Article Text

Original research
COVID-19 in French nursing homes during the second pandemic wave: a mixed-methods cross-sectional study
  1. Morgane Dujmovic1,
  2. Thomas Roederer1,
  3. Severine Frison1,
  4. Carla Melki2,
  5. Thomas Lauvin2,
  6. Emmanuel Grellety1
  1. 1Department of Epidemiology and Training, Epicentre, Paris, France
  2. 2Emergency Cell, Médecins Sans Frontières, Paris, France
  1. Correspondence to Thomas Roederer; thomas.roederer{at}


Introduction French nursing homes were deeply affected by the first wave of the COVID-19 pandemic, with 38% of all residents infected and 5% dying. Yet, little was done to prepare these facilities for the second pandemic wave, and subsequent outbreak response strategies largely duplicated what had been done in the spring of 2020, regardless of the unique needs of the care home environment.

Methods A cross-sectional, mixed-methods study using a retrospective, quantitative data from residents of 14 nursing homes between November 2020 and mid-January 2021. Four facilities were purposively selected as qualitative study sites for additional in-person, in-depth interviews in January and February 2021.

Results The average attack rate in the 14 participating nursing facilities was 39% among staff and 61% among residents. One-fifth (20) of infected residents ultimately died from COVID-19 and its complications. Failure to thrive syndrome (FTTS) was diagnosed in 23% of COVID-19-positive residents. Those at highest risk of death were men (HR=1.78; 95% CI: 1.18 to 2.70; p=0.006), with FTTS (HR=4.04; 95% CI: 1.93 to 8.48; p<0.001) or in facilities with delayed implementation of universal FFP2 masking policies (HR=1.05; 95% CI: 1.02 to 1.07; p<0.001). The lowest mortality was found in residents of facilities with a partial (HR=0.30; 95% CI: 0.18 to 0.51; p<0.001) or full-time physician on staff (HR=0.20; 95% CI: 0.08 to 0.53; p=0.001). Significant themes emerging from qualitative analysis centred on (1) the structural, chronic neglect of nursing homes, (2) the negative effects of the top-down, bureaucratic nature of COVID-19 crisis response, and (3) the counterproductive effects of lockdowns on both residents and staff.

Conclusion Despite high resident mortality during the first pandemic wave, French nursing homes were ill-prepared for the second, with risk factors (especially staffing, lack of medical support, isolation/quarantine policy, etc) that affected case fatality and residents’ and caregivers’ overall well-being and mental health.

  • COVID-19
  • Public health
  • Epidemiology

Data availability statement

Data are available upon reasonable request. Anonymised data collected for the study and a data dictionary will be made available to other researchers following approval of a study proposal by TR ( for 5 years from publication. The study protocol, statistical analysis plan and informed consent forms are also available from TR.

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:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Our study is one of the first mixed-methods investigations of nursing homes during the COVID-19 pandemic in Europe, reporting face-to-face interviews of residents themselves, in contrast to most other qualitative investigations of the geriatric population during the COVID-19 period, which have usually been conducted remotely or via surrogates (caregiving staff or family members).

  • Our study is also the first in the world to describe the second wave of the pandemic in this setting, both quantitatively and qualitatively.

  • We report in-depth quantitative data analysis of 585 COVID-19 cases from 14 nursing homes while 47 qualitative interviews were conducted in person, from December 2020 to February 2021.

  • Study site selection was not random, thus, comparing the included facilities to others in Provence and Occitania (or France) should be made with care.

  • Moreover, only residents who were fully capable of interacting with investigators and were able to give informed consent could be interviewed, thus excluding anyone with major cognitive disorders (a relatively frequent condition in nursing homes).


