Introduction

Glaucoma is one of the most common causes of blindness, and as vision loss from glaucoma is irreversible, it is important that it is detected at an early stage [1, 2]. A challenge, however, is that glaucoma is asymptomatic until substantial loss of vision has occurred and therefore early detection relies on opportunistic testing. In Scotland, all residents are entitled to a free eye examination under the general ophthalmic services contract, including screening for glaucoma, at least every 2 years, with community optometrists fulfilling the role of primary eye care providers [3]. National guidance has established the criteria for referral of patients with suspected glaucoma to the hospital eye service (HES) [4] and according to the Scottish Intercollegiate Guidelines Network all patients with suspicious or glaucomatous appearances of the optic disc, abnormal visual fields (confirmed on repeat testing) and/or intraocular pressure (IOP)Ā± central corneal thickness (CCT) measurements above a certain threshold, should be referred. The guidelines also recommend that patients with an optic disc haemorrhage (DH) are referred irrespective of whether or not there are other risk factors or features of glaucoma, with DH defined as a haemorrhage that occurs close to the optic disc within the retinal nerve fibre layer (RNFL) [4].

Although the major risk factor for glaucoma is raised IOP, the vascular theory of glaucoma hypothesises that impaired blood flow to the optic nerve head is likely to contribute to its pathogenesis. Vascular dysregulation and factors including endothelins and vascular endothelial growth factor have been implicated in the development of glaucoma and have also been associated with DHs [5, 6]. Glaucoma has been proposed as the primary cause of DHs [7,8,9] and previous studies have found the prevalence of DHs to be higher in patients with glaucoma (4.4ā€“17.6%) compared with those without (0ā€“1.4%) [10,11,12,13,14,15]. However, DHs may also be present in a large proportion of non-glaucomatous subjects and are associated with a number of ocular and systemic conditions including posterior vitreous detachment [16], hypertension [17], diabetes [18], migraine [7], and the use of medications including antiplatelet [19] and antihypertensive drugs [7, 10], particularly systemic beta-blockers.

To the best of our knowledge there are few studies examining the outcome of referral for DHs by community optometrists. A population-based study in China found a positive predictive value (PPV) of only 20% for the presence of a DH in glaucoma [20]. Given the limited capacity of hospital glaucoma clinics, there is debate regarding whether patients with isolated DHs and with no other features suggestive of glaucoma, require referral or would otherwise be more suitable for monitoring by community optometrists [21]. The aim of this study was to investigate the outcome of patients referred from primary optometrists to the glaucoma clinic due to DH and to determine the PPV for diagnosis with open angle glaucoma (OAG).

Methods

This was a retrospective review of all consecutive new patient referrals from community optometrists to the hospital glaucoma service at Princess Alexandra Eye Pavilion in Edinburgh, Scotland from June to November 2016. The study methods were approved by the Princess Alexandra Eye Pavilion Quality Improvement Committee and all methods adhered to the tenets of the Declaration of Helsinki.

1070 consecutive new referrals were identified from the electronic patient administration system. Optometrists completed a standardized referral form which requested information including gender, date of birth, visual acuity, refraction, IOP, type of tonometer used (applanation or non-contact), results of slit lamp biomicroscopy and optic disc examination, and the reason for referral. Optometrists were also requested to include colour optic disc photographs and the results of automated threshold visual fields with all referrals, and to stipulate the reason for referral. Reasons for referral included the presence of a visual field defect, changes to the optic nerve suspicious of glaucoma or elevated IOP, and decisions regarding whether or not an abnormality was present were left to the discretion of the referring optometrist. We determined that optometrists had deemed an optic disc suspicious if the referral included any comment about possible abnormal optic disc appearance. For example, if the referral included any of the terms ā€œenlarged cup disc ratioā€, ā€œsuspicious discā€, ā€œretinal nerve fibre defectā€, ā€œoptic disc asymmetryā€, ā€œneuroretinal rim thinningā€ or ā€œbayonettingā€.

