A cost-consequence analysis of the Xpert Xpress CoV-2/Flu/RSV plus test strategy for the diagnosis of influenza-like illnesses

Abstract

Aims

Influenza-like illnesses (ILI) affect millions each year in the United States (US). Determining definitively the cause of symptoms is important for patient management. Xpert Xpress CoV-2/Flu/RSV plus (Xpert Xpress) is a rapid, point-of-care (POC), multiplex real-time polymerase chain reaction (RT-PCR) test intended for the simultaneous qualitative detection and differentiation of SARS-CoV-2, influenza A/B, and respiratory syncytial virus (RSV). The objective of our analysis was to develop a cost-consequence model (CCM) demonstrating the clinico-economic impacts of implementing PCR testing with Xpert Xpress compared to current testing strategies.

Methods

A decision tree model, with a 1-year time horizon, was used to compare testing with Xpert Xpress alone to antigen POC testing and send-out PCR strategies in the US outpatient setting from a payer perspective. A hypothetical cohort of 1,000,000 members was modeled, a portion of whom develop symptomatic ILIs and present to an outpatient care facility. Our main outcome measure is cost per correct treatment course.

Results

The total cost per correct treatment course was $1,131 for the Xpert Xpress strategy compared with a range of $3,560 to $5,449 in comparators. POC antigen testing strategies cost more, on average, than PCR strategies.

Limitations

Simplifying model assumptions were used to allow for modeling ease. In clinical practice, treatment options, costs, and diagnostic test sensitivity and specificity may differ from what is included in the model. Additionally, the most recent incidence and prevalence data was used within the model, which is not reflective of historical averages due to the SARS-CoV-2 pandemic.

Conclusion

The Xpert Xpress CoV-2/Flu/RSV plus test allows for rapid and accurate diagnostic results, leading to reductions in testing costs and downstream healthcare resource utilization compared to other testing strategies. Compared to POC antigen testing strategies, PCR strategies were more efficient due to improved diagnostic accuracy and reduced use of confirmatory testing.

Keywords:

• Influenza-like illness
• diagnostics
• PCR
• costs
• cost-consequence

JEL CLASSIFICATION CODES:

• I10
• I1
• I
• I19

1. Introduction

Influenza-like illness (ILI) is an umbrella term describing a basket of individual infections that manifest a fever of 100 degrees Fahrenheit or higher, a cough, and/or a sore throat affecting between 9 and 49 million people every year in the United States (US)^1,2. Influenza, Sars-CoV-2, and respiratory syncytial virus (RSV) are among the viruses resulting in an ILI classification; determining definitively which pathogen is the cause of symptoms in a particular case is important for patient management and can be challenging^1,3. Coinfections have been observed between influenza, SARS-CoV-2, and RSV, while not comprehensively studied in the literature⁴, further complicate patient management^5–7.

According to the US Centers for Disease Control and Prevention (CDC), from October 1, 2022 through April 30 2023, there were an estimated 27–54 million influenza cases, 12–26 million influenza-related medical visits, 300,000–650,000 influenza-related hospitalizations, and 19,000–58,000 influenza deaths⁸.

In the 2020–2021 season, there were an estimated 146.6 million total SARS-CoV-2 infections (as high as 40% of the US population), of which 124 million were symptomatic. This resulted in approximately 7.5 million hospitalizations and 921,000 deaths⁹. Since then, rates of illness have decreased significantly. Currently, SARS-CoV-2 is estimated to affect approximately 6% of the population annually¹⁰.

RSV is a common respiratory illnesses, resulting in 2.1 million outpatient visits, 58,000–80,000 hospitalizations, and 100–300 deaths among children younger than 5 each year¹¹. In adults 65 and older RSV results in 60,000–160,000 hospitalizations and 6,000–10,000 deaths each year.¹¹ In the US, RSV accounts for approximately one quarter of hospitalizations for lower respiratory tract infections for children under 5 years old. Furthermore, RSV is also the leading viral cause of death among children under the age of 5 in the U.S.¹² RSV in nursing homes also poses a threat to older adults, with an incidence rate of 5–10 infections per 100 person-years and a 2–5% annual risk of death¹³. In a systematic review of infections occurring among adults in long-term care, RSV incidence per year ranged from 1.1 to 13.5%, whereas influenza incidence rates per year ranged from 5.9 to 85.2%¹⁴.

Historically, respiratory season in the US was associated with increased rates of respiratory illnesses in the Fall and Winter, but off-season illness circulation differs from year to year, by location, and may be impacted by climate change^15–17. Additionally, patients with two or more concurrent infections may present complex diagnostic and treatment challenges^5–7. While many people experience these illnesses with mild symptoms, other patients could incur severe medical sequelae and stand to gain meaningful benefits from prompt clinical management and treatment. Early and accurate detection, as well as appropriate treatment of viral respiratory infection, reduces downstream clinical sequelae and healthcare resource utilization (HRU)¹⁸.