In France, state-funded nursing and care homes are the most common living arrangement for both independent seniors and those who need daily care and support. These institutions were deeply affected by the first wave of the COVID-19 pandemic, with an estimated 38% of all residents (247 000 cases) infected with SARS-CoV-2 and 5% (30 395) succumbing to the disease from March to July 2020. The workforce that staffs these facilities was also seriously affected, with an estimated 22% of all workers (90 000 cases) testing COVID-19 positive from late February to late May 2020.1 2

In October of 2020, when rising caseloads suggested a second pandemic wave, nursing homes again braced for the worst, since no vaccine was yet approved in France (this occurred in December 2020) and some variants had begun circulating. In November of that year, the non-governmental organisation Médecins Sans Frontières (MSF) began partnering with select nursing homes in Provence and Occitania provinces, in southern France, to bolster their COVID-19 prevention and care procedures in the midst of rapidly growing medical needs, strained facilities and understaffing (often aggravated by absenteeism spurred by workplace-acquired infections). As nursing homes transformed into places providing hospital-level care, staff were required to perform more advanced technical procedures and increased disease surveillance at a moment when human resources were depleted due to illness and overwork. Concurrently, health authorities recommended strong lockdown measures for elderly care home residents, including bans on going outside, prohibiting family visits and confining residents to their rooms.

Despite the devastating mortality rates seen in care homes around the world throughout the pandemic, scientific literature has not yet described the second wave of COVID-19 in this environment. Published research is mostly focused on the first pandemic wave period, almost exclusively on quantitative studies or systematic reviews on specific topics. Several articles report best practices for infection prevention and control (ie, frequent testing for staff, residents and visitors, staff cohorting and strict isolation policies), or recommended better evaluation of the consequences of lockdown restrictions.3–15 Other lessons from the initial crisis period were that more staff,6 8 support,8 9 protective equipment and overall preparation8–10 could prevent or reduce outbreaks. Lately, articles focused only on the impact of vaccination on transmission among staff and residents.16 17 The little qualitative research conducted during the first wave was rarely able to conduct in-person interviews,12 18–24 but found that lockdowns had a significant and deleterious impact on residents, staff well-being and staff turnover.20 21

Our research attempts to understand the risk factors that influenced the second pandemic wave, the impact of that wave, and how staff and residents experienced this period of the pandemic in a nursing home setting.


In this mixed-methods, cross-sectional study, we analyse retrospective COVID-19 data from 14 nursing homes being reinforced by support from MSF to assess the impact of the second pandemic wave as well as the effects of prevention measures on resident mortality and comorbidity. These results are given depth and detail through a qualitative investigation into staff and resident experiences. Quantitative data were collected, cleaned and primarily analysed by a senior field epidemiologist who joined the MSF team for 3 months, while the qualitative survey was conducted by a social geographer working alongside MSF in nursing homes from December 2020 to March 2021.


Autonomy Evaluation Score (AES) measures a care home resident’s level of autonomy. An AES of 1 reflects the lowest level of autonomy (ie, confinement to a bed or armchair, serious mental function impairment, continuous caregiving required), while an AES of 6 refers to people who have fully retained their autonomy in their daily lives. The Average Weighted Autonomy Score (AWAS) is the overall AES score for a facility. This score is a proxy for the financial and human resources that a nursing home needs and has access to: the higher the AWAS, the more resources needed (staff-to-residents ratio, equipment, etc) and the more dependent the residents. AES and AWAS are mandatory metrics required by French authorities to allocate funds and evaluate nursing homes’ needs (further details and references in online supplemental appendix 1).

Geriatric failure to thrive syndrome (FTTS): specific to old age, this syndrome is defined by the rapid deterioration of the general state with anorexia, disorientation and social withdrawal, alongside a more or less directly expressed will to die, a passive giving up on life, an active refusal of care and/or food. It usually evolves towards death in a few days to a few weeks (80% of cases). It is triggered by physical events (acute illnesses, surgery, trauma) or psychological events (death of a loved one, social isolation, hospitalisation) (further details and references are in the online supplemental appendix 1).

