Impact of Victorian COVID-19 restrictions on emergency department presentations and hospital admissions for ophthalmic conditions

ABSTRACT Clinical Relevance Assessing the extent to which COVID-19 impacted hospitals can provide important learnings for future pandemics. Background This study aims to determine the impact of the 7-month duration COVID-19 pandemic-related lockdown orders on ophthalmology-related hospital admissions and emergency department (ED) presentations, during 2020 in Victoria, Australia. Methods Analysis was performed on Victorian statewide data from the Victorian Emergency Minimum Dataset (VEMD) and Victorian Admitted Episodes Dataset (VAED), between 1 January 2018 and 31 October 2020. Numbers of presentations and admissions for key ophthalmic conditions were stratified by age, socioeconomic status, location (metropolitan versus rural), and triage category. From the observations occurring in the pre-pandemic period (January 2018 to March 2020), a linear regression prediction model was built for each diagnosis which predicted what the presentation number in the COVID-19 period would have been if the pandemic had not occurred. Results Based on pre-COVID-19 trends, the largest decreases in expected admissions were for glaucoma (32.9%) and retinal breaks and detachments (21.2%). For the ED data, the most apparent changes were: an increase in presentations for foreign bodies (22.6%); a decrease in retinal detachments (35.5%); and a decrease in keratitis (18.4%) relative to predictions. Conclusions Hospital admissions decreased and patterns of ED attendances changed during lockdown. The findings suggest the need for the following: increased safety messaging to avoid eye injuries around the home; improved pathways for safe and rapid triaging of eye conditions in the community to ensure effective use of ED resources; and messaging to ensure that people do not delay care when they notice signs of sight-threatening conditions such as retinal detachment.


Introduction
In response to the COVID-19 pandemic, in the Australian state of Victoria (population of 6.6 million), metropolitan areas were "locked down" from March to October, 2020.This lockdown period was substantially longer than elsewhere in Australia, with restrictions present in regional Victoria for part of this duration.During lockdown, people were allowed to leave home for five reasons only: food, medical care, exercise, approved essential work that could not be undertaken from home, and carer duties (see Supplementary Material A).During much of this period, in metropolitan areas, community optometry and ophthalmology services (both public and private) were restricted from providing aspects of care, with management limited to cases considered either "urgent" or "essential" depending on the lockdown phase.
Within the COVID-19 context, the definition of urgent clinical care by the Victorian Chief Allied Health Officer was: "Direct clinical care (and other clinical support activity necessary to support direct clinical care) that prevents a significant change/deterioration in functional independence necessitating escalation of care or is required as an adjunct to other urgent medical/surgical interventions".More complete descriptions of "urgent" and "essential" clinical care are provided in Supplementary Material B. Medicare data demonstrates a 31% reduction in optometry services in the state of Victoria between March and October 2020 relative to March and October, 2019. 1 The Medicare data is combined across metropolitan and regional areas, however more significant restrictions on practise were implemented within the Melbourne metropolitan area.Both optometry and ophthalmology services switched aspects of care to telehealth, leading to acute changes to service delivery modes.
Pandemics can alter behaviour of people seeking medical care.In the ophthalmological context, the COVID-19 pandemic has altered the number and patterns of eye emergencies presenting to emergency departments (ED) in many countries including: the United Kingdom, 2-5 India 6,7 and the USA, 8,9 amongst others.Within Australia, ED ophthalmic presentations to the Western Sydney Local Health District in New South Wales decreased by 16% in the 9 weeks from March 29 to 31 May 2020, relative to the same dates in 2019. 10With few exceptions, 8 previous reports of COVID-19 shifts in eyecare behaviour have analysed data from single sites or locally grouped hospital networks, so may reflect peculiarities of local healthcare pathways, regional demographics or other site-specific influences.Many of these reports provide data for periods of 4-10 weeks of stay-at-home restrictions. 2,4,9,11hanges in ED attendance patterns may reflect alterations in the ability to access community eye care, changes in workplace behaviour or other activities that either reduce or increase the risk of eye injuries and infections, or alternately, represent delays in care access.For example, key findings from the emergency eye department at the University Hospitals of Leicester, UK, included reduced eye infections and trauma-related injuries during lockdown, but more severe presentations of macula-off retinal detachments. 2he authors suggest that improved hygiene practices associated with COVID-19 infection control guidelines may explain reductions in infective conjunctivitis, and that increased sight threating retinal detachment presentations could relate either to attendance hesitancy, and/or a lack of access to optometric services that can identify retinal tears in the community.Decreases in presentations for atraumatic retinal detachments, conjunctivitis, and ocular pain were also reported during the first COVID wave in Sydney. 10ther reports have confirmed decreases in ophthalmic trauma presentations during lockdown orders in a variety of countries. 6,8,11mergency departments become first points of care when community options are limited.In Victoria, there was a particularly lengthy period of lockdown, not only relative to other jurisdictions in Australia, but within the international context (see Supplementary Material B).The impact of the COVID-19 pandemic and lengthy movement restrictions on eyecare in Victoria has not been reported.This study presents Victorian statewide data on presentations for key emergency eye-care conditions during the 7-month COVID-19 community public health movement restrictions.The aim was to quantify the changes in the number of presentations and admissions during the lockdown period for critical, sightthreatening, ophthalmic conditions that presented to Victorian emergency departments, and to determine whether changes differed by eye condition, socioeconomic status, or geographical location.Understanding the impact of pandemic restrictions on sight-threatening conditions is critical to plan for future pandemic responses.This will better equip ophthalmologists, optometrists, healthcare managers, advocacy organisations and policymakers, to plan for eye-care needs within the current and future pandemics.

