Web-based telemonitoring of visual function and self-reported postoperative outcomes in cataract care: international multicenter randomized controlled trial

Cataract patients in this study effectively used a web-based platform to independently assess visual function and report postoperative outcomes, highlighting its potential for telemonitoring after cataract surgery.

Although national protocols vary, conventional cataract surgery follow-up typically involves routine face-to-face examinations to assess postoperative visual function and detect adverse events. 4Advances in surgical techniques and routine use of intraocular antibiotics have greatly improved safety of this procedure, and most of the postoperative check-ups are uneventful. 5,6This renders cataract surgery follow-up a compelling field for using telehealth for remote care delivery.
This trial is the first to evaluate cataract surgery follow-up involving remote self-assessments of visual function and self-reported outcome measurements collected at home, compared with conventional clinical practice where those assessments are conducted at clinics and obtained by trained staff.

METHODS Study Design
A detailed overview of the study design can be found in the published study protocol. 7This randomized controlled trial was performed at 6 eye clinics: 4 located in the Netherlands (University Medical Center Utrecht, Maastricht University Medical Center+, Amphia Hospital Breda, Oogcentrum Noordholland), 1 in Austria (Vienna Institute for Research in Ocular Surgery), and 1 in Germany (Augenklinik Sulzbach).The study was registered at ClinicalTrials.gov,number NCT04809402, and the protocol for this study was designed according to the SPIRIT 2013 guidelines. 8he study was approved by the Medisch Ethische Toetsingscommissie Utrecht, the Netherlands (NL74625.041.21); the Ethikkommission der Stadt Wien, Austria (EK 20-334-0121); and the Ethikkommission Saarbrücken, Germany (Ha 44/18).Written informed consent was obtained from all participants.
Between April 2021 and January 2023, consecutive patients planned for bilateral cataract surgery (either delayed sequential or immediate sequential) were invited to participate.Exclusion criteria were cataract surgeries combined with other procedures; presence of ocular comorbidities that negatively influence postoperative visual acuity (VA), such as amblyopia, age-related macular degeneration, diabetic retinopathy, glaucoma, or uveitis; an insufficient command of the Dutch, German, or English language; and not being able to access the web-based eye test (Easee BV, Amsterdam, the Netherlands) at home.The latter required access to a computer or tablet and a smartphone (possibly obtained by a relative) and an internet connection.Before enrollment, all interested patients were requested to try a demo version of the webbased eye test (a shortened test version).Those who were unable to access this demo were excluded from participation.Assistance by a relative was always recommended, though not strictly required.
After enrollment, patients were randomly assigned (1:1) to 1 of the 2 follow-up groups, using a web-based system stratified by center and age (<69 and ≥69 years).Participants of the telemonitoring group performed postoperative self-assessments at home (remote vision self-assessments and questionnaires) while those of the usual care group had conventional postoperative consultations dictated by local preferences and regulations.The postoperative self-assessments were offered at 3 distinct points in time: an early assessment within 1 to 7 days postoperatively, a late assessment 4 to 6 weeks postoperatively, and a final assessment at 3 months postoperatively.All participants underwent a remote assessment of patient-reported outcome measurements pre and postoperatively, and all participants underwent an in-office clinical assessment 4 to 6 weeks postoperatively.The telemonitoring group was invited to perform the remote assessments within 5 days before the in-office assessment at 4 to 6 weeks postoperatively.As such, a method comparison study design was embedded in the telemonitoring arm, comparing the home-based self-assessment with the conventional in-office clinical assessment of VA and refractive error.An overview of the study assessments is presented in Figure 1.Invitations for questionnaires and/or webbased eye tests were sent through email.Any unexpected consultations or adverse events were recorded for both groups.

