The results of the deterministic decision tree analysis are shown in Table 2.

In the deterministic decision tree analysis, the additional use of multiplex PCR (strategy 1) resulted in the case of an infected person (scenario A) costs €3,460.26, while the use of established diagnostics alone (strategy 2) costs €3,053.91. Strategy 1 was therefore €406.35more resource-intensive in this scenario. In the case of a non-infected person (scenario B), strategy 1 costs €954.60, while strategy 2 still costs €3,053.91. In this case, the additional use of multiplex PCR was €2,009.31 more cost-efficient.

Similar to the decision tree analysis, the PSA for infected individuals (scenario A) shows that the additional use of multiplex PCR was on average €408.75 (range: € 207.4 to € 1,083.9)) more costly than the use of established diagnostics alone. Most simulations showed a cost difference between €200 and €600, as shown in Fig. 2, and strategy 1 was more costly than strategy 2 in 100% of the simulations.

Distribution of the cost difference Δinf in euros for infected persons, shown as a frequency distribution of the 10,000 Monte Carlo simulations of the probabilistic sensitivity analysis

For non-infected individuals, strategy 1 resulted in the PSA in average cost savings of €2,098.40 (range: € range € 1,097.0 to € 3,371.9). Most simulations result in a cost difference Δnot of approximately €2,500 to €1,500, as shown in Fig. 3, and here too, Strategy 1 was more cost-effective than Strategy 2 in 100% of simulations.

Distribution of the cost difference Δnot in euros for non-infected individuals, shown as the frequency distribution of the 10,000 Monte Carlo simulations of the probabilistic sensitivity analysis

In the one-way sensitivity analyses, it was found that, in infected individuals, the sensitivity of multiplex PCR in particular has a major impact on the amount of the cost difference. When all other parameters used are set to their base values, this varies between €283.42when using the 97.5% percentile (Se_M97,5%= 99,83) within the distribution and € 629.56 when using the 2.5% percentile (Se_M2,5%=99.18). The second strongest influence on the cost difference \(\:}_}\) are the costs of the Multiplex PCR. When these costs vary between the 97.5% percentile (c_M97,5%= €301.31) and the 2.5% percentile (c_M2,5%= €202.35), \(\:}_}\) varies between €359.54 and €458.49. The other cost parameters (c_iso_M, c_iso_C, c_contact) have a significantly lower impact. The corresponding tornado plot for infected patients is shown in Fig. 4.

Effects on the cost difference Δinf when varying the individual parameters between the 2.5% and 97.5% percentiles within the distribution of the parameter

For non-infected persons (scenario B), the costs of isolation for the duration of the established diagnostics (c_iso_C) have a major impact on the cost difference \(\:}_}\). Using the 2.5% percentile in its distribution (c_iso_C2,5%= €2,489.03) gives a \(\:}_}\) of €-1541.08, while using the 97.5% percentile (c_iso_C97,5%= €3,686.29) results in a \(\:}_}\) of €-2,724.25. The variation in the other parameters has a significantly lower impact: specificity of multiplex PCR (Sp_M; range of \(\:}_}\): €-2,099.28 to – €2,094.78 ), cost of isolation for the duration of multiplex PCR (c_iso_M; range of \(\:}_}\): €-2,227.38 to €-1,956.83), cost of multiplex PCR (c_M; range of \(\:}_}\): €-2,146.12 to €-2,099.28). In all cases, the additional use of multiplex PCR was cheaper (strategy 1) than the use of established diagnostics alone. The corresponding tornado plot is shown in Fig. 5.

Effects on the cost difference Δnot when varying the individual parameters between the 2.5% and 97.5% percentiles within the distribution of the parameter

No sensitivity threshold (Se_M) could be identified at which the additional use of multiplex PCR (strategy 1) would become less cost-intensive than established diagnostics alone in infected patients. Even under the assumption of perfect test performance (Se_M = 1), strategy 1 remained €249.46 more expensive than strategy 2, corresponding approximately to the material cost of a single multiplex PCR kit.

Conversely, the threshold analysis for specificity in the “patient not infected” scenario showed that the specificity of the multiplex PCR (Sp_M) could fall to approximately 0.11 before the cost difference would reverse in favour of strategy 2, making established diagnostics alone more cost-effective. Within the clinically realistic range of specificities, strategy 1 thus remained cost-saving in non-infected patients.

