Research objectives: The aim of this project is to compare the pandemic response to COVID-19 in locations of Brazil, Canada, China, France, Japan and Mali during the first and second waves of the pandemic. This comparison will be used to identify strengths and weaknesses in the response, including challenges for health professionals and health systems. The research questions are:
Q1. How was the response planned, organized, and implemented in different COVID-19 referral hospitals?
Q2. What disruptions were encountered, what strategies were adopted and what is the resiliency of professionals and hospital?
Q3. How were social and health inequalities considered in the design and planning of public health interventions to prevent the spread of SARS-CoV-2?
Q4. What collective and practical lessons learned from the COVID-19 crisis that can be developed for better preparation and response in the future?
Q5. What are the factors that facilitated or hindered the dissemination and the use of these lessons learned between the countries?
Q6. How does the COVID-19 burden differ between each country and what are the similarities and differences in spatial and temporal trends?
Research design: Multiple case study approach
In the field of health systems research, comparative approaches are recommended  and are essential to develop operational, transferable lessons. We will use a multiple case study approach with multiple levels of nested analysis . Each hospital and public health intervention will be considered a single case.
For the hospital case studies (Q1 and Q2), the analysis will correspond to varying importance of different configurations. The configurations will be identified using a comparative perspective based on the conceptual framework (Fig. 1) . We chose these six countries as they represent the diversity of continents, contexts, and COVID-19 burden and we have longstanding scientific and practice collaborations. We will focus on eight major hospitals in Recife and Manaus (Brasil), Zhejiang (China), Paris (France), Bamako, (Mali), Montreal and Laval (Canada), Tokyo (Japan), see Fig. 1.
For the public health case studies (Q3), we have chosen to focus on understanding if and how inequalities have been taken into consideration during the planning of two major SARS-CoV-2 infection prevention interventions: contact tracing and testing (general population). Each intervention at each site is considered a case study.
For Q4, to generate high quality lessons learned (LL), they will be derived from a systematic approach to collecting, compiling, and analyzing data from multiple sources, and reflecting both positive and negative intervention experiences . To develop this process of LL, we conducted a rapid review of the literature which led to a ten steps guide: 1. Identification and mobilization of stakeholders; 2. Formulation of the aims of the process; 3. Identification of the events targeted to develop the LL; 4. Choosing the moment to start the process of developing the LL ; 5. Select the methods; 6. Develop interview grids; 7. Choose data source; 8. Data verification and revisions of the aspects to be covered ;9. Analysis and formulation of preliminary LL; 10. Verification of the quality of LL.
To share the research results and validate the preliminary LL (step 10), we will organize one deliberative workshop  in each country and one international deliberative workshop between the six countries with national institutions and international organizations (WHO, GloPID-R, PAHO, AFRO, TDR, PHAC, European/African CDC). The aim of the workshops is to discuss the practical implications of the various findings and recommendations, in terms of preparation, interventions, training, and communication. The workshops will be supported by preliminary drafts of policy briefs (PB)  to share the research results that led to the preliminary LL, in an accessible format and in multiple languages. This is part of our an action-oriented approach for decision-makers. Other knowledge transfer (KT) tools (infographics, videos, etc.) will be developed to disseminate these lessons to different audiences, once the LL have been finalized. The project's website will be used for information dissemination and communication (https://u-paris.fr/hospicovid) in English and French.
To evaluate these KT activities (Q5), a mixed-method design, combining quantitative and qualitative data, will be used.
For Q6, subnational (e.g., provincial, district, or department) COVID-19 portraits will be constructed for each site which will correspond, geographically, to the selected hospitals and public health organizations of Q1-Q3.
