Imaging requests

Patients present to primary care with more undifferentiated symptoms and earlier in their history. In this study, only 33 per cent of IAM MRI scans (95 per cent CI ± 6.5 per cent, n = 66) confirmed to established guideline indications. When acoustic neuroma was suspected, imaging of the brain rather than IAM was often selected. Patient requests for imaging were rarely noted; ordering patterns more likely reflect clinician heuristics than patient demand. Low recognition of tinnitus character and poor guideline compliance are in keeping with the literature as a primary care learning need.^{Reference El-Shunnar, Hoare, Smith, Gander, Kang and Fackrell21} Variation in how radiology providers vet requests was also apparent.

Investigation of dizziness was the commonest spurious use of IAM MRI. Retrocochlear pathology is not a primary suspicion in dizziness. Whilst neuroimaging can rule out infarction, the head impulse, nystagmus and test of skew (‘HiNTS’) clinical examination has superior sensitivity.^{Reference Muncie, Sirmans and James23} More importantly, central neurological pathology associated with dizziness would demand same-day referral to acute services rather than out-patient imaging with a turnaround time of weeks.

Magnetic resonance imaging findings

Only one acoustic neuroma was diagnosed, in a patient who had already been referred to ENT prior to MRI; this represents a diagnostic yield of 0.5 per cent (95 per cent CI ± 1 per cent, n = 1). Incidental findings were demonstrated in 44.5 per cent of IAM MRI scans (95 per cent CI ± 6.9 per cent, n = 89), in keeping with the existing literature.^{Reference Hoekstra, Prijs and Van Zanten11,Reference Papanikolaou, Khan and Keogh12} Incidental findings triggered cascades in 18.0 per cent of cases (95 per cent CI ± 5.3 per cent, n = 36), including further specialist referrals (with no ultimate change in management) for 15.5 per cent of all imaged patients (95 per cent CI ± 5 per cent, n = 31). Incidental findings may generate more cascades when identified in primary care, a setting with less expertise in advanced imaging along with access to numerous referral pathways.

Cerebral small vessel disease was the most prevalent incidental finding; this is important in the context of an ageing population, where up to 95 per cent of patients may demonstrate such changes.^{Reference Das, Regenhardt, Vernooij, Blacker, Charidimou and Viswanathan24} Some of these patients were referred to secondary care, and some were immediately started on hypertensive or statin therapy. Radiological cerebral small vessel disease may be a risk for future stroke or cognitive decline; however, many will never develop any symptoms. The evidence is mixed, and it is unclear whether treatment of asymptomatic cerebral small vessel disease is beneficial, particularly in the elderly.^{Reference Das, Regenhardt, Vernooij, Blacker, Charidimou and Viswanathan24,Reference Mok and Kim25} Whilst the negative psychological effects of disease labels could not be fully captured, in four cases, records documented that asymptomatic patients were distressed at being informed of cerebral small vessel disease.

Structural changes, noise exposure and psychological factors all work synergistically to trigger tinnitus, which has a bi-directional relationship with neuro-emotional circuitry.^{Reference Jang and Yi26–Reference Carpenter-Thompson, Schmidt, McAuley and Husain29} Tinnitus is associated with mental health symptoms,^{Reference Krog, Engdahl and Tambs30} noted in this cohort, and around 75 per cent of episodes might relate to emotional stress rather than cochlear pathology.^{Reference Han, Lee, Kim, Lim and Shin31} Over-medicalising, or anxiety-provoking incidental diagnoses and unnecessary referrals, is especially harmful considering that distress may already play a causative role in tinnitus amongst emotionally vulnerable patients.

Referral to specialist services

Studies show that approximately 23–37 per cent of primary care patients with tinnitus are referred to specialist services.^{Reference El-Shunnar, Hoare, Smith, Gander, Kang and Fackrell21,Reference Hall, Láinez, Newman, Sanchez, Egler and Tennigkeit32} Whilst one economic model, based on a variety of sources, suggested that 56 per cent of these patients are ultimately referred.^{Reference Stockdale, McFerran, Brazier, Pritchard, Kay and Dowrick2} Our sample of only imaged cases may not reflect all tinnitus patients presenting to primary care, so they cannot be directly compared. Nonetheless, 65.1 per cent of the 106 tinnitus patients in this cohort (95 per cent CI ± 9.1 per cent, n = 69) were referred to specialist services, with mostly unremarkable imaging findings. These results do not support a commonly held assumption that the availability of direct-access imaging reduces general practitioner specialist referrals. Of 164 patients in whom the general practitioner felt there was nothing of clinical relevance on imaging, 63.2 per cent (95 per cent CI ± 7.3 per cent, n = 107) were still referred to specialists, demonstrating that a ‘normal’ test does not necessarily avoid specialist referral and the further support needed beyond a test for those with persisting symptoms.