Death (outcome for quantitative analysis): death as an outcome in the linelists is either a resident’s death directly linked to COVID-19 or death while the resident was a confirmed COVID-19 case. Death was validated by the nursing home coordinating physician and recorded in the nursing home registries as well as in the online national COVID-19 database put in place during the crisis by the French Ministry of Health.

Study design and population

This cross-sectional, mixed-methods study used a fixed convergent design.25 The use of qualitative and quantitative methods was predetermined: the procedures for collecting and merging data were planned at the start of the study according to the identified problem. Quantitative and qualitative data were simultaneously collected during fieldwork, separately analysed and then brought together for interpretation. However, the interaction between qualitative and quantitative components occurred during study implementation. Data were integrated through data transformation (codebook of qualitative findings), multidisciplinary team discussions and comparative writing (online supplemental appendix 2). The study used a unique dataset that was made accessible thanks to the operational role of MSF in the field. The study used quantitative data from residents living in 14 nursing homes from November 2020 to mid-January 2021. The 14 nursing homes were not randomly selected but retained for analysis if they could provide a full COVID-19 linelist (out of the 22 facilities that MSF supported during this period). Four nursing facilities were purposively selected as qualitative study sites for additional in-person, in-depth interviews (IDIs) conducted between January and February 2021. Qualitative study sites were selected based on whether they had passed their epidemic peak, had high attack and fatality rates, were public or private facilities, and their geographical location.

Data collection

Administrative data about the facilities (number of beds and staff, job categories, staff-to-resident ratios, AWAS, resident mean age, etc) and COVID-19-related data at the facility level (dates and magnitude of COVID-19 outbreaks, confirmed cases among residents and staff, attack rates, episode duration, number of deaths, resident fatality ratios, etc) were retrieved by the field epidemiologist from nursing home managers. Individual, anonymised, COVID-19 case data gathered into linelists (age, sex, AES, date of COVID-19 positive confirmation, outcome, date of death, date of transfer to hospital, oxygen therapy, palliative care, comorbidities such as dementia, neurodegenerative diseases, diabetes, cancer, cardiovascular diseases, etc) were collected by the field epidemiologist. Sources of these data were residents’ electronic records and registries maintained by head nurses and coordinating physicians. Diagnoses of all comorbidities followed national guidelines and were operationalised by nursing home clinicians and coordinating physicians. Diagnoses were then recorded in resident registries and transferred to the linelists by the field epidemiologist with the help of the head nurses. Facilities and linelists data were used for quantitative analysis. Qualitative data were gathered using semistructured, IDIs during 1-week ethnographic immersions in each of the four qualitative study sites. The lead investigator targeted four groups of actors, including facility administrators (directors, coordinating physicians and nurses), clinical and facilities staff (nurses, caregivers, educators, physical therapists, maintenance crews), the residents themselves and the residents’ visiting family members. Participants were purposively selected to obtain a maximally heterogeneous sample of interview participants and reflect the spectrum of opinions and experiences of everyday life in nursing homes. Across the 4 qualitative study sites, a total of 47 IDIs were conducted with facility directors (4), staff members (36) and residents (7). Among the 36 staff members, 29 were caregivers and 7 provided other support functions (human resources, maintenance, cleaning, cooking). All interviewed residents were women, as were the majority of study participants overall (82.9%). Interview length varied from 12 to 171 min (54 min average) (online supplemental appendix 3).

Telephone and face-to-face interviews were also conducted with 10 residents’ family members, though family interviews are not included here to focus on experiences from within the nursing homes during the lockdown. Nine residents refused to participate (due to fatigue, discomfort with interviewing or COVID-19-related reasons). Caregiver participation was constrained by understaffing, overwork, fatigue or disease, which left them with very little time or energy for interviewing.

Vulnerable residents were preselected under the advisement of the coordinating nurse on the permanent caregiver teams. Participants had to be able to give informed consent, capably interact and have no major cognitive disorders. The level of autonomy (AES) did not constitute an a priori criteria for participant selection. Whenever a legal guardian or curator was designated, the latter was contacted before the interview to verify that consent could be obtained from the interviewee.

