Clostridioides difficile Near-Patient Testing Versus Centralized Testing: A Pragmatic Cluster Randomized Crossover Trial

Abstract

Background

Management of suspected Clostridioides difficile infection (CDI) in the hospital setting typically results in patient isolation, laboratory testing, infection control, and presumptive treatment. We investigated whether implementation of rapid near-patient testing (NPT) reduced patient isolation time, hospital length of stay (LOS), antibiotic usage, and cost.

Methods

A 2-period pragmatic cluster randomized crossover trial was conducted. Thirty-nine wards were randomized into 2 study arms. The primary outcome measure was effect of NPT on patient isolation time using a mixed-effects generalized linear regression model. Secondary outcomes examined were hospital LOS and antibiotic therapy based on a negative binomial regression model. Natural experiment (NE), intention-to-treat (ITT), and per-protocol (PP) analyses were conducted.

Results

During the entire study period, a total of 656 patients received NPT for CDI and 1667 received standard-of-care testing. For the primary outcome, a significant decrease of patient isolation time with NPT was observed (NE, 9.4 hours [P < .01]; ITT, 2.3 hours [P < .05]; PP, 6.7 hours [P < .1]). A significant reduction in hospital LOS was observed with NPT for short stay (NE, 47.4% [P < .01]; ITT, 18.4% [P < .01]; PP, 34.2% [P < .01]). Each additional hour delay for a negative result increased metronidazole use (24 defined daily doses per 1000 patients; P < .05) and non-CDI-treating antibiotics by 70.13 mg (P < .01). NPT was found to save 25.48 US dollars per patient when including test cost to the laboratory and patient isolation in the hospital.

Conclusions

This pragmatic cluster randomized crossover trial demonstrated that implementation of CDI NPT can contribute to significant reductions in isolation time, hospital LOS, antibiotic usage, and healthcare cost.

Clinical Trials Registration. NCT03857464.

A pragmatic cluster randomized crossover trial in a large urban hospital demonstrated that near-patient testing for Clostridioides difficile reduces patient isolation time, hospital length of stay, antibiotic usage, and healthcare cost.

Clostridioides difficile infection (CDI) remains one of the most common cause of nosocomial diarrhea and places significant clinical and economic burden on health systems around the world [1–3]. The United States Centers for Disease Control and Prevention lists C. difficile as an urgent threat for antimicrobial resistance [4]. Rapid and accurate laboratory diagnosis is crucial for CDI management, targeted antibiotic therapy, and effective infection prevention and control [1, 5]. Current diagnostic technologies do not provide an accurate stand-alone diagnostic test for the detection of CDI, leading to recommendations for 2-step testing algorithms to increase accuracy while reducing overall cost in screening [6–9]. A commonly used 2-step algorithm includes screening upon suspicion of CDI for the presence of C. difficile glutamate dehydrogenase (GDH), a metabolic enzyme produced by C. difficile. Screening is commonly followed by a nucleic acid amplification test (NAAT) such as quantitative polymerase chain reaction (qPCR) targeting toxin-producing genes [8–10]. A GDH screen followed by NAAT combines the benefits of a screening assay with a high negative predictive value with a confirmation test that is highly specific for the detection of toxigenic C. difficile [11–13].

Hospitals typically place patients with acute diarrhea of unknown cause in isolation under additional contact precautions [10, 14]. Contact precautions can impose both economic and psychological burdens on the individual and the healthcare systems [15–18]. Investigation for gastrointestinal pathogens is typically initiated via laboratory tests upon suspicion of CDI, and the duration of patient isolation depends largely on the time to acquire test results to support or refute clinical suspicion [19]. Studies have reported that the majority of patients with suspected CDI are confirmed CDI negative [20–22]. When severe or complicated CDI is suspected, empiric treatment is commonly initiated. The Infectious Diseases Society of America (IDSA) and the Society for Healthcare Epidemiology of America (SHEA) recommend vancomycin or fidaxomicin for an initial CDI episode, with metronidazole being recommended for initial CDI episode only if non-severe and access to vancomycin or fidaxomicin is limited [10, 23]. Among hospitalized patients with acute diarrhea, delays in diagnosis and empiric antibiotic regimens have the potential to exacerbate antimicrobial resistance [24].

