The Effectiveness of Teleglaucoma versus In-Patient Examination for Glaucoma Screening: A Systematic Review and Meta-Analysis

Background Glaucoma is the leading cause of irreversible visual impairment in the world affecting 60.5 million people worldwide in 2010, which is expected to increase to approximately 79.6 million by 2020. Therefore, glaucoma screening is important to detect, diagnose, and treat patients at the earlier stages to prevent disease progression and vision loss. Teleglaucoma uses stereoscopic digital imaging to take ocular images, which are transmitted electronically to an ocular specialist. The purpose is to synthesize literature to evaluate teleglaucoma, its diagnostic accuracy, healthcare system benefits, and cost-effectiveness. Methods A systematic search was conducted to help locate published and unpublished studies. Studies which evaluate teleglaucoma as a screening device for glaucoma were included. A meta-analysis was conducted to provide estimates of diagnostic accuracy, diagnostic odds ratio, and the relative percentage of glaucoma cases detected. The improvements to healthcare service quality and cost data were assessed. Results Of 11237 studies reviewed, 45 were included. Our results indicated that, teleglaucoma is more specific and less sensitive than in-person examination. The pooled estimates of sensitivity was 0.832 [95% CI 0.770, 0.881] and specificity was 0.790 [95% CI 0.668, 0.876]. The relative odds of a positive screen test in glaucoma cases are 18.7 times more likely than a negative screen test in a non-glaucoma cases. Additionally, the mean cost for every case of glaucoma detected was $1098.67 US and of teleglaucoma per patient screened was $922.77 US. Conclusion Teleglaucoma can accurately discriminate between screen test results with greater odds for positive cases. It detects more cases of glaucoma than in-person examination. Both patients and the healthcare systems benefit from early detection, reduction in wait and travel times, increased specialist referral rates, and cost savings. Teleglaucoma is an effective screening tool for glaucoma specifically for remote and under-services communities.

2 to 4 years for adults between 40 to 64 years and every 1 to 2 years when aged 65 and older [7]. Patients regularly seen by ophthalmologist for other ocular conditions may also be referred for glaucoma diagnostic testing if signs appear. In-patient care for glaucoma (passive ''in-person screening'') is performed at specialized clinics and includes detailed history, slit lamp examination, visual field testing, and fundus photography performed by the optical technician followed by consultation with the ophthalmologist [8].
Teleglaucoma is a relatively new screening and diagnostic tool for targeting remote or under-serviced communities. It uses stereoscopic digital imaging to take ocular images which are transmitted electronically to an ocular specialist. The ocular specialist will then assess the images, identify risk factors and diagnose for glaucoma. If necessary the ocular specialist will refer identified glaucoma cases for medical consultations or to ophthalmologists for follow-up treatment. Unlike other teleophthalmology tools, teleglaucoma requires more sophisticated diagnostic tests. The main tests are optic nerve photographs, Optical Coherence Tomography (OCT), Intraocular Pressure (IOP) measurements, central corneal thickness (CCT) measurements, and visual field tests [9]. The combination of examinations and equipment required can vary based on organizational resources, target goals and populations. However, the more diagnostic tools used during screening for glaucoma the greater the accuracy and effectiveness of the screening process. The equipment required for teleglaucoma are the ophthalmic examination equipment, cameras, and computer imaging software. The A full list of the standard equipment and components of teleglaucoma can be found in Table 1 [9].
The advantages reported in the literature include convenience, decreased travel time to medical clinics, increased access to specialized care for glaucoma, and decreased patient costs [10,11] The benefits are mainly seen in remote or underserviced communities such as Aboriginal communities and rural or remote areas where there is limited ocular specialists. Arora et al. reported improved access time (time from patient being referred to the date visit is booked) with teleglaucoma versus standard in-person examinations: 45 days for teleglaucoma versus in-person exam which had 88 days [12]. Teleglaucoma had reduced cycle time (time from registration until patient leaves clinic) of 78 minutes versus inperson exam of 115 minutes [12]. The pioneer teleglaucoma study conducted in Finland reported reduced absence from work by 50% with teleglaucoma versus inperson examination, and in addition reduced traveling (97%), costs (92%), and time (92%) [11].