Currently, there are a variety of different testing strategies available for identification of the cause of ILI that fall into broad categories including: point-of-care (POC) or send-out. POC testing encompasses tests that are collected and analyzed near the patient (bedside, in office, within-facility laboratory) and typically produce results during the patient’s visit, allowing for prompt medical action. Send-out tests are collected and sent out to a central laboratory, apart from the facility in which the sample was obtained. Therefore, send-out tests typically take longer to produce results than POC tests (24 h or more)^19,20. This delay may result in missing the prescribed treatment window of oseltamivir and nirmatrelvir/ritonavir, for influenza and SARS-CoV-2 infections, respectively. However, certain types of POC tests, like antigen testing, are less accurate than send-out PCR tests, potentially leading to missed or incorrect diagnoses^21,22. Additionally, with both methods, unless a panel is ordered (influenza A/B, SARS-CoV-2, and RSV), RSV may be overlooked and undetected by many testing strategies leading to potential secondary cases and additional burden²³.

Advancements in diagnostic instruments and technology seek to provide a POC system that provides a comprehensive and highly accurate test result. One such diagnostic tool, the Xpert Xpress CoV-2/Flu/RSV plus test, is a rapid, POC, multiplexed real-time RT-PCR test intended for the simultaneous qualitative detection and differentiation of SARS-CoV-2, influenza A, influenza B, and RSV²⁴. Given that the Xpert Xpress CoV-2/Flu/RSV plus yields qualitative detection and differentiation of each virus, it produces a separate identification and result (positive or negative) for each virus in the assay.

In this study, we have developed a de-novo Cost-Consequence Model (CCM) to demonstrate the potential clinical and economic outcomes of implementing the Xpert Xpress testing strategy in the diagnosis and management of patients suspected of an ILI compared to other standard testing strategies.

2. Methods

2.1. Model overview and structure

A decision tree model developed in Microsoft Excel, with a 1-year time horizon, was used to compare the Xpert Xpress testing strategy to each comparator test strategy individually in terms of clinical outcomes, costs, and cost-consequence outcomes from the perspective of a US payer. A 1-year time horizon was chosen for this analysis to adequately account for the potential differences in testing strategies. Figure 1 showcases the decision tree structure.

Figure 1. Model structure.

The population of interest was a hypothetical insured cohort of one million members, a portion of whom develop symptomatic ILIs and present to an outpatient care facility for a medical visit. The modeled population is considered to be a randomized sample of a generalizable US population and, therefore, mirrors the overall US demographic distribution. These patients are then modeled with each testing strategy, with all conditions remaining the same.

The underlying disease, presence of coinfections, test results, and treatment mix informed the number of tests, costs, rates of hospitalization, intermediate-term outcomes, intensive care unit (ICU) admission, and mechanical ventilation (MV). Model inputs and assumptions are informed by published literature and are not based on clinical trials.

2.2. Epidemiology and underlying disease

An underlying illness distribution informed the results of the test strategies applied in the model. The population of 1,000,000 members were first stratified by one of three common viral causes of ILI: influenza, SARS-CoV-2, and RSV. Disease stratification was informed by annual disease incidence calculated from US data sources (Table 1)^9,12,32. The base case analysis utilized the most recent epidemiological values per key opinion leader guidance. The incidence of RSV was calculated as a weighted average of the incidence for adults and the incidence for children because of the large difference between the adult and child population estimates. The remaining portion of the population who did not have influenza, SARS-CoV-2, or RSV were assumed to have an “unknown ailment”, such as a bacterial infection. The number of total cases of ILI was assumed to be the sum of influenza, SARS-CoV-2, RSV and unknown ailment cases.

Table 1. Essential model inputs.

Parameter	Model input		Source
Incidence rate (per 100,000)
Patients with ILI symptoms	81.42%		Assumption; sum of influenza, SARS-CoV-2, RSV, and unknown ailment incidence rates
Patients with influenza	2.7%		CDC 2023 estimate^25
Patients with SARS-CoV-2	5.98%		Johns Hopkins SARS-CoV-2 dashboard 2023^26
Patients with RSV	Adults	1.37%	Tong 2010^12
	Children	3.37%
Patients with influenza & SARS-CoV-2	0.05%		Dadashi 2021^5
Patients with influenza & RSV	0.09%		Reina 2018^7
Patients with SARS-CoV-2 & RSV	0.31%		Kim 2020^6
Patients with influenza, SARS-CoV-2, & RSV	0.002%		Steponavicene 2023^27
Test specifications	Sensitivity	Specificity
Influenza
Xpert Xpress	0.97	0.97	Liu 2022*^28
Multiplex PCR	0.90	0.99	Genilotti 2022^29
Influenza PCR + SARS-CoV-2 PCR	0.94	0.99	Genilotti 2022^29
Duplex antigen	0.69	0.97	Genilotti 2022^29
Influenza antigen + SARS-CoV-2 antigen	0.69	0.97	Genilotti 2022^29
SARS-CoV-2
Xpert Xpress	0.99	0.97	Lee 2021^30
Multiplex PCR	0.88	0.98	Jarrom 2022^┼^26
Influenza PCR + SARS-CoV-2 PCR	0.88	0.98	Jarrom 2022^26
Duplex antigen	0.76	0.99	Brummer 2022^31
Influenza antigen + SARS-CoV-2 antigen	0.76	0.99	Brummer 2022^31
RSV
Xpert Xpress	0.96	1.00	Liu 2022^28
Multiplex PCR	0.91	0.99	Genilotti 2022^29
Costs
Diagnostic tests
Xpert Xpress (MultiPlex)	$142.63		Clinical Laboratory Fee Schedule (23CLABQ1) and (22CLABQ4); CPT 0241 U
Multiplex PCR	$142.63		Clinical Laboratory Fee Schedule (23CLABQ1) and (22CLABQ4); CPT 87636/87637
Singleplex PCR Influenza	$95.80		Clinical Laboratory Fee Schedule (23CLABQ1) and (22CLABQ4); CPT 87502
Singleplex PCR SARS-CoV-2	$51.31		Clinical Laboratory Fee Schedule (23CLABQ1) and (22CLABQ4); CPT 87501/87635
Duplex antigen test	$30.94		Clinical Laboratory Fee Schedule (23CLABQ1) and (22CLABQ4); CPT 87428
Influenza A/B antigen test	$14.13		Clinical Laboratory Fee Schedule (23CLABQ1) and (22CLABQ4); CPT 87400
SARS-CoV-2 antigen test	$35.33		Clinical Laboratory Fee Schedule (23CLABQ1) and (22CLABQ4); CPT 87426

