How Accurate Are Screening Tools for ADHD in Children and Adolescents?

What you need to know

• Most ADHD screening tools have good overall diagnostic accuracy but a single measure completed by a single reporter is unlikely to be sufficiently accurate for clinical use or population screening.
• The accuracy of screening tools varies across reporters and populations, with parent reports generally being most accurate and community samples showing higher accuracy than clinical samples.
• None of the tools met minimal standards for acceptable sensitivity (0.8) and specificity (0.8) needed for effective screening.

Background and Objectives

Attention-deficit/hyperactivity disorder (ADHD) is a common neurodevelopmental disorder affecting about 5% of children and adolescents globally. However, recognition and diagnosis rates vary considerably between and within countries, leading to concerns about both under-recognition and misdiagnosis. Efficient screening could help maximize identification of possible cases for further assessment, but the accuracy of available screening tools has not been comprehensively evaluated.

This systematic review and meta-analysis aimed to:

1. Determine the accuracies of a broad range of screening tools for ADHD in children and adolescents
2. Compare the diagnostic accuracy of tools between population-based and clinical/high-risk samples, and across different reporters (e.g. parents, teachers, self-report)

Methods

The researchers conducted a systematic search of major medical and psychological databases for studies reporting on the diagnostic accuracy of ADHD screening tools in children and adolescents under 18 years old. They included 75 studies published between 1985-2021, covering 41 different screening tools.

The main measures of accuracy examined were:

• Area under the curve (AUC): A measure of overall diagnostic accuracy, with 0.7-0.8 considered acceptable, 0.8-0.9 excellent, and >0.9 outstanding.
• Sensitivity: The proportion of true ADHD cases correctly identified by the screening tool.
• Specificity: The proportion of non-ADHD cases correctly identified as not having ADHD.

The researchers conducted meta-analyses to provide pooled estimates of these accuracy measures for groups of similar screening tools. They also compared accuracy between different reporters and sample types.

Key Findings

Overall diagnostic accuracy:

• Most screening tools showed excellent overall diagnostic accuracy based on AUC values, with a pooled AUC of 0.82 for ADHD symptom scores.
• Accuracy varied considerably across reporters (AUC range 0.67-0.92) and populations (AUC range 0.60-0.95).

Sensitivity and specificity:

• None of the measures met minimal standards for both acceptable sensitivity (0.8) and specificity (0.8) needed for effective screening.
• Tools tended to have good sensitivity at the expense of specificity, or vice versa.

Differences across reporters:

• Parent reports were generally the most accurate, followed by self-reports.
• Teacher reports were typically the least accurate, often falling below the acceptable range.

Differences across populations:

• Community-based samples showed higher accuracy than clinical/high-risk samples.
• Case-control studies (clinical cases vs. community controls) had the highest accuracy.

Specific tools:

• The Achenbach System of Empirically Based Assessment (ASEBA) DSM-oriented ADHD subscale showed the best balance of sensitivity (0.75) and specificity (0.81) at a cut-off of 5.
• The Strengths and Difficulties Questionnaire (SDQ) and DSM-IV symptom scales showed varying levels of accuracy depending on the specific subscale and cut-off used.

Conclusions

While most ADHD screening tools show good overall diagnostic accuracy, a single measure completed by a single reporter is unlikely to be sufficiently accurate for clinical use or population-wide screening. The trade-off between sensitivity and specificity means that adjusting cut-offs to maximize identification of true cases leads to more false positives, and vice versa.

The high degree of variability in accuracy across studies makes it difficult to predict how these tools would perform in real-world settings. More rigorous reporting standards are needed in diagnostic accuracy studies of screening tools.

Implications for Practice

• Caution should be used when relying on questionnaires to screen for ADHD in both clinical and research settings.
• Whenever possible, screening measures should be completed by multiple reporters rather than relying on a single informant.
• Parent reports appear to be most accurate for community-based screening, while parent or self-reports may perform similarly in clinical settings.
• Teacher reports alone should not be used for ADHD screening.
• Two-stage screening approaches or combining multiple measures may improve accuracy over single-stage screening.
• Further research is needed to identify optimal approaches to ADHD screening that balance accuracy, feasibility, and cost-effectiveness.

The limitations in current screening tools highlight the ongoing challenges in efficiently identifying ADHD cases while avoiding over-diagnosis. Clinicians and researchers should be aware of these limitations when interpreting screening results and making diagnostic decisions.