The population consisted of adults with ILI presenting at ambulatory care centres in Brazil, modeled under different scenarios of expected influenza (15 %, 20 % and 25 %) and COVID-19 (5 % and 10 %), to reflect potential levels of underlying disease in an ILI cohort. The influenza upper limit of 25 % reflects the peak of the January 2022 H3N2 outbreak in Brazil.13 The base case analysis assumed 20 % underlying influenza prevalence combined with 5 % COVID-19.

The intervention in this analysis is a combination antigen RDT for COVID-19, influenza A and influenza B (Panbio™ Flu/COVID-19 Rapid Panel, Abbott Rapid Diagnostics Jena GmbH). Studies have shown that presumptive clinician diagnosis is standard practice for influenza;[2,3] therefore, the comparator used in this setting was usual care defined as a COVID-19 RDT and clinical judgment for influenza (i.e., no diagnostic testing for influenza).

Outcomes included accuracy of diagnosis and expected Health Care Resource Utilization (HCRU), consisting of hospital bed days and Intensive Care Unit (ICU) occupancy. An additional analysis explored inappropriate use of antibiotics. Outcomes are reported at the cohort level and are presented as disaggregated outcomes (i.e., as a cost-consequence analysis). A Brazilian private health insurer perspective was used for the analysis.

A dual decision tree model was developed in Microsoft Excel to capture multiple outcomes and track parallel testing for COVID-19 and influenza in this population.

A structured literature search was carried out in PubMed to identify existing economic evaluations of COVID-19 and influenza diagnostic tests to inform the model structure. Studies were included if they reported a simple replicable model; more complex simulation models were not reviewed. Four studies/model structures were shortlisted for review.17-20 These provided clear structures for test-treat modeling in COVID-19 and influenza; however, no studies were retrieved which tracked dual diagnosis in an ILI population, a key component of evaluating the use of a dual RDT. Therefore, a de novo model was developed incorporating the ability to include diagnoses for both COVID-19 and influenza. The key premise of the analysis followed the published literature, in that timely antiviral treatment is assumed to positively impact subsequent resource utilization. Following You et al. (one of the only models identified that evaluated the ambulatory setting),19 resource utilization included hospitalizations, ICU admissions, and associated length of stay. The costs of antiviral medications were not included in this analysis.

The model structure is provided in Fig. 1. Individuals are classified according to their true disease status and results of the combined RDT or standard practice diagnosis. ILI patients can have influenza, COVID-19, COVID-19/influenza co-infection, or neither disease (ILI only). Outcomes are dependent on the accuracy of the test/clinical judgment and underlying disease status. Patients with a positive diagnosis (true or false) receive antiviral treatment (assumed not to be reimbursed under private insurance) while patients with a negative diagnosis do not receive treatment. The receipt of timely treatment is assumed to reduce the likelihood of resource utilization and associated costs. This assumption is in line with the approach taken in previous models.17-20

Fig. 1.

Model Structure.

(0.14MB).

*Antibiotic use was not included in the base case scenario but was assessed in an additional analysis. Abbreviations: COVID-19, Coronavirus Disease 2019, ICU, Intensive Care Unit, ILI, Influenza-Like Illness.

Although the intervention combination RDT provides separate test results for influenza A and influenza B, as influenza A accounted for the majority of cases in Brazil in 2022,21 a simplifying assumption was made to only consider influenza A in the analysis.

Model inputs were defined based on a pragmatic review of the literature. Local data for Brazil were used where available, followed by region-specific data where possible. Table 1 outlines the parameter base case values and ranges used in the sensitivity analyses. The base case analysis assumed an influenza prevalence of 20 % and a COVID-19 prevalence of 5 %.

Inputs for test accuracy were taken from the combination diagnostic test package insert and published sources for presumptive diagnosis.2,22 The base case analysis assumed that diagnostic accuracy for COVID-19 RDTs were identical across the individual COVID-19 test and COVID-19/influenza combination test, effectively modeling the impact of testing vs. clinical judgment for influenza. No confirmatory testing was assumed for either COVID-19 or influenza.

Inputs for disease management were estimated from the literature and are reported separately for influenza and COVID-19. Co-infection inputs were estimated as a function of COVID-19 based on data reported in Alosaimi et al.23 where the likelihood of mortality for coinfection was estimated at 1.78 times that of COVID-19.

Unit costs for influenza hospitalizations and ICU stays were estimated in local currency using costs from the Brazilian public healthcare system (SUS) as a benchmark with an inflator used to approximate the private insurance market. This inflator was set to 2.8 based on the default used in a recent model with a private insurance perspective in Brazil.24 Unit costs for COVID-19 hospitalizations and ICU stays were obtained from a recent study reporting Brazilian private payer costs.25 Costs for influenza antiviral treatment were not included as most oral medications are not reimbursed by Brazilian private insurance plans. Costs were estimated over a short-term time horizon reflecting the acute nature of COVID-19 or influenza infection and are reported in 2022 US dollars (using an exchange rate of 0.2 USD to 1 BRL).26 No discounting was applied.