Of all imaged patients, 64.5 per cent (95 per cent CI ± 6.7 per cent, n = 129) were still referred to secondary care. It is unclear whether direct-access advanced imaging is helpful when patients are also referred to a secondary care specialist capable of organising such tests when necessary. In the vast majority of patients, the community-organised MRI was not available at their consultant review, and regardless of imaging results, only some patients can be discharged after a single appointment. Furthermore, additional MRI capacity for general practitioner requested imaging represents an increased cost pressure beyond the largely fixed supply of specialist imaging. The pre-hospital scans for these patients, with near-zero therapeutic yield, would seem to represent little utility for primary care management or for patient outcomes.

Only 10.5 per cent of imaged patients (95 per cent CI ± 4.2 per cent, n = 21) were appropriately scanned according to guidelines and received no additional referrals. Based on reference prices,^{Reference Stockdale, McFerran, Brazier, Pritchard, Kay and Dowrick2} an MRI expenditure of £26 000 (200 scans × £130) for this cohort effectively ruled out possible acoustic neuroma in 21 cases without specialist input, whilst generating 31 low-value specialist referral cascades as a result of incidental findings. Unfettered MRI access generates greater out-patient costs through incidental cascades than it saves through appropriately ruling out acoustic neuroma without referral.

Of all imaged patients, 35.5 per cent (95 per cent CI ± 6.7 per cent, n = 71) required no specialist referral. We could assume that the absence of general practitioner requested imaging may have resulted in all these cases being referred onwards. In this sector, primary care MRI represents additional scanning supply procured beyond the largely bottle-necked secondary care capacity. Remembering that 31 cascade referrals were generated through incidental findings, the additional cost pressure of £26 000 primary care MRI scans could have conceivably reduced by approximately only £4880 ((71 − 31) × £122, based on national tariff prices³³) in secondary care referrals. This scenario, however, requires considerable unindicated and clinically ineffective use of advanced diagnostic imaging. Furthermore, not all patients who were referred would necessarily have received IAM MRI in specialist care, and many patients imaged in the community were re-scanned by secondary care. It is therefore highly unlikely that general practitioner imaging access represents any cost-containment benefit for commissioners.

Because hospital out-patient capacity is fixed, even if some referrals are reduced, savings are not released, as out-patient capacity is consumed by other demand. Supplying general practitioners with direct-access imaging represents an additional cost pressure. The clinical threshold for a non-invasive and easily accessible test is likely to be lower than that required for a general practitioner to trigger a specialist referral. It is therefore likely that general practitioners will order more scans than they would have otherwise referred to specialists. It is unclear how many more specialist referrals would have occurred in the absence of direct-access MRI and whether the added cost of such (mostly inappropriate) imaging warrants a potential small referral reduction, particularly with so many low-value cascades.

Referral efficiency from pre-hospital diagnostics was rarely realised, as community-delivered MRI scans were often unavailable at a patient's specialist consultation because of fragmented services lacking digital integration. Therefore, further specialist follow up beyond the initial appointment was often still required.

Benefit versus harm

The diagnostic yield from this primary care strategy was 0.5 per cent. Given that as few as 31 per cent of acoustic neuromas receive intervention,^{Reference Mackeith, Kerr and Milford6} 645 (1 ÷ (0.005 × 0.31)) individuals are imaged in primary care to identify 1 acoustic neuroma candidate for intervention. This diagnostic strategy therefore costs approximately £83 850 (645 × £130) to achieve therapeutic yield.

