Measurement of Subcategories of Repetitive Behaviors in Autistic Adolescents and Adults

Abstract

Background: Restricted and repetitive behaviors (RRBs) are core features of autism. Factor-analytic studies composed primarily of children have provided evidence for two domains of RRBs: Repetitive Sensory Motor (RSM) and Insistence on Sameness (IS) behaviors. The present study explores the validity of the Autism Diagnostic Interview-Revised (ADI-R) and the Repetitive Behavior Scale-Revised (RBS-R) for assessing these RRB subtypes in autistic adolescents and adults.

Methods: The sample included 293 participants (M_age = 19.89 years, SD_age = 4.88 years) whose RRBs were assessed via ADI-R, RBS-R Caregiver Report or RBS-R Self-Report. Confirmatory factor analysis (CFA) was conducted to assess the validity of the two-factor structure for each instrument. Cronbach's alpha was computed to assess subscale reliability. Correlations were examined between instrument subscales and nonverbal intelligence quotient (NVIQ) and age.

Results: Exploratory correlations were modest and provided weak evidence in favor of the utility of a CFA for the ADI-R. The RBS-R Caregiver and Self-Report CFA and internal consistencies supported the two-factor RSM and IS model tested. Consistent with previous literature, NVIQ was negatively correlated with the RBS-R Caregiver RSM subscale, but not meaningfully associated with IS. Neither RBS-R Self-Report subscale was meaningfully correlated with NVIQ. Across instruments, RSM subscales were correlated, but associations between IS were minimal.

Conclusions: The present study provides initial support for the use of the RBS-R Caregiver and Self-Report to measure dimensions of RSM and IS behaviors in autistic adolescents and adults. The present data did not support the use of the ADI-R to assess these RRB subtypes in older individuals. Conclusions must be interpreted cautiously in light of the present study's sample limitations. Additional research is needed to understand differences in caregiver-reported and self-reported RRBs. Further research on RRBs in autistic adolescents and adults, particularly in samples with more gender and racial/ethnic diversity, is critical to inform community understanding and knowledge of autism in adulthood.

Lay summary

Why was this study done?

Restricted and repetitive behaviors (RRBs) are features necessary for the diagnosis of autism spectrum disorder and are assessed at all ages. It cannot be assumed, however, that instruments designed for assessing children are appropriate for use with older individuals. Therefore, we explored whether the Autism Diagnostic Interview-Revised (ADI-R) and the Repetitive Behavior Scale-Revised (RBS-R) questionnaire, instruments used to evaluate RRBs in children, can be used to assess adolescents and adults. This information is important to inform clinicians and researchers about the best ways to assess RRBs in older individuals.

What was the purpose of this study?

The purpose of this study was to determine if the ADI-R and RBS-R are appropriate to assess RRBs in autistic adults. Since RRBs are broad, we focused on how well these instruments measure two categories of RRBs: Repetitive Sensory Motor (RSM) and Insistence on Sameness (IS) behaviors. RSM includes behaviors such as hand flapping and lining up objects. IS includes behaviors such as negative responses to change and adherence to specific routines. Research supports the ADI-R and RBS-R for assessing these categories in children; however, little is known about their use in older samples. This study aimed to address this research gap.

What did the researchers do?

We used existing data from autistic adolescent and adults assessed using ADI-R, RBS-R Caregiver, or RBS-R Self-Report. We assessed the relationships between items on each instrument, whether items were meaningfully related to support categories of RSM and IS, and whether instrument subscales were related to age and nonverbal intelligence.

What were the results of the study?

Results provided support for use of the RBS-R Caregiver and Self-Report, but not the ADI-R, for assessing RSM and IS behaviors in adults. Scores on the RBS-R forms were not strongly related, suggesting that caregiver report and self-report may reflect different behaviors. This could be explained by caregivers not being aware of some behaviors their adult son or daughter exhibit. This would be consistent with previous research suggesting that autistic adults may hide or mask certain behaviors; however, this was not tested in this study.

What do these findings add to what was already known?

Findings demonstrate that the RBS-R Caregiver, commonly used to assess RRBs in children, can also be used to measure RSM and IS behaviors in autistic adolescents and adults. Results also provide initial evidence for the use of the RBS-R Self-Report to assess these behaviors in adults.

What are potential weaknesses in the study?

The sample was small and did not include equal representation of females and nonbinary identities, non-White or non-Hispanic ethnicities, individuals with intellectual disability, or older adults. Also, participants did not have data from all three instruments.

How will these findings help autistic adults now or in the future?

Findings inform clinicians and researchers about methods available to assess RRBs in adults. While further research is needed, especially in samples with more gender and racial/ethnic diversity, these results will contribute to more appropriate assessment of adults in clinical practice and research.

Introduction

Restricted and repetitive behaviors (RRBs) are core features of autism.¹ Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria outline four domains of RRBs encompassing a range of behaviors, including stereotyped or repetitive behaviors, Insistence on Sameness (IS), restricted, fixated interests, and sensory interests and responses. Longitudinal studies suggest that different RRB domains may show differing patterns of onset and developmental trajectories.^2–4 RRBs have been understudied relative to social communication diagnostic criteria,⁵ with little attention to their manifestation in autistic adults. For many autistic individuals, identification of RRBs may inform development of supports or environmental modifications to promote community integration. For example, sensory sensitivities in the workplace may be addressed by dimming of lights or allowing the adult to wear a headset or earbuds to minimize sound inputs. Further research on RRBs is also needed to improve community knowledge and understanding of autism in adults. To advance this area of research, the present study explores the validity of the Autism Diagnostic Interview-Revised (ADI-R⁶) and Repetitive Behavior Scale-Revised (RBS-R⁷) for assessing subtypes of RRBs in adolescents and adults. Novel comparisons of caregiver-reported and self-reported RRBs on the RBS-R are also examined in a small overlapping sample. Much of the previous RRB research has particularly focused on two subtypes or dimensions: Repetitive Sensory Motor (RSM) and IS behaviors. The RSM domain encompasses motor mannerisms, repetitive or stereotyped use of objects, and sensory interests. These behaviors are commonly observed in young children without autism, as well as individuals with intellectual disability.^8,9 Within autistic samples, RSM is usually moderately negatively correlated with nonverbal intelligence quotient (NVIQ).^10–12 IS includes compulsive or ritualistic behaviors and difficulties with changes in routine. In contrast to RSM, studies of IS suggest that this domain is not correlated with age or NVIQ.^10,11 Independence from developmental level makes IS an attractive candidate for studies striving to identify biological mechanisms underlying features of autism.^13–15 For example, structural covariance of gray matter volume (i.e., coupling among brain regions) has been related to the intensity of IS behaviors in autistic adults.¹⁴

