The Autism Mental Status Exam Indonesian Version(AMSE-INA): A Cross-Culturally Validated Pre-Diagnostic Tool for Autism Spectrum Disorder in Resource-Limited Specialist Settings

Introduction

Autism spectrum disorder (ASD) is a complex, lifelong neurodevelopmental condition characterized by persistent deficits in social communication and interaction, alongside restricted, repetitive patterns of behavior, interests, or activities.¹ Global prevalence has increased significantly over recent decades, with current estimates suggesting approximately 1–2% of children worldwide are affected.^2,3 This rising prevalence, coupled with the critical importance of early intervention for optimizing long-term outcomes, has placed considerable pressure on diagnostic systems globally.^4,5

In high-income countries, diagnosis is typically achieved through comprehensive multidisciplinary assessment involving gold-standard instruments such as the Autism Diagnostic Observation Schedule-2 (ADOS-2) and the Autism Diagnostic Interview-Revised (ADI-R).^6,7 These tools, while robust, require extensive training, are time-consuming (often ≥2 hours per assessment), and involve significant costs for materials and licensing.^8,9 Consequently, their availability is severely limited in low- and middle-income countries (LMICs), including Indonesia.¹⁰

Indonesia, the world’s fourth most populous nation, faces profound challenges in ASD diagnosis. Data on prevalence are limited, but estimates align with global figures, with significant under-identification in rural areas due to scarce specialist services, lack of trained personnel, and financial constraints.^11,12 Specialists working in secondary or tertiary referral centers often lack access to ADOS-2 or ADI-R and must rely solely on clinical judgment, which can be subjective and variable, leading to potential over- or underdiagnosis, especially in complex cases with comorbidities.^13–15 Furthermore, widely used screening tools such as the Modified Checklist for Autism in Toddlers, Revised (M-CHAT-R/F) are limited to children under 30 months and rely on parent report, making them unsuitable for older children or as a pre-diagnostic aid for specialists.^16,17 There is an urgent, unmet need for a tool that can structure the specialist’s clinical examination, making it more objective, efficient, and reliable within the constraints of the Indonesian healthcare system.

The Autism Mental Status Exam (AMSE) is an 8-item, observation-based examination designed to structure the clinical assessment of ASD core symptoms in under 15 minutes.^15,18 It was developed specifically to enhance the accuracy of clinical judgment in settings where gold-standard tools are impractical.¹⁸ By systematically guiding the observation of social interaction, communication, and repetitive behaviors—and incorporating key caregiver report—it operationalizes DSM-5 criteria within a routine clinical encounter.¹⁹ Validated in multiple countries (USA, Brazil, China, Turkey, Chile, Sweden and France.), the AMSE has demonstrated consistently high sensitivity (79–98%) and specificity (67–100%).^20–25 As an open-access instrument requiring minimal training, it presents a promising solution for specialist diagnostic settings in LMICs.

A critical and often underexplored aspect of diagnostic tool validation is performance across different developmental stages. ASD symptoms manifest and evolve with age.²⁶ In toddlers(<3 years), core symptoms may be more categorical and overt (e.g., absence of joint attention, lack of response to name). In preschool and school-aged children (≥3 years), symptoms often become more nuanced, variable, and influenced by compensatory strategies or comorbid conditions.²⁷ Tools validated on mixed-age samples may mask important age-related differences in sensitivity and specificity. Therefore, a comprehensive validation must include age-stratified analysis to inform clinicians about the tool’s appropriate use across the intended age spectrum.

However, the use of observational tools is influenced by cultural norms surrounding social behavior, eye contact, communication styles, and play.²⁸ A direct translation is insufficient; a rigorous cultural adaptation is essential to ensure diagnostic accuracy.²⁹ No such adaptation or validation exists for the Indonesian context.

This study aimed to: (1) perform a comprehensive cross-cultural adaptation of the AMSE into the Indonesian language and context (AMSE-INA), and (2) evaluate its psychometric properties and diagnostic accuracy specifically for use as a pre-diagnostic tool by specialists in Indonesian referral settings where gold-standard instruments are not available.

Methods

Results

Validity

• • Content Validity: The scale-level CVI (S-CVI/Ave) was 1.00, and all item-level CVIs (I-CVI) were 1.00, indicating unanimous expert agreement on relevance.

