Study staff conducted daily, prospective, active surveillance for incident diarrhea cases (>3 stools with Bristol scale >5 in previous 24 hours) among eligible inpatients (Louisville residents >50 years of age) by visiting inpatients, reviewing medical charts, and meeting with nursing staff. Surveillance was conducted on all 119 wards (including 26 intensive care units [ICUs]) at 8 of 9 Louisville adult hospitals from October 14, 2019, through October 13, 2020, with a surveillance pause during April 12, 2020‒August 16, 2020 because of hospital restrictions enacted in response to COVID-19. Participating hospitals had 84.4% (2,596/3,077) of the Louisville adult hospital beds. Population demographics of Louisville and the United States are similar (12).

All eligible inpatients with incident diarrhea were invited to participate in the study. After written informed consent was obtained, a study stool specimen was collected by study staff from inpatients with diarrhea and screened by using C. Diff Quik Chek Complete, a rapid, membrane, enzyme-linked immunosorbent assay (Alere Techlab, https://www.techlab.com) (13). Before the COVID-19 pause, study stool specimens that were glutamate dehydrogenase (GDH)‒positive or GDH-negative/toxin-positive were sent to the Pfizer laboratory in Pearl River, New York, USA, for NAAT and, if NAAT positive, for automated cell cytotoxicity neutralization assay (CCNA) testing. CCNA testing at the Pfizer laboratory has been validated to provide specific, sensitive, and reproducible results to support CDI clinical and epidemiology studies (14).

After the pause, all study specimens were tested by Quik Chek and sent to the Pfizer laboratory for NAAT and CCNA testing. Shipped specimens were frozen on dry ice (replenished daily) and stored, upon receipt, in −80°C freezers. Throughout the study, patients with GDH-positive or GDH-negative/toxin-positive specimens were followed up by study staff for 90 days to determine outcomes, and SOC stool specimen collection by hospital staff occurred in parallel and independent of study procedures. For SOC CDI testing at the 8 participating hospitals, 2 used Quik Chek only, 5 used Quik Chek with NAAT of Quik Chek discordant specimens (but during the study, 1 changed to NAAT-alone testing and 1 to NAAT with toxin enzyme-linked immunosorbent assay testing of NAAT-positive), and 1 used NAAT-alone testing. Actual SOC test results (i.e., NAAT-positive or NAAT-negative) were provided to clinicians. For our analysis, we defined NAAT-positive/toxin-negative as SOC C. difficile carriage.

A study CDI case-patient was a patient who had a NAAT-positive/CCNA-positive specimen or a patient who had a NAAT-positive specimen and pseudomembranous colitis (PMC). A CDI recurrent case was the occurrence of diarrhea <56 days after resolution of a previous case of diarrhea. A primary CDI case was a nonrecurrent CDI case. CDI in hospitalized case-patients was classified as hospital-onset (i.e., positive stool specimen collected >3 days after admission), community-onset (i.e., positive stool specimen collected in an outpatient setting or <3 days after admission), healthcare-associated (i.e., positive stool specimen in a person with hospital-onset or in a person with community-onset with a documented overnight stay in the 12 weeks before stool specimen collection), or community-associated (i.e., positive stool specimen in a person with community-onset with no documented overnight stay before the current admission in the 12 weeks before stool specimen collection) (15).

We reviewed medical records to compare the characteristics of inpatients with diarrhea who were enrolled and not enrolled in the study and to compare the characteristics of enrolled inpatients with diarrhea who had and did not have a study CDI test. We used χ² tests (for categorical or binary variables) and t-tests (for continuous variables) to evaluate differences between the groups by using SAS Studio 3.71 (https://www.sas.com).

