Digital technologies for applying algorithms to spirometry to support asthma and COPD diagnosis in primary care and community diagnostic centres: early-use assessment

The EAG identified 11 studies on ArtiQ.Spiro, 9 studies on LungHealth and 1 study on GoSpiro. No relevant evidence was identified by the EAG (or submitted by the company) for EasyOne Connect. NuvoAir was removed from the assessment because it is no longer available to the NHS. MIR Spiro was removed from the guidance following the second committee meeting because it was not considered to be relevant to the scope of the assessment.

No studies were identified that reported long-term outcomes for any of the included technologies. These outcomes include:

• mortality

• morbidity

• time to diagnosis

• staff time and resource use

• number of secondary care referrals for diagnosis, and

• hospital admissions.

Diagnostic accuracy evidence was available for ArtiQ.Spiro and LungHealth. Comparative diagnostic accuracy evidence (for example, compared with a reference standard) was lacking for LungHealth.

Diagnostic accuracy evidence was most comprehensive for ArtiQ.Spiro. This included SPIRO-AID, a UK randomised controlled trial. It included 133 primary care healthcare professionals (who refer for, perform, or interpret spirometry) who were randomised to review 50 retrospective spirometry records with or without ArtiQ.Spiro. The EAG calculated sensitivity and specificity of ArtiQ.Spiro using SPIRO-AID data, which was provided as academic-in-confidence by the study authors. The committee acknowledged that sensitivity was higher when using ArtiQ.Spiro to support asthma and COPD diagnosis, but that specificity was similar in both trial arms. The committee discussed that the sample of 50 records reviewed in the SPIRO-AID study was small, but also had a low proportion of people without disease (only 20% had a 'normal' spirometry pattern). The committee noted the importance of evidence showing a pre-test probability of disease that is in line with the populations in which ArtiQ.Spiro would be used in primary care and community diagnostic centres. The committee agreed that the UK diagnostic accuracy validation study of ArtiQ.Spiro (Sunjaya et al. 2025) also had a low proportion of people with a 'normal' spirometry pattern (30 out of 1,113 people). Sensitivity of ArtiQ.Spiro was low for identifying 'normal' cases (reported as 33.3%), with 66.7% of the 30 true 'normal' cases misclassified mainly to be interstitial lung disease or asthma. This could lead to a high number of people having treatment for, or a life-long label of, a condition that they do not actually have.
The committee noted that there are risks associated with both underdiagnosis and overdiagnosis of lung conditions. If people have a false-positive diagnosis, they may be offered unnecessary and potentially harmful treatment. For example, bronchodilators may be associated with increased cardiovascular risk. If people have a false-negative diagnosis, they could miss out on treatments to help manage their condition and symptoms. Algorithmic support for diagnostic spirometry could influence prescribing decisions following diagnosis. Therefore, diagnostic accuracy of the technologies is important to reduce incorrect prescribing decisions. So, more evidence is needed on the diagnostic accuracy of ArtiQ.Spiro when used as part of care pathways in primary care and community diagnostic centres.

There was limited evidence available to conclude whether diagnostic accuracy of the technologies was affected by inter-reader variability of different grades of staff doing and interpreting spirometry. One of the value propositions of the clinic-based intervention technologies is an increase in access to spirometry. This is because more types of staff would be able to perform and interpret spirometry. The committee concluded that diagnostic accuracy evidence should capture staff who would use the technologies in primary care and community diagnostic centres, including people with different levels of experience and specialist knowledge. This would show whether diagnostic accuracy differs depending on who is using the technology.

The committee noted that even if the technologies, when used by different grades of staff, have equivalent (non-inferior) diagnostic accuracy to that of current practice, there may still be benefits such as improved access to spirometry. This could reduce the burden on secondary care services if fewer referrals from primary care are needed to make a diagnosis. The committee agreed that more evidence was needed on this.

NICE's guideline on asthma notes that people should be offered treatment immediately if they are acutely unwell or highly symptomatic at presentation. It also notes that objective tests that may help support a diagnosis of asthma should be performed if the equipment is available (for example, eosinophil count, FeNO, spirometry or peak expiratory flow before and after bronchodilator). A specialist committee member noted that spirometry results can be influenced by the presentation of a person's symptoms at the time of testing. If spirometry is done while a person is not experiencing an asthma exacerbation (or after they have taken medication), then their spirometry pattern may appear to be 'normal'. If testing is done when the condition is unstable, this may result in overdiagnosis and unnecessary treatment. The committee discussed that asthma diagnosis may be more accurate if repeat spirometry measurements are taken over a period of time. It agreed that technologies bringing diagnostic spirometry closer to a person's home may increase the number of asthma diagnoses that can be based on a series of spirometry measurements. Evidence of diagnostic accuracy based on a series of diagnostic spirometry tests was not available.

