonstrated the association between the QOC (estimated by QOC_PS) and psychosocial well-being among children. The improvement of QOC needs to be integrated into mental health services for families or communities affected by HIV/AIDS or in other disadvantaged situations, providing them with training and counselling about parenting and caring. In addition, it is notable that vulnerable children reported lower QOC_PS scores than AIDS orphans. This finding further confirmed the relevance of QOC_PS measure with actual QOC as the existing literature has suggested that children living with HIV-positive parents might be in a particularly higher risk for inadequate care compared with AIDS orphans who were mostly under care in government-sponsored group settings (Chi & Li, 2013). Future interventions should pay more attentions to this subgroup of children living with HIV-infected parents who may have difficulties in providing adequate care to children due to challenges with their own disease management and limited financial or other support from government and local communities.

While our results suggested utility of the QOC_PS as an alternative measure of QOC among children affected by parental HIV, the findings need to be interpreted with cautions as physical symptoms can be influenced by a wider array of factors, beyond just the quality of care being received, such as genetic characteristics, general wellbeing, mental health and co-morbidities of the children. The use of QOC_PS needs to take into consideration of other factors of children, family, and their living environment. For example, the current analysis revealed some differences of QOC_PS by age, gender, and family SES, although the differences by gender and family SES did not reach statistical significance. Future studies are also needed to further validate the scale among children in different living environments across various cultural settings.

There are several limitations in the current study. First, although existing literature suggested that young children were able to provide self-reported health information, the data collection was still subject to errors in recall and potential social desirability bias. Particularly, social desirability bias might have occurred among younger children (e.g., ≤8 years of age) or children with low literacy as they completed the survey via face-to-face interviews. Although other studies and our own pilot-testing suggested that younger children were able to adequately report the parental education attainment and parental occupation, the self-reported parental demographic characteristics by younger children can be inaccurate. Second, we were not able to assess the validity of the QOC_PS using other objective measures of QOC as criteria. Future research should seek further validation of QOC_PS with objective measures of QOC when they are available. Third, the reliability estimates for QOC_PS measure are relatively low (e.g., Cronbach alphas <.70). However, this may be acceptable given QOC_PS measure is a checklist of multiple physical symptoms rather than a psychological measure. Fourth, we were unable to examine some key factors (besides age, gender, and family SES) that may contribute to the association between quality of care and physical symptoms. For example, the stage of parental illness could have been associated with children’s physical symptom scores but the data regarding the stage of HIV disease progress among parents were not available in the current study. Fifth, due to constraints of the data, we were not able to assess other reliability estimates such as test-retest reliability for the QOC_PS measure in this sample. Finally, the sample was recruited from rural counties with prevalent HIV infections through unhygienic blood collection in central China. The findings therefore may not be generalizable to children affected by HIV/AIDS in other regions or countries and also among children in other disadvantaged situations.

Conclusion

Despite these limitations, this study provided preliminary evidence on the psychometric properties of QOC_PS measure for children affected by HIV/AIDS in China. Based on common physical symptoms among children, the 7-item measure is short, relatively objective, easy to understand, and free of some possible bias (e.g., the items are not linked to any sensitive terms such as HIV/AIDS, or caregivers). Although it is very important to employ objective measure to assess QOC for children affected by HIV/AIDS, a physical symptoms based proxy measure can be used as an alternative estimate of QOC in resource-poor settings where other objective measures are unavailable or unfeasible. Given the accumulating evidence suggesting that deficient care may have a lasting and passive impact on well-being of children, we believe that this brief QOC_PS measure will be useful in this important area. Meanwhile, we need to further validate QOC_PS in various care settings and explore potential cofounders between physical symptoms and QOC in order to objectively and robustly assess QOC for disadvantaged children, especially those in resource-poor settings.