Evaluating the Cost-Effectiveness of Proposed Algorithms for C. difficile Infection in Different Pretest Probability Settings

Abstract

Background: The use of real-time polymerase chain reaction (RT-PCR) as a first-line test for the diagnosis of Clostridioides difficile may result in overdiagnosis and overtreatment because the test is not capable of distinguishing infection from carriage. Toxin EIA assays have impeditive low sensitivity. Some algorithms using enzyme immunoassay for glutamate dehydrogenase (GDH) antigen and toxins A and B as the first step have been proposed to increase diagnostic performance. However, cost-effectiveness of different diagnostic algorithms would depend on the cost of each test and on the pretest probability in different settings. The objective of the present study was to evaluate the cost-effectiveness of 2 algorithms proposed by current guidelines to diagnose C. difficile infection by developing a mathematical model that would take into account the epidemiology and costs in our hospital. Methods: The study was conducted in a 480-bed tertiary-care teaching hospital in So Paulo, Brazil. All suspected C. difficile infection cases from January to December of 2017 were evaluated for pretest probability analysis. All stools collected from patients with a requested PCR test for suspected C. difficile infection were selected for additional testing to measure the specificity and sensitivity of each different test: C. diff GDH/Toxin A/B combined test, Toxin A/B Microplate Assay, GDH, and PCR. Toxigenic stool culture for C. difficile was considered the gold standard. A mathematical model was developed and simulations were done. The outcomes evaluated were: final annual costs with diagnostic tests in US dollars and number of patients receiving a false-positive or a false-negative diagnosis in a year simulation. Results: In total, 1,441 stool samples were tested by PCR for C. difficile in our institution from January 2017 to December 2017. Overall, 206 had a positive result, with a pretest probability of 14.3%. In our simulations, the PCR-based algorithm had an annual cost of US$279,914.25, with 4 false-negative results and 8 false-positive results. The implementation of a GDH/Toxin/PCR stepwise algorithm would have reduced the annual cost to US$160,488.75, with 6 false-negative results and 1 false-positive result. Simulations of annual cost and performance of the 2 algorithms have shown that the stepwise algorithm would still be advantageous in settings with higher pretest probabilities (Fig. 1). Conclusions: A stepwise algorithm based on GDH/Toxin before PCR seems to be more cost-effective, even in settings with higher pretest probabilities.

Funding: None

Disclosures: None