A prototype protocol for evaluating the real‐world data set using a structured electronic health record in glaucoma

As the first step in monitoring and evaluating day‐to‐day glaucoma care, this study reports all real‐world data recorded during the first full year after the implementation of a prototype for glaucoma‐specific structured electronic healthcare record (EHR).


| I N T RODUC T ION
While randomized controlled trials (RCTs) evaluate interventions in one eye of selected patients in optimal conditions (efficacy), the evaluation of interventions and care practices in real-world conditions (effectiveness) requires different data sets and evaluation tools (Franklin et al., 2021;Porzsolt et al., 2015;Thompson, 2021).For example, patients in everyday practices have two eyes with variable comorbidities and may be cared by practitioners with unwarranted variations in the delivery of eye care (MacEven et al., 2019).
When targeting to improve the real-world costeffectiveness, what is done clinically in everyday practices needs to be measured and evaluated routinely, efficiently and non-selectively (WHO, 2023).Therefore, the Finnish University Eye Clinics have developed a holistic, yet simple and comprehensible aces-rwm™ ecosystem model (automation in care and evaluation of system with real-world monitoring) to deal with complex challenges in everyday eye care (Tuulonen et al., 2022).The overall aim of the ecosystem is to produce the best possible well-being and eyesight with the available resources.

O R I G I N A L A R T I C L E
A prototype protocol for evaluating the real-world data set using a structured electronic health record in glaucoma Sakari Sulonen 1  The aces-rwm™ ecosystem advocates a strategy to optimize real-life cost-effectiveness, sustainability and outcomes of the service delivery in ophthalmology.The ecosystem consists of three components: (1) resource governing principles to deal with increasing demand and limited resources, (2) real-world monitoring to collect structured real-world data (RWD) using structured electronic health records (EHRs) as well as measuring health-related quality of life and costs and (3) digital innovation strategy to evaluate and benchmark realworld outcomes and cost-effectiveness between eye care units nationally and internationally.The core value and strength of the ecosystem lie in the consensus and collaboration of all Finnish University Eye Clinics to collect and evaluate the uniformly structured outcome data using EHR (Tuulonen et al., 2022).
RWD collection represents the first step in the development of a framework for routine monitoring and evaluation of what gets done in everyday glaucoma care with respect to the defined strategies.The 4-year Finnish University Eye Clinics' project to develop tailor-made digital tool package for real-world monitoring in glaucoma, age-related macular degeneration, diabetic retinopathy and cataract was completed on 31 March 2023 and is ready to put to use.The experiences of Tays Eye Centre's prototype for structured EHR in glaucoma, implemented in 2018, were utilized during the development project.
In accordance with the CODE-EHR best-practice framework preferring a published protocol for each item on the checklist (Kotecha et al., 2022), the purpose of this study was to report all real-world data recorded during the first full year (2019) after introducing a prototype of glaucoma-specific structured electronic healthcare record.

| Background
Tays Eye Centre, Tampere University Hospital, Finland, is the only public unit providing eye care services for the population of 0.53 million in Pirkanmaa Wellbeing Services County.Due to the continuously increasing demand for glaucoma services, the first structured paper data collection tool was implemented in Tays Eye Centre in 2012.The collected data were typed out by secretaries into the hospital's general unstructured digital patient record.Simultaneously, nurses in the Medical Glaucoma Clinic were trained to fill in all examination data, preevaluate the fundus photographs and visual field (VF) tests and suggest a treatment and follow-up plan to be checked by a glaucoma specialist, either face-to-face or virtually (Tuulonen et al., 2016).The first prototype of glaucoma-specific EHR was tailor-made for Tays Eye Centre and implemented in autumn 2018 after a 2-year development process.The first full-year EHR data are available for 2019.
The use of structured EHR offers a unique opportunity to develop a learning healthcare system by providing a new evidence generation for large-scale clinical research, including cost-effectiveness (Kotecha et al., 2022).Therefore, a global multistakeholder group developed a 5-item CODE-EHR best-practice framework and checklist for researchers when designing and reporting big data.In addition, CODE-EHR can be used by journal editors and reviewers by requesting the authors to fill in the checklist when submitting structured healthcare data, thus embracing values of transparency, reciprocity, inclusivity and service for the common good.Consequently, we decided to follow the CODE-EHR best-practice framework (including the underneath headings) and fill in the checklist when reporting our RWD (Kotecha et al., 2022).

