SARS-CoV-2 cases reported from long-term residential facilities (care homes) in South Africa: a retrospective cohort study

We implemented a retrospective cohort analysis of SARS-CoV-2 positive cases in LTCFs across South Africa from 5 March 2020–31 July 2021.

DATCOV, a hospital surveillance system for COVID-19 admissions, was initiated on 1 April 2020 and then subsequently expanded to include sentinel surveillance in LTCFs, implemented on 4 June 2020. Data are submitted by LTCFs that have agreed to report COVID-19 cases via the DATCOV LTCFs module. Participation in the LTCFs surveillance was voluntary and included a small number of sentinel facilities. When new facilities enrolled, they captured historical cases going back to their first recorded SARS-CoV-2 case.

A range of LTCFs or congregate settings were included in the sentinel surveillance, including old age homes (21, 46.7%), retirement villages (11, 24.4%), mental health facilities (7, 15.6%), substance abuse recovery facilities (4, 8.9%) and frail care facilities 2 (4.4%). An old age home is defined as a LTCF where residents require daily care in a comfortable, safe and active environment [9]. Retirement villages are defined as accommodating places that provide an independent lifestyle for those who do not need additional living assistance [10]. A frail care centre is defined as a place giving care to those who are unable to care for themselves as a result of a motor-vehicle accident, physical disability, severe stroke or old age [9]. Psychiatric and mental hospitals are defined as specialized hospital-based facilities that provide inpatient care and long-stay residential services for people with mental disorders [9]. These facilities are usually independent and stand-alone, although they may have some links with the rest of the healthcare system. Substance abuse rehabilitation treatment facilities are defined as centres rectifying maladaptive behaviours and providing help with recovery from substance abuse disorders [11].

The study population included all LTCF residents and staff in participating LTCFs in South Africa.

For the purpose of this study, we defined outbreaks in a LTCF as follows: [12]

The wave periods were defined by the case incidence data with a national weekly incidence of 30 cases per 100,000 [13] as cut off for start and end of wave periods:

Data collection was either through direct entry onto the DATCOV online platform or through importation of electronic data via bulk-upload for LTCFs that did not have a stable internet connection. The dedicated data entry clerk in the LTCF would complete the electronic data form for any person who tested positive for SARS-CoV-2. The case reporting form was adapted from the World Health Organisation (WHO) SARS-CoV-2 case reporting tool and included basic demographic data, pre-existing health conditions and outcomes (died and recovered).

Data imports contained validation checks to identify data errors. Routine checks were performed on all data. Missing and discrepant data were followed up telephonically or by email with the submitting person.

COVID-19 mortality was defined as a death related to SARS-CoV-2 that occurred at the LTCF or while being admitted to hospital, excluding deaths that occurred due to other causes or after recovery. A COVID-19 death is defined for surveillance purposes as a death due to a clinically compatible illness, in a confirmed COVID-19 case, unless there is a clear alternative cause of death that cannot be related to COVID disease (e.g. trauma), with no period of complete recovery from COVID-19 between illness and death [14].

Descriptive statistics including frequencies and percentages were used for categorical variables, while for continuous variables a median and interquartile range (IQR) were calculated.

A random effect multivariable logistic regression model was used to assess risk factors for mortality amongst LTCF residents with laboratory-confirmed SARS-CoV2. Age, race, sex and comorbidities (hypertension, diabetes mellitus, chronic cardiac disease, asthma, other chronic respiratory disease, chronic renal disease, malignancy in the past 5 yrs, HIV, past and current tuberculosis), smoking, obesity, province, type of LTCF, and wave period, were included in models assessing risk factors for COVID-19 mortality. We assessed all variables that were significant at p < 0.2 on univariate analysis and dropped non-significant factors (p ≥ 0.05) with manual backward elimination. Pairwise interactions were assessed by inclusion of product terms for all variables remaining in the final multivariable additive model. We also reported the univariate association of all covariates evaluated in the analyses described above to the main outcome (mortality in individuals with COVID-19). The statistical analysis was implemented using Stata 15 (Stata Corp®, College Station, Texas, USA).

The data used for this study were de-identified to ensure confidentiality. All personal information of the residents and staff, concerning health status, treatment or stay in a health establishment, were kept confidential and stored in a secure server. For analysis, patient identifiers were de-linked from other data and stored separately. Ethical approval for this study was obtained from the University of the Witwatersrand Human Research Ethics Committee (M160667). The study was performed in accordance with all relevant ethical guidelines and regulations and with good clinical practice.

Disease surveillance is a critical function of the NICD as a statutory body in South Africa. The NICD has a national mandate to conduct COVID-19 hospital surveillance. The amended regulations that accompany the declaration of a national disaster (Disaster Management Act 2002), provides for healthcare institutions to submit data to NICD on notifiable medical conditions, which includes COVID-19. This encompasses the submission of patient details, their treatment and outcomes. In this case individual consent is therefore waived.

The proportion of cases in staff decreased over the surveillance period (Fig. 1 b).