* Weighted mean of influenza A and B.

^┼ In absence of real-world data for the duplex antigen test, it was assumed to have the same sensitivity and specificity as the single antigen tests.

^{^} The Influenza PCR + SARS-CoV-2 PCR, Duplex antigen, and Influenza antigen + SARS-CoV-2 antigen test strategies do not test for RSV.

It was also possible for modeled patients to have coinfections. The incidence of coinfections was informed by several studies listed in Table 1.

2.3. Comparators

The intervention of interest is the Xpert Xpress CoV-2/Flu/RSV plus, a POC, multiplex PCR test, using one nasopharyngeal swab to produce results in approximately 36 min. The intervention of interest is modeled as a standalone intervention and does not include a combination of other testing strategies. The model incorporates four comparator testing strategies:

1. Multiplex PCR: A send-out, multiplex PCR test that uses one nasopharyngeal swab and produces results in 24 hours^20,33

2. Influenza PCR + SARS-CoV-2 PCR: Co-administered, send-out, single-PCR test strategy that use two nasopharyngeal swabs (one to test for SARS-CoV-2 and the other to test for influenza A/B) and produces results in 24 hours

3. Duplex antigen test: A POC, duplex antigen test that uses one nasopharyngeal swab to test for SARS-CoV-2 and influenza A/B. This test produces results in 15–30 minutes³³

4. Influenza antigen + SARS-CoV-2 antigen: Co-administered, POC antigen test strategy that uses two nasopharyngeal swabs (one to test for SARS-CoV-2 and the other to test for influenza A/B) and produces results in approximately 15–30 minutes³³

It was assumed that the intervention test strategy and the multiplex PCR test strategy would not be followed with confirmatory follow-up testing because they are both able to detect SARS-CoV-2, influenza A/B, and RSV with a relatively high degree of accuracy.

It was assumed that the influenza PCR + SARS-CoV-2 PCR test strategy would not be followed with confirmatory follow-up testing either because it is able to detect SARS-CoV-2 and influenza A/B with a relatively high degree of accuracy. While this strategy excludes the detection of RSV, it is still commonly used in clinical practice, as validated by an expert advisor.

It was assumed that the duplex antigen test strategy and the influenza antigen + SARS-CoV-2 test strategies may go on to receive additional confirmatory testing. As per the CDC testing guidelines, patients with a negative SARS-CoV-2 result at baseline would have up to 2 additional, confirmatory antigen retests if the result continued to be negative. Following negative antigen retests, the CDC recommends additional testing with either antigen or PCR tests. We assumed that a confirmatory PCR test was performed in this case given the increased diagnostic accuracy compared to antigen tests. There was no confirmatory test performed after a positive result for influenza, SARS-CoV-2, or both.

2.4. Testing pathway

Test results were comprised of four possible outcomes per illness tested:

• True positive: Patient has the illness and tests positive for the illness. Patients were assumed to receive correct treatment and symptoms were assumed to resolve in most patients.

• False positive: Patient does not have the illness and tests positive for the illness. Patients were assumed to receive incorrect treatment (futile treatment), but symptoms were assumed to persist in most patients. Patient would need to be re-evaluated for an alternative diagnosis.

• True negative: Patient does not have the illness and test negative for the illness. Patients do not receive treatment, symptoms persist, and patient would need to be evaluated for an alternative diagnosis.

• False negative: Patient has the illness and tests negative for the illness. Patient does not receive treatment, symptoms persist, and patient would need to be evaluated for an alternative diagnosis.

The rate of each outcome was driven by the epidemiologic characterization of the population and each test’s accuracy (sensitivity and specificity). These specifications are detailed in Table 1.

2.5. Test specifications

Test specifications, including sensitivity and specificity values for each test, can be found in Table 1. Xpert Xpress CoV-2/Flu/RSV plus was found to have sensitivity values ranging from 96% to 99% and specificity from 97% to 100% across influenza, SARS-CoV-2, and RSV^28,30. Each test was associated with a pre-specified turnaround time. Turnaround time, in addition to the timing of healthcare seeking behavior, determines which modeled patients are eligible for treatment.

2.6. Treatment

As a simplifying model assumption, each modeled illness was treated with one treatment regimen. In accordance with published guidelines, patients with influenza were assumed to be treated with oseltamivir, while patients with SARS-CoV-2 were assumed to be treated with Paxlovid (nirmatrelvir/ritonavir), if diagnosed within their prescribed treatment windows^34,35. Although RSV can result in severe illness for some populations, care for RSV does not rely on a specific prescription treatment, but rather more prudent management of symptoms and over-the-counter (OTC) medications. We have not included the cost of OTC treatments in this analysis because it is from the perspective of a US payer, although RSV treatments are quantified to and count towards patients with a true positive RSV test result.

