Trip Database Blog

Trip has spent nearly 30 years helping people find evidence to inform healthcare decisions. But having useful evidence in an index is only part of the challenge: users also need to be able to find it.

Traditional lexical search looks for matches between the words in a search and the words in documents. It works well in many situations, but it can struggle when a search is broad, when authors use different terminology, or when the same phrase has more than one meaning.

We have been exploring whether AI could help with these problems and also find important material beyond Trip’s index.

Starting with the user’s intention

Someone searching for “prostate cancer screening” might be interested in benefits and harms, screening tests, guidelines, or screening people at higher risk.

Similarly, a search for “asthma in children” could be about diagnosis, treatment, attacks, or triggers and risk factors.

One idea is to offer suggestions after the initial search, allowing the user to select the aspect they are interested in. Trip could then use that hint to run additional searches and reorganise the results.

What might happen behind the scenes?

Our experimental process has six steps:

1. Run a broad search of Trip.
2. Run one or more focused searches based on the user’s selected interest.
3. Search for country-specific material, such as national guidance.
4. Ask an AI system, such as ChatGPT with web search, to find additional relevant documents outside the results retrieved by Trip.
5. Check those external discoveries against Trip’s full index, then combine and deduplicate the results.
6. Use AI to rerank the combined list, taking account of relevance, evidence type, recency and local context.

The user would see one integrated list. Documents found externally that have no confident match in Trip’s index could be labelled “Beyond Trip” (as they are in AskTrip)

Example 1: prostate cancer screening

For this test, the initial search was “prostate cancer screening”, with a focus on people at increased risk. We imagined the user was a Canadian health professional, so Canadian guidance and policy were particularly relevant.

The broad search produced a strong first result: a Cochrane review of PSA screening. The narrower search also found relevant material, including a study of risk-adapted screening in a high-risk population.

But the narrower search showed the difficulty of interpreting “high risk”. Its third result was about treatment for already diagnosed high-risk prostate cancer, while its fourth concerned pancreatic cancer screening. Both contained matching words, but neither addressed the intended question.

The combined, reranked list brought forward Canadian guidance and research on screening people at increased risk. One particularly striking example was the IMPACT study of screening in BRCA1/BRCA2 carriers. It appeared at 54 in the broad search and 92 in the narrow search; in our illustrative reranking, it rose to fourth.

Knowing the user was in Canada also shaped the external search. It helped surface the Canadian Urological Association’s screening recommendations, a British Columbia government guideline and a Canadian review of screening in people at increased risk. These were presented alongside relevant international research.

Our external search also found a systematic review on family history, ethnicity and genetic risk. It wasn’t in either top-results export, but a check showed it was already in Trip – further down. That demonstrates why an external discovery should be checked against the whole index before being labelled “Beyond Trip”.

Example 2: asthma in children

For the second test, the search was “asthma in children”, with the user selecting “triggers and risk factors”. This time, we imagined a UK health professional.

The broad results included the BTS/NICE/SIGN asthma guideline, but the section ranked first concerned diagnosis in children under five. The focused search placed a NICE Clinical Knowledge Summary on risk factors and triggers second: a much closer fit to the chosen interest.

The focused search also found two recent systematic reviews of childhood asthma risk factors, but they appeared at positions 76 and 77. After combining and reranking the results, they appeared at sixth and fifth respectively.

The UK context helped the external search find NICE guidance on indoor air quality, NHS resources for professionals and government guidance on damp and mould. These sit alongside international research, including a systematic review of environmental exposures and asthma exacerbations in children. They have not yet been checked against Trip’s full index, so we cannot say which are genuinely beyond Trip.

In both examples, the country acted as a signal to look for locally useful material, not as a filter excluding evidence from elsewhere.

What happens next?

These are early, manually assisted examples, not evidence that an automated system will consistently perform better. The ranking involves judgement, and important questions remain about speed, reliability and how best to verify and deduplicate external discoveries.

The next step is to test the approach across a wider range of searches. We want to know not only whether the first page looks better, but also whether the process finds useful evidence that would otherwise have been buried or missed.

The aim is a search that combines Trip’s curated collection with AI-assisted discovery, while helping users find evidence that better matches what they actually wanted.

Trip has evolved over the years thanks to the significant feedback from our users. I’d love to hear your ideas for what we should improve next, small tweaks or big ambitions.

Are there search features you wish we had? Content you’d like us to include? Something that takes five clicks when it ought to take one? What would make finding useful evidence easier?

And then there’s AI. There’s plenty of it being bolted onto products, sometimes seemingly for the sake of it. For Trip, the question is: would it help people find the evidence they need?

