Study overview

This study was based on previous health economic evaluations of PCT-guided antimicrobial stewardship in various countries, ^{Reference Mewes, Pulia, Mansour, Broyles, Nguyen and Steuten16–Reference Garnfeldt, Vincent and Gruson18} adapting parameters for the Thailand setting with validation by a local expert panel (see below).

The model simulated cost and clinical outcomes for two treatment arms. The intervention arm consisted of PCT-guided antimicrobial stewardship added to routine practice to initiate and/or discontinue antibiotic treatments; full adherence to the use of PCT-guided antimicrobial stewardship was assumed. Meanwhile, the standard-of-care arm did not use PCT-guided antimicrobial stewardship (referred to hereafter as ‘standard of care’).

The primary outcome was the annual budget impact from the Thai public hospital perspective for the intervention arm compared with standard of care. Secondary outcomes included the number of days of antibiotic treatment, the number of antibiotic-resistant cases, and the number of C. diff cases per year in Thailand.

Model parameters

Model parameters were informed by a focused, non-systematic, local, national, and international literature search, validated by a panel of seven local experts (including pulmonologists, emergency care specialists, and infectious disease experts). Validation was performed virtually and in stages, as follows. A questionnaire was developed containing the results of the literature search, which each panel member validated anonymously based on their own scientific judgement. Responses were then shared, and a second round of responses to the original questionnaire was completed. The last stage was a group meeting where discussions were held and consensus reached.

Where the literature search did not identify any relevant literature/data, model parameters were based on the insights and consensus of the expert panel.

The model parameters included base-case values to obtain results for the deterministic base case, as well as minimum and maximum values for sensitivity analyses. Parameters for suspected sepsis (Supplementary Table 1) and LRTI (Supplementary Table 2) were related to epidemiology, resource use, and effectiveness of PCT-guided antimicrobial stewardship.

In terms of resource use, the base-case assumption of three PCT tests per patient was validated by the expert panel; as noted above, full adherence to the use of PCT-guided antimicrobial stewardship was assumed. In terms of the effectiveness of PCT-guided antimicrobial stewardship, the expected reduction in the hospital-acquired C. diff rate was taken from a meta-analysis of the effect of antibiotic stewardship on C. diff incidence, ^{Reference Baur, Gladstone and Burkert19} which reported an average reduction of 32% (range 12%–47%); the model applied this reduction to C. diff incidence and to the associated hospitalization costs. Meanwhile, the expected reduction in antimicrobial resistance followed the method of Mewes et al. 2019, ^{Reference Mewes, Pulia, Mansour, Broyles, Nguyen and Steuten16} applying an absolute 3.2% reduction in the rate of antimicrobial resistance rate per percent reduction of days on antibiotic treatment (range 2.4–4.0). Relationships were assumed to be proportional.

Following the same approach as Mewes et al. 2019, ^{Reference Mewes, Pulia, Mansour, Broyles, Nguyen and Steuten16} the model did not include any impact of PCT-guided antimicrobial stewardship on mortality: published studies have suggested that PCT-guided antimicrobial stewardship is associated with no impact on mortality ^{Reference Dark, Hossain and McAuley10,Reference Schuetz, Christ-Crain and Thomann11,Reference Andriolo, Andriolo, Salomão and Atallah20} or a reduced mortality risk. ^{Reference Huang, Peng, Weng, Wang, Jiang and Du13,Reference Papp, Kiss and Baka21,Reference Schuetz, Wirz and Sager22}

Estimated unit costs of hospital stays, antibiotic treatment, and PCT testing are summarized in Supplementary Table 3. The cost of PCT testing was based on the average acquisition cost across Thailand public sector hospitals; this estimate was informed by local procurement data and manufacturer estimates (Roche Diagnostics Thailand, 2025 data on file). Costs of hospital stays were estimated based on data from six studies (Supplementary Table 4). A microcosting exercise was carried out to estimate the costs of antibiotic treatment, based on expert opinion on the most common antibiotic treatments for different sepsis and LRTI settings and the frequency of these settings (Supplementary Table 5). Based on the opinion of the expert panel, data for sepsis were weighted as follows: respiratory tract infections, 40.4% (community-acquired, 37%; hospital-acquired, 63%); urinary origin, 15.2% (community-acquired, 37%; hospital-acquired, 63%); and other, 44.4%. Data for LRTI were weighted as follows: community-acquired, 64%; hospital-acquired, 36%.