| Data set construction and data fitting the purpose
The glaucoma prototype EHR was used in all 4618 patients visiting Tays Medical Glaucoma Clinic in 2019.The standardized care process in Tays follows the Finnish Current Care Guidelines, which have been available for 20 years (Tuulonen et al., 2003), with the latest update in 2023 (The Finnish Medical Society Duodecim, 2023).These guidelines also form the basis for national access to care criteria (Ministry of Social Affairs and Health, 2019a).Collected EHR data are presented in Tables 1-3.Diagnostic codes for glaucoma follow the International Classification of Diseases (Tenth Revision, ICD-10).As laser and surgical procedures were not included in the prototype, we only report their rates for 2019.
The Finnish healthcare system covers the whole population, and its services are primarily tax-financed.Glaucoma medications are reimbursed for patients with an ICD-10 code for glaucoma and in accordance with the '2 out of 3' rule (defined underneath in Disease and outcome definitions), as well as for patients with IOP ≥30 mmHg.The data for reimbursements of glaucoma medication expenses including number of recipients as well as prescriptions and cost data were received from the Social Insurance Institution of Finland, which also include the number of patients followed solely in private practice, including their number of VF and imaging tests but without clinical data (Kela, 2023).
A flow diagram of different data sets is presented in Figure 1, including missing data and their proportion for each variable, and specifications for linked data sets.Regarding the data fitting the purpose and to ensure transparency, the data include all structured clinical data, which were recorded in EHR of all patients visiting the Medical Glaucoma Clinic at Tays Eye Centre in 2019.

| Disease and outcome definitions
The glaucoma care process in Tays Eye Centre follows the 20-year-old Finnish Current Care Guideline for glaucoma in which glaucoma is defined using the '2 out of 3' rule, that is, when at least two concomitant findings are evaluated as glaucomatous in optic nerve head (ONH), retinal nerve fibre layer (RNFL) images and VF tests 1.9 ± 0.9 2.0 ± 0.9 2.0 ± 0.9   2a and 2b).
In accordance with the Finnish Guideline, the glaucoma test set in Tays Eye Centre is evaluated clinically and consists of ONH and RNFL digital photographs (Tuulonen et al., 2000) using Canon CX-1 camera (Canon Medical Systems Europe BV) and VF tests using Humphrey 24-2 fast or faster programs (Carl Zeiss Meditec Inc.).The test set is taken in Tays Eye Centre at diagnosis and thereafter every 1-2 years depending on the patients' risk profile.The test set does not include optical coherence tomography (OCT) because the Finnish Current Care Guideline (The Finnish Medical Society Duodecim, 2023) and the European Glaucoma Society Guideline do not recommend glaucoma diagnosis or follow-up using OCT only (European Glaucoma Society, 2021).In accordance with the Finnish guidelines, central corneal thickness measurements are not included in the glaucoma test set in Tays Eye Centre.
For patients with stable glaucoma, a 1-to 2-year treatment and monitoring plan is created as recommended by the Finnish Glaucoma Guideline.Patients are instructed to have their IOP measured in their local opticians' shops with a rebound tonometer.The frequency of IOP measurements follows each patient's individual monitoring plan, typically every 6 months in stable glaucoma.An 'ideal' target pressure is recorded for all patients, that is, at least 25% reduction from the untreated IOP level taking additionally into account the IOP level at which glaucomatous changes might have progressed.Simultaneously, an estimate for maximum tolerable IOP is defined, that is, an IOP level, which is considered to require a change in treatment.In case patient's IOP reaches the pre-defined maximum tolerable level, the patient knows what to do as their individual monitoring plan includes also the next two interventions on how to proceed.In this case, patients are instructed to call Tays Eye Centre to have the pre-defined treatment changes implemented.The Finnish Glaucoma Guideline has separate directions on how to treat and follow-up patients with high-risk glaucoma (The Finnish Medical Society Duodecim, 2023; Tuulonen et al., 2003).
Patients with ocular hypertension (OHT) < 30 mmHg or glaucoma suspicion without elevated IOP and without glaucomatous damage are not routinely followed in Tays Eye Centre.Despite normal ONH, RNFL and VF findings, some of these patients may be invited to have another glaucoma test set within 1-2 years to rule out progression.The OHT patients with IOP < 30 mmHg are instructed to have their IOPs measured typically once a year in opticians' shops.In case the IOP increases, their monitoring plan instructs them to call Tays Eye Centre for a control glaucoma test set.In patients with ≥30 mmHg, treatment is initiated in all patients favouring selective laser trabeculoplasty as primary therapy.
The overall goal of Tays Medical Glaucoma Clinic is to organize the high-volume care for 'usual' glaucoma patients and to be able to detect the progressive cases early enough for glaucoma surgery.Although the glaucoma EHR prototype did not include data collection tools for surgical consultations and procedures, such tools have been built in the next EHR version that will be implemented in 2023 (Tuulonen et al., 2022).