Generally, it was assumed that patients would only receive treatment after test results were reported. Based on expert opinion, it was estimated that physicians would empirically treat 20% of patients testing negative for influenza, SARS-CoV-2, or RSV with comparator test strategies, but who still exhibit symptoms consistent with ILI. Empiric treatment was not applied to patients receiving the Xpert Xpress test given the short time to test results and the accuracy of testing. As no real-world literature was identified which described frequency of Paxlovid use in patients with false negative test results, assumptions regarding rates of treatment for false negatives (empiric treatment) is based on clinical expert opinion.

Prescribing information for oseltamivir and nirmatrelvir/ritonavir advise initiation of treatment within 2 days and 5 days of symptom onset, respectively. Therefore, any patients positively diagnosed after the treatment indication windows would not receive treatment, unless treated empirically.

Treatments for patients with “unknown ailments” were tabulated as a composite of the antibiotics prescribed for the most common alternative diagnoses to influenza, SARS-CoV-2, and RSV: sinusitis³⁶, ear infection³⁷, and bacterial pneumonia³⁸. Patients within the unknown ailment arm of the model who had concurrent false positive test results for influenza, SARS-CoV-2, or RSV were also treated for these illnesses accordingly.

Although antibiotics are not recommended for treatment of influenza, SARS-CoV-2, or RSV, prior research has shown that patients with these diagnoses nonetheless receive antibiotic treatment^25,39. The likelihood to prescribing antibiotics for patients with influenza was assumed to be 22%, 62% for patients with SARS-CoV-2, 54.6% for patients with RSV, and 100% for patients who fell within the basket of unknown ailments^25,39,40. Patients in the unknown ailment arm of the model are assumed to receive antibiotic treatment given that these patients will have worsening symptoms and treatments for viral infections will not be effective. The model does not account for the development of antibiotic resistant bacteria.

Number of correct and incorrect treatments was output as a result of the model. Proportion of correct treatments was defined as the number of correctly prescribed treatments divided by the number of total treatments prescribed (e.g. the number of oseltamivir treatment courses prescribed to patients with influenza divided by total number of oseltamivir prescriptions).

2.7. HCRU and costs

The proportion of patients who seek medical care ultimately drives the number of tests received and treatments given. It was estimated that that 47.6%, 48.0%, and 81.5% of patients with symptomatic influenza, SARS-CoV-2, and RSV would seek medical care, respectively^9,12,41. More generally, it was assumed based on expert opinion, that one quarter of all patients with ILI, outside of influenza, SARS-CoV-2, and RSV, would seek medical care.

The timing of healthcare seeking behavior was based on a real-world study of influenza patients seeking treatment (mean: 1 day, SD: 0.5 days)⁴². The timing of healthcare seeking behavior was assumed the same for all patients with ILI given a paucity of data. A gamma distribution (left skewed and bound by 0 to infinity) was assumed to be a good fit to predict the distribution of days in between symptom onset to medical visit. Most patients were predicted to engage a medical consultation between 0 and 2 days of symptom onset, but a subset would seek treatment after the treatment eligibility window of oseltamivir and nirmatrelvir/ritonavir, respectively, and will therefore not be able to receive treatment. These assumptions were validated by a clinical expert.

Model inputs regarding percentages of hospitalizations, ICU admissions, and MV were informed from published literature; these particular cost categories were included as they represented some of the most common cost drivers among respiratory illnesses^42–49. Each model input was stratified by illness and by treatment status. Additional details are available in Supplement Table S1.

Test costs were sourced from the Clinical Laboratory Fee Schedule and HRU costs were sourced from the CMS Physician Fee Schedule, as well as various real-world studies, when not available from the CMS Physician Fee Schedule^{43,46,50–53}. Additional costing input data can be found in the Supplemental Materials (Supplement Table S2).

Treatment costs are outlined in Supplement Table S2. Treatment costs were not included for patients diagnosed with RSV given that treatments include OTC drugs, which would not be covered by a payer. Treatment costs for oseltamivir and nirmatrelvir/ritonavir include wholesale acquisition costs and were sourced from ProspectoRx and published literature.

2.8. Sensitivity analysis

In addition to the base case economic analysis, a deterministic sensitivity analysis (DSA) was also run. The DSA is used to understand which model parameters have the greatest impact on model results. The DSA varied each parameter independently by 10%, which was deemed to be a plausible range to explore parameter uncertainty.

Multiple scenario analyses were also undertaken to address parameter uncertainty. These scenarios addressed uncertainty in test specifications, empiric treatment assumptions, and epidemiological estimates. The first scenario utilizes sensitivity and specificity estimates reported by test manufacturers, as opposed to real-world data. The second scenario, informed by expert opinion, reduces the percentage of patients who receive empiric treatment in the PCR testing strategies to 18%, while modifying this percentage in the antigen testing strategies to 35% to explore an alternative treatment possibility. The final scenario increased the incidence rates of influenza and SARS-CoV-2 to their highest reported values over the past decade to obtain a better sense of how changes in these rates of illnesses over time impacts model results.

