The impact of the COVID-19 pandemic on microbial keratitis presentation patterns

Background Microbial keratitis (MK) is the most common non-surgical ophthalmic emergency, and can rapidly progress, causing irreversible sight-loss. This study explored whether the COVID-19 (C19) national lockdown impacted upon the clinical presentation and outcomes of MK at a UK tertiary-care centre. Methods Medical records were retrospectively reviewed for all patients with presumed MK requiring corneal scrapes, presenting between 23rd March and 30th June in 2020 (Y2020), and the equivalent time windows in 2017, 2018 and 2019 (pre-C19). Results In total, 181 and 49 patients presented during the pre-C19 and Y2020 periods, respectively. In Y2020, concurrent ocular trauma (16.3% vs. 5.5%, p = 0.030) and immunosuppression use (12.2% vs 1.7%, p = 0.004) were more prevalent. Despite proportionately fewer ward admissions during the pandemic (8.2% vs 32.6%, p<0.001), no differences were observed in baseline demographics; presenting visual acuity (VA; median 0.6 vs 0.6 LogMAR, p = 0.785); ulcer area (4.0 vs 3.0mm2, p = 0.520); or final VA (0.30 vs 0.30 LogMAR, p = 0.990). Whilst the overall rates of culture positivity were similar in Y2020 and pre-C19 (49.0% vs. 54.7%, p = 0.520), there were differences in the cultures isolated, with a lower rate of poly-microbial cultures in Y2020 (8.3% vs. 31.3%, p = 0.022). Conclusions Patient characteristics, MK severity and final visual outcomes did not appear to be affected in the first UK lockdown, despite fewer patients being admitted for care. Concurrent trauma and systemic immunosuppression use were greater than in previous years. The difference in spectra of isolated organisms may relate to behavioural changes, such as increased hand hygiene.


Conclusions
Patient characteristics, MK severity and final visual outcomes did not appear to be affected in the first UK lockdown, despite fewer patients being admitted for care. Concurrent trauma and systemic immunosuppression use were greater than in previous years. The difference in spectra of isolated organisms may relate to behavioural changes, such as increased hand hygiene.

Introduction Methods
This retrospective study of medical records was conducted in accordance with the Declaration of Helsinki, and in accordance with local institutional policy. All data were anonymised prior to analysis, and the need for consent was waived. Approval was obtained from Sandwell and West Birmingham Hospitals NHS Trust Department of Clinical Effectiveness (registration #1512) to undertake this project as a service evaluation.

Study design & population
All cases of MK requiring corneal scrapes, presenting between the 23 rd March and 30 th June 2020 (Y2020) were identified through the regional microbiology service (Black Country Pathology Services Supporting Sandwell and West Birmingham Hospitals NHS Trust) database, and cross-checked with BMEC ED electronic medical records. Patients presenting during the equivalent time windows in the preceding three years (2017, 2018, and 2019) were also identified, and included as the comparator cohort (pre-C19), to reflect the variation of the disease. During the period being studied in 2020, all first-time face to face appointments were replaced with an initial telephone consultation, followed by a face-to-face consultation, where indicated. This had the effect of reducing the overall number of face-to-face appointments [18].

Routine clinical practice
Clinical assessments and decisions, such as the need for investigations, admission and follow ups, were undertaken by the BMEC ED attending ophthalmologists, in accordance with local guidelines. During the first national lockdown period, the BMEC in-patient ward was closed to allow nursing staff to be redeployed to specified medical wards to undertake general nursing duties, as well as to provide specialist care for admitted ophthalmic patients. Patients requiring admission for urgent care were initially admitted to amber wards (COVID status unknown), before relocating to specified Green (COVID-Free) and Red (COVID positive) wards with designated ophthalmic beds. The decision to admit patients was based on factors including clinical severity, risk of adverse events (e.g., perforation), social care needs (e.g., the ability to diligently administer all drops, proximity to clinic, and ability to attend for daily visits), with the final decision being taken by the lead clinician for any given session.
Where MK was suspected on presentation, corneal scrapes were taken to confirm the diagnosis. The typical corneal sampling kit consisted of a sterile needle (e.g. 23G) or scalpel blade for corneal tissue acquisition, one each of chocolate and blood agar plates, and a Sabouraud's agar plate for fungus. Nutrient depleted agar seeded with Escherichia coli was used for Acanthamoeba cultures, where indicated. Samples were placed on glass slides for microscopy and Gram staining. Dry swabs were also acquired for microbe polymerase chain reaction (PCR) typing. All cultures were incubated according to departmental protocols for at least one week.
A positive isolate was defined as a growth along the line of inoculation on solid media, and poly-microbial keratitis was confirmed if more than one clinically significant organism was isolated. Significant isolates were tested against antibiotics, in accordance with local protocols, using both disc diffusion (the British Society of Antimicrobial Chemotherapy methodology; www.BSAC.org.uk) and Vitek AST systems (www.biomerieux.co.uk). Isolates identified as contaminants in the microbiology reports were excluded from analysis.

