Main

Pandemics have disrupted societies and impacted public health throughout human history4. Today, almost 3 years after SARS-CoV-2 was first identified and more than 1.5 years after the first vaccines became available, pandemic fatigue5 threatens to undercut our vigilance and the effectiveness of our responses to ongoing and new pandemic-related challenges. As of September 2022, more than 620 million cases of COVID-19 and over 6.5 million deaths have been reported6, although mortality estimates range as high as 20 million7,8. The healthcare for millions more people has been delayed, often as a result of overwhelmed health systems9,10,11,12. Highly transmissible variants continue to spread globally, while surveillance for variants of concern remains largely inadequate13,14,15. Reinfection risks are not fully understood. Low vaccination rates16 may compound the risk from waning immunity17,18. Long COVID has emerged as a serious chronic condition19,20,21 that represents a considerable burden of disease and still lacks adequate understanding and appropriate preventive or curative solutions. In addition to its direct health consequences, COVID-19 has disrupted economic activity, social interactions and political processes, affected civil liberties and interrupted education at all levels22,23,24,25,26. Although many governments and individuals no longer have the same level of concern as earlier in the pandemic27, many public health leaders, including members of this panel28, continue to regard COVID-19 as a persistent and dangerous health threat29,30,31.

Responses to the COVID-19 pandemic have been hindered by interrelated factors that include false information32, vaccine hesitancy33,34, inconsistent global coordination35, and the inequitable distribution of supplies36, vaccines37,38 and treatments39. Despite increased levels of trust in science during the pandemic23,40, there is information fatigue4 and waning compliance with those public health and social measures41,42,43 that remain in place, particularly those that affect daily lives44. Meanwhile, during periods of high community transmission, needs for services continue to exceed the capacity of many health systems45, which also are challenged by ongoing risks to the health of their workers46,47,48. Furthermore, long-standing social inequities have caused some populations to experience greater risk of COVID-19 infection, severe disease and death37. Many of these populations continue to have less access to COVID-19 vaccines37,49 and treatment39, as well as to resources to mitigate the mental health, social and economic consequences of the pandemic50,51,52.

Beneficial knowledge about COVID-19 aetiology, pathophysiology, prevention, vaccination, treatment and care has rapidly advanced through rigorous scientific, medical and public health inquiry, debate and collaboration53,54,55,56. Notwithstanding these advances, the responses of individual countries have been heterogeneous and often inadequate, in part because they lack coordination and clear goals.

To develop a global consensus regarding these ongoing problems, we carried out a Delphi study with a multidisciplinary, geographically diverse panel of 386 academic, health, non-governmental organization (NGO), government and other experts in COVID-19 response from 112 countries and territories (Table 1 and Methods). We achieved response rates of 85% in the second round (R2) and 82% and 81% in the third round (R3) surveys of the 41 statements and 57 recommendations, respectively. The mean levels of combined agreement (agree + somewhat agree) increased across the three rounds of the consensus statements (R1, 89%; R2, 90%; R3, 96%) and the two rounds of recommendations (R2, 93%; R3, 98%). The resulting consensus statements and recommendations (Fig. 1) can serve as a strong basis for decision-making to end COVID-19 as a public health threat, and permit a more durable resumption of social, cultural, religious, political, healthcare, economic and educational activities, with less burden on vulnerable populations.

Fig. 1: Delphi panel generation and data collection.
figure 1

Study methodology, including sample and data collection. Top, the iterative sampling approach used to generate a large, diverse Delphi panel (n = 386): four project co-chairs identified a core group of 40 academic, health, NGO, government and policy experts from 25 countries; the core group identified individuals with expertise in COVID-19; under-represented countries (that is, with fewer than one invitee) were identified and targeted through PubMed/Medline searches for authors of COVID-19 research studies in these countries. Bottom, the iterative digital data-collection process, including two survey rounds (R1 and R2) of draft statements; an online consensus meeting of the core group (Supplementary Discussion 3); one round of draft recommendations (R2); and a final survey round (R3) of the consensus statements and recommendations. Earlier rounds included text boxes for panellists to provide comments and suggest edits to individual statements (R1, R2) and recommendations (R2); the final statement and recommendations round (R3) allowed for overall comments at the end of each domain. For the final set of recommendations in R3, panellists ranked the top half in each of the six domains. RR, response rate.

Table 1 Expert panel characteristics (n = 386)

Top-ranked consensus recommendations

This multidisciplinary and multinational consensus study yielded 41 statements (Tables 2 and 3) and 57 forward-looking recommendations (Tables 47) on ending COVID-19 as a threat to public health grouped into six domains. Although we suggest that policymakers and other interested stakeholders review and consider the entire study findings, for expediency, we break out the top 10 recommendations ranked by the panellists in Table 8.

