how childhood ADHD got counted
the definition changed in 2013 and the count barely moved. the measurement changed, and the count jumped every time. this is the longest clean public series on any mental-health diagnosis in the united states — 22 unbroken annual years — and it says something more surprising than the story usually told about it.
this page is about how a number is made, not about whether anyone should be diagnosed or treated. it is not medical advice, and it is not a reason to question a diagnosis or change a medication. stopping ADHD medication without a plan can be destabilising; if anything here raises a question for you, it is a question for your prescriber, not a decision to make on your own. if you're in crisis, call or text 988 (u.s.), 24/7, free.
the chart, as the agency publishes it
CDC publishes childhood ADHD prevalence as four separate series and declines to connect them. We have adopted that segmentation rather than inventing one, which means if this presentation is wrong, it is wrong in the same way the federal agency's own presentation is wrong. CDC's note, verbatim:
“National surveys are occasionally redesigned to best reflect changes in the U.S. population. When survey design changes occur, data from different versions of the same survey should not be directly compared.”— CDC, Trends in ADHD Among U.S. Children, data notes [1]
Every number on this page is ages 3–17. That label is not decoration. Three different age bases circulate for these statistics — 0–17, 2–17 and 3–17 — and they are close enough to look interchangeable and are not. The widely quoted “6.1 million children in 2016” figure is ages 2–17, while the “7.1 million in 2022” figure is ages 3–17. Quoting those two side by side as published is a mistake, and it is made constantly.
parent-reported ADHD diagnosis, children aged 3–17
Four segments. The red bands are survey redesigns — the agencies that run these surveys say not to compare across them, so the line stops and restarts. The teal marker is DSM-5, a change to the definition rather than to the ruler, so the line runs straight through it.
show the numbers
| year | NHIS 1997–2018 | NHIS 2019–2024 (redesigned) | NSCH 2003–2011 (quadrennial — points only) | NSCH 2016–2024 (redesigned) |
|---|---|---|---|---|
| 1997 | 5.5% | — | — | — |
| 1998 | 5.9% | — | — | — |
| 1999 | 5.6% | — | — | — |
| 2000 | 6.6% | — | — | — |
| 2001 | 6.3% | — | — | — |
| 2002 | 7.2% | — | — | — |
| 2003 | 6.3% | — | 7.2% (7%–7.6%) | — |
| 2004 | 7.4% | — | — | — |
| 2005 | 6.6% | — | — | — |
| 2006 | 7.4% | — | — | — |
| 2007 | 7.2% | — | 8.9% (8.4%–9.4%) | — |
| 2008 | 8% | — | — | — |
| 2009 | 8.6% | — | — | — |
| 2010 | 8.4% | — | — | — |
| 2011 | 8.4% | — | 10.4% (9.9%–10.8%) | — |
| 2012 | 9.5% | — | — | — |
| 2013 | 8.8% | — | — | — |
| 2014 | 8.9% | — | — | — |
| 2015 | 9.8% | — | — | — |
| 2016 | 9.4% | — | — | 9.7% (9.1%–10.3%) |
| 2017 | 9.4% | — | — | 9.8% (8.9%–10.7%) |
| 2018 | 9.8% | — | — | 9.7% (9%–10.4%) |
| 2019 | — | 8.8% (8%–9.6%) | — | 9.6% (9%–10.3%) |
| 2020 | — | 10.3% (9.1%–11.5%) | — | 10.1% (9.5%–10.7%) |
| 2021 | — | 9.6% (8.8%–10.3%) | — | 10.6% (10%–11.2%) |
| 2022 | — | 10.2% (9.3%–11.1%) | — | 11.4% (10.9%–12.1%) |
| 2023 | — | 12% (11%–13%) | — | 11.4% (10.9%–12%) |
| 2024 | — | 12% (11.1%–12.9%) | — | 12.6% (12%–13.2%) |
what DSM-5 did, and what it did not do to the count
The story usually told is straightforward: DSM-5 loosened the ADHD criteria in 2013, so more children qualified, so the numbers went up. The first half of that is roughly true. The second half is not visible in the data.