Question guides focused on three primary topics: the outbreak chronology, adaptation to the crisis and the individual experience of the second pandemic wave (online supplemental appendix 4). Individual guides were adapted for those living in the nursing home (residents) or working there (facility administrators and staff). All interviews were voice recorded and direct observations were written in the investigator’s field book. All written data were anonymised upon collection. Participants’ data were assigned a study number that was set on a correspondence table kept separately from other data. Written informed consent was obtained prior to interview.

Preventive measures were implemented with all participants to decrease COVID-19 transmission risk: systematic FFP2 face mask use, social distancing, hand and space disinfection, and weekly Realtime-PCR tests for the two field investigators.

Statistical analysis

Patient data were explored using univariate analysis to highlight possible mortality risks. Univariate unadjusted Cox HRs, Kaplan-Meier estimations and log-rank tests were used for multivariate analysis. A stepwise procedure was followed, retaining factors with a log-rank test value of <0.3. COVID-19 mortality was estimated using a multilevel mixed-effects Cox model using selected factors identified in the univariate analysis. Random effects on individual variables were considered and nested at the facility level.26 Interactions between potentially correlated factors (comorbidities, FTTS, autonomy level, time-related variables) were accounted for while robust SEs were computed (online supplemental appendix 2). The 95% CIs are presented and a significance threshold of 5% was chosen for p values. Statistical analyses were conducted with Stata V.15 and R Studio V.1.4.

Qualitative analysis

Data analysis was performed from January to March 2021, similar to the fieldwork period (January–February) and reporting phase (March–April). The qualitative analysis combined grounded theory and hypothetico-deductive analysis. Preliminary observation in five nursing homes and MSF team reports were used to create an initial checklist for systematic direct observation. In January and February 2021, 36 semistructured IDIs were conducted in three nursing homes, in combination with ‘external participatory observation’.27 Questions were adjusted iteratively after preliminary analysis was conducted on these initial interviews. Data saturation was sought throughout the interview process and discussed within the research every week. In February 2021, 11 semistructured IDIs were conducted in a fourth nursing home to assess data saturation.

Interview data were processed gradually through professional transcription and verified with the interviewees when necessary. De-identification occurred during transcription (names, places, dates, distinctive personal data, etc). Interview data were written, analysed and coded in Excel spreadsheets. The first codebook with 39 data codes emerged from interview transcripts. Five themes were initially analysed and refined into a final set of 33 across four key categories. Three of these were cross-cutting and had up to three subthemes (table 3). Results are reported following the Standards for Reporting Qualitative Research guidelines28 and the COnsolidated criteria for REporting Qualitative research checklist.

Patient and public involvement

Administrators and coordinating physicians from 14 nursing homes were actively involved in collecting and anonymising study data from their residents/patients. During the exploratory phase of research (December 2020–January 2021), any feedback from qualitative study site administrators was included in the study protocol. During data collection (January–February 2021), the research methodology was discussed with MSF nurses and facilities staff and adapted to each nursing home’s context and caregiver guidance. At the beginning of each IDI, caregivers and residents were encouraged to further participate in the research by contacting the lead investigator with any suggestions. In the reporting phase (from the 1 March to June 2021), internal reporting was sent to interviewees who wanted to be contacted for this purpose. This report was sent to prominent political COVID-19 crisis management actors (such as the French Ministry of Health). A summary letter will be brought to resident study participants and facility staff to inform them of the results and gather their comments on possible follow-up.


