Abstract
Background Patient-facing online triage tools potentially offer GPs a way to recognise cancer-related symptoms and features to facilitate appropriate suspected cancer referrals. Little is known about their impact on cancer diagnostic process and outcomes.
Aim To describe patterns of cancer outcome and process indicators associated with the use of online triage tools in primary care.
Design and setting This cross-sectional study examined the association between publicly reported practice-level data on 3053 English general practices. An online triage tool was used that linked to practice-level data and 31 233 594 triage forms completed using an online triage tool in 2022–2023.
Method A cross-sectional analysis (mixed-effects regression) was conducted to evaluate the association between the annual rate of eConsult forms received (weighted to practice registrations) and various cancer diagnosis process and outcome measures.
Results Increased use of online triage forms was associated with higher rates of urgent suspected cancer (USC) referrals and investigations; for example, there was a 7.75% difference in colonoscopy rate (95% confidence interval [CI] = 3.83% to 8.98%) across the 90% mid-range of practices. Despite this, there was no evidence of an association with USC referrals resulting in diagnosis or emergency presentations.
Conclusion Greater use of online triage forms is associated with more investigations and referrals for suspected cancer, but not associated with better diagnostic outcomes. This may indicate that any additional referrals are of lower quality and overall there is an increase in inappropriate use of USC referrals. Research is needed to understand how this impacts investigations and referrals for individual patients.
How this fits in
Variation in cancer process and outcome indicators of general practice activity in England is well documented. This is the first known study, to the author’s knowledge, to report how such indicators are associated with the volume of use of a patient-facing online triage tool at a general practice. Greater use of the tool is associated with more investigations and referrals, but this may reflect more inappropriate use of urgent suspected cancer (USC) referrals and has very little impact on emergency diagnoses. This study highlights the need for more granular analysis to understand how the investigations and referral are targeted when clinicians use outputs from the tool.
Introduction
In 2020, 58.6% of UK patients with cancer were diagnosed following a general practice referral.1 Commonly, patients present in general practice with either cancer-specific or non-specific symptoms potentially indicative of cancer.2 When GPs are faced with non-specific symptoms potentially indicative of cancer, they have to decide whether to refer the patient, conduct investigations, or wait to see how symptoms develop, potentially employing safety netting, but delaying further investigation.3 In recent years, the urgent suspected cancer (USC) referral (formerly known as 2-week wait or 2WW referral), has become the leading route to cancer diagnosis in England4 with 38.9% of cancer cases diagnosed following USC referral in 2020,1 opposed to other routes such as emergency presentation, which sits at 22.5%,1 and routine GP referrals where cancer is not suspected, which sits at 19.7%.1 This demonstrates an increased burden on GPs to identify patients eligible for USC referrals and reduce routine and urgent referrals where cancer is not suspected.
The 10-year health plan for England promised a shift from analogue to digital, including giving patients control over their access to GPs and preparing digitally enabled GPs to increase the use of online triage platforms or tools.5 Online triage platforms or tools allow the patient, their carer, or non-clinical reception staff at the general practice to fill in a form outlining the reason(s) for contacting their general practice.6 Practices manage incoming forms differently, with some subject to initial triage by administrative staff and others screened by GPs as they arrive. Ultimately, responsibility is with the practice to determine the appropriate route for care. In English general practice patients do not decide on consultation modality. The platforms then provide the GP with information, in a standard format, to potentially speed up the GP’s ability to make a clinical decision on patient management. Adoption of online triage is assumed to save workload by triaging patients to the right appointment with the right person at the right time, shifting the GP’s time to focusing on more urgent patients, although evidence is equivocal.7
Online triage tools may potentially help GPs recognise cancer-related symptoms and features and refer the patient to appropriate cancer referral services. Being asynchronous and written, triage tools may empower the patient to better articulate their issues, potentially leading to increased GP confidence in making decisions.8 Additionally, GPs may be more cautious with online triage if they find it difficult to identify key concerns, leading to a follow-up call or appointment to prompt more details from the patients.8 However, there are potential downsides to online triage, such as the patients downplaying their symptoms, or providing few details, which can lead to delayed decisions.8 Online triage platforms may inappropriately highlight escalation leading to referral for a minor issue.8 Therefore, it is important to understand the impact of online triage tool use on cancer diagnostic processes and outcomes.
This study examines the association between the volume of eConsult online triage use at a general practice and outcome markers, such as cancer detection and conversion rates, and process markers, such as cancer examination rates (for example, colonoscopy and endoscopy) of the cancer diagnostic process for patients registered at the same general practice.
Method
There are a number of online tools available to general practices in the UK. One major clinician-curated online triage system is eConsult,9 established by the Hurley group.10 eConsult allowed access to their usage data, which were linked to publicly available datasets on process and outcome measures. This cross-sectional analysis at general practice level examined the association between the number of eConsult forms in a year and various cancer diagnosis process and outcome indicators.