Table 2 Consensus statements to end COVID-19 as a public health threat
Table 3 Consensus statements to end COVID-19 as a public health threat
Table 4 Recommendations to end COVID-19 as a public health threat
Table 5 Recommendations to end COVID-19 as a public health threat
Table 6 Recommendations to end COVID-19 as a public health threat
Table 7 Recommendations to end COVID-19 as a public health threat

The top three recommendations focus on whole-of-society1 action and maintaining, or in some cases returning, to a vaccines-plus approach2. First, to avoid the inefficiency and ineffectiveness of fragmented efforts, pandemic preparedness and response should adopt a whole-of-society strategy that includes multiple disciplines, sectors and actors. Second, going forward, whole-of-government approaches (such as interministry coordination) can identify, review and address resilience in health systems to make them more responsive to people’s needs. Third, all countries should adopt a vaccines-plus approach, which includes a combination of COVID-19 vaccination, other prevention measures, treatment and financial incentives such as support measures. Infection rates tend to increase when governments discontinue social measures, including non-pharmaceutical interventions, regardless of the level of vaccination57,58.

The degree of consensus achieved for statements and recommendations, along with a ranking exercise in the final round, informed our synthesis of the study’s findings into six cross-cutting themes (Box 1) to which we believe decision-makers should pay particular attention: (1) SARS-CoV-2 is still present among us—despite some governments moving on—requiring continued efforts and resources to save lives; (2) vaccines are an effective tool against COVID-19 but will not alone end COVID-19 as a public health threat; (3) multisectoral collaboration that centres on communities and fosters trust is needed; (4) responsive health systems are crucial for responding to the COVID-19 pandemic and require coordinated government support; (5) adverse forces challenge efforts to end the COVID-19 public health threat; and (6) none of us is safe until everyone is safe. For ease of review, we report the tophalf ranked recommendations within each domain (Extended Data Fig. 1).

Areas of less agreement

The Delphi process involves a review and revision methodology that can result in relatively greater agreement among statements and recommendations over successive survey rounds while also identifying areas of disagreement that may require special efforts going forward. In addition to its the four-point Likert agreement–disagreement response options available in this study, panellists could select ‘not qualified to respond’ for items that they perceived as falling outside their expertise (see the ‘Delphi expert panel member sample’ section in the Methods). Although our study reflects relatively few areas of disagreement, we believe that highlighting the key areas of disagreement may be instructive for decision-makers in their own prioritization processes addressing the COVID-19 pandemic.

Extended Data Table 1 presents the six recommendations reflecting 5% or greater disagreement (disagree + somewhat disagree). Of those six, only two recommendations had greater than 10% disagreement: 18% of panellists disagreed with the recommendation to consider further economic incentives to potentially address vaccine hesitancy (REC3.6) and 11% disagreed with the recommendation that providers adopt a syndromic approach to COVID-19 diagnosis in settings with lower access to testing (REC2.18). The remaining four recommendations broadly relate to the use of governmental regulatory and enforcement powers in disease control efforts.

For statements and recommendations with response rates of ‘agree’ alone (that is, not combined with ‘somewhat agree’) below 67%, we conducted bivariate analyses to examine potential associations with panellist demographics; six statements (STMT1.2, STMT1.3, STMT2.1, STMT2.3, STMT3.5, STMT6.6) and one recommendation (REC4.5) demonstrated significant differences. Respondents who disagreed were significantly more likely to work in low- and middle-income countries than in high-income countries (P < 0.05; Supplementary Discussion 2). Few differences in agreement were identified by sector or field of employment, except for STMT1.1, for which greater disagreement was identified among those working in the health policy/advocacy field, and for STMT1.3, for which the academic and public sectors evidenced greater disagreement than other sectors.

Key statements and recommendations

The following six domains summarize the main areas of agreement, with a particular focus on the recommendations. The quantitative results on agreement and disagreement for the statements and recommendations are reflected in the tables and are further illustrated in Supplementary Discussion 1.

Communicate effectively

Substantial combined agreement among the panellists (range, 88–100%) indicates that communication issues remain a key area of risk and opportunity for ending COVID-19 as a public health threat. Policymakers and public health agencies should take special care when communicating the causation of and continuing accountability for the pandemic (Tables 2 (STMT1.7) and 4 (REC1.1)). The lowest level of agreement in this domain (agree, 57%; combined agreement, 88%) was found for a statement about government accountability receiving less attention when unvaccinated individuals are blamed for the pandemic’s continuation (Table 2 (STMT1.6)).

