Reducing Inappropriate Testing for the Evaluationof Diarrhea Among Hospitalized PatientsChad E. Tewell, MD,a,1 Thomas R. Talbot, MD, MPH,a,b George E. Nelson, MD,a Bryan D. Harris, MD,a

Whitney A. Jones, PharmD,c Narinder M. Midha, MS,d David P. Mulherin, PharmD,c,e Eric B. Stephens, MBA,e

Anuj Thirwani, PharmD,c Patty W. Wright, MDa

aDepartments of Medicine, Nashville, Tenn; bHealth Policy of Vanderbilt University School of Medicine, Nashville, Tenn; cDepartment ofPharmaceutical Services of Vanderbilt University Medical Center, Nashville, Tenn; dDepartments of Pathology, Microbiology, andImmunology, Nashville, Tenn; eBiomedical Informatics of Vanderbilt University School of Medicine, Nashville, Tenn.

ABSTRACT

BACKGROUND: Diarrhea is one of the most common illnesses in the United States. Evaluation frequentlydoes not follow established guidelines. The objective of this study was to evaluate the effectiveness of acomputerized physician order entry-based test guidance algorithm with regard to the clinical, financial, andoperational impacts.METHODS: Our population was patients with diarrheal illness at a tertiary academic medical center. Theintervention was a computerized physician order entry-based test guidance algorithm that restricted the useof stool cultures and ova and parasites testing of diarrhea in the adult inpatient location vs noninterventionsites, which were the emergency department, pediatric inpatient and adult and pediatric outpatient loca-tions. We measured stool culture, ova and parasites, and Clostridium difficile testing rates from July 1, 2012to January 31, 2016. Additionally, we calculated advisor usage, consults generated, accuracy of informa-tion, and cost savings.RESULTS: There was a significant decrease in stool culture and ova and parasites testing rates at the adultinpatient (P = .001 for both), pediatric (P < .001 for both), and adult emergency department (P < .001; P = .009)locations. The decrease at the intervention site was immediate, whereas the other locations showed a delayedbut sustained decrease that suggests a collateral impact. A significant increase in the rate of stool cultureand ova and parasites testing was observed in the outpatient setting (P = .02 and P = .001). We estimatethat $21,931 was saved annually.CONCLUSIONS: A point-of-order test restriction algorithm for hospitalized adults with diarrhea reduced stooltesting. Similar programs should be considered at other institutions and for the evaluation of other conditions.© 2018 Elsevier Inc. All rights reserved. • The American Journal of Medicine (2018) 131, 193–199

KEYWORDS: Advisor; Diarrhea; Inappropriate; Testing; Unnecessary; Waste

There has been an appropriate focus within the medicalcommunity in recent years on raising awareness of reduc-ing the incidence of unnecessary medical tests, treatments,and procedures. Eliminating this waste is an important patientsafety intervention, as excessive testing can lead to inappro-priate diagnoses (eg, attributed to false-positive results in thesetting of low pretest probability) and unnecessary thera-pies that carry their own risk of adverse events and side effects.For common medical symptoms and conditions in particu-lar, addressing this waste can have a marked positive impacton patients by reducing harm and lowering costs.

Funding: Internal institutional funds were used for this study. This workwas presented at the Society for Healthcare Epidemiology of America (SHEA)Spring Meeting, Atlanta, Ga., May 18, 2016.

Conflicts of Interest: None of the authors have any conflicts ofinterest.

Authorship: All authors certify that they have participated sufficientlyin the work to take public responsibility for the content, including partici-pation in the concept, design, analysis, writing, or revision of the manuscript.

1Current affiliation: Infectious Disease of Indiana, P.S.C., Indianapolis.Requests for reprints should be addressed to Patty W. Wright, MD, Clin-

ical Affairs, Division of Infectious Diseases, Vanderbilt University Schoolof Medicine, 1161 21st Avenue South, A2200 Medical Center North, Nash-ville, TN 37232-2582.

E-mail address: Patty.W.Wright@vanderbilt.edu

CLINICAL RESEARCH STUDY

0002-9343/$ - see front matter © 2018 Elsevier Inc. All rights reserved.https://doi.org/10.1016/j.amjmed.2017.10.006