Data
All data were aggregated at general practice level at source. Data on the number of eConsult forms that had been completed between 1 April 2022 and 31 March 2023 was provided by eConsult. Publicly reported data on cancer diagnosis indicators and general practice characteristics were accessed in September 2025 from the Cancer Services Public Health Profiles and the General Practice Profiles within the NHS Digital Fingertips public website for the fiscal year 2022–2023 (now published by the Department of Health and Social Care).11
In England, the Cancer Services Public Health Profiles were developed in response to the increased attention to the role of primary care in cancer diagnostic evaluation, and include indicators relating to diagnostic processes that relate to large numbers of patients most of whom will not have cancer, such as rate of 2WW or colonoscopy tests, as well as to diagnostic outcome indicators based around those diagnosed, such as the proportion of cancer diagnoses in a general practice following 2WW referrals (Table 1).12,13
All indicators available on the Cancer Service Public Health Profiles were used in the analysis. Practice characteristics extracted were the number of full-time equivalent (FTE) clinical staff per 10 000 patients relating to the fiscal year 2022–2023, and socioeconomic deprivation measured by the Index of Multiple Deprivation (IMD) score of the practice area relating to 2019. The analysis is limited to the indictors that are published in the General Practice Profiles within the NHS Digital Fingertips public website.
Table 1. Publicly reported general practice diagnostic process or outcome cancer indicators Rates of eConsult form usage were calculated using each general practice’s total eConsult form usage and data on the registered practice population. To account for the fact that the very young, females, and older patients consult more frequently, these rates were normalised to a weighted practice population (Supplementary Information S1). The findings of a large study examining primary care workload informed the weightings used based on GP and nurse consultation rates for each age and sex strata.14 The practice-level weighted rate of eConsult use was then linked to the cancer diagnostic process and outcome indicators given in the form of a numerator (for example, the number of incidents, referrals, or investigations per year in a practice) divided by a denominator (for example, the number of practice registered patients).
The data used in this study derived from routinely collected data as part of patient care, and as such may be imperfect. The data on the number of eConsult forms was provided directly by the system supplier and thus should be complete. The cancer diagnostic indicators are based on data combined from multiple sources including the National Cancer Registration data that undergoes rigorous quality assurance and is considered to be of high standard,15 as well as the cancer waiting times data that have been shown to have a high degree of completeness, particularly for those patients diagnosed via a USC referral.16
Analysis
Stata/SE (version 18.0) was used for analysis.
To examine the degree to which observed variation between general practices (that is, variation in publicly reported indicators) reflect variation in practice characteristics, variation in rates of eConsult use among the registered patients, or chance, the study used mixed-effects regression models. These models partition the variation between practices into variation attributable to practice characteristics and eConsult usage (fixed effects), variation attributable to chance, and variation otherwise unexplained (random effect). Separate models were used for each indicator, with logistic models for proportion indicators and Poisson models for rate indicators following previous work12,13 (using numerators and denominators, rather than derived percentages or rates).
Three sets of models were employed for each indicator (all of which account for the influence of chance) in order to estimate the between-practice variation: without adjustment; adjusted for general practice characteristics (percentage of male patients, percentage of patients aged ≥65 years, percentage of patients in lowest 40% deprivation, and the number of FTE clinical staff per 10 000 patients); and adjusted for general practice characteristics and the weighted rate of eConsult use. Between-practice variation was quantified as a variance on the log-odds or the log-rate scale for proportion and rate indicators, respectively. Models were compared to estimate the proportion of the between-practice variance that is explained by the practice characteristics, and by the weighted rate of eConsults.
We estimated rate ratios (for rate indicators) or odds ratios (for proportion indicators) comparing the 95th and 5th percentiles of the distributions of practice eConsult usage rates (90% mid-range).12
Results
A total of 31 233 594 eConsult forms were completed between 1 April 2022 and 31 March 2023 by 3053 general practices in England. Practices using eConsult are spread geographically across English regions (26.87% in North of England, 17.36% in the Midlands and East of England, 26.83% in London, and 28.93% in South of England). When comparing practices that offer eConsult with other practices, the proportion of practices using eConsult are 43.23%, 27.85%, 71.67%, and 67.03% in London, the South of England, the Midlands and East of England, and the North of England, respectively.
Descriptive statistics
Data on cancer diagnostic indicators and general practice populations was available for 6254 general practices in England (Table 2). The 3053 practices that offered eConsult to their patients (48.8% of all practices) had very similar patient (weighted list size, age, sex, deprivation) and staff (clinical staff FTE per 10 000 patients) characteristics when compared with those without eConsult. Practices that offered eConsult forms received on average 10 230 (standard deviation [SD] 17 349) completed eConsult forms within the 1-year observation period (median 4867, interquartile range [IQR] 1420 to 12 052). The average weighted rate of eConsult forms was 0.89 (SD 1.03) per registered patient year (median 0.57, IQR 0.20 to 1.17).
