Early detection of autism spectrum disorder in young isiZulu-speaking children in South Africa

Abstract

Culturally appropriate tools are needed for detecting symptoms of autism spectrum disorder in young South African children. The objectives of this study were to (1) adapt and translate into isiZulu existing measures for detecting early signs of autism spectrum disorder, (2) use the measures to characterize and compare behavioural profiles of young isiZulu-speaking children with and without autism spectrum disorder and (3) compare symptom profiles across sampling procedures. Measures were translated and adapted into isiZulu and used to evaluate 26 isiZulu-speaking children, 15 children with no reported developmental concerns and 11 referred for suspected autism spectrum disorder. A video-recorded observation of children and caregivers in their home environment was also made. Based on best-estimate diagnoses, 10 children were classified as autism spectrum disorder and 16 as non-autism spectrum disorder. The children with autism spectrum disorder presented with significantly more autism spectrum disorder red flags than the non-autism spectrum disorder group according to parent report and systematic ratings of red flags. Significant correlations between parent report and observational measures of red flags were observed. More red flags were observed during structured evaluations than home observations in the autism spectrum disorder group. Findings provide a foundation for tool translation and adaptation in South Africa and identifying social communication markers to detect autism spectrum disorder in young isiZulu-speaking children.

Keywords

autism spectrum disorder isiZulu red flags South Africa toddlers

Introduction

In a recent review of global epidemiological studies, Elsabbagh et al. (2012) presented a global estimate of autism spectrum disorder (ASD) of 62/10,000 children (0.62%). Others suggest that this figure is closer to 1%–1.5% of the population (Schendel et al., 2012). This increase in estimated prevalence over the last two decades has raised awareness of ASD worldwide. There are no prevalence studies of ASD in South Africa (SA), or indeed in sub-Saharan Africa as a whole, due in part to a lack of standardized screening and diagnostic tools validated for African populations (Harrison et al., 2014) and limited access to appropriate intervention in the public health and education sectors in Africa and SA in particular (Malcolm-Smith et al., 2013). The development of appropriate and accessible services depends to a large extent on valid estimates of ASD in SA, understanding of the phenotype and early developmental course of ASD symptoms in the SA context and the development or adaptation of valid and reliable screening and diagnostic tools for SA children in their home language. Given the mounting evidence for the effectiveness of early intervention in changing developmental trajectories of children with ASD, healthcare practitioners need tools to detect ASD as early as possible (i.e. under 3 years of age; Wetherby et al., 2014).

This article reports on a foundational study of early social communication markers of ASD in a relatively homogeneous linguistic and cultural setting in SA; specifically, isiZulu-speaking children from the KwaZulu-Natal (KZN) province. There are no validated measures available for detecting ASD in young isiZulu-speaking children. This community also experiences significant challenges in seeking early detection of ASD, including lack of access to healthcare due to poverty, lack of knowledge of ASD or the importance of early social communication milestones and social stigma associated with having a child with a disability.

If families do seek help, it is unknown how Zulu interactional styles may affect clinical observations of ASD symptoms. Anecdotal evidence suggests that Zulu styles of social interaction are shaped by social structure. Zulu society is organized around a gender-based, age-cohort system of socialization in which the care of toddlers is often in the hands of older siblings; hence, opportunities for extended interactions with adults, particularly unfamiliar adults, may be more limited than in some Western communities (Kvalsvig et al., 2013). Given the extent of male out-migration for work, as well as the high number of children orphaned as a result of HIV/AIDS, children are often raised by their mothers or other close female relatives in female-headed households. Although the rate of fosterage is high compared to other African countries, fostered children typically live with multiple consanguineous relatives (McDaniel and Zulu, 1996). Conventions of direct eye gaze with adults, such as those reported in rural Kenyan children where they are taught it is disrespectful to make eye contact with adults in authority (Carter et al., 2005), are also anecdotally ascribed to Zulu children. Researchers in the education field have noted the tendency for children to take the initiative in speaking only when confident they will not experience shame or a loss of dignity, and in school settings prefer to communicate with superiors by ‘chorusing’ rather than speaking alone (Chick, 1996). Such features may still influence the early social communication profiles of today’s Zulu toddlers in clinical settings, although it is understood that cultures can change rapidly due to political and economic circumstances (Kvalsvig et al., 2013). It is not known whether these distinctive cultural styles of early social interaction between parents and toddlers in this cultural group may impact the presentation of early red flags of ASD, or the ability to detect red flags using tools developed in the West; regardless, like all societies, isi Zulu-speaking South Africans are able to report deficits in social communication based on their expectations of child development.

