Introduction

The COVID-19 pandemic has placed significant strain on the NHS in the UK, particularly during surges in SARS-CoV-2 community transmission. In response, novel methods of delivering care have been required to avoid overwhelming hospital admissions. COVID-19 virtual wards are utilised to remotely manage COVID-19 patients in the community and were officially recommended by NHS England since November 2020 [1]. These services are designed to ensure safe and early discharge of COVID-19 patients from hospital.

The COVID-19 virtual ward service was set up at University College London Hospital (UCLH) following a promising preliminary study in April 2020 [2]. An evaluation of this service during September-October 2020 showed a reduction in the number of patients who had to unexpectedly reattend the UCLH Emergency Department (ED) from 22.6% to 4.7%, when comparing to rates prior to the service’s initiation [3]. This showed that virtual wards could help in managing those at risk of COVID-19 deterioration.

Since strain on healthcare services is higher when case numbers rise, we decided to evaluate the COVID-19 virtual ward service at UCLH during the Winter months of late 2020/early 2021. Assessing the service during this second wave of COVID-19 in the UK (i.e., the Alpha variant-of-concern) we were able to see if it still reduced the rate of unplanned ED reattendance.

In settings such as this, with high case numbers and finite healthcare resources, risk stratification can help identify those mostly likely to deteriorate. In other COVID-19 virtual wards, prognostic scores (such as the RECAP-V1 model) have been used to identify the highest risk patients to give them more regular follow up and/or hospital reassessment [4].

The ISARIC-4C mortality score is a prognostic score which is widely used in NHS hospitals across the UK [5]. Produced from measurements collected in most COVID-19 patients attending hospital (sex, number of comorbidities from a predefined list, Glasgow Coma Scale {GCS}, age, respiratory rate, admission oxygen saturation, urea, CRP, lymphocytes- see online tool for calculator [6]), it gives a prediction of inpatient risk of death from COVID-19 [7]. We believe it could also be used as a measure of COVID-19 deterioration risk for patients attending virtual wards.

In this study we investigated whether the COVID-19 virtual ward at UCLH was able to identify patients who were at higher risk of deterioration from COVID-19 and facilitate safe ED reattendance. We also explored whether the ISARIC-4C mortality score could be used to facilitate the identification of these higher risk COVID-19 patients.

Methods

Results

Total cohort description

Between 24/10/2020 and 12/2/2021 there were a total of 649 patient, who received a total of 1182 virtual ward service appointments, with a median of 1 (range 1–10) appointments per patient.

The median age was 47 years (IQR 36–60). The largest ethnic group was White at 39.6% (257/ 649).

Mean CRP was 36.5 mg/L (SD: 48.1), median respiratory rate 18/min (IQR 18–20) and median oxygen saturations 95% (IQR 96–98). The median day of COVID-19 illness was 9 days (IQR 6–13 days. 5/ 649 (0.8%) patients were discharged with dexamethasone, after first ED attendance [Table 1]. All of these had been discharged from inpatient wards in late January 2021 rather than presenting to ED de novo.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Demographics and clinical characteristics at ED presentation.

https://doi.org/10.1371/journal.pone.0284997.t001

Reattendance and hospital admission rate

After an initial virtual ward service appointment, 112/ 649 (17%) patients reattended ED [Table 2].

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Clinical risk factors and outcomes of each of the reattendance groups.

https://doi.org/10.1371/journal.pone.0284997.t002

Compared with patients that did not reattend ED, patients that reattended ED (admitted or not) had a slightly higher median age (54 vs 47 years), higher mean CRP (47.8 vs 34.0 mg/L), lower mean lymphocyte count (1.2 vs 1.4 x1000/μL) and earlier median day of COVID-19 illness (8 vs 9 days) S1 Table in S1 File.

Compared with patients that did not reattend ED, patients admitted to hospital following ED reattendance, had a higher median age (59 vs 47 years) and a higher median comorbidity number (1 vs 0 comorbidities). Ethnicity proportions were similar between these groups [Table 2].

