By now, you’ve probably heard it more than once: “AI can do all this now.” Plenty of medical records retrieval services are saying it about artificial intelligence (AI) and medical record retrieval. It’s not entirely accurate, though. AI can do a lot with medical records and health records once they’ve been retrieved. Getting them still takes a person.
What Does Medical Record Retrieval Involve?
Unless you’ve handled the medical record retrieval process yourself, most of the process of gathering medical documentation happens out of sight. It includes:
- Checking that the authorization is filled out correctly and confirming what it allows
- Finding the right facility and department to request records from
- Sending requests, usually to several providers per case
- Following up when a provider doesn’t respond or sends only part of what was requested
- Handling rejected requests
- Paying provider fees
- Checking that the records are complete
- Getting certifications
If we could count on everything to go smoothly, parts of this process could be automated. Retrieval rarely goes smoothly, and when it doesn’t, a person must step in to deal with real-world problems that complicate the process.
Some offices only accept fax. Certain practices have closed or merged. Older records have moved to storage vendors. Some providers keep billing, imaging, and chart records in separate places. Artificial intelligence is not yet capable of managing these issues when they arise. However, human retrievers already know exactly how to handle them.
Where Does AI Help With Medical Records?
While AI is not yet ready for use during the retrieval process, it does have a role in handling medical records and health records.
Artificial intelligence is most useful after the records arrive. It can:
- Create medical record summarizations and build medical chronologies
- Spot possible gaps, like a treatment date with no matching record
- Sort large record sets into categories like office notes, labs, prescriptions and imaging
- Find and remove duplicate pages
- Flag possible pre-existing conditions
- Draft case summaries
- Point out contradictions, like different injury descriptions or treatment dates
All of these tasks rely on data quality, however. Smudged, cut off, messy, illegible, or otherwise ambiguous medical documentation can cause the AI to make mistakes that can cost your client compensation. This is where humans handling the data first is still beneficial.
How do humans resolve ambiguous documentation in medical records?
There are several ways that human oversight resolves ambiguous documentation. First, a retriever will review the rest of the file to confirm whether the ambiguity is resolved. For example, the provider may have realized a page was cut off and re-sent it. If the retriever finds that the ambiguity isn’t resolved, they will contact the provider for clarification. This clarification may be provided in the form of sending the record again if it is smeared or cut off. If the information is incorrect or illegible, clarification may be added to the record in the form of an addendum that provides the corrected information while leaving the ambiguous information unaltered and in place.
Where Does AI Fall Short?
AI is useful once the records are in hand, but it struggles with the retrieval itself. Here’s where people still do the work:
- Relationships: Retrievers get to know records clerks at hospitals, doctors’ offices and urgent care centers. Those relationships often get results that an automated request won’t.
- Communication: Medical providers are unlikely to take a call from AI. They can usually tell when they’re receiving an automated email. Those messages are easy to ignore. A real person on the phone or in the inbox is much harder to brush off.
- Judgment: Artificial intelligence can only work with the pages in front of it and has a hard time with cross-record reconciliation. It might notice that an office note mentions an MRI with no report in the file. It won’t know that a hospital keeps ER notes separate from the main chart, or that the imaging center across town holds the films. A retriever who knows where records actually live can reach out to the right person or department to get them.
- Completeness: Even when AI flags a gap, it can’t patch incomplete or missing records the way a human retriever can. Filling those gaps means contacting the provider, requesting the missing pages and following up until they arrive.
- Problem-solving: Rejected authorizations, providers who have moved, and “we never received your request” are everyday obstacles. Getting past them takes phone calls, persistence, and judgment that artificial intelligence doesn’t have yet.
- Accuracy: AI summaries save time, but they can miss details or invent them. In litigation, one missed page can matter, so someone has to check AI output against the actual records.
- Privacy and accountability: HIPAA duties and chain of custody need someone who is responsible for them. Using an AI tool raises questions about where health information goes and who answers for it if something goes wrong.