Data source
Hospital separation data (information recorded at the time of discharge) were obtained from the Victorian Admitted Episodes Dataset (VAED) and ED presentations data from the Victorian Emergency Minimum Dataset (VEMD) for the period 1 January 2018, to 31 October 2020.The VAED data comprises de-identified demographic, clinical and administrative details for all admitted episodes in Victorian hospitals, both private and public. 12The VEMD data comprises deidentified demographic, administrative and clinical data for presentations at Victorian public hospitals with designated emergency departments. 13The reason for admission or presentation was selected based on the principal diagnosis recorded using the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification (ICD-10-AM) codes. 14he study obtained organisational approval from The Royal Children's Hospital for an exemption from ethical review (Project ID: 70069).

Study population
The linked data (VAED and VEMD) was used at an episode level.All episodes with admissions or presentations between 1 January 2018, and 31 October 2020 were included.

Conditions of interest
The six most-commonly represented in the data ophthalmology conditions were investigated for both admission and emergency presentations.For VAED (admissions data), these six conditions were: glaucoma, retinal tears and retinal detachments, ophthalmic complications associated with type 2 diabetes mellitus, injury of eye/orbit, optic neuritis, and vitreous haemorrhage.For VEMD (emergency department presentations), the six conditions were: ocular foreign body, injury of eye/orbit, keratitis, retinal detachments and breaks, iridocyclitis, and glaucoma.Supplementary Table A provides a list of relevant ICD-10 AM codes.

Outcome
Number of admissions (VAED) or presentations (VEMD) for each condition of interest.

ED presentation urgency
ED presentation urgency is a binary variable derived from ED triage category which indicates whether the presentation was an urgent or non-urgent episode.ED triage category in VEMD is a six-level triage algorithm that provides clinically relevant stratification of patients into six groups, according to urgency of need for medical and nursing care, 1 (resuscitation), 2 (emergency), 3 (urgent), 4 (semi-urgent), 5 (non-urgent), and 6 (dead on arrival).ED presentations that had a triage l or 2 classifications in VEMD were defined as urgent episodes, while the rest were classified as non-urgent.

Remoteness
Remoteness is a derived binary variable indicating whether the episode occurred in regional Victoria or metropolitan Melbourne.Remoteness was defined based on The Australian Bureau of Statistics (ABS) Remoteness Areas Structure within the Australian statistical geography standard (ASGS). 15ASGS divides Australia into five classes of remoteness based on a measure of relative access to services: major cities, inner regional, outer regional, remote, and very remote.Inner regional, outer regional, remote, and very remote areas were aggregated and classified as regional Victoria, which covers all areas of Victoria except metropolitan Melbourne.

Socioeconomic status
Socioeconomic status (SES) is a derived binary variable indicating whether the episode occurred in a lower or higher SES area.SES was derived using the ABS developed Socio-Economic Indexes for Areas (SEIFA).SEIFA is a product developed by ABS that ranks areas in Australia using postcode according to relative socio-economic advantage and disadvantage.The Index of Relative Socio-economic Disadvantage (IRSD) is one of the four indexes in SEIFA that summarises a range of information about the economic and social conditions of people and households within an area by area-based deciles. 16Area-based deciles are calculated by dividing the areas, ordered by disadvantage, into 10 equally sized groups with decile 1 containing the most disadvantaged areas.Lower SES was defined as areas with a decile number of 1-5, and higher SES included areas with deciles of 6 to 10.

Length of stay
Length-of-stay data had a large right skew.It was assumed that the distribution of length-of-stay is log normal, thus log (length-of-stay) is normally distributed.