3. Results

3.1. Population results

Out of the hypothetical cohort of one million insured members, an estimated 203,548 patients engaged in a medical visit and were tested for ILI per year. Of these patients, there were 13,159 influenza cases, 29,735 SARS-CoV-2 cases, and 14,784 RSV cases. These values include coinfection cases. The model quantified 3,488 people with 2 or more infections. Model results are reported in additional detail in Table 2.

Table 2. Cost-consequence model results.

Description	Xpert Xpress	Multiplex PCR	Influenza PCR + SARS-CoV-2 PCR	Duplex antigen	Influenza antigen + SARS-CoV-2 antigen
Testing
Tests (n)	203,548	203,548	407,095	680,872	884,420
Tests averted with Xpert Xpress	N/A	0	203,548	477,325	680,872
Costs
Testing	$29,031,980	$29,031,980	$29,943,873	$39,814,731	$43,584,431
Treatment	$53,099,369	$88,652,191	$92,142,464	$90,266,420	$90,266,420
Hospitalizations	$55,702,933	$58,524,394	$58,971,026	$57,773,297	$57,773,297
ICU	$6,610,537	$7,588,243	$7,611,701	$7,218,748	$7,218,748
MV	$6,478,020	$7,054,077	$7,101,656	$6,856,193	$6,856,193
Total costs	$150,922,839	$190,850,884	$195,770,720	$201,929,389	$205,699,089
Budget impact estimates
Testing cost (PMPM)	$2.42	$2.42	$2.50	$3.32	$3.63
Incremental testing PMPM	N/A	$0.00	$0.08	$0.90	$1.21
Total cost PMPM	$12.58	$15.90	$16.31	$16.83	$17.14
Incremental total cost PMPM	N/A	$3.33	$3.74	$4.25	$4.56
Percent of treatments correctly prescribed*
Oseltamivir	74%	28%	27%	22%	22%
Nirmatrelvir/ritonavir	87%	46%	44%	48%	48%
RSV management	100%	32%	9%	29%	29%
Azithromycin	79%	78%	76%	79%	79%
Healthcare resource utilization
Hospitalizations (n)	4,942	5,130	5,173	5,089	5,089
ICU (n)	936	996	1,004	977	977
MV (n)	601	623	629	619	619
Deaths (n)	207	227	228	216	216
Pneumonia (n)	2,235	2,453	2,870	2,335	2,335
Number needed to test (to prevent an…)
… additional correct treatment	N/A	25	12	30	30
… additional inappropriate treatment	N/A	2.8	2.5	2.5	2.5
… additional hospitalization	N/A	1,083	879	1,384	1,384
… additional ICU admission	N/A	3,375	3,003	4,876	4,876
… additional MV	N/A	9,495	7,473	11,785	11,785
… additional death	N/A	10,178	10,094	22,618	22,618
… additional case of pneumonia	N/A	932	320	2,028	2,028
Cost-consequence outcomes
Net correct treatment courses	133,419	53,610	35,930	44,156	44,156
Total cost per correct treatment course	$1,131	$3,560	$5,449	$4,573	$4,658
Test cost per correct treatment course	$218	$542	$833	$902	$987
Avoided hospitalizations**	N/A	188	232	147	147
Avoided ICU**	N/A	60	68	42	42
Avoided MV**	N/A	21	27	17	17
Deaths averted**	N/A	20	20	9	9

* Proportion of correct treatments was defined as the number of correctly prescribed treatments divided by the number of total treatments prescribed (e.g. the number of oseltamivir treatment courses prescribed to patients with influenza divided by total number of oseltamivir prescriptions).

**Relative to Xpert Xpress.

Testing with Xpert Xpress resulted in a range of averted tests from 0 compared with the multiplex PCR strategy, and up to 680,872 averted tests (including confirmatory tests) compared with the Influenza antigen + SARS-CoV-2 antigen test strategy because this strategy necessitated multiple test kits and possible rounds of confirmatory testing.

3.2. Overall cost results

The Xpert Xpress test was found to be the least expensive testing strategy with total costs summing to $151 million, inclusive of testing, treatment, and downstream HCRU costs. Total cost differentials between the Xpert Xpress test strategy and comparators range from $40 million in the multiplex PCR strategy to $55 million in the Influenza antigen + SARS-CoV-2 antigen test strategy. See Table 2 for additional details.

Despite a higher per-test cost for the Xpert Xpress and multiplex tests, the reduction in number of total tests administered offset total testing costs compared to the other three test strategies. The Xpert Xpress testing strategy also demonstrated lower treatment costs compared with other testing strategies.

3.3. Treatment stewardship

The Xpert Xpress testing strategy demonstrated the highest proportion of correct treatments compared to all other testing strategies. The Xpert Xpress test led to correctly prescribing 74% of Oseltamivir treatments, 87% of nirmatrelvir/ritonavir treatments, and 100% of RSV treatments.

Correct prescription of oseltamivir ranged from 22% to 28% among comparators, while correct prescription of nirmatrelvir/ritonavir ranged from 44% to 48%, and correct prescription of RSV OTC treatments ranged from 9% to 32%. Proportions of correct prescriptions can be found in Table 2.