| Analysis
For the analysis, the distributions are presented in two ways: (1) according to the recorded diagnosis in the EHR (Table 1) and (2) in four groups based on the '2 out of 3' rule (Tables 2 and 3): • '≥ 2/3 glaucoma' refers to 2-3 compatible glaucomatous findings in ONH, RNFL and/or VFs and an ICD-10 code for glaucoma.• '1/3 glaucoma' refers to 0-1 glaucomatous findings and an ICD-10 code for glaucoma.• '≤ 1/3 suspected glaucoma' refers to 0-1 glaucomatous findings in either eye and no ICD-10 for glaucoma.• Mixed group: All remaining eyes not included in the above groups, that is, mixed diagnosis between eyes indicating either nonmatching diagnoses in the left and right eyes, and/or the two eyes belonged to different subgroups based on the '2 out of 3' rule.
Progression is evaluated clinically by comparing the most recent ONH and RNFL images and VFs to the first ever taken images and VFs as well as all tests taken during the follow-up (Figures 2a and 2b).If the tests are only compared to those taken 1-2 years earlier, the progressive changes may be hard to detect unless they are significant.Progressive changes in any of the three tests (ONH, RNFL and VFs) are marked as 'progression' in the EHR.Progression leads to increase in medication, laser therapy and/or surgery, taking simultaneously into account the severity of the damage, the progression rate and patient's age, as recommended in the Finnish Guidelines (The Finnish Medical Society Duodecim, 2023).
Although reimbursement of medication requires the 2 out of 3 criteria for glaucoma, they may not have always been strictly followed, which may lead to over-diagnostics and over-treatment.For example, patients may be referred to Tays by private practitioners who may have already started the treatment and given a glaucoma diagnosis to the patient, sometimes even years before visiting Tays.

| Ethics and governance
This study was approved by the Tays Research Services (number R21519/2021) and was conducted in accordance with the Declaration of Helsinki.The Ethics Committee of Pirkanmaa Wellbeing Services County does not require ethics approval for this study as it evaluates aggregated unidentified real-world data of glaucoma patients.
The use of EHR data in this study complies with the Finnish legislation, data security regulations of Tampere University Hospital and the Finnish recommendations on research ethics.In Finland, a separate law has been decreed on the Act on the Secondary Use of Health and Social Data (552/2019) (Ministry of Social Affairs and Health, 2019b).The purpose of the Act is to facilitate effective and safe processing and access to the personal social and health data for steering, supervision, research, statistics and development in the health and social sector.A second objective is to guarantee an individual's legitimate expectations as well as their rights and freedoms when processing personal data.According to the Act on Secondary Use of Health and Social Data, patients' personal data can  The EHR data have been transferred and are stored in a secured environment based on Azure cloud technology in Tays Research Workspace.The research group has been provided a virtual server in its own sandbox with no direct internet connections.The authentication of each member of the research group is verified with unique username and password.Remote access to the sandbox is done using secure, encrypted virtual private network (VPN) connection.The service provider loads both source data and needed software programs into the sandbox, allowing also transferred learning for the artificial intelligence (AI) algorithms.All data and AI programs used in the Workspace will be audited/inspected before entered or taken out.

