Clostridium difficile (C. diff), the spore-forming, anaerobic, Gram-positive bacillus, is at times part of the normal flora in the digestive tract. However, under certain circumstances, C. diff will produce toxins A and B with the resulting symptoms ranging from diarrhea to life-threatening inflammation of the colon. Clostridium difficile infection (CDI) is defined as the presence of clinical findings (diarrhea, pseudomembranous colitis), in concert with a positive laboratory finding (positive result for toxigenic C. diff), and no other documented cause for diarrhea.¹ Since 2000, the incidence of CDI has been steadily increasing and is the leading cause of hospital associated diarrhea, and attendant increases in health care costs.²

Diagnosing CDI infection continues to be a challenge,¹ due to the variety in specificity and sensitivity among the various diagnostic methods with false negative and false positive results leading to the wrong course of treatment.³  Recently published guidelines concur that only unformed stool should be tested for CDI (unless ileus is suspected), repeat testing should be discouraged, and test of cure should not be performed (see SIDEBAR). In order to enact these guidelines and foster an evidence-based, comprehensive approach to managing C. diff and CDI, we developed a multi-step algorithm with the input of several practitioners. This algorithm is supported by a robust interface between the electronic health record (EMR), computerized provider order entry (CPOE), and the laboratory information system (LIS). 

Maintaining a Broad Project View
MedCentral Health System in Mansfield, Ohio, is a 351-bed, non-profit health system comprising two hospitals (one of which is critical access) and serving six counties in Northeast Ohio. Heading into this project, we knew pre- and post-implementation retrospective data analysis would be valuable for the implementation and validation of the C. diff testing algorithm.

Pre-algorithm State
Prior to implementation of the C. diff testing algorithm, stool specimens were exclusively analyzed using the toxin enzyme immunoassay test. However, in recent years, this test has been found to be relatively insensitive (higher incidents of false negative results) and is no longer recommended as a stand alone test.⁴  In addition, both formed and unformed stools were accepted for C. diff toxin testing, and repeat testing and test of cure were also accepted as routine practices. The combination of these activities directly contributed to increased lengths of stay, worse outcomes, and higher costs.

Goals for the Algorithm
According to the guidelines,^1,4,5 it was imperative to replace the C. diff toxin enzyme immunoassay with one of the three alternatives proposed by the ASM: Either a three-step, two-step, or one-step algorithm.⁴ The one-step algorithm—a nucleic acid amplification test (NAAT)—used as a stand-alone test seemed to be the most convenient; however, it is also the most expensive single operation. Furthermore, NAAT is more accurate at detecting CDI than enzyme immunoassays only when used for the right pretest probability patients—patients with more than three unformed stools within 24 hours and no other cause for diarrhea.^1,6 It is important to keep in mind that colonization does not necessarily indicate infection according to the current guidelines. Thus, testing the wrong pretest probability patients with NAAT could lead to false positive results.^1,5 

To avoid this scenario, MedCentral chose to produce a two-step algorithm, the first test of which is a GDH-Toxin A/B combination lateral flow assay. If both tests (GDH antigen and toxin A/B) are positive, the final result posted to the EMR will be Positive for toxigenic C. difficile. If both tests are negative, the final result posted will be Negative for toxigenic C. difficile. In the case of a discrepancy (one positive and one negative), the test result will post as Indeterminate, and the molecular test will be ordered automatically (reflex testing). This second part of the algorithm, the confirmatory test, would be performed by NAAT with the results also appearing in the EMR (See Figure 1).



