Impact of an Electronic Hard-Stop Clinical Decision Support Tool to Limit Repeat Clostridioides difficile Toxin Enzyme Immunoassay Testing on Test Utilization

Abstract

We performed an intervention evaluating the impact of an electronic hard-stop clinical decision support tool on repeat Clostridioides difficile (CD) toxin enzyme immunoassay (T-EIA) testing. The CD testing rate and number of admissions with repeat tests decreased significantly post-intervention (p<0.01 for both); the percentage of positive tests was unchanged (p=0.27).

INTRODUCTION

The diagnosis of Clostridioides difficile infection (CDI) requires the presence of appropriate clinical signs and symptoms in combination with a positive test for toxigenic C. difficile.¹ Historically, toxin enzyme immunoassays (T-EIA) were the most commonly used test. Due to concern regarding low sensitivity of toxin T-EIA tests, there is a common misconception that three sequential tests are necessary to “rule out” CDI.^2,3 This practice leads to a decrease in the positive predictive value of each subsequent test, and therefore an increase in the likelihood of false positive results; previous data at our institution indicates this may elevate the “CDI” rate by up to 32%.^3,4 This is significant because false positive tests can lead to unnecessary treatment, which can lead to adverse drug side effects, unnecessary costs, and a paradoxical increase in the risk of developing actual CDI.⁵

Diagnostic stewardship can be used to improve the appropriate utilization of the C. difficile T-EIA test. Specifically, the use of a computerized clinical decision support (CDS) tools have been shown to improve practitioner laboratory test ordering practices.⁶ Our objective was to improve C. difficile T-EIA test utilization by implementing a hard-stop CDS tool to limit repeat C. difficile T-EIA testing within 96 hours of a negative T-EIA test, and measuring the impact of this intervention on test utilization.

MATERIALS AND METHODS

Setting:

This quality improvement project was performed at Barnes-Jewish Hospital, a 1,250-bed tertiary care hospital in St. Louis, Missouri, from January 2015 to August 2015. Inpatients ≥18 years old were eligible if they had a C. difficile T-EIA test ordered. Publication of results was approved by the Washington University Human Research Protection Office.

Intervention:

A hard-stop intervention was placed in the electronic medical record (EMR) system during the project period that limited repeat C. difficile T-EIA testing within 96 hours of a previous negative test.^4,7 If the clinician felt that a repeat test was indicated, he/she could contact the laboratory medicine resident on-call to request a repeat test. Appropriate testing parameters, including the presence of clinically significant diarrhea (CSD)¹ and the absence of a laxative, were discussed with the provider; ultimately the decision to order the repeat test was at the discretion of the treating clinician. A hard-stop clinical decision support (CDS) intervention limiting repeat C. difficile T-EIA testing within 10 days of a positive test was also implemented.

Education/Washout Period:

Institution-wide education was provided to ordering providers on appropriate C. difficile test utilization and the quality improvement intervention. In-person education and presentations were provided to any individual, division, or department who requested it. Prior to implementation and ongoing through the intervention, education and training were provided to lab medicine residents.

Laboratory Testing for C. difficile:

Fecal samples submitted to the clinical laboratory were tested with the TechLab^® Toxin A/B II™ EIA (Alere™, Blacksburg, VA); testing was rejected on formed fecal specimens. If available, remnant feces was frozen at −80°C. In the pre- and post-intervention period, culture for toxigenic C. difficile and Xpert C. difficile/Epi polymerase chain reaction (PCR) testing (Cepheid^®, Sunnyvale, CA) was retrospectively performed on index and repeat stool specimens from patients for whom repeat testing was requested and remnant stool was available. C. difficile culture and identification were performed per previously published procedures.⁸

Statistical Analysis:

Patient demographics, C. difficile testing frequency and rates, and patient outcomes during the three months pre-intervention (January to March 2015) and the three months post-intervention (June to August 2015) were compared. The education phase was considered a washout period (April to May 2015) and was excluded from all analyses. Data were collected electronically from the hospital’s Medical Informatics databases and through chart review. Data obtained included receipt of an ICD-9 code for CDI (008.45), discharge location, death within 30 days, colectomy due to CDI, and antimicrobials for CDI. ICD-9 codes were used to classify comorbidities according to the Quan adaptation of the Charlson-Deyo index.⁹ Chi-square analyses and univariate logistic regression were used for analyses of categorical variables, and Mann-Whitney U tests for continuous variables (SPSS version 21, IBM Statistics, Armonk, NY).

RESULTS

Demographics:

There were 1,146 admissions with C. difficile tests in the pre-intervention period (rate=6.86 per 100 admits), and 982 in the post-intervention period (rate=5.67 per 100 admits). There was no difference in the Charlson composite score between the two groups (data not shown).

