VA Claim Evidence: What 42,902 Board Decisions Show About What Works
Which evidence shows up in granted appeals, and which does not. Claim Raven holds and searches a database of more than 501,000 Board of Veterans' Appeals decisions; the figures on this page come from 42,902 analyzed decisions whose evidence discussion was extracted, predominantly 2024-2025, which yield 80,128 condition records. The single most important finding: a private medical opinion the Board described as weak was granted 2.4% of the time; described as strong, 93.7%. What the opinion says matters. That it exists does not.
- 42,902 decisions with evidence extraction
- 80,128 condition records
- 501,000+ Board decisions in the library
- Updated August 3, 2026
A weak nexus letter did not help. At all.
The most marketed piece of VA claim evidence is the paid medical opinion: the nexus letter or private independent medical examination (IME). The raw numbers below show decisions that discussed one granted at well above average rates, and that number alone is exactly how these documents get sold. It does not survive one control.
Split the same decisions by how the Board rated the medical-link evidence, and the association concentrates entirely in opinions the Board found persuasive. Where the Board called that evidence weak, decisions with a private opinion were granted 2.4% of the time, slightly below weak-tier decisions with no private opinion at all (3.3%).
Grant rate of decisions with a private medical opinion, by the Board's rating of the medical-link evidence
| Board's rating of medical-link evidence | Granted, no private opinion | Granted, private opinion present |
|---|---|---|
| Strong | 86.9% (n=6,889) | 93.7% (n=5,050) |
| Adequate | 67.3% (n=13,066) | 72.5% (n=5,531) |
| Weak | 3.3% (n=14,197) | 2.4% (n=3,530) |
| Missing | 1.4% (n=9,911) | 2.0% (n=807) |
Source: Claim Raven analysis of 80,128 condition records with evidence extraction, from 42,902 Board of Veterans' Appeals decisions (predominantly 2024-2025). Board outcomes on appealed claims, not initial-claim outcomes. Associations, not causes.
The descriptive reading: a private medical opinion was associated with modestly higher grant rates when its reasoning held up (5 to 7 points within the strong and adequate tiers), and with nothing when it did not. Veterans who obtained strong opinions may differ from other veterans in ways this dataset cannot see, so none of this measures what buying a letter causes. What it does measure is that the Board's own rating of the opinion's reasoning, not the opinion's existence, is what tracked with outcomes.
Outcome rates by evidence type
Each row counts the decisions whose written text discussed that evidence type; a decision usually discusses several, so rows overlap. Baseline across all 80,128 decisions with evidence extraction: 35.9% granted, 32.2% denied, 30.9% remanded. Do not read this table without the section above it: the nexus-letter and private-opinion rows are exactly the numbers that get quoted alone, and alone they mislead.
| Evidence discussed | Decisions | Granted | Denied | Remanded |
|---|---|---|---|---|
| Work impact statement | 610 | 62.8% | 19.0% | 18.2% |
| Nexus letter | 4,189 | 61.7% | 13.5% | 24.6% |
| Private medical opinion (IME) | 13,871 | 54.5% | 22.4% | 22.5% |
| Buddy statement | 3,063 | 49.7% | 21.0% | 28.8% |
| Deployment records | 9,269 | 45.4% | 16.0% | 38.3% |
| DBQ | 7,235 | 45.1% | 27.2% | 26.9% |
| Medical literature | 16,535 | 43.9% | 25.8% | 29.5% |
| Treating physician opinion | 32,594 | 43.6% | 27.5% | 28.3% |
| Lay statement | 43,610 | 39.5% | 29.1% | 30.7% |
| Personnel files | 18,070 | 39.4% | 25.2% | 34.7% |
| VA C&P exam | 71,018 | 36.4% | 32.4% | 30.2% |
| Service treatment records | 68,192 | 35.9% | 32.2% | 31.1% |
Rows do not sum to 100% because dismissals and rare outcomes are not shown as columns. Source: Claim Raven analysis of 80,128 condition records with evidence extraction, from 42,902 Board of Veterans' Appeals decisions (predominantly 2024-2025). Board outcomes on appealed claims, not initial-claim outcomes. Associations, not causes.
16 extractor labels are excluded (unnormalized duplicates of canonical categories, or below the 500-decision publication threshold): VA treatment records (473), VA examination (387), Symptom Log (367), Award Documentation (329), VA examination reports (293), Lay statements (276), Va Treatment Records (245), Service treatment records (223), Treatment records (212), VA examination records (160), Private medical records (150), Medical records (147), Lay testimony (139), Veteran testimony (139), Service records (134), Private medical opinion (127).
The ordering has a plain reading: it is roughly "evidence a veteran adds" above "evidence every file already has." Service treatment records and the VA C&P exam sit at the baseline because nearly every decided appeal discusses them; they do not distinguish outcomes. The types above baseline are the ones that appear in some files and not others.