Statistical analysis
To examine the change in health service use before and after the COVID-19 restrictions, the number of admissions (VAED) or presentations (VEMD) were converted into a monthly time series format and the entire observational period was divided into a pre-COVID-19 period and a COVID-19 period.The COVID-19 period was defined from the start of the first stage 3 lockdown on 31 March 2020, to the end of the second lockdown on 28 October 2020. 17Observations that occurred before the start of COVID-19 period were defined as the pre-COVID-19 period.Thus, the study period comprised 27 "pre-COVID-19" and 7 "COVID-19" months of observations.
A linear regression prediction model was used to predict what the monthly number of presentations in the COVID-19 period would be if the COVID-19 pandemic and associated restrictions had not happened and the pre-COVID trend had continued.The linear regression model, was built using the 27 data points occurring in the pre-COVID-19 period, with the number of presentations as the outcome variable (y) and the time passed from the start of the observational period as the predictor (t).To quantify the uncertainty in the randomness associated with the point being predicted as well as in the coefficient estimates, 95% prediction interval (PI) of the predicted admissions (VAED) or ED presentation (VEMD) number was also reported.The association between COVID-19 restrictions and health services usage change was assessed by checking if the observed number of presentations fell in the 95% PI of the predicted values.
To investigate the validity of using a linear model, the residual error series were examined after fitting the linear regression.The error series did not display a remarkable amount of remaining trend, indicating that the behaviour of the trend was accounted for by the linear model. 18In addition, the distribution of the outcome (number of presentations and admissions) followed a normal distribution approximately.Taken together, these features provided confidence that the linear prediction model was appropriate to approximate the data trend.
Difference between the observed and the predicted number was computed using the observed minus the predicted.Relative difference was calculated by subtracting the predicted from the observed then divided by the predicted number.Then for each condition, a mean difference was computed by averaging the sum of the monthly difference, and a mean relative difference was computed by averaging the sum of the relative difference of each month in the COVID-19 period.
To investigate any characteristic change related to COVID-19 restrictions, subgroup analysis by age, remoteness, SES, and sex were performed for both VAED and VEMD.Subgroup analysis by presentation urgency and admitted proportion were investigated for VEMD, while length-of-stay was examined for VAED.All analyses were performed in R version 4.0.3.

Victorian admitted episodes dataset (VAED)
There was a sharp drop in the number of admissions in the first month of the COVID lockdown period followed by a rebound.The largest decrease was observed for glaucoma cases, followed by retinal breaks and detachments (Figure 1).Table 1 shows the admissions data for each condition during the pre-COVID and COVID periods, along with the demographic data.In general, there was a similar distribution of admissions in terms of SES status, geographical location, gender and age, during the COVID and pre-COVID periods.There was no notable change in length-of-stay (Table 1).
Table 2 compares the observed number of admissions to those predicted if the COVID-19 pandemic and associated restrictions had not occurred.Months and conditions where the number of admissions fell outside the prediction interval are indicated in the table.Based on pre-COVID-19 trends, the authors would have expected, on average, 32.9% more admissions for glaucoma, 21.2% more admissions for retinal breaks and detachments, 11.1% more admissions for ophthalmic complications from type 2 diabetes mellitus, 10.3% more admissions for injury of eye/orbit, and 12.8% more admissions for vitreous haemorrhage, each month from April to October 2020.For optic neuritis, there was no difference between the observed and predicted admission number on average.This data is graphically represented in Figure 2, which plots the observed data alongside the linear prediction model.

Victorian emergency minimum dataset (VEMD)
Figure 3 shows the number of presentations to ED for the selected ophthalmology conditions for the period from 1 January 2018 to end October 2020.Table 3 presents the data according to age (above or below 18 years), regional versus metropolitan areas, and according to SES grouping.
The proportion of emergency presentations for adults versus children was relatively stable (Table 3), as were the other demographic features (SES area, regional versus metropolitan Melbourne).Most presentations were deemed as non-urgent (Table 3) both during and prior to the pandemic period.Men were more likely to present to ED than women, with the increases in ophthalmic foreign body presentations being driven by men (Table 3).
Table 4 compares the observed number of admissions to those predicted if the COVID-19 pandemic and associated restrictions had not occurred.Based on the pre-COVID-19 trends, the authors would have expected, on average, 22.6% fewer presentations for foreign bodies, 3.9% more presentations for injury of eye and orbit, 18.4% more presentations for keratitis, 35.5% more presentations for retinal detachments and breaks, and 2.9% more presentations for iridocyclitis, each month from Apr to Oct 2020.For glaucoma, there was no difference between the observed and predicted presentation number on average.This data is depicted graphically in Figure 4.
Supplementary Table B presents the locations and activities associated with ophthalmic foreign body presentations and ocular injuries from the VEMD dataset.In 2020, there were fewer presentations related to workplaces, places of education and sports than in prior years.