Twenty-two nursing homes were originally included in the study, though data were available for only 14 of them (the others did not send data in time for analysis or the data were not electronically recorded). The 14 participating nursing facilities were largely state-supported entities (79%) with an average of 68 residents (median=65; IQR: 58–73). Results varied considerably from one nursing home to another. COVID-19 outbreak duration averaged 39 days (median=40; IQR: 30–50 days), while infected residents’ individual COVID-19 episodes averaged 24 days (median=30; IQR: 14–51 days). The average attack rate was 39% (median=39%; IQR: 29%–54%) among staff and 61% (median=60%; 50%–73%) among residents. One-fifth (median=20%; IQR: 17%–23%) of the residents who were infected ultimately succumbed to COVID-19 and its complications. The mean AWAS was 770 (median=763 ; IQR: 722–804) and the average staff-to-resident ratio was 0.82 (median=0.86 ; IQR: 0.72–0.90). The average time to universal masking policies being implemented was 9.6 days (median=6.5; IQR: 2–15 days) and the average time until a facility was bolstered with MSF support (staff or resources) was 17.5 days (median=15; IQR: 13–28 days) (online supplemental appendix 5).

Patient risk factors

Retrospective COVID-19 data were obtained for 14 nursing homes, finding 585 COVID-19 cases among 930 residents (61% attack rate) (table 1). Cases were mostly women (78%) who were >85 years old (68%). Individual autonomy scores (IAS) were low (<2) in a majority of cases (60%), indicating a very low level of autonomy overall. One-fifth (21%) of cases were transferred to a hospital, while half (46%) were put on oxygen therapy. One-tenth (12%) of COVID-19 cases received palliative care, and nearly one-quarter (22%) died. FTTS was diagnosed in nearly one-quarter (23%) of COVID-19-positive residents. At least one other comorbidity was found in over half (61%) of infected residents. AWAS, nursing home size and staff-to-resident ratios were all strongly correlated, as were time-related variables (time until external MSF support was received, time until universal masking policies were applied and duration of COVID-19 episode) (table 1).

Table 1

Univariate analysis of nursing home resident and facility data, Provence and Occitania provinces, France, 2021

Univariate analysis using Cox modelling (table 1) and Kaplan-Meier estimations (figure 1) suggested that individual characteristics like gender (log-rank p<0.001) and IAS (p=0.008) were associated with COVID-19 mortality, while age and specific comorbidities were not. Survival curves also suggested that facility characteristics like low AWAS (p<0.001), the absence of a permanent physician on-site (<0.001), larger nursing home size (>70 residents) (p=0.036) and a high staff attack rate (p=0.025) were also associated with resident mortality. Predictably, hospitalisation (p<0.001), palliative care (p<0.001) and oxygen therapy (p<0.001) were all strongly correlated with the risk of death, as was the presence of FTTS (p<0.001) and the presence of more than four comorbidities (risk increased with the number of comorbidities present, p=0.045). Additional Kaplan-Meier curves for non-significant factors can be found in the online supplemental appendix 6.

Figure 1

Likelihood of survival by resident and nursing facility characteristic, univariate (Kaplan-Meier) analysis, Provence and Occitania provinces, France, 2021. On the x-axis: number of weeks from 15 October 2020; on the y-axis: the probability of resident survival. FTT, failure to thrive; MSF, Médecins Sans Frontières.

Multilevel Cox hazard modelling highlighted mortality-associated factors adjusted for potential confounders (figure 2). Those at highest risk of death were men (HR=1.78; 95% CI: 1.18 to 2.70; p=0.006), with an FTTS diagnosis (HR=4.04; 95% CI: 1.93 to 8.48; p<0.001) or in facilities with delayed implementation of universal masking policies (HR=1.05; 95% CI: 1.02 to 1.07; p<0.001). The lowest mortality risk was found in residents of facilities with a partial (HR=0.30; 95% CI: 0.18 to 0.51; p<0.001) or full-time physician on staff (HR=0.20; 95% CI: 0.08 to 0.53; p=0.001), with individual AES >3 (HR=0.38; 95% CI: 0.16 to 0.89; p=0.026). Noticeably, higher AWAS (a proxy for staff-to-resident ratios and a nursing home’s overall means) was associated with a lower risk of death (HR=0.99; 95% CI: 0.99 to 1.00; p=0.020) (table 2). Sensitivity analysis can be found in the online supplemental appendix 7.