The panel focused primarily on the role of trust in government (Table 2 (STMT1.5)), the consequences of false information (Table 2 (STMT1.2, STMT1.3, STMT1.4)) and the rapid production of large volumes of new COVID-19-related information (Table 2 (STMT1.1)). That said, governments themselves may be a source of misinformation, for example, in the context of identifying transmission mechanisms (Table 6 (REC4.3)) and when stating that the COVID-19 pandemic has ended (Table 2 (STMT1.7)).

To counteract the infodemic and false information, governments should monitor false information (Table 4 (REC1.7)), expose networks of false information (Table 4 (REC1.9)) and consider holding publishers of false information liable (Table 4 (REC1.10)). Furthermore, public health professionals and other authorities should use clear, culturally responsive messaging to combat false information (Table 4 (REC1.3)). In parallel, social media companies should implement controls that reduce the publication and dissemination of false health information (Table 4 (REC1.8)).

Institutions and individuals should advance public trust by seeking training on building trust and developing trust-oriented communication strategies (Table 4 (REC1.4)), expanding collaboration with community leaders and the scientific community (Table 4 (REC1.1)), and working with individuals and organizations that have established trust in communities (Table 4 (REC1.2)). Using the preferred means of communication for different populations was unanimously recommended to further earn trust (Table 4 (REC1.1)).

Multidisciplinary research should assess the impact of the COVID-19 infodemic on health behaviours and outcomes (Table 4 (REC1.5)). Research funders should commission more reviews that synthesize, evaluate and disseminate COVID-19-related evidence to inform needed interventions (Table 4 (REC1.6)).

Strengthen health systems

Health systems have experienced wide-ranging circumstances throughout the pandemic, from periods of relative calm to periods of near collapse. The broad agreement among panellists strongly suggests that, although many health systems will remain at risk of once again being overwhelmed, those risks can be mitigated. Certain sources of risk to health systems are essentially structural, such as the lack of implementation of an evidence-based, globally agreed-upon set of minimum COVID-19 pandemic response standards (Table 2 (STMT2.1)).

As noted above, health systems recommendations with respect to whole-of-society (Table 4 (REC2.5)) and whole-of-government approaches (for example, multiministry coordination) (Table 4 (REC2.6)) were among the most highly ranked by the panel.

As community transmission of SARS-CoV-2 continues to present a risk to health systems, particularly through variants of concern, extensive virological surveillance should be used (Table 5 (REC2.8)). Public health policies should take better account of the potential long-term impact of the unchecked spread of COVID-19 given the ongoing uncertainties about the prevalence, severity and duration of post-COVID-19 morbidity (long COVID) (Table 5 (REC2.9)). Member States should authorize the World Health Organization (WHO) to lead a large, inclusive, multistakeholder, global effort to provide public health and clinical targets pertaining to SARS-CoV-2 and COVID-19, with an emphasis on cases, vaccination, morbidity and mortality (Table 5 (REC2.17)).

Economic impacts, notably costs borne by consumers (Table 2 (STMT2.5)), create risks to health systems. To address these risks, structural and economic recommendations include removing economic barriers to SARS-CoV-2 tests, personal protective equipment, treatment and care (Table 4 (REC2.1)), supporting the development of regional manufacturing hubs for COVID-19 supplies, treatments and vaccines (Table 4 (REC2.2)), and considering legislative and regulatory reforms to address market failures (Table 5 (REC2.16)). Where access to PCR or antigen tests is limited, providers should consider adopting a syndromic approach (Table 5 (REC2.18)). Notably, REC2.18 is the health systems recommendation with the highest percentages of panellists disagreeing as well as panellists indicating ‘not qualified to respond’.

To reduce the burden on hospitals, the role of primary health care should be strengthened (Table 5 (REC2.10)), while health care workers’ physical, mental and social well-being should be supported (Table 4 (REC2.4)).

With respect to digital health, the recommendations encourage increasing investments in digital health infrastructure (Table 5 (REC2.13)), adapting user interfaces and experience to expand access, particularly for vulnerable groups (Table 4 (REC2.3)), and leveraging implementation science to determine which digital health solutions can be quickly scaled (Table 5 (REC2.12)).

With respect to procurement practices, engaging continuous improvement disciplines for intercountry procurement, pooling and supply-chain management was urged (Table 5 (REC2.11)). To best leverage community-based interventions and services, community-based organizations and students pursuing degrees in health-related fields should be engaged in providing COVID-19 education, testing and vaccination services (Table 5 (REC2.14)).