Creating and Implementing the Algorithm
The new, evidence-based algorithm was prepared based on the recommendations of the 2010 SHEA/IDSA guideline1 and the practical guidance document prepared by the ASM.⁴ In proposing this algorithm, the laboratory department defined the analytical performance aspect of the algorithm (testing methodology and availability), as well as the clinical performance aspect of the algorithm (which patients should be tested). In summary, the proposal established the following basic criteria:

• Only unformed stool should be tested unless ileus is suspected 
• Repeat testing should not be performed for seven days unless clinically indicated (result of physician-laboratory communication) 
• Test of cure should not be performed
• The first of the two-part C. diff testing algorithm should be performed in both locations (GDH and Toxin A/B) with a TAT goal of two hours, followed by an NAAT, if necessary, performed in the main hospital’s laboratory daily on first shift 
• The laboratory is responsible for the algorithm’s analytical and clinical performance 

The evidence-based C. diff algorithm proposal was endorsed by the chief medical officer, infectious disease physicians and infection prevention personnel, as well as the internal medicine department, the open-heart surgery department, and the general surgery department, with final approval made by the medical executive committee. True interdisciplinary teamwork was vital to the success of this project, as agreement among many practitioners is required for such an initiative to be successful.

Value of IT Integration and Process Review
The algorithm for C. diff testing is enabled by our health system’s information technology (IT) infrastructure. In fact, IT integration is one of the cornerstones of this project, as the algorithm relies on the electronic medical record (EMR), the laboratory information system (LIS), the interface engine, and the laboratory’s clinical decision support (CDS) system. 

The second cornerstone is the aforementioned interdisciplinary teamwork of laboratory, education, IT, nursing, administration, social work, infection control, and physician staff.
Before implementing the algorithm, educational tools were developed for nurses, physicians, social workers, area nursing homes, and laboratory personnel. In tailoring these education and training tools, we stressed the review of prevention methods for stopping the spread of C. diff. Through this process, we determined that testing should be limited to patients presenting greater than three non-formed stool specimens per 24-hour period, or patients who present to the hospital having had diarrhea for 24 hours or three diarrhea stools prior to admission.^1,4 If a patient is readmitted after having tested positive for C. diff in the last 90 days, an EMR alert will inform the nurse of the patient’s history upon admission. If the patient is experiencing loose stools at the time of admission, he or she will be placed in isolation and a stool specimen will be obtained to test for C. diff, assuming this was not performed within the past seven days. An EMR alert was also set up to inform infection control and nursing staff (inpatient or outpatient) of any positive results in real time.

Putting the Algorithm to Work 
The algorithm starts in the EMR/CPOE system with order templates for C. diff testing developed specifically to guide the prescriber to adhere to the 2010 SHEA-IDSA guidelines. During the order process, a summary of the guideline recommendations is provided, as well as links to the actual guidelines and to the medical-staff-approved protocol. As an automatic part of the process, when a prescriber places an order using the EMR/CPOE interface, the clinical reason for the order (diarrhea vs ileus) is documented, and this information is attached permanently to the order used by nursing when collecting stool samples. Nursing staff are trained and empowered not to send formed stools when the order is Clostridium difficile Routine - Diarrhea Stool.


Once the sample is collected and received in the lab, the first step of the C. diff algorithm is to evaluate the sample’s consistency (formed vs unformed) to match the clinical indication (diarrhea vs ileus). If the stool is formed and the clinical indication is diarrhea, laboratory personnel will respond NO in the Result field, which will automatically reflect the first comment shown in Figure 2. This will be the result sent to the EMR and consequently, the patient’s account will not be charged for the C. diff test. 



In the event the sample and the clinical indication are in agreement, laboratory personnel will respond YES. The system will then check for previous results within seven days by medical record number. If results are found, no test will be performed (see  Previous Interpretation in Figure 2), and the patient’s account will not be charged; however, the system can be overridden when repeat testing is clinically indicated.^1,6 For example, assume a patient is on antibiotics for a UTI (tested negative for C. diff infection four days prior) and comes to the emergency department (ED) with intractable diarrhea because they were exposed to a relative with CDI (two days prior to the ED visit). If the prescriber follows the exact process outlined in the algorithm, the patient will not be tested for C. diff for an additional three days, which would cause a delay in diagnosis and treatment. In this case the ordering physician can consult with laboratory personnel and communicate the extenuating circumstances for repeat testing and a new test will be performed.