Here is the American Psychiatric Association describing its own change [6]:
“The diagnostic criteria for attention-deficit/hyperactivity disorder (ADHD) in DSM-5 are similar to those in DSM-IV. The same 18 symptoms are used as in DSM-IV, and continue to be divided into two symptom domains (inattention and hyperactivity/impulsivity), of which at least six symptoms in one domain are required for diagnosis.”— American Psychiatric Association, Highlights of Changes from DSM-IV-TR to DSM-5, p. 2
Note the order of that sentence. The APA leads with “similar to those in DSM-IV” and “the same 18 symptoms.” What changed sits at the margins of eligibility: the onset criterion moved from before age 7 to before age 12, the adult threshold moved from six symptoms to five, and a comorbid diagnosis with autism became permitted. One change cuts the other way — the cross-situational requirement was strengthened. So “DSM-5 loosened the definition” is a flatter claim than the primary document supports.
And there is a reason to expect a DSM change to reach this particular number when it would not reach others. Transmission is plausible here in a way it is not for survey-algorithm measures: the NSCH and NHIS items ask whether a clinician ever told the parent the child has ADHD, and clinicians are the population that applies DSM criteria. So the change reaches the measured quantity through a real, nameable mechanism. Whether it actually did is a separate question, and the series answers it: no.
Now look at what the series actually did across 2013:
2012: 9.5% → 2013: 8.8% → 2014: 8.9% → 2015: 9.8% → 2016: 9.4% → 2017: 9.4% → 2018: 9.8%
NHIS, parent-reported ever-diagnosed ADHD, ages 3–17 [1]
There is no step. 2013 went down. The series wanders inside roughly one percentage point for the whole 2012–2018 stretch. Whatever DSM-5 did to childhood ADHD diagnosis, it did not produce a discontinuity in the longest clean public series — and no honest annotation of this chart can claim it did.
That is why the DSM-5 marker on the chart above is a thin vertical line the data runs straight through, rather than a shaded band. The two annotation styles mean different things, and the difference is most of the teaching: a band says the ruler changed and you must not read across it; a marker says the thing being counted changed and you should look hard at the slope. Here, we looked, and there is nothing there.
what did move the number
Both survey redesigns produced visible downward steps — against a rising underlying trend, which makes them harder to dismiss as noise:
- NHIS: 9.8% (2018) → 8.8% (2019). A one-point drop at the redesign.
- NSCH: 10.4% (2011) → 9.7% (2016). A 0.7-point drop across five years in which the NHIS series rose.
The Census Bureau, which fields the NSCH, is blunt about what that means — and this sentence is the thesis of this entire page, written by a federal statistical agency [2]:
“An accurate comparison of estimates between the redesigned NSCH and prior surveys cannot be made due to substantive changes in sampling frame and data collection modes, in addition to questionnaire content and wording in many cases. Observed differences in estimates before and after the redesign, or lack thereof, may be attributable to design changes and it is not possible to distinguish the contribution of actual or real changes over time. Therefore, combining or comparing data from before or after the redesign is not recommended.”— U.S. Census Bureau, 2024 NSCH Frequently Asked Questions, p. 11
When the same question was put to Census in 2016 — can we do trend analysis across the redesign? — the published answer began with a single word: No. [2]
NCHS's language about the 2019 NHIS redesign is weaker, and we are going to be precise about that rather than borrow the stronger phrasing [4]:
“Since 1957, the content of the survey has been updated about every 10-15 years to incorporate advances in survey methodology and coverage of health topics. In January 2019, NHIS launched a redesigned content and structure that differs from the 1997-2018 NHIS.”
“These changes to the nonresponse adjustment approach and the calibration methods have the potential to impact comparisons of the weighted survey estimates over time.”— NCHS, 2024 National Health Interview Survey Survey Description, p. 11
A precision point that protects this page. NCHS's own language is a hedge, not a prohibition. A flat NCHS sentence saying 'do not compare 2019+ to 1997-2018' does not exist; it was searched for and not found. The blunt 'should not be directly compared' phrasing belongs to CDC's ADHD programme page, not to NCHS survey documentation. We attribute each quote to the right agency and do not upgrade the hedge.
the third break, which is invisible
The two redesigns are at least visible as steps. There is a third break that leaves no mark in the data at all, and it is the one that should make you most careful about any ADHD statistic you read.