Outcome Measures
Accuracy of the Web-Based Vision Self-Assessment (Telemonitoring Group Only) The web-based eye test evaluated in this study allows self-assessments of visual function at home.Easee (Amsterdam, the Netherlands) uses an ISO 13485 Quality Measurement System, and the tool is classified as Conformité Européenne class 2A medical device according to the Medical Device Regulation 2017/745. 9An overview of the test flow and assessed determinants of visual function is presented in Figure 2.
The web-based test's accuracy was evaluated by comparing the home-assessed uncorrected distance VA and refraction to the outcomes obtained by trained staff during the validation visit at 4 to 6 weeks postoperatively.These assessments only refer to the telemonitoring group, and comparisons are made of data acquired in the same subject.Repeatability of the web-based test was evaluated by comparing the outcomes obtained 3 months postoperatively with those obtained 4 to 6 weeks postoperatively.
Participants were asked whether they had experienced a change in vision before the start of the final eye test, and if so, these cases were excluded from the analysis.][12][13] The web-based VA assessment consists of 3 parts.In each part, a different optotype is used (Figure 2).This results in 3 individual scores that are averaged to calculate the final web-based VA score.A human over-read was performed after data collection.All web-based assessments were checked on individual levels to detect a within-test variation >0.2 logMAR between the 3 separate VA tests of 1 monocular assessment.As the true web-based test score cannot be reliably determined for these eyes, these were excluded from the comparisons and discussed separately.Conventional VA was assessed by Early Treatment Diabetic Retinopathy Study (ETDRS) charts at 4 meters (distance) and Sloan ETDRS charts at 40 centimeters (near).
The test's algorithm will interpret any VA poorer than À0.1 logMAR (ie, 1.25 Snellen decimal) to be caused by a refractive error.Astigmatic dials will determine the presence and magnitude of a cylinder.The type of refractive error (myopia or hyperopia) is based on an adapted red/green duochrome test combined with a short questionnaire.Algorithm updates had taken place after the start of the trial, resulting in a new method of calculating the spherical equivalent (SEQ).This updated algorithm performance is reported alongside the performance of the original algorithm.Subjective refraction was obtained at the clinic by trained optometrists masked for the web-based outcomes.After this assessment, corrected distance VA (CDVA) was measured using the prescription of both the subjective refraction and the web-based assessment.
Safety For both study groups, all adverse events and unexpected consultations were registered.Participants of the telemonitoring group were requested to fill out a triaging questionnaire before each web-based vision assessment.Outcomes of these questionnaires were evaluated.
[16] Statistical Analysis Statistical analyses were performed using IBM SPSS Statistics for Windows v. 29.0 (IBM Corp.).All quantitative variables were summarized.The data were tested for normal distribution, and the appropriate (non-)parametric test was used.For all analyses, a P ≤ .05 is considered statistically significant.Comparisons between the web-based assessments and the conventional clinical assessments were analyzed with paired t tests and in line with the Bland-Altman methodology. 17Mean differences (ie, bias) and 95% limits of agreement (LoA) (ie, sampling error) were presented.Residual refractive error outcomes were converted to absolute values of the SEQ.Associations between clinical characteristics and accuracy of the web-based test were evaluated by generalized estimating equation (GEE) analysis correcting for bilaterality, age, sex, myopic target refraction, and self-reported vision symptoms.Scores of the self-reported outcome measurements were calculated using official scoring manuals and conversion sheets.Differences between the 2 follow-up groups regarding these outcomes were evaluated with an independent samples t test.

Recruitment
A total of 391 patients were invited to participate.The participant recruitment is depicted in Supplemental Figure 1 (available at http://links.lww.com/JRS/B155).The willingness and ability of invited patients to enter the study varied between countries: 41% in the Netherlands, 14% in Germany, and 9% in Austria.There was no difference in age between those who were willing and able to participate and those who were not (70 ± 7 vs 71 ± 11, P < .001).
A total of 94 participants (188 eyes) were enrolled at baseline.Table 1 presents their baseline characteristics.