As social, political and economic sector risks continue to have spillover effects on health systems, key multisectoral indicators for systemic risks to health systems should be identified and assessed (Table 5 (REC2.7)).

Finally, health systems should identify and, where possible, reduce diagnostic, treatment and care backlogs for non-COVID-19-related medical conditions (Table 5 (REC2.15)).

Emphasize vaccination, but not exclusively so

Even assuming continued innovation of vaccines and interventions that reduce vaccine hesitancy, 97% of the panel agrees that vaccination alone is insufficient to end the COVID-19 pandemic as a public health threat (Table 2 (STMT3.6)). Thus, the panel places a strong emphasis on additional prevention measures, particularly, as noted above and in the ten highest-ranked recommendations (Table 8), for countries to adopt a vaccines-plus approach, as discussed in the next domain.

Regarding the key role of vaccines, the panel made a range of recommendations. Government, philanthropic and industry funding should invest in developing vaccines that provide long-lasting protection against multiple SARS-CoV-2 variants (Table 6 (REC3.4)). As waning immunity remains a risk, calculations for immunity should consider the time after the date of vaccination and/or infection and be regularly updated with new scientific evidence (Table 6 (REC3.5)).

Vaccine hesitancy, which ranges from delay to refusal despite availability of vaccine services, remains a major challenge (Table 2 (STMT3.3)). To reduce vaccine hesitancy and increase uptake, several interventions are recommended: engaging trusted local leaders and organizations in vaccination efforts (Table 6 (REC3.2)), providing information that clearly explains the efficacy and limitations of current vaccines (Table 6 (REC3.1)) and tailoring messages to address the underlying bases of various populations’ specific concerns through targeted public health communications (Table 6 (REC3.3)). Vaccine hesitancy may also be associated with false information, which is addressed in the communication domain above.

On the one hand, panellists largely agree that medical autonomy of individuals with decision-making ability extends to the right to make one’s own decisions regarding vaccination (Table 2 (STMT3.2)). On the other hand, panellists also acknowledge that, when the risk of harm to others is sufficiently severe, governments may determine that the right of all individuals to good health overrides the autonomy of any one individual to choose not to be vaccinated (Table 2 (STMT3.1)). These statements reflect among the highest levels of combined disagreement (Table 2 (STMT3.1, 9%; STMT3.2, 16%)). Civil liberties implications are further discussed in the next domain.

Promote preventive behaviours

As noted above, vaccination alone will not end COVID-19 as a public health threat (Table 2 (STMT3.6)) for all people. Infection rates tend to increase when governments discontinue social measures, including non-pharmaceutical interventions, regardless of the level of vaccination (Table 3 (STMT4.5)). Thus, all countries should adopt a vaccines-plus approach, including a combination of COVID-19 vaccination, other prevention measures, treatment and possibly financial incentives (Table 6 (REC4.5)).

Although the nature and vectors of SARS-CoV-2 transmission were not clearly understood early in the pandemic, current evidence guided the panellists to near-unanimous agreement that SARS-CoV-2 is an airborne virus that presents the highest risk of transmission in indoor areas with poor ventilation (Table 3 (STMT4.1)). Risk-related communications from all actors should clearly emphasize that transmission of SARS-CoV-2 is primarily caused by inhalation of the virus (Table 6 (REC4.3)). Considering the airborne nature of transmission, governments should regulate and incentivise structural prevention measures, such as ventilation and air filtration (Table 6 (REC4.1)), and high priority should be given to preventing SARS-CoV-2 transmission in the workplace, educational institutions and commercial centres (Table 6 (REC4.6)).

Mammal-to-mammal transmission represents a reservoir for future zoonotic variants (Table 3 (STMT4.3)). Thus, substantial virological surveillance based on whole-genome sequencing of positive samples in human and high-risk mammal populations is an essential component of the continued pandemic response and preparedness (Table 5 (REC2.8)).

National and international travel restrictions should be based on current scientific knowledge and prevailing transmission rates of all variants that consider relevant, health-based factors (Table 6 (REC4.4)). Measures that are no longer scientifically valid for COVID-19 prevention should be immediately removed from COVID-19 guidance and policy (Table 6 (REC4.2)). Going forward, governments should consider imposing broad restrictions on civil liberties only in the event of variants of concern presenting risk of high rates of transmission and severity, coupled with waning immunity or vaccine resistance (Table 6 (REC4.7)).

Expand treatments

Panellists had substantially high agreement regarding all aspects of treatment and care, indicating that treatment will continue to be an area of major importance both for ending COVID-19 as a public health threat and for individual patient care. Notably, a statement addressing the risk of prioritizing treatment over prevention (Table 3 (STMT5.1)) had the highest level of combined disagreement (7%) for this domain.