This study was a collaboration between researchers at the Autism Institute at Florida State University (FSU) and George Washington University (GWU) in the United States (US) and the Department of Paediatrics at the University of KwaZulu-Natal, SA. The team of collaborators included experts in early detection of ASD in toddlers based on the screening and evaluation methods of the FIRST WORDS^® Project at FSU, the epidemiology of neurodevelopmental disorders in sub-Saharan Africa and global expertise in the role of culture in ASD studies. The objectives of the study were to (1) adapt and translate into isiZulu existing measures for detecting early signs of ASD, (2) use the measures to characterize and compare behavioural profiles of a sample of young isiZulu-speaking children with and without ASD and (3) compare symptom profiles across sampling procedures.

Methods

Socio-demographic profile of study population

Participants were recruited in urban Pietermaritzburg and peri-urban areas of Sweetwaters and Edendale. These three areas (along with Bishopstowe and Raisethorpe) comprise the Msunduzi municipality (population 620,000), the largest municipality in the uMgungundlovu District. The SA government classifies residents in Msunduzi as 81% Black African, 10% Indian/Asian, 6% White and 3% mixed race/Coloured people (Statistics South Africa, 2011). The majority speak isiZulu (71%) followed by English (19%), with other national languages accounting for less than 2% each. Sweetwaters and Edendale are situated within 10 and 12 km of the Pietermaritzburg city centre, respectively. Sweetwaters consists largely of what the government terms ‘traditional tribal lands’ where residents lease land from the local tribal chief and build their own dwellings from mud or bricks, often in the traditional ‘rondavel’ style (round hut with a thatched or metal roof). Edendale also contains free government-built housing, usually small houses out of concrete blocks or bricks. Census 2011 data suggest that 73% of the population in this municipality live in formal dwellings, and 92% have access to electricity for lighting. However, only half of domociles have access to a flush toilet or piped water. There is an overall unemployment rate of 33%, 43% in younger adults, with 16% of households in the area reporting no income. KZN is also the province with the highest overall prevalence of HIV and AIDS in SA, which has significant implications for overall health outcomes and child-rearing practices in KZN (Kvalsvig et al., 2013), especially given the high level of fosterage and residence in orphanages.

Participant recruitment

Purposive sampling was used to recruit participants between 12 and 48 months of age from the population. To recruit typically developing toddlers, our trained isiZulu-speaking community liaison invited families with no concerns about their children’s development to participate. Over time, families also volunteered to participate after hearing of the study by word of mouth. As lack of parental concern about development was a broad inclusion criterion, we anticipated that some children may have mild developmental or language delays but children with moderate-to-severe delays were not likely to be included.

In an effort to recruit children at risk of, or already suspected of, having ASD, primary, secondary and tertiary health centres and the KZN branch of Autism South Africa (ASA), the national society for people with autism, were approached for possible candidates. The first author (N.J.C.) held several training workshops on ASD for medical and paramedical professionals in the area to raise awareness about early red flags for ASD and the study. An Autism Diagnostic Observation Schedule (ADOS) training held in Pietermaritzburg at the start of the study to train study personnel and other interested medical professionals was attended by over 60 professionals. Families were offered small incentives for their participation: small toys for the children following each clinical evaluation, and a R200 (±US$20) shopping voucher for a local grocery store chain. All transport costs to the clinic for evaluations were also reimbursed.

Participant characteristics

Participant characteristics are presented in Table 1. Based on a best-estimate diagnosis made through consensus using information from all evaluation measures (process described below), 10 children were classified as ASD and 16 as non-ASD. All children ultimately classified as ASD were referred to the study by staff at public healthcare facilities (including two speech therapists, from a specialist assessment clinic and a tertiary hospital, and a paediatrician) or the KZN representative of ASA (a mother of a child with ASD) and had been identified with potential red flags of developmental delay or ASD; although none had been given a definitive diagnosis of ASD by the referring agents. All of these referral sources had been informed of the study and given a brochure listing early red flags of ASD in toddlers (Wetherby et al., 2007). All of the non-ASD group, except one, were recruited by our community liaison. The exception was referred by a nurse whose own child has ASD.

Table 1.

Demographic characteristics of isiZulu-speaking participants (n = 26).

	ASD (n = 10)	Non-ASD (n = 16)
Gender (male): n (%)	7 (70%)	13 (81%)
Mean (SD) age in months	38.02 (5.28)	22.79 (4.82)**
Referral source
Invitation	0	15
Hospital/clinic	4	1
Private sector	4	0
Autism South Africa	2	0
Mean (SD) parental age in years
Mother	27.81 (3.51)	27.03 (8.69)
Father	33.09 (6.54)	30.41 (11.81)
Mean (SD) years parental education
Mother	14.00 (1.76)	9.94 (3.64)**
Father	13.89 (1.83)	11.92 (2.29)*
Source of income (n)
Formal sector	5	0
Informal sector	1	3
Grants	4	12
None	0	2
Type of dwelling
Self-built	3	14
Government built	2	2
Privately built	5	0
Piped water/tap in house	7	2
Flush toilet connected to sewage	6	1
Electricity	10	16

ASD: autism spectrum disorder; SD: standard deviation.

p < 0.01; *p < 0.05.

