Who gets an attention diagnosis is changing faster than attention itself — and a mechanism for why focus drains with time on task
The headline evidence this week is epidemiological rather than mechanistic: a Danish registry case-control study of 71,317 children diagnosed with ADHD or autism shows that, between 2012 and 2022, diagnosed children became progressively more similar to their undiagnosed peers on perinatal, socioeconomic and health-care markers — the clearest quantitative signal yet that rising diagnosis rates partly reflect changing diagnostic practice, not only changing risk. An All of Us cross-section of 13,527 adults with ADHD adds that adult-diagnosed cases carry a milder clinical and polygenic profile than youth-diagnosed cases, yet remain far from controls. On mechanism, a bioRxiv preprint from the Lavie lab offers a neurophysiological account of the vigilance decrement — time on task steepens the aperiodic 1/f slope, consistent with falling excitatory drive, and perceptual load makes it worse. The attention-economy items are deliberately mixed: a preregistered 21-day teen smartphone detox produced only transient gains, a JAMA Network Open review finds the evidence base for school phone bans thin, and a twelve-week field RCT shows that letting people choose their own reduction target beats assigning one. One preprint, one non-randomized observational study and one small EEG sample are flagged as such.
Odds ratio for low birth weight among children later diagnosed with ADHD/ASD, 2012–13 vs 2020–22 — diagnosed children are converging on their peers (Danish registry, 71,317 cases)
Where the human evidence is
Count by statusEvidence-stage map
Stage × promise (1–10)Attention domain impact
Strength 0–10Why focus drains with time on task (preprint mechanism)
MechanismThe Lavie-lab bioRxiv preprint proposes that sustaining attention consumes a resource needed for excitatory cortical signalling. As minutes pass, the EEG aperiodic 1/f slope steepens (read as a falling excitation/inhibition ratio) and alpha rises; detection sensitivity drops, responses grow variable and the mind wanders more. Demanding stimuli (high perceptual load) suppress mind-wandering in the short run but drain the resource faster. Not yet peer-reviewed; the E/I reading of the 1/f slope is itself debated.
The diagnosis-rate debate gets its best data yet
Children diagnosed with ADHD or autism became steadily more similar to undiagnosed peers between 2012 and 2022 — Danish registry, 71,317 cases and 713,170 matched controls
Tarp and colleagues followed everyone born in Denmark 1994–2022 to age 18, matched each child diagnosed with ADHD or ASD between 2012 and 2022 to ten controls on sex, birth year, municipality and country of birth, and tracked how 19 prediagnostic characteristics (parental, perinatal, health-care use) related to diagnosis by year. Diagnosed children differed from controls on every characteristic, but the odds ratios attenuated toward the null over the decade, most sharply for socioeconomic and perinatal factors: the OR for low birth weight fell from 1.54 (95% CI 1.41–1.68) in 2012–13 to 1.17 (1.10–1.24) in 2020–22, and the household-income OR for ADHD shrank by a mean 6.5% per year versus 0.7% for ASD. Attenuation was larger for ADHD than ASD and for diagnoses at ages 10–17 than in younger children, and held in cases with higher diagnostic certainty. The authors' reading is that diagnostic practice and capacity — not underlying risk alone — drive the rise; they are explicit that this says nothing about impairment at the individual level. The design is register-based and cannot identify which practice changes are responsible.
Read the paperAdults first diagnosed with ADHD in adulthood show a milder clinical and polygenic profile than youth-diagnosed adults — but both sit far from controls (All of Us, n=13,527)
Leffa, Cortese, Rohde, Molina and colleagues split self-reported ADHD in the US All of Us cohort by age at first diagnosis (28.5% youth-diagnosed, 71.5% adult-diagnosed; both ~72–75% female) and matched each group 1:2 to controls. Adult-diagnosed participants were older, better educated and higher-income, and had lower adjusted probabilities of most psychiatric comorbidities (largest gaps: PTSD −7.5, bipolar −7.4 percentage points) and somatic conditions, with a lower ADHD polygenic risk score (Cohen d = −0.17 vs youth-diagnosed). Yet both groups showed large excesses over matched controls — depression +39 and +37 percentage points respectively — and higher PRS than controls (d = 0.21 youth-diagnosed, 0.12 adult-diagnosed). The authors argue adult-diagnosed ADHD is a milder but clinically real subgroup that may be recognised late precisely because it is milder. Limitations: self-reported diagnosis and age at diagnosis, a volunteer cohort, and no ability to separate late-recognised childhood ADHD from a distinct adult-onset entity.
Read the paperThe mechanism — why attention drains with time on task
Time on task steepens the aperiodic 1/f slope and raises alpha power alongside falling sensitivity and rising mind-wandering — and perceptual load makes the drain worse
Barne and colleagues from Nilli Lavie's group had participants detect rare mountain scenes in a stream of city scenes under low or high perceptual load (with or without overlaid noise) while EEG was parameterised into periodic and aperiodic components; the 1/f slope is used as a proxy for excitation/inhibition balance, steeper slopes indicating a lower E/I ratio. With time on task, alpha power rose broadly and the 1/f slope steepened in a left temporo-parietal cluster, in step with reduced detection sensitivity, more variable responses, more thought-probe-reported mind-wandering and less detailed thoughts. High load reduced mind-wandering — the established load-theory result — but exacerbated the time-on-task decline in sensitivity and the slope steepening in a right parieto-occipital cluster. The authors' interpretation is that sustaining attention depletes a resource needed for excitatory signalling and that demanding stimuli drain it faster. This is a single-lab preprint; the sample size is not in the abstract, the E/I interpretation of the 1/f slope is itself contested, and no peer-reviewed version exists yet.