Compared with patients who reattended ED but were not admitted, patient who were admitted had a higher mean CRP (54.2 vs 42.1 mg/L), lower lymphocyte count (0.97 vs 1.41 x1000/μL) and earlier median day of COVID-19 illness (7 vs 8 days) at time of initial COVID-19 assessment. One of the 5 patients who were prescribed dexamethasone after initial ward discharge reattended ED [Table 2].

Compared to patients with unfacilitated ED reattendances, patients with facilitated reattendances had a higher mean CRP (53.63 vs 41.67 mg/L), later median day of COVID-19 illness (8 vs 7 days) and higher percentage likelihood of being non-white ethnicity (32% vs 21%). A higher number of patients with facilitated reattendance were admitted to hospital (31, 60%) compared with patients with unfacilitated reattendance (20, 39%). The above and the level of care data is available in S2 Table in S1 File.

ISARIC-4C score

Mean ISARIC-4C score was significantly higher in patients who reattended ED than patients who did not reattend ED (3.87 vs 3.48, p = 0.003). It was also significantly higher in patients who were admitted to hospital after ED reattendance than patients who did not reattend ED (5.56 vs 3.48, p = 0.04) [Fig 1].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. A figure showing the comparison between ISARIC- 4C mortality scores of non-reattender, reattender and admitted patients.

https://doi.org/10.1371/journal.pone.0284997.g001

There was no significant difference between the ISARIC-4C score of unfacilitated and facilitated reattendances (4.82 vs 4.60, p = 1).

Comparison within our service

Compared to previous service evaluations performed in July [2] and September [3] 2020, patients in this evaluation had a higher mean CRP (18.8 vs 8.4 vs 2.2mg/L), a higher number re-attended ED (112 vs 12 vs 32) and were more admitted to hospital (51 vs 5 vs 7) [Table 3].

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Comparison of cohort demographics, clinical characteristics, and outcomes.

https://doi.org/10.1371/journal.pone.0284997.t003

There were no deaths in previous cohorts.

Discussion

In this study we found that risk stratification of COVID-19 patients attending virtual wards, via subjective clinical assessment and objective ISARIC-4C mortality scores, was beneficial in identifying those most likely to deteriorate from COVID-19.

We observed that those patients with more risk factors for disease progression and evidence of more severe COVID-19, i.e. older, comorbid, higher CRP, lower lymphocyte count, low oxygen saturations and earlier day of COVID-19 illness, [9, 10] were more likely to require ED reattendance.

Additionally, we have shown that there is a difference between the ISARIC-4C scores of ED reattenders and non-reattenders, so that it can be used to help identify those more likely to deteriorate with COVID-19.

In this study the ISARIC- 4C mortality score was used to determine risk of deterioration, rather than the ISARIC -4C deterioration score, due to the availability of required parameters. We have shown that there were statistically significant differences between the groups in our cohort however, there was significant overlap between these groups. It is plausible that the ISARIC -4C deterioration score may have offered an improved discrimination of risk [11]. As developed and validated prior to widespread COVID-19 vaccination availability, further investigation into these prognostic scores current applicability is warranted. As seen with the lack of difference between unfacilitated and facilitated reattendances in this study, there may have been patients who deteriorated more than expected or who reattended without recommendation. This highlights the importance of close patient evaluation on a case-by-case basis to assess risk of deterioration, and not just relying on risk stratification tools.

We noted a higher number of ED reattendances, hospital admissions and deaths in this study compared with earlier analyses of the UCLH virtual ward service [Table 3] [3]. Extrapolation of these results may be difficult given the different cohort sizes, but they are likely explained by the prevalence of more clinically severe disease, evidenced by higher mean CRP and lower mean oxygen saturations in this cohort.

The difference in reattendance rate likely reflects a change in management in response to the worsening clinical severity seen at this point in the COVID-19 pandemic. During the surge of cases in Winter 2020, unprecedented pressures were placed on inpatient hospital beds [12]. At UCLH, as elsewhere, the raised threshold for hospitalisation meant more unwell patients were discharged into the care of the COVID-19 virtual ward service. This may explain the higher ED reattendance and mortality rates, compared to previous cohorts. This increase in disease severity is reflected in other UK COVID-19 virtual wards, which showed comparable ED reattendance rates (4–36%) and mortality rates (0–29%) [13].