In October 2023 the Census Bureau retroactively re-weighted and re-imputed the 2016–2021 NSCH files. Settled years moved. In its own words [3]:
“Modifications to the data by race and ethnicity are sufficiently substantial that we do not recommend comparing estimates using original data released from 2016 to 2021 against estimates using data released in or after October 2023. … The files released prior to October 2023 are no longer available for download and should be replaced by the enhanced files for future analyses.”— U.S. Census Bureau, NSCH Data File Enhancement Technical Document, 2024-04-10, p. 2
The cleanest demonstration is that CDC now publishes two different values for identical statistics. Its chart file and a CDC-authored paper in a CDC journal give, for the same survey, same ages, same “ever” definition [1][7]:
| year | CDC chart file | Leeb et al. 2024 (CDC journal) |
|---|---|---|
| 2016 | 9.7% | 9.9% |
| 2018 | 9.7% | 9.9% |
| 2020 | 10.1% | 10.2% |
Neither is wrong. They sit on different file vintages. Everything published before roughly 2024 rests on the old weights; everything after rests on the new ones. Most secondary coverage of ADHD prevalence mixes them freely, because the break is invisible unless you go looking for the technical document that announces it.
This is why every figure on this page carries a pull date, and why the underlying data files are checked into version control with their provenance attached. A silently stale graph here would not merely be old — it would be sitting on a superseded weighting scheme.
two surveys, same question, same years, different answers
For three years — 2016, 2017 and 2018 — two federal surveys measured the same construct, in the same age band, at the same time. This is the cheapest possible demonstration that a prevalence number is an estimate with a method attached rather than a fact.
the overlap years: NHIS and NSCH measuring the same thing
Parent-reported ever-diagnosed ADHD, ages 3–17, 2016–2018. Same question, same ages, same years, two federal surveys.
show the numbers
| year | NHIS | NSCH |
|---|---|---|
| 2016 | 9.4% | 9.7% (9.1%–10.3%) |
| 2017 | 9.4% | 9.8% (8.9%–10.7%) |
| 2018 | 9.8% | 9.7% (9%–10.4%) |
A few tenths of a point apart, every year. That gap is not error in either survey. It is what “measured” means.
a second, independent modern series
NCHS publishes a separate childhood ADHD series in its Child Summary Health Statistics, independent of the NSCH [5]. It is a useful corroboration panel, and it produces the same lesson a second time: for 2021 it reports 9.5% where CDC's ADHD chart file reports 9.6% for the same survey year. Both are CDC/NCHS products. We show the discrepancy rather than picking one.
two modern series, plotted side by side
NSCH 2016–2024 and the NHIS Child Summary Health Statistics 2019–2024. Separate surveys, separate lines, no join. Shaded ribbons are the published 95% confidence intervals.
show the numbers
| year | NSCH 2016–2024 | NHIS Child Summary Health Statistics 2019–2024 |
|---|---|---|
| 2016 | 9.7% (9.1%–10.3%) | — |
| 2017 | 9.8% (8.9%–10.7%) | — |
| 2018 | 9.7% (9%–10.4%) | — |
| 2019 | 9.6% (9%–10.3%) | 8.8% (8%–9.5%) |
| 2020 | 10.1% (9.5%–10.7%) | 10.3% (9.2%–11.4%) |
| 2021 | 10.6% (10%–11.2%) | 9.5% (8.8%–10.4%) |
| 2022 | 11.4% (10.9%–12.1%) | 10.2% (9.4%–11.1%) |
| 2023 | 11.4% (10.9%–12%) | 12% (11%–13%) |
| 2024 | 12.6% (12%–13.2%) | 12% (11.1%–13%) |
Within-era comparison is sanctioned, with conditions. Census, on the redesigned NSCH: “Is it possible to conduct time-series trend analyses with NSCH data from 2016 to 2024? Yes, when data sets are using the same weighting methodology.” And from its analytic guide: “Only estimates from the redesigned survey (2016 forward), without substantive changes in question wording, can be compared over time.” Census adds an editorial rule we have adopted: Census adds a rule we adopt: "Changes over time should be sustained over multiple assessments to confirm a change versus a single fluctuation or random error." [2]
One more piece of honesty that cuts against the obvious reading of the rising line: across 2016–2021, the ADHD trend in the NSCH was not statistically significant. CDC-affiliated authors report an average annual percent change of 1.3 (−0.6 to 3.4), P = .13 [7]. In the same table, over identical years and with identical question wording, childhood anxiety and depression did rise significantly. Any account that treats “diagnostic inflation” as one undifferentiated phenomenon is refuted by CDC's own table.
