Design and Setting

An institution-based cross-sectional study was done among parents accompanying a child diagnosed with ASD to three specialty clinics, one intermediary facility, and one basic facility in Trivandrum district in Kerala state, India. Kerala state is known for good health indicators including child survival and very good access to biomedical care. But Kerala does not have the same saturation of multidisciplinary care that is characteristic of high-income countries.12

Participants

The accompanying parent of a child up to 18 years of age with ASD visiting each institution during the study period (June 2018 to August 2018) was approached for participation in the study. Parents of children with a recent (within 1 month) diagnosis of ASD were excluded. The study could recruit 99 mothers but only 11 fathers. As the current paper focused on mothers of children diagnosed with ASD, we excluded fathers from further analysis.

Data Collection

A peer reviewed semi-structured interview schedule was used to collect data. Prospective parents were first informed about the study by the primary therapist at the institution they were visiting and referred to the investigator if they were willing to participate. Study description, consenting, and interviews were conducted by the same investigator for the sake of uniformity.

Variables

The interview schedule comprised of questions on socio-demographic characteristics including age, socioeconomic status based on state allotted categorization as below-poverty line or not, basic aspects of the diagnosis and treatment of the child, and potential factors associated with stress and coping strategies including sources of support parents received for various responsibilities. We incorporated 10 items based on the Cohen’s and Williamson’s Perceived Stress Scale (PSS-10),13 and 16 items covering four domains of emotion-based coping strategies, namely positive reinterpretation, acceptance, religiosity, and emotional support received.14 We prioritized on the emotion-based coping strategies in this study as we were focusing on the emotional experiences of the parents rather than the actions taken.

Analysis

Data entry, cleaning, and quantitative analysis were done in IBM Statistical Package for the Social Sciences (SPSS) version 25. Spousal support was assessed based on the reported role of the spouse and level of engagement in the care of autistic child and accompanying the child for therapeutic sessions. Spouses were considered supportive if mothers mentioned them as having equal or more responsibility in any of the aforementioned roles. Scale reliability of the stress and coping scales were assessed using Cronbach’s α. As raw scales had very low α values, underlying factors related to stress and coping were assessed to derive stress and coping scores. Principal component analysis with varimax rotation was used to reduce the number of variables and get a combination of factors with better internal consistency. The remaining items were added up to get summative scores. A receiver operator characteristic (ROC) curve was plotted to look at the ability of the computed stress score to identify mothers who need screening for anxiety disorder. Mothers were categorized as needing screening for anxiety disorder if they reported avoiding social gatherings “all the time” and labeled their anxiety level as “much higher.” With respect to coping, only the factors that were extracted with eigenvalue over one were considered for developing summative domain scores. These domain scales were categorized as low or high based on the median score for the respective domain.

Summative stress scores were categorized into three based on terciles to obtain the outcome variable—low, intermediate, or high stress level. Chi-square tests (or Fisher exact test) were used to compare the proportions of mothers across different stress categories for each independent variable. Those variables that had a p-value less than 0.01 were carried forward to ordinal logistic regression analysis with a constrained cumulative logit model (proportional odds model). Variables significant in one-to-one regression analysis with the outcome variable were incorporated to a multiple ordinal regression model. Assumptions of proportional odds (Chi-square test for model fit), goodness of fit, and parallelism (Chi-square test for parallel lines) were checked and odds ratios with 95% confidence intervals were estimated. Only the main effects were computed as the sample numbers were inadequate to explore possible interactions with regression analysis. However, we report stratified proportions to suggest possible interactions between spousal support, stress levels, and other variables.

Ethical Considerations

Consenting processes were in line with the relevant regulations of the Indian Council for Medical Research.15 Participants were assured privacy, confidentiality, and noneffect on availed services for the child/ parent whether they chose to participate or not. The study protocol received clearance from the Institutional Ethics Committee (IEC) of host institution (the Sree Chitra Tirunal Institute for Medical Sciences and Technology, Trivandrum.). The IEC is registered under the relevant legislation in India.16

Stress Levels

The scale based on the raw items of the PSS Score had a Cronbach’s α of 0.323. The item on anger (item 9) had a very high ceiling effect with almost one-third of the mothers choosing the highest possible response. The factor structure that emerged in dimension reduction analysis comprised of four items (items 1, 2, 3, and 10). These items, shown in the upper factor structure in Fig. 1, had good covariance. Sampling adequacy criteria for items and correlation matrix were met. Extracted communalities were 0.5 or higher and one component with eigenvalue 2.1 was extracted. The four items together had a Cronbach’s α of 0.69. When added together, they gave a variable ranging with values from one to 16, with a mean of 10.0 (standard deviation or SD 3.4). Based on tercile values of nine and 12, mothers were divided into three groups—31 with low stress, 34 with medium stress, and 34 with high stress.