The children with ASD were significantly older at the time of assessment than the non-ASD group (t = 7.55, p < 0.001), and despite their older age, the ASD group was also lower functioning than the non-ASD group as measured by the Communication and Symbolic Behavior Scales (CSBS) Symbolic Composite (t = −5.35, p < 0.001), presented in Table 2. Based on language used during the CSBS, only 10% of the ASD group used single words and 90% were nonverbal; in contrast, 62.5% of the non-ASD group used single words and 37.5% were nonverbal. None of the children used word combinations. As anticipated, the non-ASD group consisted of typically developing and mildly developmentally delayed children.

Table 2.

Comparison of weighted raw scores on study measures.

Measure	ASD (n = 10)	Non-ASD (n = 16)	F (ANCOVA)	p	d	95% CI
ESAC
Total	32.60 (8.40)	12.50 (8.85)	4.34	0.048	2.33	(12.87, 27.33)
CSBS clusters
Emotion/eye gaze	4.30 (3.06)	14.25 (3.84)	37.78	0.001	−2.87	(−12.91, −6.98)
Communication	9.57 (3.53)	16.08 (3.32)	21.66	0.001	−1.90	(−9.34, −3.69)
Gestures	4.10 (2.02)	10.75 (4.83)	14.94	0.001	−1.80	(−9.99, −3.31)
Sounds	3.50 (4.74)	12.56 (7.62)	17.70	0.001	−1.42	(−14.62, −3.50)
Words	0.65 (2.06)	5.59 (6.10)	7.46	0.012	−1.09	(−9.09, −0.79)
Understanding	0.90 (2.85)	11.25 (8.16)	24.19	0.001	−1.69	(−15.91, −4.79)
Use of objects	5.90 (5.22)	15.13 (3.95)	29.70	0.001	−1.99	(−12.94, −5.51)
CSBS composites and total
Social	17.97 (7.29)	41.08 (10.64)	33.21	0.001	−2.53	(−31.03, −15.20)
Social AE	9.50 (1.08)	16.13 (4.69)	11.55	0.002	−1.95	(3.67, 15.11)
Speech	4.15 (6.58)	18.16 (13.30)	13.62	0.001	−1.34	(−23.97, −4.64)
Speech AE	10.90 (3.21)	16.88 (4.98)	19.22	0.001	−1.43	(6.68, 18.62)
Symbolic	6.80 (5.25)	26.37 (10.73)	48.84	0.001	−2.32	(−27.12, −12.03)
Symbolic AE	11.30 (2.21)	17.81 (3.49)	43.75	0.001	−2.23	(8.39, 16.02)
Total	28.92 (16.26)	85.61 (29.73)	42.71	0.001	−2.37	(−77.94, −35.46)
Total AE	10.10 (1.91)	16.94 (3.82)	37.17	0.001	−2.26	(8.17, 16.57)
SORF
Systematic observation (CSBS)	40.90 (4.56)	7.63 (7.42)	67.80	0.001	5.40	(27.87, 38.67)
Naturalistic observation (home)	29.41 (7.76)	8.06 (7.05)	20.64	0.001	2.88	(15.24, 27.43)

ASD: autism spectrum disorder; CI: confidence interval; CSBS: Communication and Symbolic Behavior Scales; ESAC: Early Screening for Autism and Communication Disorders; SORF: Systematic Observation of Red Flags; AE: age equivalent; ANCOVA: analysis of covariance (controlling for age).

The families with children with ASD were generally characterized by higher socio-economic status (SES) indices compared to the families of the children without ASD. Their parents reported significantly more years of education (fathers t = 2.14, p < 0.038; mothers t = 3.28, p < 0.003), and half of the parents of the ASD group were employed in the formal sector. In contrast, none of the parents of the non-ASD group were employed in the formal sector but instead relied on temporary work or social grants for income. The other socio-economic indicators listed in Table 1 indicate that the non-ASD group represented a low socio-economic group typical of the recruitment region.

Measures

The following measures were selected based on experience and expertise of the collaborators, their appropriateness for the young age of the target sample and emerging evidence of these measures being suitable for cultures other than mainstream US samples.