With current public health policies reflecting greater tolerance for community transmission and increased rates of infection, research into COVID-19 must adapt and develop further evidence to understand the cumulative effect of COVID reinfection (Table 7 (REC5.4)). Research is needed to determine whether infection from distinct variants of SARS-CoV-2 is associated with significant differences in long-term morbidity (Table 3 (STMT5.4)). Additional research funding, particularly for long COVID, should be prioritized (Table 7 (REC5.6)), and multisectoral collaboration should accelerate new therapies across all stages of COVID-19 (Table 7 (REC5.2)). Moreover, global case definitions should be standardized (Table 7 (REC5.1)).

Echoing some statements and recommendations in the pandemic inequities domain (discussed below), clinical trials and longitudinal cohorts should be more inclusive and statistically representative regarding age, gender and vulnerable populations (Table 7 (REC5.3)).

Eliminate inequities

The substantial agreement of the panellists suggests that addressing inequities remains a global challenge. Immediate efforts should be made to reduce vaccine wastage (Table 7 (REC6.8)), addressing the need for cold storage, transport and other infrastructure-based barriers in low-resource settings (Table 7 (REC6.4)), addressing the affordability of testing and treatment for people in all countries (Table 7 (REC6.2)), as well as accelerating efforts to distribute vaccines in low- and middle-income countries (Table 7 (REC6.10)).

Transfer agreements to increase production capacities in low- and middle-income countries should be expedited (Table 7 (REC6.6)). Pre-existing social and health inequities must be considered in pandemic preparedness and response going forward (Table 7 (REC6.7)). The findings call special attention to two vulnerable populations: children (Table 7 (REC6.5)) and those living within or fleeing from conflict zones (Table 7 (REC6.9)).

The pandemic has illustrated the risk of over-reliance on experts from a small number of disciplines (Table 3 (STMT6.8)), often excluding the expertise of community members (Table 4 (REC1.2)) and vulnerable groups (Table 3 (STMT6.7)). Instead, vulnerable groups should be sought out and actively engaged (Table 7 (REC6.3)). As noted in the communication domain, community leaders should also be engaged (Table 4 (REC1.1)). Multidisciplinary experts who understand local contexts should be included in developing national operational plans for ending COVID-19 as a public health threat (Table 7 (REC6.1)). COVID-19 tests and treatments should be affordable for all people in all countries (Table 7 (REC6.2)).

Discussion

Wide-ranging pandemic control measures59,60,61,62 have not ended COVID-19 as a public health threat63,64,65,66,67,68. Although this study echoes some earlier findings—for example, the Independent Panel for Pandemic Preparedness and Response35, the European Union 2022 communication on preparedness and response69 and WHO’s 2022 plan on strategic preparedness53—it is distinct from previous efforts22 given its design, which emphasized consensus building and the reporting of disagreement through the Delphi method, panellist diversity with regard to geography and disciplines, and the large sample size. The study’s focus—ending COVID-19 as a public health threat—is defined as being evidenced by the resumption of pre-pandemic social, cultural, religious, political, healthcare, economic and educational activities in each country’s context. Some retrospective matters (for example, pandemic root-cause analysis), theoretical questions and modelling were judged to be beyond the scope of the study.

Where possible, the study emphasizes recommendations that can be implemented in the short term (that is, in months, not years) to end COVID-19 as a public health threat. Although examples of countries implementing multiple recommendations exist (for example, free tests70, combining widespread testing and free treatment of positive cases along with digital technologies71, the development of vaccines providing long-lasting protection against variants72,73), the exceptions accentuate global challenges and provide new opportunities for action. Certain statements and recommendations resulting from this consensus process address gaps in WHO’s strategic plan31, most strikingly, the failure to directly address the airborne nature of transmission. Initially, the WHO incorrectly labelled airborne transmission of SARS-CoV-2 as ‘misinformation’. Only much later, after multidisciplinary scientific efforts, did the WHO recognize airborne transmission to be a predominant mode of transmission74,75,76. By contrast, this panel recommends that ‘risk communications clearly emphasize’ (Table 6 (REC4.3)) the causal link between inhalation of SARS-CoV-2 and the transmission of COVID-19 as well as policy incentivizing ‘structural prevention measures (for example, ventilation, air filtration) to mitigate airborne transmission’ (Table 6 (REC4.1)).

The WHO’s slow pace in directly addressing the airborne nature of transmission underscores why public health policy and risk communications should be based on evidence. For example, supposing that endemicity will result in lower virulence is an erroneous assumption77,78,79 that may exacerbate disproportionate risks of COVID-19 among vulnerable groups80. By extension, engagement with communities through effective risk communication should remain a priority for all countries.