diagnosis rose. the medicated share fell.
If rising diagnosis mechanically drove rising prescribing, this panel would slope upward with the one above it. It does not.
treatment among children who currently have ADHD
Note the denominator: this is a share of children with a current ADHD diagnosis, not a share of all children, and not a prescription count.
show the numbers
| year | taking ADHD medication | receiving behavioral treatment |
|---|---|---|
| 2016 | 57.7% | 45.4% |
| 2017 | 59.1% | 41.6% |
| 2018 | 56.5% | 43.2% |
| 2019 | 54.5% | 48.5% |
| 2020 | 53.1% | 44.2% |
| 2021 | 52.4% | 39.9% |
| 2022 | 49.9% | 41.8% |
| 2023 | 49.7% | 45.9% |
| 2024 | 53.2% | 46.1% |
CDC-affiliated authors state the finding directly, and we quote rather than paraphrase it because it is the single best inoculation this page has against the accusation that it is anti-psychiatry with charts [8]:
“ADHD diagnosis prevalence estimates were stable from 2016 to 2019 (8.6% to 8.8%), rising after the onset of the COVID-19 pandemic in 2020 to 10.5% in 2023. Yet, from 2016 to 2023, overall ADHD treatment rates declined (76.9% to 70.8%), particularly for medication treatment (62.5% to 53.0%), while behavioral treatment stayed steady (47.3% to 48.2%).”— Katz et al., J Clin Child Adolesc Psychol 2026 [8]
These two CDC sources do not agree, and we are not going to hide it. The chart above uses CDC's chart file, which gives medication treatment as 57.7% in 2016 and 49.7% in 2023. The Katz paper — also CDC — gives 62.5% and 53.0% on two-year increments. We could not retrieve a published reconciliation and are not going to invent one. The direction is the same in both, which is the part the argument rests on.
the workforce panel, and the four things it must not be read to say
Provider counts appear here because they are part of the same system's growth, and for no other reason. No causal claim is made in either direction. Both these lines and the ones above rise across the same decades; so does nearly everything measured since 1990.
Read the y-axis label carefully. This is wage-and-salary employment — jobs at surveyed establishments — not practitioners. The survey structurally excludes the self-employed, which is exactly where behavioral health concentrates. For psychiatrists, it reports 24,830 for May 2023 against roughly 46,000–52,000 from AMA Masterfile-based counts. It is capturing about half [9].
behavioral-health wage-and-salary employment, May 1999 – May 2025
Jobs at surveyed establishments, not people and not practitioners. Code changes are marked; the psychiatrist line is drawn as two segments because the occupation code changed in 2019.