Reducing the duration between clinical suspicion of CDI and diagnostic results can reduce unnecessary and costly contact precautions, change decisions on antibiotic prescription (improved stewardship), and decrease hospital stays due to targeted clinical treatment [11]. The quaternary care hospital in our study utilizes a centralized microbiology testing facility for inpatient testing of CDI. While this decreases laboratory testing costs, it can increase time to results due to transportation and processing. Our center presents a unique opportunity to examine the effect of decreasing CDI diagnostic turnaround time by decentralizing CDI testing to near-patient testing (NPT). NPT implemented in a rapid response laboratory can deliver results within hours, while typically off-site testing takes up to 1 day. For patients without CDI, the antibiotics, and in many cases, the isolation, are unnecessary.

METHODS

Study Design and Clinical Setting

A 2-period, 2-intervention, cluster randomized crossover trial was conducted at Foothills Medical Center (FMC) in Calgary, Alberta. FMC is a large quaternary care center in western Canada with 1087 acute care beds and more than 46 000 discharges and 385 000 inpatient-days per year. This study leveraged the existing infrastructure of one of the largest centralized laboratories in Canada (Calgary Laboratory Services, now called Alberta Precision Laboratories), which provides testing to a population of 1.5 million people. The study design relied on each FMC ward as the cluster unit with 2 groups of wards in each arm, which underwent a crossover of the NPT and standard of care (SOC) intervention in order to mitigate confounding effects such as ward type and patient disease severity. Ethical approval was obtained through the University of Calgary's Health Research Ethics Board (REB-18-0397). The trial was also registered at ClinicalTrials.gov (identifier: NCT03857464). Information was protected as per Alberta Health Services protocol and data were de-identified prior to analysis.

Interventions

Within FMC, there were 39 inpatient wards that were randomly assigned to a study arm (A or B). All efforts were made to keep wards in randomly assigned arms, but some modifications to ward assignment had to be made due to clinical workflow and pragmatic considerations for NPT implementation. For instance, some ward distinctions are merely administrative with nonexistent or minimal physical barriers or difference in staff between wards. These wards were grouped to minimize unintentional contamination between study arms and interventions. Study arm A received rapid NPT for CDI for the first period (period 1) while study arm B underwent the SOC, off-site testing at a centralized laboratory. In the second period (period 2), study arm B received the NPT and study arm A the SOC. Study outcomes during period 1 and period 2 interventions were compared in both arms. The trial was initiated in January 2019. There was a washout period between 1 January and 15 March 2019. Period 1 began 15 March 2019; the crossover occurred in October 2019 with period 2 ending May 2020. Following the crossover, a 2-week wash-in period was applied to allow for implementation of the NPT intervention to take place.

Patient samples were identified for the study arm by a label placed on the stool container collected from patients in the NPT arms. Labels were placed on the container by nursing staff prior to sending the sample from the wards to the laboratory for processing. Nurse training and compliance was therefore key for study implementation and consisted of in-service training on the wards prior to NPT intervention, instructional signage on the wards and in staff areas, and periodic refreshers of study protocols by email and in person from the clinical coordinator prior to and during the study. Clinicians ordering the test were blinded to the NPT or SOC allocation. In the NPT arm, stool samples from suspected CDI patients were processed on site at the rapid response laboratory (RRL) using the C. DIFF QUIK CHEK COMPLETE (TechLab, Radford, Virginia). The RRL was located at the hospital site within walking distance of all wards. Samples identified as negative by C. DIFF QUIK CHEK COMPLETE, negative by both GDH and toxin, were reported out immediately whereas positive and inconclusive results (positive for only toxin or GDH) were confirmed using GeneXpert C. difficile (Cepheid, Sunnyvale, California) qPCR at the centralized laboratory. This was determined by the study team to be a clinically equivalent diagnostic algorithm based on a validation study (data not shown). Samples processed by the SOC were directly routed by vehicle to Calgary's centralized microbiology testing facility operated by Alberta Precision Laboratories. In the central lab, samples were first screened by Liaison C. difficile GDH (DiaSorin, Cypress, California) with negative results reported immediately and positive results confirmed by GeneXpert C. difficile. All testing was done according to the manufacturer's guidelines by certified personnel in an accredited facility.