Paediatric spirometry has higher technical failure rates and greater interpretative uncertainty. More caution is therefore necessary when using technologies in populations of children and young adults (where indicated for use in these populations). The committee agreed that more research on diagnostic accuracy of the technologies in populations of children and young people is needed.

The committee discussed that, in practice, technologies that apply algorithms to spirometry could also be used to support a corrective diagnosis of asthma or COPD, if it is suspected that the initial diagnosis was incorrect.

Comparative evidence for the other outcomes defined in the scope was available for ArtiQ.Spiro, and was limited for the other technologies included in this assessment. The studies used real-world primary care datasets and included primary care healthcare professionals who perform and interpret spirometry. The evidence showed examples of changes in practice through using the technology in primary care settings. Findings included a non-significant difference in accuracy of spirometry interpretation (for example, recognising an obstructive or restrictive spirometry pattern) using ArtiQ.Spiro (Doe et al. 2025). An increase in the quality of spirometry (Doe et al. 2025) was also reported, along with reductions in the time to perform and interpret spirometry (Hayes et al. 2025 [PDF only] and Adams et al. 2024 [PDF only]). Regional variation in resource and capacity in primary care settings and community diagnostic centres may mean that there is regional variation in the magnitude of change seen through the use of ArtiQ.Spiro. Further evidence generation could show the benefits that are seen in practice.

Evidence from 1 abstract suggested that using ArtiQ.Spiro led to an increase in testing capacity (75 more tests per month) and improvement in wait times (before and after not reported). It also projected full-test-backlog resolution within 8 months. The committee acknowledged the potential of the technologies to address unmet system need by increasing access to diagnostic spirometry. The committee noted that this evidence comes from a single abstract, that is not peer-reviewed, on 1 of the technologies. It concluded that further evidence of the impact of all technologies on access to diagnostic spirometry in care pathways is needed.

The committee discussed that a key factor in access to spirometry is local commissioning decisions. Without local commissioning, services are unable to provide spirometry due to the lack of funding. There is regional variation in spirometry commissioning in primary care settings, with commissioning more common in community diagnostic centres. So, implementation of the technologies included in this guidance is unlikely to improve access to diagnostic spirometry in areas where spirometry is not commissioned.

There was a lack of longitudinal outcomes across all included technologies. This includes mortality, morbidity, time to diagnosis, staff time and resource use, number of secondary care referrals for diagnosis and hospital admissions. The EAG reported that the conceptual economic model was not sensitive to small differences in these outcomes (see section 3.32), but there was uncertainty in some the model inputs because of the lack of available clinical evidence. The committee concluded that evidence generation on long-term outcomes (along with other outcomes related to diagnostic accuracy and resource use) could help to address uncertainties.

Lay specialist committee members emphasised the importance of early and accurate diagnosis of lung conditions. This could lead to earlier access to effective treatment options, and enhanced quality of life for people with asthma or COPD and their family or carers.

The committee heard that while many people would feel comfortable trusting a diagnosis made by a staff member with less experience (with algorithmic support), some people may prefer to have their diagnosis made by a staff member with more experience.

Technologies that enable spirometry to be done closer to a person's home could benefit them. For people living with an undiagnosed lung condition, their symptoms may mean they find it difficult to attend an in-person appointment in secondary care. For example, if their journey would require them to walk for a period of time. Neurodivergent people may also be more comfortable doing spirometry in a setting that is more familiar to them. There may also be cost barriers for some people when attending an in-person hospital appointment. People living in areas of deprivation may have an increased risk of lung conditions, and may not be able to travel to an in-person hospital appointment because of their financial situation. Technologies that would bring spirometry into primary care and community diagnostic centres could allow easier access to diagnostic testing for these people.

Evidence related to healthcare professional acceptability, ease of use, experience and satisfaction was available from surveys on ArtiQ.Spiro and LungHealth. This evidence generally indicated positive healthcare professional experience with using the technologies, highlighting potential benefits in terms of workflow efficiency. One study (Hayes et al. 2025 [PDF only]) highlighted that users felt that additional training and support was needed for interpreting artificial intelligence reports from ArtiQ.Spiro. The committee noted that support may be needed to ensure seamless integration of the technologies in current workflows and care pathways, to maximise the benefits seen in practice.