Of the 1070 referrals, 301 were excluded as they were not referred by a community optometrist or had a previous diagnosis of glaucoma or a diagnosis of secondary glaucoma. An additional 13 were excluded due to missing referral letters, 33 due to a non-attendance at the clinic and 10 due to the lack of information about their glaucoma clinic visit. 95 patients were referred due to suspected angle closure glaucoma and were also excluded. This resulted in 618 patients eligible for inclusion in the analysis, of which 52 (8%) had been noted by their optometrist to have at least one DH at the time of referral.

All 618 patients had a comprehensive history and an eye examination performed at the hospital eye clinic. The examination was performed either by a consultant ophthalmologist (four in total) or a hospital optometrist specialising in glaucoma (two in total), under direct supervision of an ophthalmologist and with at least two yearsā€™ experience of working regularly in the hospital glaucoma service. All patients had a comprehensive past medical history recorded, in particular enquiring about hypertension, diabetes, migraine/Raynaudā€™s phenomenon, cerebrovascular accident (CVA), ischaemic heart disease (IHD) and smoking. A medication history was also taken with special regard to use of antihypertensives, anticoagulants, or antiplatelets. Tests performed in the glaucoma clinic included slit lamp biomicroscopy, Goldmann applanation tonometry, gonioscopy, optic disc and fundus examination and visual field testing (SITA standard 24ā€“2 strategy using the Humphrey Field Analyzer, Carl Zeiss Meditec). Gonioscopy was performed in a dark room prior to dilatation using a one mirror Magnaview (Oculus) gonioscope. All subjects also had optical coherence tomography (OCT) imaging of the circumpapillary RNFL using the Heidelberg Spectralis OCT (Heidelberg Engineering, Heidelberg, Germany).

The primary outcome was whether or not the patient was diagnosed with OAG in at least one eye at the first hospital visit. The diagnosis of OAG was determined by the examining clinician and defined by the presence of glaucomatous optic neuropathy detected on examination of the optic nerve head and RNFL by clinical examination and/or OCT, with the exact diagnostic criteria dependent on the clinicianā€™s judgement. Patients with OAG were required to have an open angle on gonioscopy (i.e., the posterior trabecular meshwork visible for >90Ā°). and have no secondary cause for glaucoma identified, although pigmentary and pseudoexfoliative glaucoma were permitted. Suspect glaucoma was defined as a suspicious optic disc appearance or visual field test result suspicious of glaucoma but where the probability of glaucoma was deemed insufficient to commence treatment and ocular hypertension (OHT) was defined as an IOP >21ā€‰mmā€‰Hg with no evidence of glaucoma and with an open angle on gonioscopy. A secondary outcome was whether or not the patient was discharged from the glaucoma clinic at the first visit (first visit discharge rate, FVDR). All patients who were discharged at the first visit were deemed by the examining clinician to not have glaucoma and to not require treatment, for either glaucoma or OHT.

Statistical analysis

Normality assumption was assessed by the inspection of histograms and using the Shapiroā€“Wilk test. For comparison of continuous non-normal variables, the Wilcoxon rank-sum test was used. Fisherā€™s exact test was performed for the comparison of categorical variables. Univariable logistic regression was used to examine whether the presence of a DH was associated with increased odds of glaucoma diagnosis. Similar analyses were performed to see if other factors increased the odds of glaucoma diagnosis including age, gender, use of antihypertensive, anticoagulant, or antiplatelet medication; a diagnosis of diabetes, hypertension, CVA, IHD, or Raynaudā€™s phenomenon/migraines; spherical equivalent; and smoking status. Most analyses were performed on a patient-basis, but three referral criteria that can affect a single eye (high IOP, abnormal visual fields, and suspicious optic discs) were also analysed by individual eyes. When analysing by patient, the analysis considered the patient to have glaucoma if they had glaucoma in either eye.

Logistic regression analyses were performed using Stata Version 13.0 (Stata Corp, College Station, TX). All other analyses were carried out using SPSS Version 24 (IBM Inc., Chicago, IL). All tests were two-sided and a P valueā€‰<ā€‰0.05 was considered statistically significant.

