Article Text
PDF
Abstract
Objectives The WHO designated individuals with low oxygen saturation, SpO2<94%, as severe SARS-CoV2 infection (COVID-19) and recommendations to seek care in a hospital setting were advised. A rapid, office-based method to select patients with severe COVID-19 who need intensive care was necessary during the peak of the pandemic.
Design, setting and participants This is a prospective cohort study of patients with confirmed severe COVID-19 between September 2020 and April 2021.
Outcome measures and analysis Oxygen saturation was obtained at rest (SpO2r), following exertion from a 20 m walk test (SpO2e), and the difference was calculated (SpO2Δ). Radiographs and laboratory values were obtained and recorded. Logistic regression models were used to determine variables associated with hospitalisation. A lung injury score was used to quantify pulmonary involvement.
Results Out of 103 patients enrolled with severe COVID-19 infection, 19 (18.4%) were admitted to the hospital (no deaths). Patients managed as outpatients had a standard treatment protocol. The SpO2Δ and SpO2e were associated with hospitalisation (p<0.005) while SpO2r was no different between non-hospitalised and hospitalised patients (90.7%±2.7% vs 90.8%±2.3%, p=0.87). By contrast, exertional SpO2e was significantly different between non-hospitalised and hospitalised (87.3%±2.6% vs 84.4%±3.4%, p=0.0005). The mean lung injury score was 11.0±3.5 (18-point scale) and did not discriminate against those who would need hospitalisation. Lower lung fields were significantly more involved than the upper (p<0.0001). All patients had elevated biomarkers of inflammation, C reactive protein (CRP) median 82.5 IQR (43–128.6) mg/L and evidence of elevated liver enzymes. A logistic regression model was constructed including SpO2Δ, CRP and alanine aminotransferase to predict hospitalisation. Only SpO2Δ was significant, p=0.012, 95% CI (1.128 to 2.704) and correctly classified 85.71% of patients who could remain at home or would need to receive treatment in the hospital.
Conclusion An office-based, 20 m walk test can help diverge patients with severe COVID-19 who need escalated care. Further, an aggressive standardised treatment protocol can be used to successfully manage patients outside of hospitals despite having severe COVID-19.
- COVID-19
- INFECTIOUS DISEASES
- Patient Reported Outcome Measures
- Pulmonary Disease, Chronic Obstructive
Data availability statement
Data are available on reasonable request. Detailed data are available on request from the corresponding author.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
STRENGTHS AND LIMITATIONS OF THIS STUDY
Study covering a minority group that can be applicable nationwide.
A 20 m walk test used in this study is rapid, easily repeatable and inexpensive.
Outcomes enable a predictive model to help in future pandemic-like scenarios, which include physiological, radiographic and biochemical parameters.
The data lack a size, a validation dataset and other demographic groups for direct comparison.
There is potential for selection bias as the decision to refer to the hospital was not specifically protocolled and was based on clinical judgement.
Introduction
Background
Primary care and other points of entry into the medical system were often overwhelmed during the height of the SARS-CoV-2 pandemic.1–5 The COVID-19 is cause by SARS-CoV-26; an enveloped RNA virus that causes respiratory, enteric, hepatic and neurological diseases.7 8 Clinical and diagnostic tools for care developed rapidly but implementation within local communities was challenging.9–13 Hospital diversions, long waits and other delays led to negative health consequences.14 Guidance for treatment caused confusion to both physicians and patients.15
Alivio Medical Center, an organisation that treats an urban, Hispanic population, adopted a triage system and remained open to care for both infected and non-infected individuals during the COVID-19 pandemic.16 Hispanics not only had higher infection rates but also had worse outcomes. Hispanics were less likely to be vaccinated and COVID-19 vaccination appeared to have a lesser impact on this minority population17–19; stable vaccine effectiveness of the Johnson & Johnson (J&J) vaccine is 76% in the USA population but has been found to be as low as 14% in minority cohorts.19 Our population is largely uninsured, employed in essential labour and many experience immigration-related stress. For a variety of complex reasons, many were reluctant to go to the hospital even when advised to do so.20 To reduce the burden on urgent care and the patient population with non-urgent COVID-related needs, the implementation of rapid and efficient methods is imperative to diverge only necessary patients to urgent care.