This UCLH COVID virtual ward service evaluation is part of a growing literature on a subject that has expanded enormously since the pandemic began. Due to their essential, but rapid development, understanding and measuring the benefit of virtual wards has faced many challenges and produces mixed results.

In a similar timeframe to this study, a retrospective single centre study of 43 patients at a virtual ward of a UK district general hospital (January—March 2021), showed a reduced mean length of hospital stay for COVID-19 virtual ward attendees (10.3 days +/- 9.7 days) compared to all COVID-19 positive patients at this hospital (11.9 days +/- 11.6 days) [14]. Moreover only 9% required admission, like admission rates in these systematic review papers [13, 15]. These review papers also report good clinical outcomes for COVID-19 virtual wards. Published on the 26^th May 2021, a systematic review of 6 databases across 7 countries on COVID-19 virtual wards reported mortality rates of 0–3.1%, reattendance rate of 4–36% and admission rates between 0–29% [13]. A multi-site mixed qualitative and quantitative methods study between July and August 2020 of 8 study sites (looking at 1737 patients) found a similarly low mortality rate (1%) [15].

However, neither of these studies were able to confidently assess the effectiveness of virtual wards, due to lack of standardised reporting and control groups. Limited exploration of virtual wards’ cost-effectiveness and resource allocation, considering the concurrent non-COVID-19 workload outside of the initial stages of the pandemic, may limit their utility [13, 15].

The provision of home oximeters has often been promoted as a core aspect of COVID-19 virtual ward services, [16] however one study indicated pulse oximetry results do not always accurately reflect arterial oxygen saturation [17]. Moreover a multivariate analysis from national observational data of 123 English hospital trusts between August-February 2021 showed where COVID-19 virtual wards were available, inpatient stays were 5% longer. The authors hypothesise that this may be related to earlier admissions and improved bed capacity from simultaneous virtual ward set up, at the tail-end of the UK’s second wave [18].

In our study, although pulse oximeters were given to patients fulfilling specific ‘high risk’ criteria, the data was not available and so cannot comment on their effectiveness. We believe that oximetry is probably still useful in remote clinical assessment, for normal oxygen saturations in mild COVID-19 cases may be reassuring enough to not require ED reattendance.

The COVID-19 vaccination status of patients was not included in the study, as it was not available from the clinical records. At the time of the study only a small proportion of the UK population would have received a COVID-19 vaccine. These would have been frontline healthcare and social workers, or those with increased risk of serious disease or death from COVID-19 (i.e. organ transplant recipients, those with specific cancers, the immunosuppressed, those with severe chronic respiratory conditions or significant congenital diseases) [8]. Even within this cohort the 12 week gap between COVID-19 vaccine doses recommended at the time, meant that most would’ve only had a single dose [19]. It’s lack of inclusion in the study, therefore, may not have affected results. The widespread COVID-19 vaccination status (including boosters) of the population, with the Omicron and Delta variants and the emphasis on the importance of mass vaccination [20], means it should now be included in further evaluations to assess impact on deterioration risk [21].

Limited data also prevented accurate evaluation of level of care received on admission and discharge dexamethasone provision (as well as other outpatient treatments), so these should be studied further.

When incorporating the experiences of our study with that of healthcare services in the UK, and internationally with other similar risk stratification tools, it is important that the differences in COVID variants, community prevalence, access to COVID-19 testing, outpatient COVID-19 treatments, and access to COVID-19 vaccination are all considered when building on COVID-19 virtual wards. When evaluating COVID-19 virtual wards it is important that an assessment of their clinical effectiveness, impact on healthcare systems and cost must be made and properly evaluated, to best adapt them for the contemporary phase of the COVID-19 pandemic.

Conclusion

The COVID-19 virtual ward service at UCLH identified patients with demographic and clinical risk factors associated with increased risk of deterioration. The ISARIC-4C score may be a useful additional method of stratifying risk that might improve the focussed management patients at highest risk of clinical deterioration.

COVID-19 virtual ward services have helped to safely manage patients outside of the hospital setting in the UK prior to widespread vaccination. Prospective studies are needed to better define patient groups likely to benefit from COVID-19 virtual support in contemporary and future phases of the pandemic.

Supporting information