Read the paperIn medication-naïve adults with ADHD, mind-wandering is not accompanied by the stable posterior alpha rise seen in controls — alpha is instead more variable
Horáková, Weibel and colleagues (Strasbourg) probed mind-wandering during a Go/No-Go task and compared pre-stimulus alpha power and its variability across attentional states. Controls showed higher global alpha power during mind-wandering without increased variability, consistent with a stable, internally directed state; ADHD participants showed no alpha modulation between states but greater alpha variability during mind-wandering, which the authors read as unstable internal distraction rather than a coherent inward shift. Deliberate and spontaneous mind-wandering had distinct spatial alpha patterns, with deliberate episodes keeping attentional systems relatively active. This is a small single-site sample and the thought-probe method cannot separate the timing of the lapse from its report, so the marker is descriptive rather than diagnostic.
Read the paperThe attention economy — trials, policy and a thin evidence base
A preregistered 21-day smartphone and social-media detox in 82 UK teenagers produced tentative gains in sleep, mood, stress and working memory — none of which survived to two-month follow-up
Sullivan, Cairney and Henderson (York) randomised adolescents to a total smartphone/social-media detox (n=26), a detox with a basic 'brick' phone (n=26) or business-as-usual (n=30), with surveys, wearables and cognitive tasks before and after. Attrition was 13%. At 21 days one or both detox arms showed tentative improvements in sleep duration, daytime sleepiness, depressive symptoms, perceived stress and working memory relative to controls, but no changes persisted at two months. Qualitative reports mixed benefits with boredom, functional reliance on the phone and fear of missing out; 83% of participants nonetheless endorsed government action. As a pilot it is underpowered for efficacy and the outcomes are many, so the appropriate reading is feasibility plus a caution that short abstinence does not produce durable change. The authors suggest fully powered trials test targeted measures such as switching off mobile internet at bedtime.
Read the paperChallenged: the evidence that school phone bans, age verification or warning labels improve adolescent outcomes is thin — 39 eligible papers, mostly observational or expert opinion
Nagata, Moreno and colleagues (UCSF/Wisconsin) reviewed policies and legislative proposals from 2003 to March 2026 across warning labels, age restriction and assurance, parental consent and monitoring, school phone policies, platform design regulation and deepfake laws. Only four included papers examined school phone policies and two examined warning labels; age-assurance measures were commonly circumvented; parental monitoring was associated with lower problematic use in survey data; and for phone restrictions the review states plainly that evidence on academic, social or mental-health effects is limited. Warning-label support consisted largely of expert opinion (a separate JAMA Health Forum within-participants RCT this month found warnings rated more effective than a control message, but that measures perceived effectiveness, not behaviour). The point for this digest: the popular claim that phone bans demonstrably restore attention or wellbeing is not yet supported by evaluation studies — consistent with the null Lancet Regional Health cohort covered in the 27 July edition.
Read the paperLetting people choose their own phone-reduction target cut use by 26 more minutes a day than assigning an equivalent target — twelve-week field RCT, n=149
Sachdeva, Somasundaram and Sharma randomised adults to self-select a 10%, 20% or 30% daily reduction, to an assigned 14% reduction (matching the average self-selected goal), or to no goal, with identical incentives. Self-selectors reduced use by 26 minutes more per day (a 73% larger reduction) and hit their goals 11 percentage points more often than the assigned group. Reductions correlated with improvements in perceived addiction, depressive and anxiety symptoms, but those were secondary endpoints, underpowered, and not significantly different between arms — the authors say so. The behavioural effect is the finding; the wellbeing effect is not established. It is a single trial with monetary incentives, so generalisation to unincentivised self-regulation is untested.
Read the paperWhen a patient or colleague cites rising ADHD prevalence as evidence of an epidemic, the Danish registry data are the counterweight: diagnosed children have converged on undiagnosed peers across perinatal and socioeconomic markers, which points to practice change, not only risk change — and says nothing about individual impairment.
Adult-diagnosed ADHD is real but milder on average (All of Us): expect lower comorbidity and polygenic load than in childhood-diagnosed patients, but do not read a late diagnosis as a false one — both groups sit far from controls.
If the Lavie-lab preprint holds up, the vigilance decrement is not just boredom but a measurable neurophysiological drain that demanding input accelerates; treat load as something to budget across a long task, not simply a way to stay engaged.
For phone-use change, structure beats abstinence: a 21-day teen detox produced no durable gains, whereas letting people pick their own reduction target produced 26 more minutes a day of reduction than assigning one (RCT, n=149).
Weekend methylphenidate holidays in adults rest on almost no evidence; a small non-randomised comparison associates them with worse diary-rated functioning and more reported BP elevation, so the default should be an explicit clinical reason rather than habit.