Accuracy of Web-Based Vision Self-Assessment (Telemonitoring Group Only)
Comparison of Web-Based vs Conventional VA Assessment The graphical comparison between the VA assessments at home and at the clinic at 4 to 6 weeks postoperatively is depicted in Figure 3, A. These comparisons were made within the same subject and only refer to the telemonitoring group.The mean difference was À0.03 ± 0.14 logMAR (P = .07),indicating that there is no systematic under or overestimation of the VA by the web-based tool.The 95% LoA ranged from À0.30 to 0.24 logMAR, and most assessments fall within the predetermined acceptable range of ±0.15 logMAR.Repeatability of the VA assessment showed a fairly similar precision: a mean difference of 0.02 ± 0.14 logMAR (P = .15),with 95% LoA ranging from À0.25 to 0.29 logMAR (1 vs 3 months postoperatively, see Supplemental Figure 2, available at http://links.lww.com/JRS/B156).Six participants were excluded from this repeatability comparison as they had reported a change in vision over the 2-month interval.GEE analysis revealed no associations between any of the examined clinical variables and accuracy of the web-based VA assessment (see Supplemental Table 1, available at http://links.lww.com/JRS/B157).
Comparison of Web-Based Refraction vs Manifest Refraction at the Clinic A graphical comparison between the web-based and subjective refraction outcomes at 4 to 6 weeks postoperatively is depicted in Figure 3, B. These comparisons were made within the same subject and only refer to the telemonitoring group.The mean difference in SEQ was 0.15 ± 0.67 D (P = .07),with 95% LoA ranging from 1.15 to 1.46 D. The updated algorithm generated a reduced variability, yet with a higher and significant systematic bias: a mean difference of 0.32 ± 0.43 D (P < .01),with 95% LoA ranging from À0.54 to 1.16 D. GEE analysis revealed that age and the presence of selfreported symptoms in the triage questionnaire before the web-based test are significantly associated with the accuracy of the SEQ assessment, although the extent of these effects (b = 0.01 and b = 0.17) are not considered clinically relevant (see Supplemental Table 1, available at http://links.lww.com/JRS/B157).
The data demonstrate that the algorithm tends to overestimate the absolute residual refractive error.The web-based test's ability to determine a residual refractive EuroQol 5-Dimension 5-level questionnaire [14][15][16] error >0.5 D has a sensitivity of 88% and specificity of 67%, meaning that there will be a substantial number of false positives (ie, patients with emmetropia will falsely be classified as ametropic).The attained mean CDVA achieved with the algorithm's web-based refraction increased compared with the mean preoperative CDVA (0.1 vs 0.3 logMAR), yet was inferior to the mean CDVA achieved with the prescription of the subjective refraction assessed at the clinic: 0.1 vs À0.1 logMAR (ie, 0.8 vs 1.2 Snellen decimal).An improvement of 1 or more lines from the preoperative VA was achieved for 63% of the eyes using the web-based refraction and for 96% of the eyes using the manifest refraction.
Human Over-read An unacceptable high within-test variation (ie, during the same monocular examination) was identified for 15 assessments (in 13 patients); these were excluded from the overall analyses.This subgroup was not evidently different at baseline in terms of age (71 ± 5 years) or sex (46% female) when compared with other telemonitoring participants.Interestingly, 7 of these eyes had a myopic target refraction (ie, half of all myopic eyes included in the trial) and 1 eye was aphakic because of an interoperative complication; underlining that a poor uncorrected distance VA is associated with a high within-test variation.GEE analysis confirmed this with a positive association between a myopic target refraction and a high within-test variation.No other clinical parameters were identified in this aspect (see Supplemental Table 1, available at http://links.lww.com/JRS/B157).

Safety
Adverse Events and Postoperative Management Changes Most of the participants did not have a face-to-face consultation in the first week postoperatively: Only 18 participants visited the clinic (36% of the usual care group).These visits had been scheduled preoperatively in line with the clinic's usual practice and were all uneventful.
In addition to the scheduled consultations, 12 usual care and 7 telemonitoring participants contacted or visited the clinic (24% vs 16% of each group total) (as defined in Figure 1).These unexpected consultations and consecutive management changes are presented in Table 2. Three surgical complications were registered: 1 residual cortical lens matter in the anterior chamber, 1 wrong lens implantation, and 1 eye was left aphakic because of intraoperative difficulties.All of these underwent successful reoperations.
Triage Questionnaires In addition to the web-based eye tests, the telemonitoring participants were asked to fill out triage questionnaires.The reported symptoms are presented in Table 3.