Objective
Referral to a hospital for intensive care should be reserved for those in most need. Therefore, a suitable and standardised method is needed to rapidly provide the critical decision at the level of primary care to refer and escalate care. For asymptomatic patients or those with mild to moderate disease, outpatient management was appropriate whereas those in obvious clinical respiratory distress required prompt referral to a hospital. However, there was a large subset of patients with severe COVID-19 infections but were not in respiratory distress.21 22 The WHO defines severe COVID-19 as, ‘pneumonia (fever, cough, dyspnoea) involving 50% of lung fields plus: respiratory rate>30 breaths/min or an SpO2<94% on room air’.23 24 The majority of these patients had what is referred to as asymptomatic hypoxaemia with relatively low oxygen saturations and radiographic evidence of pneumonia but clinically short of respiratory distress.25–27 It is well known that these patients have poorer outcomes than those without hypoxaemia but it is not known where best to manage their disease, hospital or outpatient. The objective of this manuscript, therefore, was to identify a rapid triage strategy that would allow us to decide whether to send them to the hospital or to manage them as outpatients.
We implemented a standardised protocol using early oxygen therapy at home, chest physiotherapy, nebulisation and prophylactic anticoagulation with close follow-up. Close follow-up was defined as patient contact every day to assess progress. Patient contact was performed either through office visits, physical therapy visits, phone calls or the use of remote monitoring devices. Early use of oxygen has been shown to improve outcomes in a large cohort in China.28
The protocol we implemented was based on the best medical evidence at the time of the study and clinical judgement.29–32 The authors acknowledge the level of evidence for any individual treatment in this protocol was not strong. However, the clinicians involved felt strongly that it was better to err by commission rather than omission for this cohort of patients.
The most important criteria to determine the need for COVID-19-related hospitalisation appears to be the O2 saturation of the patient.33 Patients with severe COVID-19 are expected to have significantly lower SpO2 after exertion than patients at rest or compared with mild COVID-19 patients conducting similar exertion. This study aims to provide a method of prioritising severe COVID-19 patients for urgent care when the supply of care cannot meet the demand.
Therefore, the goal of this present investigation is to determine if there is a subset of patients with WHO-defined severe COVID-19 infection who can be managed in the outpatient setting and if so, how can they best be identified using tools readily available to primary care physicians.
Methods
Study design
This is a prospective, cohort, follow-up study on outpatients with confirmed severe COVID-19 infections. Resting oxygen saturation was obtained at baseline (‘SpO2r’), exertional oxygen saturation was obtained following a 20 m walk test (‘SpO2e’) and the difference was calculated (‘SpO2Δ’). All patients were treated with home oxygen and given a target SpO2 of >94% by pulse oximetry. Aggressive chest physiotherapy (hourly incentive spirometry at home and chest percussion with breathing exercises were performed by a licensed physical therapist daily, aggressive nebulisation (inhaled steroids budesonide two times per day and bronchodilators, ipratropium/albuterol every 6 hours), methylprednisolone intramuscular injection of 80 mg followed by an oral prednisone taper starting with 60 mg/day and decreasing by 10 mg/day), macrolide-based antibiotic regimen, prophylaxis anticoagulation with either low-molecular-weight heparin (LMWH) (40 mg/day subcutaneous) or rivaroxaban (10 mg/day orally). The choice of rivaroxaban versus LMWH was related to patient preference and cost of drug. Follow-up office visits and phone calls were conducted daily or every other day until patients are deemed to be fully recovered.
Setting
Alivio Medical Center, Indianapolis, was the host for this study. Alivio Medical Center is an organisation that treats an urban, Hispanic population. The recruitment of this study was between September 2020 and April 2021. All study staff in this investigation spoke Spanish fluently and conducted interviews, collected demographic and socioeconomic information, and provided treatment in the Spanish language.
Participants
All patients presented at the Alivio Medical Center who were diagnosed with severe COVID-19 between September 2020 and April 2021. The inclusion criteria were as follows:
Confirmed COVID-19 infection.
Severe COVID-19 infection defined by the WHO definition; SpO2<94% for this study.
Greater than 18 years of age and able to sign informed consent.
Demonstrate the willingness and ability to comply with our protocol.