Finally, in the event the sample and the clinical indication are in agreement and no previous testing was performed within seven days, the LIS will automatically reflex the orders for the GDH-Toxin A/B testing, and NAAT, when necessary. Because the testing is performed in this case, the patient’s account will be charged accordingly.

Assessing the Impact
This evidence-based algorithm went live on October 16, 2012. To assess its impact, three months of post-implementation data were compared retrospectively to three months of pre-implementation data (see Table 1). During pre-implementation, 910 C. diff tests were performed, versus 423, post-implementation, resulting in a testing decrease of 52.2%. 

Click here to view a larger version of this Table

To adhere to the 2010 SHEA-IDSA guidelines, a two-point strategy was created to support the algorithm. The first point guides prescribers to order C. diff testing under the right clinical circumstances and at the same time, document the clinical indication for testing. This strategy has proven to be effective in the inpatient environment where sample collection dropped 49.6% (from 472 to 238) as shown in Table 1; however, this did not work as well in the outpatient environment, which only saw an 8.2% decrease, because physicians do not order the test through the EMR/CPOE system and nursing is not empowered in this environment to collect appropriate samples. 

The second point of the algorithm is to empower lab personnel through IT to perform a test only when clinically indicated. As mentioned, once the sample reaches the lab, it is checked for consistency related to the clinical indication. After implementation of the algorithm, only 12.2% of orders were found to be inappropriate for testing when the sample was collected by inpatient nurses, versus 31.8% (138 samples) among the outpatient samples. In order to address this gap, MedCentral is currently implementing a health information exchange (HIE) for its 27 nursing home clients and 24 group homes/assisted living facilities, as the intention is to leverage the HIE to eventually work as an outpatient CPOE in order to avoid unnecessary collections and pickups.

Among other interesting post-implementation data, we found that 9.2% of the inpatient samples and 9% of the outpatient samples received by the lab were repeat testing as indicated by our CDS system. To address this, after performing the stool-consistency evaluation, the order is checked for any previous results within the last seven days by medical record number (MRN). The MRN was chosen to take advantage of the longitudinal record-retention capability of the EMR, regardless of provider change. When a test result is found within the last seven days, a C. diff previous interpretation result will be sent to the EMR (see Figure 2) that will include the previous result and the date and time of the collection. Of note, the repeat testing criteria were only overridden once during the first three months of this project due to clinical indication on that particular patient. Since the end of the trial period, there have been a few more similar incidents, especially when the endoscopy showed pseudomembranous colitis (an inflammation often, but not always, caused by C. diff) and the previous test was negative. In all of these cases, the final result was negative. 

Final Results and Continuous Improvement
The laboratory performed 423 tests with the two-step algorithm during the study management phase. Of these, 88.4% (374 out of 423) delivered results after first-step testing (GDH-Toxin A/B) with a TAT maximum of two hours; 11.6% (49 of the 423) needed further confirmation by NAAT. This result agrees with previously published data.⁷ In the inpatient setting, nursing staff is responsible for maintaining sample integrity and they make sure the sample reaches the lab within 30 minutes. It is worth noting that C. diff toxin degrades within two hours at room temperature, which can lead to false negative results for the first test (GDH-Toxin A/B). Such false negatives will then trigger the need for NAAT, which in turn delays the diagnosis of CDI. Therefore, in the nursing home setting, nurses are positively impacting the quality of the samples received in the lab by immediately refrigerating the samples after collection. When this is done, 74% of positive results are determined with the first step testing (GDH-Toxin A/B). Overall, the performance of this algorithm has helped the lab decrease the cost of C. diff testing per month from $5,468.17 to $3,972.66, even though the new assays are more expensive.