Have you seen any particularly good uses of AI in search or information retrieval that you think could make Trip better? Perhaps something that helps you refine a question, explore results or find evidence you might otherwise miss.

You don’t need a fully formed idea. A frustration, a wish or an example from elsewhere would all be useful.

What should we work on next?

We’re very pleased with this graph:

We’re clearly doing something right! We don’t advertise AskTrip or Trip itself, so this growth is organic – a positive sign that users are returning, discovering us and perhaps recommending us to others.

AskTrip was designed primarily to help people answer clinical questions. But a substantial minority of the questions submitted to it are not really about choosing a diagnosis or treatment for an individual patient. They are about how healthcare works: staffing, communication, education, inclusion, ethics, carers, policy and the organisation of services.

We analysed the first 20,000 questions submitted to AskTrip. Using a keyword-and-context classifier, followed by iterative checking to reduce obvious false positives, we identified 2,433 core questions – 12.2% of the total – as primarily non-clinical or systemic. A further 476 questions combined clinical and organisational elements; these were kept separate and are not included in the headline figures below.

The largest theme was not policy or ethics, but how services work

The dominant theme was service organisation and quality, appearing in 988 of the 2,433 core questions. These questions concerned such things as continuity of care, staffing models, care pathways, waiting times, referral systems, missed care and quality improvement.

One question makes the connection particularly clear: “How does nurse staffing level impact the incidence of missed care in hospital settings?” The focus is staffing, but the concern is the care patients receive or miss.

This matters because “service organisation” can sound remote from the bedside. In the questions themselves, it rarely was. Staffing levels were linked to safety and missed care; communication failures to patients’ experiences and outcomes; staff morale, breaks and supervision to the reliability of care. The organisational question was often another way of asking a clinical question: what conditions allow good care to happen?

Education, training and development was the second-largest theme, with 471 questions, followed by communication and patient experience with 402. Equity, inclusion and social determinants appeared in 327 questions, while 299 concerned workforce and working conditions. These were not marginal curiosities. Together they show a substantial demand for evidence about the people, relationships and structures around clinical decisions.

The education questions also went beyond acquiring clinical knowledge. One asked: “What are the perceived barriers and facilitators to effective supervision and mentoring for newly qualified occupational therapists during their transition to practice?” This asks about the support needed to put knowledge into practice, making the working environment part of the evidence question.

Evidence needs extend beyond the patient-clinician pair

Several themes suggest that the imagined user of evidence should be broader than an individual clinician choosing a treatment.

There were 238 questions about carers and family support, including advocacy, guilt, young carers, support at school and transitions into care homes. One asked: “How can schools effectively recognize young carers and support their challenges?” Here, the person needing support is a child with caring responsibilities, and the setting for action is a school. The question brings education and family life into the scope of a healthcare evidence tool.

Questions about equity and inclusion dealt with interpreters, access to services, racial abuse, neurodivergent staff and students, cultural safety and protected characteristics. Some involved concrete service choices: “How does the use of professional translation services compare to ad-hoc translation by bilingual staff in terms of accuracy and patient satisfaction?” That single question spans communication, equity and service organisation. It also illustrates why the themes overlap: assigning it to just one category would lose something important.

Ethics appeared not only in abstract debates, but in everyday decisions about consent, capacity, restraint, confidentiality, prescribing for relatives and end-of-life care.

Climate change also emerged as a recurring strand within the 215 public-health, environmental and community questions: heatwaves, climate as a “threat multiplier”, and its effects on older or homeless populations. This is a useful reminder that the boundary of clinical evidence is porous. Questions arriving through a medical evidence tool may begin with health, but quickly move into housing, employment, climate, law, education and social care.

There were also signs of uses that are not direct clinical care at all. Some questions resembled coursework or research preparation, for example, requests to select and critique qualitative studies. AskTrip is therefore being used not only to support decisions, but also to learn how to find, interpret and apply evidence.

Workforce questions often connect wellbeing to care quality

One question asked: “How do racially minoritised staff in medium secure forensic hospitals experience racism, and what are the perceived impacts on wellbeing, professional identity, and clinical practice?” Its wording connects staff experience with professional practice, while asking for evidence about lived experience. A useful response would need to take that qualitative knowledge need seriously.

Within the workforce theme, burnout was recurrent, particularly among nurses. In the broader core set, explicit nurse-burnout questions outnumbered physician or doctor burnout questions by 13 to 5. Three identical questions about mindfulness-based stress reduction for nurse burnout appeared at different times.

