The Telesafe archive: creating a database of UK primary care telephone consultations

This study demonstrates the feasibility of creating a reusable archive of UK primary care telephone consultations linked to electronic health records and patient questionnaires, though it highlights low consent rates—particularly in deprived areas—when using retrospective remote recruitment methods.

The Gist

Imagine a giant library where, instead of books, the shelves are filled with the recorded voices of everyday conversations between doctors and patients. For years, researchers have studied face-to-face visits, but with the rise of phone calls in healthcare, a huge chunk of these conversations has been happening "off the record" or without a clear understanding of what actually takes place.

This paper is about building a new, special section in that library called the "Telesafe Archive." Here is how the researchers built it and what they found, explained simply:

The Mission: Catching the "Ghost" Conversations

In 2024, about one out of every three doctor appointments in England happened over the phone. It's like a massive, invisible city of conversations. The researchers wanted to know: What actually happens in these calls? To find out, they needed to record them, link them to the patient's medical file, and get the patient's permission to study them later.

Think of this like trying to film a movie set where the actors (doctors and patients) are already performing. The researchers had to sneak in, ask for permission to keep the footage, and then organize it all neatly.

The Construction: How They Built the Archive

The team went to seven different doctor's offices in South West England. They set up a system to catch phone calls as they happened.

• The Invitation: After a patient finished a call, the doctors sent them a letter or a text message asking, "Can we keep a copy of your call for research?"
• The Collection: They gathered the audio, the patient's medical records, and a short survey from the patient about how they felt the call went.

The Results: A Small but Valuable Collection

Despite their best efforts, building this archive was harder than expected.

• The "Low Turnout" Problem: Out of 2,053 people they asked, only 123 said "Yes." That's a 6% success rate. It's like inviting 100 people to a party and only 6 showing up.
• The Deprivation Gap: The researchers noticed that people in poorer neighborhoods were much less likely to say yes compared to those in wealthier areas. It's as if the "party" was happening in a place where some people felt less comfortable attending.
• The Final Archive: They ended up with 91 usable phone calls from 88 unique patients. While small, this is a "gold mine" because it links the sound of the conversation with the written medical notes and the patient's feelings.

What Was Inside the Archive?

When they listened to the 91 calls, they found some interesting patterns:

• The Sound of Typing: In nearly 7 out of 10 calls, you could hear the doctor furiously typing on a keyboard while talking. It's like trying to have a conversation while someone is frantically hammering a typewriter next to you.
• The Length: The average call lasted about 7 minutes.
• The Topics: Patients usually talked about one or two main issues. The most common topics were general health checks, stomach issues, and muscle or joint pain.
• The Mood: Most patients were happy with the call (82% said they were "happy" with the phone appointment). However, if you asked them what they usually prefer, most still wanted to see the doctor in person (59%). It's like saying, "I'm glad I got my package delivered, but I'd still prefer to go to the store myself."

The Big Takeaway

The researchers proved that it is possible to build a library of recorded phone calls with all the extra details attached. They showed that the "Telesafe Archive" exists and is ready for other scientists to use.

However, the paper also warns that the method they used (sending letters and texts after the fact) didn't work very well to get people to join, especially in poorer areas. The paper concludes that while the archive is a success, future attempts to build bigger libraries will need to try different ways to invite people, perhaps by talking to them directly or finding new ways to reach communities that felt left out this time.

In short: They successfully built a small, high-quality museum of phone calls, but they learned that getting the public to donate exhibits is much harder than they thought, and they need a better strategy to make sure everyone gets a chance to participate.

Technical Summary

Technical Summary: The Telesafe Archive

Problem Statement
Despite telephone consultations accounting for approximately one-third of all General Practitioner (GP) appointments in England as of 2024, the specific interactional dynamics and content of these calls remain under-described. Existing large-scale analyses rely on coded data, which may suffer from incomplete or biased clinician coding, while qualitative interviews are limited by small sample sizes and recall bias. Although the "One in a Million" (OiaM) archive successfully established a dataset of face-to-face video consultations, a comparable, reusable archive for telephone consultations did not exist. This gap hinders the ability to study communication quality, safety-netting advice, and documentation practices in remote primary care.