show the numbers
| year | psychiatrists (SOC 29-1066) | psychiatrists (SOC 29-1223) | counselors (SOC 21-1018) | marriage and family therapists (SOC 21-1013) | mental health and substance abuse social workers (SOC 21-1023) |
|---|---|---|---|---|---|
| 1999 | 18k | — | — | 19k | 73k |
| 2000 | 21k | — | — | 19k | 80k |
| 2001 | 22k | — | — | 20k | 86k |
| 2002 | 20k | — | — | 22k | 91k |
| 2003 | 20k | — | — | 22k | 97k |
| 2004 | 22k | — | — | 21k | 109k |
| 2005 | 23k | — | — | 19k | 120k |
| 2006 | 25k | — | — | 21k | 115k |
| 2007 | 22k | — | — | 23k | 119k |
| 2008 | 22k | — | — | 25k | 131k |
| 2009 | 22k | — | — | 26k | 127k |
| 2010 | 23k | — | — | 33k | 120k |
| 2011 | 23k | — | — | 34k | 115k |
| 2012 | 24k | — | — | 34k | 110k |
| 2013 | 25k | — | — | 29k | 110k |
| 2014 | 25k | — | — | 30k | 109k |
| 2015 | 24k | — | — | 32k | 110k |
| 2016 | 25k | — | — | 37k | 114k |
| 2017 | 25k | — | 242k | 43k | 112k |
| 2018 | 26k | — | 268k | 49k | 117k |
| 2019 | — | 26k | 284k | 59k | 118k |
| 2020 | — | 26k | 294k | 61k | 117k |
| 2021 | — | 26k | 311k | 55k | 114k |
| 2022 | — | 27k | 345k | 62k | 108k |
| 2023 | — | 25k | 398k | 63k | 115k |
| 2024 | — | 25k | 440k | 66k | 126k |
| 2025 | — | 28k | 492k | 67k | 133k |
The agency that publishes this data does not want you to do this. BLS, verbatim: “The OEWS survey methodology is designed to create detailed cross-sectional occupational employment and wage estimates by geographic area or industry, but it is less useful for looking at changes over time. Challenges in using OEWS data as a time series include changes in the occupational, industry, and geographic classification systems; changes in the OEWS methodology and data collection procedures; and permanent features of the OEWS methodology, including the three-year pooled sample design. The Bureau of Labor Statistics does not encourage the use of OEWS data for time-series analysis. If users choose to make such comparisons, we caution them to note the comparability issues and data limitations described below.” — BLS, Occupational Employment and Wage Statistics FAQ, section F (https://www.bls.gov/oes/oes_ques.htm) [9]
We are doing it anyway, which is defensible only because the warning is displayed here in full rather than buried. If we were not willing to show it, we should not show the chart. Two further constraints ride along: the OEWS estimates pool three years of survey data, so this panel is a moving average by construction and cannot show a sharp response to anything; and the May 2025 psychiatrist jump to 27,980 is a single-year move that should be read as provisional, not as growth.
What the panel does show is a shape: the prescribing-physician workforce did not grow, and the workforce around it did. That is more interesting than “provider counts went up,” and it is a statement about payroll jobs, not about practitioners.
One line the brief asked for is missing, and its absence is a finding. OEWS does not break out psychiatric nurse practitioners at any aggregation level, in any year. We verified that by enumerating every occupation code in the May 2024 and May 2025 national files. The federal statistical system does not count that workforce separately, so we are not going to approximate it [9].
the panel that is deliberately empty: adults
no continuous public series exists for adult ADHD diagnosed prevalence
Checked source by source: NSDUH has no ADHD diagnosis variable and no ADHD screen anywhere in the adult instrument. BRFSS has no ADHD item, verified across 1998–2024. NHANES dropped its childhood item after 2004 and never had an adult module. MEPS covers ages 5–17. NSCH covers ages 0–17. The NCHS Rapid Surveys System carried ADHD content in one round only. NHIS added adult ADHD questions in 2026; nothing is published yet.
The two defensible adult figures cannot be joined and are not drawn as a line: 6.0% of adults reported a current ADHD diagnosis in October–November 2023 [10], and 4.4% of adults aged 18–44 met 12-month criteria in 2001–2003 under a structured interview. Those are two different instruments measuring two different things across two different age ranges twenty years apart. Drawing a line between them would be indefensible.
The asymmetry is the payload. Childhood ADHD has a 22-year annual series. Adult ADHD has essentially nothing — at precisely the moment when adult diagnosis and adult stimulant prescribing are the fastest-moving part of the story. What gets counted reflects what the system decided to look at.
what this page does not show
- That ADHD became more common. These series measure diagnostic behaviour — whether a clinician ever said so — not disease incidence.
- That the additional diagnoses are wrong. The DSM-5 changes were argued toward accuracy, with under-diagnosis of adults and of children with autism as the stated rationale.
- That DSM-5 caused the trend. The rise long predates 2013 and continues after it, and there is no step at 2013.
- That rising diagnosis drove rising prescribing. CDC's own data contradicts it.