We compared laboratory results of study and SOC specimens collected from inpatients with diarrhea to determine the frequency of discordant results indicating SOC laboratory testing CDI overdiagnosis or underdiagnosis compared with study testing. We used laboratory results of specimens from inpatients with diarrhea who had a study specimen tested but did not have a SOC specimen collected to determine the frequency of SOC CDI underdiagnosis caused by lack of SOC specimen collection. We compared laboratory results of specimens collected after the pause from inpatients with diarrhea who had a study specimen screened by GDH and tested by NAAT to determine the negative predictive value (NPV) of the GDH test compared with the NAAT.

We estimated population-based hospitalized CDI incidence among Louisville residents >50 years of age (n = 276,456), >65 years of age (n = 127,864), and >75 years of age (n = 51,509) after adjusting the number of study CDI cases for the percentage of Louisville adult hospital beds that were present in the nonparticipating hospital (multiplying by the inverse of the percentage of Louisville beds in participating hospitals [84.4%]) and for the percentage of inpatients with diarrhea who did not have a study specimen collected (multiplying by the inverse of percentage of participants with specimen collected [67.9%]). In a sensitivity analysis, we adjusted further the number of study CDI cases for specimens that were NAAT positive but with quantity not sufficient (QNS) for CCNA testing (multiplying by the inverse of the percentage of NAAT positive with CCNA testing [93.3%]) and, using the NPV of the GDH test compared with NAAT, for specimens that were not tested by NAAT (multiplying by the inverse of NPV [93.8%]).

This study was approved by the institutional review boards at each hospital. It was also conducted in accordance with the study protocol.

Study staff identified 1,541 incident diarrhea cases among 85,719 patient-days before the COVID-19 pause (Figure 1). SOC specimens were collected from 680 inpatients with diarrhea, for a SOC CDI testing density of 79.3/10,000 patient-days. Study specimens were collected from 1,047 (67.9%) inpatients with diarrhea, for a study CDI testing density of 122.1/10,000 patient-days; the study enrollment rate was 83.3% (1,283/1,541) and the specimen collection rate among enrolled participants was 81.6% (1,047/1,283) (Table 1). Patient transfer or discharge was the main reason that a specimen was not collected from all enrolled participants. Study staff identified 319 incident diarrhea cases among 27,373 patient-days after the COVID-19 pause. After the pause, study specimens were collected from 144 inpatients with diarrhea for a study CDI testing density of 52.6/10,000 patient-days; the study enrollment rate was 59.6% (190/319), and the specimen collection rate was 75.8% (144/190) (Figure 2).

Before the pause, 222 specimens were sent to the Pfizer laboratory after Quik Chek testing (74 GDH-positive/toxin-positive, 146 GDH-positive/toxin-negative, and 2 GDH-negative/toxin-positive); 163 (74.8%) were NAAT-positive. Of the NAAT-positive specimens, 103 (63.2%) were CCNA-positive, 49 (30.1%) were CCNA-negative, and 11 (6.7%) were QNS for CCNA testing. Of 103 NAAT-positive/CCNA-positive cases, 99 (96.1%) were primary CDI cases, and of the 49 NAAT-positive/CCNA-negative cases, 10 (20.4%) were primary CDI cases (7 had PMC and 3 had a specimen collected during the 90-day follow-up period that was NAAT-positive/CCNA-positive), for a total of 109 study-identified primary CDI cases. Before the pause, SOC testing identified 128 CDI cases, of which 126 were primary CDI cases (15 were identified by NAAT alone). After the pause, study testing identified 10 primary CDI cases, and SOC testing identified 22 primary CDI cases.

