Disability NLP Evidence Detection with Tenasol
Disability NLP is the use of NLP to detect evidence of disability in structured and unstructured medical records. This capability has profound impacts on processing the high volumes of data that analysts must parse to grant disability benefits to individuals, usually through the Social Security Administration (SSA) or Veterans Affairs (VA). Tenasol performs these operations as a regular part of its business.
A few notes to consider:
Disability approvals are becoming increasingly slow and complex. 2 million people apply each year, 30% are approved, but take on average 7.5 months and can take as long as 4 years [1]. Medical records contain an increasing amount of data that needs to be parsed, further slowing the process.
The SSA is attacking the problem from all angles. Ultimately it will likely be solved by a mix of policy changes and technology.
Disability definitions and types are indexed in a resource called The Blue Book which has listings for 14 adult disabilities and 15 childhood disabilities.
Disability Application Process
You can apply for benefits online via the SSA. Generally an applicant must meet the definitions of being disabled and have met disability insured status (typically worked 5 out of the 10 last years). The process after that goes like this:
Person discovers they may qualify for disability [Disability NLP may occur here]
Application is submitted (form SSA-16). In rare instances a patient may be immediately approved if they have critical condition (CAL/QDD).
SSA requests permission to review your medical records via interop (form SSA-827). This may also involve a consultative exam by an M.D. [Disability NLP may occur here]
If a patient has been denied in any previous step they may undergo up to 4 successive legal appeals [Disability NLP may occur here]
Why Disability NLP?
While health plans are often interested in discovering gaps in care or maximizing HEDIS measure ratios, the federal government is highly interested in making sure that maximum number of people who need benefits get them. Believe it or not, not everyone who is entitled to benefits has applied for them and is receiving them. So really there are two goals:
Find Disability Gaps (step 1): using existing evidence on hand, discover people who are likely disabled (disability gap) but not receiving benefits so they may be notified that they may qualify for disability benefits. There is indeed a noticeable percentage of people who are eligible for disability benefits but are not receiving them.
Find Disability Evidence (step 3): Using gathered patient records for a patient applying for disability, extract evidence that they are disabled. Tenasol has also developed technology for the parsing of SSA-827 forms for granting data collection approvals more quickly to the SSA.
Interoperability (step 3): Tenasol has further used NLP for validating that new data connections were reliable between the SSA and partnering facilities. Making sure data is exchanged properly and with high quality is a process permitted by NLP systems.
Assist in Appeal Decisions (step 4): These may be long and drawn out and there are many opportunities where NLP tools are capable of efficiently parsing through large amounts of diversely formatted language to seeking answers more quickly. This may include evidence detection or summarization via Healthcare LLM systems.
What is Different about Disability NLP?
Disability NLP systems are different from health plan uses cases like risk adjustment and HEDIS in:
Different Definitions: the definitions of disability are far more complex than the narrow definitions of diagnosis codes used by health plan reimbursement systems.
Age dependency: There are different criteria for adults than for children in disability.
Higher data volumes: Because the lookback period is larger than health plan NLP use cases or prior authorization, records float around 500 pages or equivalent of medical data - significantly higher than what is seen in other populations in healthcare NLP.
Table of SSA Blue Book adult disability definitions.
Tenasol processing pipeline
In addition to normal extraction, Tenasol performs disability-specific detection of language by flagging instances of high confidence and linking them to their indicated disability type detected based on age group and calibrated confidence score. In total, the system performs:
suspected disability language detection
associated code detection tied to disability
autonomous patient and patient information detection
summarization
Disability NLP Evidence Output
In output, Tenasol can:
PDF Highlighting Create a highlighted document identifying disability NLP evidence
Raw Data: Supply JSON representation of evidence files, pages, and coordinates with additional metadata. This includes autonomous detection of patient information.
API output: Tenasol is capable of delivering HL7 FHIR (USCDI or Base R4) or raw data via API upon request.
Generalized Output in CSV, XLSX, Pipe-delimited or other custom formats if requested.
Conclusion
Disability NLP has the potential to revolutionize how disability benefits are processed, addressing long-standing inefficiencies and improving accessibility for millions of applicants. By leveraging Tenasol’s advanced technology, the complexities of detecting disability evidence within massive volumes of structured and unstructured medical records can be effectively managed. This ensures that qualified individuals, whether applying through the SSA or VA, receive the support they need without unnecessary delays.
Tenasol’s approach—combining comprehensive data extraction, disability-specific detection, and outputs tailored to meet diverse needs—positions it as a leader in solving critical challenges faced by disability determination processes. Whether through PDF highlighting, JSON metadata, or seamless API integrations, Tenasol’s output formats ensure flexibility and accessibility for end-users, including analysts, policy-makers, and healthcare professionals.
As disability approvals become more complex due to growing data volumes and stringent criteria, tools like Tenasol’s Disability NLP pipeline will play an essential role in transforming the application and appeals process.
Contact our sales team for more details.