Test Utilization:

The testing rate pre-intervention was 9.12 per 100 admits vs.6.94 per 100 admits post-intervention (p<0.01; Table 1). The hard stop CDS alert fired a total of 293 times during the post-intervention period. Of these, 156 were “duplicate alerts,” where the alert was shown repeatedly to the same ordering provider multiple times on the same calendar day for the same patient (median 2 alerts, range 1–25 alerts per patient). There was no significant difference in the overall percentage of all positive assays pre- and post-intervention (6% vs. 7%; p=0.27) or in the percentage of first assays per admission that were positive (7% vs. 7%; p=0.56). There was a significant reduction in the number of admissions with repeat tests <96 hours from an initial negative test, from 124 (11%) admissions (among those with a test) pre-intervention vs. 15 (2%) post-intervention (p<0.01). Among admissions during which a test was ordered, there was a significant reduction in the mean number of tests per admission (1.42 vs. 1.25; p<0.01) and the mean number of assays <96 h apart per admission (0.13 vs. 0.02; p<0.01), and a significant increase in the days between the first and second test, (7.6 vs 9.5; p<0.01).

Table 1.

Testing practices pre- and post-clinical decision support intervention

Variable	3 months pre-intervention N(%) or mean (range)	3 months post-intervention N(%) or mean (range)	OR (95% CI)	p
Testing practices
Number of assays	1525	1203
Negative assays (overall)	1432 (94)	1117 (93)	Reference
Positive assays (overall)	93 (6)	86 (7)	1.2 (0.9 – 1.6)	0.27
Negative assays (first per admission)	1074 (93)	910 (93)	Reference
Positive assays (first per admission)	77 (7)	72 (7)	1.1 (0.8 – 1.5)	0.56
Number of admissions with an assay	1146	982
Number of assays per admit	1.42 (1 – 14)	1.25 (1 – 6)	Not applicable	<0.01*
Time (days) between first and second T-EIAs	7.6 (0 – 64)	9.5 (<1 – 41)	Not applicable	<0.01*
Admissions with T-EIAs <96 hours apart	124 (11)	15 (2)	0.1 (0.1 – 0.2)	<0.01

^*Mann-Whitney U test

Clinical Outcomes:

There were no significant differences post-intervention in patient discharge location, patients who received the CDI ICD-9 code, all-cause death within 30 days, or colectomy due to CDI (Table 2). There were no significant differences in the proportion of patients on an antibiotic targeting C. difficile before or after the stool collection date, either overall, among patients with a negative test first, or among patients with repeat tests (Table 2).

Table 2.

Clinical outcomes and C. difficile treatment pre- and post-clinical decision support intervention

Variable	3 months pre-intervention (N=1146 admissions) (n[%])	3 months post-intervention (N=982 admissions) (n[%])	OR (95% CI)	p
Clinical outcomes
Discharge location
Home, including with home health	708 (62)	639 (65)	Reference
Healthcare facility	304 (27)	229 (23)	0.8 (0.7 – 1.0)	0.08
Died or discharged on hospice	127 (11)	111 (11)	1.0 (0.7 – 1.3)	0.82
Unknown	7 (1)	3 (<1)	0.5 (0.1 – 1.8)	0.28
Received ICD-9 code for CDI	104 (9)	92 (9)	1.0 (0.8 – 1.4)	0.82
Died within 30 days of T-EIA	161 (14)	108 (11)	0.8 (0.6 – 1.0)	0.04
Colectomy due to CDI	0	0
CDI antibiotic stopped within 48 hours of T-EIA result date
Metronidazole	123 (11)	114 (12)	1.1 (0.8 – 1.4)	0.52
Oral/rectal vancomycin	52 (5)	45 (5)	1.0 (0.7 – 1.5)	0.96
Any CDI antibiotic	151 (13)	140 (14)	1.1 (0.9 – 1.4)	0.47
Among patients whose initial T-EIA was negative:
First CDI treatment started after T-EIA result date	106 (10)	91 (10)	1.0 (0.8 – 1.4)	0.95
CDI antibiotic within 48 hours before T-EIA collection date	167 (16)	162 (18)	1.2 (0.9 – 1.5)	0.19
CDI antibiotic stopped within 48 hours of T-EIA result date	137 (13)	132 (15)	1.2 (0.9 – 1.5)	0.27
Among patients with T-EIAs <96 hours apart whose initial T-EIA was negative:
First CDI treatment started after T-EIA result date	28 (24)	5 (33)	1.6 (0.5 – 5.2)	0.52
CDI antibiotic within 48 hours before T-EIA collection date	37 (31)	3 (20)	0.6 (0.1 – 2.1)	0.55
CDI antibiotic stopped within 48 hours of T-EIA result date	30 (25)	4 (27)	1.1 (0.3 – 3.6)	1.00

^aMetronidazole orders may have been for conditions other than CDI.

