Measurement Equivalence of the ASRS Across the Adult Lifespan: A Differential Item Functioning Analysis

This study reveals that the Adult ADHD Self-Report Scale (ASRS) exhibits significant differential item functioning across the adult lifespan, with older adults showing reduced endorsement of hyperactivity symptoms but increased endorsement of inattention symptoms, suggesting that current screening practices may systematically underestimate ADHD severity in older populations and highlighting the need for age-adjusted norms or the use of the full 18-item scale.

The Gist

The Big Idea: The "ADHD Thermometer" is Broken for Older Adults

Imagine you have a thermometer designed to measure body temperature. It works perfectly for a 25-year-old. But if you use that exact same thermometer on an 80-year-old, it starts giving you the wrong reading. It might say the older person has a fever when they don't, or miss a fever entirely.

This study found that the Adult ADHD Self-Report Scale (ASRS)—the most common "thermometer" used to screen for ADHD in adults—works differently depending on how old you are.

The researchers discovered that as people get older, their symptoms of ADHD change shape, but the test doesn't know how to adjust for that. This means doctors might be missing ADHD in older adults because the test is looking for the wrong things.

---

The Cast of Characters

• The Test (ASRS): A 18-question checklist.
  • Part A (The Gatekeeper): The first 6 questions. This is the "quick scan" doctors use to decide if you need a full check-up.
  • Part B (The Deep Dive): The remaining 12 questions. This gives a fuller picture.
• The Participants: 600 people, evenly split into age groups from 20 to 80 years old.
• The Problem: The test assumes a 25-year-old and a 75-year-old experience ADHD symptoms in the exact same way. The study says: No, they don't.

---

The Discovery: The "Symptom Shuffle"

The researchers used a sophisticated statistical method (think of it as a high-tech magnifying glass) to see if the questions meant the same thing to a 20-year-old as they did to an 80-year-old.

They found that while the underlying ADHD severity might be the same, the way symptoms show up changes as we age. It's like a musical instrument changing its tune over time.

1. The "Hyperactivity" Fade (Part A)

In the "Gatekeeper" section (Part A), the test asks about physical restlessness, like "fidgeting," "feeling driven by a motor," or "squirming."

• The Finding: Older adults are much less likely to say "Yes" to these questions, even if they have the same level of ADHD as a younger person.
• The Analogy: Imagine a young ADHD brain is like a racing car with the engine revving loud. An older ADHD brain is like that same car, but the engine has been tuned down. It's still moving fast internally, but it's not making as much noise or shaking the seat.
• The Result: Because older people stop saying "I fidget," the test thinks they don't have ADHD. The score drops, and the doctor might say, "You're fine," when they aren't.

2. The "Inattention" Rise (Part B)

In the "Deep Dive" section (Part B), the test asks about mental organization, like "making careless mistakes," "misplacing things," or "interrupting people."

• The Finding: Older adults are more likely to say "Yes" to these questions as they age.
• The Analogy: If the racing car's engine is quieter, the driver is getting more tired. The older brain struggles more with organizing files, remembering where keys are, or waiting its turn in a conversation.
• The Result: These symptoms become more visible and easier to admit to as we get older.

---

The "Score Gap"

The study calculated exactly how much this matters. If a 20-year-old and an 80-year-old have the exact same amount of ADHD inside them:

• The 20-year-old gets a high score on the "Gatekeeper" (Part A) because they admit to fidgeting and restlessness.
• The 80-year-old gets a much lower score on that same Gatekeeper because they don't fidget as much.

The Gap: The 80-year-old's score is about 1.36 points lower just because of their age. In the world of medical screening, that is a huge difference. It's the difference between passing a test and failing it.

Why This Matters (The "So What?")

Right now, many doctors and researchers only look at the first 6 questions (Part A) to screen for ADHD.

• The Risk: Because the "Hyperactivity" questions (Part A) drop off as we age, the test acts like a net with holes in it. It catches young people, but older people slip right through the holes.
• The Consequence: Older adults with ADHD are being told they are "just getting old" or "forgetful," when they actually have a treatable condition. Their symptoms are being misdiagnosed as normal aging or memory loss.

The Solution: Change the Lens

The authors suggest two simple fixes:

1. Don't just use the "Gatekeeper": When checking older adults, doctors should use the full 18-question test, not just the first 6. The full test includes the "Inattention" questions that older adults are still struggling with.
2. New Rules for Old Ages: We need to create "age-adjusted" scores. Just like a thermometer might need a different setting for a baby vs. an adult, the ADHD test needs to know that "fidgeting" isn't the main sign of ADHD in a 70-year-old.

The Bottom Line

ADHD doesn't disappear when you turn 60; it just changes its outfit. It stops wearing the "hyperactive" jacket and starts wearing the "disorganized" sweater.

If we keep using a test that only looks for the "hyperactive" jacket, we will keep missing the people who are wearing the "disorganized" sweater. To help older adults, we need to update our tools to see the whole picture.

Technical Summary

1. Problem Statement

Adult Attention-Deficit/Hyperactivity Disorder (ADHD) is increasingly recognized as a lifelong condition, yet it remains significantly underdiagnosed in older populations (ages 50+). Current prevalence estimates suggest a sharp decline in ADHD rates after age 50, but this is likely confounded by measurement bias rather than a true biological remission.

The primary screening tool, the Adult ADHD Self-Report Scale (ASRS), was originally validated on adults aged 18–44. There is no evidence that the instrument functions equivalently across the entire adult lifespan (20–80 years). Specifically, the study addresses whether the ASRS exhibits Differential Item Functioning (DIF), where items are interpreted differently by older adults compared to younger adults despite having the same underlying level of ADHD severity. If DIF exists, the standard 6-item Part A screener may systematically underestimate ADHD severity in older adults, leading to missed diagnoses.