For Q1 and Q2, we will describe how countries have planned, organized, and implemented hospital responses to COVID-19, to describe the resilience of hospitals and their staff. Several empirical data collection techniques will be used (observation, interviews, document analysis,). For the observations, the researchers will conduct lengthy observation sessions, when it is safe to do so, over several weeks in some of the hospitals. The aim is to observe the functioning of services, meetings, interaction between professionals, etc. While these sessions provide empirical data through systematic note-taking , they will also be instrumental for the interviews and developing the interview guide. Qualitative interviews will be conducted with stakeholders using a diversification sampling strategy  within each stakeholder group (decision-makers, managers, medical staff, non-medical staff). We anticipate conducting approximately 30 interviews per site/institution, until empirical saturation is reached. The conceptual framework will inform the development of interview guides, which are discussed below. Interview guides will be developed collaboratively and piloted in each jurisdiction prior to use.
For Q3, the two public health measures, contact tracing and SARS-CoV-2 testing, will first be described using the TIDier-PHP guidelines for each site. The descriptions will inform the interview guide as well as the conceptual dimensions of the REFLEX-ISS tool, which enables stakeholders to question the way in which social and health inequalities are taken into account in their interventions. The interview guides will be drafted and tested prior to use for qualitative interviews. These interviews will also be conducted with stakeholders using a diversification sampling strategy  within each stakeholder group (decision-makers, managers, public health practitioners). We anticipate conducting approximately 30 interviews per site/institution, until saturation is reached.
For Q4 and Q5, two questionnaires will be used i) to assess the policy briefs and their use, and ii) to assess the participants' intention to use the lessons learned. This questionnaire adapted from the tool developed by Légaré et al. is based on the Theory of Planned Behavior  and Triandis’ theory. Approximately 30 semi-structured interviews will be conducted on-line three months after the national workshops. All members of the research team and at least three key-informant in each country will be invited to participate. The interview grid is partly constructed according to the essential organizational components of a deliberative workshop as formulated by Boyko et al..
For Q6, publicly available COVID-19 disease surveillance data will be collated for each site along with other data sources such as Census (e.g., population, sociodemographics) and mapping files.
Our empirical research for Q1 and Q2 will be supported by an original analytical framework on health system resilience (Box 1).
Box 1 : Health system resilience definition
|The capacities of a health system faced with shocks, challenges/stress or destabilizing chronic tensions (unexpected or expected, sudden or subtle, internal or external to the system), to absorb, adapt and/or transform in order to maintain and/or improve universal access to comprehensive, relevant and quality health care and services without pushing patients into poverty.
We have incorporated an analytical framework from the UK Department For International Development  as well as aspects of another conceptual framework that emphasizes the importance of interactions between the health system and the population in achieving access to care [36–39]. The rationale for our framework (Fig. 2) is that we need to first understand the disruptions encountered by COVID-19, by addressing the question "resilience to what?". The COVID-19 pandemic is at the core of our analysis and it represents a series of shocks to the system (1.). All types of events (or “situations” to be dealt with) are possible, such as sudden shocks (e.g., COVID-19 pandemic), unique stresses and challenges (e.g. staff and inputs availability) and chronic stresses (e.g. drug shortages). Second, we wish to answer the question "resilience of what?". We wish to uncover the effects (positive or negative) of these events, then the strategies deployed to deal with them (by describing them and explaining their rationale) as well as their impacts on organisational routines and system dimensions (2.). While the 10 dimensions, at the center of the figure (e.g governance, human resource, logistic), have been identified in a scoping review, our empirical analysis will be adapted to each hospital’s context. . It is important to understand health system “resilience processes” at the individual/team level. We will focus on how managers, health and non-health staff have mobilized resources to respond to the disruptions in the hospital (1.). Third, following our definition of resilience (Box 1), we need to understand how the mobilization (or not) of these strategies in the face of events that influences access to health care (3.) and in particular the five health systems abilities (approachability; acceptability; availability; affordability; appropriateness) of access to care. We will distinguish between these five dimensions for COVID-19 patients from other patients present in hospitals. We recognize that one of the limitations of our research, due to a lack of resources, is the difficulty of taking into account the other five demand dimensions of access to care (approachability, acceptability, availability, affordability, appropriateness). Finally, we will examine the combined impacts of these different resilience processes (absorption, adaptation, transformation) and outcomes (improvement, recover, deterioration, collapse) regarding four different scenarios of hospital resilience (4.).
