# Verification log — UK aphasia provision evidence base (card t_592b3c4a)

Verification wave over the three parent lanes (L1 treatments t_5d9a6009, L2 services t_4db1dbca,
L3 barriers t_7e7e7ab2). Each load-bearing number was re-fetched at source on 2026-10-16. Verdicts:
**CONFIRMED** (source states the figure), **CORRECTED** (figure or attribution wrong, replacement given),
**KILLED** (claim not supported at all), **UNVERIFIABLE** (no primary-source access this run).

Totals: 70 items checked. 55 CONFIRMED, 4 CORRECTED (3 distinct corrections; rows C6 and C7 are the same defect), 1 KILLED, 10 UNVERIFIABLE.
(CORRECTED/KILLED items are enumerated with replacements in kill_list.json; UNVERIFIABLE items and the
confidence that should attach to them are in residuals.md.)

## A. Lane 1 — treatments (t_5d9a6009)

| # | Claim as stated | Cited source | What the source says now | Verdict | Quote (verbatim) |
|---|---|---|---|---|---|
| A1 | Palmer 2018 surveyed 278 community-dwelling stroke survivors with aphasia in 21 UK SLT departments, 2014–2016 | [3] PLOS ONE e0200096 | "278 patients in 21 UK speech and language therapy departments in 2014–2016" | CONFIRMED | "This paper revisits provision based on what was received by 278 patients in 21 UK speech and language therapy departments in 2014–2016." |
| A2 | 45% of participants were in receipt of therapy in the three-month window | [3] | "Forty-five percent of participants were in receipt of therapy in the three month window observed." | CONFIRMED | as quoted |
| A3 | Median amount of therapy in three months was 6.3 hours | [3] | "The median amount of therapy received in three months was 6.3 hours at an average of one 60-minute session every two weeks." | CONFIRMED | as quoted |
| A4 | 77% of sessions delivered by qualified SLTs, 23% by assistants | [3] | "Seventy-seven percent of therapy sessions were delivered by qualified speech and language therapists and 23% by assistants." | CONFIRMED | as quoted |
| A5 | Therapy falls with time post-stroke: ~11 h/3 months at 1–3 months to 2.75 h at >5 years | [3] | "those between one and three months post stroke receiving 11 hours of therapy … whilst those more than five years post stroke received 2.75 hours of therapy" | CONFIRMED | as quoted |
| A6 | Vault synthesis reported Palmer 2018 n=114 | vault synthesis-aphasia-therapy-evidence | Published paper reports 278, not 114 | KILLED | see A1 quote |
| A7 | Monnelly 2023 e-survey of 227 UK SLTs | [18] Wiley 1460-6984.12918 | "Two hundred twenty-seven respondents engaged in the e-survey." | CONFIRMED | as quoted |
| A8 | Mean therapy delivered 128 min/week | [18] | "The average amount of therapy provided per person with aphasia per week was 128.3 min (SD 86.5)" | CONFIRMED | as quoted |
| A9 | 37% of SLTs defined intensive therapy at/above the NICE standard | [18] | "37% of SLTs defined IAT in ways that met or exceeded NICE guidelines." | CONFIRMED | as quoted |
| A10 | 16.5% felt their service could be reconfigured to deliver an ICAP; 50% aware; 15 involved | [18] | "50% of respondents were aware of ICAPs and 15 had been involved in ICAP provision. Only 16.5% felt their service could be reconfigured to deliver an ICAP." | CONFIRMED | as quoted |
| A11 | Most-delivered approaches: functional language 85%, impairment-based 70% | [18] | "the majority delivered functional language therapy (85%), impairment-based therapy (70%) and involved family and others (58%)" | CONFIRMED | as quoted |
| A12 | Big CACTUS randomised 278 people across 21 UK NHS SLT departments | [4] PMC6700375 | "278 (34%) participants were randomly assigned" from 21 SLT departments | CONFIRMED | "From Oct 20, 2014, to Aug 18, 2016, 818 patients were assessed for eligibility, of which 278 (34%) participants were randomly assigned" |
| A13 | Word-finding improvement 16.2% over usual care | [4] | "Word finding improvement was 16·2% (95% CI 12·7 to 19·6; p<0·0001) higher in the CSLT group than in the usual care group" | CONFIRMED | as quoted |
| A14 | No benefit for conversation or quality of life | [4] | "did not result in an improvement in conversation"; TOMs difference −0·03 (p=0·709) | CONFIRMED | "CSLT plus usual care resulted in a clinically significant improvement in personally relevant word finding but did not result in an improvement in conversation." |
| A15 | Big CACTUS: 20–30 min of daily practice over six months, volunteers/assistants ≥1 h/month | [4] | Not stated in the retrieved abstract/full-text extract; the intervention description is in the paper's TIDieR table | UNVERIFIABLE | n/a |