Figure 2

Final Cox model: forest plot of mortality-associated factors in French nursing facilities, Provence and Occitania provinces, 2021. On the x-axis: adjusted HRs are represented by a diamond. Full lines in red for 95% CIs of significant risk factors (HR >1), full lines in green for protective factors (HR<1) and dashed lines in grey for 95% CI of non-significant factors. AWAS, Average Weighted Autonomy Score; FTT, failure to thrive.

Table 2

Multivariate Cox hazard-adjusted analysis of mortality-associated factors in French nursing facilities, Provence and Occitania provinces, 2021 (Akaike Information Criteria: AIC=1171; Bayesian Information Criteria: BIC=1226)

Qualitative results

The qualitative approach richly described interviewees’ lived experiences during the COVID-19 crisis, revealing difficult-to-quantify social influences on the outbreak’s evolution and impact. Three significant themes emerged from our discussions (table 3).

Table 3

Representative quotes for the three themes

Structural, chronic neglect of nursing homes

Staff members described a long-standing lack of physicians in nursing homes, exacerbated by lockdowns and growing medical needs during a period of rising COVID-19 infections. One nurse explained, “the nursing home was almost like a hospital ward at one point…There was more supervision [needed], more care…We didn’t have the staff to do all that.” All groups of interviewees emphasised that working in precarious and understaffed conditions was a substantial difficulty that became a critical risk during the COVID-19 outbreak and compromised the response. Assistant nurses described extremely challenging working conditions: “When they ask you to help 13 people to bath before noon, you don't work well.” This situation was worse during the second pandemic wave when, as one psychologist explained, “no one counted the hours. We had to be there, we put our private lives on hold, but it was important to do it.” All directors described a structural lack of a ‘permanent medical presence’ and the need for a ‘strict staffing ratio’.

Top-down crisis management

Personnel highlighted the ‘top-down’ approach of French health authorities, including a lack of communication and time-consuming processes for staff and administrators alike, “The ARS [Regional Health Authorities] have been absent during the whole crisis. (…) Since March, I haven’t seen the authorities giving us any support, nor any real help, except for claiming statistics back.” These officials worked far from the frontline environment of a nursing home and were removed from the suffering of residents and staff. As a result, it was felt that they encouraged ill-informed, unrealistic and inconsistent crisis response measures: limiting contact with residents, confining them to their (small) rooms, abruptly relocating them to new rooms (very disturbing for them) or even physically restraining residents in distress. A psychologist described how ‘some people had to be uprooted from their rooms’ where they had ‘spatial-temporal and autobiographical markers’, while others ‘had to be restrained’ by assistant nurses. All of these were deeply disheartening to staff and residents, creating feelings of shame and guilt among caregivers and the potential for cognitive disorders among residents. A resident explained that ‘it was hard, staying in the room for a whole day, without going out’, and that ‘anyone would become nuts!’ Weak crisis response mechanisms also manifested as poor prevention measures (a lack of universal masking requirements initially, facemask shortages during the first wave), lack of state medical relief staff and such an extreme lack of preparedness that assistance from a non-state humanitarian actor like MSF was needed. As a director told us, calling MSF, a disaster response organisation, ‘showed what a disaster we were experiencing’.

Counterproductive effects of lockdowns

Finally, participants described the counterproductive effects of lockdowns, including negative medical outcomes and even violence. Physiotherapists described ‘a decline in motor skills, but even more in cognitive skills’ and ‘completely accelerated FTTS’ which corroborates other descriptions of ‘bedridden patients, depressive states, failure to thrive’ because ‘the residents haven’t gone out for a year’. Participants were discouraged that lessons from the first pandemic wave did not translate into better preparedness and smoother, more nuanced, and less restrictive lockdown policies during the second. Despite feeling secure in their nursing home environment during the pandemic period, interviews with residents revealed the depth of their dislike for the extreme physical and social isolation they faced while alone in their rooms, especially when facilities’ social activities, family visits, and outings were suspended or strictly supervised with social distancing measures. Extreme fatigue occurred after a year of lockdown and social restrictions, as one nursing home’s 90-year-old resident explained: “if we could go out, we would bear it better.” Since facility administrators were urged to follow the ARS recommendations, only a few directors or staff were willing to soften lockdown measures, allow family visits, or take residents’ end-of-life wishes or needs for social interaction into account.

