Clinical, economic, and humanistic burden of community acquired pneumonia in Europe: a systematic literature review

ABSTRACT

Background

Community-acquired pneumonia (CAP) is an infectious lung inflammation contracted outside the hospital. CAP is a leading cause of death among young children, elderly, and immunocompromised persons. Incidence can reach 14 cases/1,000 adults. Up to 50% of cases require inpatient hospitalization. Mortality is 0.7/1,000 cases or 4 million deaths per year. We sought to summarize multi-dimensional burden of CAP for selected European countries.

Methods

We conducted a systematic literature review of literature published from 2011 to 2021 whereby we sought information pertaining to the epidemiologic, clinical, economic, and humanistic burden of CAP. Findings were summarized descriptively.

Results

CAP incidence in Europe is variable, with the highest burden among those of advanced age and with chronic comorbidities. Etiology is primarily bacterial infection with Streptococcus pneumoniae being the most frequently implicated. Direct medical costs are primarily attributable to inpatient stay, which is exacerbated among high-risk populations. Higher mortality rates are associated with increasing age, the need for inpatient hospitalization, and antibiotic resistance.

Conclusions

A better understanding of CAP is needed, specifically the economic and quality of life burden on patients and caregivers. We recommend further assessments using population-level and real-world data employing consistent disease definitions.

KEYWORDS:

• Community acquired pneumonia
• disease burden
• Europe
• economic
• epidemiology
• quality of life

1. Introduction

Community-acquired pneumonia (CAP) is an infectious pneumonia acquired outside the hospital setting [1]. Among infectious diseases, CAP is the leading cause of death despite improvements in prevention and care [2], and outcomes among young children, the elderly, and otherwise immunocompromised persons are particularly poor [3]. Other known risk factors include (but are not limited to) smoking, malnutrition, previous CAP episode, chronic bronchitis or chronic obstructive pulmonary disease (COPD), asthma, diminished functional impairment, poor dental health, and treatment with immunosuppressive drugs [3].

CAP incidence among all adults can reach 14 cases per 1,000 [4]. Up to 50% of cases require inpatient hospitalization and the mortality rate reaches 0.7 per 1,000 persons per year [3,5]. The World Health Organization has reported that CAP accounts for 4 million deaths per year and 7% of the total annual mortality rate [6]. The economic burden of CAP is high as well, with Europe accruing EUR 10.1B per year not including indirect costs such as presenteeism, absenteeism, and impacts on caregivers [7,8].

Notwithstanding the considerable impact of CAP globally, Europe – with its aging population – is particularly susceptible [9]. Though the literature pertaining to its burden is somewhat fractured owing to inconsistent definitions of CAP disease, differences among countries in how well CAP is monitored, and inconsistencies in how CAP effects are measured clinically and economically. The objective of the present review and narrative synthesis is to develop a cohesive summary of this multi-dimensional burden of CAP for selected European countries.

2. Materials and methods

This systematic literature review was conducted according to a pre-specified protocol PROSPERO (# CRD42021274335). Our objective was to summarize the epidemiology, clinical and economic burden, and quality of life impact of CAP in the EU27, UK, Norway, Switzerland, and Iceland. This objective was operationalized through and organized by the following six research questions:

• (Epidemiology) What is the incidence of CAP?

• (Epidemiology) What is the distribution of causative pathogens?

• (Clinical) What is the survival of CAP patients?

• (Clinical) What is the length of stay for hospitalized patients with CAP?

• (Economic) How does CAP drive direct medical, direct non-medical, and societal costs?

• (Humanistic) How does CAP impact patient quality of life and that of caregivers?

The selected European countries are as follows: Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, Switzerland, United Kingdom.

2.1. Search and selection

We applied a structured search strategy (Appendix A) to MEDLINE and Embase databases executed through the ProQuest electronic search portal. Databases were for articles published from 2011 to 2021 or conference abstracts published in 2019 to 2021. Note that we chose this time horizon for conference abstracts so that our search would yield important information that had not yet been fully published. Our strategies for Embase/MEDLINE were populated with key words and themes based on recommendations from the Canadian Agency for Drugs and Technologies in Health [10] and the Scottish Intercollegiate Guidelines Network [11].