Data collection
All data were recorded in an adaptation of a validated data collection proforma used in a previous study [14], using the secure web application Research Electronic Data Capture (REDCap© v9.6.3 2020 Vanderbilt University, Nashville, TN, USA). Data collected included patient demographics (sex, age, ethnicity and Index of multiple deprivation [IMD] score) and clinical details (presenting features, underlying risk factors, past ocular history, medications). The IMD score combines information from seven differentially weighted domains, to classify the relative deprivation of small areas around the UK; and scores were obtained from a government website [19].
Underlying risk factors were grouped as follows: contact lenses wear; active ocular surface disease (complete list in S2 Table); previous keratitis (infective and marginal); previous trauma (healed before the onset of MK) or previous ocular surgery; concurrent trauma (related to the onset of MK); foreign bodies associated with the current episode; as well as the systemic conditions: diabetes mellitus; rheumatoid arthritis; thyroid eye disease; and the use of systemic immunosuppression medication.
Details of clinical assessments were also recorded. The best corrected Snellen visual acuity (VA) at presentation was collected, and converted to LogMAR VA for analysis [20]. In addition, the final VA was also recorded, based on assessments performed at clinical follow up appointments (1, 3, 4 and 12 weeks after presentation). Where patients attended a clinical follow up at week 12, the VA at this appointment was used, with the latest available assessment used instead in patients that were discharged from the service prior to this.
Slit-lamp biomicroscopy was used to assess the size of the epithelial defect, infiltrate, or scar, using standardised methodology adapted from the Herpetic Eye Disease Study [21], by measuring the longest and the longest perpendicular dimensions. The area was then calculated by multiplying these readings together. Epithelial defect, infiltrate, and scar size were not differentiated, henceforth this measurement is referred to as the "ulcer area", which was also the summation of all single areas of involvement in the cornea.
The corneal involvement score (CIS) was retrospectively derived from the clinical notes, based upon the validated corneal opacification score described by Ong et al. [22]. Briefly, the locations of the corneal ulcer are documented according to the number of quadrants involved (temporal, superior, nasal, inferior), which are each assigned 1 point, with involvement of the central 4mm zone being assigned 5 points. The numbers of points are then added, to give a final CIS out of 9.

Statistical analysis
Comparisons of patient characteristics by presentation period (Y2020 or pre-C19) were performed, using Fisher's exact tests for nominal variables, and Mann-Whitney U tests for ordinal and continuous variables. Continuous variables were reported as mean ± standard deviation if approximately normally distributed, with median (interquartile range; IQR) used otherwise. Cases with missing data were excluded from the analyses of the affected variables, and the sample sizes included in each analysis are reported in the associated tables. All analyses were performed using IBM SPSS 26 (IBM Corp. Armonk, NY), with p<0.05 deemed to be indicative of statistical significance throughout.

Included cases
A total of 230 MK patients were identified, comprising 63, 50, 68 and 49 patients from the time windows in the years 2017, 2018, 2019 and 2020, respectively. Total numbers of attendances to the BMEC for any indication were 12,128 during the time window in 2018 and 12,239 in 2019, compared to only 5,759 in 2020 (accurate data were not available for 2017). As such, MK comprised 0.5% of attendances in 2018-19, which increased significantly to 0.9% in 2020 (p = 0.001). Comparisons between the years 2017-2019 found no significant differences in patient characteristics (S3 Table). As such, the 181 cases from these three years were combined into a single cohort for subsequent analysis (pre-C19), and compared to the 49 cases from the year 2020 (Y2020).