The literature suggests teleglaucoma has comparable diagnostic accuracy. Teleglaucoma technology demonstrated moderate agreement in its ability to diagnose glaucoma (Kappa statistic 0.55% (0.48, 0.62)) [13]. When disc damage had Vertical Cup Disc Ratio (VCDR) greater than 0.7 the Frequency Doubling Technology (FDT) had a substantial agreement with ability to diagnose glaucoma (kappa statistic 0.84) [13]. In addition, a study conducted in rural India compared the ability of teleglaucoma to detect glaucoma compared to standard in-clinic examination and found that there was good agreement in detecting glaucoma [14]. For glaucoma the kappa scores were 0.61 with standard screening versus 0.59 for teleglaucoma [14]. In comparison to the in-person slit lamp examination, the positive predictive value was 77.5% for positive teleglaucoma diagnosis and had a negative predictive value of 82.2% for negative teleglaucoma diagnosis [13]. However, a cohort study conducted by the University of Alberta found 24% of teleglaucoma photographs were deemed unreadable from media opacities, patient cooperation, and unsatisfactory photographic techniques [13].
In this study, a systematic review and meta-analysis was conducted on teleglaucoma screening for patients with glaucoma to evaluate the following: the effectiveness of teleglaucoma as a screening device, its diagnostic accuracy, its diagnostic odds ratio, and its cost-effectiveness in comparison to in-person examination. Section 2 will explain the methods, section 3 provides the detailed analysis, and section 4 concludes with a discussion and implications for future research.  (Table S1). Alerts were set up for each database to receive publication notifications for new related articles.

Inclusion and exclusion criteria
Articles included were from any country, all in English, published from 1999 to current, and were research articles. The articles included study population that are adults in the general population or populations at risk of glaucoma. The study population included those with or without glaucoma. Articles on teleglaucoma intervention for glaucoma screening were included, both in-comparison to inperson screening and analyzing teleglaucoma on its own. Outcome measures of teleglaucoma articles selected contained efficiency measures, specificity, sensitivity, and its ability to detect glaucoma, as well as patient benefits and cost data. Economic evaluations such as cost-effectiveness analysis and studies with costing data were also included. The exclusion criteria was articles published prior to 1999 since teleglaucoma is fairly new and to be consistent with the teleglaucoma screening procedure, year 1999 was selected as a cut-off year. Additionally, non-research articles such as methodology papers, editorials, review articles, commentaries, and letters were excluded. Articles on diagnosis or prognosis, genetic screening, and teleophthalmology for ocular conditions other than glaucoma were eliminated.
A total of 11237 articles were retrieved by searching various databases and an additional 526 were retrieved from hand searching and grey literature search which were then imported into EPPI 4.0 reference manager. Based on the inclusion and exclusion criteria, two reviewers independently reviewed all articles.
After removing duplicate articles, 8157 articles were included for screening. Articles were screened by title, abstract, and full text in level 1, 2, and 3 screening respectively. After each level of screening, kappa statistics was calculated to measure reviewer's agreement. Additionally, if consensus was not reached by the two reviewers' then a third reviewer intervened to solve disagreements on article eligibility. The agreement between the two reviewers was excellent (kappa 50.86). The PRISMA diagram demonstrating the selection process is displayed in Figure 1 ( Table S3).

Quality Assessment Strategy
Articles were assessed using the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) guidelines for publication bias, risk of bias, imprecision, inconsistency, and indirectness [15][16][17][18][19][20]. Articles were graded as either low, moderate, or high quality of evidence. The results indicated that 17 articles were high quality, 13 were moderate quality, and 15 articles were graded as low quality of evidence. Despite the quality of evidence, all articles were included in the analysis.

Data Extraction Strategy
Qualitative and quantitative data necessary for analysis was obtained from each article. Information on study location, design, effect measures (sensitivity and specificity), percentage of glaucoma diagnosed, service times, image quality, visual acuities, ophthalmic characteristics, and costs were collected. One reviewer extracted data using an excel template. Authors were emailed to obtain missing relevant information. All databases were updated with new information from respective authors. Additional current costing data was provided by ophthalmic equipment vendors INNOVA, Topcon, and Ocular Health Network. Costs were converted to 2014 US dollars [21]. This research study has no financial relationships, investments, or sponsorship related to the cited commercial vendors.