Discussion
This study quantifies the changes in the number of presentations and admissions during the key lockdown period from March to October 2020, in Victoria, Australia.There was a sharp drop in hospital admissions, in particular for glaucoma and retinal detachments.In terms of emergency presentations, marked increases in ocular foreign body presentations were evident, with reductions in presentations  for keratitis and retinal detachments.Many of these observations are similar to those reported in other jurisdictions internationally, providing a strong consolidated evidence base for pandemic impacts on eyecare that are not unique to one particular healthcare environment.These are discussed below.
As shown in Table 4, the results of the linear regression prediction model show that ocular foreign body presentations were 22.6% higher than expected based on pre-pandemic trends.The trade/construction industry was able to continue to work, although with reduction in on site personnel, during much of Victorian lockdown, hence workplace presentations from the building and construction sector were likely to remain constant.With lockdown orders and curfews, there was an increase in do-it-yourself home and garden improvement projects.One study from the USA noted a 29% increase in injuries arising from home improvement projects during stay-at-home period, 9 and data from South  Australia demonstrated a 20% increase in trauma in the home, largely driven by an increase in falls from heights, in keeping with increased home improvement activities.
With work and home-related activities the top identified reasons for ocular foreign body (Supplementary Table B), and workplace mandating eye protective goggles, there is a clear need for increased education in the community regarding the need for eye-related personal protective equipment particularly at home.The increased presentations to ED could also be related to reduced access to community eye-care services, as suggested by the decrease in presentations to ED during the less severe stage of the lockdown period (May, June 2020) when community optometry services were temporarily re-opened.
There were reduced emergency department presentations for keratitis.Reduced infective eye conditions has also been reported in other studies from the United Kingdom, 2 along with reports of differences in the nature of organisms cultured from microbial keratitis presentations during the pandemic. 20It has been suggested that reductions in the number of infective presentations may result from changes to hygiene and general infection control in the community, such as increased hand-washing. 2,20Contact lens wear is a significant risk factor for keratitis.Reduced contact lens wear during stay-at-home orders may also contribute to decreased keratitis presentations.According to surveys of contact lens wearers, 67% (Spain) and 56% (UK) of people reported spending less time wearing their contact lenses 21,22 during stay-at-home orders.
There was a marked decrease in the number of presentations for retinal detachments, by 35.5% fewer presentations than expected (Table 4).This finding is in line with other reports.In United Kingdom, a 62% reduction in presentations for retinal detachments was reported during a much shorter, 1 month, duration of stay-at-home orders by the Moorfields Eye Hospital. 2A reduction of approximately 50% in presentations for retinal detachments has been noted in the Scottish Retinal Detachment study. 23It is unclear whether these reductions arose from reduced underlying presentations, or whether people are neglecting symptoms of impending vision loss.Alternately, people may be unclear of pathways for referral during periods of limited access to community eyecare.It is possible that people may have misunderstandings regarding permissible activities under lockdown orders, or are fearful of hospital attendance leading to coronavirus infection.Interestingly, the present data showed an increase in retinal detachment admission during the period between hard lockdowns (Figure 1, May-June 2020), suggesting the above assumption may be true.
During most of the studied period in Victoria, community sport was halted, and night curfews instigated, possibly resulting in reduced avoidable injuries, which may also contribute to the observation of fewer traumatic retinal  detachments.Internationally, decreases in ocular trauma during the pandemic period have also been noted.A 68.4% decrease in eye trauma cases were reported over a single month of lockdown at one Italian ophthalmological emergency department, 11 and a population review of eye injuries in the USA demonstrated a reduction in trauma cases during March to April, 2020 (coinciding with stay-at-home orders) relative to the same time period in previous years. 8The present data showed a slight decrease in eye injuries, and a marked decrease in retinal detachments over a much longer 7-month COVID lockdown period, compared to the previous 27 months of pre-COVID data.A change in the nature and severity of retinal detachment presentations has also been noted internationally, with more macula-off (worse outcomes) retinal detachments suggesting delayed presentations. 3,23The reduction in presentations for retinal detachment identified in many international studies, and in the present study, does suggest delayed access to eyecare, which generally leads to poorer outcomes and blindness. 5,10,23istorically, telehealth for eyecare has been patchy in Australia, largely limited to remote and rural services.The current pandemic situation provides an acute incentive to significantly improve and innovate models of eyecare.In response to the pandemic, internationally, numerous urgent eye-care services have been reconstructed to enable triaging of conditions via videocall.The aim of these programs has been to reduce attendances at emergency eyecare departments for stable or non-urgent cases, with trials being deemed as highly successful. 24,25Patient acceptance of telehealth for the triaging and management of eye conditions appears high and may reduce delays in seeking care. 26reative use of telehealth has significant potential to enable the efficient use of healthcare resources.Community optometry access should be kept open and a telehealth consult with an optometrist used for triage prior to in-person slit lamp examination with an optometrist, then subsequent referral and in-person assessment by an ophthalmologist where indicated.
The main strength of the present study was the availability of statewide data across multiple hospital systems including all major trauma centres, the Royal Victorian Eye & Ear Hospital -which is a dedicated eye emergency hospital for Victoria, as well as rural hospitals.Comparative data was also available from January 2018, which meant inclusion of 27 months of pre-COVID data to ensure good prediction estimates.Additionally, different to other Australian states and territories, and internationally, Victoria had a particularly lengthy and strict 7-month lockdown period from 31 March 2020, through to 28 October 2020, in an attempt to reach COVID-zero cases.This is in stark comparison to many of the early international studies that report on data only from one month of acute stay-at-home orders at the beginning the pandemic.
The availability of 7 months of COVID lockdown data can better guide public health advice, behaviours and messaging, particularly as ongoing limited staged restrictions may continue for some time into the future.Limitations of the present study include potential issues with the accuracy of coding, and absence of nuanced coding, using the ICD-10-AM codes.The present data includes unique episodes but does not provide detailed clinical data, or qualitative data/experience to support the interpretation.
It is also important to recognise that the data analysed herein represents all hospital presentations; however, these may not represent the true incidence in the population of these conditions.A further limitation is that the COVID-19 period was conservatively defined here as commencing from March 31 st , 2020; however, health behaviours were likely changing prior to that date as concern regarding the impending pandemic increased throughout March 2020.
There are several recommendations that arise from this study, both for acute pandemic situations but also with implications for eyecare into the future beyond COVID-19.Consideration should be given to: (a) increased safety messaging regarding the need for personal protective equipment to avoid eye injuries around the home; (b) strong public health messaging to ensure that people do not delay health care when they notice changes to vision including flashes, floaters and other signs of possible retinal detachment; (c) communication to the public regarding how they can safely have their eyes examined during the pandemic to minimise delays in due to fear of virus contagion or poor understanding of which health services are permitted during stayat-home orders; and (d) improved pathways for safe and rapid triaging of eye conditions in the community, including innovative use of telehealth pathways.