• • Concurrent Validity: AMSE-INA total scores correlated strongly with CARS-2 ST total scores (Spearman’s rho = 0.85, p < 0.001).

• • Discriminant Validity: AMSE-INA scores significantly differed across diagnostic groups (Kruskal-Wallis H = 125.28, p < 0.001), with highest scores in the ASD group AMSE-INA scores were significantly higher in the ASD group (median = 8) compared to the non-ASD group (median = 3) (Mann-Whitney U, p < 0.001).

Diagnostic Accuracy

ROC analysis indicated outstanding diagnostic performance ( Figure 1). The AUC was 0.978 (95% CI: 0.958–0.998, p < 0.001). The optimal statistical cut-off was ≥5.5 (Youden’s Index = 0.886). For clinical utility, integer cut-offs were evaluated ( Table 2).

Figure 1. Receiver operating characteristic (ROC) curve of AMSE-INA scores for predicting autism spectrum disorder diagnosis against the best estimate clinical diagnosis (BECD) reference standard.

The blue curve represents the AMSE-INA performance across different cut-off scores. The diagonal dashed line represents the reference line (area = 0.5), indicating no discriminatory power. The Area Under the Curve (AUC) was 0.978 (95% CI: 0.958–0.998, p < 0.001), demonstrating outstanding diagnostic accuracy. The optimal cut-off score of ≥6 (Youden’s Index = 0.886) is indicated by the point nearest the top-left corner of the curve.

Table 2. Diagnostic accuracy of AMSE-INA at different cut-off scores (N = 174).

Parameter	Cut-off ≥5	Cut-off ≥6
Sn [%, (95% CI)]	93.9 (90.4–97.5)	88.9 (84.2–93.6)
Sp [%, (95% CI)]	86.7 (81.7–92.7)	94.7 (91.3–98.0)
PPV [%, (95% CI)]	90.3 (85.9–94.7)	95.7 (92.5–98.6)
NPV [%, (95% CI)]	91.5 (87.4–95.7)	86.6 (81.5–92.7)
Accuracy [%, (95% CI)]	90.8 (86.5–95.1)	91.4 (87.2–95.6)
LR+ (95% CI)	7.1 (4.0–12.6)	16.7 (6.4–43.7)
LR- (95% CI)	0.07 (0.03–0.15)	0.12 (0.07–0.20)

Age-Stratified Diagnostic Accuracy:

• • Children <3 years: The diagnostic accuracy was perfect, with an AUC of 1.00.

• • Children ≥3 years: The diagnostic accuracy remained excellent, with an AUC of 0.97 (95% CI: 0.93–0.99).

A cut-off of ≥6 provided an optimal balance for a pre-diagnostic tool in a specialist setting, maximizing specificity (94.7%) and PPV (95.7%) to minimize false positives, while maintaining high sensitivity (88.9%). A cut-off of ≥5 could be considered for high-sensitivity screening in lower-prevalence settings ( Table 3).

Table 3. Diagnostic accuracy of AMSE-INA at cut-off ≥6, Overall and by age group.

Parameter	Overall sample (N = 174)	Children <3 years (n = 41)	Children ≥3 years (n = 133)
Sn [%, (95% CI)]	88.9 (84.2–93.6)	94.1(73.0–99.0)	87.8 (79.0–93.2)|
Sp [%, (95% CI)]	94.7 (91.3–98.0)	100 (85.7–100)*	92.2 (81.2–97.8)
PPV [%, (95% CI)]	95.7 (92.5–98.6)	100 (79.4–100)*	94.7 (86.8–98.5)
NPV [%, (95% CI)]	86.6 (81.5–92.7)	96.0 (80.5–99.3)	82.5 (69.8–91.3)
Accuracy [%, (95% CI)]	91.4 (87.2–95.6)	97.6 (87.4–99.6)	89.5 (83.1–93.6)
LR+ (95% CI)	16.7 (6.4–43.7)	>100 (5.0- ∞)	11.3 (4.3–28.8)
LR- (95% CI)	0.12 (0.07–0.20)	0.06 (0.002–0.29)	0.13 (0.07–0.24)

* Note: Confidence intervals for proportions were calculated using the Wilson score method, except for values of 100% where exact binomial (Clopper–Pearson) intervals are reported.