That repetition is interesting, but should not be over-interpreted. The data contain questions, not reliable user identities. Repetition could represent the same person returning, different people asking the same question, teaching activity, or a shared concern prompted by an external event. What can be said more safely is that workforce wellbeing was repeatedly framed as an evidence problem and often as one with consequences for retention, safety and quality, not simply personal resilience.

The proportion appears to have grown – but not smoothly

The monthly share of core non-clinical questions rose from roughly 6–7% in the earliest months to around 13–14% in January to March 2026. At first sight, that might suggest AskTrip’s use was broadening beyond bedside decision support.

There are good reasons to be cautious. June 2025 and June 2026 are partial months. More importantly, May 2026 contains concentrated runs of related questions: 109 about carers and residential transitions, 54 about visual-impairment training for care staff and 37 about newly qualified occupational therapists. Those clusters account for 200 of May’s 530 core questions and possibly suggests a training event related to AskTrip.

The time series is therefore consistent with broader use, but it does not prove that the audience itself broadened. It may also reflect repeated refinement, teaching sessions, organisational projects or batches of related enquiries. A future analysis with privacy-preserving session information could distinguish genuine audience growth from concentrated patterns of use.

These questions were not associated with obviously poorer quality labels

The supplied quality metadata did not show a penalty for questions beyond individual clinical care. Among the 2,433 core questions, 49.7% were labelled High or Good; across all 20,000 questions, the figure was also 49.7% after rounding. Core questions had a slightly larger Good share and a slightly smaller Limited share, but the overall distribution was remarkably similar.

That should not be taken to mean that the answers were equally accurate or useful. The dataset did not include the answers, references, scoring rubric or user feedback, so the meaning of the quality label cannot be independently tested. The safer conclusion is simply that the available metadata provides no obvious signal that AskTrip struggled more with this group.

What this means for AskTrip

The clearest finding is not that AskTrip is being used “off label”. It is that healthcare evidence needs do not divide neatly into clinical and non-clinical boxes. A question about burnout may also be about patient safety. A question about interpreters may be about equity, communication and diagnostic accuracy. A question about carers may concern service design as much as family support.

For AskTrip, this suggests an opportunity to recognise different answer modes. A response to an organisational question may need to make context, implementation barriers, resource requirements, stakeholder perspectives and limits to transferability more explicit than a conventional clinical answer. It may also need to draw on a wider evidence base, including qualitative research, service evaluations, policy documents, economics and implementation science.

The first 20,000 questions therefore reveal something larger than an unexpected category of use. They show that people do not approach an evidence tool with clinical trials neatly separated from the realities of delivering care. They ask about the whole system – and about the human conditions under which evidence can actually make a difference.

Methodological note

The analysis used a rule-based keyword and context classifier, refined through manual review of samples and common false positives. For example, “de-escalation” can refer either to managing aggression or to reducing antibiotic treatment, while “diversity” can refer to a workforce or to a patient population. The 2,433 core questions were judged primarily organisational, workforce, ethical, educational, public-health or otherwise systemic. A further 476 mixed questions were retained separately. Themes are non-exclusive, and all counts describe questions rather than individual users. The five illustrative questions quoted in the text reproduce the wording recorded in the dataset; they were selected to clarify the themes, not as a representative sample.

Three recent questions submitted to AskTrip made me think about the relationship between the evidence we make available and the knowledge people actually need. They all came in this morning – one at 6.46am, the next at 6.52am and the final one at 7.01am – my assumption is these were all from the same person!

All three concerned sexual and reproductive rights: how they differ across countries, the distinction between sexual rights and reproductive rights, and how sexual rights are defined in the literature.

AskTrip produced answers, but the material it found was limited or only partly relevant. Reading them, I wondered whether a general-purpose AI model, drawing on broader sources, might have been better suited to these questions.

Trip is primarily focused on evidence to inform healthcare. These questions sit at the intersection of health, law, ethics and public policy. That makes them useful examples of what happens when a knowledge need extends beyond the strongest coverage of an evidence resource.

But looking more closely, there was something else going on.

The answers sometimes assessed the available material using criteria that did not fit the question.

The first question asked how sexual and reproductive rights compare across countries. The answer opened:

“Evidence on sexual and reproductive rights shows significant disparities in service quality and access across countries…”

It went on to discuss sexual health services, STI screening and sexual harassment. These are relevant to the realisation of rights, but they provide only part of the comparison requested. A reader could finish the answer knowing that access to services varies, while still knowing little about how countries’ legal protections differ.