After deduplication, all citations underwent title and abstract screening by two reviewers. At the title and abstract screening stage, articles were selected if it was indicated that they describe adults with CAP (i.e. ≥18 years old) or their caregivers. Studies selected in the title/abstract screening step proceeded to the full-text review step where they were fully evaluated against the PICOS criteria (Table 1). Studies selected in the full-text review step were retained for data extraction. Citations from relevant systematic reviews and meta-analyses were searched for additional studies of interest.

Table 1. PICOS criteria for study selection.

	Inclusion Criteria	Exclusion Criteria
Population	Adults with CAP in the target countries^A	Non-adults (i.e. <18 years old) Persons diagnosed with hospital-acquired pneumonia
Interventions/Comparators	Any or none	Not applicable
Outcomes	Epidemiology: Incidence, prevalence, demographic characteristics, population characteristics (stratified by microbial etiology and serotype to the extent that the data permit) Clinical: Mortality, morbidity (i.e. subsequent clinical impacts) Humanistic burden: quality of life (any instrument) Economic burden: direct medical costs, direct non-medical costs, and societal costs	Outcomes pertaining to persons with hospital-acquired pneumonia
Study design	Any not excluded^B,C	Interventional/clinical trials Non-systematic reviews Editorials, comments, notes Case studies
Data sources		Not applicable
Electronic databases	Articles indexed in Embase or MEDLINE
	Conference abstracts indexed in Embase^D
Search limits	Articles published 2011–2021^E; conference abstracts from 2019–2021	None

^ABelgium, Bulgaria, Czech Republic, Denmark, Germany, Estonia, Ireland, Iceland, Greece, Spain, France, Croatia, Italy, Cyprus, Latvia, Lithuania, Luxembourg, Hungary, Malta, Netherlands, Norway, Austria, Poland, Portugal, Romania, Slovenia, Slovakia, Finland, Sweden, Switzerland, United Kingdom.

^BSystematic literature reviews will be retained separately to hand-check the studies included in those reviews against the studies identified in the present review. Any relevant studies identified in previous systematic literature reviews but not by the present search strategy will be evaluated and retained if applicable.

^CThis search strategy captured published Cochrane reviews.

^DThese databases contain individual records for individual abstracts and conference presentations.

^EThe search excluded records published before the year 2011; although, it is expected that some records published on or after the year 2011 will include information collected before 2011.

2.2. Data collection and synthesis

We extracted data into a centralized, structured database. For binary outcome data, the numerator and denominator were extracted. For continuous items, the mean/median value was extracted as well as the standard error, standard deviation, or confidence interval. Baseline, endpoint, and change-from-baseline values were extracted as well as mean differences (and their respective standard error, standard deviation, or confidence interval) between cohorts if reported. This evidence synthesis does not include formal analysis, hypothesis-driven analysis, or meta-analysis. Note that such analyses might be possible but were not further considered given the complications expected to arise from the variable study designs included here and because our interest was in understanding the key trends and drivers of the chosen outcomes in a descriptive manner. Information was analyzed descriptively, and findings reported in a narrative style. Throughout this manuscript, we refer to ‘high risk’ cohorts based on the contents of the included studies, which might have defined ‘high risk’ inconsistently.

2.3. Risk of bias and study quality

Risk of bias and study quality were assessed for all studies retained in the review using recognized quality assessment tools. Scores were applied by one reviewer and verified by a second reviewer, with discrepancies being resolved by consensus or involvement of a senior reviewer. Given the broad nature of this review, we applied multiple quality assessment tools published by the Joanna Briggs Institute (https://jbi.global/critical-appraisal-tools). Specifically, we applied the checklists for cohort studies, economic evaluations, and prevalence studies. Our findings in this regard are available upon request in a data supplement.

3. Results

The structured literature search yielded 2,222 unduplicated studies for title and abstract screening (Figure 1). Of these studies 1,819 were excluded at this initial screening step. Aside from an initial exclusion step that removed early conference abstracts, the most common reasons for exclusion during the initial screening step were general irrelevance (n = 225), wrong study design-commentary (n = 163), and not target country (n = 158).

Figure 1. PRISMA diagram.

