Existing cost-effectiveness analyses for diseases caused by Group A Streptococcus: A systematic review to guide future research [version 1; peer review: awaiting peer review]

Background: Group A Streptococcus (Strep A) causes a broad spectrum of disease manifestations, ranging from benign symptoms including throat or skin infections, to fatal illness such as rheumatic heart disease, or chronic renal failure. Currently, there is no vaccine available against Strep A infections. Despite the high burden of Strep A-associated infections worldwide, little attention has been paid to the research of these diseases, including standardized surveillance programs, resulting in a lack of economic evaluations for prevention efforts. This study aims at identifying existing cost-effectiveness analyses (CEA) on any Strep A infections. Methods: A systematic literature review was conducted by searching the PubMed electronic database. Results: Of a total of 321, 44 articles met the criteria for inclusion. Overall, CEA studies on Strep A remain limited in number. In particular, a number of available CEA studies on Strep A are disproportionately lower in low-income countries than in high-income countries. Decision-analytic models were the most popular choice for CEA on Strep A. A majority of the models considered pharyngitis and acute rheumatic fever, but it was rare to observe a model which covered a wide range of disease manifestations. Conclusions: Future research is needed to address missing clinical outcomes, imbalance on study locations by income group, and the transmission dynamic of selected diseases.


Introduction
Group A Streptococcus (Strep A), also known as Streptococcus pyogenes (S. pyogenes) is a Gram-positive bacterium, often identified in the throat or on the skin. Strep A is a major public health concern causing significant morbidity and mortality worldwide. While the World Health Organization (WHO) prioritized Group A Strep vaccine development in 2014, there are no vaccines available. Strep A infections include a broad spectrum of diseases. Relatively minor infections can be a precursor for acute and invasive diseases, both of which can lead to long-term morbidity. Acute conditions include throat and skin diseases, as well as toxin-mediated diseases. If relatively benign infections (i.e. Strep throat or skin infections) are not properly treated, the infection may further develop into post infectious autoimmune diseases (i.e. acute rheumatic fever (ARF), glomerulonephritis), which can lead to chronic diseases such as rheumatic heart disease (RHD) and chronic renal failure.
The absolute numbers of episodes of Strep A throat infections and skin infections are much higher than those associated with more severe illness 1 . This raises the possibility that although the symptoms of pharyngitis or skin infections may not be as severe as the ones of ARF or RHD, their economic and social burdens at the population level could be noticeably high considering direct treatment costs and indirect costs 2,3 .
While vaccines against Strep A are absent, the use of antibiotics such as oral or intramuscular penicillin has proved effective and been recommended to treat patients with Strep A infections. In addition, several prevention strategies were developed. Primary prevention of ARF involves the detection and timely treatment of streptococcal pharyngitis 4,5 . In order to identify patients with Strep A infections, the following test options are often considered: clinical diagnosis (with or without the use of clinical scoring algorithms), throat or skin cultures, and point of care testing (e.g., rapid Strep A antigen detection test and nucleic acid amplification tests). Once confirmed positive, antibiotics are prescribed to treat patients. However, this process may result in prescription and consumption of unnecessary antibiotics for those tests who are false positives or in missing true cases (false negative). While this primary prevention strategy is effective, this approach is also costly considering only 10-20% of pharyngitis is caused by Strep A 5

.
A study estimating the rates of inappropriate outpatient antibiotic use in the United States reported that 56.2% and 72.4% of ambulatory care visits for pharyngitis were associated with antibiotic prescribing in children (0-19 years) and adults (20-64 years), respectively. However, streptococcal prevalence for pharyngitis was 37% for children and 18% for adults, indicating the existence of inappropriate use of antibiotics in pharyngitis treatment 6 . The secondary prevention strategy is to use intramuscular antibiotics as a prophylaxis to prevent recurrent ARF, but it was shown that increasing patients' compliance to the recommended schedule of injections occurring every 28 days over a minimum of 10 years is challenging 5 . The tertiary prevention scenario involves increasing the coverage of valve surgery by building local surgical capacity especially in resource-limited settings 4 . While the tertiary approach will enhance local health capacity to treat patients with severe illness, this process will require a long-term plan and consensus among decision makers who need to consider various competing health problems in a nation.
Few health economic studies have been conducted on the entire spectrum of Strep A diseases. This may be due in part to the significant reduction in the rates of autoimmune diseases (i.e., acute rheumatic fever) and its sequelae (i.e., RHD) in high-income countries (HIC) during the late 20 th century. The reduction was mainly attributable to improvements in socioeconomic conditions and to the increase in the use of antibiotics [7][8][9] . However, the diseases are still highly prevalent in lower-and middle-income countries 10 , and the burden of Strep A throat or skin infections is not negligible in HIC either 11 . It is also worth noting that Strep A causes a wide range of disease manifestations, and there is a lack of available data points for each disease category, making it difficult to establish a universal model that covers all symptoms which progress over time.
The primary interest of the current review lies in identifying existing health economic models (i.e., those used in costeffectiveness analyses) for Strep A-associated diseases. This review aims to summarize the types of model structures and evaluation perspectives which have been frequently emphasized by others, as well as to identify the gap in existing literature.