Outcome Measures

The primary outcome was the differential effect of NPT on the duration of contact precautions (patient isolation time). Secondary outcomes included changes in hospital length of stay (LOS), quantity of antibiotic therapy, and cost-benefit analysis associated with NPT. Time from specimen collection to testing was also measured for each arm using data from the laboratory information system. Contact precautions due to CDI was defined as an order by healthcare workers for contact precautions within a 7-day infection window of CDI diagnosis. The length of time with contact precautions was defined as the calendar date and time contact precautions were completed or patient discharge date minus the calendar date and the time contact precautions were ordered. Contact precaution orders were matched to CDI diagnostic requests by date window to ensure records were correctly paired. Hospital LOS was defined as the difference between the admission and discharge date. Antibiotic therapy was defined as the administration of antibiotics initiated on the day of and up to 48 hours after specimen collection for CDI testing and was collected from physician order entry system. Antibiotic use was requested from the Pharmaceutical Information Network database for 30 days postdischarge for patients in the cohort. Total days of antibiotic therapy during the acute care encounter was defined as the calendar date antibiotic therapy was completed minus the calendar date antibiotic therapy was ordered and then quantified in milligrams. Cost analysis included reagents, consumables, and technologist time to process a single test, as well as costs associated with patient isolation such as contact precaution measures.

Statistical Analysis

Analyses were conducted in Stata software for statistics and data science (version 14.0; StataCorp, College Station, Texas) or R-Studio software (version 1.4.1103; R Foundation for Statistical Computing, Vienna, Austria) [25]. Further statistical details are described in the Supplementary Data.

RESULTS

Participant Characteristics

The patient demographics are summarized for the natural experiment (NE), intention-to-treat (ITT), and per-protocol (PP) populations in Table 1. No significant difference in age (mean range, 57.71–59.68 years) or gender (female proportion: range, 44.09%–47.22%; male proportion: range, 52.77%–55.91%) was observed for the NE, ITT, or PP populations when comparing the NPT and SOC arms. White blood cell count (WBC), albumin, and creatinine for each analysis and intervention arm showed no significant differences with the exception of WBCs in the NE population. Laboratory testing confirmed that 258 patients in both arms were positive for CDI, with positive patients comprising 9.44% of patients (132/1397) in arm A and 10.5% of patients (126/1191) in arm B. Thirty-nine inpatient wards were randomized to either NPT (23 wards) or SOC (16 wards) followed by crossover, resulting in 2 periods. During the entire study period, a total of 656 patients received NPT for CDI and 1667 received SOC testing (Figure 1). Of patients testing negative (n = 2330) for CDI, 2323 had matching inpatient records. The NPT arm was comprised of 656 patients, and 1667 patients were tested by the SOC. More individuals received SOC due to imperfect compliance with the NPT algorithm implemented by nursing staff on the wards. Of the SOC patients, 538 of these patients were erroneously assigned to the SOC due to lack of compliance in sample labeling protocols on the inpatient wards, a function of the pragmatic nature of the study design. In the NE analysis, noncompliance was ignored and patients in the NPT arm receiving SOC were assigned to the SOC arm. In the ITT analysis, we included noncompliers (NPT-designated individuals who received SOC) in the intervention group, whereas in the PP analysis we excluded all noncompliers from our sample. Isolation orders attributed directly to suspected CDI were matched with the testing data, resulting in 234 patients with isolation data in the NPT arm and 627 patients with isolation data in the SOC arm.

Table 1.

Patient Demographics and Markers of Clostridioides difficile Infection Disease Severity in the Study Population Based on Natural Experiment, Intention-to-Treat, and Per-Protocol Populations and Based on Study Arm (Near-Patient Testing Versus Standard of Care)