Close

Fig. 1 Items contributing to computation of composite stress score and score for acceptance level; results of principal component analysis and ROC curve analysis. ROC, receiver operator characteristic.

Export to PPT

Fig. 1 Items contributing to computation of composite stress score and score for acceptance level; results of principal component analysis and ROC curve analysis. ROC, receiver operator characteristic.

Coping

Only two factors could be extracted from the four potential domains of coping assessed. The first one had an eigenvalue of 2.6 and corresponded to the domain religiosity and all four items were retained. The second one had an eigenvalue of 2.1 and corresponded to the domain acceptance. Only three items, as shown in lower factor structure in Fig. 1, corresponding to acceptance levels were retained. Two additive scores with the upper tercile 13.6 for religiosity and 11 for acceptance were obtained. Two new variables were derived to indicate a high level of religiosity and coping based on the mothers who scored more than the upper tercile value.

Spousal Support

Paternal support was predominant in accompanying the child for therapeutic intervention sessions as compared with care of the affected child at home, with 44 fathers playing an equal or larger role than the mothers in accompanying the child to treatment sessions while only 12 fathers took up the role of the main caregiver of the child at home. Only nine fathers had a dominant role in both aspects of care of the child diagnosed with ASD by being equally or more responsible than the mother for care of the child at home as well as accompanying the child for therapeutic intervention sessions. Three mothers reported absolutely no support of this form—one was divorced and one separated while the third was still living within her marriage.

Increasing Stress Levels and Spousal Support

Mothers were likely to fall into a higher category of stress if spousal support was low (p = 0.05), as shown in Table 2. Additionally, they were likely to have moderate or high stress if they were from poor households, had a child who was completely dependent on them due to the severity of symptoms (p < 0.05), had a child diagnosed with mixed disabilities (p = 0.08), had difficulty in interacting with other persons around her or had possible anxiety disorder (p = 0.06). Mothers who were deemed as having a high level of acceptance were less likely to be in the moderate or the high stress group (p = 0.05).

The results of ordinal logistic regression are shown in Table 3. The first model (Model 1) included all variables significant in bivariate analysis, but violated several assumptions of ordinal regression. There were many cells with zero values. The goodness-of-fit Chi-square values suggest inadequacy of the sample size. The second model (Model 2) met these assumptions, but we could adjust the effect of spousal support only for two other variables—level of dependence of the child and level of acceptance by the mother. In both the models, low spousal support was independently associated with mothers falling in the moderate or high stress category as opposed to the low stress category. The risk was almost three times (adjusted odds ratio [AOR] 2.80; [95% confidence intervals or CI 1.28–6.11]). Having a completely dependent child probably continued to have a very strong gradient of risk, and low acceptance remained significant too.

Possible two-way and three-way interactions were exp-lored with the variables in model 2. However, no significant interactions were obtained. Only the model with no interactions is reported. However, some interaction results suggested possibility to generate some hypotheses. For instance, level of dependence and spousal support had a p-value of 0.07. We explored such possibilities by stratifying the mothers into those with low spousal support and with high spousal support.

Table 4 depicts the comparison of risk factors for high stress stratified based on the level of spousal support. When spousal support was high, the stress levels in mothers seem to be determined by severity of the child’s condition. Mothers with higher degrees of stress were likely to have a fully dependent child or a child diagnosed with a mixed disorder. If spousal support was low, then there was less variation in maternal stress levels with respect to the child’s condition. In such mothers, low levels of stress were associated with her level of acceptance and ability to interact socially. Additionally, in the absence of spousal support, mothers belonging to a nonpoor socioeconomic group were more likely to fall under the low stress level category than the mothers from poor households.

Limitations

Our sample selection was institutional as there is no available registry of autism in Kerala at present. The sample size may have been inadequate to capture interactions, or the measurement may not have permitted enough variance to explore interactions. Also, there is a possibility of selection bias in this study, as we selected mothers recruited who were already part of therapeutic programs and recruited by their immediate care providers. Mothers who would not have reached such institutions due to the delay in diagnosis in the child, or not able to access supportive services may be facing different levels or patterns of stress. However, in Kerala, children with developmental disorders are likely to be diagnosed and treated at reasonably high levels. All institutions with good case load catering for children with disabilities in selected district were included. Thus, the findings are quite likely to be generalizable, at least within the state of Kerala. Also, we did not study the degree of disability of the child, the level of behavioral problems in the child, or the depth of the relationship between the child and the mother, or the actual actions mothers were taking to cope with the stress. These may have important implications on the stress levels of the participating mothers.

Future Research

Parental self-efficacy, a feeling of parents that they are competent in their parental role, has been found to be a mitigator for high stress levels in parents with low spousal support.30 Self-efficacy of mothers with low spousal support can be explored in future studies.