Early Screening for Autism and Communication Disorders (ESAC; Wetherby et al., 2012): The ESAC is a parent-report screening tool developed to detect absence or delay in typical social communication skills and presence of unusual behaviours associated with ASD using the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5) framework (American Psychiatric Association (APA), 2013). The ESAC consists of 30 items selected on the basis of their psychometric properties and clinical utility. Higher scores indicate more red flags. This measure demonstrated good sensitivity (0.88), good specificity (0.81) and excellent AUC (area under receiver operating curve) (0.92) using a cutoff score of 18. Scores on the ESAC have been examined across ethnic groups in the US and found to be comparable when controlling for maternal education (Stronach, 2013).

CSBS-Developmental Profile Behavior Sample (Wetherby and Prizant, 2002). The Behavior Sample of the CSBS was selected for the study and scored using the conventional scoring and a systematic rating of red flags (described in 3 below). The conventional scores were calculated as these yield profiles of social communication skills shown to differentiate children with ASD from those with developmental delays and typical development in the US (Wetherby et al., 2007). Higher conventional scores indicate better functioning. Conventional scores have also shown promise for accurately profiling ASD symptoms in minority populations within the US (Stronach, 2013) as well as English-speaking children in Australia (Eadie et al., 2010) and more relevantly, SA (Chambers et al., 2016). Chambers recently described the performance of 70 typically developing English-speaking SA toddlers from a range of cultural backgrounds on the CSBS. They found no significant differences on cluster, composite or total scores between the SA sample and the US standardization sample from 15 months of age suggesting that the CSBS holds promise for more widespread testing in SA samples. See Chambers et al. (2016) for further description of this measure’s psychometric properties and inter-rater reliability.

The Systematic Observation of Red Flags (SORF) of ASD (Wetherby et al., 2016). The SORF is an observational screening measure consisting of 22 items in two behavioural categories based on the DSM-5 (APA, 2013): social communication and interaction and repetitive movements and restricted interests. The SORF was used to code red flags of ASD in the CSBS clinical sample as well as the home observation (see 5 below), for example, poor eye gaze, less interest in people than objects and repetitive movements with objects. Higher scores on the SORF indicate more red flags for ASD. SORF coding on the CSBS samples of US toddlers has also been shown to distinguish children with ASD from those with developmental delay (DD) and typically development (TD) (Wetherby et al., 2004). Using a cutoff composite score of 20, this measure has good sensitivity (0.80), specificity (0.78) and AUC (0.87) for toddlers aged 16–24 months (Dow et al., 2016). To account for source and magnitude variance of scorer rating and child behaviours, g intraclass correlation coefficients were calculated for a larger set of 101 US SORF samples. For adequate inter-rater reliability, g coefficients should be 0.60 or higher (Mitchell, 1979). SORF sample g coefficients were 0.84 for social communication items, 0.76 for repetitive behaviours and restricted interests and 0.86 for the total composite.

The Autism Diagnostic Observation Schedule Second Edition (ADOS-2; Lord et al., 2012a, 2012b). The ADOS is considered a gold standard component of a diagnostic test battery to evaluate children considered to be at risk of ASD. No data has been published regarding the suitability of this tool with children in SA, particularly non-English-speaking children, who form the majority of the SA population. Participants received the Toddler Module or Modules 1 or 2 depending on their age and expressive language level.

Naturalistic home observation. A video-recorded observation of the children and caregivers in their home environment was obtained to sample behaviour across a variety of naturalistic everyday activities. The home observations were coded using the SORF to quantify the presence of red flags in the home environment as research suggests that the presentation of red flags may differ across structured clinical and naturalistic home contexts (Stronach and Wetherby, 2014). Given the expected lack of familiarity with structured evaluation situations in the children in this study, it was deemed important to include a naturalistic home observation as an important indicator of validity of the both the structured clinical measures and the parent-report measure. The home observation also gave necessary information for prioritizing intervention goals relevant for families in their natural environments.

The home observation lasted a minimum of 30 min (typically 45–60 min) and included a snack or meal time, caregiving routine, (e.g. washing hands, brushing teeth), family chore (e.g. sweeping, feeding chickens, helping with washing up), reading activity (e.g. looking at a newspaper, book or magazine), play with people (e.g. songs, rhymes, tickle games) and play with objects (i.e. toys or materials the family allowed their toddlers to play with naturally at home). No outside materials or toys were provided to families for this observation.

Procedures

Phase 1 involved foundational work to prepare for participant recruitment and evaluations

Community engagement

Key informant interviews and focus groups were carried out before and during data collection to engage members of the KZN community, including parents and teachers of children with ASD, day-care teachers, nurses at primary health clinics, traditional healers, doctors, speech therapists, psychologists, occupational therapists and paediatricians. Barriers to participation in the study were identified, specifically, a lack of knowledge of ASD and a fear among primary health (e.g., nurses) and educational personnel (e.g. crèche teachers) that informing parents of a developmental aberration would lead to stigma and withdrawal from the school or clinic. Other findings from these meetings are reported elsewhere (Grinker et al., 2012).