| A16 | LUNA: 20 h total, 2 sessions/week of 60 min, over 10 weeks | [5] PMC11178191 | "LUNA treatment was delivered twice weekly for 10 weeks via the videoconferencing technology, Zoom" (20 sessions × 1 h = 20 h) | CONFIRMED | as quoted |
| A17 | LUNA: 85% consent, 86% retention, 92% fidelity | [5] | "85% of those eligible consented to the trial; trial retention was 86%; … treatment fidelity was high at 92% adherence" | CONFIRMED | as quoted |
| A18 | TeleCHAT: 50 h over 8 weeks; composition 14+14+14+8 h; 9/12 full dose | [14] PMC12615849 | "50 h of goal-directed aphasia therapy"; "14 h each of impairment, functional, and computer-based therapy, and 8 h of psychosocial group therapy, delivered over 8 weeks"; "Nine participants received the intended dose of 50 h" | CONFIRMED | as quoted |
| A19 | SUPERB: six 1-hour visits over three months, plus two optional sessions | [12] PMC8273536 / [21] PMC8330561 | "Six 1-hour peer-befriending visits over three months"; "Two optional sessions" | CONFIRMED | "Six 1-hour peer-befriending visits over three months." |
| A20 | SUPERB: 10 peer-befrienders, trained 4–6 h | [21] PMC8330561 | "10 peer-befrienders at least 1-year poststroke"; "All peer-befrienders received 4–6 hours training over 2–3 days as intended." | CONFIRMED | as quoted |
| A21 | SUPERB: ~£57.24 per visit; cost-effective in 39–66% of simulations | [6] PMC8941719 | "£57.24 (including training and supervision costs). The probability of peer-befriending being cost-effective ranged 39% to 66%." | CONFIRMED | as quoted |
| A22 | ASK: 373 participants consented | [7] PMC10507993 | "screened 1744 people with aphasia and 373 participants consented to intervention" | CONFIRMED | as quoted |
| A23 | ASK: only 41 of 171 who received treatment met the minimum dose | [7] | "Of those 171 who did receive treatment, only 41 met the prescribed minimum dose." | CONFIRMED | as quoted |
| A24 | ASK: null for mood | [7] | "ASK showed no benefit over attention control in improving mood and preventing depression" | CONFIRMED | as quoted |
| A25 | Cochrane 2016: 57 RCTs, 3002 participants, SMD 0.28 (0.06–0.49) | [17] CD000425.pub4 | "We included 57 RCTs … involving 3002 participants"; "SMD 0.28, 95% confidence interval (CI) 0.06 to 0.49" | CONFIRMED | as quoted |
| A26 | NICE NG236 1.2.16: at least 3 h/day, at least 5 days/week | [19] NICE NG236 | "for at least 3 hours a day, on at least 5 days of the week" | CONFIRMED | as quoted |
| A27 | BCPPA plus: 4–6 × 1-hour sessions; prior BCPPA 4 × 1 h face-to-face | [15] PMC13475093 | "four to six 1-hour sessions"; "delivered face to face over four 1-hour sessions" | CONFIRMED | "refined to be delivered online via telehealth in four stages over four to six 1-hour sessions" |
| A28 | BCPPA plus: >80% of delivery by a highly specialist HCPC-registered SLT | [15] | "the majority (over 80%) of the intervention will be delivered by the RA" (RA is a HCPC-registered SLT) | CONFIRMED | as quoted |
| A29 | BCPPA plus provider: "…or a trained assistant under supervision" | [15] | Protocol's alternative deliverer is a local collaborator SLT, not a trained assistant | CORRECTED | "An RA…or…local collaborator SLT" (see kill_list.json V3) |
| A30 | Better Conversations intensive groups: 26 h total, 2.5–3 h/day, 4 days/week over 2 weeks | [13] PMC13425045 | "Participants received a total of 26 h of Better Conversations CPT, delivered for 2.5–3 h per day, over 2 weeks (4 days a week)." | CONFIRMED | as quoted |
| A31 | UK ICAP RCT NCT07386652: 334 participants; not yet recruiting; completion 2030 | [16] CT.gov API | "ENROLL count 334 (estimated)"; status NOT_YET_RECRUITING; primary completion 2030-03-31 | CONFIRMED | registry record |
| A32 | NCT07386652: 4-week 100-hour ICAP | [16] | Registry arm/outcome fields retrieved do not state a 4-week/100-hour dose | UNVERIFIABLE | n/a |
| A33 | CILT: original model ~30 h over 2 weeks | StepByStep manual (L1 ref: StepByStep manual) | PDF page 19 needed OCR; not re-read at source this run | UNVERIFIABLE | n/a |
| A34 | COMPARE/CIAT-Plus: 15 h/week × 2 weeks vs 6 h/week × 5 weeks | L1 ref Pierce 2024 (abstract-only) | Abstract-only; dose specifics not re-verified. Arithmetic is internally consistent (both = 30 h) | UNVERIFIABLE | n/a |
| A35 | Monnelly 2024 SR: 21 studies, 13 ICAPs, N=485 | L1 ref Monnelly 2024 (abstract-only) | Paywalled abstract; not re-verified | UNVERIFIABLE | n/a |