Discussion

This study successfully adapted and validated the AMSE for use in Indonesia. The resulting AMSE-INA is not merely a screening questionnaire but a structured pre-diagnostic examination tool with robust psychometric properties, specifically designed to aid clinical decision-making by specialists in settings where gold-standard instruments like ADOS-2 are inaccessible. The age-stratified analysis offers novel and crucial insights for its application across different developmental stages.

The cultural adaptation process was vital to this success. Modifications like clarifying “pointing toward an object” rather than “at an object” account for local nonverbal communication norms, ensuring behavioral anchors are meaningful.^28,30 The development of a detailed, culturally relevant scoring guide based on pre-test feedback was crucial for achieving high inter-rater reliability, transforming the tool from a simple translation into an operational clinical instrument.³⁰

The perfect content validity (CVI = 1.00) is a direct outcome of the rigorous adaptation process and confirms that AMSE-INA is culturally and conceptually appropriate for Indonesia. Key adaptations, such as refining the description of pointing behavior, mirror the essential work done in other cultural validations, ensuring the tool measures ASD symptoms rather than cultural misfit.^18,23 This step is non-negotiable; a tool’s ecological validity in a new setting depends on such careful localization of behavioral anchors.

The observed internal consistency (Cronbach’s α = 0.66) warrants a nuanced interpretation that is common in the validation of multidimensional diagnostic tools. A moderate alpha is not indicative of a flawed tool but reflects its intentional design to capture the distinct, heterogeneous domains of ASD as defined by the DSM-5. The AMSE is not a unidimensional scale measuring a single latent trait; it is a structured clinical checklist that operationalizes two relatively independent symptom domains: social-communication deficits and restricted, repetitive behaviors (RRBs). This pattern is consistent with global AMSE validations, in which Cronbach’s alpha typically ranges from 0.61 to 0.80 rather than exceeding 0.90. For instance, studies in Chile and China reported alphas of 0.61 and 0.65, respectively,^21,23 while Brazil reported 0.74,²⁰ and Turkey reported 0.80.²²

The moderate internal consistency (α = 0.66) is consistent with previous AMSE validations and reflects its intentional design to measure related but distinct clinical domains of ASD (social communication and restricted/repetitive behaviors).^18,21,22 The low contribution of the “Pragmatics” item is a known characteristic of the tool across cultures and may relate to the challenge of reliably observing this domain in a brief examination.^20,34 The high content validity confirms that Indonesian experts view the AMSE-INA items as fully representative of core ASD symptoms within the local context.

The consistently low item-total correlation for the “Pragmatics of Language” item across studies (including ours) is a known characteristic, likely due to its conditional scoring and the subtle, culturally-influenced nature of pragmatic deficits. Therefore, the internal consistency result should be interpreted as evidence that AMSE-INA’s total score is a composite of semi-independent domains, which is appropriate for its purpose. This stands in clear contrast to its excellent inter-rater and test-retest reliability, which confirms that clinicians can apply the scoring rules for each individual item with very high consistency.

The core strength of AMSE-INA lies in its excellent reliability and high diagnostic accuracy. The inter-rater and test-retest ICCs (>0.95) exceed those of many established clinical tools and indicate that AMSE-INA can yield consistent scores across clinicians and over time, a critical feature for a reliable diagnostic aid.^35,36 The AUC of 0.98 is exceptional and aligns with findings from validation studies in Brazil (AUC = 0.99) and Turkey (AUC = 0.98).^22,24 This confirms that the adapted instrument retains a near-perfect ability to discriminate between ASD and non-ASD cases within a referral population.