Data Analysis
Data was synthesized and analyzed using STATA 13. When studies reported estimates as range or p-value or multiple estimates, mean and standard deviation (SD) were derived. Hierarchical logistic regression was used to determine the pooled estimates of sensitivity and specificity of teleglaucoma and in-person examination. A graphical representation of the summary estimates was presented in a Hierarchical Summary Receiver Operating Characteristic (HSROC) curve with 95% confidence intervals and 95% prediction regions.
The positive/negative likelihood ratios (LR+/LR2) were calculated using bivariate models to generate estimates of the likelihood of a positive/negative test in a glaucoma/non-glaucoma patient. From this result the diagnostic odds ratio (DOR) was calculated to determine the relative diagnostic effectiveness of The Effectiveness of Teleglaucoma for Glaucoma Screening  teleglaucoma. DOR is the ratio of the odds of a positive screen test in a glaucoma case relative to the odds of a negative screen test in a non-glaucoma case [22]. Due to the variability of study effectiveness measures, not one article had a complete set of data. Missing data was treated as statistically missing values and not included in the analysis. Only articles with complete data were included in each analysis. Publication bias was assessed using funnel plots.

Results
A total of 45 studies were included in this meta-analysis. Table 2 and Table 3 display the baseline characteristics of each study. Studies were conducted in fourteen different countries with representation in each continent. All articles were published between 1999 and 2014. The cumulated individuals of all studies were 101,512 participants. All studies were observational studies, as there were no randomized controlled trials conducted. Three studies contained economic evaluations or cost-effectiveness analysis. Of the 45 studies, 16 compared teleglaucoma to in-person examination. The other 29 studies analyzed teleglaucoma without comparison or was an evaluation of different teleglaucoma equipment. There was minimal variation in study populations; they included either glaucoma patients or patients who were at risk of glaucoma (based on diabetes status, family history of glaucoma, age, or ethnicity). Table 4 displays additional study details on demographics and study methods (glaucoma definition, pupil dilation, and number of field tests examined). Although there was some variation, less than 10% of studies reported these details. The main outcome measures were specificity and sensitivity (Table 5). Other included outcome measures (percentage of glaucoma diagnosed, referral rate, and proportion of images with poor quality) are displayed in table 5.
Costing data was given by nine studies and the quality of analysis of costing is displayed in table 6. Teleglaucoma costs vary by the capacity of the service and the type and amount of equipment. The current vendor estimate shows that the total costs for standard glaucoma equipment range from 89,703.53 to 123,164.55 US dollars (Table 6) [23,24]. Additionally, to transfer images and patient test results securely to ophthalmologists electronically a service exists costing $62.13 US/ month [21,25]. This service allows teleglaucoma technicians and ophthalmologists to login electronically to attach, send, view and assess retinal images and patient test results.
Costing data from the literature shows the cost per detected case of glaucoma ranged from $13.03-2020.96 US after conversion to US dollars and adjusted for inflation to 2014 costs (Table 7) [26]. The mean cost is $1098.67 US for every case of glaucoma detected (n53) ( Table 7). The mean cost of teleglaucoma per patient screened was $922.77 US (n52) ( Table 7).
Another necessary costing aspect is the ophthalmologist fee for glaucoma consultation. The ophthalmologist may be compensated for each teleglaucoma referral or time spent on teleophthalmology consultations. Compensation varies  The Effectiveness of Teleglaucoma for Glaucoma Screening  [27,28]. In Ontario, Canada, the compensation for the fee-for service model, is $16.00 CAN per ophthalmic referral [29]. The physician liable for teleglaucoma consultations must be a licensed ophthalmologist in both the area of the service and the patient. Physicians must hold liability coverage appropriate to state/provincial laws. In Canada, the Canadian Medical Protective Association provides ophthalmologists with liability coverage for teleophthalmology [30]. Ten studies had complete data necessary to conduct the analysis for teleglaucoma diagnostic accuracy. The summary estimate for sensitivity was 0.833 [95% CI 0.77, 0.88] and specificity was 0.79 [95% CI 0.668, 0.875] for glaucoma screening using optic nerve examinations ( Figure 2). The summary estimates indicate that teleglaucoma correctly detects 83.3% of glaucoma cases and correctly classifies 79% of those without glaucoma as glaucoma-negative. Figure 3 displays each study estimate and the summary estimate with its associated confidence intervals and the generated HSROC curve. The distribution of the studies in the  The Effectiveness of Teleglaucoma for Glaucoma Screening plot demonstrates the variability of both specificity and sensitivity amongst studies. Six studies fall outside of the 95% confidence interval of the summary estimate. The 95% prediction region is the estimate of future observations. The results demonstrate a fairly wide prediction region for both true predictions of specificity and sensitivity, with greater variability expected for specificity. The study populations used to assess diagnostic accuracy were those at-risk of glaucoma (based on diabetes status, family history, age, ethnicity, etc.), optometrist and ophthalmic clinic patients, and patients who were glaucoma suspects (Table 1). One study reported its study population as glaucoma patients only (Table 1) and contrary, this study had one of the lower reported scores for diagnostic accuracy: specificity was 71.5% and sensitivity was 67% (Table 5) [31].