3.4. HCRU

The Xpert Xpress testing strategy also resulted in the fewest hospitalizations, ICU admissions, MV uses, and secondary pneumonia cases (see Table 2). Of the initial cohort of 1 million insured individuals, the number of hospitalizations ranged from 4,942 in the Xpert Xpress test strategy to 5,173 Influenza PCR + SARS-CoV-2 PCR test strategy. ICU admissions ranged from 936 to 1,004 and MVs ranged from 601 to 629 in the Xpert Xpress test strategy and Influenza PCR + SARS-CoV-2 PCR test strategy, respectively.

3.5. Mortality

A small proportion of the initial 1 million cohort died due to ILI. The number of modeled deaths ranged from 207 in the Xpert Xpress arm of the model to 228 in the Influenza PCR + SARS-CoV-2 PCR test strategy (see Table 2).

3.6. Cost-consequence results

All cost-consequence results include the total costs incurred by patients throughout the time horizon. Total cost per correct treatment course was $1,131 for the Xpert Xpress. Comparator testing strategies were less cost-efficient at $3,560, $5,449, $4,573, and $4,658 per correct treatment course for the multiplex PCR, Influenza PCR + SARS-CoV-2 PCR, duplex antigen, and Influenza antigen + SARS-CoV-2 antigen test strategies, respectively.

The number of patients needed to test to prevent one additional inappropriate course of treatment with Xpert Xpress ranged from 2 tests to 3 tests in the antigen testing strategies and the PCR testing strategies, respectively. The number of patients needed to test to prevent an additional death with Xpert Xpress range across comparators from 10,094 to 22,618 in the Influenza PCR + SARS-CoV-2 PCR test strategy and the antigen testing strategies, respectively. Additional cost-consequence outcomes can be found in Table 2.

3.7. Sensitivity analysis

The tornado diagram in Figure 2 shows the results of the DSA. The top three most influential model parameters on total costs associated with Xpert Xpress include the intervention specificity of SARS-CoV-2, nirmatrelvir/ritonavir cost per treatment course, and the intervention specificity of influenza.

Figure 2. Deterministic sensitivity analysis.

3.8. Scenario analyses

Several scenario analyses were developed to understand the impact of model parameter uncertainty including seasonality of ILI. Table 3 showcases the description, model inputs, and results of each scenario analysis.

Table 3. Cost-consequence model scenario analyses.

Scenario	Description	Updated model inputs	Model outputs
1	Test specifications (sensitivity and specificity) are replaced with estimates from package insert estimates for each modeled strategy.	Influenza (sensitivity, specificity) Xpert Xpress (100%, 100%) Multiplex PCR (100%, 100%) Singleplex PCR (100%, 100%) Duplex antigen (100%, 99.75%) Influenza A/B antigen (92.20%, 100%) SARS-CoV-2 Xpert Xpress (100%, 100%) Multiplex PCR (100%, 100%) Singleplex PCR (100%, 100%) Duplex antigen (95.24%, 99.35%) SARS-CoV-2 A/B antigen (95.50%, 99.20%) RSV Xpert Xpress (100%, 100%) Multiplex PCR (100%, 100%)	Total cost per correct treatment course Xpert Xpress = $1,033 Multiplex PCR = $3,104 Influenza PCR + SARS-CoV-2 PCR = $5,469 Duplex antigen = $3,892 Influenza antigen + SARS-CoV-2 antigen = $4,050
2	Empiric treatment assumptions are removed for PCR testing strategies and modified for antigen testing strategies.	Percentage of patients receiving empiric treatment of influenza with oseltamivir within antigen testing strategies: 35.0% Percentage of patients receiving empiric treatment of SARS-CoV-2 with nirmatrelvir/ritonavir within antigen testing strategies: 17.5%	Total cost per correct treatment course Xpert Xpress = $1,131 Multiplex PCR = $1,164 Influenza PCR + SARS-CoV-2 PCR = $1,353 Duplex antigen = $3,602 Influenza antigen + SARS-CoV-2 antigen = $3,670
3	Incidence estimates of influenza, SARS-CoV-2, and RSV are replaced with higher estimates, corresponding to 10-year historical maximums.	Medical visits among patients with ILI: 50% (assumption) Influenza incidence: 12,611 per 100,000^45 SARS-CoV-2 incidence: 37,764 per 100,000^9	Total cost per correct treatment course Xpert Xpress = $1,272 Multiplex PCR = $2,240 Influenza PCR + SARS-CoV-2 PCR = $2,461 Duplex antigen = $2,040 Influenza antigen + SARS-CoV-2 antigen = $2,074

Replacing the test specification inputs with those provided on package inserts for all tests results in a decrease in total costs per correct treatment course from the base case analysis. In this scenario, total costs per correct treatment course were found to be $1,033 in the Xpert Xpress testing strategy with a range in the comparator testing strategies from $3,104 to $5,469 in the multiplex PCR strategy to the Influenza PCR + SARS-CoV-2 PCR testing strategy, respectively. Similarly, when the empiric treatment assumptions are varied, the total costs per correct treatment course decreased from the base case across model comparators, ranging from $1,131 to $3,670 in the Xpert Xpress and Influenza antigen + SARS-CoV-2 antigen testing strategy, respectively. The total costs per correct treatment course range from $1,272 to $2,461 in the Xpert Xpress strategy and Influenza PCR + SARS-CoV-2 PCR strategy, respectively, when the influenza and SARS-CoV-2 incidence rates are increased to their 10-year historical maximums.