Figure 1 .
Figure 1.Number of admissions to Victorian hospitals according to the VAED dataset for the period from 1 st January 2018, to the 31st of October 2020.The grey section to the right hand sides denotes the periods of COVID-19 movement restrictions in Victoria, with the dark grey sections indicating the periods of severe "lockdown" in metropolitan areas.

Figure 2 .
Figure 2.Observed versus predicted admissions across the time period from 1 st January 2018 to the 31 st of October 2020, according to the VAED dataset.The grey section to the right hand side denotes the periods of COVID-19 movement restrictions in Victoria, with the dark grey sections indicating the periods of severe "lockdown" in metropolitan areas.The green shaded area shows the 95% prediction interval for admissions according to the linear prediction model (i.e. the 95% prediction interval for admissions if the COVID lockdown had not occurred).

Figure 3 .
Figure 3. Number of presentations to Victorian emergency departments according to the VEMD dataset for the period from 1 January 2018 to the 31st of October 2020.The grey section to the right hand sides denotes the periods of COVID-19 movement restrictions in Victoria, with the dark grey sections indicating the periods of severe "lockdown" in metropolitan areas.

Figure 4 .
Figure 4. Observed versus predicted admissions across the time period from 1 January 2018 to 31 October 2020, according to the VAED dataset.The grey section to the right hand side denotes the periods of COVID-19 movement restrictions in Victoria, with the dark grey sections indicating the periods of severe "lockdown" in metropolitan areas.The green shaded area shows the 95% prediction interval for admissions according to the linear prediction model (i.e. the 95% prediction interval for admissions if the COVID lockdown had not occurred).

Table 2 .
Observed versus predicted number of admissions (with those falling outside the 95% PI denoted by an asterisk).

Table 3 .
Emergency department presentations data pre-COVID versus COVID lockdown period in Victoria.

Table 4 .
Observed versus predicted number of ED presentations (with those falling outside the 95% PI denoted by an asterisk).