Self-Reported Outcome Measurements
The response rate of the questionnaires at baseline and 3 months postoperatively was 100%.An overview of the self-reported outcome measurements is presented in the Supplemental Table 2 (available at http://links.lww.com/JRS/B158).Visual functioning (indicated by the Catquest-9SF and NEI-VFQ-25 scores) improved postoperatively for all participants (with a mean improvement of À0.80 and 16.70, respectively), and performing vision self-assessments did not materially affect these scores, as there were no significant differences between the 2 follow-up groups.The effects of cataract surgery on EQ-5D-5L index scores were marginal in both groups.

DISCUSSION
Our study evaluated remote cataract surgery follow-up involving self-assessments of visual function and self-reported outcome measurements using telehealth technology.The studied web-based postoperative VA assessment, performed independently at home, showed a mean difference of 0.03 ± 0.14 logMAR when compared with conventional ETDRS testing.This negligible mean difference indicates that there is  12) 74 (15) General health condition: self-rated health score at baseline (scale 0%-100%), mean (SD) 83 ( 12) 82 ( 15) EQ-5D-5L Index Values (scale 0.00-1.00),mean (SD) 0.96 (0.06) 0.91 (0. no fixed bias, meaning that the web-based test does not systematically over or underestimate VA.The 95% LoA (ie, the distribution of differences) ranged from À0.30 to 0.24 logMAR, indicating that outlier assessments occurred beyond the predetermined clinical deviation limit of ±0.15 logMAR.The web-based refraction algorithm resulted in an overestimation of the residual refractive error and will falsely indicate a residual refractive error in patients with emmetropia.Adverse events were rare and occurred in both groups.Telemonitoring and self-reporting led to a more detailed overview of postoperative symptoms and complaints while unscheduled consultations were recorded less often.We acknowledge that the study was not powered on adverse events and postoperative management changes.
Vision-related quality of life improved postoperatively with no apparent differences between groups.
The mean difference between the web-based VA assessment and the conventional chart was clinically negligible.However, when evaluating agreement, it is crucial to look into the distribution of differences.It is important to realize that a certain degree of variation is unavoidable when comparing different charts, given the diversity in optotypes and scoring criteria. 11,18An authoritative study comparing repeated VA assessments with various charts and observers concluded that differences exceeding ±0.15 logMAR should, therefore, be considered clinically relevant. 11In our study, we identified outcomes beyond this range.It should be noted that the occurrence of variation in the absence of clinical changes is a common feature of VA testing, attributed to behavioral factors such as fatigue or intrinsic motivation.This has been demonstrated by repeatability (ie, test-retest studies) that evaluate repeated VA assessments using the same chart in the same person.For the ETDRS chart, repeatability studies have shown 95% LoA ranging up to ±0.18 logMAR. 12,19For the Snellen chart, the variability is higher, with 95% LoA ranging up to ±0.24 logMAR using the singleletter method and up to ±0.33 logMAR using the line assignment method (ie, the test is terminated when half of the letters on the line are misread). 12,19The wide distribution of measurement variability indicated by the 95% LoA underlines that the occurrence of outlier measurements is an inherent aspect of assessing VA.Notwithstanding, the herefound agreement is subpar to the precision reported in ETDRS test-retest studies, prompting future efforts to increase the measurement precision of the web-based test.
Multiple telehealth tools for self-assessing visual function are available on the internet, although many have not been validated. 20Those that are validated have mostly been evaluated in controlled settings, prone to observer bias. 19A previous study evaluated the performance of the herestudied web-based test in a supervised hospital setting and reported a slightly better precision regarding the uncorrected VA assessment after cataract surgery (95% LoA À0.26 to 0.15 logMAR). 21Unsupervised home assessments will arguably be more affected by behavioral and environmental factors than assessments in controlled settings.In this study, the web-based test was performed  independently by patients at home, reflecting a real-world scenario.
Future software updates should be aimed at improving the measurement precision of the web-based test.This involves new technical integrations, for instance, by using webcam images and providing AI-guided live feedback to optimize testing conditions at home (eg, considering lighting conditions and distance to the screen).Moreover, the ease by which repeated measurements can be performed is an interesting feature of home-based self-assessments, potentially reducing the variability of VA results.The web-based test could benefit from automated detection of within-test inconsistencies indicative of poor performance, with the subsequent functionality to let patients repeat a specific part of the test or to redo the complete test at a later point in time under optimal testing conditions.