Evidence was available for ArtiQ.Spiro on healthcare professional confidence in interpreting spirometry results and making a diagnosis. Studies by Doe et al. (2025) and Adams et al. (2024 [PDF only]) did not report significant differences in healthcare professional confidence when using ArtiQ.Spiro. Willaert et al. (2023) reported that GPs using ArtiQ.Spiro recognised the need for more objective findings before making a diagnosis or altering treatments. Concerns about unfamiliarity with the spirometry procedure and limited time and resources were considered barriers to implementation. The committee recommended evidence generation on how the technologies would be implemented in care pathways in primary care and community diagnostic centres. This includes demonstrating efficiencies and resource impact.

The EAG developed a conceptual economic model in which the general structure could apply to all asthma, COPD and restrictive lung disease populations included in the scope. The structure incorporated a decision tree to model the diagnostic phase, which is embedded within a 'testing state' of a Markov model to model the wider care pathway of diagnosis and management. The model has a 10-year time horizon with monthly cycles (with alternative time horizons considered in sensitivity analysis). Further details of the economic modelling are in section 6 of the external assessment report. Many of the model inputs were derived from assumptions and clinical opinion, given the limited evidence available. The committee agreed that the model structure and input parameters were appropriate for this early-use assessment.

Two base-case scenarios were modelled, based on the 2 value propositions of the included technologies: increased diagnostic accuracy, and faster access to objective testing. Assumptions were made to model these value propositions. For improved diagnostic accuracy, a 10% increase in sensitivity was assumed in the intervention arm. For access to spirometry, the intervention arm assumed 70% of people had testing within 6 months, compared with 63.2% for the comparator arm. The committee noted that these assumptions were suitable given the limited evidence available for diagnostic accuracy (see sections 3.11 to 3.15) and access to testing (see section 3.20). It concluded that more evidence was needed on diagnostic accuracy and efficiencies when the technologies are used in care pathways.

The base-case scenarios assumed technology costs for LungHealth, with available costs for other technologies applied in the sensitivity analysis (see section 3.32).

Results of the economic model were presented for each base-case scenario, for populations of adults with asthma, children with asthma, and people with COPD. The committee discussed that, in the base-case scenarios (which used technology costs for LungHealth), there were small differences in total costs between the intervention and comparator arms. It was also noted that over the 10-year time horizon, incremental quality-adjusted life years (QALYs) appeared to be small. Base-case incremental cost-effectiveness ratios appeared below a threshold of £20,000 per QALY gained in each of the base-case scenarios, for each of the modelled populations. But, there is uncertainty in many of the model inputs because of the limited available evidence (see section 3.28). The committee concluded that cost-effectiveness estimates are therefore only to be treated as illustrative, and cannot be used to determine whether the interventions are cost effective.

The committee noted that the purpose of this conceptual modelling work was to highlight key evidence gaps and key drivers of differences in costs and utilities of the intervention technologies when compared with standard care. Key drivers of cost effectiveness in the model were identified as being diagnostic accuracy of the interventions and technology costs per patient. Other parameters to which the model was sensitive to include initial prevalence of disease, time horizon and costs of further testing. The EAG reported that small differences in long-term outcomes may not significantly impact the overall cost effectiveness of the technologies, but there was a lack of evidence to inform model parameters on long-term outcomes for all technologies.

Technology costs for ArtiQ.Spiro were applied, along with diagnostic accuracy (sensitivity and specificity) calculated from the SPIRO-AID trial. In this scenario, the intervention arm would be considered dominant for each of the populations modelled. Scenarios were also tested in which technology costs for GoSpiro were applied, with all other parameters applied as they were in the base case, because of the lack of evidence on this technology to inform model parameters. The intervention arm was again dominant for each of the populations modelled. As in section 3.31, these results are to be treated as illustrative only. No technology costs were available for EasyOne Connect, so cost-effectiveness estimates were not reported for these technologies.

The EAG tested a scenario in which the technology costs were assumed to include spirometry performance and interpretation by a GP, instead of a band 5 practice nurse as in the base case (see section 6.3.2.3 of the external assessment report). The EAG concluded that the model is sensitive to technology costs, and this is influenced by the grade and time of staff using the technologies. The committee discussed that, in practice, other grades of staff may use the technologies, so more evidence is needed on who in the care pathway may use the technologies.

The EAG noted that restrictive lung disease was not included in its economic analysis, because of a lack of evidence available for this population. The committee discussed that spirometry is not typically used as a first-line diagnostic test for suspected restrictive lung disease. Spirometry may be used to exclude obstructive disease, and as a prognostic tool for staging the severity of disease once restrictive disease has been diagnosed. So, the committee concluded that it was acceptable for the illustrative model results to be in populations with suspected obstructive diseases only (asthma and COPD).