Translation and adaptation of measures

A systematic process of tool adaptation and translation was followed with publisher permission (Bracken and Barona, 1991). The scripts of the ESAC, CSBS and ADOS (Toddler, 1 and 2 Modules) were submitted to the following steps.

English-to-English adaptation. The questions and statements in each instrument were assessed for comprehensibility and relevance to the Zulu culture through a set of cognitive interviews (Yucel et al., 2011). Nine bilingual isiZulu parents were recruited to evaluate 193 statements and questions of the screening and diagnostic assessments in one-on-one cognitive interviews. Participants were female, ranging in age from 22 to 39 years (average age 29). The cognitive review of some statements was handled by reading the statement to the respondents and having them paraphrase using their own words what they understood the statement to be saying. Wherever the content had the potential to be culturally irrelevant, respondents were asked whether the content in question was common or typical among Zulu people (e.g. ‘Is it typical for Zulu children to have a birthday cake to celebrate their birthday?’). Based on the results of the cognitive interviews, recommendations for English-to-English revisions were made. All of these recommendations were reviewed by test administrators in the US and the SA research team, including the manager of the study and the isiZulu-speaking clinician who would be administering the tools. In all, 70 (36%) of the 193 statements/questions were revised, 45% were simplified English wording for greater clarity, 28% replaced English idiomatic expressions, 15% replaced ambiguous English words with more specific ones and 12% eliminated or replaced content that was not culturally relevant. For example, the original ESAC item ‘Is it easy to get your child to respond to your instructions with gestures (for example, when you say “give it to me” while holding out your hand or while pointing to the object, does your child readily give you the object)?’ was simplified to ‘Does your child follow your instructions when you use gestures? For example, when you say “give it to me” and hold out your hand or point to the object, does your child give it to you?’

Forward translation into isiZulu. The adapted English scripts were forward translated into isiZulu.

The first isiZulu versions were reviewed by the SA research team and revisions were recommended, mostly relating to more regionally appropriate isiZulu expressions that were common in the study area and to replace formal (‘high’) isiZulu expressions in the original translation. The revised isiZulu versions were then blind back-translated into English.

The first English back-translations were reviewed by the SA and US research teams, and further revisions to the isiZulu script were suggested where it was clear that the translator had misunderstood the concepts or awkward expressions were used. An example of an awkward back-translated statement follows from the previous example. The simplified ESAC item ‘Does your child follow your instructions when you use gestures?’ was back-translated as: ‘Does your child follow your instructions when you use your parts of your body?’ as there is no equivalent isiZulu term for ‘gestures’.

The revised scripts were submitted to a second blind back-translation into English by an independent translator. The scripts for the ESAC and CSBS were considered appropriate for use at this point.

The revised version of the ADOS script was then submitted to a review by an ADOS expert to ensure that the original intent behind the ADOS statements had been maintained throughout this process. The ADOS expert highlighted a number of inconsistencies. The ADOS script was then revised by the SA research team via a process of consensus.

Adaptations to equipment or activities

With respect to test adaptations, our goal was keeping as close to the original administration as possible, while being sensitive to potential cultural issues. For the ESAC, the only adaptation was in the mode of presentation. Although designed as a self-report questionnaire, it was administered as an interview to account for varying levels of parental literacy and also to begin establishing rapport and consensus with families about their child’s development. No adaptations to the CSBS equipment or activities were made. In the ADOS, participants in the cognitive interviews identified two activities and equipment that they considered culturally inappropriate, specifically, the birthday party in Module 1 and aspects of the bath-time activity in the Toddler Module. Annual birthday parties with cakes and candles are not a part of traditional Zulu home life in KZN. With this information, we decided to ask parents at the start of an ADOS Module 1 evaluation whether their child was familiar with the event of a birthday party, with cake and candles. If yes, parents were also asked whether their child was familiar with the English or isiZulu birthday song and, if so, the relevant song was used. If not, this activity was substituted with the bath-time routine from the Toddler Module. Two adaptations were made in the bath-time routine: if the child was from a home with no indoor plumbing, we used a small bucket to pour water into the baby bathtub rather than pretending with taps. Also, we replaced the unfamiliar ‘Rubber Ducky’ song with a well-known isiZulu lullaby ‘Thula Thula Mntwana’ and sang this while putting the doll to sleep after the bath. Participants and families in the study also debated the use of a white versus brown-skinned doll in the ADOS, with some participants advocating a brown baby, and others saying that Zulu children were familiar only with white dolls. We used both interchangeably and noticed no differences in responses of the participants.