Patient characteristics
Comparisons between Y2020 and pre-C19 found no significant differences in the age, sex, laterality of eye, ethnicity or IMD scores between the groups ( Table 1). The duration of symptoms at presentation was also similar in the Y2020 and pre-C19 groups, with medians of 4 days (IQR 2-7) and 3 days (1-6), respectively (p = 0.201). Of the risk factors considered, concurrent ocular trauma (16.3% vs. 5.5%, p = 0.030) and systemic immunosuppression (12.2% vs. 1.7%, p = 0.004) were both significantly more prevalent in the Y2020 group. The full list of causes of concurrent trauma is reported in S4

Clinical assessments
The severity of disease at presentation was quantified using the VA, ulcer area and CIS, none of which were found to differ significantly between the Y2020 and pre-C19 groups (Fig 1A and  Table 2). These comparisons were also repeated after excluding the 67 patients with pre-existing visual impairment at presentation (based on their ophthalmic history), with the difference between groups remaining non-significant. Despite the similarities in patient characteristics, admission rates were found to be significantly lower in Y2020, at 8.2% compared to 32.6% for the Pre-C19 group (p<0.001, Table 3). However, the disease course was found to be similar in the two groups, with no statistically significant differences noted in the final VA (Fig 1A), or in complication or intervention rates (Table 3).

Microbiology
Rates of culture positivity were similar in the two groups, at 49.0% in Y2020 and 54.7% in Pre-C19 (p = 0.520, Table 4 and Fig 1B). However, the distribution of culture isolates was found to vary between the groups (Fig 1C), with a significantly lower rate of poly-microbial infections in Y2020, compared to pre-C19 (8.3% vs. 31.3%, p = 0.022), and a non-significant tendency for higher rates of gram-negative mono-microbial infections in Y2020 (33.3% vs. 18.2%, p = 0.160). Fungal infections comprised similar proportions of culture positive cases in both groups (4.2% vs. 5.1% in Y2020 vs. pre-C19, p = 1.000).
Assessment of the most frequent gram-positive isolates found a preponderance of Staphylococcus aureus infections in the Y2020 group, being isolated in 25.0% of those with positive cultures, compared to 11.1% in pre-C19 (p = 0.099). Of the gram-negative isolates, it was notable that no cases of P. aeruginosa were detected in Y2020, compared to 11.1% in previous years (p = 0.120). However, neither of these differences reached statistical significance, largely as a result of the small sample sizes in these subgroups.

Discussions
The pandemic's negative impact on ophthalmic services [10,11,13,23,24] has raised concerns about patients' well-being. This study evaluated the impact of the first COVID-19 lockdown on the outcomes of patients with MK. In a survey completed by the British public, our group identified how concerns about the pandemic would lead individuals to consider seeking healthcare for their eye symptoms less urgently than if there was no pandemic [17]. The present study demonstrates a strong similarity between patients with MK in the first UK lockdown and those from previous years, with respect to time-to-presentation, presenting VA and ulcer area, complications, interventions, and final VA. However, the prevalence of concurrent trauma and use of systemic immunosuppression were greater than in previous years, while fewer poly-microbial infections and ward admissions occurred. Thus, patients presenting to this centre during the lockdown appear to be accessing services on time, did not have worse MK, and perhaps had milder disease in a more vulnerable group of patients.
Disease epidemiology and health care services vary geographically. Whilst Agarwal et al. [24] reported an increase in MK incidence during the lockdown at their unit in India, Poyser at al. [10] report a decrease in contact lens associated keratitis of more than 50%, compared to the same period in 2019, although the proportions remained similar in both study periods. The present study's results identified an increase in the proportion of MK patients seen in the department compared to previous years, whilst fewer patients were seen in the department overall [18]. The average time-to-presentation and number of patients attending in Y2020 compared to pre-C19 indicate no change in the patterns of the public accessing services for  (Fig 1A) are summarised using boxplots, with outliers indicated with circles or asterisks for those outside the box by 1.5-or 3-times the interquartile range, respectively. Microbiology is summarised as the total proportion of the cohort that were culture positive (Fig 1B), and the distribution of culture isolates from these positive cases ( Fig 1C); unlabelled bars consist of <5% of cases. Further details of the definitions used for the microbial cultures are reported in Table 4. https://doi.org/10.1371/journal.pone.0256240.g001