Methodology
The study employed a retrospective observational design to create the "One in a Million Telesafe" archive.

• Setting and Recruitment: Seven primary care practices in South West England (Bristol, North Somerset, and South Gloucestershire) using the EMIS electronic health record (EHR) system were recruited. Eligibility required routine audio recording of telephone calls and at least two clinicians willing to participate.
• Data Collection: The study utilized a two-wave patient recruitment strategy to obtain consent for data retrieval:
  1. Wave 1 (Postal): Patients were invited via post 1–3 weeks after their consultation.
  2. Wave 2 (Electronic): Due to low initial uptake, a text-message-based electronic consent process was introduced for faster notification.
• Data Linkage: For consenting patients, the study aggregated four distinct data types:
  1. Audio Recordings: Retrieved from practice systems, screened for usability (excluding answerphones, cut-offs, or excessive third-party involvement).
  2. Electronic Health Records (EHR): Extracted for the index consultation and a 4-month window (1 month prior to 3 months post).
  3. Patient Questionnaires: Post-consultation surveys covering demographics, satisfaction, and safety-netting recall.
  4. Clinician Surveys: Demographic and professional background data.
• Analysis: Consultation problems were coded using the Complex Consultations Coding Toolkit and the International Classification of Primary Care, version 3 (ICPC-3). Statistical analysis included mixed-effects logistic regression to model consent rates based on practice deprivation (IMD) and recruitment mode.

Key Contributions
The primary contribution is the establishment of the Telesafe archive, a controlled, reusable dataset comprising:

• 91 usable, consented audio recordings of GP telephone consultations from 19 clinicians across 7 practices.
• Linked multi-modal data: Each recording is linked to corresponding EHR entries (86/91) and post-consultation patient questionnaires (89/91).
• Granular coding: 161 distinct clinical problems were identified and coded, providing a detailed breakdown of consultation content.
• Methodological insights: The paper documents the feasibility and specific challenges of retrospective remote consent for audio data, highlighting significant disparities in participation rates based on socioeconomic status.

Results

• Recruitment and Consent: Out of 2,053 invitations sent, 123 patients (6.0%) returned consent forms. Consent rates were significantly lower in more deprived practices (IMD deciles 1–2) compared to the least deprived (IMD 9–10), with an odds ratio of 0.22 (95% CI 0.09–0.54).
• Dataset Composition: The final archive contains 91 consultations involving 88 unique patients. The mean patient age was 63.5 years, with 63% female and 92% White ethnicity.
• Consultation Characteristics:
  • Duration: Mean duration was 7 minutes 13 seconds.
  • Content: The mean number of problems discussed was 1.77 per consultation. The most common problem categories were General (19.3%), Digestive (10.6%), Genital (10.6%), and Musculoskeletal (10.6%).
  • Interaction: Audible typing was present in 69% of consultations, typically occurring while the GP documented patient information.
• Patient Experience: While 82% of respondents were happy with the telephone appointment on the specific occasion, 59% reported a general preference for face-to-face appointments. Satisfaction with the specific appointment was high (67% very satisfied, 28% satisfied).

Significance and Claims
The paper claims that it is feasible to assemble a reusable archive of GP telephone consultations with linked EHR and patient-reported data, subject to ethical approval for secondary analysis. The authors position the Telesafe archive as a platform to examine communication quality, documentation accuracy, and safety-netting advice in remote primary care.

However, the paper maintains a modest tone regarding the generalizability of the dataset due to recruitment limitations. It explicitly notes that the low consent rate (6%) achieved through retrospective remote contact, particularly the disparity in deprived populations, limits the representativeness of the estimates derived from this specific dataset. The authors suggest that future work must explore alternative recruitment approaches, such as prospective enrolment or researcher-led contact, to improve engagement and diversity. The study does not claim to have solved the issue of remote care safety but rather provides the necessary infrastructure (the archive) to enable future research into these critical questions.