Phase 2: participant evaluations

Phase 2 evaluated participants in the following manner: Following participant invitation and consent, the community liaison team member visited the family and completed a participant information form, conducted the ESAC in an interview format and scheduled a second home visit with the family including the provision of oral and written information about the upcoming home observation. At the second visit, the isiZulu-speaking speech-language pathologist, trained in the administration of all instruments, accompanied the family liaison to oversee the home observation video recording and administer the CSBS. At the final clinical appointment, the same clinician administered the ADOS in a standard room, to reduce anxiety due to lack of familiarity and keep the evaluation context consistent across all families.

The ESAC forms, and video recordings of the CSBS, home observation and ADOS were disseminated to the US and SA research teams. All isiZulu vocalizations in the clinical evaluations were translated by the study clinician. The CSBS and home observations were scored by coders blind to diagnostic group. Using all measures and video ratings, the teams independently made a clinical diagnosis for each participant. Comparison revealed 100% agreement between the teams. However, although the study clinician was research reliable in her scoring of videos of English ADOS evaluations, her reliability on scoring the isiZulu ADOS evaluations could not be confirmed by the English-speaking members of the team, including the US ADOS consultant. As a result, the team decided not to calculate or report ADOS scores. Video recordings of the ADOS evaluations were viewed to assist a clinical diagnosis, but no diagnosis was made on the basis of any quantitative rating.

Children identified as having ASD were offered six intervention sessions focused on equipping caregivers with strategies to enhance their child’s active engagement in everyday activities in their natural environments. Intervention included both clinic and home visits to coach caregivers. All children with ASD were also receiving standard care through the public health service.

Analyses

Parametric tests were used for all analyses as no obvious violations of normality were noted (see results below). Differences in scores between the groups with and without ASD were calculated using analysis of covariance (ANCOVA) to control for differences in age. Cohen’s d was used as an effect size index with conventions of small (0.20), medium (0.50) and large (0.80). As the sample sizes in this study were small, and the ASD and non-ASD groups differed on a number of parameters (e.g. age and SES), confidence intervals (CIs) were also calculated and reported. CIs provide information about a range in which the true value lies with a certain degree of probability, as well as about the direction and strength of the demonstrated effect, useful when groups have not been randomly sampled or assigned from the same population (Shakespeare et al., 2001). Relationships between measures were calculated using Pearson’s r. A paired sample t-test was used to compare SORF ratings across evaluation contexts.

Ethical approval

The study was approved by the Biomedical Research Ethics Committee at the University of the KZN (BF258/09), the GWU Committee on Human Research (IRB#120910) and the FSU Institutional Review Board (HSC no. 2011.6731) before data collection.

Results

Comparison of behavioural profiles between children with and without ASD

Indices of normality were reviewed for the CSBS cluster, composite, and total, SORF, and ESAC scores and skewness (−0.46 to 1.41) and kurtosis (−1.76 to 0.75) values fell within an acceptable range (−2 to 2). Scores on the ESAC, CSBS and SORF are provided and compared across groups in Table 2. Higher scores on the ESAC and SORF reflect more red flags, while higher scores on the CSBS indicate better functioning. Table 3 provides a summary of correlations among measures.

Table 3.

Correlations among measures.

	ESAC	CSBS	SORF CSBS
CSBS	−0.62
SORF CSBS	0.66	−0.85
SORF HOME	0.80	−0.75	0.82

ESAC: Early Screening for Autism and Communication Disorders; CSBS: Communication and Symbolic Behavior Scales; SORF CSBS: Systematic Observation of Red Flags in CSBS; SORF HOME: Systematic Observation of Red Flags for Home Observation.

Controlling for age, there were significant group differences with large effect sizes on all scores on the ESAC, CSBS conventional scores and SORF ratings of the CSBS and home observations. In other words, children with ASD presented with significantly more red flags according to parent report and during systematic coding for red flags of the CSBS and the naturalistic home observation. The children with ASD also scored significantly lower on the conventional scoring of the CSBS on all cluster, composite and total scores, indicating significantly poorer social communication skills despite their older age. In addition, Table 3 indicates that scores across measures were all significantly correlated, with moderate-to-large effect sizes, suggesting good agreement across measures.

Comparison of behavioural profiles across sampling context

Paired sample t-tests indicated that the ASD group demonstrated significantly more red flags according to the SORF during the CSBS Behavior Sample than the naturalistic home observation (t = 4.33, degree of freedom (df) = 9, p = 0.002, d = 1.81, 95% CI = (5.49, 17.51)). This suggests that the semi-structured Behavior Sample was able to reveal more red flags than the naturalistic home observation in this sample. No differences in SORF scores were noted across contexts in the non-ASD group.