Characteristic	NE			ITT			PP
	NPT	SOC	P Value	NPT	SOC	P Value	NPT	SOC	P Value
Age, y, mean	58.08	59.68	.41	59.35	57.71	.42	58.08	57.71	.41
21–30	4.1%	1.3%		2.57%	2.31%		4.10%	2.31%
31–40	4.86%	6.84%		5.30%	7.52%		4.86%	7.52%
41–50	8.33%	9.12%		7.87%	10.03%		8.33%	10.03%
51–60	13.88%	19.90%		14.31%	21.61%		13.88%	21.61%
61–70	26.38%	24.11%		26.50%	22.20%		26.38%	22.20%
71–80	31.94%	25.08%		27.81%	22.97%		31.94%	22.97%
81–90	8.33%	13.02%		13.50%	9.84%		8.33%	9.84%
≥91	2.08%	3.58%		2.09%	3.47%		2.08%	3.47%
Sex
Male	52.77%	54.39%	.76	54.65%	0.559	.09	52.77%	55.91%	.52
Female	47.22%	45.6%	.76	45.04%	0.4409	.09	47.22%	44.09%	.52
Severity markers (WBCs × 109/L; Albumin g/L; Creatinine μ mol/L)
WBC	4.57 (0.49)	6.93 (0.57)	.03	6.79 (0.71)	5.7 (0.38)	.19	4.57 (0.49)	5.7 (0.38)	.16
Albumin	28.04 (1.09)	27.05 (0.59)	0.38	26.8 (0.74)	28.2 (0.75)	.2	28.04 (1.09)	28.2 (0.75)	.89
Creatinine	110.33 (10.70)	104.78 (4.71)	0.61	110.73 (6.85)	101.51 (6.10)	.32	110.33 (10.70)	101.51 (6.10)	.44
No.	170	415		304	281		170	281

Abbreviations: ITT, intention-to-treat; NE, natural experiment; NPT, near-patient testing; PP, per-protocol; SOC, standard of care; WBC, white blood cell count.

Figure 1.

Data collection and matching workflow for the eligibility, randomization, testing of patients, and outcomes in the trial. Thirty-nine inpatient wards acted as clusters and were randomized into 2 study arms in a 2-period crossover trial. Arm A received the near-patient testing (NPT) intervention during period 1, with arm B receiving the standard of care (SOC). During period 2, the NPT intervention switched to the opposite arm. Patients negative for Clostridioides difficile infection (CDI) were evaluated for NPT effects on isolation time. Patient laboratory data were matched with hospital data containing isolation orders for patients with suspected CDI and the primary outcomes determined. Patients positive for CDI contributed to secondary outcome analysis.

Primary Outcomes

For patients having a negative test result, the NE analysis showed a reduction in patient isolation time by 9.22 hours (P < .01, n = 585) (Table 2). The ITT analysis suggests that even in the presence of low adherence, NPT led to a significant (P < .05, n = 585) reduction in contact precautions hours by 2.34 hours. The PP analysis suggests that using NPT in place of SOC testing led to a significant (P < .01) reduction in contact precaution hours by 6.44 hours (n = 451) during the entire study period. In the NE analysis when comparing pre- and post-crossover periods, a decrease was observed in patient isolation time (pre: 9.4 hours, P < .01; post: 8.1 hours, P < .01). When looking at the time of sample collection, the reduction in testing turnaround time was magnified for NPT during the day shift (Figure 2). A concomitant diurnal pattern of improved patient isolation time reduction was observed (Figure 3).

Table 2.

Primary Outcome Reduction in Patient Isolation (Hours) in the Near-Patient Testing Arm of the Cluster Randomized Trial Versus Standard of Care for the Natural Experiment, Intention-to-Treat, and Per-Protocol Populations

Analysis	Reduction in Patient Isolation for NPT Arm, h (SEM)	P Value	Sample Size, No.
NE	9.22 (1.21)	<.01	585
ITT	2.34 (1.02)	<.05	585
PP	6.44 (1.306)	<.01	451

A mixed-effects generalized linear regression model was applied.

Abbreviations: ITT, intention-to-treat; NE, natural experiment; NPT, near-patient testing; PP, per-protocol; SEM, standard error of the mean.

Figure 2.

Mean time to negative result (turnaround time) in the near-patient testing (NPT) and standard of care (SOC) testing arms by hour of day when sample collection occurred on the hospital ward for the entire study period.

Figure 3.

Mean time to removal of patient isolation from sample collection for the near-patient testing (NPT) arm and standard of care (SOC) arm by hour of day when sample collection occurred on the hospital ward for the entire study period.