At this initial screening, there were many reasons that a study could be considered generally irrelevant and these reasons which could also overlap with the more specific reasons for exclusion. Typically, a study was marked as irrelevant when it had multiple exclusionary characteristics simultaneously. The most common reason for a study being judged as irrelevant is that it did not include at least one outcome of interest. Full texts of the remaining 403 studies were then screened against the PICOS criteria (Table 1), and 248 studies were subsequently excluded. The most common reasons for exclusion were not target country (n = 73), included only irrelevant outcomes (n = 58), and CAP was not the focus of the study (n = 45). A total of 155 studies qualified for extraction initially. Of these 155 qualifying studies, 82 studies were conducted from a national or relevant sub-national perspective and were retained for summary in the present report.

3.1. Epidemiology

Target outcomes for the epidemiology topic included national incidence and relative contribution of causative pathogens (i.e. etiology). With respect to the incidence of CAP nationally for the selected countries (Figure 2), we identified only 12 studies with national-level observations, while three studies estimated national-level CAP incidence within the context of population or economic models [12–14] (Appendix, S2). Some studies distinguished between incidence of CAP requiring hospitalization versus CAP not requiring hospitalization. For example, in a German study by Kolditz et al. (2016), incidence of CAP requiring hospitalization ranged from 50 per 100,000 among persons aged 18–29 years old to 2,940 per 100,000 persons aged ≥90 years old [15]. A similar trend was identified in a study of the Belgian population 106 per 100,000 among persons aged 65–74 to 332 per 100,000 persons aged ≥85 years old [14]. The distinction was more pronounced when comparing incidence of CAP requiring hospitalization among younger persons with low risk (e.g. lower age, fewer comorbidities) (64 per 100,000 persons) versus older patients with high risk (e.g. higher age, more comorbidities) (1,733 per 100,000 persons).

Figure 2. Summary of CAP incidence (per 100,000) by age group and healthcare setting.

We observed similar age and baseline risk trends in the incidence of CAP not requiring hospitalization. In a model of the Swedish healthcare system, for example, the incidence of CAP not requiring hospitalization ranged from 315 per 100,000 persons aged 65–74 years old to 525 per 100,000 persons aged ≥75 years old [12]. Other notable subgroups included increased incidence among Dutch persons living in the vicinity of livestock farms (1,800 per 100,000 persons) [16] and increased incidence among patients with diabetes (2,058 per 100,000) versus those without (727 per 100,000) in Spain [17].

A summary of the etiology of CAP (i.e. relative proportion of causative pathogens) is given in (Figure 3, individual data in Appendix S3, based on data reported Shoar et al 2020² now updated with more recent studies). Here, we found across many countries and settings that the most frequently implicated bacterial pathogen was Streptococcus pneumoniae (implicated in 43% of cases) followed by Haemophilus influenzae and Staphylococcus aureus (16.1% and 9.6%, respectively). Viral infections were less common (6.8%). Note that this was not the case for all studies [18,19].

Figure 3. Summary of CAP-causative pathogens.

3.2. Clinical burden

The clinical burden of CAP was assessed in terms of mortality. The highest mortality was reported among elderly patients who were admitted for inpatient treatment for CAP. Aliberti et al. (2013) conducted an analysis of CAP cases in the United Kingdom (17% of whom had been admitted directly from nursing homes and who had a median age of 69 years old). Mortality at 30 days was 39% [20]. Likewise, Torner et al. (2017) found the highest 30-day mortality in this review: 44.5% among persons aged >84 years old who were admitted for inpatient CAP treatment in Spain [21].

Alternatively, persons who were treated on an outpatient basis – even those with advanced age – had lower mortality at 30 days: 0% to 10%. A comparison of patient subgroups for survival at 30 days (Figure 4) indicated that higher mortality was associated with inpatient treatment and age. Other factors associated with higher mortality at 30 days included selected comorbidities (e.g. chronic obstructive pulmonary disease (COPD), diabetes, cardiovascular disease (CVD), or immunocompromised) and multi-drug resistance. There were insufficient mortality observations at 90 or 180 days to draw meaningful inferences.

Figure 4. Summary of CAP-associated mortality by country and patient age.