- That provider growth caused diagnosis growth, or the reverse. No causal claim is made in either direction, and the panels share nothing but an x-axis.
- Anything about you or your child. This is a page about a statistic.
how this page is built
Every series is a checked-in JSON file under data/system-education/, each carrying its source URL, its pull date, its measurement-type label, and the publishing agency's own comparability warning as text. The annotation layer — which breaks exist, which register each one belongs to, and whether a DSM change actually reached the instrument doing the counting — is a separate versioned file, because DSM-5-TR now receives criteria updates between editions (most recently September 2025) and a hardcoded marker set would rot.
The charts are server-rendered inline SVG with no charting library and no client JavaScript. That was a deliberate choice: the rule that two segments must never be joined is easier to guarantee structurally in our own code than to configure in someone else's.
Retrieval note, for anyone reproducing this. www.cdc.gov returns HTTP 403 to non-browser clients. CDC's data pages render their charts from JSON config files named in the page HTML, and those files carry the complete underlying year-by-year table — that is how the full 1997–2024 series here was obtained, via the Internet Archive. ftp.cdc.gov is not blocked and serves NCHS documentation directly.
questions worth asking
Did DSM-5 cause the rise in childhood ADHD diagnoses?
The longest clean series says no. Parent-reported ADHD in the National Health Interview Survey went 9.5% in 2012, then 8.8% in 2013 — the DSM-5 publication year — then wandered between 8.8% and 9.8% through 2018. There is no step at the definition change. The rise long predates 2013 and continues after it.
What did move the number?
Survey redesigns, visibly and every time. NHIS dropped from 9.8% (2018) to 8.8% (2019) at its redesign. NSCH dropped from 10.4% (2011) to 9.7% (2016) at its redesign — across five years during which the NHIS series rose. A third break is invisible in the data: an October 2023 re-weighting of the 2016–2021 files that shifted settled years.
Does this mean more children have ADHD?
The chart cannot tell you that. It measures diagnostic behaviour — whether a clinician ever told a parent their child has ADHD — not disease incidence. It is also an "ever" measure, which drifts upward mechanically as diagnosis becomes more common at younger ages.
Did rising diagnosis drive rising prescribing?
CDC’s own data contradicts that. Among children with current ADHD, the share taking medication fell from 57.7% in 2016 to 49.7% in 2023 before rising to 53.2% in 2024, while diagnosis rose across the same window. Diagnosis and medication moved in opposite directions for most of that period.
Why is there no adult ADHD chart here?
Because no US federal survey has tracked adult ADHD diagnosis year over year. We checked NSDUH, NHIS, BRFSS, NHANES, MEPS and the Rapid Surveys System. NHIS added adult ADHD questions in 2026 and nothing is published yet. The absence is a finding about the measurement system, not a gap in our research.
Why do two federal surveys give different answers for the same year?
Because a prevalence figure is an estimate with a method attached, not a fact read off nature. In 2016, 2017 and 2018 both NHIS and NSCH measured parent-reported ever-diagnosed ADHD in ages 3–17. They got 9.4/9.4/9.8 and 9.7/9.8/9.7. Close, and not the same.
this is not medical advice, and it is emphatically not a reason to question a diagnosis or stop a medication. a number being constructed does not make a condition unreal, and nothing on this page tells you anything about a specific child or adult. stimulant medication in particular should not be stopped or restarted without a prescriber — changes in dose or timing can affect sleep, appetite, mood and blood pressure. if this page raises a question for you, bring it to the appointment. if you're in crisis, call or text 988 (u.s.), 24/7, free.