Estimation of budget impact

Budget impact was estimated using a previously published decision tree model, ^{Reference Mewes, Pulia, Mansour, Broyles, Nguyen and Steuten16,Reference Kip, Kusters, IJzerman and Steuten23} with parameters adapted to the Thai healthcare system to allow assessment from the Thai hospital perspective (Figure 1). Each pathway in the decision tree model was assigned corresponding healthcare resource utilization and costs. For every cost component (hospital stay, antibiotic use, laboratory tests [including PCT], and costs related to antibiotic resistance and C. diff infections), total costs were calculated by multiplying resource utilization by unit prices. Only relevant costs, potentially affected by the intervention, were included to calculate the differential costs. The average duration of treatment for each antibiotic included in the microcosting methodology was accounted for, with an assumption that 52% of patients with antimicrobial resistance stay in an intensive care unit (ICU) and 48% stay on a general ward (based on Phodha et al. ^{Reference Phodha, Riewpaiboon, Malathum and Coyte24} ). Calculations were performed for all items and summed to obtain the overall costs for PCT-guided care and standard of care. Incremental cost per patient was determined by subtracting the standard of care cost from the PCT-guided care cost. Population-level costs were estimated by multiplying the annual number of patients hospitalized with suspected sepsis or with LRTI by the expected per-patient cost for each strategy. The number of patients experiencing C. diff infection or antibiotic-resistant infection was derived by applying event probabilities (eg, proportion of LRTI patients admitted to the ICU) to the size of the patient population.

Figure 1.

Model structure. Decision tree adapted for Thailand (adapted from Mewes et al. ^{Reference Mewes, Pulia, Mansour, Broyles, Nguyen and Steuten16} ).

¹Includes community- and hospital-acquired. ICU, intensive care unit; LRTI, lower respiratory tract infection; PCT, procalcitonin.

For reporting the economic assessment results, the study followed the International Society for Pharmacoeconomics and Outcomes (ISPOR) budget impact analysis principles of good practice. ^{Reference Sullivan, Mauskopf and Augustovski25}

Sensitivity and scenario analyses

Deterministic sensitivity analyses were performed to analyze the uncertainty of parameters. This allowed drivers of results to be identified through univariate analyses, in which one parameter was altered to the extreme of its range of variability whilst all other parameters remained unchanged. The findings of the univariate analyses were summarized in tornado plots.

Scenario analyses were performed to demonstrate the potential budget impact of different scenarios. These analyses were based on published literature as well as input from the expert panel to reflect plausible variations in local clinical practice patterns and associated outcomes. For sepsis, scenarios were community-acquired sepsis only (based on average antibiotic treatment duration of 8 d, antimicrobial resistance rate of 20% [vs 52% for base case], and daily antibiotic costs of 472 THB [vs 832 THB for base case]), hospital-acquired sepsis only (based on average antibiotic treatment duration of 8 d, antimicrobial resistance rate of 70% [vs 52% for base case], and daily antibiotic costs of 1,166 THB [vs 832 THB for base case]), inclusion of antibiotics-related cost savings only (excluding the modeled impact on rates of C. diff and antimicrobial resistance), reduction in duration of antibiotic treatment due to PCT-guided antimicrobial stewardship (reduction to .99 d based on Peng et al. ^{Reference Peng, Chang, Xie, Sun, Qiu and Yang26} or 1.28 d based on Andriolo et al. ^{Reference Andriolo, Andriolo, Salomão and Atallah20} [vs 1.79 d for base case]), reduction in length of stay on general ward (reduction of 3.5 d based on Andriolo et al. ^{Reference Andriolo, Andriolo, Salomão and Atallah20} [vs 0 d for base case]), and reduction in length of stay in ICU (reduction of 2.05 d based on Andriolo et al. ^{Reference Andriolo, Andriolo, Salomão and Atallah20} [vs 0 d for base case]). Approaches were based on data relating to the effectiveness of antibiotic treatment reduction from meta-analyses; ^{Reference Huang, Peng, Weng, Wang, Jiang and Du13,Reference Andriolo, Andriolo, Salomão and Atallah20,Reference Papp, Kiss and Baka21,Reference Peng, Chang, Xie, Sun, Qiu and Yang26} note that these meta-analyses did not generally report separate estimates for hospital- and community-acquired conditions, so the same antibiotic treatment reduction was assumed for both for the purposes of creating the scenarios. For LRTI, scenarios were community-acquired LRTI only (based on average antibiotic treatment duration of 9.4 d, antimicrobial resistance rate of 23.4% [vs 49.2% for base case], and daily antibiotic costs of 340 THB [vs 457 THB for base case]), hospital-acquired LRTI only (based on average antibiotic treatment duration of 9.4 d, antimicrobial resistance rate of 73.2% [vs 49.2% for base case], and daily antibiotic costs of 666 THB [vs 457 THB for base case]), and inclusion of antibiotics-related cost savings only.