Hospital acquired COVID-19 infections amongst patients before the rollout of COVID-19 vaccinations, a scoping review

Eligibility criteria for studies was formulated using the Population-Exposure-Comparison-Outcome-Time (PECOT) [19]. We included peer-reviewed primary studies of experimental and descriptive designs. The inclusion criteria were as follows: Participants—patients receiving health care of any kind; Exposure—admission at a healthcare facility of any type and in any country; Comparison—not important for this study; Outcome —COVID-19 positive result using any approved test; Time—before the availability of COVID-19 vaccines. The exclusion criteria were: studies reporting HAIs in healthcare workers only, non-English publications, non-peer reviewed grey literature, systematic and scoping reviews. Given the varied range of descriptive studies conducted during the rapid response to the COVID-19 pandemic, the most applicable definition for HAI with respect to the pathogenesis of COVID-19 was not applied in the eligibility criteria. One example of a strict definition for any hospital-acquired (i.e., nosocomial) infection of a patient is: ‘a positive test/symptoms within 48 h of hospitalization or within 3 days of discharge or within 30 days after an operation’ [20]. A strict definition which has been used for COVID-19 nosocomial infection of patients based on the pathogenesis profile of SARS-CoV-2 is: ‘a positive SARS-CoV-2 Reverse transcriptase polymerase chain reaction (RT-PCR) result on hospital day 3 or later or within 14 days of discharge’ [21]. It was not expected for many studies to adhere to this definition given many challenges and delays in acquiring and maintaining sufficient RT-PCR diagnostic resources in some countries, for example. The inclusion criterion for the exposure was therefore relaxed in this scoping review, in cognisant of these obvious limitations, rather, the definitions and criteria used by authors to refer to ‘COVID-19 HAI’ were accepted.

An initial search of PubMed, Scopus, Web of Science, Cochrane library and Google scholar for grey literature was conducted to assess the availability of information and the types of terminology used to describe the Outcome, Participants and Exposure of interest. This output was used to formulate the search strategy using identified key words combined using Boolean operators. The initial search strategy was piloted for each database in duplicate by two authors during May 2021, and was supervised by information management expert author NL through discussions with the author pairs running the search, practical review of the search results, and reiterations of search strategies, until consistent results were obtained and agreed upon by senior authors NKN, NL and DN. On June 1, 2021, one author used the finalized search strategies and conducted the final search in PubMed, Scopus and Web of science. The search terms for Participant included: ‘patients’ OR ‘hospitalized’; Exposure search terms included: ‘hospital acquired’ OR ‘nosocomial’ OR ‘healthcare associated’. Outcome search terms included queries that mention COVID-19 and SARS-CoV-2 related key words: ‘Wuhan coronavirus’ OR ‘2019-nCoV’ OR ‘coronavirus disease 2019’ OR ‘severe acute respiratory syndrome coronavirus 2’. The comprehensive list of key words and search terms is provided in Additional file 2.

The search output was imported into RAYYAN software for screening [22]. Duplicate records were removed within RAYYAN prior to screening. The screening of titles and abstracts was conducted from 3 June 2021 to 20 August 2021. Two pairs of authors (TMM & RD; AT & DN) supervised by NKN, independently screened each record (title and abstract) to identify potential eligible articles using the Participant, Exposure and Outcome criteria. The title/abstract was classified as either ‘Include’, ‘exclude’ or ‘maybe’ within the RAYYAN tool to shortlist articles for full text review. Three separate databases corresponding to abstracts classified by both reviewers as ‘include’ or ‘maybe’ or had conflicting classifications, i.e., ‘conflict’, were exported into EndNote. Abstracts classified as ‘maybe’ and with disagreement between the screening pair were discussed and resolved with the lead author (NKN). Full texts for the included records were retrieved and the same peer-screening and supervision process to assess the eligibility of titles and abstracts were followed to screen the full texts.