| R E SU LT S
Table 1 presents the patient data and distributions in different diagnostic groups as recorded in EHR, history of procedures prior to 2019, other eye diseases, medications and the interval for the patients' next visit to Tays Eye Centre.The mean were higher for the patients with primary open-angle, normal-tension and exfoliative glaucoma (74-80 years) compared to glaucoma suspects (65 years) (Table 1).The patients with a glaucoma secondary to another eye disease were the youngest (mean ages 57-61 years).Female gender dominated in all groups (59%-68%) except for patients with secondary glaucoma due to another eye disease (40%-47%).44% of the follow-up glaucoma test sets were appointed within 1 year and another 44% within 2 years.The data entry for the next control visit was missing in 11% of all eyes, most commonly (37%) in suspected glaucoma (Table 1).
RNFL was recorded normal in 47% of all eyes and glaucomatous in 33%.ONH images were recorded to be glaucomatous more often (44%) than VFs (30%) and RNFL images (33%) (Table 2).No cases were excluded from the data of patients visiting the Medical Glaucoma Clinic.Other than a glaucomatous abnormality was recorded in 8% of VFs and in 2% of both ONH and RNFL images.Poor quality or no VF was recorded in 12% of all eyes, and poor picture quality or no picture was recorded in 3% for ONH and 19% for RNFL, respectively (Table 2).
The mean number of medications in the '≥2/3 glaucoma group' was 2.2, with the highest mean of 2.5 medications in eyes with secondary glaucoma due to another eye disease.In the '≥2/3 glaucoma group', 31% of eyes had received laser treatment and 10% glaucoma surgery (Table 3).The evaluation of progressive changes in any of the tests (ONH and/or RNFL and/or visual tests) over time was recorded in 82% of all eyes.The highest progression rate was reported in the '≥2/3 glaucoma group' (35%) compared to 2%-9% in the other groups (Table 2).The mean IOP was 14-15 mmHg (median 13-15 mmHg) in all groups except for suspected glaucoma with the mean and median of 17 mmHg.Target and maximum tolerable IOPs were recorded for 94%-95% of eyes (Table 2).
In glaucoma suspects, the IOP distributions of the untreated IOP were ≥27 mmHg in 21% of eyes, between 22 and 26 mmHg in 31% of eyes and <22 mmHg in 43% of eyes.In these eyes with glaucoma suspicion, 86% were evaluated as ≤1/3 abnormal findings in ONH, RNFL and VFs and 2% were recorded to have progressed compared to previous visits.Similar to eyes with a glaucoma diagnosis, ONH was evaluated to be glaucomatous more often (10%) compared to VF tests (2%) or RNFL (3%) in glaucoma suspects.

| DI SC US SION
The European Expert Panel on Effective Ways of Investing in Health has published recommendations for digital transformation of health services (ExPH, 2018): 1. develop a strategy for the digital transformation and evidence-informed policy measures to support decisionmaking, 2. develop and invest in coherent framework for monitoring and evaluation methodology, 3. create an environment that can adopt innovations and 4. be progressive with caution.As a follow-up for the published decision-making and monitoring strategies in Tays Eye Centre (Tuulonen et al., 2009(Tuulonen et al., , 2016)), the current article describes the first step in the development of a framework for continuous monitoring and evaluation of what gets done in everyday glaucoma care with respect to the defined strategies, which was further developed into the aces-rwm™ ecosystem by all five Finnish University Eye Clinics (Tuulonen et al., 2022).
When using EHR systems, the rate of recorded complete examination findings has been reported to be significantly higher than in paper records (Sanders et al., 2013).Systematic collection and comparison of EHR-RWD between different providers and countries also create a potentially efficient platform for future innovations, for instance, using AI for predicting the  2b) and for genetic association studies (Restrepo et al., 2015).So far, there has been limited evidence of incorporating EHR-RWD in AI algorithms and no prospective studies, for example, to demonstrate how well AI algorithms could predict the development of glaucoma (Ting et al., 2019).Most studies have collected EHR data from surgical cases of glaucoma only (e.g., Sun et al., 2022;Wang et al., 2022).The updated version of the structured glaucoma EHR (Figure 3) will be implemented into routine use in 2023 and will also enable prospective RWD study designs.
In England, EHR data of 45 309 patients from five different regions showed that less than one-fifth of OHT patients converted to glaucoma over a 5-year period, suggesting that many patients may require less intensive follow-up (Kelly et al., 2020).In 2023, Fu et al. published another UK study with at least 1-year EHR data in 2013-18 in five clinics in the National Health Service, including 43 742 patients.They also suggested that glaucoma care within the NHS Hospital Eye Service seems to be disproportionately directed towards patients with mild, low-risk glaucoma, who may be more appropriately managed using alternative models of care.Fu et al. (2023) proposed measures to reduce demand in hospital glaucoma clinics such as virtual clinics to remove the need for face-to-face clinician consultations, longer intervals between follow-up visits when appropriate and discharge of low-risk patients and ocular hypertensive patients, that is, all measures already implemented in Tays Eye Centre.Interestingly, the UK 2016 EHR data (Fu et al., 2023), including cost level, were very similar to our 2019 data (Table 4), except for the proportion of suspicions being smaller in Tays.In another UK study (Kelly et al., 2020), the interval between VF tests was 10-11 months for glaucoma suspects, OHT and glaucoma patients, that is, shorter compared to the test intervals of 1 year (44%) and 2 years (44%) in Tays Eye Centre (Table 1).
In spite of the differences, the yearly reports between Tays and the UK seem however surprisingly similar to each other in many aspects (Table 4), as well as between different years during the UK study (Fu et al., 2023).This raises an incentive to search for ways to further improve the long-term reporting and analysis protocols on how to compare and improve cost-effectiveness on everyday clinical outcomes.Although the resource allocation (e.g., longer test intervals, smaller proportion of suspicions and implementation of virtual clinics) in Tays Eye Centre seems different compared to the UK data set, obviously our data as presented in this article cannot be used to assess whether the glaucoma care protocol in accordance with the Finnish Guidelines would lead to same or different long-term outcomes in respect of preventing glaucoma-induced visual disability.
Similar to efficacy trials, it is also extremely important to create minimum standard framework for big data sets to be used by researchers and clinicians to improve the design of RWD studies and enhance transparency of study methods (Kotecha et al., 2022).The CODE-EHR bestpractice framework prefers a pre-published protocol for each of its 5 items in the checklist (Kotecha et al., 2022).Publishing our first version of the EHR reporting protocol according to the CODE-EHR checklist, to the best of our knowledge for the first time in eye care, offers an opportunity for current and future collaborators to evaluate and compare RWD sets and to suggest how the reporting and benchmarking protocols could be further improved.