The WHO recognizes the infodemic as a key challenge to effective communication for general populations53,81,82,83, vulnerable groups84 and scientists85. Governments, health authorities and healthcare providers should especially take care in the accuracy of their communications. The panel also emphasized that institutions should proactively monitor false health information and collaborate with trusted community leaders to refute it and enhance trust86.

Given the disproportionate impact that the pandemic has had on vulnerable groups to date87,88,89, the panel voiced concern that policy decisions must aim to find ways of lowering risk within these groups after resumption of the aforementioned activities (STMT6.1). As those vulnerable to COVID-19 in many countries can no longer rely on other individuals practising basic prevention measures (such as the use of face masks and isolating after testing positive), the structural changes recommended in this study (for example, indoor ventilation and filtration) assume heightened importance. Furthermore, COVID-19 continues to prompt global discussion and vigorous debate, particularly about tensions among medical ethics, civil liberties and pandemic control measures80. This study is no exception, with statements STMT1.6 (blaming unvaccinated individuals) and STMT3.2 (individual decisions regarding vaccination) receiving the highest levels of disagreement, underscoring the need for equitable structural interventions. In countries with widespread availability of vaccines, it is important for health authorities to distinguish between those who have clearly refused and are unlikely ever to seek vaccination and those who remain hesitant and continue to delay vaccination90. In the latter case, specific factors prolonging the delay can be addressed by targeted interventions. Finally, continued uncertainty about the widespread consequences of long COVID and its implications for public health policy (REC2.9) is an ongoing concern91,92.

Some innovations, notably vaccines37,38, have not been equitably distributed to low- and middle-income countries, and others, such as high-quality facemasks, have not been widely adopted in high-income countries despite their availability93. Some recommendations addressing pandemic inequities remain underleveraged; for example, providing more vaccines94 to countries with a low percentage of people vaccinated (REC6.10). Other recommendations may necessitate increased funding and time— for example, calls for continued vaccine and treatment innovations (REC2.12, REC5.2, REC5.6).

Importantly, the single significant difference in levels of panel agreement between those working in high-income countries and those working in low- and middle-income countries pertained to the role of economic incentives (REC3.6), probably reflective of sociocultural distinctions or perhaps disagreement over feasibility in implementation and ethics concerns95,96. Furthermore, 14% of the panellists considered themselves to be not qualified to respond to STMT4.3 concerning zoonotic variants, which probably indicates a lower understanding of biological vectors and the aetiology of variants among some of the disciplines included in the panel compared with the other topics covered97.

As noted above, the panellists nearly unanimously agreed on and prioritized whole-of-society and whole-of-government approaches98,99,100,101 (Table 8). The panellists also prioritized recommendations for communicating effectively with the public and developing technologies (for example, vaccines, therapies and services) that can reach target populations (Table 8). Failure to use these approaches risks not only prolonging COVID-19 as a public health threat, but also further diversion of resources from efforts to achieve other extant public health goals102,103.

Strengths and limitations

One of the strengths of this study is its use of Delphi methodology. By demonstrating increased agreement with each subsequent round, this method enabled us to determine whether our incorporation of feedback was successful in refining the statements and recommendations, increasing the degree of consensus and, in some cases, reaching unanimity. The consistently increasing mean levels of agreement with the consensus statements and recommendations observed across all three survey rounds strengthens our confidence in the relevance of the iterative Delphi process in eliciting feedback to improve subsequent rounds. This is particularly noteworthy given that the effort to incorporate feedback from the expert panel may have resulted in more complex (for example, multiple item) statements and recommendations. Generally, there may be concerns as to the clarity of such statements; however, levels of agreement tended to be either maintained or increased, providing greater confidence in their resonance with the panel. The overall high response rates across three survey rounds speaks to both the rigorous implementation of the method and the commitment of the assembled panel of experts. Endorsement of the resultant consensus statements and recommendations by 184 organizations in 72 countries (Supplementary Table 2) at the time of publication further testifies to their global relevance.