The author team developed a data extraction form, adapted from the EPOC Good Practice Data Collection guidelines (EPOC 2017b), and piloted it with five randomly selected shortlisted articles [23]. The data extraction items included author and publication year, date and country of data collection, type of healthcare facility and setting (i.e., clinical service/illnesses), age distribution of study sample population, definition used for the exposure, measure and type of the outcome, any factors associated with the outcome and recommended prevention strategies. The final tool was used independently by two reviewers (TMM & NKN) to extract the data, between 16 September and 20 October 2021. Disagreements were discussed and resolved between the reviewers and further reviewed and supervised by a third author DN.

We summarised the extracted data descriptively in Microsoft Excel. We presented ranges of estimates of COVID-19 HAIs by timing of data collection relative to implementation of prevention protocols and by the type of healthcare setting. In addition, the data were collated by country and sampled age-groups. We further discussed the external validity of observed results and the definitions used for ‘hospital-acquired’ infections with respect to SARS-CoV-2.

Use of the JBI checklist for quality of evidence was explored [24]. The study design of each eligible article was critically reviewed against the corresponding JBI checklist to obtain a quality score.

The use of JBI checklist for quality of evidence could not be used for any of the included studies due to the nature of convenience sampling and study designs tailored for outbreak investigation and to provide urgent response to control the spread of COVID-19 [24]. For example, although many studies were described by authors as cohort designs, they did not meet the epidemiology definition of a cohort. However, the studies were observational in nature using some form of retrospective or prospective cohort approaches, outbreak investigation descriptions or cross-sectional reports. None of the studies performed a sample size calculation to report on both internal and external validity of their results. Some sample sizes were as small as 11 yet some were large enough to offer some level of external validity such as hospital sample sizes of over 6000 patient records.

The included studies (Additional file 3) were conducted in 17 countries in Europe, Asia and North America, and including one study from Brazil. The highest number of studies at country-level was from the United Kingdom (UK) (10/45) followed by the United States of America (USA) (6/45). Besides Brazil, no studies from South America, Australia, and Africa, met the inclusion criteria for this review. Most of the data were collected between March and May 2020 from a diverse array of healthcare settings including long-term care facilities, dialysis units, cancer and other operative surgery wards including neurology units, psychiatric centers, academic and general hospitals. Age groups varied, although most studies reported data from adults over the age of 50 years.

COVID-19 HAIs were largely reported as simple proportions out of the total number of screened hospitalized patients. Only two studies reported incidence rates: a study in France reported a COVID-19 HAI rate of 155.6 cases per 100,000 in-hospitalized patients [27], and another in the USA [28] reported an HAI rate of 0.8–5.0 cases per 10,000 patient days. The remaining studies reported proportions between 0 and 65.00%. Studies with similar sample age-groups did not necessarily report similar estimates of HAIs either (Additional file 4). The 29 studies with adults-only samples reported a wide range of HAI estimates from 0 to 65.00%. The estimates from the two studies reporting on children/neonates also fell within this range, i.e., 27.08–35.00%.

Four studies did not provide specific dates for the data collection period. Out of the remaining 41 studies, 95% (39/41) initiated data collection before May 2020 during the very early stage of the global pandemic and 37 of them completed data collection by June 2020 (Additional file 3). Besides the remaining 2/41 studies conducted in July (Middle East) and December 2020 (China), there were no studies identified after this early phase. It was difficult to compare estimates between countries due to the varying number of studies but appeared to overlap between countries in cases where several studies were reported. For example, the ranges of HAI estimates were 6.48–35.00%, 0–46.29% and 0.02–53.60% in Spain, UK, and USA, respectively (Additional file 5). Within the North American region, Canada had a lower range of estimates (4.60–19.00%) but from only two studies compared to the USA (0.02–53.60%) with six studies. Other European countries, besides the UK and Spain, reported results from only one, two or three studies and the estimates were anything between 0.16 and 65.00%. Although most studies were conducted over the same time-period and overlapping calendar months, the HAI estimates were not similar between studies from the same country, e.g., the two studies in Belgium reported HAI proportions of 2.84% (conducted in March–May 2020) and 65.00% (conducted in March–April 2020).