Table 2. General practice characteristics and cancer diagnostic activity indicators Practices using eConsult have slightly less variation (SD) than practices not using eConsult in the following indicators: colonoscopy, sigmoidoscopy, upper gastrointestinal (UGI), and emergency presentations but slightly more variability in USC referral, conversion, and detection rates.
Between practice variation
Table 3 compares the variation in cancer diagnostic indicators between practices with eConsult using three different models. These quantify the variation attributable to differences between practice after accounting for chance and that explained by the factors adjusted for in the model. Reduction in this variance implies that some of this variation was owing to the factors adjusted for in the random effects model. For all indicators considered, there is a slight increase in the percentage of between-practice variance explained by practice factors when the weighted rate of eConsult forms is included indicating that adjustment for eConsult form usage explains little variation. While estimates of between-practice variances are useful to understand how much variation is explained by various factors, they do not convey intuitively the magnitude of variation, and so these have been used to estimate associations in terms of odds ratios in the next section.
Table 3. Comparison of between practice variance estimated using a random intercept on the log-odds or log-rate scales for eConsult general practices: unadjusted versus adjusted for practice factors only versus adjusted for practice factors and weighted rate of eConsult forms received The association between eConsult use and primary care cancer diagnostic activity
The odds or rate ratios for the weighted rate of eConsult forms usage are shown for different primary care cancer diagnostic indicators in Table 4. The ratios show the relative increase (or decrease) in the various activity indicators associated with each additional eConsult form per weighted patient. For example, each additional eConsult form per weighted patient is associated with an increase in colonoscopy rate of 2.6% (95% confidence interval [CI] = 1.3% to 3.9%), and a decrease in referrals resulting in a diagnosis of cancer of –0.9% (95% CI = -2.1% to 0.3%) (Table 4).
Table 4. Odds ratio (OR) or rate ratios (RR) (95% confidence interval [CI]) P-value for general practices offering eConsult to their patients There is evidence of associations between practice’s eConsult form usage and the following indicators: colonoscopy (P<0.0001), sigmoidoscopy (P = 0.026), UGI endoscopy (P<0.0001), 2WW referrals (P<0.0001), and new cancer cases treated resulting from 2WW (P = 0.001) (Table 4). Neither emergency presentations nor 2WW referrals resulting in diagnosis of cancer had a statistically significant association with the weighted rate of eConsult forms (P = 0.582 and P = 0.145, respectively). Comparison of odds ratios and rate ratios (unadjusted and adjusted models) for all general practices in England vs general practices without eConsult vs general practices with eConsult are provided in Supplementary Information S2. Odds ratios and rate ratios for control variables are provided in Supplementary Information S3.
When comparing between general practices with the 5th and 95th percentiles of weighted number of eConsult forms completed in a year, the percentage changes in indicator variables is as follows: colonoscopy 7.75% higher (95%CI: 3.83% to 8.98%) , sigmoidoscopy 5.03% higher (95%CI: 0.58% to 8.98%), UGI endoscopy 8.37% higher (95%CI: 4.73% to 12.09%), 2WW referrals 5.63% higher (95%CI: 2.35% to 5.93%), and new cancer cases treated resulting from 2WW –6.27% lower (95%CI: -9.85% to -2.88%).
Discussion
Summary
Practices where higher numbers of eConsult forms were used (for a given population) had higher cancer diagnostic activity (referrals and investigations), when compared with the practices where fewer forms are used. It is important to distinguish between correlation and causation, and to avoid interpreting the increase or referrals as a result of the use of eConsult. The paper intends to explore how much of the variation in the referrals and diagnosis can be explained by variation in use of eConsult, but not overstate a causal relationship. However, such a relationship was not seen for diagnostic outcome indicators with detection rates slightly lower when more eConsult forms were used.
These findings indicate that more use of eConsult forms may lead to more investigations and referrals, but may in turn increase the inappropriate use of USC referrals and have very little impact on emergency diagnoses.
A potential explanation for this is that the investigations and referrals are being targeted in a less efficient way, in other words, it is the wrong patients who are receiving the additional referrals.
Despite these associations being present, eConsult use may not be a factor in driving the overall variation in diagnostic activity between practices, suggesting that other wider factors relating to practice operations could be driving most of the differences between practices.