Grant rate rose with every additional evidence type discussed
Decisions whose text discussed one evidence type were granted 16.7% of the time. Decisions discussing six or more were granted 56.9%, and the gradient between them is monotonic.
Grant rate by number of evidence types discussed in the decision
This is the chart on this page most likely to be misread as advice, so the caveat goes here and not in a footnote: well-developed cases differ from thin ones in many ways at once, and the count of evidence types may simply mark cases that also had stronger content. "Submit six things" is not what this measures.
Lay statements mattered most where nothing else was added
In decisions with no private medical opinion, those that discussed a lay or buddy statement were granted about twice as often as those that discussed neither. Where a private opinion was present, the lay evidence association nearly flattened:
| No lay or buddy statement | Lay or buddy statement discussed | |
|---|---|---|
| No private medical opinion | 17.7% (n=49,206) | 34.3% (n=33,264) |
| Private medical opinion discussed | 53.9% (n=6,249) | 56.1% (n=10,517) |
Source: Claim Raven analysis of 80,128 condition records with evidence extraction, from 42,902 Board of Veterans' Appeals decisions (predominantly 2024-2025). Board outcomes on appealed claims, not initial-claim outcomes. Associations, not causes.
This complements the lay-evidence gradient on the main statistics page, which measures how the Board treated lay evidence it quoted. Both point the same direction: lay evidence appeared in a large share of granted decisions, and its association was strongest in files that had little else added to them.
Definitions
- Nexus letter
- A written medical opinion, usually from a private clinician, addressing the link between a condition and military service.
- Private medical opinion (IME)
- An independent medical examination or opinion obtained outside VA, including paid expert reviews.
- DBQ
- Disability Benefits Questionnaire, a standardized VA form a clinician completes about a condition's severity.
- Lay or buddy statement
- Non-expert written statements from the veteran or people who know them describing symptoms, events, or changes over time.
- VA C&P exam
- The compensation and pension examination VA schedules to evaluate a claimed condition.
- Board's rating of medical-link evidence
- Claim Raven's classification of how the Board's decision characterized the nexus evidence before it: strong, adequate, weak, or missing. It reflects the Board's stated reasoning, not Claim Raven's judgment of the underlying medicine.
- Evidence discussed
- The evidence types Claim Raven's extraction pipeline found discussed in the decision's text. Not a list of everything in the claims file.
Methodology and limitations
- Source: 80,128 condition records with a non-empty evidence extraction, drawn from 42,902 Board of Veterans' Appeals decisions in the analyzed subset of Claim Raven's 501,000+ decision library. Decision dates are predominantly 2024-2025.
- Unit of analysis: the condition record, not the decision. The Board often rules on several claimed conditions in one decision, and the extraction stores one row per condition per decision. Every rate on this page is a share of condition records, not a share of veterans or decisions.
- These are Board outcomes, not initial claim outcomes. Every case in the dataset was appealed, which is a selected population. Nothing here is a "VA claim approval rate."
- Evidence counts as present when the Board's written decision discussed it. Evidence in the claims file that the decision never mentioned is not captured. This measures what surfaced in adjudication, not what a veteran submitted.
- Evidence type labels come from Claim Raven's extraction taxonomy. Synonym labels are merged before publication and the merges are disclosed; unnormalized or sub-threshold labels are excluded and listed rather than silently dropped.
- All relationships are associations, not causes. Decisions with a given evidence type differ from other decisions in ways beyond that evidence.
- The baseline on this page (35.9% granted among evidence-extracted decisions) differs from the all-decisions outcome distribution on the main statistics page because evidence extraction coverage is not uniform. Comparisons belong within one page's base, not across pages.
Cite this research
This page is free to cite and link. The statistics and charts are published under a Creative Commons Attribution 4.0 (CC BY 4.0) license: reuse them anywhere with attribution to Claim Raven. Charts can be downloaded with their source line included. Suggested citation:
Claim Raven (2026). VA Claim Evidence Statistics: Evidence Effectiveness in Board of Veterans' Appeals Decisions. Analysis of 80,128 condition records from 42,902 Board decisions. https://claimraven.com/va-claim-evidence-statistics
Questions about the data or requests for other cuts: contact Claim Raven. For journalists, the methodology above travels with every number, and specific breakdowns can be re-run on request.
Update history
- August 7, 2026: corrected the unit of analysis. This page previously called 80,128 the number of Board decisions, including in its headline and its suggested citation. That figure counts condition records: the extraction stores one row per claimed condition per decision, and those records come from 42,902 decisions. The old figure was larger than the entire analyzed decision count, which is how the error was caught. No rate or chart changes, because every one of them was computed over condition records already; only the labels were wrong.
- August 3, 2026: first publication. 80,128 condition records with evidence extraction.
Research your own claim
The dataset behind this page powers free self-service tools: condition research, evidence checklists, and Resume Raven, a free military-to-civilian resume builder. Companion release: VA Disability Claims Statistics. Claim Raven is not a VSO or law firm and does not file claims for anyone.