Although the Delphi method is a robust approach (Methods) to assess levels of agreement on specific issues and explore whether a consensus can be reached, it is not without limitations. A main concern pertains to the construction of a truly representative expert panel. The sequential, multimethod sampling approach that we used (see the ‘Delphi expert panel member sample’ section in the Methods) minimized potential bias from purposive sampling of a small group and, instead, generated a large, geographically and disciplinarily diverse panel from multiple sources (that is, the core group, nominees from the core group and corresponding authors of key COVID-19 literature). While potential panellists were identified from their work related to COVID-19, infectious diseases, public health preparedness and other fields, the chairs further confirmed their appropriateness for the study by instructing them to not participate if they felt they lacked expertise concerning the pandemic. This approach appears to have been appropriate, as only 5–14% of the panellists felt they were not qualified to respond to just 5 out of the 41 statements, and 3 of the 57 recommendations. Although conducting the study in English limited the participation to English speakers, the inclusion of experts from 112 countries and territories strengthens our confidence in the potential broad applicability of these recommendations to a range of cultures and countries. With regard to the mid-study convening of the core group to discuss issues raised in the initial survey rounds, another limitation may have been that we conducted it virtually rather than in person (see the ‘Delphi data collection’ section in the Methods).

Conclusions

The multidisciplinary panel’s emphasis on actionable, near-term recommendations guided the Delphi consensus-building process and increased the relevance of the study’s findings to a broad group of stakeholders, including governments, public health authorities, NGOs, community-based organizations, industry, and social media platforms and other media. This consensus study advances a global vision of informed decision-making on how the world can end COVID-19 as a public health threat without a return to sweeping limitations on civil liberties, without risking the health and lives of vulnerable groups, and without exacerbating economic burdens.

Methods

Delphi expert panel member sample

We used an iterative sampling approach to generate a large panel for this Delphi study (Fig. 1). The four co-chairs (J.V.L., A.B., A.K. and A.E.-M.) identified a core group of 40 academic, health, NGO, government and policy experts from 25 countries and territories. Selection by the co-chairs was primarily based on publication record and engagement on COVID-19 issues as well as online biographies. Twenty-nine of these experts were well known to the chairs while seven were suggested through snowball sampling to result in geographical and gender equity among the core group of 40. Furthermore, a concerted effort was made towards multidisciplinary representation in the core group, including medical sciences (such as infectious diseases, public health and vaccinology), engineering, and social sciences (such as policy, law and ethics). The core group proposed additional experts to create a global panel of approximately 400 experts. The lead chair (J.V.L.) and methodologist (D.R.) led this core group through implementation of the project. Snowball sampling was then used as core group members identified individuals with expertise in COVID-19 from their professional networks to generate an initial list of potential Delphi panel members with the goal of broad representation. In proposing experts, co-chairs focused on identifying at least one representative from at least 100 countries. One co-chair (J.V.L.) took responsibility for reviewing the suggestions, with support from a research assistant who shared recent publications and a professional biography for every proposed co-author. Many initial suggestions were of leading experts with whom the co-chairs had previously collaborated.

The core group then reviewed the panel list for under-represented countries and PubMed/Medline searches were conducted using the search term ‘COVID-19’ in combination with the names of under-represented countries to identify authors of COVID-19 research studies involving primary data collection in these countries. Authors of relevant studies were invited to participate in the Delphi panel to further increase geographical diversity and include panellists beyond the core team members’ networks. All of the panel participants were carefully vetted; most had published in one or more relevant fields.

To further validate the expertise of the panel, the study was described to the invitees (n = 696) with the following instructions: “If you consider your professional training and expertise applicable to the subject matter of this global consensus statement project, we encourage you to participate in the panel.” Informed consent was obtained for each panellist after explaining the purpose of the study and their expected contributions, including review and approval of the submitted manuscript, by accession to the Round 1 (R1) survey. Our objective was for invited participants to explicitly consider whether they had the necessary level of expertise before joining the Delphi panel. We do not have specific information regarding the basis of invitees’ non-participation but expect that these instructions enabled a substantial portion of non-respondents to self-select out of the study. We know that 84 invitees began the R1 survey but did not complete it; thus, if we assume that they did consider themselves to be eligible to participate but then decided not to do so, that would result in an estimated response rate of 82.1% (386 out of 470). The resultant expert panel is diverse in terms of demographic, disciplinary and geographical characteristics (Table 1).

Table 8 Ten highest ranked recommendations

Delphi statement domains

The core group reviewed the published literature available up to January 2022 to draft initial statements for the first Delphi survey round, grouped in the following domains: (1) communication; (2) health systems; (3) vaccination; (4) prevention; (5) treatment and care; and (6) pandemic inequities. No formal systematic review with stringent criteria for levels of evidence was performed owing to the sheer volume of COVID-19-related published studies and the frequency at which they were and continue to be published. However, all of the authors and panellists were invited to suggest relevant papers, which were reviewed by the core group members based on journal rankings, paper citations and other metrics. In R1, panellists considered draft consensus statements based on the literature before moving to the next step of recommendations in round two (R2), which emanated from the panellists’ feedback on the statements as well as new research findings over the course of data collection from 18 February 2022 to 28 April 2022.