Full results of the additional deterministic analysis are available in the supplementary material.

In the additional deterministic pre-test probability analysis, the multiplex PCR strategy was associated with lower expected costs than conventional diagnostics alone across all epidemiological anchor scenarios and all hypothetical HCID pre-test probabilities below the threshold of 83.78%. Above this threshold, the expected cost difference reversed. At the epidemiological anchor values, the expected cost difference remained strongly in favour of the multiplex PCR strategy, with cost differences of −€2,099.31 for the Spain CCHF notification-rate anchor (0.00001%), −€2,099.28 for the Türkiye CCHF notification-rate anchor (0.001%), and −€2,099.18 for the EuroTravNet returning-traveller HCID anchor (0.005%). Full results are provided in the Supplementary Material. At these very low pre-test probabilities, the NPV remained high, whereas the PPV was low, reflecting the rarity of HCID in the evaluated anchor scenarios.

In the contact-management scenario analysis, assumptions regarding the number of exposed contacts, the duration of contact management and the isolation setting affected only the infected-patient scenario, because contact-management costs occur only after false-negative multiplex PCR results. Home quarantine or active monitoring was assumed not to generate additional hospital costs for contact isolation. In contrast, HLIU-based contact isolation generated additional costs that increased with the number of exposed contacts and the duration of isolation. Across the evaluated HLIU-based contact-isolation scenarios, the cost difference in infected patients ranged from €318.95 to €668.54, whereas the cost difference in non-infected patients remained unchanged at −€2,099.31. The corresponding HCID pre-test probability threshold above which strategy 1 became more costly than strategy 2 ranged from 75.85% to 90.07%.

In the workflow scenario analysis, strategy 1 remained less costly in non-infected patients when early discontinuation or de-escalation of HLIU isolation was assumed. The cost difference was −€2,099.31 with complete de-isolation after a negative multiplex PCR result, −€1,483.93 with delayed de-isolation after 1 h, and −€2,096.35 when HLIU isolation was discontinued but contact isolation was continued outside the HLIU. In contrast, when no early de-isolation from the HLIU was assumed, strategy 1 became more costly in both conditional scenarios, reflecting the additional multiplex PCR test cost without an isolation-related cost offset.

In infected patients, strategy 1 remained more costly across all workflow scenarios. However, costs were lower when contact isolation was continued outside the HLIU than after complete de-isolation, because no unprotected contacts were modelled in this scenario. Across workflows involving early discontinuation or de-escalation of HLIU isolation, the pre-test probability threshold above which strategy 1 became more costly ranged from 78.31% to 90.05%.

In the differentiated costing scenario analysis, the cost advantage of strategy 1 depended strongly on the extent to which staff-related HLIU resource use was included. When only PPE-related consumables were considered, strategy 1 was nearly cost-neutral in the non-infected-patient scenario and became more costly above an HCID pre-test probability of 0.49%. Including the nursing risk allowance increased the threshold to 11.61%, whereas the full allocated resource-use base case resulted in a threshold of 83.78%.

In the platform-cost scenario analysis, adding device acquisition costs increased the expected costs of strategy 1 in both conditional scenarios, with the magnitude depending on the assumed useful life and annual test volume. Across the evaluated scenarios, strategy 1 remained associated with lower expected costs in the non-infected-patient scenario, whereas it was more costly in the infected-patient scenario. The corresponding pre-test probability threshold above which strategy 1 became more costly ranged from 63.70% to 78.55%.

In the invalid-rate scenario analysis, increasing the proportion of invalid multiplex PCR results reduced the expected cost advantage of strategy 1 in the non-infected-patient scenario, because fewer patients benefited from early de-isolation after a valid negative multiplex PCR result. Across invalid rates from 0% to 5%, strategy 1 remained less costly in non-infected patients, while it remained more costly in infected patients. The corresponding pre-test probability threshold above which strategy 1 became more costly was only minimally affected, ranging from 83.38% to 83.90%.

In the test-performance scenario analysis, lower assumed multiplex PCR sensitivity increased expected costs in the infected-patient scenario, reflecting higher false-negative-related contact-management costs. Using the alternative performance estimates from Borgans et al. and Manabe et al., strategy 1 remained less costly in the non-infected-patient scenario, while the pre-test probability threshold above which strategy 1 became more costly decreased from 83.78% in the base case to 38.20% and 53.91%, respectively.