Several factor-analytic studies have provided empirical support for these RRB subtypes as measured using the ADI-R⁶ and the RBS-R.⁷ In ADI-R studies, there has been some variation regarding the specific items loading on each factor,^9,11,16–20 with some studies suggesting inclusion of circumscribed interests on the IS factor and others separating interests into an additional third factor.^18,20 In the RBS-R, results have been more varied, with one study reporting RSM and IS factors¹⁸ and others supporting three- to five-factor solutions that included subscales resembling RSM and IS.^21,22 In a large sample of 1825 school-age children, Bishop et al. expanded upon previous work by exploring the validity of these RRB constructs in both the ADI-R and caregiver-reported RBS-R and by conducting cross-measure comparisons.¹¹ Their exploratory factor analysis supported a two-factor RSM and IS solution for the ADI-R and a five-factor RBS-R solution that was very similar to previous research.²¹ The RBS-R Sensory Motor and a Ritualistic/Sameness subscales (which approximated RSM and IS, respectively) showed good convergent validity with their respective ADI-R factors, and subscales from both instruments exhibited comparable associations with age and NVIQ. Consistent with previous literature, both ADI-R and RBS-R RSM subscales were negatively correlated with NVIQ, whereas neither instrument's IS subscale was meaningfully associated with NVIQ.

While studies supporting the use of the ADI-R and RBS-R to assess RSM and IS behaviors have sometimes included adults, the average age of the samples generally fell in the school-age range.^16,20,22,23 There is evidence, however, of age-related changes in RRBs measured by both instruments. In a sample of autistic children and adults (M_age = 22 years, SD_age = 10 years) followed over a 4.5-year period, Shattuck et al. reported decreases in caregiver endorsement of RRBs on the ADI-R.²⁴ In a cross-sectional study of autistic children and adults, adults had lower levels of caregiver-reported RSM and IS behaviors on the RBS-R compared with children.²³ Similarly, in a sample of 34 autistic adults, current caregiver-reported RSM and IS levels were lower than lifetime ratings (i.e., whether a behavior had ever been present) on both the ADI-R and RBS-R.²⁵ There is also evidence that change over time in RRB may differ by domain; in one longitudinal study, RSM behaviors remained relatively high from 2 to 9 years of age, whereas IS behaviors started low and increased over time.²⁶ Although the extant literature provides support for the use of the ADI-R and RBS-R to assess these subdomains of RRBs in children, the validity of these measures for assessing RSM and IS behaviors in older samples has not yet been established. This is an important next step to foster exploration of possible age-related differences in RRB subtypes.

It is also important to note that the studies described above have focused on caregiver reports of RRBs. To the best of our knowledge, only one study has investigated the validity of these constructs based on self-report. Conducting a principal component analysis of the self-report Adult Repetitive Behavior Questionnaire-2, Barrett et al.²⁷ identified a two-dimensional structure, with components corresponding to RSM and IS behaviors. The RSM component was negatively correlated with age, suggesting a decrease in these behaviors as individuals grew older. NVIQ estimates were not available due to the online nature of the study. In a recent qualitative study, many autistic adults reported that some repetitive rhythmic behaviors were used as self-regulatory mechanisms in response to sensory overload, dysregulated or excessive thoughts, or experiencing an abundance of emotion that could not be contained.²⁸ Insights into the different causes of these behaviors highlight the importance of using self-reports to assess RRBs. Self-reports will likely capture nuanced differences in the nature of the behavior and/or the cause of the behavior itself that may be overlooked by caregivers, who may perceive a specific behavior as being the same or similarly motivated across contexts.

The focus of the present study is to explore the factor and construct validity of the ADI-R and RBS-R in a convenience sample of autistic adolescents and adults aged 16–46 years. Specifically, we tested the two-factor RSM and IS structure, which has been theoretically and empirically supported in primarily child samples. In addition to assessing the validity of caregiver report instruments, this study extends previous work that has focused exclusively on caregiver report by assessing the factor and construct validity of the RBS-R Self-Report form. We hypothesized that confirmatory factor analyses (CFA) would provide support for the two-factor structure in all three instruments (ADI-R, RBS-R Caregiver, and RBS-R Self-Report). We also hypothesized that subscales assessing the same construct would be correlated across instruments and that RSM, but not IS, would be negatively associated with NVIQ.

Methods

Participants and procedures

In total, 149 participants were drawn from existing samples of autistic adolescents and adults recruited to studies conducted at specialty autism clinics in Michigan, New York, California, and Tennessee (details regarding recruitment are provided in Appendix A1 and Appendix Table A1). An additional 144 adolescents aged 16–17 years were drawn from the Simons Simplex Collection²⁹; 85 of these participants were also included in a previous factor-analytic study of the RRB.¹¹ The primary analyses included these participants; sensitivity analyses excluding these participants yielded highly similar results (data available upon request). The final sample comprised 293 participants (M_age = 19.89 years, SD_age = 4.88 years; range = 16.00–45.80 years). The sample was primarily male (75.5%), White (87.4%), non-Hispanic (91.8%), and had NVIQ above 70 (76.1%).