These interviews show some overlap with the risk factors that were highlighted in the quantitative data (mortality risks linked to understaffing, the absence of a permanent staff physician, low staff-to-resident ratios and lockdowns linked to FTTS). Other qualitative factors associated with better pandemic management also appeared in interviews, such as reliable communication with local health authorities, the presence of an effective national health strategy and collaboration with other medical sectors.


Our study is the first mixed-methods investigation of nursing homes during the COVID-19 pandemic in France, and one of the first in Europe. MSF staff’s close, in-person work with these care facilities gave investigators privileged access during a challenging period and led to particularly rich interviews. This lies in contrast to most other qualitative investigations of the geriatric population during the COVID-19 period, which have usually been conducted remotely or via surrogates (caregiving staff or family members), without being able to interview residents themselves. These results show clearly that the second wave looked largely similar to the first wave in French nursing homes, in both response and impact, and that these facilities were not sufficiently prepared and supported when facing subsequent threats to their vulnerable tenants.

Nursing home data are not routinely collected by French national health information services because residents are considered to ‘live at home’. Thus, considering how difficult it is to access even the most basic data from these facilities (such as the number of cases or deaths), we managed to construct a large dataset containing detailed information about COVID-19 cases, which affected 30% of all residents in the 14 participating nursing homes. The study also allowed a thorough examination of COVID-19 as experienced by the staff and residents who most suffered from the pandemic. To the best of our knowledge, French crisis management measures during the second pandemic wave were never informed by qualitative data. In this study, patients’ risk factors could be explored about influential social and structural determinants of health, such as understaffing, strict lockdown measures, isolation from other medical actors/lack of medical support, or the top-down and bureaucratic crisis management by health authorities.

Our multivariate analyses confirmed mortality trends seen in other settings. Similar to other studies, we found that men died more often despite being a minority of nursing home residents and that residents’ autonomy was a strong factor in their survival, with those who were more reliant on staff for daily support most likely to succumb to their disease.8–10 29–33 Living with multiple comorbidities (especially diabetes and dementia) was also strongly predictive of COVID-19 mortality in our group.8 10 29–33 The negative effects of understaffing (seen as sick leave or AWAS in our data) were similar to those reported in the USA,8 Spain33 and the UK,34 35 and constitute a vicious cycle: during periods of high transmission, more staff needed sick leave. Yet, the medical and staffing needs of residents were simultaneously surging, forcing many sick (and infectious) caregivers back into the workplace. The cycle was compounded by the destructive effects that an enormous workload and an anxiety-producing work environment are known to have on caregivers’ well-being.12 18 20 21 36 37

The efficacy of universal masking to prevent respiratory disease is well established,38 39 though we were not able to measure the impact of staff/resident masking because mask mandates were often put in place at the same time that extra resources and support from MSF arrived and bolstered the nursing facility overall. Nevertheless, our results do suggest that higher transmission and case fatality were associated with delays in mandatory mask requirements for staff, confirming the utility of these rules in uniquely vulnerable and high-risk nursing home settings. The facemask issue is not easy, however, in a nursing home context. The health benefits of masking have trade-offs with other social needs: care home residents may live with hearing or cognitive disorders, and masking may prevent voice and facial recognition or communication. The absence of others’ daily smiles or expressions may have led to cognitive decline, a point that has been shown in previous research and was emphasised in our interviews with caregivers, managers and residents alike.40 41