Discussion

This was an exploratory study to begin the process of validating tools to detect ASD in young isiZulu-speaking children. The project included foundational qualitative steps, as well as detailed clinical evaluations and analysis using quantitative methods. The early steps were necessary to ensure community engagement in the process and to begin building ASD expertise and capacity in the KZN region. Although the sample size was small and the ADOS could not be employed quantitatively, the findings revealed significant differences between young isiZulu-speaking children with ASD and those without ASD on all other measures and across informants (parents vs skilled coders) and sampling contexts. The large differences in groups (as indicated by the large effect sizes) suggest that potential cultural differences in parent–child interaction styles did not preclude the ability to detect red flags for ASD and also that parents and trained coders were able to agree on the presence of red flags in the children. These results support previous research that indicated the utility of the measures used in this study across race and ethnicity in the US (Stronach, 2013).

The CSBS provided more opportunities to observe red flags than the naturalistic home observation. This finding is consistent with previous research. Stronach and Wetherby (2014), for example, found significantly higher measures of restricted and repetitive behaviours, particularly repetitive movements with objects, clutching objects and sticky attention to objects during the CSBS compared to a home observation. It was hypothesized that the availability of specific objects and high rates of transitions between activities and objects contributed to this between-context difference. The presentation of more red flags in the CSBS in this sample would not be attributable to discomfort or anxiety related to an unfamiliar clinic context, as the CSBS was conducted in the child’s home. Despite differences in observed red flags between the two contexts, CSBS and SORF scores from both contexts demonstrated good agreement with parent report. The findings suggest that the measures used in the study are sensitive to the presence of red flags of ASD in this sample of young isiZulu-speaking children and provide a foundation for further and wider-scale testing of the measures.

Selection bias is an important limitation of the study, as participants were not randomly selected or detected via population-based screening. All children diagnosed with ASD were referred to the project with suspected ASD and therefore displayed symptoms of sufficient severity to detect easily. The study clinician was also aware of the referral source of each participant at the time of her evaluations. It is important to apply these tools in a broader community-screening approach to determine their utility for detection of children with milder presentations of ASD. A larger, more representative sample of children with ASD is essential for validation of the tools prior to their use for clinical or epidemiological purposes. Another limitation is that the same evaluation contexts were used to first classify the children to groups and then to compare behavioural profiles across the groups. However, a strength of this study is that the CSBS and SORF scores were coded blind to diagnostic group. It is unfortunate that the ADOS scores could not be calculated in the study, given the extensive process of translation, adaptation and training that took place. All members of the US and SA research teams, however, felt that the structured sampling context of the ADOS provided valuable qualitative information that informed clinical diagnosis. Despite not reporting ADOS scores, a description of the process of translation and adaptation has value for other research teams involved in such endeavours generally and provides a solid foundation for future research with isiZulu-speaking samples in particular. Building clinical and research capacity in SA, and ultimately serving families throughout SA more effectively, requires an increase in the number of research reliable isiZulu-speaking (and indeed any of the other 10 official SA languages) clinicians.

In conclusion, the findings suggest that existing tools for detecting ASD have potential for characterizing symptoms of ASD in young isiZulu-speaking children following appropriate translation and adaptation procedures. The process reported here may serve as a useful guide for other researchers navigating the complex process of translation, adaptation and application of ASD evaluation tools, and facilitating early detection of ASD, across cultures, and especially in low-resource settings. With detection comes the need for access to evidence-based early intervention services, a crucial area for future research. Planned future studies include comparisons of isiZulu-speaking children to international samples of children with and without ASD. Cross-cultural studies have the potential to inform characterization of the ASD phenotype in very young children and, by extension, underlying gene–environment interactions.

Footnotes

Acknowledgements

The authors thank all of the families who have participated in the project and the US and SA staff for their assistance with data collection and coding.

Declaration of conflicting interests

Author Amy M Wetherby receives royalties for the CSBS. All other authors have no conflict of interest to declare.

Funding

The author(s) received no financial support for the research, authorship and/or publication of this article.

References

American Psychiatric Association (APA) (2013) Diagnostic and Statistical Manual of Mental Disorders. 5th ed. Washington, DC: APA.

Bracken

Barona

(1991) State of the art procedures for translating, validating and using psychoeducational tests in cross-cultural assessment. School Psychology International 12: 119–132.

Carter

Lees

Murira

. (2005) Issues in the development of cross-cultural assessments of speech and language for children. International Journal of Communication Disorders 40: 385–401.

Chambers

Stronach

Wetherby

(2016) Performance of South African children on the Communication and Symbolic Behavior Scales-Developmental Profile (CSBS-DP). International Journal of Language and Communication Disorders 51: 265–275.

