Study setting

This quality improvement intervention took place at 2 hospitals: UPMC Presbyterian (hospital A) and UPMC Shadyside (hospital B), located in Western Pennsylvania. Hospital A is a 695-bed, level 1 regional resource trauma center that specializes in solid-organ transplants; hospital B is a 520-bed, tertiary-care hospital specializing in oncology care. This study was granted approval as a quality improvement project by the UPMC Quality Improvement Review Committee (project no. 3808). The study population included inpatients at hospitals A and B who had C. difficile testing ordered between January 10, 2022, and October 14, 2022.

Prior to and throughout this intervention, C. difficile testing was recommended only after 2 criteria were met: (1) at least 3 loose stools in a 24-hour period not explained by laxatives, enteral feeding, enemas, or bowel preparation, and (2) at least 1 clinical indication of infection was present including antibiotic exposure within the prior 60 days (ie, body temperature >38°C, abdominal tenderness, cramping or distention, peripheral blood total white blood cell count >10,000 cells/µL, recent chemotherapy or current immunosuppression, or history of C. difficile infection). Since 2018, these criteria have been included in education, facility guidelines, and as a decision support tool in the electronic health record. C. difficile testing is only performed on stool that conforms to the shape of the sample container. C. difficile testing was performed using a 2-step testing algorithm. ^{Reference McDonald, Gerding and Johnson10} Enzyme immunoassay is performed to detect glutamate dehydrogenase and the toxin produced by C. difficile (Techlab, Blacksburg, VA) and a polymerase chain reaction test is performed to detect toxin gene production (Cepheid, Sunnyvale, CA) for discordant results.

Study intervention

The intervention added the requirement for review by an infection preventionist for inpatient C. difficile orders prior to laboratory processing. After providers placed C. difficile orders and the sample was submitted to the clinical laboratory, specimen processing was held until the order was approved by an infection preventionist. All existing orders were reviewed by infection preventionists from 1:00–3:00 p.m. daily; orders placed after 3:00 p.m. were reviewed the next day. Infection preventionists performed chart reviews and reviewed the patient’s clinical status with the bedside nurse, as necessary, to determine whether the order met the existing ordering criteria. Orders meeting testing criteria were approved. If ordering criteria were not met, the infection preventionist contacted the ordering provider to discuss the order indication and provide education on diagnostic stewardship of C. difficile. If the provider believed after education that the order was still indicated, or the infection preventionist could not reach the ordering provider, the order was approved. Order status (“approved” or “not approved”) for held specimens were submitted to the laboratory at 3:00 p.m. daily and were processed the following morning. Specimens not approved for testing were discarded.

Inclusion criteria for the intervention included all inpatients being tested for C. difficile. The emergency departments and outpatient locations were excluded from the intervention and analysis.

Contact precautions were applied at the time of the order, and providers were encouraged to consider empiric therapy for C. difficile pending the test result.

This quality improvement intervention began on the oncology units at hospital B where an opportunity for diagnostic stewardship was identified through event reviews. A pilot project was initiated on hospital B oncology units from September 12, 2021, through October 1, 2021, and was reimplemented on October 22, 2021 and continued subsequently. After its success, the decision was made to expand the intervention to all inpatient units at hospitals A and B on January 10, 2022.

Outcomes and data sources

The outcomes of the investigation were the proportion of completed tests meeting testing criteria and the primary outcome of C. difficile healthcare-associated infections, which are any positive C. difficile sample collected on or after hospital day 3. ⁴ Additional outcomes included potential adverse events of canceled testing, characterized by testing positive for C. difficile after having an order canceled with the intervention, and the exploratory outcome of antimicrobial use targeting C. difficile infection. For antimicrobial use, we hypothesized that the rate of antimicrobial use may not change because we did not expect to decrease true infections and because providers are educated not to treat C. difficile colonization cases.

The intervention period for this study was January 10, 2022, through October 14, 2022. Because this study was a quality improvement project that started with the pilot project in October, baseline (preintervention) data on C. difficile orders were collected from October 1, 2021, through January 9, 2022. The baseline period for HAI and antimicrobial usage outcomes was March 1, 2021–January 9, 2022. C. difficile HAIs and antimicrobial usage were reported monthly. The additional outcomes representing process measures were analyzed using weekly intervals consistent with how these data were used for quality improvement efforts. Intervention weeks are oriented according to intervention date: weeks −15 through −1 for the preintervention period and weeks 1 through 40 for the intervention period. Weeks −15 and 40 were partial weeks because the overall quality improvement analysis included events occurring October 2021 through October 14, 2022. Data from units participating in the pilot project were included with hospital-wide preintervention data.

During the preintervention period, and to assess order appropriateness in the preintervention period only, a random sample comprising 15% of completed inpatient C. difficile orders (resulting positive or negative) was reviewed for order appropriateness, and was assessed using the described C. difficile ordering criteria. By the nature of the intervention, all orders during the intervention period were evaluated for appropriateness. To identify adverse outcomes attributed to delayed diagnoses of C. difficile, records were reviewed for C. difficile infection diagnoses for 30 days following the canceled order among patients with an order that was not approved for testing. Facility-wide prescription of antimicrobials targeting C. difficile including vancomycin oral formulation and fidaxomicin were quantified for the preintervention and intervention periods, calculated as monthly days of therapy per 1,000 patient days. Data were collected from the electronic health record and existing infection prevention and antimicrobial stewardship quality improvement databases and were stored and analyzed using Microsoft Excel (Microsoft, Redmond, WA). For orders that were canceled with the intervention, reasons testing was no longer indicated as well as subsequent C. difficile testing or diagnoses within 30 days were recorded.

Statistical methods

To demonstrate the ability of the intervention to decrease nonindicated orders, we compared the proportion of completed orders not meeting ordering criteria before and after intervention. We compared the rate of C. difficile HAI (events per 10,000 patient days) in the preintervention and intervention periods by calculating an incidence rate ratio. Because this was a quality improvement intervention with prespecified trial periods, we did not perform sample size or power calculations prior to the intervention. P values were calculated using median-unbiased estimation. We describe potential adverse events attributable to tests canceled by the intervention. Because C. difficile–targeted antimicrobial therapy was examined in an exploratory analysis, we have reported days of therapy for vancomycin and fidaxomicin in the preintervention and intervention periods. Statistical calculations were performed using Stata version 12.1 software (StataCorp, College Station, TX).