All individuals participated in a direct assessment, which included cognitive testing and the Autism Diagnostic Observation Schedule (ADOS³⁰). In addition, parents or caregivers were interviewed using the ADI-R (n = 219) and completed the RBS-R (n = 241). A subset of adults completed the RBS-R Self-Report (n = 57). ADOS and ADI-R were administered by research reliable staff. The mean age and NVIQ for participants who completed each measure and the percentage of measure overlap are represented in Table 1. All participants received best estimate clinical Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR) or DSM-5 diagnoses of an autism spectrum disorder based on all available information made by a supervising psychologist or an advanced graduate student. Procedures were in accordance with the requirements of the institutional review board at each university.

Table 1.

Sample Size, Age, and Nonverbal Intelligence Quotient for Each Instrument

	Total sample	ADI-R	RBS-R-C	RBS-R-SR
n	293	219	241	57
% Male	76	82	80	59
M Age (SD)	19.89 (4.88)	18.30 (3.02)b	18.79 (3.42)c	24.69 (6.57)b,c
% <18 years	56	75	66	0
M NVIQ (SD)	87.42 (29.83)	85.45 (29.63)b	83.72 (30.83)c	102.65 (18.48)b,c
% ADI-Ra	75	100	81	0
% RBS-R-Ca	82	89	100	52
% RBS-R-SRa	20	0	12	100

^aPercent indicates proportion of participants within that column who were also administered the instrument noted for that row.

^bADI-R and ^cRBS-R-C participants were younger and had lower NVIQ than RBS-R-SR participants (p ≤ 0.001).

ADI-R, Autism Diagnostic Interview-Revised; NVIQ, nonverbal intelligence quotient; RBS-R-C, Repetitive Behavior Scale-Revised Caregiver Report; RBS-R-SR, Repetitive Behavior Scale-Revised Self-Report; SD, standard deviation.

Measures

Autism Diagnostic Interview-Revised

The ADI-R⁶ is a semi-structured, standardized diagnostic interview conducted with a caregiver. Ninety-three items are scored on a 0 (criteria not present) to 3 (criteria met to a degree that interferes with daily function). In keeping with Bishop et al.,¹¹ “current” behavior items were used to assess RSM (5 items) and IS (6 items; Table 2).

Table 2.

Within-Domain Item Correlations and [95% Confidence Interval] for the Autism Diagnostic Interview-Revised (n = 219)

ADI-R itema	RSM					IS
	67.	69.	71.	77.	78.	70.	72.	73.	74.	75.	68.
RSM
67. Unusual preoccupations	—	0.17 [0.03 to 0.31]	0.22 [0.08 to 0.35]	0.12 [−0.02 to 0.25]	0.10 [−0.03 to 0.33]	0.04 [−0.09 to 0.18]	0.10 [−0.05 to 0.22]	0.11 [−0.03 to 0.24]	0.09 [−0.04 to 0.23]	0.11 [−0.03 to 0.25]	0.04 [−0.11 to 0.19]
69. Repetitive use of objects		—	0.29 [0.17 to 0.42]	0.28 [0.15 to 0.41]	0.12 [−0.02 to 0.26]	0.13 [0.00 to 0.27]	0.14 [−0.01 to 0.26]	0.02 [−0.12 to 0.16]	0.15 [0.02 to 0.29]	0.09 [−0.04 to 0.22]	0.07 [−0.06 to 0.20]
71. Unusual sensory interests			—	0.29 [0.15 to 0.42]	0.25 [0.13 to 0.38]	0.19 [0.05 to 0.32]	0.20 [0.07 to 0.32]	0.16 [0.02 to 0.28]	0.21 [0.08 to 0.34]	0.19 [0.04 to 0.32]	0.18 [0.05 to 0.31]
77. Hand and finger manners				—	0.14 [0.00 to 0.29]	0.13 [−0.01 to 0.26]	0.10 [−0.03 to 0.23]	0.10 [−0.03 to 0.23]	0.10 [−0.04 to 0.24]	0.07 [−0.06 to 0.21]	0.07 [−0.06 to 0.20]
78. Other complex manners					—	0.05 [−0.09 to 0.19]	−0.02 [−0.15 to 0.11]	0.14 [0.01 to 0.28]	0.11 [−0.03 to 0.24]	0.05 [−0.09 to 0.19]	0.00 [−0.15 to 0.13]
IS
70. Compulsions/rituals						—	0.16 [0.02 to 0.29]	0.23 [0.10 to 0.35]	0.33 [0.20 to 0.45]	0.27 [0.13 to 0.40]	−0.01 [−0.16 to 0.11]
72. Sensitivity to noise							—	0.24 [0.11 to 0.36]	0.23 [0.10 to 0.36]	0.12 [−0.02 to 0.26]	0.15 [0.02 to 0.28]
73. Abnormal/idiosyncratic response								—	0.20 [0.07 to 0.34]	0.25 [0.11 to 0.38]	0.12 [−0.02 to 0.25]
74. Difficulties with change									—	0.34 [0.23 to 0.44]	0.12 [−0.02 to 0.25]
75. Resistance to change										—	0.06 [−0.07 to 0.20]
68. Circumscribed interests											—

^aItem number as it appears on the ADI-R instrument.

IS, Insistence on Sameness; RSM, Repetitive Sensory Motor.

Repetitive Behavior Scale-Revised

The RBS-R⁷ is a self-report or caregiver-completed questionnaire designed to assess RRBs. Forty-three items are scored on a 0 (behavior does not occur) to 3 (behavior is a severe problem) scale. The RBS-R includes six subscales⁷; however, the present study used the Sensory Motor (7 items) and Ritualistic/Sameness (11 items) factors reported by Bishop et al.¹¹ to reflect RSM and IS, respectively, for both caregiver (Table 3) and self-report in this study (Table 4).

Table 3.