The proposed cut-off of ≥6 is strategically chosen for the tool’s intended use. In a specialist referral clinic where the pre-test probability of ASD is high (57% in this sample), the priority is to confirm the diagnosis with high confidence to justify resource-intensive interventions and family counselling. A specificity of 94.7% and a PPV of 95.7% mean that a child scoring ≥6 has a very high probability of having ASD, effectively “ruling in” the disorder. The high positive likelihood ratio (LR+ = 16.7) signifies a “conclusive” shift in post-test probability.³⁷ This addresses a key need: reducing diagnostic uncertainty and subjective variability when gold-standard tools are absent.^14,38

The age-stratified analysis provides one of the most significant and actionable findings of this study. The perfect diagnostic accuracy (AUC = 1.00) in children under 3 years is exceptionally promising. At the recommended cut-off of ≥6, AMSE-INA achieved 94.1% sensitivity, 100% specificity, and PPV in this young cohort. This indicates that when a toddler scores 6 or higher on AMSE-INA, the clinician can be virtually certain of an ASD diagnosis, a powerful asset for early, confident identification.

This superb performance in very young children likely stems from the nature of early ASD symptoms. In toddlers, core deficits (e.g., absent joint attention, lack of response to name, limited social smiling) are often more categorical, severe, and easily observable within a brief clinical interaction.^27,39,40 The structured observation of AMSE-INA effectively captures these clear behavioral markers. This finding is crucial for Indonesia, as it suggests AMSE-INA is not just a general tool but one with particular strength in facilitating early detection within specialist settings, directly addressing the national problem of late diagnosis.

For children aged 3 years and older, accuracy remained excellent (AUC = 0.97), though specificity at the ≥6 cut-off was slightly lower (92.2%) than in the younger group. This is clinically understandable. As children develop, ASD symptoms can become more nuanced, variable, and intertwined with comorbid conditions (e.g., ADHD, anxiety) or intellectual disability.²⁶ Some children may develop compensatory strategies that mask core social deficits during brief observations. Furthermore, the “Pragmatics of Language” item, which is more frequently scorable in this verbal older group, has known reliability challenges, potentially contributing to score variability. This does not diminish the tool’s value for older children but underscores that clinical judgment must always integrate the AMSE-INA score with a comprehensive developmental history. A score just below 6 in an older child with strong historical evidence of ASD should not rule out the diagnosis.

The strong correlation with CARS-2 ST supports the convergent validity of AMSE-INA with an established ASD assessment tool. However, future validation against gold-standard instruments like ADOS-2 would strengthen the evidence, though pragmatic constraints in LMICs often limit such comparisons.

AMSE-INA fills a specific niche in Indonesia’s diagnostic ecosystem. Widely-used tools like M-CHAT-R/F are parent-reported and age-restricted, serving as initial population screens.^15,41 In contrast, AMSE-INA is a clinician-driven, specialist-level observational tool. It does not replace a comprehensive clinical evaluation but structures and objectifies it, directly addressing the over-reliance on unstructured clinical judgment that can lead to diagnostic delays and errors in resource-limited settings.^13,42,43 For a specialist who cannot administer an ADOS-2, the 15-minute AMSE-INA provides a standardized framework for systematically assessing DSM-5 criteria, integrating caregiver reports, and generating a quantifiable score that strongly predicts the final diagnosis.

AMSE-INA’s role must be understood within the existing Indonesian diagnostic landscape. Tools like the M-CHAT-R/F are vital for Level 1 screening in community or primary care settings. In contrast, AMSE-INA is a Level 2, specialist-administered, observational tool. It does not replace a full clinical evaluation but structures and objectifies it. For specialists who cannot access an ADOS-2, AMSE-INA provides a standardized, evidence-based framework for efficiently gathering and synthesizing diagnostic information during a consultation. Its 15-minute administration time makes it feasible in busy public hospital settings, and its zero cost eliminates a major barrier to adoption. It effectively bridges the gap between a positive screening result and a definitive, specialist-level diagnosis.

Conclusion

This study provides a comprehensive cross-cultural adaptation and validation of the Autism Mental Status Exam for Indonesia. AMSE-INA is a reliable, valid, and diagnostically accurate pre-diagnostic tool across a wide age range that equips specialists with a structured, rapid, and accurate method for assessing ASD when access to gold-standard instruments is constrained. By providing a standardized framework to guide observation and clinical judgment, it has the potential to significantly improve diagnostic consistency and accuracy, reduce delays, and facilitate earlier intervention pathways for children with ASD in Indonesia and similar resource-limited settings. Its open-access nature and minimal training requirements make it a feasible and scalable solution for bridging the diagnostic gap in specialist care. Its widespread implementation in secondary and tertiary healthcare settings is strongly recommended.