4. Discussion

ILIs affect millions of Americans every year, yet the diagnosis of the underlying pathogen may be missed by current testing strategies. Testing for respiratory pathogens enables appropriate and timely patient management, improving infection control and clinical outcomes. This model demonstrated that the Xpert Xpress test strategy facilitated the provision of higher proportion of correct treatments. This was driven both by its accuracy and its speed of turnaround. Due to the high accuracy of the Xpert Xpress PCR test, more patients were correctly diagnosed and due to its short turnaround time, more patients were diagnosed within the limited treatment window of oseltamivir and Paxlovid treatments. Compared to multiplex PCR, use of Xpert Xpress was more likely to result in a correct treatment both due to higher sensitivity of the Xpert Xpress test, and because longer turnaround time for multiplex PCR (24 h vs to 15–30 min) results in missed opportunities for diagnosis within the treatment window, particularly in the case of oseltamivir.

This diagnostic efficiency enabled efficient healthcare resource utilization and resulted in fewer clinical consequences and healthcare costs related to undetected, untreated, and inappropriately treated viral infections. Early and accurate treatment resulted in cost savings of $52.9 M vs antigen POC test strategies and $42.4 M savings vs send-out test strategies. The key drivers of cost differences include reduced hospitalizations, treatment costs, and testing costs among patients who received prompt and correct medical treatment. For example, a study by Shah et al. found that adults who took Paxlovid within 5 days of a COVID-19 diagnosis had a 51% lower hospitalization rate within the next 30 days, than those who were not given the drug. Additionally, they found that patients appropriately given Paxlovid were 89% less likely to develop severe illness and death compared to participants who were given placebo. Among influenza patients, Treanor et al. found that timely treatment with oseltamivir reduced the duration and severity of acute influenza in adults and may decrease the incidence of secondary complications such as bronchitis and sinusitis⁵⁴.

Xpert Xpress also produced the lowest treatment costs compared to the other four test strategies. This was determined to be partially driven by the empiric treatment decisions associated with other comparators. In the comparator test strategies, we assumed that clinicians may have to make empiric treatment decisions if test results are delayed or unreliable. This resulted in a greater number of incorrect treatments and elevated treatment costs among comparators. Our model assumed that clinicians would not have to treat patients empirically in the Xpert Xpress testing strategy because results would be both accurate and available in time to make “on-demand” treatment decisions.

Through the scenario analyses, we were able to test alternative values for parameters with more than one valid estimate. In the base case of the model, test specification data was drawn from real-world studies with large study populations. In the first scenario, when manufacturer estimates of test specifications (from clinical trials) are used for all comparators, including the intervention, Xpert Xpress still demonstrated total cost savings vs. all comparators.

Additionally, the assumption of 20% empiric treatment was examined through a scenario analysis. These results strengthen our base case findings that testing with Xpert Xpress will result in cost savings for the payer and will likely persist as individual model parameters continue to change.

Lastly, the incidence rates of influenza and SARS-CoV-2 have fluctuated vastly in the past few years due to the COVID-19 pandemic. Prior to initiating lock-down, COVID-19 cases were surging. During lock-down and shortly after, SARS-CoV-2 and influenza rates decreased substantially^9,32,55. Therefore, we tested the outcomes of this model in a high-infection-rate scenario. Xpert Xpress remained in most efficient test strategy in terms of cost per correct course vs. comparator strategies.

In a review of previously published models, there was one study on a molecular point-of-care test (POCT-PCR) for influenza among elderly patients in an ambulatory care setting. In this analysis the POCT-PCR test strategy was compared to clinical judgement (empiric treatment without test). The POCT-PCR group had a lower hospitalization rate (1.38% vs 2.85%) and mortality rate (0.08% vs. 0.16%) compared to the empirically-treated group and cost an additional $33.20 per ILI patient tested. You et al. (2017) reported that the model results were sensitive to hospitalization rates without prompt therapy and the prevalence of influenza⁵⁶. The reduced HRU and health consequences are aligned with the findings of our analysis, while our sensitivity analysis appeared to be more centered around the test specifications for the Xpert Xpress strategy.

Another model by Abbasi et al. compared the cost effectiveness of a rapid test to that of a PCR test among a population with acute respiratory syndrome. The authors found that the rapid test was less costly and less effective than the PCR test, but the cost difference was much larger than the effectiveness difference, resulting in the conclusion that the rapid test was more cost-effective than the PCR test. This result was different from the findings in our analysis. One primary reason for this difference could be because the Abbasi study did not include downstream HRU or treatment costs that could potentially offset the added cost of the PCR test, as was the case in our analysis.

There is currently a paucity of cost-effectiveness studies describing SARS-CoV-2 testing strategies that consider antiviral treatment, rather than isolation alone. A study in an Iranian cohort reported that PCR tests are more cost effective during times when SARS-CoV-2 is not as prevalent (5%), while antigen tests are more cost effective during periods of high disease prevalence (10–50%)⁵⁷. These results are also directionally aligned with what was observed between the base case and scenario 3 (varied test accuracy, changes to empiric treatment assumptions, and maximum ILI incidence rates) results of this study.

4.1. Strengths

One strength of this model is that it is the first model that we are aware of which includes coinfections. Previous studies on economic burden and cost-effectiveness of test strategies and treatments either examine influenza specifically or the overarching basket of ILI diagnoses^58–60. The former does not consider other pathogens that circulate in the same seasons and environments as influenza and could therefore over-simplify the real world situation. The latter inherently lacks the ability to identify virus-specific outcomes and possible sequelae.