All participants had SpO2 less than 94% and lung infiltrates greater than 50%. Patients presenting in respiratory distress were immediately transported to the hospital without delay and were excluded from enrolment. Other exclusion criteria included subjects whom the investigators felt would not be able to comply with the treatment protocol or had inadequate access to transportation and communication.34 The decision to transfer was based on the clinical judgement of the provider caring for the patient. Enrolment for the study was from September 2020 to April 2021. All patients were called daily by a member of the COVID-19 care team. All patients were placed on a standard regimen as described above (table 1). Intramuscular and oral steroids were used unless glucose levels precluded them.29 Isotonic saline was used liberally when volume deficit was present. If evidence of bacterial pneumonia was present, then levofloxacin was given. COVID-19 lung disease is normally characterised as a dry non-productive cough or with scant clear sputum. If the patients had a cough productive of purulent sputum, we assumed it was caused by a bacterial infection. Follow-up was for 30 days or until full recovery. Full recovery was determined by the treating physician and defined as cessation of all COVID-19 medications and return to normal activities of daily living. Death from any cause within 30 days was censored as death. The schedule of follow-up visits was determined by the severity of the illness. In some cases, it was every day either by phone call or by office visit; in others, it was every other or every third day.
Treatment protocol for patients with severe COVID-19 infections at Alivio Medical Center
Variables
All laboratory values were run by HealthLab Clinical Laboratory. Aspartate aminotransferase (AST) and alanine aminotransferase (ALT) were expressed as IU/L. C reactive protein (CRP) is expressed as mg/L and D-dimer is expressed as D-DU ng/ml. CRP and ALT are reported to be associated with mortality in COVID patients35 36; CRP is a broad marker of the overall inflammatory state of the patient and ALT indicates liver dysfunction. The upper limits of normal in our lab are 10 mg/L for CRP, 243 ng/mL for D-dimer, 54 IU/L for AST and 51 IU/L for ALT.
Data source
Electronic medical records in iSalus were used to compile patient’s baseline demographics, comorbidities, medications, laboratory and socioeconomic data. Chest radiographs and clinical courses were documented.
Study size
Out of 103 patients enrolled with severe COVID-19 infection in the time frame, 19 (18.4%) were admitted to the hospital. There was no sample size calculation performed as this was a prospective cohort design. The sample size was determined by the inclusion/exclusion criteria and all patients eligible were enrolled and included in this analysis.
Quantitative variables
The primary means of confirming COVID-19 infection in this cohort was via PCR and rapid antigen test from a nasal swab. The PCR test was the Bridge Diagnostics SARS-CoV-2 PCR Method, which is performed using an automated version of the ThermoFisher Scientific TaqPath SAR-CoV-2 Combo Kit. The antigen test was a rapid point-of-care, lateral flow assay.
Oxygen saturations were measured with a finger pulse oximeter; Henry Schein is the distributor of the pulse oximeter used in this study. Radiological lung injury scores were based on a method developed by a group in Brescia Italy.37 This system is known as the Brixia system and consists of two steps in image analysis. First, the lung is divided into six zones on the anteroposterior projection. The right and left upper zones are above the aortic arch, the middle zones are below the inferior wall of the aortic arch and above the inferior wall of the pulmonary veins (this includes the hilar structures), and the lower zones consist of all pulmonary structures below the wall of the pulmonary vein. Each lung zone is then given a score from 0 to 3 based on lung abnormalities detected as follows:
Score 0: no lung abnormalities.
Score 1: interstitial infiltrates.
Score 2: interstitial and alveolar infiltrates (interstitial predominance).
Score 3: interstitial and alveolar infiltrates (alveolar predominance).
The scores of each of the six zones are then added to obtain a total score of 0–18. All chest radiographs were reviewed and scored by Chris Owens and Alfredo Lopez were two of the primary physicians caring for the COVID-19 clinic during the pandemic at Alivio Medical Center. Together they oversaw the care of more than 10 000 office visits for COVID-19 infections. This scoring system was shown to be significantly higher in patients who died than those who survived the infection.