Results

The analysis included 618 patients referred to the glaucoma clinic by community optometrists, including 54 (8.7%) referred with a DH (DH+) and 564 (91.3%) with no DH (DHāˆ’). DH+ patients were significantly older than DHāˆ’ patients (75.1ā€‰Ā±ā€‰10.4 versus 66.8ā€‰Ā±ā€‰13.7 years respectively, Pā€‰<ā€‰0.001). Patients with DH were also more likely to be female and on average had lower IOP than those without DHs (TableĀ 1). Patients with DH also had thinner CCT, were less likely to have a visual field defect noted by the optometrist at the time of referral and were less likely to have other suspicious optic disc features or a family history of glaucoma (TableĀ 1). Average spherical equivalent among patients with DH was āˆ’0.96 dioptres (medianā€‰=ā€‰0, IQ range āˆ’1.00 toā€‰+1.00, range āˆ’6.75 toā€‰+3.50 dioptres). A substantial proportion of patients without DHs did not have information regarding refraction completed in the referral meaning comparison with the DH group was not possible. One patient was noted to have DHs in both eyes, whereas all other patients had DH noted in one eye only. Although optometrists are recommended to include optic disc photographs with all glaucoma referrals, only 221 of 618 patients (35.8%) had the disc photograph included, though there was no significant difference in the proportion of patients with and without DH in whom photographs were provided (Pā€‰=ā€‰0.163) (TableĀ 1).

Table 1 Summary of characteristics of patients referred to the glaucoma clinic with DH compared with those referred with suspected open angle glaucoma or OHT and no DH
Full size table

21 of 54 patients (38.9%) referred with DHs were referred due to DHs alone, meaning they had normal IOP (defined as IOPā€‰ā‰¤ā€‰21ā€‰mmā€‰Hg) and no other suspicious features of glaucoma detected by the referring optometrist. 33 patients had a DH plus another reason for referral, including 6 with high IOP (defined as IOPā€‰>ā€‰21ā€‰mmā€‰Hg), 24 with abnormal visual fields (defined according to the referring optometrist) and 16 with optic disc appearance suspicious of glaucoma (according to the referring optometrist) (TableĀ 2). 6 of 54 patients (11.1%) had a family history of glaucoma in a first degree relative. For the purpose of analysis, patients with DH were divided into three groups; DH alone; DH plus one other feature suggestive of glaucoma, e.g. high IOP, abnormal visual field (according to the referring optometrist or optic disc appearance suspicious of glaucoma (according to the referring optometrist); and DH plus two features suggestive of glaucoma.

Table 2 Positive predictive value (PPV) for OAG diagnosis and first visit discharge rate (FVDR)
Full size table

7 of 54 (13%) patients with DHs were using antiplatelet agents, 5 (9.3%) were taking anticoagulants, and 22 (41%) antihypertensive medications. 3 of 54 patients (5.6%) were diabetic, 24 (44.4%) had hypertension, 22 (40.7%) had a history of migraines or Raynaudā€™s phenomenon, 10 (18.5%) had a history of CVA or IHD, and 3 (5.6%) were current smokers.

Overall, of the 54 patients referred with a DH, 29 (53.7%) were diagnosed with OAG at the first hospital visit, for a PPV of 53.7% (95% CI 39.6ā€“67.4%) (TableĀ 2). The PPV was expectedly lower at 23.8% (95% CI 8.2ā€“47.2%) for those referred with a DH alone. Therefore, of patients referred with a DH but no other features of glaucoma noted by the referring optometrist, almost one in four was diagnosed with OAG at the first visit to the glaucoma clinic.

The PPV increased with the increasing number of concurrent features suggestive of glaucoma, reaching 92.3% (95% CI 64.0ā€“99.8%) for those with a DH and two other features suggestive of glaucoma. The overall FVDR was 35.2% (95% CI 22.7ā€“49.4%), increasing to 57.1% (34.0ā€“78.2%) in those referred due to DH alone. The FVDR for those referred with DH alone was significantly greater than the FDVR of 25% among the 564 patients referred with suspected glaucoma without a DH (Pā€‰=ā€‰0.001). The FVDR decreased to 0% for patients with a DH and at least two other features suggestive of glaucoma (TableĀ 2).