There were varying patterns of outbreaks among LTCFs, with six (13.3%) reporting no outbreaks (only sporadic cases), 10 (22.2%) reporting one outbreak and 29 (64.5%) reporting more than one outbreak (Supplementary Table 1). The six (13.3%) LTCFs reporting sporadic SARS-CoV-2 positive cases included four old age homes and two retirement villages (Supplementary Figure a). There were 48 (66.7%) small outbreaks and 24 (33.3%) large outbreaks reported (Supplementary Figure b). The 24 (61.5%) LTCFs that reported large SARS-CoV-2 outbreaks included 8 (33.3%) retirement villages, 6 (25.0%) psychiatric facilities, 5 (20.8%) old age homes, 4 (16.7%) substance abuse recovery centres and 1 (4.0%) frail care centre (Supplementary Figure c). There were 30 outbreaks reported in wave 1, 21 in wave 2 and 15 in wave 3, while 6 outbreaks were reported between waves.

The median age of COVID-19 cases among residents was 56 years (IQR 38–73) and 812/1504 (54.0%) were male (Table 1). Of the 1501 (99.8%) residents for whom race was known, 734 (49.0%) were Black African, 610 (40.6%) were White, 101 (6.7%) were Coloured and 56 (3.7%) were Indian. Among 1473 (97.9%) residents for whom there were data on comorbidities, 274 (18.6%) had comorbid conditions. Of these, 191 (69.7%) had one comorbid condition, 60 (21.9%) had two comorbid conditions and 23 (8.3%) had three or more comorbid conditions. The most common comorbid conditions among residents were hypertension (194, 15.2%), diabetes mellitus (64, 4.5%) and chronic cardiac disease (48, 3.4%). Of the 1504 SARS-CoV-2 positive residents with outcomes, 1308 (87.0%) recovered, 59 (3.9%) were active cases and 137 (9.1%) had died, giving a case fatality ratio (CFR) of 9.1%. The CFR excluding active cases amongst residents was 137/1445 (9.5%).

The median age of COVID-19 admissions amongst staff was 42 years (IQR 35–51) and 705/820 (86.0%) were female (Table 1). Of the 767 (93.5%) staff for whom race was known, 640 (83.4%) were Black African, 79 (10.3%) were Coloured, 8 (1.0%) were Indian and 40 (5.2%) were White. Among 793 (96.7%) staff for whom there were data on comorbidities, 104 (13.1%) had comorbid conditions. Of these, 82 (78.8%) had one comorbid condition, 20 (18.2%) had two and three (2.9%) had three or more comorbid conditions. The most common comorbid conditions among staff were hypertension 69 (8.7%), diabetes mellitus 28 (3.5%), HIV 15 (1.9%) and asthma 14 (1.8%). Of the 820 SARS-CoV-2 positive staff with outcomes, 762 (92.9%) cases recovered, 54 (6.6%) were active cases and 4 (0.5%) had died, giving a case fatality ratio (CFR) of 0.5%. The CFR excluding active cases amongst staff was 4/766 (0.5%).

The number of SARS-CoV-2 deaths and CFR among residents in the various epidemic periods were 22/115 (19.1%) in pre-wave 1; 72/793 (9.1%) during wave 1; 5/60 (8.3%) during post-wave 1; 23/241 (9.5%) during wave 2, 4/18 (22.2%) during post-wave 2 and 16/218 (7.3%) during wave 3. (Fig. 3).

The number of SARS-CoV-2 deaths and CFR among staff in the various pandemic periods were 1/91 (1.1%) during pre-wave 1; 2/438 (0.5%) during wave 1; and 1/112 (0.9%). No deaths occured in satff during post-wave 1, post-wave 2 and wave 3.

Among residents the CFR was 46/252 (18.3%) in old age homes, 15/112 (13.4%) in frail care centres; 21/389 (10.5%) in retirement villages; 47/649 (7.2%) in psychiatric facilities; and 9/232 (3.9%) in substance abuse recovery centres.

Of the four staff (0.5%) who died, two each were female and male. The median age was 48.5 years (IQR 44.5–57.0). One female had hypertension and diabetes mellitus. The remaining three individuals who died did not have any reported comorbidities. Each of the four deaths occurred in an old age home, psychiatric facility, retirement village and a substance abuse recovery centre.

On multivariable analysis, factors associated with SARS-CoV-2 mortality among LTCF residents were age 40–59 years (aOR 2.4, 95% CI 1.0–5.5), 60–79 years (aOR 7.6, 95% CI 3.6–16.5) and ≥ 80 years (aOR 18.2, 95% CI 8.1–41.2) compared to < 40 years; and being a resident in a LTCF in Free State (aOR 4.2, 95% CI 1.4–12.8) or Northern Cape (aOR 3.8, 95% CI 1.2–12.6) compared to Western Cape (Table 2). Compared to pre-wave 1, there was a decreased risk of mortality in wave 1 (aOR 0.3, 95% CI 0.1–0.9), post-wave 1 (aOR 0.2, 95% CI 0.05–0.8), wave 2 (aOR 0.3, 95% CI 0.1–1.0), post-wave 2 (aOR 0.1, 95% CI 0.02–0.9) and wave 3 (aOR 0.2, 95% CI 0.006–0.6).

Table 2 Multivariable analysis of factors associated with SARS-CoV-2 mortality among LTCF residents, South Africa, 5 March 2020–31 July 2021 (n = 1504).

We could not find any other studies in Africa describing SARS-CoV-2 in LTCFs. DATCOV is a sentinel surveillance system that has only a small number of reporting sites for LTCFs which may not be generalizable across the country, but it does include sites in most provinces and across different LTCF types. There is some incomplete data in the surveillance system, particularly that of smoking (1348/2324, 58.0%). These variables have therefore been excluded from the multivariable analysis.