Repeat Testing for C. difficile Pre-Intervention Period:

There were 124 admissions pre-intervention during which a repeat test within 96 hours of an index negative test was performed. Of these, remnant stool from the index C. difficile test was available from 88 patients (all T-EIA negative) for toxigenic culture and PCR testing; 70 (80%) had a negative T-EIA, toxigenic culture, and PCR result from the index fecal specimen, and a negative T-EIA from the repeat sample. Nine (10%) had a T-EIA and toxigenic culture negative index fecal specimen and a negative T-EIA from the repeat fecal specimen; however; not enough remnant specimen was available for PCR. Nine patients had discordant test results via T-EIA, toxigenic culture, and/or PCR. Of those 9 patients, 1 patient had a negative index T-EIA result but a positive index toxigenic culture and PCR. This patient had a subsequent positive repeat T-EIA result and was diagnosed with CDI. None of the other 8 patients were diagnosed with CDI during their index hospitalization.

Repeat Testing for C. difficile Post-Intervention Period:

There were 15 admissions post-intervention with a repeat test within 96 hours of an index negative test. Of these admissions, remnant stool from the index test was available from 11 patients for toxigenic culture and PCR. All 11 patients were negative on their index T-EIA. Of the 11 patients, 7 were negative for C. difficile via T-EIA, PCR, and toxigenic culture on both their index and repeat; none were diagnosed with CDI. There were 4 patients with either discordant test results or not enough specimen available for further testing, and of those, 3 were diagnosed with CDI via T-EIA on repeat testing.

DISCUSSION

The primary concern surrounding the use of T-EIA tests alone for C. difficile detection is poor sensitivity;¹⁰ thus repeat testing may seem to “protect” against missed CDI diagnoses due to false negative T-EIAs. Multiple studies have shown that repeating T-EIA tests within a short period of time has limited diagnostic utility.^7,11,12 The purpose of this quality improvement study was to evaluate the impact of an electronic hard-stop clinical decision support intervention limiting repeat C. difficile T-EIA testing. Our intervention resulted in significant decreases in CDI testing rates and mean number of tests per admission. There were no significant differences in patient discharge location or increases in 30-day mortality post-intervention. Antibiotic utilization data indicated clinicians were not treating empirically for missed cases of CDI. In addition, the overall rate of positive tests did not change post-intervention, but the rate of CDI diagnosis after repeat test increased, from one patient (out of the 88 with remnant stool available) pre-intervention to three (out of 11 with remnant stool) post-intervention. These numbers are small, but suggest the intervention improved the selection of patients for repeat testing. Together, these data suggest that an EMR-based hard-stop intervention effectively reduced unnecessary testing without negatively impacting the variables measured.

Most previous studies evaluating EMR-based interventions designed to reduce repeat C. difficile testing have been performed in the context of molecular testing for C. difficile detection.^13–16 Our intervention was unique in that it focused on a toxin T-EIA test. One prior study evaluated repeated T-EIA testing for C. difficile, but their conclusion was the same: positive T-EIA results after initial negative results were rare (1.9%).¹⁴ The results of our study also suggest that even among facilities that use T-EIA tests alone, without a PCR or other molecular test, the practice of repeat testing may result in unnecessary testing. From a hospital administration perspective, C. difficile infection is a Centers for Medicare and Medicaid Services (CMS) value-based program; false positive test results can falsely elevate hospital CDI rates, thereby reducing reimbursements.

There are several limitations to this study. It was conducted in a limited time frame at a single institution. There is a possibility that some cases of CDI may have been “missed,” or empirically treated despite a negative test; however, clinicians were given the opportunity for repeat testing if clinically indicated, and antibiotic prescribing practices do not suggest this occurred. Further studies of the outcomes related to interventions to reduce repeat C. difficile T-EIA testing are necessary to draw more definitive conclusions about patient outcomes.

Our study supports the use of a hard-stop CDS tool to reduce repeat T-EIA testing for C. difficile within 96 hours of an initial negative test. As there are scenarios where a repeat test is clinically indicated, a mechanism to allow for testing in a clinically appropriate setting is necessary. Interventions to improve diagnostic stewardship for C. difficile in the EMR system should be considered even among hospitals that use a toxin T-EIA instead of a PCR-based test for C. difficile identification. In era of increasing focus on diagnostic stewardship and appropriate test utilization, this study provides evidence supporting the role of clinical decision support in improving C. difficile test utilization, regardless of whether the diagnostic method used is PCR or T-EIA.