Diarrhea is one of the most common chief complaints forpatients seeking health care, with an estimated 211 millioncases of acute gastroenteritis occurring in the United Statesannually.1 It is estimated that in 2010 in this country, therewere 3.7 million emergency department visits and 1.3 millioninpatient hospitalizations with a diagnosis of infectious en-teritis or gastrointestinal symptomssuggesting foodborne illness.2 Theeconomic impact of acute diar-rhea is substantial, being estimatedat $23 billion in the United Statesalone in 1988.3 Diarrhea can becaused by a wide array of infec-tious and noninfectious etiologies,many of which have their own spe-cific diagnostic testing, which canmake the initial assessment of thepatient with acute diarrhea verychallenging. Consequently, guide-lines were developed to aideclinicians in the evaluation of pa-tients with diarrhea4; however, adherence to theserecommendations is often suboptimal, resulting in in-creased costs without clinical benefit and waste of laboratoryresources.4-6 Two specific tests utilized in the diagnosis of di-arrhea are bacterial stool cultures and stool examination forthe presence of ova and parasites. In certain clinical situa-tions, however, the utility of these tests in the diagnosticevaluation of diarrhea is poor. For example, many clinical labo-ratories will reject specimens sent for these tests in hospitalizedpatients admitted for more than 3 hospital days (known asthe “Three-Day Rule”), given the very low incidence of thesepathogens in this setting. Voluntary implementation of the“Three-Day Rule” has been shown to reduce inappropriatestool testing, improve diagnostic yield, and increasecost-effectiveness.7-9

In addition to clinical guidelines and policies, novel toolsthat target ordering physicians to provide education andfeedback,10,11 as well as using computerized physician orderentry (CPOE)-based interventions, can markedly reduce un-necessary testing.12-14 Use of such interventions to reduceunnecessary testing for Clostridium difficile infection amonghospitalized patients has been reported,14,15 but broader toolsto address the additional testing among patients hospital-ized with diarrhea are less prevalent. In response to variabilityin diarrheal testing practices among clinicians at VanderbiltUniversity Medical Center (VUMC), the Vanderbilt Antimi-crobial Stewardship Program, in conjunction with the VUMCMicrobiology Laboratory leadership, developed a CPOE-based testing algorithm for adult inpatients with diarrhea basedon published guidelines and patient-specific characteristics.The approach of using a CPOE-based testing guidance al-gorithm is not well described in the literature, although thereare suggestions that this strategy could be useful.16 Follow-ing the implementation of the testing algorithm, the VanderbiltAntimicrobial Stewardship Program team conducted an eval-uation of the tool’s impact in the adult inpatient units, as well

as an examination of any collateral impact in other clinicalareas of the medical center where the intervention was notexplicitly activated (eg, inpatient and emergency pediatric,adult emergency department, and adult and pediatric outpa-tient locations). The results of this assessment are describedbelow.

METHODS

Study DesignThis study is a retrospective, inter-rupted time series analysisexamining rates of diarrhea-relatedtest ordering and the impact of anovel test algorithm and restric-tion strategy (“testing advisor”). Itwas approved by the institutionalreview board at VUMC as a qualityimprovement investigation.

InterventionThe diarrhea testing advisor (Figure 1) is a clinical deci-sion tool that is incorporated in the electronic order entrysystem. It was developed after examining institutional testingpractices and a review of the relevant literature, with a basisin the logic of the “modified Three Day Rule.”9 Specific goalsof the advisor were to reduce orders for bacterial stool cultureand stool ova and parasites testing and to guide clinicians toseek consultation with subspecialty consultation for more com-plicated patients. For the initial implementation, adult inpatientunits were included. When a provider orders a stool culture,stool ova and parasites, and several other specialized stooltests for suspected infectious diarrhea on an inpatient in theselocations, they are automatically directed into the advisor. Aseries of questions are then posed to the provider to capturedetails of the clinical illness and underlying patient charac-teristics (eg, symptom duration, travel history, presence ofimmunocompromise, and presence of hematochezia). The ap-propriate tests or consults are recommended depending onthe information entered. For example, the advisor directs theuse of bacterial stool cultures and stool ova and parasites ex-aminations to specific patients with acute diarrhea ofcommunity onset (ie, patients with history of hospitaliza-tion of ≤3 days). If the clinician wishes to order a restrictedtest for a patient who does not meet the defined criteria, thenapproval by the infectious diseases or gastroenterology ser-vices is required. The advisor does not restrict C. difficiletesting, as a CPOE-based tool to direct appropriate C. difficiletesting (ie, limits testing to unformed stools, prohibits repeattesting in patients for 7 days) was already in place.

Evaluation of Advisor ImpactIn order to assess the impact of the advisor, the monthly testingrate of bacterial stool cultures and stool ova and parasites ex-aminations were assessed. Because C. difficile testing could

CLINICAL SIGNIFICANCE

• Diarrhea evaluation frequently does notfollow established guidelines, leading tounnecessary testing and treatment.

• A point-of-order test restriction algo-rithm for hospitalized adults reducedunnecessary stool testing, including acollateral impact at noninterventionsites.

194 The American Journal of Medicine, Vol 131, No 2, February 2018

be ordered independent from the advisor, rates of testing wereassessed as a surrogate for changes in the incidence of pa-tients presenting with diarrhea during the study period.