COVID-19 HAI estimates appeared to differ by timing of implementation of NPPIs relative to when a study was conducted (Table 1) as well as by type of healthcare service (Table 2). Only 3/14 studies conducted before the implementation of NPPIs at study sites, reported low (< 7%) HAI proportions while the remainder 11/14 studies reported proportions between 18.50 and 65.00% (Table 1). On the contrary, the majority (18/22) of studies conducted post-initiation of NPPIs at study sites, reported low (< 7%) HAIs proportions and only 4/22 reported high proportions between 15.40 and 53.60% (Table 1).

Low estimates (< 7%) of COVID-19 HAIs were also observed in most of the studies conducted in healthcare settings servicing cancer patients (6/9), neurology (3/4) and other operative surgery (4/5) (Table 2). Two of these studies (one in a cancer setting and the other in operative surgery) were amongst the only three with low HAIs but conducted before NPPIs were implemented at study sites [29‐31]. Although there were fewer studies grouped by other healthcare services, most studies in psychiatry hospitals (2/2), dialysis units (3/4) and long-term care (2/2), reported high HAI proportions, all over 25.00% (Table 2). Two of these studies, one conducted in psychiatry [24], and the other in a long-term care setting [25], were amongst the only four studies with high HAIs after initiation of NPPIs at study sites. The remainder two studies reporting high HAIs during use of NPPIs were conducted in general hospital wards. One of these was conducted twelve months into the pandemic in China and this unexpected outcome was related to a possible relaxation in adherence to protocols amongst hospital personnel [32].

Some recommendations for tailored NPPIs to maximize efforts to prevent HAIs were noted from the reviewed studies and could be adopted across the globe. These include the following: In all healthcare settings, adherence to strict spatial separation of patient beds and frequent intensified training of healthcare workers are recommended alongside the standard recommended prevention procedures [44]. Cleaning of touch surfaces was suggested to be as frequent as every two hours in order to be effective [45]. Strategies to reduce airborne transmission included use of portable particulate filters in wards and regular meticulous cleaning of non-invasive ventilation apparatus which have been reported to spark super-spreading events [32, 46, 47]. Universal testing of COVID-19 infection on admission appeared to be effective in lowering HAIs, with some hospitals recommending daily screening and testing of symptomatic hospital habitants [41, 45]. This requires properly designed holding wards while waiting for results, without promoting spread amongst these waiting patients. Whether this is feasible in low-resource and high-volume hospital settings is of concern.

In operative surgeries, universal recommendations for different types of surgery operations included screening and testing within 48–72 h pre-operative; isolation and delayed elective/non-urgent procedures if symptomatic, and only proceeding after a confirmed negative RT-PCR test result [28, 48, 49]. Given the short period of monitoring before surgery, two different sources of swabs (e.g., throat and nasal) or a swab and a chest X-ray appeared to be preferred for accuracy of COVID-19 diagnosis [50]. Delay in surgery procedures by at least 14 days if symptomatic or conducting double sequential testing for those who are asymptomatic was also recommended [51, 52]. Post-operative monitoring with at least one additional RT-PCR test is recommended for at least 14 days and until 30 days post-operative [28, 48, 49]. The recommended measures were reported to be effective and observed perioperative HAIs in this context were associated with specific risk factors including complicated procedures such as transplant surgery or presence of chronic comorbidities, advanced malignancies and recent chemotherapy [49, 53].