The presence or absence of a DH was not associated with increased odds of diagnosis with OAG among the 618 patients referred to the glaucoma service during the study period (ORā€‰=ā€‰0.88, 95% CI 0.40ā€“1.96, Pā€‰=ā€‰0.77). Logistic regression analysis showed the gender, use of antiplatelets, anticoagulants, or antihypertensives, diabetes, hypertension, Raynaudā€™s, migraines, stroke or IHD, refractive error and smoking status had no significant effect on the odds of diagnosis of glaucoma in patients with DHs (TableĀ 3). This information was not collected for patients without DH which prevented comparison of DH+ and DHāˆ’ groups.

Table 3 Logistic regression analysis examining odds of diagnosis of glaucoma at first visit in patients with DH
Full size table

Discussion

In Scotland, the current glaucoma referral guidelines recommend that community optometrists refer patients with DHs to secondary care regardless of the presence or absence of other signs of glaucoma [4]. In this study, 54% of patients with DHs were diagnosed with glaucoma at their first visit to the glaucoma clinic, indicating a high glaucoma risk if a DH is noted by a community optometrist. However, the majority of patients with a positive diagnosis had other signs of glaucoma noted at the time of referral. The primary aim of this study was to determine if optometrists should continue referring patients with isolated DHs and the PPV for this group of patients was lower at only 24%, indicating that only one in four patients with an isolated DH were diagnosed with OAG. Those referred with an isolated DH also had a high FVDR, with almost 60% discharged at the first visit, compared with a FVDR of only 25% for the overall cohort.

Previous studies examining community optometry referrals for suspected glaucoma have reported PPVs (for glaucoma diagnosis or high suspicion of glaucoma requiring follow up in the hospital eye clinic) as low as 37% [22, 23]. We found a higher overall PPV for all referrals of almost 60%, perhaps higher than similar studies due to the requirement for optometrists in Scotland to perform supplementary tests prior to referral, such as applanation tonometry, pachymetry and automated threshold perimetry, which likely improves referral accuracy [4].

As expected, the PPV was higher when DHs were associated with other features of glaucoma, with a PPV of 92.3% for patients with DHs and at least two other features suggestive of glaucoma noted by the referring optometrist. Although it is intuitive that patients with more abnormal signs are more likely to be diagnosed with glaucoma, there were low numbers of patients with a combination of these features, which reduces the ability to draw firm conclusions. The results suggest though that when a DH is present in an eye with other suspicious features of glaucoma, the probability of glaucoma is high.

It is also important to appreciate predictive values that should be interpreted with caution as they are strongly influenced by disease prevalence. Although this makes PPVs useful for examining the value of a test in different populations, translating results from a hospital-based study to community screening is problematic. When used in low prevalence settings even a test with a high sensitivity and specificity can have poor predictive value and therefore be of limited utility. The prevalence of glaucoma in Scotland is estimated at 1ā€“2% in those aged over 40 years, and 5% of those aged over 75 years [4]. Were a test with a sensitivity and specificity of 98% used in a setting where the prevalence of disease is only 2%, although the negative predictive value would be high, the testā€™s PPV would be only 49% [23, 24]. If the same test was performed in a setting where the prevalence of disease is 5%, the PPV would be 71%. The PPV can therefore be increased by conducting the test in a subset of a population with higher disease prevalence or by combining the test with supplementary testing.

The PPV for diagnosis with glaucoma will be higher in patients attending for opportunistic glaucoma screening at a community optometrist, where the prevalence of glaucoma is higher, than in the population as a whole. In the current study, the prevalence of glaucoma in patients referred to the hospital eye clinic was almost 60% and the PPV for glaucoma diagnosis in those with a DH was 53.7% (95% CI 39.6ā€“67.4%). In a population with a prevalence of glaucoma of 2%, the PPV would reduce to only 1.6% (95% CI 1.0ā€“2.6%). For patients with an isolated DH, the PPV for glaucoma diagnosis was 23.8% (95% CI of 8.2ā€“47.2%), however, in a population with a prevalence of glaucoma of 2%, the PPV would reduce to 0.4% (95% CI of 0.2ā€“1.2%). The effect of disease prevalence on PPV illustrates the importance that diagnostic tests are used in the appropriate setting.