## B. Lane 2 — services (t_4db1dbca)

| # | Claim as stated | Cited source | What the source says now | Verdict | Quote (verbatim) |
|---|---|---|---|---|---|
| B1 | 92,414 admissions, 250 hospitals, 195 community services, 184 six-month follow-up providers (2024/25) | [1] SSNAP SoN 2025 | "92,414 patients were admitted across 250 hospitals, 195 community services and 184 six-month follow up providers" | CONFIRMED | as quoted |
| B2 | 66.6% discharged to a stroke/neurology-specific community rehabilitation service | [1] | "66.6% of patients were discharged to a stroke/neurology specific community rehabilitation service" | CONFIRMED | as quoted |
| B3 | 23.1% discharged to a combined ESD-CRT service | [1] | "23.1% of patients were discharged to a stroke/neurology specific combined ESD-CRT service" | CONFIRMED | as quoted |
| B4 | 35.1% six-month review (2024/25); 39% in 2023/24 | [1] | "35.1% of applicable patients received a 6 month follow-up"; "declined further to 35%, from 39% in 2023/24" | CONFIRMED | as quoted |
| B5 | Six-month review mode: 14.7% in person, 18.8% telephone, 0.4% post | [1] | Figure 10, 2024/25 row: in person 14.7, telephone 18.8, post 0.4 (online blank) | CONFIRMED | Figure 10 |
| B6 | ~one fifth of patients report moderate-to-severe mood issues at six months | [1] | "approximately one fifth of patients report moderate to severe mood issues" | CONFIRMED | as quoted |
| B7 | ESD/CRT discharge growth 41% (2013/14) → 66.5% (2024/25) | [1] | Figure 9: 2013/14 = 41; 2024/25 = 66.5 | CONFIRMED | Figure 9 |
| B8 | RCSLT vacancy: 14% UK, 13% NHS, 14% NHS adult, 10% children's, 24% learning disabilities, 23% independent | [2] RCSLT vacancy survey | all six figures stated on the page | CONFIRMED | "24% of posts for SLTs in the NHS supporting people with learning disabilities were vacant." |
| B9 | RCSLT by nation: England 17%, London 21%, Scotland 12%, NI 13%, Wales 6% | [2] | all five stated | CONFIRMED | as quoted in source |
| B10 | 23,590 adults waiting for SLT in England; 7,281 (30.9%) over 12 weeks (end Nov 2024) | [9] HoC Library CDP-2025-0010 | "23,590 adults were waiting for speech and language therapy. Of these, 7,281 (or 30.9%) had been waiting over 12 weeks." | CONFIRMED | as quoted |
| B11 | Waiting >12 weeks: London 37.8%, East of England 18.7% | [9] | Table: London 37.8%, East of England 18.7% | CONFIRMED | as quoted |
| B12 | No ringfenced central funding for SLT | [9] | "There is no ringfenced central funding for speech and language therapy" | CONFIRMED | as quoted |
| B13 | Post-acute audit: 664 services (591 England, 42 Wales); over 6,700 WTE staff | [8] Post-acute Organisational Audit 2021 | "A total of 664 post-acute services participated…of which 591 were in England, 42 in [Wales]"; "over 6,700 whole-time equivalents (WTE)" (total 6,710.85) | CONFIRMED | as quoted |
| B14 | Fewer than one third of community-based teams met recommended staffing levels | [8] | "Fewer than a third of community-based multidisciplinary teams met recommended staffing levels." | CONFIRMED | as quoted |
| B15 | Recommended SLT establishment in community teams: "0.4 WTE per unit" | [8] | Table 3.3: SLT = 0.4 WTE **per 100 patients** | CORRECTED | "Speech and language therapist | 0.4 | …" (staffing levels: WTE per 100 patients) |