It is important to address the suboptimal performance of the web-based refraction assessment in this study population.While an improvement in VA is indicative of surgical success, most cataract guidelines also recommend a refraction assessment postoperatively. 22Our findings indicate that the current version of the webbased refraction assessment is not able to reliably determine refractive error in this patient population.The current version has a tendency to falsely identify a residual refractive error in patients with emmetropia.This is an important limitation.The algorithm, therefore, requires a recalibration before it can be useful to collect surgery outcomes for mandatory registries.It was trained on a population of young adults (aged 18 to 40 years) with an excellent VA (>1.25 Snellen decimal). 23VA has been reported to decrease with age, even in healthy eyes, particularly after 55 years of age. 24The assumption in the algorithm that any VA poorer than À0.1 logMAR requires a refractive correction will, therefore, by design result in an overestimation of the residual refractive error in older adults.A data set comprising normative best corrected VA scores after uneventful post-cataract surgery (based on 490.240 records) has been obtained from the European Registry of Quality Outcomes for Cataract and Refractive Surgery and will be used to recalibrate the algorithm.An updated and accurate self-assessment of residual refractive errors postoperatively would be an important step for automated data collection and quality registration.
The quality of vision as perceived by the patient may be influenced by parameters other than VA.Patient-reported data on vision-related quality of life have become important indicators for surgical success. 25These parameters are often under-reported in conventional care.Interestingly, the response rate of these questionnaires in this study was 100%.We believe that collection of self-reported data relies on adequate instructions, a perceived benefit of registration, and a good and accessible digital infrastructure.Therefore, engaging patients in their eye health by self-monitoring their vision might build on this perception of usefulness and lower thresholds to fill out these outcome-related questionnaires.In addition, a well-integrated web-based platform could be of great value for maintaining feedback loops to cataract surgeons, by collecting data that are relevant for quality control and benchmarking.
Public support is of paramount importance for successful adoption of a remote follow-up practice.Some eyecare professionals might be reluctant to adopt remote care practices as they fear vision-threatening postoperative complications.Notwithstanding, a substantial body of clinical experience and scientific evidence has demonstrated the safety and efficacy of telephone follow-up alternative to face-to-face visits. 5,26Unexpected management changes are unlikely in asymptomatic patients after uncomplicated surgery, and a structured set of clinical questions (such as pain, redness, and decreased vision) has been proven sufficient to risk-stratify patients. 22,27A recent innovation is an automated telephone consultation with AI-driven clinical assistants, combining a low-tech method of delivery with a high-tech large language modeldriven solution. 28A potentially more accessible and scalable alternative could be a self-administered digital questionnaire, as in this study, or a combination of both: an interactive questionnaire providing personalized feedback based on the responses.Patients could, for instance, be reminded to apply their lubricating drops as most reported symptoms in the triage questionnaires relate to dry eyes.An automated alert system, integrated with the clinic's electronic health record should identify those in need of a follow-up consultation.
Finally, we would like to underline that conventional follow-up care involving face-to-face visits should always remain accessible.0][31] Regional differences in the ability and willingness to participate were identified, with participation rates being highest in the Netherlands.This was mainly rooted in technology adoption barriers.In Germany and Austria, a notable proportion of invited patients could not participate because of a lack of access to mobile devices or insufficient proficiency to perform selfassessments at home.The participant rates align with Eurostats data indicating that the Netherlands surpasses both Germany and Austria by having the highest levels of internet usage and greatest number of citizens possessing basic digital skills in Europe. 32Internet access has, however, been growing rapidly across Europe over the past decade, and we expect the group of digitally skilled patients with cataract to grow exponentially in the upcoming years.In reality, a group of patients will remain not suitable for remote follow-up.Preoperative counselling and identifying those at risk of complications and postoperative management changes is crucial.4][35][36] In addition, even when the patient was initially planned for remote follow-up, thresholds for call-backs to the clinic should be low.
In conclusion, our study showed that a group of digitally skilled patients with cataract were able to complete unsupervised self-assessments of visual function and selfreport postoperative outcome measures.This underlines the potential of telemonitoring after cataract surgery.Adverse events or (vision-related) quality of life did not differ between the 2 follow-up groups.Future adjustments should aim to further improve the precision of the webbased test.The biggest challenges revolve around the inevitable occurrence of outlier assessments, which could be more frequent when vision assessments are performed unsupervised by patients at home.Software updates of the web-based tool should be aimed at identifying and limiting these outliers, and design automated feedback loops that invite to retest when an outlier is suspected.The current version of the web-based refraction assessment does not deliver reliable outcomes in this population and requires a recalibration, with future training building on a normative data set derived from the European Registry of Quality Outcomes for Cataract and Refractive Surgery.We believe that remote self-care can be a promising avenue to comply with increasing demands of cataract care, and improvements of technology and public support could make web-based self-collection of postoperative outcome data become a reality.