F I G U R E 3
The clinical data in 2012-22 of the patient presented in Figures 2a and 2b were entered into the structured glaucoma-specific EHR, which has been developed using the prototype described in this study (Tuulonen et al., 2022) (Avenue Flow, Optomed Plc, Finland).On the IOP graph on the left, the number of lines refers to the number of medications, which can be read by clicking the visit.The triangle refers to events (e.g., laser trabeculoplasty, disc haemorrhage or progression).On the right, the clinical data are presented in the last visit in 2022.In each visit, previous entries are available enabling to change only required parameters.The swim lane on the top presents the previous visits and the planned next visit with the yellow circle indicating a procedure.The patient responded to health-related quality of life 15D (Sintonen, 2001) questionnaire in 2020.Both total and vision scores were normal (1.0).
The flow diagram positions our real-world data set, collected using the electronic health record (RWD-EHR), of 4618 patients visiting Medical Glaucoma Clinic in Tays Eye Centre in 2019, and presents its representativeness among different background data sets in Pirkanmaa Wellbeing Services County, including missing data and their proportion for each variable and specifications for linked data sets.bemanaged informed consent in officially audited and secured environments under the permission granted by the organization responsible for patient care, in our case by Tays, Tampere University Hospital.The audit of Tays secured environment was carried out in April 2022 by auditing companies approved by the National Supervisory Authority of Welfare and Health.The secured environment was added to the Toini database (the National Supervisory Authority of Welfare and Health, 2023).

F
I G U R E 2 2a (right eye, above) and 2b (left eye, below).A 27-year follow-up of ONH and RNFL pictures and the latest visual field taken in 2022.The patient was examined in Tays Eye Centre for the first time in 2012 and brought a picture taken in 1995.Treatment had been started in 2006.The white arrows in the ONH and RNFL pictures indicate the areas of progressive changes, the red arrows optic disc hemorrhages and the yellow arrows epiretinal membrane.Glaucoma Hemifield Test of the visual field indicated abnormality in 2022 in the right eye and since 2015 in the left eye.The glaucoma risk regression model G-RISK of the ONH (Hemelings et al., 2023) yielded a positive trend of about 0.01 risk increase per year which matches with the clinically detected progression.

T A B L E 1
Distributions according to the diagnosis as recorded in the electronic health record (patient-specific data for right eyes).

Distributions in different diagnoses according to '2 out of 3' definition All eyes ≥2/3 glaucoma with glaucoma diagnosis ≤1/3 glaucoma or OHT >30 mmHg with glaucoma diagnosis ≤1/3 glaucoma with suspected glaucoma Mixed/ undefined diagnosis
T A B L E 2 Findings in groups classified according to the Finnish Glaucoma Guideline's '2 out of 3' definition.

Distributions in different diagnoses according to '2 out of 3' definition All eyes ≥2/3 glaucoma with glaucoma diagnosis ≤1/3 glaucoma or OHT >30 mmHg with glaucoma diagnosis ≤1/3 glaucoma with suspected glaucoma Mixed/ undefined diagnosis
T A B L E 2 (Continued)T A B L E 3 Distributions of diagnostic according to the Finnish Glaucoma Guideline's '2 out of 3' rule.

Distributions in different diagnoses according to '2 out of 3' definition All eyes ≥2/3 glaucoma with glaucoma diagnosis ≤1/3 glaucoma or OHT >30 mmHg with glaucoma diagnosis ≤1/3 glaucoma with suspected glaucoma Mixed/undefined diagnosis
Comparison between Tays and UK data, which are partly differently reported.
T A B L E 4 a 1.4 visits per patient when also 2500 visits related to surgical consultations and procedures are included for which no clinical data were available in 2019.progression of et al., 2019, Figures 2a and