For Q3, we will use the conceptual reflections that guided and informed the development of the REFLEX-ISS tool , which are the following: a) What is the philosophy of social inequalities in health (SIH) in the context at hand? Is there a shared vision of SIH (i.e., a shared vision for the respondent, its institution, and its partners) based on a common analysis of the context and supported by evidence?; b) Was the intervention planning done in consultation with major stakeholders, including community-based organizations (and ensuring their ongoing commitment) and to what extent does the consultation process effectively enact the intersectoral approach at the institutional level (e.g., formalization of consultation spaces, e.g., setting up partnerships and steering committees) vs. only at the individual level (e.g. mobilization of one’s own network of contacts)?
All interviews (for Q1, Q2 and Q3) will be transcribed and coded using computer-assisted (or aided) qualitative data processing software guided by the approach and principles of framework analysis, i.e., using a deductive-inductive approach to coding. 
From data collected in the hospitals (i.e., interview transcripts and observation notes), descriptive accounts of hospitals’ configurations (10 to 15 for each hospital) will allow us to highlight how adapted dimensions of the framework below are revealed through empirical data. We will uncover the recurrence, according to the specific and historical contexts of hospitals, of configurations between problem situations, the effects on organisational routines, and the strategies deployed to deal with them and the impacts caused (Fig. 3). These configurations will be organised with regard to the dimensions usually analysed in the resilience of health systems using a comparative perspective.  The three dimensions of processes from a resilience perspective (effects/strategies/impacts) could give rise to the existence of three types of configurations, for example: reactions (effects/strategies/impacts), anticipations (strategies/impacts before any effects) or inactions (effects but no strategies).
The analytical approach of the multiple case studies will be carried out in two stages. In the first stage, an intra-case analysis will be carried out for each hospital in the study. This analysis will be global and exploratory using a framework analysis approach (Fig. 2).  Throughout the analysis, we will crosscheck and validate the data collected and interpretations with key stakeholders involved in the response. A case study report will be written for each hospital following the plan in Appendix 1. In addition, we will conduct several scoping reviews on the resilience of health systems during the COVID-19 pandemic to ensure that both the historical contexts and up-to-date evidence are incorporated into our analyses. The protocols will available online at (https://www.protocols.io). The second stage of the analysis will be a comparative analysis between the cases, using the configurations as a heuristic tool, and to generalize the analysis (Fig. 4).
Based on findings at the hospital level and using our resilience framework (Fig. 2), we will then synthesize the results to make generalizations similar situations in other locations . This will inform a middle-range theory on the resilience of health care systems and hospitals, i.e. “theories that lie between the minor but necessary working hypotheses that evolve in abundance during day-to-day research and the all-inclusive systematic efforts to develop a unified theory” . The goal is to analyze whether configuration patterns from the different case studies provide the same or different understanding of hospital resilience processes and outcomes. In other words, we will attempt to identify consistencies in the processes and configurations of hospital resilience in the context of a pandemic.
For public health analyses (Q3), we will use the conceptual reflections that guided and/or informed the development of the REFLE-ISS tool .
For Q4 and Q5, due to the small number of workshop participants, quantitative data from the two questionnaires will be subject to descriptive statistics, such as mean and standard deviation. This will provide a descriptive understanding of the reactions of the participants following their participation in the deliberative workshop as well as their intention to use the knowledge and LL. For the semi-structured interviews, all interviews will be recorded after the consent of the interviewee and then fully transcribed. The transcribed data will then be coded using QDA Miner software and then analyzed using a content analysis . During the content analysis, we will seek to understand general trends and discrepancies with an emphasis on comparison between different stakeholders. Triangulation of the results of the qualitative and quantitative analyses will be carried out through a triangulation-convergence approach , which involves comparing and contrasting the same object from both sources of data to increase the richness of the interpretation and conclusions.
For Q6, epidemiologic curves will be constructed for each site as will COVID-19 burden maps at the highest spatial resolution possible (e.g., neighbourhoods). We will create descriptive tables that will include the characteristics of COVID-19 (when possible) as well as timelines of the sequence of events and measures for each site.