There is a wide range of interventional procedures available for acoustic neuromas, each with varying associated risks.^{Reference Starnoni, Giammattei, Cossu, Link, Roche and Chacko34} Studies show that between 42 and 58 per cent of patients may see benefit in tinnitus or hearing preservation following surgery.^{Reference Coughlin, Willman and Gubbels8,Reference Kameda, Shono, Hashiguchi, Yoshida and Sasaki10,Reference Kohno, Shinogami, Yoneyama, Nagata, Sora and Sato35} With the most optimistic improvement rate of 58 per cent, we can therefore postulate that only 0.09 per cent (0.5 per cent × 31 per cent × 58 per cent) of patients undergoing IAM MRI in primary care stand to benefit clinically. This means that 1111 (1 ÷ 0.0009) primary care patients need to be imaged for 1 patient to achieve a clinical improvement from treatment. The cost of this diagnostic strategy equates to £144 430 (1111 × £130) per benefitting patient. Even without factoring in the cost of surgical or radiosurgical interventions, the diagnostic spend alone would exceed UK incremental cost-effectiveness thresholds of £20 000 to £30 000 per quality-adjusted life year, considering that the gain in quality-adjusted life years from surgery or radiosurgery above conservative treatment is thought to be often minimal.^{Reference Gait, Frew, Martin, Jowett and Irving36,Reference Zygourakis, Oh, Sun, Barani, Kahn and Parsa37}

The one case of acoustic neuroma in this cohort had already been referred to ENT prior to his MRI, so there is no evidence that general practitioner access to imaging results in an improved diagnostic yield, or sensitivity, compared with usual general practitioner care without imaging.

Incidental findings were seen in 44.5 per cent of imaged patients (95 per cent CI ± 6.9 per cent, n = 89), and 18.0 per cent of imaged patients (95 per cent CI ± 5.3 per cent, n = 36) suffered as a result of overdiagnosis. This included negative labelling with conditions such as small vessel disease, and its associated treatment burden of unclear benefit, and the unnecessary cascades of referrals or further investigations, with no change in management. This means only approximately five patients need to be imaged for one to experience negative consequences of overdiagnosis. Such iatrogenic harms of psychological distress and anxiety, financial or insurance implications, as well as healthcare burden on patients may not be immediately tangible to clinicians. The psychological distress caused by being informed of incidental intra-cranial structural abnormalities is particularly important, considering the bi-directional relationship between tinnitus and mental health.

Strengths and limitations

We were able to audit a significant consecutive sample of IAM MRI scans over a two-year period across a large primary care population and a wide array of practices and clinicians, with mixed demographics, likely reflective of UK general practice. Only nine cases were excluded because of incomplete records, minimising any sampling or attrition bias. Adequate sample size provided an acceptable CI for sampling error. The rates of positive and incidental findings, as well as claustrophobia, were consistent with previous literature. There were no other published data on IAM MRI in primary care for comparison.

Primary care records provide robust data on associated healthcare utilisation across different secondary care services and sectors; however, they may not reflect all real-world symptoms. We could not capture patient-orientated outcomes or quality-of-life data, focusing instead on healthcare utilisation. We could not quantify the potential reassurance from scans with unremarkable findings, or the potentially enduring psychological harm associated with incidental findings and their cascades. We were unable to capture the full range of potential negative consequences of tests across physical, psychological, social, financial, treatment burden and dissatisfaction domains.^{Reference Korenstein, Chimonas, Barrow, Keyhani, Troy and Lipitz-Snyderman38}

Only one evaluator, the regional clinical lead, reviewed the indications for scanning, which could introduce measurement bias. ‘Appropriateness’ of imaging can be subjective, and certain patient neurological features may not have been documented in clinical records. Despite such potential limitations, the low proportion of requests meeting guidelines provides a convincing signal regarding a problem of unindicated imaging overuse.

We did not have a comparator group without MRI access, to contrast management decisions and cascades. We considered comparing rates of ENT secondary uses service (SUS) data against IAM MRI referral rates. However, whilst around a quarter of practices requested no IAM MRI scans, the volume of IAM MRI scans was too small to expect a detectable difference in ENT referrals, as audiological symptoms represent only a small proportion of total ENT referrals.^{Reference Jacups and McConnon39,Reference Mahalingam, Seymour, Pepper, Tostevin and Oakeshott40}

Specialist referrals triggered by incidental findings have a clear temporal causal relationship to patient MRI. Patients undergoing direct-access imaging may reflect those with confounding factors, representing a more complex group than most patients attending primary care with tinnitus. Whilst we therefore cannot comment on whether imaging itself is related to a higher rate of specialist ENT referral, this evaluation certainly shows no evidence to support the assumption that primary care direct access to imaging reduces specialist referrals.