Statistical methods
Statistical analyses were performed on STATA/SE v.11.0 (StataCorp). Baseline characteristics are presented as mean±SD or median and IQR depending on the normality of their distribution. Differences in means were determined by Student’s t-test, either paired or unpaired, and proportions between groups were compared with the χ2 test. Logistic regression was performed with the logistic regression (LOGIT) function with hospitalisation modelled as a binary outcome. There was no sample size calculation as the study is descriptive in design; a relaxed rule of thumb for logistic regression was used to accommodate the sample size.38 Positive and negative predictive values were generated by the output command (LSTAT).
Patient and public involvement
None.
Results
Participants
Between September 2020 and April 2021, 4112 patients confirmed with COVID-19 were treated at the Alivio Medical Center. This study consisted of 103 consecutive subjects meeting inclusion criteria of severe COVID-19 infection and hypoxia, mean SpO2r was 90.7%±2.6%, chest radiographs demonstrated abnormal ground glass infiltrates. Nineteen (19 of 103) patients, 18.4%, were hospitalised during the study with the remainder (84) recovering fully from the infection. The 20 m walk test was done consistently in all patients at the time of presentation. Among these 19 patients, the mean time from initial trial of outpatient management to referral to the hospital was 3.64±2.34 days. There were no deaths among the patients who eventually went to the hospital, hence no apparent detriment due to delayed referral.
Descriptive data
The baseline characteristics are present in table 2. In brief, all were Hispanic, about one third had type 2 diabetes mellitus, and the mean temperature was 37.9°C±1°C. There were no missing participant data for any of the variables of interest.
Patient demographics and vital statistics at presentation to Alivio Medical Center
Outcome data
The decision to refer to the hospital was made by the treating physician. There were no specific criteria for referral; criteria were based on deteriorating clinical status and clinical judgement determined by the licensed medical professional. There were no demographic features independently predictive of hospitalisation in our sample such as advanced age or diabetes.
Main results
There was no difference in SpO2r between those hospitalised versus those not (90.7%±2.7% vs 90.8%±2.3%, p=0.87, table 2). By contrast, exertional SpO2e was significantly different between non-hospitalised and hospitalised (87.3%±2.6% vs 84.4%±3.4%, p=0.0005), indicating that those who eventually were referred to the hospital had lower pulmonary reserve function and, therefore, less tolerance to exercise. Further, the SpO2Δ between the groups were also different, 4.1%±1.8% vs 6.4%±2.2% (p=0.0001).
Overall, the lung injury score was 11±3.5 out of a possible of 18, indicating all patients had severe pulmonary involvement. However, there was no difference in lung injury scores between those hospitalised and those not, 11.11± vs 10.9±3.5, p=0.84. Lung injury due to COVID-19 preferentially affects the lower lung fields as seen in table 2. The mean lung injury score for the lower lung fields was 2.2±0.7 while those in the upper lung fields were just 1.3±0.55, p<0.0001 (figure 1).37 Lung injury scores did not correlate with resting or exertional SpO2.
Lung injury scores as defined by Borghesi and Maroldi.37 The lower lung fields were more involved and had higher injury scores than the upper lung fields (1.3±0.55 vs 2.2±0.7, p<0.0001). Solid dark-grey line represents the right lung and the dashed light-grey line represents the left lung.
Other analyses
Subjects had significant biochemical evidence of system inflammation at presentation characterised by high plasma levels of CRP, mean 98.8±121.7 mg/L median and IQR is 82.5 (43–128.6) mg/L and elevated liver enzymes (International Classification of Diseases diagnostic code, Tenth Revision (ICD-10), associated with it R74.0), ALT 66.9±63.6 mg/dL and AST 56.7±45.4 mg/dL, however, these were not significantly different between those hospitalised and those not (table 2). The white cell count, per cent lymphocytes or D-dimer did not offer further discriminatory value.
A logistic regression model was developed using a univariate association with hospitalisation at a statistical level of p<0.1 as the entry criteria into the multivariable model (table 3). No demographic or radiographic variable met this entry criteria. The final model consisted of both biochemical (CRP, ALT) and physiological parameters, SpO2Δ. Out of these, only SpO2Δ remained significant at the multivariable level (OR=1.75, p=0.012 95% CI 1.128 to 2.704); for ALT the OR is 1.018 with an SE of 0.014 and for CRP the OR is 1.005 with an SE of 0.005 (see table 3). This model correctly classified 85.7% of patients into hospitalisation or not and had a positive predictive value of 83.3% and a negative predictive value of 86.1%.