Within-Domain Item Correlations and [95% Confidence Interval] for the Repetitive Behavior Scale-Revised Caregiver Report (n = 241)

RBS-R itemsa	RSM
	2.	3.	4.	5.	6.	43.
RSM
1. Whole body	0.41 [0.26 to 0.54]	0.50 [0.39 to 0.61]	0.38 [0.24 to 0.51]	0.21 [0.07 to 0.34]	0.37 [0.26 to 0.49]	0.28 [0.15 to 0.43]
2. Head	—	0.36 [0.23 to 0.48]	0.23 [0.08 to 0.38]	0.14 [−0.01 to 0.30]	0.36 [0.23 to 0.48]	0.13 [−0.01 to 0.27]
3. Head/finger		—	0.44 [0.31 to 0.57]	0.25 [0.10 to 0.37]	0.37 [0.24 to 0.49]	0.21 [0.07 to 0.34]
4. Locomotion			—	0.19 [0.05 to 0.33]	0.35 [0.31 to 0.46]	0.21 [0.05 to 0.34]
5. Object usage				—	0.27 [0.12 to 0.39]	0.35 [0.20 to 0.48]
6. Sensory					—	0.30 [0.17 to 0.42]
43. Fascination with movement						—

	IS
	27.	30.	31.	32.	33.	34.	35.	37.	38.	39.
IS
26. Travel/transportation	0.54 [0.41 to 0.66]	0.42 [0.30 to 0.54]	0.42 [0.29 to 0.53]	0.29 [0.12 to 0.43]	0.53 [0.41 to 0.65]	0.43 [0.30 to 0.56]	0.39 [0.25 to 0.50]	0.47 [0.36 to 0.57]	0.52 [0.40 to 0.63]	0.58 [0.46 to 0.69]
27. Play/leisure	—	0.38 [0.25 to 0.51]	0.39 [0.28 to 0.49]	0.33 [0.17 to 0.47]	0.35 [0.21 to 0.49]	0.40 [0.27 to 0.53]	0.34 [0.17 to 0.48]	0.42 [0.30 to 0.53]	0.48 [0.30 to 0.59]	0.52 [0.39 to 0.64]
30. Objects to visiting new places		—	0.43 [0.30 to 0.55]	0.11 [−0.04 to 0.26]	0.36 [0.22 to 0.49]	0.46 [0.33 to 0.58]	0.38 [0.25 to 0.50]	0.51 [0.41 to 0.62]	0.42 [0.28 to 0.54]	0.33 [0.20 to 0.45]
31. Becomes upset if interrupted			—	0.28 [0.16 to 0.39]	0.42 [0.29 to 0.53]	0.44 [0.33 to 0.55]	0.32 [0.20 to 0.42]	0.64 [0.55 to 0.73]	0.45 [0.34 to 0.56]	0.45 [0.35 to 0.56]
32. Insists on walking in pattern				—	0.42 [0.28 to 0.54]	0.22 [0.06 to 0.36]	0.32 [0.12 to 0.50]	0.26 [0.13 to 0.38]	0.34 [0.22 to 0.46]	0.30 [0.16 to 0.43]
33. Insists on sitting at the same place					—	0.49 [0.35 to 0.61]	0.55 [0.45 to 0.65]	0.47 [0.36 to 0.58]	0.49 [0.37 to 0.60]	0.47 [0.33 to 0.59]
34. Dislikes changes in other people						—	0.42 [0.27 to 0.57]	0.48 [0.37 to 0.58]	0.50 [0.39 to 0.61]	0.43 [0.30 to 0.56]
35. Insists on using a particular door							—	0.37 [0.27 to 0.46]	0.35 [0.22 to 0.47]	0.28 [0.12 to 0.41]
37. Difficulty with transitions								—	0.59 [0.49 to 0.69]	0.53 [0.43 to 0.63]
38. Insists on same routine									—	0.71 [0.62 to 0.79]
39. Insists things at specific times										—

^aItem number as it appears on the RBS-R instrument.

Table 4.

Within-Domain Item Correlations and [95% Confidence Interval] for the Repetitive Behavior Scale-Revised Self-Report (n = 57)

RBS-R itemsa	RSM
	2.	3.	4.	5.	6.	43.
RSM
1. Whole body	0.57 [0.33 to 0.76]	0.29 [0.04 to 0.52]	0.26 [−0.01 to 0.51]	0.29 [0.04 to 0.55]	0.56 [0.36 to 0.75]	0.34 [0.08 to 0.57]
2. Head	—	0.53 [0.31 to 0.73]	0.39 [0.04 to 0.61]	0.39 [0.13 to 0.60]	0.59 [0.39 to 0.75]	0.34 [0.05 to 0.58]
3. Head/finger		—	0.11 [−0.16 to 0.38]	0.57 [0.31 to 0.76]	0.47 [0.22 to 0.66]	0.38 [0.15 to 0.61]
4. Locomotion			—	0.09 [−0.18 to 0.32]	0.23 [−0.01 to 0.46]	0.09 [−0.19 to 0.38]
5. Object usage				—	0.52 [0.27 to 0.71]	0.42 [0.17 to 0.64]
6. Sensory					—	0.47 [0.24 to 0.66]
43. Fascination with movement						—