Side-by-Side Comparison
| Human Retriever | AI | Verdict | |
| Speed | Limited by how fast providers respond, which is the same for everyone | Speedy for sorting, indexing, tracking, and summaries, but can’t make providers respond faster | Both |
| Organizing large record sets | Slow and labor-intensive by hand | Sorts, indexes and makes thousands of pages searchable in minutes | AI |
| Accuracy | Can check records against the request and catch illegible or cut-off pages | Fast, but can miss or invent details | Both: AI speeds it up, people verify |
| Getting records from unresponsive providers | Calls, follows up, escalates and uses relationships with records staff | Can send requests and reminders, but can’t negotiate or problem-solve, and automated messages are easy for providers to ignore | Human |
| Spotting missing records | Knows where records are actually stored and catches gaps nothing in the file points to | Flags gaps the records mention, like a referenced MRI with no report | Both |
| Filling in incomplete or missing records | Contacts the right provider or department and follows up until the missing records arrive | Can flag a gap, but can’t go get what’s missing | Human |
| Privacy protections | Trained on HIPAA and follows chain-of-custody procedures | Depends on the tool’s security settings and where the data goes | Human |
| Accountability | A named person responsible for your file | No one to call when something goes wrong | Human |
| Cost of mistakes | Errors can be traced and fixed by the person responsible | Errors can be hard to spot and can carry into summaries and chronologies | Human |
What Can Human Retrievers Do That AI Can’t?
We know law firms want a medical records solution that offers both experience and technology. That’s why Records On Time pairs experienced retrievers with technology, so you get speed without sacrificing accuracy.
Our outsourced record retrieval for lawyers is designed to keep things simple. You submit a request through your case management system (CMS), and we take it from there. Our retrievers contact providers and follow up until the records come in. Every communication, with your firm and with medical providers, is handled by a real person.
One of our retrievers’ most important jobs is checking each record for accuracy and completeness before it’s delivered to your firm. We review each set for correct dates, gaps and for smeared or cut-off pages. Then, we apply optical character recognition (OCR) so the records are searchable. The finished records go straight into your CMS, ready to review.
We know fast record turnaround for lawyers matters. Our average turnaround is 21 days for traditional retrieval and one day for electronic health records (EHR).
Optical Character Recognition
Applying OCR allows you to search for specific documents, key words or phrases, dates, and other information in the medical records without having to read each page. OCR also bridges the gap between workflows that require human effort and AI workflows. Optical character recognition allows artificial intelligence to generate more comprehensive document summaries, create chronologies, surface patterns, and provide higher-quality legal document drafts. It even allows you to use AI to perform simple yet time-consuming tasks such as identifying relevant names, dates, and other relevant case details.
FasterOutcomes Partnership
Once records are in hand, artificial intelligence can play a meaningful role in organizing, summarizing, and preparing them for review. Rather than building those capabilities ourselves, we’ve partnered with FasterOutcomes. Its platform is designed specifically for that work, so your firm maximizes efficiency by using both.
FasterOutcomes’ AI builds the chronology once the records are received and immediately flags treatment gaps. It also offers AI solutions for the rest of the case process, including viability screening, fault mapping, demand drafting, and doing the math to help you decide to litigate or settle when it’s time to negotiate.
What Questions Should You Ask a Retrieval Vendor?
You probably have a lot of questions about medical record retrieval. Whether you’re considering one record retriever vendor or comparing several, these questions are worth asking. Here’s how we answer them.
Who Follows Up When a Provider Doesn’t Respond?
We follow up with every provider until we either get the records or get a clear answer that there are none.
How Do You Confirm Records Are Complete?
Firms still need to do their own completeness check, since they have the case timelines and billing statements. We review the records for gaps, look for Certificates of Authenticity or affidavits with the records we receive, and request a Certificate of No Records when a provider says there aren’t any.
Where Is Our Clients’ Health Information Stored?
We deliver retrieved records directly into your CMS through our integration with most systems, where they stay unless you delete them. We also keep a copy in our secure HIPAA-compliant system for 90 days in case a problem comes up.
Is a Real Person Accountable for Our Records?
Yes. You have a dedicated customer service representative to help with problems or questions.
Keeping the Human Side of Record Retrieval
Technology can organize medical records and health records, but people are the ones who get them. This isn’t an anti-AI position. We use technology where it adds real value, and we don’t think retrieval is there yet. If that changes, we’ll gladly build it into our process. Until then, our retrievers will keep doing the work that gets records in the door.
If you’re looking for a record retrieval solution that keeps people at the center and you’d like to see how it works, schedule a demo.