Multivariable regression model
Discussion
Key results
The data suggest that most individuals presenting with severe COVID-19 infection, when carefully selected, can be managed as outpatients with an aggressive office-based treatment regimen and close follow-up. These patients must comply with treatment, have reliable transportation and phone as well as another person at home to assist in their care. This finding is especially pertinent given the overburdened hospital system and the reluctance of Hispanic patients to seek care in tertiary settings. It was remarkable to us that many subjects, despite the degree of hypoxaemia and radiographic evidence of severe disease, were relatively mildly symptomatic. This may be due to a predominantly young, working demographic.
Early oxygen therapy was the cornerstone of our treatment protocol. In a study from China, initiation of home oxygen more than 2 days after the onset of dyspnoeic symptoms increased the risk of death while prior to 2 days was protective.28 Hence, we felt that the earlier-the-better approach with oxygen therapy was important. Indeed, as this investigation progressed, we increasingly adopted a lower threshold for earlier use of oxygen in patients with moderate COVID-19. In most patients, we saw a quick response (within 2 days) to this protocol. Those who did not respond or worsened despite the protocol were sent to the hospital. All 19 patients who were sent to the hospital were discharged without any apparent increase in morbidity suggesting that we did not ‘burn any bridges’ during this time.
The most significant findings in the present investigation were that postexertional oxygen saturation and the difference between resting and postexertional oxygen saturation were significantly different between those who were referred to the hospital and those who were not. This is contrary to the findings by Goodacre et al, who did not find that postexertion oxygen was a significant predictor of adverse outcomes.39 However, there are noteworthy differences between their study and ours. First, this was conducted on patients presenting to the emergency department of the hospital, exactly what we were trying to avoid. It is possible that patients presenting to the hospital are substantially different than those in an outpatient clinic. Second, they enrolled patients with suspected COVID-19 infections and not confirmed COVID-19 as we did in the present manuscript. Therefore, it is likely that their cohort had other respiratory diseases as well. Third, and most significant, is that the baseline oxygen saturations in their cohort were much higher than ours suggesting that many patients enrolled had milder disease. Clearly, more work is needed in this area.
Study limitations
One question that remains is how to choose a priori among those with severe COVID-19 who will most likely require treatment in a more intensive setting.36 40–45 Most published models only include demographic data such as age, gender, race and comorbidities, and use all COVID-19 positive subjects, irrespective of symptomatology, to develop and calibrate the model.42 44 45 Our model, by contrast, was developed only in outpatients presenting with severe COVID-19 infections where a decision had to be made to refer to the hospital or continue to manage them as outpatients. This is the main triage question primary care providers are faced with in the setting of a pandemic and is, therefore, much more clinically relevant. Our model includes both physiological and biochemical parameters that can be quickly and easily measured. Indeed, the 20 m walk test was used to assess pulmonary reserve for the very reason that it has been shown to be robust in a variety of office and rehabilitation settings and its reliability is known.46
The limitations of this study include a small sample size and lack of a validation set for the prediction model despite the viability when adopting a relaxed rule of thumb for logistic regression was used to accommodate the sample size.38 In addition, this cohort consisted only of Hispanic patients and may not be generalisable to other populations. As previously published, our patients are primarily Spanish-speaking, monolingual, uninsured and are employed in essential labour.16 Many patients were reluctant to go to the hospital due to complex psychosocial factors such as immigration status or distrust and fear. Cervantes et al claim that worry about healthcare costs, concern about the ability to access care if uninsured or undocumented, and fear of deportation are barriers.47 This study supports these findings. Furthermore, we feel the model, while minimally over saturated, is the most clinically relevant because it can be performed rapidly in the outpatient setting with inexpensive equipment. Of course, with the next wave of the pandemic, we can expect to check this validity.
