Industries / Health / Healthcare / Behavioral health
Access and waiting lists
Getting people from needing care to receiving it: how many clinicians there are, where care is offered, who can deliver it, and how long people wait.
6
Milestones achieved, from 1963 to 2025.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1963
Asylums emptied, centers half built
The US Community Mental Health Act of 1963 funded three years of federal grants, $150 million in all, to build and staff 1,500 community mental health centers in place of large state hospitals. States built about half of them. The number of people in state mental hospitals fell by over 90% from its 1955 peak of 558,922 by the early 2000s, and many people without homes ended up on the street or circling between hospitals, jails and prisons.
LessonClosing hospital beds without paying for the community care meant to replace them moved people into jails and onto the street.
Credit
President John F. Kennedy and the US Congress
2002
Depression care in primary care
In the IMPACT trial, 1,801 adults aged 60 and over with depression at 18 primary care clinics in 5 US states got either usual care or a depression care manager, supervised by a psychiatrist and a primary care expert, who supported the family doctor's treatment. After 12 months, 45% of the care-manager group had their symptoms at least halved, against 19% in usual care. The approach is called collaborative care.
Credit
Jürgen Unützer, Wayne Katon and the IMPACT investigators
2008
Talking therapy at national scale
England's Improving Access to Psychological Therapies program, now NHS Talking Therapies, began in 2008 to train new therapists in treatments with research support and to publish outcomes for nearly every patient. In 2025-26 it received 1.81 million referrals, 1.17 million of which reached the service; 89.1% were seen within 6 weeks, 674,683 finished a course of treatment, and 50.1% of eligible referrals moved to recovery, by NHS England's count.
Credit
NHS England; David M. Clark, University of Oxford, who described the program's design
2016
Lay workers on a bench
In Harare, Zimbabwe, trained and supervised lay health workers gave six sessions of problem-solving therapy on a "Friendship Bench" at primary care clinics. In a trial across 24 clinics and 573 patients, 13.7% of the bench group had depression symptoms above the study's cutoff after 6 months, against 49.9% of the control group. Moving work like this from specialists to trained lay people is called task-shifting.
Credit
Dixon Chibanda and colleagues
2023
Ghost networks counted
Staff of the US Senate Finance Committee's majority called 120 mental health providers listed in the directories of 12 health plans in 6 states, posing as relatives of an older adult with depression. They could book an appointment only 18% of the time; 33% of the listings had non-working or wrong numbers or calls that were never returned. A 2020 national survey had found that 53% of people who used a mental health directory met an error in it.
LessonA plan's list of in-network clinicians is not evidence that care is available.
Credit
US Senate Committee on Finance majority staff; Susan Busch and Kelly Kyanko for the survey
2025
A world count of carers
The World Health Organization's reports World mental health today and Mental Health Atlas 2024 found a global median of 13 mental health workers per 100,000 people, and a median government spend on mental health of 2% of health budgets, unchanged since 2017. Fewer than 10% of countries had fully moved to care in the community. In low-income countries fewer than 10% of people with psychosis receive care, against over 50% in higher-income countries, from the 22 countries with enough data.
Credit
World Health Organization
Watch for
Chatbots at the front door
NICE, England's health technology assessor, allowed two AI chatbots that take self-referrals for NHS Talking Therapies, Limbic Access and Wysa Digital Referral Assistant, to be used for three years while more evidence is gathered. Limbic says its tool saves 12.7 minutes per assessment and Wysa says 16 to 21 minutes. In a study of 129,400 patients written by Limbic staff, services using Limbic's chatbot saw referrals rise 15%, against 6% in other services, with larger rises among nonbinary and ethnic minority people. NICE noted that most of the evidence came from routine service data or the companies' own reports, and that none of Wysa's was comparative.
Achieved whenstudies not run by the companies show time saved and wider access, and NICE recommends routine use at the end of the evidence period.
Claimed by
Limbic and Wysa, as reported by NICE, July 2025
One national provider directory
The US Centers for Medicare & Medicaid Services (CMS) is building a single national directory of doctors and hospitals with up-to-date contact and insurance details, which records software vendors and insurers can update, and expects a "beta launch" later in 2026. A CMS official told HFMA it "can serve as a source of truth." A public federal bulk file already exists, and one open-source project checks it against other federal records.
Achieved whenan outside check finds that most mental health listings in the directory are reachable and in network.
Claimed by
CMS, February and March 2026
Open
01
Ghost listings, plan by plan
A public count, for each health plan, of mental health and substance use listings in its public provider directory that point to no one: clinicians whose national provider number has been deactivated, whose state license has lapsed, or who are on the federal exclusion list. Publish the count plan by plan every quarter. Use only public data; do not call providers.
The briefCredit · unclaimed02
Outside check on referral bots
An independent analysis, from England's published NHS Talking Therapies data, of what happened to referrals, waits and recovery in services after they adopted a referral chatbot (Limbic Access or Wysa Digital Referral Assistant), compared with services that did not, using a difference-in-differences design, which compares the change in adopting services with the change in others over the same months.
The briefCredit · unclaimed
Open 01 · the brief
Ghost listings, plan by plan
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public count, for each health plan, of mental health and substance use listings in its public provider directory that point to no one: clinicians whose national provider number has been deactivated, whose state license has lapsed, or who are on the federal exclusion list. Publish the count plan by plan every quarter. Use only public data; do not call providers.
Why it is still open
Directory errors have been shown by secret-shopper calls, surveys and one claims study: Jane Zhu and colleagues found in 2022 that 58.2% of listings in Oregon Medicaid plans' directories, and 67.4% of mental health prescribers, were "phantom" providers who saw no Medicaid patients in 2018. That method needs claims data that only states and plans hold, and no regular, public plan-by-plan check from public data was found. AINPI, an open-source checker released in October 2026, tests only the federal directory file; it says plans analyze their own directories, and it publishes no mental health results.
Built with what exists
The Provider Directory APIs that CMS has required Medicare Advantage, Medicaid and CHIP plans to make public since 1 January 2021; the national provider number file from NPPES; state license lookups; the federal exclusion list (OIG LEIE); AINPI's Apache-2.0 code; Python.
First step
Pull the mental health listings from five Medicaid managed care plans' Provider Directory APIs in one state and match them to the NPPES file.
Done looks like
At least 20 plans in at least 3 states; for each plan, the number of mental health listings checked and the share failing each test, with 95% confidence intervals; code and method published; the run repeated 90 days later to show which errors persist.
Open 02 · the brief
Outside check on referral bots
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An independent analysis, from England's published NHS Talking Therapies data, of what happened to referrals, waits and recovery in services after they adopted a referral chatbot (Limbic Access or Wysa Digital Referral Assistant), compared with services that did not, using a difference-in-differences design, which compares the change in adopting services with the change in others over the same months.
Why it is still open
The main published study of a referral chatbot's effect on access was written by the maker's staff, and NICE asked in July 2025 for more evidence over three years because most of what it saw was routine or company-reported data. A search of Crossref found no outside replication.
Built with what exists
NHS Talking Therapies monthly statistics, whose activity data file gives sub-national figures; adoption dates gathered from NHS trust board papers, public tender notices and company announcements; R or Python with an open-source package for staggered difference-in-differences.
First step
Compile a dated list of NHS Talking Therapies services that use either chatbot from public sources, and download two years of monthly activity files.
Done looks like
At least 20 adopting services with a dated start; effects on referral volume, waits and recovery, with 95% confidence intervals, set beside the maker's reported 15% against 6%; the adoption list, data and code published so that someone else gets the same estimates.