| B16 | 17% of community services have appropriate access to all core disciplines | [8]/[SA] | KI3 "Team has appropriate access to all relevant disciplines": 16.7% community (41), 18.3% inpatient (17) | CONFIRMED | Table 3.2 / KI3 |
| B17 | PLOS ethnicity analysis: 80,413 of 307,054 (26.2%) attended six-month review | [11] PLOS 0354353 | "A total of 80,413 (26.2%) stroke-survivors attended their 6-month review" | CONFIRMED | as quoted |
| B18 | CHSS: n=250 people with aphasia (summer 2023); 50% could not access needed support; 24% of unmet need was more SLT; 67% difficulty accessing services; 52% mental health affected; 8% accessed psychological support; 96% white | [10] CHSS "1 in 5" | "250 responses were from people living with aphasia"; "HALF (50%) said there was support they needed but weren't able to access"; "for more speech and language therapy (24%)"; "TWO THIRDS (67%) … difficulty in accessing services"; "HALF (52%) … affects their mental health, but only 8% have accessed psychological or social support"; "96% of the sample were white" | CONFIRMED | as quoted |
| B19 | Stroke Association reach 2024-25: >66,000 reached; ~38,000 via local services; ~15,000 helpline; £2.4m research (charity self-report) | [23] SA annual report | "over 66,000 people through all our support services"; "nearly 38,000 people through our local support services"; "around 15,000 enquiries"; "£2.4 million invested" | CONFIRMED | as quoted |
| B20 | SSNAP vault point values: SLT assessment within 72 h 46.8%; within 24 h 24.8%; no assessment 7.6%; NICE 128 min/day dose 35.4% | vault source-ssnap-apt-synthesis-2026 | SSNAP annual results portfolio (XLSX) not downloaded; values not re-derived | UNVERIFIABLE | n/a (lane already flagged [UNVERIFIED]) |
| B21 | Say Aphasia is CIO 1169933 | [24]/[25] | Charity Commission: "SAY APHASIA / Charity number: 1169933" | CONFIRMED | as quoted |
| B22 | Northcott 2016 UK SLT survey n=124; 99% social participation; 50% 70-100% psychological difficulty | L2 ref Northcott 2016 (abstract) | Abstract-only; not re-fetched | UNVERIFIABLE | n/a |
| B23 | Hilari telehealth: >80% UK SLTs used telehealth in COVID; >90% planned to continue | L2 ref Hilari telehealth | Not re-fetched at source this run | UNVERIFIABLE | n/a |
| B24 | Aphasia Alliance: 13 member organisations | L2 ref Aphasia Alliance list | Not re-fetched at source this run | UNVERIFIABLE | n/a |
| B25 | Six-month review: SA "39%" vs SSNAP "35.1%" | [1]/[8] | Source anomaly is real (SA 10-Year-Plan response says 39% for 2024-25; SSNAP reports 35.1% and 39% for 2023/24) | CONFIRMED (as flagged) | lane records the contradiction, does not average it |
| B26 | Stroke Association 434 groups / 13,000 reached / 480 interpreting users | Charity Commission filing description (metadata only) | Metadata-level only; not re-fetched | UNVERIFIABLE | n/a |
| B27 | Care-home provision: audit section 4.5 exists; figures not extracted | [8] | Section 4.5 "Treatment of patients in care homes" present | CONFIRMED (section exists) | table of contents, Section 4.5 |