Another strength of this model is that it considers a post-pandemic world in which the Public Health Emergency has ended and sustainable public health practices have returned. With the initial emergence of the SARS-CoV-2 virus came a rapidly and persistently shifting landscape of ILI-related treatment, prevention, and policies^55,61. Many health economic evaluations produced during the pandemic consider practices that are out of line with current policies and treatment protocols, such as isolation, masking, and social distancing⁶². In light of the global health emergency declaration ending, it is important to evaluate current health system responses to SARS-CoV-2 and other ILIs⁶³. Our model considers the most current epidemiological data, as well as current testing and treatment reimbursement policies.

4.2. Assumptions and limitations

A majority of coinfection model inputs have been calculated or assumed due to a paucity of data in the literature. The studies that reported the incidence of these coinfections indicated that they were relatively uncommon or under-reported.

This model uses real-world evidence (RWE) for the test specification data, as opposed to clinical trial data^26,28–31. Using RWE outcomes are generally considered more realistic and generalizable compared to using the outcomes of a clinical study. The environment, procedures, and populations selected in clinical studies are highly regulated, which can result in overly-optimistic findings compared to RWE⁶⁴. Therefore, we expect the findings of this study to be as close as possible to what would be observed in clinical practice.

This model relied upon data from the Healthcare Cost and Utilization Project (HCUP) from the Agency for Healthcare Research and Quality. Medicare payment rates were used within the model, which may not reflect commercial payer costs. It is likely that a portion of the population would have commercial insurance and may have different costs and healthcare utilization than is currently reported by the model.

The model only includes costs for a simplified breakdown of the ILI bucket of illnesses (influenza, SARS-CoV-2, RSV, and general bacterial infections). Furthermore, one proxy for treatment per illness was assumed. In clinical practice, there are multiple options for treatment of the ILIs, so treatment windows and costs may differ in actuality. This simplifying assumption provides a representation of practice to estimate a generalized cost of treatments. Additionally, it was assumed that empiric treatment was only applied to patients within the comparator test strategies, and not the intervention test strategy which could have increased the number of incorrect treatments among those groups.

The DSA employed within the analysis varies each parameter independently by 10%. While this methodology does not necessarily comply with best practices, it offers a comprehensible and transparent sensitivity analysis in the face of data limitations. The 10% variation was considered to be within the plausible range for each parameter based on expert opinion.

Other simplifying assumptions were used to focus the scope of the analysis. Model outcomes are provided by individual testing strategies and a mix of testing strategies was not considered. Quality of life was also not taken into consideration for this analysis. Additionally, treatment dosing for adult patients has been applied to all modeled patients.

This model used the most recent influenza, SARS-CoV-2, and RSV incidence and prevalence data. While this would ordinarily be the best way to demonstrate real world situations, in light of the SARS-CoV-2 pandemic the incidence of influenza shifted away from the average historic incidence. The average incidence of influenza between 2010 and 2020 in the US was 8.84%; however, the incidence of influenza during the 2021–2022 season which was used in the model was 2.71%. The lower incidence of influenza used in the model may underestimate the benefits of detection and treatment of influenza given the short 2-day treatment window with oseltamivir.

A final limitation of this study is the fixed nature of the decision tree, which does not include the chain of infection and impact of different testing modalities on transmission dynamics. Prior research has found that knowledge of test result impacts intention to quarantine; in particular, knowledge of a positive test result is associated with higher intention to quarantine⁶⁵. As such, it is possible that test methodologies that are more sensitive and provide faster results to patients would decrease transmission of disease through improved quarantining practice. Further research is needed to better understand the relationship between diagnostic test results and transmission dynamics.

5. Conclusion

This cost-consequence model was used to quantify the cost-consequences of using the Xpert Xpress CoV-2/Flu/RSV plus testing strategy by considering the clinical, health, and cost implications of employing the diagnostic test compared to other commonly used testing methods. Testing with the Xpert Xpress CoV-2/Flu/RSV plus test in a hypothetical cohort of 1 million insured individuals results in significant improvements in clinical and economic outcomes at a one-year time horizon. Improvements in outcomes were driven primarily by rapid and accurate diagnostic test results, leading to reductions in testing costs and downstream HRU. The Xpert Xpress CoV-2/Flu/RSV plus enables targeted reliable test results for patients with acute respiratory symptoms resulting in appropriate and efficient testing and treatment.

Transparency

Author contributions

All authors contributed to the conception and design of the study, data acquisition and analysis, and interpretations of the results. EB and SD drafted the first draft of the manuscript. All authors critically revised the manuscript to its final stages, approved the final version of the manuscript, and take responsibility for all aspects of the study.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgements

No assistance in the preparation of this article is to be declared.

Declaration of funding

Financial support for this study was provided by Cepheid (CA, United States). The funding agreement ensured the authors’ independence in designing the study, interpreting the data, writing, and publishing the report. The content and results in this manuscript were approved by all authors and has not been subject to sponsor censorship.

Declaration of financial/other relationships

SD, EB, IJ, and CM are employees of PRECISIONheor, which provides consulting services to the diagnostic and pharmaceutical industries, including Cepheid. AB and JC are employed by Cepheid.