Methods
A systematic literature review was conducted by searching the PubMed electronic database. Search terms were divided into two groups and developed separately: (1) disease category and (2) economic evaluation terminology category. Considering that Strep A causes a broad spectrum of diseases from seemingly benign throat and skin infections to chronic RHD or CHF, search strategies in the disease category closely followed the terms previously defined in texts on the burden of group A streptococcal diseases 1,12,13 . For the current search, "group A streptococcus" was additionally included as a separate search term to expand the search to papers which did not mention the specific names of Strep A-related diseases. In addition to the disease category, all search terms related to health-economic evaluation were developed in the economic evaluation category. Table 1 summarizes the search terms used for the current review.
All lines of the search terms in each of the two categories were combined using "OR", and the two categories were eventually joined by "AND" in order to identify papers associated with health-economic models for Strep A-related diseases. The search terms were not limited to a title or an abstract. Any papers published up to May 2020 were included, and papers written in non-English languages were excluded. Additional search was done by going through bibliographies for eligible articles. The initial screening was carried out by going through all abstracts and shortlisting the papers that indicated the use of a CEA in the economic analyses. For the shortlisted papers, a more comprehensive (full-length) assessment was carried out by reviewing the full text against a list of inclusion and exclusion criteria. The search was performed independently by JSL and cross-checked by JSL and SK. Any discrepancies were discussed and resolved between the two independent reviewers.
Given that the studies were conducted under varying conditions (i.e. different country-contexts, study designs, disease types), a standardized set of criteria would be useful to make systematic assessments among the articles identified at the final stage of the literature review. First, disease category was defined. Sanyahumbi et al. (2016) previously categorized Strep Arelated diseases into four groups: superficial and locally invasive disease, immune-mediated disease, disease sequelae, and invasive-and toxin-mediated disease 1 . The same categorization was applied for the current study. Second, given that Strep A causes a wide range of disease presentations, age groups chosen for an intervention may also differ depending on disease types and the peak incidence of a disease. Thus, target cohorts were also identified. Third, given that model structure is one of the key factors that determine the final outcome of an intervention (i.e. cost-effectiveness strategy), the types of health economic models were compared. Fourth, cost perspective was identified. For the current review, any costs related to healthcare costs such as drug, hospitalization, treatment, etc. were termed "health system perspective". On the other hand, any studies which considered broader cost items such as healthcare costs, productivity losses, caregiving, etc. were defined as "societal perspective" 14 . Fifth, a CEA model often compares total costs with intervention benefits which can be measured in various ways. For example, while some studies use the Quality-Adjusted Life Year (or Day) (QALY(D)) as an outcome measure, others adopt the Disability-Adjusted Life Year (DALY).
In addition, there are studies which directly utilize the number of episodes prevented by converting into saved costs. Hence, outcome measure was described for each study. Lastly, while some studies calculated cost-effectiveness based on primary data sources obtained from a trial, many studies estimated cost-effectiveness outcomes by constructing a decision analytic model. Given that such a model often utilizes multiple health states, more details on health states were further investigated for these studies.