Secondary Outcomes

When examining hospital LOS using NE, ITT, and PP analyses, a significant reduction in LOS was observed for short-stay (<48 hours) patients (Table 3). That is, for patients who were discharged within 48 hours of CDI testing, NPT resulted in a 47.4% (P < .01) reduction in LOS for the NE analysis, 18.4% (P < .01) for the ITT analysis, and 34.2% (P < .01) reduction for the PP analysis. For longer-stay patients, only the NE population for <4 days saw a significant reduction in LOS at 26.9% (P < .01). For CDI-negative patients with a hospital stay >24 hours, every hour a CDI negative test result was delayed, the vancomycin quantity prescribed increased by 8 defined daily doses (DDDs) per 1000 patients, although not to statistical significance. For metronidazole, a similar increase in DDDs was observed at 24 per 1000 patients with statistical significance (P < .05). For non-CDI-treating antibiotics, presumably used for other infectious issues in the patient, a delay of each hour led to an increase of 70.13 mg (P < .001; Table 4).

Table 3.

Secondary Outcome Hospital Length of Stay in the Near-Patient Testing Arm of the Cluster Randomized Trial Versus Standard of Care for the Natural Experiment, Intention-to-Treat, and Per-Protocol Populations

Hospitalization Duration	Analysis	Sample Size, No.	Change in LOS for NPT Arm, %	P Value
<48 h	NE	237	−47.4	<.01
	ITT	237	−18.4	<.01
	PP	206	−34.2	<.01
<4 d	NE	601	−26.9	<.01
	ITT	601	−2.9	NS
	PP	518	−9.9	NS
<6 d	NE	917	−3.1	NS
	ITT	917	−5.7	NS
	PP	800	−4.8	NS

A negative binomial regression model was applied.

Abbreviations: ITT, intention-to-treat; LOS, length of stay; NE, natural experiment; NPT, near-patient testing; NS, not significant; PP, per-protocol.

Table 4.

Antibiotic Use in Relation to Hourly Changes in Time to Result for Patients Who Tested Either Negative or Positive for Clostridioides difficile Infection

Antibiotic	CDI Negative (>24 h in Hospital)		CDI Positive (>2 wk in Hospital)
	Change in 1 h	SE/Sample Size, No.	Change in 1 h	SE/Sample Size, No.
Vancomycin (DDDs [2 g])	0.008	0.005/106	0.014a	0.007/168
Metronidazole (DDDs [1.5 g])	0.024b	0.008/52	−0.024b	0.007/92
Non–C. difficile antibiotics (mg)	70.13b	26.21/3636	···	···

CDI treatment drugs as well as antibiotics in general are detailed.

Abbreviations: CDI, Clostridioides difficile infection; DDD, defined daily dose; SE, standard error.

P < .05.

P < .01.

Cost Analysis

Examining costs and potential benefits of NPT yielded the following results (Table 5). The cost of SOC testing was United States dollars (USD) 15.49 for both positive and negative patients. This cost includes cost of labor, reagents, and proficiency testing. The cost of NPT was USD36.84 for a positive patient and USD21.35 for a negative patient. This equates with an increased cost of testing in NPT of USD8.95. When factoring in the cost of patient isolation, NPT resulted in a net savings of USD25.48.

Table 5.

Cost Analysis of Implementing Clostridioides difficile Near-Patient Testing Compared to Standard-of-Care Testing

Costs (USD)	NPT	SOC	Sample Size, No.
Test cost if result is positive	36.84	15.49	207 (N1)
Test cost if result is negative	21.35	15.49	830 (N2)
Average test costa	24.44	15.49	…
Additional cost of NPT	8.95
Decreased isolation hours NPT (A)	9.22
Per hour cost of isolation (B)	6.62
No. of patients in isolation	…	…	585 (N3)
Benefit per patient of using NPTb	34.43
Net benefit of NPT per patient	25.48

All costs are in USD at time of testing. Data are restricted to patients who were hospitalized <48 hours after time of testing.

Abbreviations: NPT, near-patient testing; SOC, standard of care; USD, United States dollars.

(N₁ × cost of positive test + N₂ × cost of negative test)/(N₁ + N₂).

(N₃ × A × B) / (N₁ + N₂).