It could be argued that our protocol overtreats some patients or was not completely evidence based. However, we developed this protocol based on the limited evidence available for outpatient treatment at the time of enrolment. During this time, there was significant confusion surrounding best practices leading to a sense of treatment nihilism in some outpatient clinics.15 We adopted a policy of better to err on the side of commission rather than omission in the development of this protocol. Dexamethasone is supported in the RECOVERY trial,29 and a recent Cochrane review supports the use of inhaled steroids.31 Macrolides, in particular azithromycin, have been shown to be effective in vitro and in vivo.48 49 Although we acknowledge that more recent data refute some of the earlier evidence and recognise that macrolides are no longer indicated for COVID-19 infection.50 We also acknowledge decision-making regarding hospitalisation based on visible respiratory distress and the lack of specific referral criteria. This may cause bias in the results based on the treating physician’s decision-making at the time of tests such as the 20 m walk.
Another limitation is that we did not have arterial blood gas confirmation of the hypoxaemia detected by pulse oximetry. However, given that it is well known that darker skinned individuals are more likely to have falsely elevated oxygen saturation when measured with pulse oximetry thereby masking hypoxaemia, a phenomenon known as ‘occult hypoxaemia’ patients in this cohort may have actually had lower oxygen values then reported.51
Finally, it is likely that our sample size was insufficient to determine if CRP and elevated liver enzymes were independent predictors associated with hospitalisation leaving the possibility of type 2 error. However, these limitations notwithstanding, we conclude that some patients with severe COVID-19 infections can safely be treated at home with this protocol while acknowledging the caveats. A physiological pulmonary stress test such as a 20 m walk can identify most individuals a priori who are unlikely to respond to this treatment and should be referred for escalation of care.
Interpretation
This study was conducted during the peak of the COVID-19 pandemic where healthcare services were struggling to provide care for the mass influx of patients. As a means to mitigate some of the burden on hospitals, the Alivio Medical Center sought to rapidly develop a sufficient method of grouping patients into those that required minimal intervention and those requiring urgent care. Although the 20 m walk method is not perfect at distinguishing vulnerable patients, it is a rapid and inexpensive method to check a large cohort of patients, especially in extraordinary circumstances. This study was completed in a relatively short amount of time with a small sample size of only one socioeconomic demographic group (working class Hispanic patients). Other studies have not found the same results in samples representing the majority of US patients, which suggests more research is needed to demonstrate its effectiveness, particularly in minority groups.
Generalisability
Our prediction model was a simple 20 m office-based walk test that includes a standard pulse oximeter. All patients had low SpO2r and evidence of severe lung injury as evidenced by the Brixia score. This is a semiquantitative scoring system that ranges from 0 to 18. The mean score in this investigation was 11, indicating severe and bilateral lung disease. Yet, the radiographic findings were similar in all patients enrolled and had poor discriminatory ability to predict treatment response. The difference in oxygen saturation from resting to after exercise is likely a measure of the amount of pulmonary reserve and better reflects overall health and resiliency of the lungs. Therefore, a large SpO2Δ was the single greatest predictor of patients who would need care in a more intensive setting and can be used as a rapid triage tool during a pandemic in an outpatient setting. Our treatment protocol aimed to improve pulmonary reserve as all subjects were given an incentive spirometer and specific chest physiotherapy exercises, in addition to oxygen, to be performed in the office and at home.
Conclusions
To treat patients with severe COVID-19, a treatment regimen based on home oxygen and physiotherapy represents a feasible alternative to hospitalisation in some patients who meet the inclusion criteria and appear to be safe, given no deaths. An office-based 20 m walk test is crucial for the evaluation. The difference between resting and exertional oxygen saturation predicts individuals at high risk for hospitalisation. The protocol described in this study may be particularly efficacious and efficient with low-income minority populations with limited access to care and a reluctance to use available healthcare systems.
Data availability statement
Data are available on reasonable request. Detailed data are available on request from the corresponding author.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and was approved by Western IRB: IRB Pr #: 20210840; Work Order #: 1-1404432-1. Participants gave informed consent to participate in the study before taking part.
Acknowledgments
The author acknowledges Ryan Dagenais for editorial support with the manuscript.
References
Footnotes
Contributors Conceptualisation: CO. Guarantor: CO. Methodology: CO. Collecting data: JS and MQ. Laboratory analyses: JS and MQ. Data analysis: JS and MQ. Writing—original draft preparation: CO. Writing—review and editing: CL. Supervision: CL and AL-Y. Authors have read and agreed to the published version of the manuscript.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.