The diagnostic tools of the included studies varied slightly (Table 8). Eight out of the ten studies analyzed for sensitivity and specificity used at minimum optic nerve examinations as part of the screening process ( Table 8). The other two studies reported using IOP or visual field defects as the methods to detect glaucoma suspects (Table 8). For these studies which did not include fundus photographs, the sensitivity and specificity were 81.5% and 95.5% respectively for glaucoma screening using only visual field and 38.1% and 98.8% respectively for glaucoma screening using IOP and Orbscan Topography (Table 5) [32,33]. Three studies reported sensitivity and specificity of in-person examination. The weighted mean of sensitivity was 74.9¡27.6% (n53) and specificity was 88.8¡10.3% (n53) for in-person examination. The summary estimates indicate that in-person examination correctly detects 74.9% of glaucoma cases and correctly classifies 88.8% of those without glaucoma as glaucoma-negative.
The positive likelihood ratio was 3.97 [95% CI: 2.3-6.7] while the negative likelihood ratio was 0.21 [95% CI: 0.14-0.32] (Figure 2). This demonstrates that the likelihood of a positive screen test in a glaucoma case is greater than the likelihood of a negative screen test in a non-glaucoma case. In addition, the positive likelihood ratio is greater than one and thus the positive screen test is associated with glaucoma. Since the negative likelihood ratio is less than one, the negative screen test is associated with the absence of disease [22]. The effectiveness of the diagnostic accuracy of teleglaucoma was given by the DOR, which was 18.7 [95% CI: 7.9-44.4] (Figure 2). The relative odds of a positive screen test in glaucoma cases are 18.7 times more likely than a negative screen test in a non- glaucoma case. Since the DOR was greater than one the test is discriminating between true positives and true negatives correctly [22]. There was insufficient data to conduct hierarchical logistic regression on the percentage of glaucoma diagnosed. Three of the 45 studies reported percentage of glaucoma diagnosed in both teleglaucoma and in-person examination necessary for analysis. The mean percentage of glaucoma diagnosed was 13.4% for teleglaucoma and 7.8% for in-person examination which suggests that teleglaucoma is capable of detecting more cases of glaucoma.
Other effectiveness measures of teleglaucoma were analyzed such as variables of healthcare service quality. The mean percentage of patients referred to specialist for consultation was 12.5¡7.8% (n56). The mean percentage of images that were of poor quality was 10.4¡6.7% (n57). It took a mean time of 75.6¡87.7 seconds (n54) to process the teleglaucoma images. Timing associated with teleglaucoma service is another measure of quality. The mean time for screening was 8.8¡5.1 minutes (n53). The time reported for ophthalmologist to make diagnosis was 34 minutes (n51). The mean reporting time was 7.6¡2.6 minutes (n56). Teleglaucoma gave a reduction for patient travel time of 61.23 hours (n51). Teleglaucoma had a mean access time (time from patient being referred to the date visit is booked) of 59.7¡9.9 minutes (n54) in comparison to 73.7¡29.8 minutes (n54) for in-person examination. The mean cycle time (time from registration until patient leaves clinic) for teleglaucoma was 81.7¡6 minutes (n52), which was less than that of in-person examination, 116¡2.5 minutes (n52). The mean proportion of patient satisfaction with teleglaucoma was 47.3¡8.8% (n52) while only 42% (n51) were satisfied with in-person examination.