Delphi method data collection

The study design consisted of digital data collection: two survey rounds (R1 and R2) of draft statements; an online consensus meeting of the core group (16 March 2022) to discuss salient issues; one round of draft recommendations (in R2); and, a final, third survey round (R3) of the consensus statements and recommendations (Fig. 1). The core group decided a priori to use a supermajority (that is, ≥67% combined agreement) minimum cut-off for consensus. This more demanding cut-off (relative to a simple majority of greater than 50%) was considered to be necessary given the project goal of supporting global policy and programmatic actions to address the COVID-19 public health crisis. We used the QualtricsXM platform to develop and distribute the surveys (round duration ranged from 1.5 to 3 weeks) with four-point Likert-type categories for measuring the level of agreement with the statements and recommendations (that is, agree, somewhat agree, somewhat disagree, disagree); a fifth ‘not qualified to respond’ option was provided given the panel’s range of COVID-19 expertise. Panellists could provide comments and suggest edits to individual statements and recommendations in text boxes, which followed each of the statements and recommendations. All rounds allowed for overall comments at the end of the survey, and the researchers reviewed 1,409, 755, and 188 comments associated with the statements in R1, R2 and R3, respectively, and 1,025 and 2,156 comments associated with the recommendations in R2 and R3, respectively. Summaries of changes based on panellist input from a previous round were available in text boxes next to each statement and recommendation in the subsequent round. Similarly, the definition for “Ending COVID-19 as a public health threat as evidenced by the resumption of social, cultural, religious, political, healthcare, economic and educational activities in each country’s context” was presented during each round so that panellists could respond to statements on the basis of a shared understanding of how the phrase “ending COVID-19 as a public health threat” was defined for the purpose of this study. In R3, panellists also ranked the top half of recommendations within each of the six domains, which were automatically randomized to mitigate order-effect bias. Using Microsoft Excel (v.16), scores were calculated and normalized using the Dowdall system to compare rankings across domains by accounting for weighting bias due to differences in the total number of recommendations in each domain137,138.

An important component of the data-collection process involves the discussion among core group members of issues that emerge from the early survey rounds and how best to incorporate such feedback in subsequent rounds. Given the geographical distribution of panel members and COVID-19-related travel and health concerns, we convened the core group virtually for in-depth, real-time deliberation. This web-based approach is different from in-person discussion of complicated or contentious issues; however, panel members had multiple opportunities to provide open-ended comments in the absence of dominant voices that can inhibit the expression of minority viewpoints during in-person convenings. Thus, the combination of real-time feedback (from core group members) and written feedback (from the entire panel) probably resulted in more comprehensive contributions overall.

Delphi data analysis

Data analysis reflected the multiple-methods nature of Delphi studies and was managed by an analytic team of core group members, the study methodologist and research assistants. Across the three rounds, we ran frequencies of all statements and recommendations (Supplementary Discussion 2); the proportion who selected ‘not qualified to respond’ is reported in the data tables but removed from the denominator to calculate levels of agreement/disagreement from the relevant sample. The team then analysed the extensive qualitative data (that is, open-ended text-box comments). Specifically, comments were first reviewed individually by at least three core group members (J.V.L., co-chair; D.R., methodologist; and C.J.K.) and an additional co-author (T.M.W.). For each data collection round, comments were then discussed in online review meetings, including at least three core group members and an additional co-author. After review and discussion, comment suggestions were incorporated into statement and recommendation revisions for subsequent rounds. A supermajority of core group members (28 out of 40; 70%) participated in the online consensus meeting, which permitted in-depth breakout-group discussions on salient issues from R1 and R2 informing R3 revisions (Supplementary Discussion 3). Quantitative analysis of the final R3 results involved assigning each statement and recommendation a grade to indicate the level of combined agreement (agree + somewhat agree), using a system that has been used in other Delphi studies139,140,141 in which ‘U’ denotes unanimous (100%) agreement; ‘A’ denotes 90%–99% agreement; ‘B’ denotes 78%–89% agreement; and ‘C’ denotes 67%–77% agreement. Although all statements and recommendations exceeded the standard supermajority minimum of ≥67% combined agreement for consensus, we highlighted those with <67% for ‘agree’ alone for further analysis. Statements and recommendations were analysed using Fisher’s exact tests in Stata (v.16) to assess differences in agreement by the following panellist characteristics: income level (high income versus low- and middle-income) for country of birth and country where currently working, primary sector of employment and primary field of employment (Supplementary Discussion 2). The use of the terms combined agreement and combined disagreement are presented in the results.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.