If periodic monitoring of adherence to prevention measures, review and re-strengthening are not performed, there is a chance of re-introduction of outbreaks [54]. The issue regarding asymptomatic healthcare workers and patients with delayed viral shedding, which could lead to unexpected outbreaks even in well-controlled healthcare centers, remains a concern [55, 56]. Vaccinating everyone could be the best and only solution for these concerns, further emphasizing the critical urgency of increasing vaccine access and coverage across LMICs, where effectively maintaining NPPIs may not be that easy in many hospital settings. The waning of vaccine-induced immunity against COVID-19 also emphasizes the need for monitoring and evaluation of NPPIs in healthcare settings to ensure these are tailored appropriately and adhered to. Monitoring the rate of COVID-19 HAI as part of routine practice, could be another approach to evaluate the effectiveness of NPPIs as well as the gains made from the vaccination campaigns of healthcare workers and hospitalized patients.

Grey literature was not included in the study. Only peer reviewed published manuscripts written in English were considered. However, peer reviewed manuscripts provide more reliable scientific evidence and are the sources referenced by policy guidelines.

Other limitations observed are not from the study design and methodology but from the observed retrieved literature. The first is the fewer number of studies from most countries, making it impossible to understand inter-country differences, if there were any. The second is the inability to understand the possible influence of different pandemic phases because nearly all studies were conducted during the same and earliest period of the pandemic. Thirdly, the studies fell short in terms of epidemiologically rigorous research designs and reporting measures, hence quality assessment to report the strength of evidence could not be conducted.

None of the included studies conducted sample size and power calculations, probably due to the rapid outbreak response conditions. Although the observed results are useful to understand the importance of adhering to strict prevention protocols in healthcare centers, there is not direct external validity to inform specific scenarios of low-income settings.

Most of the results with clear definitions of nosocomial infections and meticulous prospective surveillance systems were from well-resourced academic hospitals. The heterogeneity observed in the definitions used for HAI can also make it difficult to compare estimates from different settings. HAI definitions used for many operative surgeries were consistent and relevant to the nature of healthcare service provided, for example, a negative result 48–72 h pre-operative followed with a positive result during hospital stay 2–14 days post-operative[48‐50]. Definitions varied widely in other healthcare service settings. Although some studies defined HAIs based on the median incubation period of SARS-CoV-2 known at the time of the outbreak, a negative diagnosis result at point of admission was not confirmed for example [27, 57]. However, other studies confirmed negative infection status at point of hospital admission followed with a positive result within 14 or more days of hospitalization and until 14 days after discharge [30, 58]. Given current knowledge about the pathogenesis profile of SARS-CoV-2 in humans, we recommend that the definition of COVID-19 HAI should include confirmed uninfected status during the first 72 h of hospital admission, followed by a positive result thereafter during hospitalization or until 14–30 days post-discharge [20, 45, 59]. In addition, given the wide range of the possible incubation period, we recommend including the approach used by Bhogal et al., 2020, to further stratify HAI into ‘definite’, ‘probable’ or ‘indeterminant’ for > 14 days, 8–14 days and 3–7 days after admission, respectively [53]. HAIs diagnosed post-discharge could be described as 0–7 days, 8–14 days, 15–30 days post-discharge for ‘definite’, ‘probable’ and ‘indeterminant’, respectively.

Evidence is clear that vaccination reduces the risk of severe COVID-19 disease and mortality and it could, speculatively, also reduce person-to-environmental spread of high viral load [2, 60‐62]. While NPPIs are essential for as long as the COVID-19 pandemic persists, vaccination as known historically [63], is the key to curbing the spread of SARS-CoV-2. General care hospitals could be the ideal points to monitor the need for and effectiveness of booster doses due to the high turnaround of patients, combined with the daily in and out shifts of healthcare staff [62, 64, 65]. The hospital setting also provides an ideal place to assess the benefits of high vaccination coverage in minimising person-environmental surface preservation of viruses. Periodic random sampling of different types of healthcare settings could be useful to monitor the ecological effectiveness of increasing vaccination coverage and of increasing booster dose coverage.