Discussion
Telemedicine has demonstrated good use for offering glaucoma services to people of remote areas. Teleglaucoma is beneficial to remote areas as the physician is not required to see patients in person, which reduces wait times and shortens the length of ophthalmic consultations. Teleglaucoma avoids long distance travel and time wasted on commute. The results of the pooled estimates for diagnostic accuracy have shown teleglaucoma to be more sensitive and less specific than inperson examinations. Teleglaucoma is advantageous at detecting true positive cases of glaucoma, but has a higher rate of false positives in comparison to inperson examination. With very high DOR estimates, it is suggested that teleglaucoma can accurately discriminate screen tests. Teleglaucoma has demonstrated capability to detect glaucoma cases that may not have been detected during in-person examination. Glaucoma progresses without patient awareness and it is usually detected at the advanced stages. Thus teleglaucoma serves as a tool for early detection of glaucoma. If caught earlier and with treatment, glaucoma can be effectively managed and can result in the preservation of vision.
Telemedicine for glaucoma can have several combinations of examinations and measurements used for glaucoma screening. Examination of fundus photographs are commonly used for teleglaucoma screening. Four of the ten studies analyzed used only fundus examinations while another four studies included IOP, CCT, visual field loss, and visual acuity, in addition to fundus photograph examinations (Table 8). Two studies did not use fundus photograph examination but rather visual acuity, IOP, CCT, and ACT (Table 8). However, this is based on studies who explicitly stated the terms for ophthalmic examination. Some studies reported ''comprehensive eye examinations'' were performed, but did not explicitly state which examinations were performed, thus assumptions cannot be made. The use of different tests for glaucoma screening can potentially bias the results as the more diagnostic tools used during screening results in a greater probability of correct diagnosis naturally. However, the results did not show any significant differences in accuracy with studies which reported using multiple diagnostic tools. Interestingly, the specificity and sensitivity values reported ranged independent of the number and the type of examination used for teleglaucoma ( Table 4 and Table 8).
The combinations of examinations are dependent on financial and resource limitations of the hosting organization and can vary from small programs to very large programs. It is dependent on the target goals and target populations of the organization. However, the standard examinations recommended for glaucoma screening are those that can evaluate visual field defects, IOP, and the biological structure and function of the optic nerve. These include HRT, OCT, optic disc photography, RNFL photography, as well as FDT, tonometry, and perimetry [34]. There were limitations within the study. Insufficient data reported was a major limitation of the meta-analysis, although authors were contacted for additional information. Nevertheless, the key goal was to systematically review the literature on tele-glaucoma and in-person screening and perform the meta-analysis. With small samples sizes there was not enough power to show statistical or clinical significance. Different comparators were reported by studies and to ensure internal validity, only studies with exact comparators were analyzed together. This was one of the reasons for reduced sample sizes for the analysis. However, our analysis does provide information on diagnostic accuracy of teleglaucoma, its capability to detect glaucoma, and to detect negative and positive cases correctly. It demonstrates teleglaucoma has the potential as a screening device to detect a greater amount of cases than in-person examination. Since teleglaucoma is an active screening, it suggests glaucoma cases are detected at earlier stages. However the significance of this difference is limited by the number of comparative studies. The majority of the studies were non-comparative which, in addition, limits the significance of the relative effectiveness to in-person examination.
Teleglaucoma has been evaluated in many different ways: diagnostic accuracy, cost reduction, technological capabilities (image quality, image transmission speed, etc.), reduction of patient and health care provider time, and convenience. Thus many studies focus on only part of the effectiveness. As a result, there is insufficient data when summarizing all of the studies together. This has proven the need for more research literature on the diagnostic accuracy of teleglaucoma and its ability to detect glaucoma in comparison to in-person examination. There is a need for research on the follow-up of detected cases and long-term effects of teleglaucoma. In addition, better quality of evidence through randomized controlled trials is recommended. There are implications for cost-effectiveness analyses. Although, costing data suggests cost savings for patients' time and travel with teleglaucoma, a thorough costing of current health care expenditure is required to determine its overall cost-effectiveness from the scope of the healthcare system.
Teleglaucoma is beneficial to offering services in underserviced regions and rural areas. It considerably reduces patient access times and cycle times. The time required for service is shorter than in-person examination and physician commitments are reduced. As a result teleglaucoma saves costs to patients and costs to the health care system as a whole.