The answer acknowledged this:

“many studies discuss service provision or healthcare outcomes rather than explicitly comparing legal frameworks”

That is an important qualification. It also identifies the central problem: the retrieved material largely answers a neighbouring question.

A more useful approach would first establish which countries and which rights the user wanted to compare, then examine legal protections alongside practical access. A country comparison needs both, with dates and sources attached.

The second question asked about the difference between sexual rights and reproductive rights. Its opening distinction was helpful:

“Sexual rights focus on the right to safe, consensual, and pleasurable experiences related to sexuality, while reproductive rights emphasize the ability to make informed decisions about reproduction…”

However, the limitations section criticised the available material for lacking:

“exhaustive empirical validation of the distinctions”

This seems to apply the wrong test. The user is asking for a conceptual explanation. Definitions need attribution, clarity and attention to areas of disagreement. Their usefulness does not depend on being validated in the same way as a treatment effect.

An illustrative alternative would be:

Sexual rights concern autonomy, equality and freedom from coercion in relation to sexuality and sexual relationships. Reproductive rights concern decisions about reproduction, including whether and when to have children. They overlap, but sexuality extends beyond reproduction. A fuller answer should explain how particular organisations define the terms and where those formulations differ.

That is a starting explanation, rather than a complete review. It addresses the distinction directly and makes clear what further sourcing would add.

Significantly, AskTrip had already searched beyond Trip for this answer and found a WHO source. Broader retrieval helped, but the answer still judged the source through an inappropriate empirical lens.

The third question asked how sexual rights are defined in the relevant literature. The answer assembled themes from literature concerning different populations, then observed:

“The literature heavily leans on international conventions and human rights frameworks”

For this question, those frameworks could be central to a good answer. WHO’s sexual health page, for example, explicitly presents a working definition of sexual rights grounded in existing human rights.

A more direct answer could begin with that definition, explain its scope, and compare it with other relevant formulations. It would distinguish an organisation’s definition from rights enforceable in a particular jurisdiction. Population-specific research could then show how those rights are understood or experienced in practice.

Across these examples, I see three questions for an evidence service:

• Do we hold the sources needed to answer the question?
• Have we retrieved material that directly addresses it?
• Are we judging that material using criteria appropriate to the knowledge being sought?

These are separate problems. Adding sources will not necessarily fix an answer that treats a definition as an inadequately validated empirical claim.

Would a general-purpose model do better? Probably. Its broader knowledge could help it explain concepts and recognise relevant legal or ethical frameworks. But fluent writing is not proof of accuracy. Current country comparisons still require checked sources, and a fair comparison would assess accuracy, relevance and attribution rather than how convincing the answers sound. The alternatives above illustrate a different approach; they are not a formal test of competing systems.

For me, the most useful lesson concerns the phrase “limited evidence”. It can describe a genuine lack of research, a gap in a database, or a search that has failed to find suitable material. Those meanings should not be interchangeable.

“The sources we found do not adequately answer this question” is sometimes the more accurate statement.

A question can concern health while requiring knowledge that biomedical research alone cannot supply. The challenge for AskTrip – and for evidence services more broadly – is to recognise the knowledge need, find appropriate sources, and be clear when the fit is poor.

AskTrip has now answered 25,000 questions. That’s a milestone we’re delighted to reach, and the question that took us there deserves a closer look. It asked whether bed-based exercises could improve pressure areas, wellbeing or discharge outcomes and whether there was research supporting video-guided exercises.

A practical question about everyday care. Could something relatively simple make a meaningful difference to someone in hospital?

For me, that makes it a fitting milestone question.

It brings together several things that matter: preventing harm, helping someone feel better and supporting their recovery towards leaving hospital. It also reflects the curiosity behind evidence-based practice: here is something we could do, but what research supports it?

The answer also illustrates why uncertainty matters. AskTrip reported that it had not identified direct evidence answering the precise question, although it found related evidence concerning broader mobility interventions.

That distinction deserves attention. Evidence about mobility in general does not automatically establish the benefits of a particular bed-based exercise programme. And whether exercises help is a separate question from whether video is an effective way to deliver them. Equally, not finding direct evidence does not mean an intervention has been shown not to work. It means the specific question remains open.

That can still be a useful starting point. It helps us think about what we would need to know next. Does video guidance make exercises easier to follow? Do patients participate more? Which outcomes should a service measure to understand whether its approach is helping?

Reaching 25,000 Q&As is a reason to celebrate the people asking the questions, too. This particular question reminds us that improving care depends on people noticing opportunities, questioning assumptions and looking for evidence.