Results
The initial search using the key words identified 321 articles from the database, as shown in Figure 1. After going through the abstracts and titles, 274 articles were omitted, resulting in 47 articles for a more comprehensive review. The full-length assessment was carried out for these articles. Of the 47 articles, nine studies were further excluded, and six articles were additionally identified through the bibliography search of the eligible articles. A total of 44 articles were selected at the final stage of the current literature review search. These final papers were assessed based upon the six criteria described above. Table 2 summarizes the final 44 articles identified by the systematic literature review. The majority of studies (93%) were done in countries classified as high-income or uppermiddle-income by the World Bank 15 . There were only three studies that were carried out in lower-middle-income economies or below: two studies from Africa 4,5 and one study from India 16 . While most studies conducted a cost-effectiveness analysis for a single country or sub-population of a country, Watkins et al. 4 and Manji et al. 5 covered multiple African countries by taking into account evidence reported in existing literature. About 25% of the studies (n = 11) solely considered superficial diseases such as throat or skin infections. Among those 11 studies, six of them were not Strep A-specific but more general, resulting in only five studies with a specific focus on Strep A. Another 34% of the studies (n = 15) included immune-mediated-(i.e. acute rheumatic fever) or locally invasive diseases (i.e. peritonsillar abscess) in addition to superficial diseases. As described above, Strep A causes a broad spectrum of diseases from benign superficial infections to severe cardiac failures. Five studies investigated disease sequelae (i.e. RHD) along with superficial-and immune-mediateddiseases, and two studies further included locally invasive diseases on top of these. There was only one study that covered at least a subset of each of the four disease categories: superficial and locally invasive, immune-mediated, sequelae, and invasive-and toxin-mediated diseases.
Forty-one percent and 43% of the studies applied a health system perspective and a societal perspective, respectively. Of the 44 articles, five studies adopted both perspectives. For example, the health system perspective was chosen for a primary analysis, but the societal perspective was also considered as a sensitivity or scenario analysis. A total of 18 studies used health-related quality of life such as QALY(D) or DALY as an outcome measure, and others directly utilized the number of episodes prevented or the number of patients free of recurrence which were in most cases converted into saved costs. While some studies conducted cost-effectiveness analyses alongside (randomized) clinical trials (n = 7) or simple comparisons between costs and benefits (n = 8), the majority of the studies (66%) used decision analytic models. Among the studies with decision analytic models, 72% of them (n = 21) adopted decision tree models, and eight studies employed Markov models.
Since decision analytic models take into account multiple health states and transition probabilities from one health state to another, more details on health states were further investigated as shown in Table 3. The most common health states chosen for the models were Strep A pharyngitis and ARF    followed by death, superficial infections prior to Strep A confirmation, and allergic reactions due to antibiotics. Health states such as RHD, and suppurative complications were also moderately selected. Six models included a health state of recurrent ARF. Eight models took account of severe RHD or other manifestations of cardiovascular disease. It should be noted that the inclusion of this health state is relatively new, reflecting that six of these models have been developed since 2015.
It was rare to observe models that included Strep A skin infection, acute post-streptococcal glomerulonephritis (APSGN), or invasive and toxin-mediated diseases, showing that there was only one model with each of these health states. This rarity reflects the complex nature of Strep A infections and implies the limited number of surveillance data points for each health state.

Discussion
The current review focuses on the identification of existing CEA studies on Strep A infection. Given a wide range of disease presentations caused by Strep A, a large variation exists across the identified models in terms of disease types covered in the studies. Overall, CEA studies on Strep A remain limited in number. In particular, the number of the existing studies was highly concentrated in upper-middle income countries or higher, and there were only four studies focused on lower-middle income countries or below. This is problematic because the burden of more severe illnesses caused by Strep A (i.e. RHD and CHF) has been greatly reduced in advanced countries but remains disproportionately high in developing countries 7,8,56 . This does not mean that high income economies are free of Strep A. Superficial diseases such as pharyngitis or impetigo are sometimes thought of as small-time players compared to the ensuing diseases that cause more severe illness. However, pharyngitis is one of the most common diseases observed globally, including in more advanced countries. In addition, there is a growing concern that Strep A skin infections may play a significant role in developing ARF 3,56  In the context of health economic models on Strep A-associated diseases, cohort-level models have been widely used by incorporating varying health-states and predicting disease progression among patient groups. While there were more decision tree models observed than Markov models in this review, a decision tree model may not be appropriate when dealing with the long-term progression of diseases and treatment effects. On the other hand, a Markov-cohort model is suitable for chronic diseases because the model can incorporate repetitive cycles. With the Markov-cohort model, events are considered stochastic processes over time, allowing to evaluate costs and effects of intervention strategies over a long time period 5 . However, Markov-cohort models are limited to the lack of memory when transitioning from one health state to another (i.e. Markovian assumption). While this property can be circumvented by setting up temporary tunnel states, this procedure results in a more complex model due to dividing one health state into multiple sub health states. Patient-level (or microsimulation) models can improve the drawback of the cohort model as patient-level models follow an individual trajectory across multiple health states. However, this type of microsimulation model often requires a high level of computational power, more input parameters, and detailed data sources at the individual-level, which is often challenging in resource-constrained settings.
While health officials have implemented various intervention strategies (i.e., primary, secondary, and/or tertiary prevention strategies) to reduce the burden of the diseases associated with Strep A, the existing control strategies almost always involve the use of antibiotics. The use of oral or intramuscular penicillin has proved effective in reducing the disease progression and treating rheumatic fever. However, it should be noted that the use of such drugs may also cause allergic reactions such as rash, anaphylaxis, or sometimes, death. In addition, antibiotics can be unnecessarily prescribed to patients who are false positive or carriers, which may contribute to the increasing trend of antimicrobial resistance. Thus, there is no doubt that preventive measures such as a safe vaccine will reduce the concerns raised by excessive antibiotic uses [57][58][59][60][61] . Currently, there is no vaccine available for Strep A infections. The development of safe, efficacious, and affordable vaccines may open a new era to control Strep A infections in a more effective manner. In other words, with a vaccine that protects populations from contracting superficial Strep A infections, vaccination will likely limit chances for benign symptoms to be developed further into more severe illnesses such as autoimmune diseases or its sequelae, and reduce not only the burden of a broad spectrum of the Strep A diseases but also antimicrobial resistance.

Data availability Underlying data
No data are associated with this article.