WHAT WAS KNOWN
Demands for cataract surgery are ever-increasing.Although national protocols vary, conventional cataract surgery followup typically consists of routine face-to-face examinations to assess postoperative visual function and detect adverse events The high volume of surgical procedures and low adverse event rates make routine cataract surgery follow-up an interesting field for using telehealth for remote care delivery.A web-based platform for collecting self-assessments of visual function and self-reported clinical outcome measures has yet to be evaluated.

WHAT THIS PAPER ADDS
Digitally skilled patients with cataract are able to complete unsupervised self-assessments of visual function and selfreport clinical outcome measures using a web-based platform.Both conventional and web-based follow-ups yielded similar patient-reported outcome measurements outcomes and adverse event rates, suggesting less unscheduled clinical visits with web-based follow-ups.Future adjustments should aim to further improve the precision of the web-based test.The biggest challenges revolve around the inevitable occurrence of outlier assessments, which could be more frequent when vision assessments are performed unsupervised by patients at home.

Figure 2 .
Figure 2. Overview of the webbased test flow.Collected parameters of the web-based test include CDVA/UDVA and refractive error (sphere, cylinder, axis).

13 )
EQ-5D-5L = EuroQol 5-Dimension 5-level; NEI-VFQ-25 = National Eye Institute Visual Function Questionnaire-25; SEQ = spherical equivalent a Preoperative VA was assessed with current prescription, if applicable b Target refraction was similar for both eyes c Potentially complicated surgical cases where not explicitly excluded from participation

Figure 3 .
Figure 3. Comparisons of web-based vs conventional assessments (telemonitoring participants only).Please note that these 2 assessments were compared within the same subject and, therefore, only refer to the telemonitoring group.The straight lines represents the mean difference; the dashed lines the 95% limits of agreement.The semitransparent blue zone in A depicts the predetermined clinically acceptable deviation range of ±0.15 logMAR.Note that an update of the web-based refraction algorithm had taken place after the trial, resulting in a new method of SEQ calculations.The blue lines and blue circles in B represent the original web-based refraction algorithm; the red lines and red circles represent the updated algorithm.SEQ = spherical equivalent

Table 2 .
Reasons for unexpected consultations (independent of the study scheme)

Table 3 .
Outcomes of self-reported triage questionnaires