PLOS ONE
Impact of the COVID-19 pandemic on microbial keratitis  MK. Although no specific restrictions were placed on ward admissions, the decrease in 2020 is likely influenced by clinicians' concerns about their patients being exposed to COVID-19 in hospital. It is interesting to note that this did not appear to have a significant impact on the measured outcomes in these patients. Before the COVID-19 pandemic, severe MK ulcers (e.g., ulcer >3mm in diameter) were admitted to the ward for intensive topical medication. However, the adjustment to self-administration of drops in 2020 appears safe and effective. The economic burden of managing MK as an in-patient is considerable [14]. While other factors (e.g. social care) may drive the need for hospital admission, judiciously increased outpatient management would help to reduce the risk of COVID-19 exposure and be significantly more cost-efficient. In this case, an estimated £150,000 of direct patient cost-savings were made in the 2020 study period [14]. Cultured isolates identified as contaminants by the microbiology department were excluded from analysis in this study; however, it can be challenging to discern contaminants from pathogenic isolates, due to the high prevalence of commensal bacteria known to cause MK [15,[25][26][27]. Corneal sample culture contamination may be influenced by face-mask wear. In their interesting study, Samarawickrama et al. [28] demonstrated the impact of study participants speaking out-loud for 30 seconds at 30cm from an open culture dish, with and without wearing a surgical mask. They found a significantly higher culture rate in the no-mask group. However, as acknowledged by the authors, their simulation likely over-estimates contamination rates compared to real-world practice, as culture plates are unlikely to have such prolonged direct exposure, and the scrape needle (or knife) surface area is considerably smaller. Although this may explain the decrease of some oral-cavity commensals such as Streptococcus, it is not supported by the results of the present study when considering the prevalence of others isolates like Staphylococcus epidermis and the increase in Staphylococcus aureus. Furthermore, if mask wear reduced contamination, it would be expected that the culture positive rate in Y2020 would have reduced, relative to the earlier period. This did not occur in the present study, with the pre-C19 and Y2020 rates being similar, and comparable to other UK studies [15,[25][26][27]. Poor hand hygiene is a known risk factor for developing MK [29]. The bacterial diversity of the hands is greater and more dynamic than other body areas, and is considerably influenced by factors such as sex, environment and hand washing [30]. Following handwashing, although the bacterial load is decreased, its diversity is retained [30]. Since bacteria have innately varying transmission potentials, hand washing may have a differential effect on the prevention of transmission of different species [31]. Thus, the microbiological findings of this study may be influenced by the increased handwashing and the impact of campaigns advising against excessive hand-face contact during the pandemic [32], which may have altered the autoinoculation of pathogenic microbes onto the ocular surface.
The urgency with which individuals seek medical attention may differ considerably between pathologies, and delays in presentation for some conditions may be more clinically significant than for others. Mild ocular surface disease symptoms are considered by the pubic to be of low seriousness, and to require medical attention less urgently during lockdown, compared to normal circumstances [17]. Ocular trauma, specifically occurring at home, has increased during the lockdown, with delays in medical review also being reported [23,33]. This is reflected in the higher prevalence of concurrent trauma in patients from the Y2020 group. Although uncomplicated mild trauma is relatively easily managed with lubricants and topical antibiotic prophylaxis, delays in initiating treatment may permit progression to infective keratitis. An explanation for the increased prevalence of systemic immunosuppression as a risk factor among the Y2020 cohort is less apparent.
The strength of this study is in its use of real-world data over four years in a large unit serving an out-of-hours population of up to 3.5million (5.25% of the UK population), where 72% of emergency room referrals are out of the local catchment area. This has helped to generate well-documented cohort of patients, thus reducing any variability introduced by dissimilar geography and clinical practice at different departments that may confound results. In this study, presenting VA, ulcer area, CIS, and final VA were utilised as proxy measures of severity. This represents an inherent limitation within the study; although severity scales for MK have been proposed [34,35], there is currently no widely accepted severity stratification system that adequately covers the entire spectrum of the disease. Complications and interventions are indicators of severity; however, these relatively rare events occurred too infrequently to offer insight here. Further assessments of disease state, including a detailed time-course of lesion morphology, assessment of final optical state (corneal scarring), and patient-reported outcomes, were desirable, but not possible here. Further limitations of this study include its external generalisability, since these observations are of one centre, hence further work from multiple centres across the UK is required to validate these findings.
This study compared the features of MK patients from the first UK lockdown to previous years. These results demonstrate the considerable similarity in the presenting severity and clinical outcomes of the two groups, despite fewer patients being admitted for care in 2020. This finding is significant, considering the persisting need to safely adapt clinical practice to manage the risk of COVID-19 transmission. While other literature supports the link between increased ocular trauma and lockdown-related lifestyle changes, an explanation for the microbiology findings is less readily identifiable. Increased handwashing practices, as well as changes in environmental factors, such as reduced contact lens wear, may have contributed to this. However, these findings must be validated on a larger scale; therefore, future work aims to connect the corneal clinician network in the UK to investigate this nationally.
Supporting information S1