The other clinical burden outcome assessed in this synthesis was length of stay among persons admitted for inpatient treatment for CAP. Here, we found that length of stay seemed to vary according to baseline risk factors, particularly age and comorbidities, but also somewhat by country when comparing seemingly similar cohorts. The shortest lengths of stay (approximately 7 days) were observed among younger persons (≤64 years old) who either had not recently received outpatient antibiotic treatment [22] or who had not developed or did not have evidence of antibiotic resistance [20]. Longer lengths of stay (approximately 19–23 days) were associated with multidrug resistance [20,23] and admittance through the intensive care unit [24].

3.3. Cost and QOL burden

Seven studies reported the cost (Table 2) and/or quality of life implications of CAP [12,14,23,25–28]. In an economic analysis of Belgian CAP patients, Marbaix et al. (2018) applied a Markov model to assess the cost per CAP case stratified by age and treatment setting. Among patients treated in an inpatient setting, the cost per case (in 2016 Euros) ranged from EUR 8,501 for those aged 65–74 years old to EUR 17,044 for those aged ≥85 years old. A similar impact of age was seen among patients treated in an outpatient setting: EUR 867 for those aged 65–74 years old to EUR 984 for those aged ≥85 years old [14]. Elsewhere in the German population, Kaier et al. (2019) used a claims database analysis to assess the difference between the cost per CAP inpatient stay versus the amount reimbursed. The study found that for patients in whom CAP was the primary diagnosis, the cost of treatment was higher than the amount reimbursed (EUR 14,066 vs EUR 13,713 in 2014 Euros) [23].

Table 2. Summary of CAP economic studies, per event costs.

Study (Country)	Study Design	Relevant Findings
Marbaix 2018 [14] (Belgium)	Cohort Model: Assessed lifetime risks, consequences, and costs of IPD and CAP associated with PCV13 vaccination for cohort aged 65–75	Per case cost of moderate-risk CAP with inpatient care: €8,501-€15,482 Per case cost of high-risk CAP with inpatient care: €9,359-€17,044 Per case cost of moderate-risk CAP without inpatient care: €867-€866 Per case cost of high-risk CAP without inpatient care: €985-€984
Wolff 2020 [12] (Sweden)	Cohort Model: Assessed lifetime risks, consequences, and costs of IPD and CAP associated with PPV23 or PCV13 vaccination for cohort aged 65–75	5.8 days in the general ward per CAP case at €603 per day 0.25 days in the intensive care unit per CAP case at €1,728 per day
Naoum 2020 [25] (Greece)	Retrospective analysis of electronic medical records	Mean cost per hospitalized CAP patient was €7,407 Mean cost per outpatient CAP patient was €111
Dupuis 2021 [26] (France)	Retrospective claims database analysis of hospitalized CAP patients	Cost of initial hospitalization, 1-year survivors: €18,613 Cost of initial hospitalization, 1-year decedents: €19,834 Cost of subsequent hospitalizations, 1-year survivors: €19,834 Cost of subsequent hospitalizations, 1-year decedents: €14,815
Antunes 2020 [27] (Portugal)	Retrospective claims database analysis of hospitalized CAP patients	Mean cost per hospitalized CAP patient was €2,628
Kaier 2019 [23] (Germany)	Retrospective claims database analysis of hospitalized CAP patients	Mean CAP case cost, CAP not main diagnosis: €26,955 Mean CAP case cost, CAP as main diagnosis: €14,066

CAP impact on quality of life was the least frequently reported outcome among the retained studies. This included only one study by Andrade et al. (2018) (a prospective cohort study among French citizens with CAP) that objectively assessed quality of life. The study applied the EQ-5D to assess quality of life longitudinally for up to 360 days in patients treated for CAP and stratified by baseline age. Overall, quality of life improved as time since CAP treatment increased, with values at 30 days (up to 0.62) being lowest and values at 360 days (up to 0.82) being highest [28]. This temporal trend of increasing quality of life with longer time since CAP treatment was consistent across age groups. The proportion of patients reporting moderate-extreme problems in individual quality of life domains (Figure 5) indicated a high burden within 30 days of CAP diagnosis, but the burden waned for all domains by 12 months of follow-up. The most severe domain-level burdens were observed in Usual Activities and Mobility, and worse overall patient utility was most strongly associated with longer hospital stay, increased age, and positive smoking status [28].