	IS
	27.	30.	31.	32.	33.	34.	35.	37.	38.	39.
IS
26. Travel/transportation	0.30 [0.02 to 0.55]	0.39 [0.12 to 0.55]	0.06 [−0.20 to 0.33]	0.39 [0.15 to 0.58]	0.39 [0.10 to 0.65]	0.34 [0.10 to 0.59]	0.41 [0.16 to 0.63]	0.27 [−0.03 to 0.54]	0.30 [0.05 to 0.53]	0.19 [−0.06 to 0.44]
27. Play/leisure	—	0.36 [0.12 to 0.62]	0.34 [0.07 to 0.58]	0.36 [0.09 to 0.57]	0.33 [0.05 to 0.59]	0.52 [0.28 to 0.73]	0.42 [0.15 to 0.64]	0.44 [0.17 to 0.67]	0.34 [0.08 to 0.56]	0.60 [0.40 to 0.77]
30. Objects to visiting new places		—	0.25 [0.01 to 0.50]	0.47 [0.22 to 0.65]	0.26 [−0.02 to 0.56]	0.19 [−0.07 to 0.46]	0.36 [0.10 to 0.58]	0.44 [0.17 to 0.69]	0.35 [0.12 to 0.55]	0.44 [0.22 to 0.63]
31. Becomes upset if interrupted			—	0.22 [−0.05 to 0.46]	0.29 [0.01 to 0.56]	0.43 [0.16 to 0.64]	0.43 [0.17 to 0.64]	0.52 [0.26 to 0.75]	0.53 [0.32 to 0.71]	0.42 [0.18 to 0.62]
32. Insists on walking in pattern				—	0.31 [0.06 to 0.52]	0.51 [0.27 to 0.70]	0.62 [0.39 to 0.78]	0.30 [0.05 to 0.51]	0.51 [0.28 to 0.69]	0.53 [0.31 to 0.70]
33. Insists on sitting at the same place					—	0.51 [0.29 to 0.72]	0.48 [0.22 to 0.68]	0.66 [0.40 to 0.87]	0.45 [0.22 to 0.63]	0.26 [0.01 to 0.48]
34. Dislikes changes in other people						—	0.61 [0.41 to 0.77]	0.52 [0.32 to 0.71]	0.55 [0.33 to 0.74]	0.60 [0.41 to 0.77]
35. Insists on using a particular door							—	0.41 [0.14 to 0.64]	0.57 [0.37 to 0.74]	0.42 [0.16 to 0.65]
37. Difficulty with transitions								—	0.52 [0.31 to 0.69]	0.46 [0.27 to 0.65]
38. Insists on same routine									—	0.55 [0.32 to 0.72]
39. Insists things at specific times										—

^aItem number as it appears on the RBS-R instrument.

Nonverbal intelligence quotient

All participants had NVIQ scores derived from a hierarchy of cognitive tests,³¹ most often the Differential Ability Scales, Second Edition (DAS-II³²; 50%), or the Wechsler Abbreviated Scale of Intelligence (WASI³³; 41%). The remaining 9% were assessed using the Mullen Scales of Early Learning,³⁴ Wechsler Intelligence Scale for Children, Fourth Edition,³⁵ the Wechsler Adult Intelligence Scale, Fourth Edition,³⁶ or the Stanford–Binet Intelligence Scales, Fifth Edition.³⁷ For participants who were not able to complete the test standardized for their age, ratio IQs were computed by averaging age equivalents for the nonverbal subtests and dividing by chronological age.³⁸

Analysis

Exploratory data analysis included the review of within-instrument item Spearman's correlations to provide evidence that at least one shared construct is measured by the items. CFA with weighted least square mean value estimator (WLSMV) were conducted. Root mean square error of approximation (RMSEA) <0.08 and comparative fit index (CFI) >0.9 were used as indicators of good fit.^39,40 Internal consistency was measured by Cronbach's alpha for each subscale; α > 0.7 was considered acceptable.⁴¹ Subscale totals were calculated (per convention, scores of 2 and 3 were collapsed for the ADI-R), and Spearman's correlations (with bootstrapped 95% confidence intervals [CIs]) were used to examine relationships between instrument domains and NVIQ and age. Factor analyses were performed in Mplus Version 8⁴² using WLSMV estimation to account for ordinality of items; all other analyses were conducted using SPSS version 25.⁴³ Following the current recommendations of the American Statistical Association,⁴⁴ we eschew the use of “statistical significance” and instead present 95% CIs and/or exact p-values.

Results

Exploratory data analysis: within-instrument Spearman's correlation matrices

Within-domain item correlations for each instrument are reported in Tables 2–4. For the ADI-R, item correlations were modest, RSM correlations ranged from 0.14 to 0.29 (median r_s = 0.25), and IS correlations ranged from 0.14 to 0.33 (median r_s = 0.22). The RBS-R Caregiver item correlations were higher; RSM correlations ranged from 0.13 to 0.50 (median r_s = 0.31) and IS correlations ranged from 0.22 to 0.71 (median r_s = 0.43). RBS-R Self-Report item correlations were comparable to the Caregiver form (RSM range r_s = 0.29–0.59, median r_s = 0.48; IS range r_s = 0.26–0.66, median r_s = 0.44).

Validity of two-factor structure: ADI-R

The results of the exploratory data analysis provided weak evidence in favor of the utility of a factor-analytic model for the ADI-R. Accordingly, internal consistency for each putative subscale was very low (α_RSM = 0.53; α_IS = 0.56). Overall, these results indicated that there was very little covariation for a latent factor structure to explain. Thus, we did not expect the CFA to be informative; indeed, the results of the CFA indicated that the item correlations were too low to test the model (the chi-square was less than the degrees of freedom of the model, which produces uninterpretable fit statistics). Because it is not possible to evaluate the fit of this model to the data, we do not include them in the main text, but they are presented in Appendix Table A2. Given that the factor validity of the ADI-R was not supported, external validity analyses were not conducted.

Validity of two-factor structure: RBS-R Caregiver

The RBS-R Caregiver CFA provided support for the two-factor model tested (RMSEA = 0.07 [90% CI 0.05–0.07]; CFI = 0.96). Table 5 shows the frequency of caregiver-reported endorsement for each RBS-R item and CFA results. Internal consistency of the RBS-R Caregiver subscales was α_RSM = 0.76 and α_IS = 0.91. The correlation between RSM and IS subscales was r_s = 0.30 [95% CI 0.18–0.42].

Table 5.

Confirmatory Factor Analysis for Repetitive Behavior Scale-Revised Caregiver Report (n = 241)

RBS-R items	% Endorseda	RBS-R-C RSM	SE	RBS-R-C IS	SE
1. Whole body	28	0.77	0.07
2. Head	22	0.53	0.09
3. Hand/finger	44	0.70	0.07
4. Locomotion	25	0.70	0.08
5. Object usage	29	0.57	0.09
6. Sensory	50	0.70	0.06
43. Fascination with movement	18	0.81	0.09
26. Travel/transportation	32			0.83	0.03
27. Play/leisure	28			0.77	0.05
30. Objects to visiting new places	36			0.65	0.06
31. Becomes upset if interrupted	70			0.78	0.04
32. Insists on walking in pattern	14			0.67	0.07
33. Insists sitting same place	30			0.81	0.04
34. Dislikes changes in others	24			0.80	0.04
35. Insists using particular door	10			0.83	0.06
37. Difficulty with transitions	70			0.84	0.03
38. Insists on same routine	49			0.88	0.03
39. Insists on specific times	36			0.88	0.03

χ²(134) = 238.66; RMSEA = 0.07; CFI = 0.96.