DISCUSSION

Improvements in technology through point-of-care testing and NPT devices necessitates a holistic evaluation of the impact of such testing algorithms compared to standard laboratory-based testing. Given the continued economic burden of CDI and the associated significant morbidity and mortality, this trial sought to quantify the benefits of NPT. The primary outcome goal of this study was to evaluate how NPT could reduce CDI diagnostic turnaround time and what effect that would have on patient isolation (contact precautions), which is typically initiated for all patients with diarrhea in the hospital. The data support that NPT reduces testing times and patient isolation times. Furthermore, a reduction in hospital LOS was observed for short-stay individuals when NPT was made available. The significant reduction of contact precaution hours can reduce adverse effects associated with contact precautions including changes in care that produce delays, less patient–healthcare worker contact, and patient depression and anxiety [17]. The reduction of hospital LOS with NPT in patients who were hospitalized only a further 48 hours provides a scenario in which quicker testing can allow earlier discharge. Furthermore, rapid negative test results have the potential for additional cost saving through significant reduction in hospital LOS. A reduction in LOS can free up hospital beds, which are often in short supply [26]. Further health economic evaluations should be pursued regarding the impact of NPT on hospital administration. Some patients remained under contact precautions long after a negative CDI test result was reported to clinical staff. This was due to 1 of the following reasons: There were delays in removing contact precautions by the medical team, a contact precaution order was not properly entered/removed into the medical record system upon de-isolation of the patient, or the patient remained under contact precautions for reasons other than suspected CDI. Lower than expected contact precaution compliance is commonly seen in hospital settings [27].

The trial evaluated the effect of NPT on antibiotic use by physicians managing patients with suspected CDI. With each hour a negative result came back sooner, we saw an expected decrease in vancomycin and metronidazole as these first-line treatment options are often started empirically. Less clear, but interestingly, a decrease in non-CDI-treating antibiotics was also seen with a faster CDI test result. These data suggest possible further benefits in improved antimicrobial stewardship with rapid testing. The IDSA recommended that rapid diagnostic testing can enhance antimicrobial stewardship by promoting appropriate antibiotic use for certain infections [28].

Limitations of the study include noncompliance of the wards to the testing protocol. This imperfect compliance is inherent to the pragmatic study design but did result in patients intended for NPT getting SOC testing. However, the crossover design of the study mitigated the effects of such noncompliance. Furthermore, the NE, ITT, and PP analyses allowed for a more granular understanding if there was any impact of this noncompliance. When matching patients with CDI diagnostic requests to hospital records indicating patient isolation orders, it was apparent that not all patient isolation orders are captured in the electronic medical record. This lack of isolation reporting in the medical records was a major cause of attrition to the study sample size. Nevertheless, in the NE, ITT, and PP analyses, primary and secondary outcome analyses did reach significance. It is possible that implementing a second shift to conduct CDI testing at the centralized laboratory in the evening could achieve the same result with faster turnaround time. However, this was not evaluated in this study.

Even with the higher cost of NPT, we identified potential financial benefits for NPT due to the significant impact on isolation time when applying NPT. To weigh this benefit against higher test implementation cost, we applied the findings of a cost evaluation study that examined the financial impact of contact precaution measures [29]. The cost of isolation materials for donning and doffing and for cleaning and disinfection consumables was calculated based on units consumed. With a holistic evaluation of cost-benefit including the test implementation and the consequent impact on infection control, a net saving was observed.

CONCLUSIONS

This study provides evidence that NPT can be implemented into an active clinical setting and provide clinically relevant improvements to patient care metrics including reduction in test turnaround time, reduction in patient isolation time, reduction in hospital LOS, and reduction in the quantity of antibiotics prescribed. NPT for CDI should be evaluated further and considered for implementation into health systems as a cost-effective way to improve clinical outcomes for patients suspected of having C. difficile infection.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. This study was conducted with support and cooperation from the laboratory management and staff at Alberta Precision Laboratories. We would like to specifically thank the nursing and laboratory staff at the Foothills Medical Center.

Financial support. This work was supported by the Canadian Institutes of Health Research (PJT 156251) and the AMR One Health Consortium University of Calgary (Major Innovation Fund).