Scenario analysis was performed to assess the impact of underlying influenza and COVID-19 prevalence on the model outcomes and to simulate situations where influenza peaks coincide with an increase in COVID-19. This was particularly important given the seasonal variation to influenza prevalence, and that the relative prevalence of COVID-19 compared to influenza for future seasons is uncertain. Influenza prevalence was varied between 15 % and 25 % and COVID-19 prevalence varied between 5 % and 10 %.

The impact on antibiotic prescribing was also explored, as prior studies have indicated that rapid testing may reduce inappropriate antibiotic use (i.e., antibiotics in influenza-positive individuals).27-31 A simple risk reduction was calculated from a systematic review of the impact of influenza rapid testing on antibiotic use[29] and combined with an estimate of baseline levels of antibiotic prescribing to explore the impact of rapid testing on antibiotic use for influenza. Antibiotic prescription costs were included as a proxy for potential patient-out-of-pocket expenses.

One-Way Sensitivity Analyses (OWSA) were conducted on key model inputs to test whether the model outputs were sensitive to variations in model inputs. Parameter values were varied by ±20 %, unless logically bounded or sources clearly specified the 95 % Confidence Intervals. Results are reported as a tornado diagram in the supplementary appendix.

The diagnostic testing market in Brazil has been very dynamic, with multiple manufacturers and rapidly changing pricing. The impact of the relative pricing of combination RDT (compared to COVID-19 RDT) was explored to look at the impact of potential fluctuations.

The results of the base case analysis are reported in Table 2. Use of the combination RDT resulted in reductions in over-diagnosis (false positives; 0 in the combination RDT arm vs. 174 in the standard practice arm) and missed diagnosis (false negatives) (22 for combination RDT vs. 139 for standard practice). Hospital and ICU days were lower in the combination testing arm than the standard practice arm, due to fewer false negative diagnoses of influenza. The combination RDT arm resulted in 53.2 hospital days per 1000 people tested compared to 78.8 days in the standard practice arm, or a relative reduction of 32 %. ICU days decreased from 14.99 in the standard practice arm to 10.3 in the combination RDT arm, or a 24 % relative reduction.

The cost of diagnostic testing increased in the combination testing arm, reflecting the higher price of the combination test ($7.6) vs. a standalone COVID-19 test ($3.8). These costs were offset by reductions in HCRU related costs (HCRU costs estimated at $22,254 for combination RDT testing vs. $26,153 for standard practice), resulting in a cost saving of $99.

Fig. 2 provides results of the scenario analyses. Under all examined scenarios, use of combination RDT testing decreased HCRU compared to standard practice. The impact on hospital bed days ranged from a reduction of 20.3 bed days per 1000 people (15 % influenza prevalence and 5 % COVID-19 prevalence) to a reduction of 35.4 bed days per 1000 people (25 % influenza prevalence and 10 % COVID-19 prevalence). Comparable results were seen for ICU data, with lower prevalence settings resulting in a decrease of 3.9 ICU days per 1000 people and higher prevalence settings in a decrease of 6.8 ICU days per 1000 people. The positive impact on HCRU offset the additional diagnostic costs in all scenarios where influenza prevalence was equal to or greater than 20 % (four of the six scenarios). A full tabulation of results is provided in the supplementary materials (Table S1).

Fig. 2.

Scenario analyses - impact of influenza and COVID-19 prevalence on HCRU.

(0.1MB).

The OWSA (reported in the Supplemental Appendix) indicates that under current model assumptions, model outputs were most sensitive to the cost of the combination RDT diagnostic and the extent to which early diagnosis of influenza is assumed to impact subsequent HCRU.

Fig. 3 illustrates how the comparative pricing of a combination RDT relative to an individual COVID-19 RDT would impact results. Our base case found that a combination RDT is cost-saving at twice the price of an individual COVID-19 RDT (100 % increase, cost savings of $99). The figure shows that relative price fluctuations would have a substantial impact on savings.

Fig. 3.

Sensitivity Analysis: Fluctuation in RDT relative costs.

(0.15MB).

In the additional analysis where the potential impact of RDT use on prescribing of antibiotics was explored, standard practice (i.e., no RDT for influenza) was estimated to result in 275 unnecessary antibiotic prescriptions in a cohort of 1000 ILI adults, and use of an RDT (in this case the combination RDT) could reduce this to 183 prescriptions. Using a provisional cost of $6 per prescription (equivalent to 30 BRL), this could result in additional savings of $554 (Table 2).