The effect that an abnormal clinical examination or test finding has on the likelihood of disease is often quantified using likelihood ratios. Likelihood ratios have the advantage of not being influenced by disease prevalence and can be used to modify the pre-test probability of disease to determine post-test probability, reflecting a clinicianā€™s normal decision-making process. Among all patients referred to the glaucoma clinic, the presence of an isolated DH decreased the probability of glaucoma diagnosis, with a positive likelihood ratio of 0.22 (95% CI of 0.08ā€“0.59). Therefore, if a patient was deemed to have a pre-test probability of glaucoma of 25%, the finding of an isolated DH, with normal IOP, normal visual fields and otherwise normal disc appearance reported by the optometrist, would decrease the probability of glaucoma to 5.5%. This should however not be interpreted as the finding of an isolated DH decreasing the probability of glaucoma in patients undergoing opportunistic screening at a community optometrist. As we did not capture information about patients not referred to the glaucoma clinic we were not able to examine this question and did not report likelihood ratios.

Whether a PPV of 24% for OAG diagnosis in patients with isolated DHs is deemed an acceptable cut off for referral is likely to depend on the healthcare setting. The goal of glaucoma treatment is to reduce patientsā€™ lifetime risk of visual loss and therefore referring an individual with an isolated DH and no evidence of other structural or functional changes may not always be necessary in those at low risk. Depending on the facilities available, monitoring in primary care may be more appropriate. Nevertheless, as one in four patients noted to have an isolated DH by their optometrist was diagnosed with OAG on the first visit to the glaucoma clinic, it seems prudent that patients with DHs have further testing. It is likely that in patients referred due to DH alone but diagnosed with glaucoma at the first visit to the glaucoma clinic, the treating clinician detected other signs of glaucoma not identified by the referring optometrist. This may reflect for example, the wider availability of OCT in secondary compared with primary care, or the specialist nature of the clinic.

It is important to acknowledge other limitations of the study, firstly, the risk that the calculated PPV is an overestimate due to not being able to capture information about patients with DH who were not referred. Optometrists are very likely to not have referred some patients with DH that they deemed to be at low risk and this selection bias would have inflated the PPV. The ideal methodology to determine whether optometrists should continue referring patients with isolated DHs would require capture of this missing data, and ideally be primary care or population-based. Furthermore, as DHs are self-limiting and may occur unobserved, it is certain that some patients in the no DH group had unobserved DHs. We also based the diagnosis of OAG on the first hospital visit but as DHs are a risk factor for glaucoma progression, some patients referred due to DHs who do not have glaucoma at present may develop it in the future, and DHs confer higher risk. It would be interesting to determine whether any of the 60% of patients with isolated DHs who were discharged at their first visit are re-referred due to glaucomatous changes in the future, as one would expect them to be at higher risk. DHs are however not infrequent in the general population and are more often than not, not associated with glaucoma. For example, The Blue Mountains Eye Study [10] and Beaver Dam Eye Study [25] population-based studies reported respectively only 27 and 4.3% of subjects with DHs to have glaucoma. When evaluating a patient with DH, it is consequently important to consider other potential causes of DHs, including hypertension [17], diabetes [18], retinal vein occlusion [26], myopia [27] and perhaps most commonly posterior vitreous detachment [16]. There appears to be no association though between DHs and smoking, CVA or IHD [10]. We had only one patient in this study who was thought to have a DH caused by a posterior vitreous detachment.

In conclusion, despite limitations, this study has confirmed DHs to be a common reason for referral to glaucoma clinics, cited as a motivation in 9% of referrals. Whether patients with isolated DHs should be referred to the HES depends on local arrangements and the ability of the optometrist to monitor for possible future glaucomatous changes.

Summary

What was known before

  • Optic disc haemorrhages are an important risk factor for glaucoma and glaucoma progression.

  • Patients with isolated disc haemorrhages are often referred to glaucoma clinics but the majority of these patients do not have glaucoma.

What this study adds

  • Almost 10% of patients referred to a hospital glaucoma clinic had an optic disc haemorrhage noted at time of referral.

  • 24% of patients with isolated disc haemorrhage were found to have glaucoma, however the predictive value of a test depends on disease prevalence.

  • Whether patients with isolated disc haemorrhages should be referred to glaucoma clinics depends on local arrangements and the ability of the optometrist to monitor for possible future glaucomatous changes.