Finally, the benefit of confining residents to their rooms is strongly questioned by these results. While such measures undeniably reduce virus transmission among residents,6–10 14 15 34 38 39 42–44 the consequences for their mental health and nutritional status have also been shown to be considerable.12 13 20–24 37 45–49 Strict lockdowns in our cohort were associated with higher FTTS incidence, triggered by individuals’ difficult living conditions over multiple months (the long duration of the crisis, an anxiety-provoking atmosphere, social isolation, other residents’ deaths, etc). We found a strong statistical association between COVID-19 case fatality and FTTS diagnoses, a result that was triangulated by qualitative interview data and is consistent with other research from France,42 the UK,43 Finland,47 the USA,48 Spain49 50 and Italy.51


Our study is limited by the fact that study site selection was not random but was instead steered by discussions with MSF. Moreover, since MSF targeted mostly struggling nursing homes, the study included only a small number that did not have major outbreaks (or contained their outbreaks early). As a result, comparing these facilities to others in Provence and Occitania (or France) should be made with care. Participant selection was biased by the fact that only residents who were fully capable of interacting with investigators and were able to give informed consent could be interviewed, thus excluding anyone with major cognitive disorders (a relatively frequent condition in nursing homes). Quantitative data were neither exhaustive nor always electronically recorded. Associations between COVID-19 deaths and FTTS were complicated by the comorbidities that many residents also lived with, though adjusted analysis attempted to control for potential confounding.


These results raise questions about French health authorities’ approach to managing the second wave of the COVID-19 pandemic, as seen through the lens of those living through the crisis. If institutional management of older age, loss of autonomy and end of life is a chronic issue for a long time in France, solutions exist to support nursing homes in times of acute crisis. Future debates about a pandemic response in this setting should take into account things like the social needs of residents, understaffing as a risk factor for higher COVID-19-related deaths, and should refine general health policies and prevention measures in nursing homes.

Moreover, once an outbreak has occurred, tough questions must be asked: Are restrictive measures for all residents worth the personal and mental health toll? How can facilities improve residents’ end-of-life conditions in a controlled, safe way that will allow them (and their families) dignity and care? Is this reasonable to do if it involves a modicum of increased risk exposure for the facility overall? These results remind us that an effective COVID-19 response should be context adapted, patient centred and humane.

Data availability statement

Data are available upon reasonable request. Anonymised data collected for the study and a data dictionary will be made available to other researchers following approval of a study proposal by TR ( for 5 years from publication. The study protocol, statistical analysis plan and informed consent forms are also available from TR.

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and received approval from the MSF Ethical Review Board (ERB; ID 2703) and the Commission Nationale de l’Informatique et des Libertés (CNIL) in France. Patient data and qualitative observations were fully anonymised. All study procedures were in line with the Declaration of Helsinki. Participants gave informed consent to participate in the study before taking part.


First and foremost, the authors are very grateful and thank Janet Ousley for her help with article editing. The authors also thank Marie Thomas, Tommaso Fabbri, Klaudia Porten, Michel-Olivier Lacharité, Marc Gastelly-Etchegorry and the whole MSF team in the field. This study would not have been possible without the collaboration of the nursing home managers, staff and residents. A very special thanks go to each and every one of them.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Twitter @melkicarla

  • MD and TR contributed equally.

  • Contributors CM, TR, MD, TL and EG conceived the study (literature search, study design, etc). MD, SF, TR, CM, TL and EG developed the study protocol. MD performed field data collection (qualitative interviews) and SF collected epidemiological data. TR and SF performed data management and statistical data analysis. MD performed interview transcription and qualitative analysis. MD and TR performed the literature search and wrote the first version of the manuscript. TR and EG verified the underlying data and performed additional analyses. TR is the guarantor of the quantitative data and overall content of the protocol and article, MD is the guarantor of the qualitative content of the study. All authors interpreted the results, contributed to writing the manuscript and approved the final version for submission.

  • Funding This study was entirely funded by Médecins Sans Frontières, France (award/grant number is not applicable).

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the Methods section for further details.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.