2. Methodology

Participants and Design

• Sample: $N = 600$ adults recruited via Prolific (US/UK), stratified into six decades (20–29, 30–39, 40–49, 50–59, 60–69, 70–80) with $N=100$ per group.
• Inclusion Criteria: English proficiency, no diagnosed psychiatric/neurological conditions, high approval rating on previous studies.
• Instrument: The WHO Adult ADHD Self-Report Scale (ASRS-v1.1), an 18-item Likert scale (0="Never" to 4="Very Often").
  • Part A: 6-item clinical screener (most predictive items).
  • Part B: 12-item supplementary checklist.

Statistical Analysis
The analysis followed a rigorous Item Response Theory (IRT) framework:

1. Data Preprocessing: Response categories were collapsed from 5 to 4 points (merging "Often" and "Very Often") to address sparse data in older age groups.
2. Latent Trait Estimation: A Bi-factor Graded Response Model (GRM) was estimated using Quasi-Monte Carlo Expectation-Maximisation (QMC-EM).
   • Structure: One General ADHD factor (loading on all 18 items) orthogonal to two specific factors: Inattention and Hyperactivity/Impulsivity.
   • Reliability: The General factor showed high hierarchical reliability ($\omega_H = .895$).
   • Latent scores ($\theta$) were extracted via Expected A Posteriori (EAP) estimation.
3. DIF Detection: DIF was assessed using Ordinal Logistic Regression comparing three nested models for each item:
   • Baseline: Response predicted by Latent Severity ($\theta$) only.
   • Uniform DIF: Response predicted by $\theta + \text{Age}$.
   • Non-Uniform DIF: Response predicted by $\theta + \text{Age} + (\theta \times \text{Age})$.
   • Adaptive Modeling: Age was modeled using natural cubic splines ($df=2$), with an automatic fallback to linear terms if cell counts dropped below $N=8$ to prevent overfitting.
   • Correction: P-values were adjusted using the Benjamini-Hochberg procedure to control the False Discovery Rate.
4. Impact Quantification: Expected Total Scores (ETS) were calculated for Part A and Part B, comparing a 20-year-old vs. an 80-year-old at equivalent latent severity ($\theta = 0$).

3. Key Results

Structural Validity
The bi-factor model demonstrated acceptable global fit. The General ADHD factor accounted for 80.2% of the common variance (ECV), confirming the unidimensionality of the construct across the lifespan.

Differential Item Functioning (DIF)

• Uniform DIF: 5 out of 18 items exhibited significant uniform DIF (all adjusted $p < .002$). No significant Non-Uniform DIF was found, indicating that item discrimination remained stable; the bias was purely in the severity thresholds.
• Part A (Screener) - Negative Bias: Two hyperactivity/impulsivity items showed negative age effects:
  • Item 5 ("fidget or squirm")
  • Item 6 ("overly active and compelled to do things")
  • Finding: Older adults were significantly less likely to endorse these symptoms than younger adults at the same latent severity.
• Part B (Checklist) - Positive Bias: Three items showed positive age effects:
  • Item 7 ("careless mistakes")
  • Item 10 ("difficulty finding things")
  • Item 18 ("interrupt others when busy")
  • Finding: Older adults were more likely to endorse these inattentive/disorganized symptoms than younger adults at the same latent severity.

Quantitative Impact (Expected Total Scores)
At equivalent latent severity ($\theta = 0$):

• Part A: The expected score for an 80-year-old was 1.36 points lower than for a 20-year-old (a decrease of ~0.27 points per decade).
• Part B: The expected score for an 80-year-old was 1.15 points higher than for a 20-year-old (an increase of ~0.23 points per decade).
• Correlation Divergence: Part A scores showed a significantly stronger negative correlation with age ($r = -.26$) compared to Part B ($r = -.12$), confirmed by Steiger's test ($p < .001$).

4. Key Contributions

1. Identification of Phenotypic Redistribution: The study provides empirical evidence that ADHD symptom expression shifts with age. Hyperactivity/impulsivity (Part A) becomes less overt and less self-reported, while inattentive/disorganized symptoms (Part B) become more prominent or easier to endorse.
2. Validation of Measurement Bias: The research demonstrates that the widely used 6-item Part A screener is not measurement invariant across the lifespan. It systematically underestimates severity in older adults because it relies heavily on items (fidgeting, motor activity) that naturally decline or change presentation with age.
3. Methodological Rigor: The application of a Bi-factor GRM combined with adaptive ordinal logistic regression offers a robust framework for detecting DIF in aging populations, addressing issues of data sparsity and non-linear developmental trajectories.

5. Significance and Clinical Implications

• Risk of Underdiagnosis: Relying solely on the Part A screener in primary care or epidemiological studies likely leads to a systematic under-identification of ADHD in adults over 50. This exacerbates the existing "ten-fold discordance" between symptom estimates and clinical diagnoses in older cohorts.
• Differentiation from Cognitive Decline: The underestimation of ADHD in older adults complicates the differential diagnosis between ADHD and Mild Cognitive Impairment (MCI), as both present with inattention and executive dysfunction.
• Recommendations:
  • Immediate Action: Clinicians and researchers should utilize the full 18-item ASRS rather than the 6-item screener when assessing older adults.
  • Future Development: There is an urgent need to develop age-adjusted norms or dedicated screening tools that account for the phenotypic redistribution of symptoms (e.g., weighting inattentive symptoms more heavily for older cohorts).

Conclusion: The ASRS does not function equivalently across the adult lifespan. The "mellowing" of hyperactivity symptoms in older adults is not merely a biological remission but a measurement artifact that, if uncorrected, obscures the true prevalence and severity of ADHD in the aging population.