An SOC specimen and a study specimen were collected and tested from 145 cases before the pause, of which 7 were recurrent CDI cases, yielding 138 for the evaluation of SOC misdiagnosis; of those cases, 75 (54.3%) were SOC primary CDI cases and 79 (57.2%) were study primary CDI cases (Table 2). Of 75 SOC primary CDI cases, 12 (16.0%) were not study-diagnosed as CDI cases; 6 NAAT-positive/CCNA-negative (SOC testing: 3 GDH-positive/toxin-positive/NAAT-positive and 3 NAAT-positive/toxin-positive) and 6 NAAT-negative (SOC testing: 3 GDH-positive/toxin-positive/NAAT-positive and 3 GDH-positive/toxin-positive). Of 79 study primary CDI cases, 16 (20.3%) were not SOC-diagnosed as CDI cases; SOC test results for those 16 were 13 NAAT-positive/toxin-negative, 2 NAAT-negative, and 1 GDH-negative/toxin-negative. In addition, 70 case-patients before the pause had a study specimen tested but without a SOC specimen collected; of those, 28 were study primary CDI case-patients. Therefore, 44 study-diagnosed primary CDI cases were not SOC diagnosed: 16 (36.4%) SOC-undiagnosed after SOC testing and 28 (63.6%) SOC-undiagnosed because of lack of specimen collection. Of 28 study-diagnosed CDI case-patients who were not SOC diagnosed because of lack of specimen collection, 11 (39.3%) were taking laxatives.

The median age of the 109 study inpatients with primary CDI identified before the pause was 72 (range 50–98) years. A total of 63 (57.8%) were women. Based on the diarrhea onset date, 23 (21.1%) were community-onset and community associated and 86 (78.9%) were healthcare associated; of the healthcare-associated cases, 55 (64.0%) were community-onset and 31 (36.0%) were hospital-onset. Of the 109 study patients with primary CDI, 18 (16.5%) had PMC, 36 (33.0%) were admitted to an ICU, and 21 (19.3%) died within 90 days of CDI diagnosis. The median patient age among the CDI cases who died was 78 (range 56–95) years.

Characteristics of 44 study-diagnosed but SOC-undiagnosed inpatients who had primary CDI before the pause and the 126 SOC-diagnosed inpatients who had primary CDI before the pause were similar except that the SOC-undiagnosed inpatients who had CDI were more likely to be hospital-onset CDI case-patients and less likely to have PMC (Table 3). Of the 19 study-diagnosed but SOC-undiagnosed hospital-onset CDI cases, 16 (84%) were SOC-undiagnosed because an SOC specimen was not collected. On the basis of the 109 study-diagnosed primary CDI cases before the pause, SOC practices underdiagnosed 40.4% (44/109) of the study-diagnosed primary CDI cases. Because SOC overdiagnosed 16.0% of the study-diagnosed primary CDI cases and underdiagnosed 40.4% of study-diagnosed primary CDI cases, SOC testing identified 24.4% fewer primary CDI cases than did the study.

After the pause, study specimens from 142 cases were screened by Quik Chek and tested by NAAT; of the 113 GDH-negative/toxin-negative specimens, 106 (93.8%) were NAAT-negative. Therefore, the NPV of the Quik Chek GDH test compared with NAAT was 93.8%, indicating that an estimated 6.2% of NAAT-positive specimens might have been missed because of the screening used during the pre‒COVID-19 period.

A review of the medical records of inpatients with diarrhea before the pause (Table 4) indicated that patients who were enrolled and those not enrolled had similar demographics and medical histories, except enrolled patients were more likely to be women and less likely to die in the 90 days after diarrhea onset. Among diarrhea case-patients who were enrolled (Table 5), patients with and without a stool specimen collected for CDI testing had similar demographics and medical histories, except that those who had a stool specimen collected were less likely to be Black and more likely to have taken antimicrobial drugs in the previous 3 months.

The study incidence was 154 hospitalized CDI cases/100,000 population/year and the SOC incidence was 121 hospitalized CDI cases/100,000 population/year for persons >50 years of age before the pause. The study hospitalized CDI incidence was 226/100,000 population/year for persons >65 years of age and 334/100,000 population/year for persons >75 years of age. In the sensitivity analysis adjusted for CCNA QNS and for stool specimens that were not NAAT/CCNA tested because of the GDH screen, the study incidence was 202 hospitalized CDI cases/100,000 population/year for persons >50 years of age, 296 hospitalized CDI cases/100,000 population/year for persons >65 years of age, and 438 hospitalized CDI cases/100,000 population/year for persons >75 years of age.