Implications for practice

Imaging guidelines do not comment on the setting of care, and this evaluation is the first to highlight IAM imaging challenges in the low-prevalence and more undifferentiated primary care landscape. General practitioners can refer onward and do not require the high-sensitivity diagnostic strategy of specialist settings. The IAM MRI diagnostic yield for acoustic neuroma was only 0.5 per cent, at a cost of approximately £26 000 per diagnosis. Considering that only 31 per cent of neuromas are resected, and only some of these achieve clinical improvement, this would place the cost per therapeutic yield of this general practitioner diagnostic strategy well above UK cost-effectiveness thresholds. Factoring in surgical complications, it is thought that there is no significant quality-adjusted life year gain from surgery or radiotherapy compared with conservative management of small-to-medium (1–20 mm) acoustic neuromas, and so a diagnostic strategy for such conditions presents no cost utility.^{Reference Gait, Frew, Martin, Jowett and Irving36}

It is debatable whether unilateral, non-pulsatile tinnitus alone warrants imaging. A systematic review found that less than 1 per cent of acoustic neuromas present with unilateral tinnitus alone, and that tinnitus correlates more with hearing loss than acoustic neuroma per se.^{Reference Sweeney, Carlson, Shepard, McCracken, Vivas and Neff4} Studies argue against imaging unilateral tinnitus, suggesting that acoustic neuroma can be reliably and more cost-effectively identified by asymmetric sensorineural hearing loss audiometric protocols.^{Reference Hoekstra, Prijs and Van Zanten11,Reference Chatrath, Frosh, Gore, Nouraei and Harcourt41,Reference Bhargava, Coyle, Wong, Masood and Qayyum42} In view of the harms of over-testing, the 2020 NICE guideline committee recommended to ‘consider’ but not ‘offer’ imaging for unilateral tinnitus. It is unclear what impact this semantic advice will have, considering the current ubiquitous unindicated imaging overuse.

Education is resource-intensive, with limited reach and mixed long-term effects.^{Reference Cadogan, Browne, Bradley, Cahill, Knottnerus and Pop43,Reference Fryer and Smellie44} System changes, such as amending electronic and paper request forms,^{Reference Goldzweig, Orshansky, Paige, Miake-Lye, Beroes and Ewing45} may more effectively limit IAM MRI to proven asymmetric sensorineural hearing loss (asymmetry of more than 15 dB at any frequency has shown 100 per cent sensitivity and 29 per cent specificity for acoustic neuroma^{Reference Obholzer, Rea and Harcourt46}). Based on this evaluation, these changes could avoid over 90 per cent of current general practitioner requested IAM MRI scans as well as resource savings from avoidable cascades as a result of incidental findings. The commissioning of UK audiometry providers would require review, as many currently exclude patients with unilateral tinnitus or asymmetric hearing loss. Readily available bedside tuning fork tests have limited sensitivity, potentially as low as 67 per cent,^{Reference Bayoumy and De Ru47} to detect asymmetric sensorineural hearing loss and are not consistently used.^{Reference El-Shunnar, Hoare, Smith, Gander, Kang and Fackrell21} However, tinnitus alone has a very low prevalence for acoustic neuroma (less than 1 per cent), and therefore negative tuning fork tests may provide an acceptable negative predictive value in this scenario for watchful waiting rather than immediate imaging, particularly for those with a low level of impairment. Because of the poor specificity of tuning fork tests for asymmetric sensorineural hearing loss, as low as 60 per cent,^{Reference Kelly, Li and Adams48} their positive predictive value to indicate imaging is poor, and audiometry first would still be advisable to reliably identify asymmetric sensorineural hearing loss prior to imaging.

Aside from ruling out rare acoustic neuroma, the most therapeutic offer, including psychological therapies, to support habituation to often chronic symptoms may be best accessed through non-surgical audiovestibular services.^{Reference Stockdale, McFerran, Brazier, Pritchard, Kay and Dowrick2} Such audiologist-led care is likely to be more clinically and cost-effective than general practitioner requested imaging and traditional surgical ENT referral,^{Reference Stockdale, McFerran, Brazier, Pritchard, Kay and Dowrick2} which has low diagnostic and therapeutic yield, with greater iatrogenic disease labelling and unintended cascades. Whilst variation exists amongst audiology and tinnitus services in the UK,^{Reference Hoare, Gander, Collins, Smith and Hall49,Reference Gander, Hoare, Collins, Smith and Hall50} it may be reasonable to reduce commissioned general practitioner IAM MRI capacity and more effectively re-invest in audiologist-led diagnostic pathways, with strict scanning protocols and expertise in interpreting advanced imaging, for the high burden of audiovestibular complaints in primary care.^{Reference Abbas, Smith and Trinidade51} Further research should focus on capturing wider domains of quality of life following IAM MRI, including psychological and financial consequences, to understand the net benefit versus harm of imaging patients with a very low likelihood of serious disease.