^aProportion with nonzero scores, reflecting endorsement that the behavior occurs to some extent.

CFI, comparative fit index; RMSEA, root mean square error of approximation; SE, standard error.

Validity of two-factor structure: RBS-R Self-Report

CFA for the RBS-R Self-Report also supported the two-factor model (RMSEA = 0.05 [90% CI 0.00–0.08]; CFI = 0.98). Table 6 shows the frequency of self-reported endorsement for each RBS-R item and CFA results. The internal consistency of the RBS-R Self-Report was α_RSM = 0.80 and α_IS = 0.90 (IS). The correlation between subscales was r_s = 0.63 [95% CI 0.44–0.77].

Table 6.

Confirmatory Factor Analysis for Repetitive Behavior Scale-Revised Self-Report (n = 57)

RBS-R items	% Endorseda	RBS-R-SR RSM	SE	RBS-R-SR IS	SE
1. Whole body	45	0.68	0.19
2. Head	34	0.83	0.08
3. Head/finger	48	0.81	0.11
4. Locomotion	17	0.48	0.18
5. Object usage	48	0.71	0.10
6. Sensory	55	0.93	0.09
43. Fascination with movement	34	0.78	0.11
26. Travel/transportation	34			0.78	0.10
27. Play/leisure	25			0.70	0.11
30. Objects to visiting new places	31			0.62	0.11
31. Becomes upset if interrupted	65			0.77	0.07
32. Insists on walking in pattern	22			0.83	0.09
33. Insists sitting same place	47			0.88	0.07
34. Dislikes changes in others	41			0.84	0.06
35. Insists using particular door	21			0.90	0.06
37. Difficulty with transitions	62			0.85	0.08
38. Insists on same routine	53			0.79	0.08
39. Insists on specific times	33			0.76	0.10

χ²(134) = 151.35; RMSEA = 0.05; CFI = 0.98.

^aProportion with nonzero scores, reflecting endorsement that the behavior occurs to some extent.

External validity of the RBS-R and relationships with NVIQ and age

Both the RBS-R Caregiver subscales were negatively correlated with NVIQ; however, the correlation with RSM was stronger (r_s = −0.42 [95% CI −0.52 to −0.31]) than the correlation with IS (r_s = −0.17 [95% CI −0.28 to −0.04]. Neither RBS-R Self-Report subscale was meaningfully related to NVIQ (RSM r_s = −0.11 [95% CI −0.41 to 0.17]; IS r_s = −0.11 [95% CI −0.36 to 0.15]). While self-reported RBS-R IS was modestly positively correlated with age (r_s = 0.21 [95% CI −0.07 to 0.47]), caregiver report of IS behaviors was modestly negatively correlated with age (r_s = −0.18 [95% CI −0.31 to −0.05]). Age was not meaningfully related to RSM on either form (Caregiver r_s = 0.01 [95% CI −0.14 to 0.13], Self-Report r_s = 0.14 [95% CI −0.15 to 0.44]).

In the subset of people with both caregiver-reported and self-reported RBS-R, the within-domain correlations (i.e., RSM with RSM and IS with IS) were positive. However, while the RSM correlation was moderate (r_s = 0.39 [95% CI 0.05 to 0.68]), the IS correlation was weak (r_s = 0.12 [95% CI −0.22 to 0.49]). As shown in Figure 1, in the subset of participants with both self-report and caregiver report (n = 30), participants tended to self-report higher levels of RSM [d = 0.40, t(29) = −1.91, p = 0.07] and IS [d = 0.19, t(29) = 0.82, p = 0.41], relative to caregiver reports.

FIG. 1.

Comparison of self-reported and caregiver-reported RBS-R for overlapping sample (n = 30). Bars represent ±1 standard error; ^#p < 0.10. IS, Insistence on Sameness; RBS-R, Repetitive Behavior Scale-Revised; RBS-R-C, Repetitive Behavior Scale-Revised Caregiver Report; RBS-R-SR, Repetitive Behavior Scale-Revised Self-Report; RSM, Repetitive Sensory Motor; IS, Insistence on Sameness.

Discussion

The results of this study provide support for the validity of the RSM and IS subscales of the RBS-R Caregiver Report in adolescents and adults, extending previous literature primarily conducted in child samples.^12,22,45 Results also provide initial evidence for the use of the RBS-R Self-Report to assess these behavioral domains. However, among participants with both caregiver and self-report forms, RSM scores were only moderately correlated and IS scores were not meaningfully related—suggesting that although the factor validity may be supported, the constructs measured by the scales might differ.

Consistent with previous studies of school-aged children,^11,12 caregiver-reported RSM behaviors (i.e., Sensory Motor subscale) were more strongly and negatively correlated with NVIQ than IS behaviors (i.e., Ritualistic/Sameness subscale). This supports the construct validity of the caregiver RSM because RSM behaviors are observed in younger children and expected to be more prevalent among people with lower IQs. However, neither self-reported RSM nor IS was meaningfully correlated with NVIQ. The lack of correlation between NVIQ and RSM based on RBS-R Self-Report might be attributable to characteristics of this self-report sample (i.e., fewer participants, older individuals with a higher/more restricted NVIQ range) compared with the sample with caregiver report. Autistic adults may also interpret items differently than caregivers. The present findings warrant further investigation in larger samples to better understand factors affecting differences in self-report and caregiver report.