Thank you to everyone who has used AskTrip, challenged an answer or offered feedback. Your questions and observations help us improve it.

Read our 25,000th Q&A: Are there studies on the impact of bed-based exercises on pressure areas, wellbeing, or discharge pathways, and is there research supporting video bed-based exercises?

A little over a week ago we published a blog about user feedback on the AskTrip Answer Score. The concern was that the colours and overall rating could give a reassuring impression even when the evidence was limited. I felt this needed addressing quickly, so we have prioritised an update.

We have removed the colour-coded overall score and replaced it with a more neutral display showing two separate ratings:

• Evidence base: how strong and directly relevant the evidence found is.
• Answer quality: how well the answer addresses the question and reflects the evidence found, including its limitations.
  

The distinction matters. An answer can explain weak or inconclusive evidence accurately and usefully, and therefore score highly for answer quality. That does not mean the evidence itself is strong.

The new design is deliberately more understated. We want these ratings to help you understand the basis for an answer, without the colours suggesting more certainty than the evidence supports.

Thank you to everyone who gave feedback. Please keep it coming—it helps us make AskTrip better.

When should you use Trip search, and when should you use AskTrip?

Trip offers two ways to explore clinical evidence: search for relevant documents, or ask a question and receive a referenced answer. Which should you choose?

A practical example helps illustrate what each offers:

Is oral or IV iron the best practice for treating iron deficiency in individuals over 65 years old?

With Trip search, a simple starting point is iron deficiency AND elderly.

This opens up the wider topic. The results include research on iron deficiency in geriatric patients, studies of investigation and outcomes, and the British Society of Gastroenterology guideline on managing iron deficiency anaemia in adults.

That breadth is useful when you want to understand what evidence is available, find a particular guideline, or choose which studies to read. You can filter by evidence type, such as systematic reviews or guidelines, and use the year range to focus your search.

It is also a starting point rather than a finished search strategy. The query does not specify oral or intravenous treatment, so not every result will address that comparison. You can refine the terms as you discover which aspects of the topic matter to your question.

AskTrip starts with the question you want answered. Rather than translating it into search terms, you can ask it in full.

In this AskTrip example, the answer presents oral iron as the usual starting option and identifies circumstances in which IV iron may be preferred, including intolerance, inadequate response or a need for rapid correction.

The answer then expands on the evidence, safety, monitoring and research gaps, with references to explore. It also highlights an important limitation: much of the available research comes from specific settings, such as surgery, and may not directly answer questions about older adults living in the community.

This makes AskTrip useful when you have a focused question and want an initial synthesis to orient your reading. The references allow you to examine the sources behind it; the summary still needs to be assessed in the context of the patient and the applicability of the evidence.

You can then take the question further. The “Explore further” feature lets you ask for more detail or challenge the answer. Suggested follow-ups in this example include when to switch from oral to IV iron and what is known about longer-term health outcomes.

You can also move to Trip search to explore a guideline in full, look for additional studies or widen your reading beyond the treatment comparison.

Equally, you might begin with search, identify an issue that needs clarification, and take that question to AskTrip.

Use Trip search when your priority is finding and selecting the documents you want to examine. Use AskTrip when your priority is a referenced synthesis of a specific question, with the option to explore it further.

Try the two links above with the same clinical problem and see how each supports a different part of finding and understanding the evidence.

We recently received some thoughtful, and slightly concerning, feedback about AskTrip’s answer-confidence display.

A clinician asked about radiation therapy for osteoarthritis. AskTrip’s Clinical Bottom Line was cautious:

Limited and indirect evidence exists regarding the use of radiation therapy for osteoarthritis, particularly in the shoulder, suggesting it remains investigational rather than mainstream treatment.

Alongside this, however, the answer-confidence graphic displayed a green box labelled “Good.”

The clinician and a medical student initially interpreted the green as indicating that AskTrip was more supportive of radiation therapy than another evidence tool they were comparing it with. It took them a little while to realise that the green referred to confidence in AskTrip’s cautious answer, not support for the intervention itself.

This raises an interesting design problem. Most of us have learned to interpret green as “go,” “positive” or “recommended.” Even when the accompanying labels say something different, that immediate visual impression may dominate, particularly when someone is scanning an answer quickly.

We are therefore wondering whether our confidence graphic is communicating what we intended. Would a neutral colour work better? Should “Answer Confidence” become “Confidence in this evidence summary”? Or should the evidence base and answer quality be presented separately without an overall green rating?

We would be interested to hear from other AskTrip users: does the current display make sense to you, or could you also interpret green as endorsing the treatment?