## C. Lane 3 — barriers (t_7e7e7ab2)

| # | Claim as stated | Cited source | What the source says now | Verdict | Quote (verbatim) |
|---|---|---|---|---|---|
| C1 | 64 barriers; 56 documented, 8 inferred | cfir_eric_map.csv, coverage.json | Recomputed from the CSV: 64 rows; documented 56, inferred 8 | CONFIRMED | computed counts |
| C2 | evidence_standing: validated_uk 0, validated_other 8, mapped_only 47, untested 9 | coverage.json | Recomputed: 0/8/47/9, sums to 64 | CONFIRMED | computed counts |
| C3 | 40 distinct ERIC strategies used | cfir_eric_map.csv | Recomputed: 40 distinct strategies | CONFIRMED | computed |
| C4 | Every ERIC strategy is from the published 73-strategy set | [22] ERIC 73 appendix | All 40 names match the published ERIC taxonomy verbatim (e.g. "Access new funding", "Alter incentive/allowance structures", "Prepare patients/consumers to be active participants", "Use mass media", "Change service sites") | CONFIRMED | published list [22] |
| C5 | ERIC cluster labels | [22] | Nine clusters match Waltz/Powell clusters ("Use evaluative and iterative strategies" … "Change infrastructure") | CONFIRMED | as quoted |
| C6 | "64 barriers stated in the vault's CFIR 2.0 vocabulary" | L3 completion metadata | The construct vocabulary is a CFIR 1.0/2.0 hybrid plus non-CFIR (COM-B) labels; ~32 of 64 rows use a label that is not a CFIR 2.0 construct name | CORRECTED | see kill_list.json V4 |
| C7 | "Deliverers: Capability (COM-B) / Knowledge", "Recipients: Opportunity (COM-B)", "Deliverers: Capability / Skills", "Recipients: Motivation / Beliefs about Consequences" | cfir_eric_map.csv | COM-B and TDF terms are not CFIR constructs (CFIR 2.0 Individuals.Characteristics = Need, Capability, Opportunity, Motivation) | CORRECTED | cfirguide constructs [20] |
| C8 | "…that distribution is the finding" (no mapped strategy has UK effectiveness evidence) | coverage.json + cfir_eric_map.csv | validated_uk = 0 is consistent with the file; the statement is an internal count, not an external claim | CONFIRMED (as internal count) | computed |

## D. Three load-bearing claims attacked

The card asked for the three claims that would most change the conclusions if wrong. These were chosen
and attacked directly.

**D1. "Routine UK provision sits far below any effective dose."** Load-bearing because the entire
provision-gap argument rests on it. Attack: re-fetched the primary source (Palmer 2018 [3]) and the
guideline standard (NICE NG236 [19]) independently. Source says median 6.3 h/3 months and ~one 60-min
session per fortnight; NICE says ≥3 h/day × 5 days/week (≈15 h/week). The gap is real and large. The
one soft point: the 6.3 h median is for the subset with time recorded (n=120), not all 278; the paper
states this. Verdict: CONFIRMED, with the denominator caveat recorded. This claim survives.

**D2. "The six-month review rate is falling."** Load-bearing for the "provision drops away after
discharge" reading. Attack: re-fetched SSNAP SoN 2025 [1]. Report states 35% in 2024/25 against 39% in
2023/24, "continuing a concerning downward trend since 2019/20". Figures 10 (mode table) support the
decline. A contradictory charity figure (39% for 2024-25) exists and the lane already flags it rather
than resolving it. Verdict: CONFIRMED. The direction (down) and both point values (35.1% / 39%) agree
with the source.

**D3. "No mapped CFIR→ERIC strategy has effectiveness evidence in a UK aphasia context."** Load-bearing
because it is the lane's honest-evidence headline. Attack: recomputed the evidence_standing column from
the CSV (validated_uk = 0) and checked whether any row was miscoded. The count holds; the statement is
an accurate summary of the lane's own coding, not an independent verdict about the world. It should be
read as "none of the strategies this map selected carries effectiveness evidence in the lanes' sources",
not "no UK aphasia implementation strategy has ever been evaluated". The lane presents it as the former.
Verdict: CONFIRMED as an internal count; the scope limit is recorded in residuals.md. A separate defect
in the same lane (C6/C7) is the CFIR vocabulary claim, corrected.

## E. Provenance failures specifically checked

- **Figure attributed to a source that does not contain it.** None found. Every figure re-fetched landed
  at the cited source. The one near-miss is B15 (0.4 WTE) where the attribution is right but the unit
  is wrong; corrected.
- **Charity self-reported reach presented as an independent estimate.** Not found. L2 labels the Stroke
  Association reach figures as charity self-report [23] and the SSNAP figures as national audit [1].
  No mixing of the two inside a single figure.
- **Date older than stated.** Not found as a defect. The CHSS report is presented as "summer 2023" and
  the source confirms the survey window; the SSNAP Post-acute audit is dated 1 April 2021 in the lane
  and the source is the 2021 audit. B20's SSNAP point values are the only ones whose currency the lane
  could not pin to a source file, and the lane already marks them [UNVERIFIED].
- **Number superseded by a newer publication.** The RCSLT 2025 vacancy survey [2] correctly supersedes
  the earlier RCSLT wave; L2 keeps the older "21%" item only as a trend note. The 2024/25 SSNAP [1]
  supersedes earlier annual reports; L1/L2 use the latest.
- **UK claim resting only on non-UK evidence.** Not found as an unlabelled defect. TeleCHAT [14], CHAT
  and ASK [7] are Australia-led and the lanes label them non-UK with the UK-equivalent status stated.