Conclusion
By testing the right pretest-probability patients, improving sample integrity and testing strategies, and introducing real-time EMR alerts to nursing and infection control of positive C. diff results, the length of stay for patients diagnosed with CDI decreased from 12.9 to 8.4 days, resulting in an average total hospital cost savings per patient of $9,849.50; this translates into a total annual savings of approximately $1.1 million per year. Further cost analysis reveals that nursing and pharmacy cost reductions are proportional to the length of stay (34.8% reduction) as shown in Table 2; however, laboratory costs are not (52.3%), indicating that the longer the patient stays in the hospital, the more laboratory tests were over-utilized (in CDI-diagnosed patients). 

Click here to view a larger version of this Table

The algorithm also had a positive impact on the discharge disposition of patients with CDI. In the pre-implementation group 18% died, 18% went home (including home health care), and 64% went to nursing home care. In the post-implementation phase, 54% went home (including home health care) and 46% went to nursing home care. This shift in discharge disposition implies improved outcomes and cost savings for the entire health care system.

Ultimately, by leveraging interdisciplinary teamwork and IT, MedCentral health system improved C. diff testing utilization, decreased hospital costs, decreased nursing sample collection volume, and decreased lengths of stay, while simultaneously improving patient outcomes for patients with CDI. 

References 

1. Cohen SH, Gerding DN, Johnson S, et al. Clinical practice guidelines for Clostridium difficile infection in adults: 2010 Update by The Society for Healthcare Epidemiology of America (SHEA) and the Infectious Diseases Society of America (IDSA). Infect Control Hosp Epidemiol. 2010;31(5):431-455.
2. O’Brien JA, Lahue BJ, Caro JJ, et al. The emerging infectious challenge of clostridium 
difficile-associated disease in Massachusetts hospitals: clinical and economic 
consequences. Infect Control Hosp Epidemiol. 2007;28(11):1219-1227. 
3. Planche T, Wilcox M. Reference assays for Clostridium difficile infection: one or two gold standards? J Clin Pathol. 2011;64(1):1-5. 
4. A Practical Guidance Document for the Laboratory Detection of Toxigenic Clostridium difficile—Revised. American Society for Microbiology. September 21, 2010.Available at: www.asm.org/images/pdf/Clinical/clostridiumdifficile9-21.pdf 
5. Surawicz CM, Brandt LJ, Binion DG, et al. Guidelines for diagnosis, treatment, and prevention of Clostridium difficile infections. Am J Gastroenterol. 2013;108(4):478-498. 
6. Swindells J, Brenwald N, Reading N. Evaluation of diagnostic tests for Clostridium difficile infection. J Clin Microbiol. 2010;48(2):606-608.
7. Sharp SE, Ruden LO, Pohl JC, et al. Evaluation of the C. Diff Quik Chek Complete Assay, a New Glutamate Dehydrogenase and A/B Toxin Combination Lateral Flow Assay for Use in Rapid, Simple Diagnosis of Clostridium difficile Disease. J Clin Microbiol. 2010;48(6):2082-2086 

Eugenio H. Zabaleta, PhD, is a clinical chemist at MedCentral Health System in Mansfield, Ohio. He is also a part-time lecturer at Cleveland State University’s graduate clinical chemistry program. Dr. Zabaleta graduated from the Catholic University of Cordoba (Argentina) with a degree in biochemistry and received his PhD in chemistry from the University of Akron.

SIDEBAR

C. Diff Testing Guidelines

• Clinical Practice Guidelines for Clostridium difficile Infection in Adults: 2010 Update¹
  Society for Healthcare Epidemiology of America (SHEA) and the Infectious Diseases Society of America (IDSA)
• A Practical Guidance Document for the Laboratory 
  Detection of Toxigenic Clostridium difficile, (2010)4The American Society for Microbiology (ASM)
• Guidelines for diagnosis, treatment, and prevention of Clostridium difficile infections (2013)⁵
  [Supplement to the 2010 SHEA/IDSA guidelines]
  American College of Gastroenterology