sources
- Centers for Disease Control and Prevention. "Trends in ADHD Among U.S. Children." Chart data file ADHD-diagnosis.json (dataFileName: "Prevalence of parent-reported ADHD diagnosis.csv"). Published 2024-05-15; updated 2026-07-07. Source of all four diagnosis segments and of CDC's own four-way segmentation. Retrieved 2026-08-25 via the Internet Archive, because www.cdc.gov returns HTTP 403 to non-browser clients. https://www.cdc.gov/adhd/data/trends-in-adhd-US-children.html
- U.S. Census Bureau. "2024 National Survey of Children's Health Frequently Asked Questions," p. 11, and "2016 NSCH Frequently Asked Questions," p. 8. Source of the redesign comparability language quoted verbatim on this page, and of the within-era permission for 2016 onward. Retrieved 2026-08-25. https://www.census.gov/programs-surveys/nsch.html
- U.S. Census Bureau. "NSCH Data File Enhancement Technical Document," 2024-04-10, p. 2. Source of the October 2023 re-weighting and re-imputation notice covering the 2016–2021 files. Retrieved 2026-08-25. https://www.census.gov/programs-surveys/nsch/technical-documentation.html
- National Center for Health Statistics. "2024 National Health Interview Survey Survey Description," p. 11 (identical framing in the 2019 edition). Source of the 2019 redesign language. Note: a flat NCHS sentence prohibiting comparison across the redesign does not exist; the blunt "should not be directly compared" phrasing belongs to CDC's ADHD programme page, not to NCHS survey documentation. Retrieved 2026-08-25 from ftp.cdc.gov. https://ftp.cdc.gov/pub/Health_Statistics/NCHS/Dataset_Documentation/NHIS/2024/srvydesc-508.pdf
- National Center for Health Statistics. NHIS Child Summary Health Statistics, "Ever having attention-deficit/hyperactivity disorder," children aged 3–17, 2019–2024. A second published series, independent of NSCH. Question verbatim: "Has a doctor or other health professional ever told you that [child's name] had attention deficit hyperactivity disorder (ADHD) or attention deficit disorder (ADD)?" Retrieved 2026-08-25. https://www.cdc.gov/nchs/nhis/shs/tables.htm
- American Psychiatric Association. "Highlights of Changes from DSM-IV-TR to DSM-5," p. 2. Source of the ADHD criteria passage quoted verbatim. Retrieved 2026-08-25. https://www.psychiatry.org/File%20Library/Psychiatrists/Practice/DSM/APA_DSM_Changes_from_DSM-IV-TR_-to_DSM-5.pdf
- Leeb RT, Danielson ML, et al. "Prevalence of Mental, Behavioral, and Developmental Disorders Among U.S. Children, 2016–2021." Prev Chronic Dis 2024;21:240142. PMCID PMC11640887. A CDC journal, 7 of 10 authors CDC/HRSA staff. Source of the ADHD average annual percent change of 1.3 (−0.6 to 3.4), P = .13 — not statistically significant — and of the values that differ from CDC's chart file for identical years. Retrieved 2026-08-25 via NCBI E-utilities. https://doi.org/10.5888/pcd21.240142
- Katz SM, Claussen AH, Alberto CK, et al. "Trends in Parent-Reported ADHD Diagnosis and Treatment Among U.S. Children, 2016–2023." J Clin Child Adolesc Psychol 2026;55(4):688–703. PMID 41686140. CDC (NCBDDD) authors. Quoted verbatim on this page for the treatment finding. Retrieved 2026-08-25 via NCBI E-utilities. https://doi.org/10.1080/15374416.2026.2620385
- US Bureau of Labor Statistics. Occupational Employment and Wage Statistics, national cross-industry estimates, May 1999 – May 2025, and the OEWS FAQ section F ("The Bureau of Labor Statistics does not encourage the use of OEWS data for time-series analysis"). OEWS counts wage-and-salary jobs and excludes the self-employed. Retrieved 2026-08-25. https://www.bls.gov/oes/oes_ques.htm
- Staley BS, Robinson LR, Claussen AH, et al. "Attention-Deficit/Hyperactivity Disorder Diagnosis, Treatment, and Telehealth Use in Adults — National Center for Health Statistics Rapid Surveys System, United States, October–November 2023." MMWR Morb Mortal Wkly Rep 2024;73(40):890–895. The one-round source of the adult self-reported current-diagnosis point estimate of 6.0%. Retrieved 2026-08-25. https://doi.org/10.15585/mmwr.mm7340a1
related on resolv
- where diagnoses come from — the DSM, in the words of the people who ran it
- who funds the advocates? — pharma money to advocacy organisations, from public records
- how long were these drugs actually tested?
- what FDA approval actually means — and what it doesn't