## F. TIDieR extraction check

The lane extracted 12 TIDieR rows. Spot-checked for inferred content against the fetched full texts:

- **Inferred cell found (1):** BCPPA plus "who provided" gave "or a trained assistant under supervision".
  The protocol [15] names a local collaborator SLT, not an assistant. Corrected (V29).
- **"Not reported" that is actually reported (1):** TeleCHAT's "when and how much" is fully reported in
  the source ("50 h … over 8 weeks"; 14/14/14/8 h) [14]; the lane does carry it, so this is a false
  alarm raised and cleared, recorded for completeness.
- **Not independently re-checkable (2):** the semantic-therapy and phonological-therapy rows rest on an
  abstract-only comparison paper (L1 ref Schwen Blackett 2025), so their "Not reported" cells cannot be confirmed or
  refuted here. Carried as UNVERIFIABLE in residuals.md.
- **Absence genuinely absent (checked and upheld):** the StepByStep/Big CACTUS "20-30 min/day" dose
  (A15) is not in the retrieved text; the lane's row is not contradicted, only unsupported at source
  this run.

## G. CFIR/ERIC check against the vault

- ERIC: all 40 strategies and all 9 clusters match the published taxonomy [22]. No invented strategy, no
  strategy outside the published set. CONFIRMED ([20] used for the CFIR side).
- CFIR: the vault's own concept-cfir page and its source-damschroder-2022 page carry the CFIR 1.0→2.0
  rename table (Intervention Characteristics→Innovation; Knowledge and Beliefs→Capability; and so on).
  Measured against that and against cfirguide.org's CFIR 2.0 construct list [20], ~32 of the 64 rows use
  a CFIR 1.0 construct name or a non-CFIR (COM-B) hybrid label, so the summary claim "stated in the
  vault's CFIR 2.0 vocabulary" is not supported. See kill_list.json V4. This is a labelling/provenance
  defect, not fabrication: the underlying barriers are real and the ERIC side is clean.

## Sources

[1] https://www.hqip.org.uk/wp-content/uploads/2025/11/SSNAP-State-of-the-Nation-2025-FINALv2.pdf
[2] https://www.rcslt.org/speech-and-language-therapy/workforce/vacancy-survey
[3] https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0200096
[4] https://pmc.ncbi.nlm.nih.gov/articles/PMC6700375
[5] https://pmc.ncbi.nlm.nih.gov/articles/PMC11178191
[6] https://pmc.ncbi.nlm.nih.gov/articles/PMC8941719
[7] https://pmc.ncbi.nlm.nih.gov/articles/PMC10507993
[8] https://www.hqip.org.uk/wp-content/uploads/2026/03/Ref.-296-SSNAP-2021-Post-acute-Organisational-Audit-National-Report-Final.pdf
[9] https://researchbriefings.files.parliament.uk/documents/CDP-2025-0010/CDP-2025-0010.pdf
[10] https://www.chss.org.uk/wp-content/uploads/F011-Coms-NLHL-1in5-Aphasia-Report-v11.pdf
[11] https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0354353
[12] https://pmc.ncbi.nlm.nih.gov/articles/PMC8273536
[13] https://pmc.ncbi.nlm.nih.gov/articles/PMC13425045
[14] https://pmc.ncbi.nlm.nih.gov/articles/PMC12615849
[15] https://pmc.ncbi.nlm.nih.gov/articles/PMC13475093
[16] https://clinicaltrials.gov/study/NCT07386652
[17] https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD000425.pub4/full
[18] https://onlinelibrary.wiley.com/doi/full/10.1111/1460-6984.12918
[19] https://www.nice.org.uk/guidance/ng236/chapter/Recommendations
[20] https://cfirguide.org/constructs
[21] https://pmc.ncbi.nlm.nih.gov/articles/PMC8330561
[22] https://kclpure.kcl.ac.uk/ws/files/194935792/Appendix_2_ERIC_Implementation_strategies.docx
[23] https://www.stroke.org.uk/about-us/annual-report
[24] https://register-of-charities.charitycommission.gov.uk/en/charity-search/-/charity-details/5089928
[25] https://www.sayaphasia.org/about-us