Chick

(1996) Safe-talk: collusion in Apartheid Education. In: Coleman

(ed.) Society and the Language Classroom. Cambridge: Cambridge University Press, pp.21–39.

Dow

Guthrie

Stronach

. (2016) Psychometric analysis of the Systematic Observation of Red Flags (SORF) for autism spectrum disorder in toddlers. Autism: International Journal of Research and Practice. Epub ahead of print 30 April. DOI: 10.1177/1362361316636760.

Eadie

Okuomunne

Skeat

. (2010) Assessing early communication behaviours: structure and validity of the Communication and Symbolic Behaviour Scales – Developmental Profile (CSBS-DP) in 12-month-old children. International Journal of Language and Communication Disorders 45: 572–585.

Elsabbagh

Divan

Koh

Y-J

. (2012) Global prevalence of autism and other pervasive developmental disorders. Autism Research 5: 160–179.

Grinker

Chambers

Njongwe

. (2012) ‘Communities’ in community engagement: lessons from autism research in South Africa and South Korea. Autism Research 5: 201–210.

10.

Harrison

Zimak

Sheinkopf

. (2014) Observation-centered approach to ASD assessment in Tanzania. Intellectual and Developmental Disabilities 52: 330–347.

11.

Kvalsvig

Taylor

Kauchali

. (2013) Acknowledging methodological complexity in assessing children in HIV-affected communities in KwaZulu-Natal province, South Africa. In: Boivin

Giordani

(eds) Neuropsychology of Children in Africa. New York: Springer, pp.37–68.

12.

Lord

Luyster

Gotham

. (2012a) Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) Manual (Part II): Toddler Module. Torrance, CA: Western Psychological Services.

13.

Lord

Rutter

DiLavore

. (2012b) Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) Manual (Part I): Modules 1–4. Torrance, CA: Western Psychological Services.

14.

McDaniel

Zulu

(1996) Mothers, fathers, and children: regional patterns in child-parent residence in Sub-Saharan Africa. African Population Studies/Etude de la Population Africaine 11(1): 1–28.

15.

Malcolm-Smith

Hoogenhout

Ing

. (2013) Autism spectrum disorders – Global challenges and local opportunities. Journal of Child and Adolescent Mental Health 25: 1–5.

16.

Mitchell

(1979) Interobserver agreement, reliability, and generalizability of data collected in observational studies. Psychological Bulletin 86: 376–390.

17.

Schendel

Diguiseppi

Croen

. (2012) The study to explore early development (SEED): a multisite epidemiologic study of autism by the Centers for Autism and Developmental Disabilities Research and Epidemiology (CADDRE) Network. Journal of Autism and Developmental Disorders 42: 2120–2140.

18.

Shakespeare

Gebski

Veness

. (2001) Improving interpretation of clinical studies by use of confidence levels, clinical significance curves, and risk benefit contours. The Lancet 357: 1349–1353.

19.

Statistics South Africa (2011) Local Municipality: Msunduzi. Available at: http://www.statssa.gov.za/?page_id=993&id=the-msunduzi-municipality

20.

Stronach

(2013) Social communication profiles of toddlers with and without autism spectrum disorder from three racial/ethnic groups. PhD Dissertation, Florida State University, Tallahassee, FL.

21.

Stronach

Wetherby

(2014) Examining restricted and repetitive behaviors in young children with autism spectrum disorders during two observational contexts. Autism 18: 127–136.

22.

Wetherby

Prizant

(2002) Communication and Symbolic Behavior Scales-Developmental Profile (CSBS-DP) (First Normed Edition). Baltimore, MD: Paul H Brooks.

23.

Wetherby

Guthrie

Woods

. (2014) Parent-Implemented Social Intervention for Toddlers with Autism: an RCT. Pediatrics. Epub ahead of print 3 November. DOI: 10.1542/peds.2014–0757.

24.

Wetherby

Woods

Lord

(2012) Early Screening for Autism and Communication Disorders (ESAC). Tallahassee, FL: Florida State University.

25.

Wetherby

Woods

Nottke

. (2016) Systematic Observation of Red Flags of ASD (SORF) Manual. Tallahassee, FL: Florida State University.

26.

Wetherby

Watt

Morgan

. (2007) Social communication profiles of children with autism spectrum disorders late in the second year of life. Journal of Autism and Developmental Disorders 37: 960–975.

27.

Wetherby

Woods

Allen

. (2004) Early indicators of autism spectrum disorders in the second year of life. Journal of Autism and Developmental Disorders 34: 473–493.

28.

Yucel

Kudumu

Barker-Cummings

. (2011) An innovative approach to cognitive interviewing for culturally adapting autism screening instruments. American Journal of Epidemiology 173(Suppl. 11): S225.