The present data did not provide support for the use of the ADI-R to assess these RRB constructs in older individuals. This was a somewhat surprising finding, considering that the ADI-R has historically been the most commonly measure used in this area of research with children.^9,11,16–20 Consistent with previous studies of adults, endorsement of some items was somewhat lower than previously reported (Appendix Table A2), but overall quite comparable to levels of endorsement in children.¹¹ However, the extremely low item correlations that were observed in this sample suggest that the ADI-R RRB items are not indicators of some shared latent construct(s). These results suggest that the current ADI-R may not be an appropriate instrument to measure these constructs in older individuals; replication in larger adult samples will be needed to inform future research. Especially considering the fact that among the same participants, we found support for the factorial and construct validity of RBS-R RSM and IS subscales, the ADI-R results should not be interpreted as evidence that these constructs are not empirically meaningful in adolescents and adults. Instead, this may serve to underscore the need for additional research to understand the manifestation and assessment of RRBs in adults.^23,25 Indeed, relatively modest ADI-R and RBS-R correlations in the study by Bishop et al. serve as a reminder that ADI-R items were designed to capture a broad range of behaviors corresponding to diagnostic criteria, whereas the RBS-R was intended to characterize different domains of repetitive behaviors using items assessing more specific aspects of behavior.¹¹ Because the ADI-R was developed to primarily assess children, it is possible that the items do not fully capture how these types of RRBs manifest in adults.

A novel aspect of this study is the consideration of RSM and IS measured by self-report. While some empirical support for these constructs in autistic adults has been demonstrated in a study using the self-report Adult Repetitive Behavior Questionnaire-2,²⁷ few studies have focused on self-reported measurement of RRBs. The results of our study suggested that although the two-factor conceptualization appeared valid, the RSM and IS subscales appeared to be measuring substantially different things across self-report and caregiver report (as indicated by low within-domain correlations across methods). Future research in larger samples should be conducted to assess the measurement invariance of the RBS-R between caregiver and self-report, to determine whether the same constructs are measured in the same way when respondents differ. While the interpretation of these results must be tempered by awareness of the small size and relatively high cognitive ability of this sample, they are consistent with descriptions of camouflaging, which may reflect adults' conscious modification or minimization of behaviors in contexts when perceived as inappropriate to a given situation or to “blend in” with others.⁴⁶ In other words, although most RSM behaviors are observable (e.g., body rocking, hand mannerisms), higher self-reported levels of these behaviors may reflect adults reporting their engagement in these acts when out of view of others (their caregivers specifically in this context). Another consideration is whether higher scores reflect a view that these RRBs are more “problematic.” On the RBS-R, ratings of 1, 2, and 3 indicate that the behavior occurs and is a mild, moderate, or severe problem, respectively. It is possible that individuals view their RRBs as problematic because others have had negative reactions when they display these behaviors or have told them to stop.²⁸ Alternatively, they may be experienced as more problematic if they feel they must camouflage these behaviors to avoid feeling marginalized.²⁸ These response categories may also promote underreporting if an adult views their behavior as a positive attribute, rather than a challenge. More information is needed on adults' perspectives of their RRBs to inform interpretation of the RBS-R. Cognitive interviewing,⁴⁷ a standard approach used in questionnaire development, may be useful to assess comprehensibility, ambiguity, and relevance of items, as well as the decision processes the individual goes through when endorsing different behaviors. Focus groups and interviews with autistic individuals will be important to inform development of new instruments. Using these approaches, Kapp et al.²⁸ found that some autistic individuals reported certain RRBs as mechanisms to self-regulate when feeling overwhelmed. Insights gained from their perspectives highlight a need to consider whether response choices on future instruments should capture another feature, such as frequency, rather than emphasizing behaviors as problematic.

While much is to be learned about the manifestation and measurement of RRBs, differences in reports highlight the importance of self-report where possible to capture a fuller picture of RRBs in adults. Discrepancies between self-report and caregiver report could provide important insights into individual experiences and may even serve as a red flag for concerns considering the possible relationship between features of depression and camouflaging.⁴⁸ Previous studies have also demonstrated a relationship between depressive features and self-reported IS behaviors on the RBS-R,⁴⁹ further highlighting the significance of assessing RRBs using self-report.

Limitations

The primary limitation of this study was its use of a small convenience sample of autistic adolescents and adult research referrals. While these may not be representative of clinical samples or the larger population of autistic adults, the studies for which they were recruited varied in focus and inclusion criteria and were conducted in multiple states, providing some protection against consistent ascertainment bias. Nonetheless, the sample was primarily composed of White, non-Hispanic males with NVIQs above 70. This last feature is especially important, as our construct validity analyses were conducted in the context of a somewhat restricted range of IQ. Although this study included adults up to 45 years of age, the sample was primarily composed of older adolescents and young adults (88% were 16–25 years old) because the studies providing these data did not recruit adults older than 45 years. It is important to acknowledge that the limited diversity with respect to gender, race, ethnicity, and age in our sample limits the generalizability of findings. It is hoped that this study represents a first step toward better understanding RRBs in adults. It will be critical for future studies to explore whether there are differences in the manifestation of RRBs in other subgroups.

The present study utilized a modestly sized sample, particularly in comparison to studies of children such as that by Bishop et al.¹¹ In addition, the lack of overlap across measures, particularly for the RBS-R Caregiver and Self-Report forms, makes it more difficult to interpret differential associations between subscales on each instrument and with NVIQ. Moreover, the sample size was insufficient to conduct CFA on the broader RBS-R (i.e., the five subscales identified in previous analyses) or exploratory analysis of a broader ADI-R RRB item set to explore whether other items may better capture constructs of RSM or IS in older individuals. It was also too small to test the invariance of items (i.e., whether they were measuring the same thing) on either instrument between older adolescents and adults. Future studies using larger data sets are needed to further advance our understanding of RRBs in adults.