Figure 5. Quality of life domain scores after CAP diagnosis.

*Adapted from a report by Andrade et al. (2017).

4. Discussion

We sought to provide an update of aggregated information pertaining to the clinical, economic, and humanistic impact of CAP in selected European countries published from 2011 to 2021. The burden of CAP has remained high for incidence, length of hospital stay, mortality, and associated direct medical costs, particularly in older persons aged ≥65 years who are more susceptible to contracting CAP and to the disease itself due to age-related factors accumulated comorbidities (e.g. COPD, diabetes, CVD) and antibiotic resistance.

Though with the most updated information, the present findings are consistent with those of previous key reviews: the relevant details of which are reported in Table 3. Herein, incidence for inpatient- and outpatient-treated CAP was 50–2,940 per 100,000 and 45–2,380 per 100,000 depending primarily on age and comorbidities. The scope of the review by Torres et al. (2018) included studies published from 2000 to 2016 and reported overall CAP incidence ranging from 68 to 7,000 per 100,000 and 16 to 3,581 per 100,000 for hospitalized cases [30]. The difference between Torres et al. (2018) and the present review (which included studies up to the year 2021) suggests – at least in terms of the volume of inpatient cases – that the burden of CAP has decreased somewhat in recent years. However, the review by Torres et al. (2018) interestingly noted that the burden of pneumonia in Europe may have been underestimated due to a lack of consideration of CAP in addition to invasive pneumococcal disease. Interestingly, recent population-level mitigation strategies against the spread of COVID-19 might further reduce CAP incidence in the most recent studies versus those from the pre-pandemic period [32]. Consistent with the previous reviews [8,30,31] S. pneumoniae infection was consistently the most common CAP etiology regardless of treatment setting.

Table 3. Previous reviews of CAP burden in Europe.

Study	Years Covered	Key Findings
Welte et al. (2012) [8]	1990 to 2007	Incidence varied by country, age, gender; higher in persons aged ≥65 years old and in men Streptococcus pneumoniae was the most common etiology Mortality (<1% to 48%) was associated with older age, co-morbidities, and CAP severity CAP was associated with high rates of hospitalization and length of hospital stay
Alimi et al. (2017) [29]	1999 to 2016	25% of CAP cases in Europe had a viral etiology 13% of CAP cases were due specifically to influenza, rhinoviruses, and coronaviruses
Torres et al. (2018) [30]	2000 to 2016	Overall incidence was 68–7000 per 100,000; hospitalized cases: 16–3581 per 100,000 Incidence increased with age and (for pneumococcal CAP) comorbidities Most cases (30%–78%) were caused by serotypes covered by PCV13 vaccine
Ferreira-Coimbra et al. (2020) [31]*	Not Reported	Incidence varied considerably among countries, but direct assessment is hindered by differences in reporting and case selection from epidemiological studies CAP is responsible for at least 23,000 deaths annually in Europe Streptococcus pneumoniae was the most common etiology in all treatment settings (outpatient, general ward, and ICU)

*Ferreira-Coimbra et al. (2020) was a comprehensive qualitative synthesis and not a systematic review. The study also included non-European countries.

We expanded on the scope of previous key reviews by summarizing the economic and humanistic burden of CAP. In this respect, we identified few studies that assessed direct medical costs, non-medical costs, or quality of life among CAP patients or their caregivers. Though there were many studies that assessed the economic impact of pneumonia, few focused on CAP specifically or were focused on invasive pneumococcal disease, therefore excluding CAP. Consistent with recent observations, no recent studies have assessed the impact of CAP on non-medical costs (e.g. lost work productivity) [33]. We identified four retrospective analyses and three cohort models. Commonly reported across these studies included the primacy of hospitalization utilization as the strongest driver of CAP-associated direct medical costs, and in particular, admission to the intensive care unit [12]. Quality of life evidence specifically for confirmed CAP patients was quite limited. The lone analysis assessed the evolution of quality of life after CAP discharge using the EQ-5D and demonstrated that quality of life deterioration after discharge from a CAP-associated hospitalization was most pronounced among the elderly and specifically those with a high comorbidity burden and with a previous history of pneumonia [28]. The analysis by Andrade et al. (2018) was also unique in its reporting of health state utility values that will facilitate pharmacoeconomic modeling in CAP.