Strengths and limitations
Appropriate modelling techniques were used to account for the variation by chance and practice-level variables, within the limits of the available data. Adjustment for confounding at patient level was not possible as data were not available or routinely collected. Without patient-level data it is not possible to know how many, if any, consultations leading to referral, investigation, or cancer diagnosis, involved an eConsult. However, the triage implementation by a practice impacts on overall access to the general practice for patients and so indirectly impacts those who do not use eConsult. In these circumstances ecological studies are appropriate.17
It is also not possible to examine variation driven by the eConsult triaging practitioner as this data were only available at practice level. The practice-level data available are not particularly nuanced, and the potential impact of policies that encouraged adoption of primary care total triage on individual practitioners is likely masked at practice level. Studies examining other measures of quality in primary care, such as variation in prescribing18 and variation in patient experience,19 have shown substantially higher variation between GPs than between general practices.
Comparison with existing literature
Previous studies quantified the size and source of variation in practice-level estimates of cancer process and outcome indicators of general practice activity.12,20–22 Those studies have examined the variation in cancer process and outcome indicators adjusting for patient characteristics in addition to a range of factors, including GP characteristics,20 access to and continuity of primary care,21 and capacity of diagnostic service in secondary care.22 This study adds to existing literature by considering the use of eConsult, which, to the author’s knowledge, has not been previously considered.
The studies found that practice population demographics (that is, age and sex) explained some of the variations in indicators, practice-level characteristics explained relatively little,20 and continuity of care showed strong associations with referrals and investigations.21 Additionally, at health-system level (that is, integrated care systems) access to secondary care diagnostic services accounts for variation in primary care referral outcomes.22,23 A study using patient-level data, indicated improvements in cancer mortality and stage at diagnosis with increased use of USC referrals.24
Access to more granular data would permit exploration of the impact of variation in use of diagnostics related to the use of eConsult, and examine how factors such as continuity of care might have been impacted by the triage following the patients reaching out to their general practice using eConsult.
Little research has examined how GPs use the reports from online triage tools to decide the best action to meet patient health needs.8 A study found that some primary care clinicians find the absence of physical examination when dealing with suspicious symptoms had very little impact, as they referred to electronic health records, and had access to patients’ history. However, others found the experience overwhelming and raised concerns that continuity of care was not possible when receiving patient symptoms online (for example, when patients shared images of skin lesions, clinicians might have found it difficult to assess risk from blurred or poor quality images).25 A systematic review of the role of GP factors in referral decisions showed that doctors ‘clinical suspicion’ and ‘gut feeling’ impacted on their decision when presented with non-specific cancer symptoms, or when patients used vocabulary different from medical terminology familiar to doctors.26 When patients use online triage, some will research their symptoms, others might use their own words, with different implications for ‘clinical suspicion’ and ‘gut feeling’.
Implications for research and practice
Greater use of eConsult is associated with more investigations and referrals, but this may reduce the quality of USC referrals and has very little impact on emergency diagnoses. It is therefore important to understand how the investigations and referrals are targeted. It is also important to understand how each practice is using eConsult; for example, some will be offering this as an addition and others as the main route of access. Access to patient-level and practitioner-level data on eConsult usage and related diagnostic activity will allow examination of any impact of individual patient characteristics, and variation between triaging practitioners.
Additionally, wider factors about the practice could be considered to explore these variations, such as conversion of eConsult forms to face to face versus telephone, and other process diagnostic activities such as magnetic resonance imaging (MRI), chest X-ray, and others.
While the increased referral owing to use of eConsult might seem to be indicating increased use of resources, the impact on patients is unclear and we did not investigate that in this study. It may be argued receiving a negative result from suspected cancer is better than not being referred for diagnosis when cancer is present. However, there is both potential psychological and physical harm that can arise from an unnecessary referral and this should be investigated further.
Notes
Funding
This study is part of a PhD that is funded through the University of Warwick, Warwick Medical School in collaboration with an industrial partner: eConsult (https://eConsult.net/). Helen Atherton, Gary Abel and Joanne Parsons and Louise Hiller are supervisors of the PhD study but are not in receipt of any funding for this.
Ethical approval
The research used entirely anonymous aggregated data at general practice level, requiring no ethics approval. The study obtained ethics approval (reference: BSREC 139/22-23) from the Biomedical and Scientific Research Ethics Committee, University of Warwick, UK.
Provenance
Freely submitted; externally peer reviewed.
Acknowledgements
The authors would like to thank Benedict Hayhoe, Chris Whittle, and Claire Robinson from eConsult for support with the data sharing agreement.
Competing interests
The lead author, Armina Paule, receives a PhD studentship via a Warwick Industrial Fellowship, in conjunction with eConsult Ltd. eConsult funds 50% of the studentship. They are not involved in the design or conduct of the research (beyond specifying the broad research area), and analysis is conducted independently of eConsult. All other authors have no competing interests to declare.
- Received November 11, 2025.
- Revision received January 19, 2026.
- Accepted April 14, 2026.