Conclusions

To the best of our knowledge, the present study is the first to investigate the validity of RSM and IS measured by the ADI-R and RBS-R in autistic adolescents and adults. These results suggest the ADI-R may not be a good measure of these RRB subcategories in adolescents and adults. Support was found, however, for the use of the RBS-R Caregiver and Self-Report, although more research is needed to establish the construct validity of these scales. Findings also add to a growing body of literature suggesting that IS behaviors may be relatively independent of developmental level. These conclusions must be interpreted cautiously in light of the sample limitations noted above, particularly limited gender and racial/ethnic diversity, and await replication in larger more diverse samples. Longitudinal studies following children through adulthood will be essential to advancing understanding of the manifestation of RRBs in later life.

Appendix

Appendix A1. Sample Ascertainment

Data were drawn from existing studies. The specific breakdown by instrument and study is provided in Appendix Table A1 and described below.

Appendix Table A1.

Sample Distribution for Each Instrument by Study

Study name	Site location	ADI-R	RBS-R-C	RBS-R-SR
		n = 219	n = 241	n = 57
Adult Diagnosis	Michigan/New York	50	50	0
Limited Language	California	0	13	0
Well-Being	Michigan	25	0	0
Adult Well-Being	Tennessee	0	34	57
Simons Simplex Collection	Across North America	144	144	0

ADI-R, Autism Diagnostic Interview-Revised; RBS-R-C, Repetitive Behavior Scale-Revised Caregiver Report; RBS-R-SR, Repetitive Behavior Scale-Revised Self-Report.

Appendix Table A2.

Confirmatory Factor Analysis for Autism Diagnostic Interview-Revised (n = 219)

ADI-R items	% Endorseda	RSM	SE	IS	SE
Unusual preoccupations	20	0.47	0.12
Repetitive use of objects	36	0.55	0.10
Complex mannerisms	31	0.36	0.10
Hand and finger mannerisms	42	0.49	0.10
Unusual sensory interests	46	0.81	0.11
Compulsions/rituals	47			0.60	0.09
Sensitivity to noise	52			0.45	0.09
Abnormal/idiosyncratic response	55			0.49	0.08
Difficulties with change	67			0.67	0.07
Resistance to change	20			0.69	0.09
Circumscribed interests	75			0.26	0.10

^aProportion with nonzero scores, reflecting endorsement that the behavior occurs to some extent.

IS, Insistence on Sameness; RSM, Repetitive Sensory Motor; SE, standard error.

Adults were recruited to the following research studies conducted at specialty autism centers:

• (1)The Development and Refinement of Diagnostic Instruments for Use with Adults (Adult Diagnosis) study was focused on improving the Autism Diagnostic Observation Schedule (ADOS) for use with adults. Specifically, the project aimed to (i) revise the ADOS Module 4 algorithm (used to assess verbally fluent adults) to be consistent with revisions to Modules 1–3 used with children and (ii) adapt the ADOS Modules 1 and 2 to be appropriate for use with adults with limited language skills. This study recruited adolescents and adults aged 15–30 years across the ability range at two sites (University of Michigan Autism and Communication Disorders Center in Ann Arbor, Michigan, and Center for Autism and the Developing Brain in White Plains, New York). Fifty participants from this study contributed data from the Autism Diagnostic Interview-Revised (ADI-R) and Repetitive Behavior Scale-Revised Caregiver Report (RBS-R-C).
• (2)The Strengths and Challenges of Adults with Limited Language Study was focused on characterizing the behavioral profiles of minimally verbal adults with autism spectrum disorder (ASD) and contribute to the validation of the Adapted ADOS Modules 1 and 2. This study recruited minimally verbal adults (i.e., individuals 18+ years who were nonverbal or used only single words or phrases to communicate verbally) from the Bay Area in California. Thirteen participants from this study contributed RBS-R-C data.
• (3)The Well-Being in ASD Study investigated associations between depressive symptomatology, participant insight into ASD symptoms and social motivation and participation. Autistic adolescents and adults aged 16–35 years with intelligence quotient (IQ) above 70 and at least fifth-grade reading level were recruited to the University of Michigan Autism and Communication Disorders Center in Ann Arbor, Michigan. Twenty-five participants from this study contributed data from the ADI-R.
• (4)The Adult Well-Being study sought to identify candidate mechanisms that contributed to depressed mood in people with ASD. Autistic adults, aged 18–35 years, with IQ above 70 and at least a fifth-grade reading level were recruited to the Vanderbilt University in Nashville, Tennessee. Thirty-four participants from this study contributed data from the RBS-R-C and 57 participants contributed data from the Repetitive Behavior Scale-Revised Self Report (RBS-R-SR) (n = 30 had both).
• (5)The Simons Simplex Collection (SSC) was a study of families with only one child with ASD focused on identifying genetic variants associated with ASD. One hundred forty-four adolescents (aged 16–17 years) who were recruited from 12 sites across North America contributed ADI-R and RBS-R-C data.

All studies excluded participants with acute psychiatric disorders (e.g., bipolar disorder, schizophrenia) and individuals with significant sensory impairments (deafness, blindness) that would impede standardized testing. The SSC study and both Well-Being studies also excluded individuals with previously identified genetic syndromes.

Authorship Confirmation Statement

The authors confirm contribution to the article as follows: all authors were involved in the study conception and design. V.H.B. and K.O.G. each contributed data for analysis. V.H.B., C.R.M., and C.F. conducted analysis and interpretation of results. All authors critically revised the article for intellectual content and approved the final version of the article. The article has been submitted solely to this journal and is not published, in press, or submitted elsewhere.

Author Disclosure Statement

K.O.G. receives royalties from the publisher of the Autism Diagnostic Observation Schedule (ADOS-2). This measure was used for confirming autism spectrum diagnosis in this sample and was not primary to outcome in this article. K.O.G. donated to charity all royalties from the Vanderbilt University Medical Center, from which a subsample of this reported sample issued. V.H.B., C.R.M., and C.F. have no conflicts to disclose.

Funding Information

This research was supported by an Autism Speaks Dennis Weatherstone predoctoral fellowship and Bay Area Autism Consortium Collaboration Counts grant to V.H.B. and the NIMH (K23MH115166-01 to V.H.B. and K01MH103500 and R01MH113576 to K.O.G).