Though the present descriptive synthesis did not include formal analysis or meta-analysis, the methods and assumptions employed here are susceptible to limitations common to any evidence synthesis. First, this study does not include formal analysis and therefore did not employ analytic methods to limit the impact of bias on aggregated outcomes. The choice to not include a formal analysis was based on the variability in study designs, case definitions, and other study characteristics that – in our opinion – would render a formal analysis untenable due excessive heterogeneity. This issue is particularly pronounced for case definitions of infection and of ‘high risk.’ We recommend that these issues be explored in subsequent research so that aggregate estimates can be reliably derived. Second, the search strategy that has inherent limitations. Most notably, the search strategy was limited to studies published after the year 2010. This was an attempt to limit information that is no longer relevant, but it is possible that relevant information (not otherwise accounted for in more recent publications) was published before the year 2010. Other limits pertaining to the study design (e.g. excluding prospective interventional studies and clinical trials) may have also excluded potentially relevant studies. Third, the search strategy was populated with terms that target the specific information being sought with respect to clinical indication, study design, outcome, and geographic scope. It is possible that some relevant studies were not captured by our search strategy because they might not have included the specific terms being sought through our search. In a related sense, we did not apply specific definitions of CAP to identify studies with consistent definitions. Rather, we included studies of CAP regardless of those definitions provided that they pertained to infections acquired outside the hospital setting. However, it is possible that our findings could have been affected by variability in the definitions of CAP. Fourth, this review excluded interventional studies and clinical trials because such studies often apply patient selection criteria that would have otherwise skewed our findings and/or rendered our findings not generalizable to broader, real-world populations sought in this review. However, such interventional studies might offer meaningful information regarding mortality and hospitalization and should perhaps be considered in a subsequent review. Lastly, an important limitation noted in the previous CAP literature [2,30] and encountered here is inconsistent methods of surveillance and reporting across studies and countries. For example, the interpretation of data pertaining to the frequency of pathogens causing CAP was here – and has been previously noted [2] - complicated by a lack of reporting of the number of samples submitted for testing in addition to the number of samples positive for a given pathogen. Likewise, the aggregation of epidemiologic information continues to be hindered by methodological inconsistencies among studies such as application of multiple diagnostic coding schemes with varying sensitivity to identifying CAP, reliance on administrative data in general, a lack of consistency with respect inclusion of CAP with a virologic etiology, and difficulties in distinguishing community-acquired from institution-acquired pneumonia.

The clinical, economic, and humanistic burden of CAP persists in the selected European countries particularly among the elderly given often accompanying chronic comorbidities and frailty [34]. Additional factors such as malnutrition, combinations of comorbidities, and even seasonal patterns exacerbate this burden [35,36:] as does the presence of SARS-CoV-2 as the causative pathogen, which itself is associated with higher mortality and healthcare resource utilization [37]. Importantly, the review does not cover the literature pertaining to recent pandemics (i.e. H1N1 and COVID). Our findings highlight a need for comprehensive vaccination strategies, treatments, and surveillance to mitigate the clinical, economic, and humanistic burden of CAP. This lingering burden belies advances in effective public health strategies that include vaccination and surveillance. However, a better understanding of CAP and its consequences is still needed specifically with respect to the economic and quality of life burden on CAP patients and their caregivers. Further study relying on population-level real-world data incorporating comprehensive etiological interrogation and employing consistent disease definitions is recommended.

Declaration of interest

E Tsoumani, S Salomonsson, and G Bencina are employees of MSD. J Carter and J Stephens are employees of OPEN Health, which received funding from MSD to conduct this study. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have received an honorarium from Expert Review of Vaccines for their review work. One of these reviewers has disclosed that they provided scientific consultancies for protocol preparation and data analyses, for Pfizer, GSK, MSD. The remaining reviewers have no other relevant financial relationships or otherwise to disclose.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/14760584.2023.2261785

Additional information

Funding

This paper was funded by Merck & Co.