The intervention was activated in mid-October 2014, andtime period for this analysis was July 1, 2012 to January 1,2016, divided into the preintervention (July 1, 2012 to October31, 2014) and postintervention (November 1, 2014 to January31, 2016) time periods. Because the advisor was activatedonly in the adult inpatient area, we assessed whether therewas any collateral impact in other clinical areas of the medicalcenters. Specifically, testing rates were examined by the lo-cation of the patient at the time the order was placed, whichincluded the adult inpatient (the intervention unit) and the 3control locations: the adult emergency department, pediat-ric hospital, and adult and pediatric outpatient locations. Testingrates were calculated per 1000 patient days for the inpatientlocations and per 1000 visits for the outpatient and emer-gency department locations. To examine whether the advisormay have prohibited appropriate testing of patients, we ex-amined the absolute number and rate of positive cultures duringthe study period, calculated as number of positive stool cul-

tures and ova and parasites tests in a given month out of thetotal number of tests sent. We hypothesized that if the advisorwas too restrictive, the test positivity rate would be signifi-cantly reduced after implementation of the intervention. Thesources of data for this analysis were the VUMC EnterpriseData Warehouse and the electronic medical record.

Because subspecialty consultation was required for somebranches of the advisor, an analysis of the number of infec-tious diseases and gastroenterology consults that resulted fromthe advisor was also performed. Finally, the financial impactof the advisor was evaluated by determining the change inthe number of stool bacterial cultures and ova and parasitesexaminations between the preintervention and thepostintervention periods at the adult inpatient location. Thedifference between these rates was multiplied by the cost foreach test, which was determined from cost data previouslycalculated by the VUMC Microbiology Department. Thesecosts included the laboratory costs of the reagents and laborinvolved in the stool testing. Costs related to patient lengthof stay, nursing time, collection supplies, and antibiotic usewere not included in this analysis.

Figure 1 Diarrhea testing advisor algorithm.

195Tewell et al Reducing Inappropriate Diagnostic Testing Using a Computerized Algorithm

Statistical AnalysisAn interrupted time series analysis with Newey-West stan-dard errors was used to compare diagnostic testing rates overthe course of the study period. The 2-sample t test was usedfor comparison of absolute stool testing rates and positivityrates. A P-value of < .05 was considered to be statisticallysignificant. Hypothesis testing was 2 sided. Stata version 14(StataCorp, College Station, Texas) was used for data analysis.

RESULTS

Diarrhea Testing Advisor UtilizationDuring the time period after the advisor was implemented,it was activated 704 times and completed 238 times. Therewere 26 infectious diseases consultations and 13 gastroen-terology consultations generated during this time as a resultof the advisor. The majority of the infectious diseasesconsults (81%) were prompted by the patient beingimmunocompromised.

Stool Testing RatesFollowing activation of the diarrhea testing advisor, there was asignificant immediate decrease in test orders and a significant con-tinued decrease in the testing rate over time for both stool culturesand ova and parasites at the adult inpatient location (Figures 2A

and 3A, Table, P = .001). In the control units, while therewas not an immediate impact of the advisor activation, therewas a significant collateral decrease in the rate of stool cultureand ova and parasites testing in both the pediatric hospitalunits (P < .001) and the adult emergency department (P < .001;P = .009) in the period after activation of the advisor in theadult inpatient units despite the fact that the intervention wasnot activated at these sites (Figures 2B and C and 3B and C).Conversely, a significant increase in the rate of stool cultureand ova and parasites testing (Figures 2D and 3D) was ob-served in the outpatient setting during the period after activationof the advisor in the adult inpatient units (P = .02 and P = .001,respectively). In the assessment of the incidence of patientspresenting with diarrhea, C. difficile testing rates showed asmall but significant decrease postimplementation at both theadult location (P = .047), with 10.6 vs 9.6 tests per 100 patientdays, and the pediatric inpatient location (P = .049), with 3.8vs 3.5 tests per 100 patient days postimplementation(Supplementary Figure 1A and B, available online). Therewere no significant changes in C. difficile testing rates inthe emergency department or outpatient locations(Supplementary Figure 1C and D, available online).

Test Positivity Rate ImpactThere was a statistically significant decrease in the numberof positive tests per month (1.63 vs 0.67 tests per month

Figure 2 (A-D) Interrupted time series analysis of stool culture rates in relation to implementation of the diarrhea advisor algorithm,which occurred only at the adult inpatient location (panel A).

196 The American Journal of Medicine, Vol 131, No 2, February 2018

preintervention and postintervention, respectively; P = .0188)in the adult inpatient location associated with advisor im-plementation; however, there was no statistically significantdifference in test positivity rate (positive tests/total tests) priorto and after the intervention (2.1% vs 2.4%; P = .69). A similardecrease in the number of positive stool cultures per month

was also seen in the pediatric (6.25 vs 2.56; P = .0006), adultemergency department (.92 vs .11; P = .004), and outpa-tient (4.83 vs 2.44; P = .018) settings. There was no significantchange in the number of positive tests per month for ova andparasites examinations in relation to the advisor activation atthe adult inpatient location (0.19 vs 0.067; P = .38).

Figure 3 (A-D) Interrupted time series analysis of stool ova and parasite testing rates in relation to implementation of the diarrhea advisoralgorithm, which occurred only at the adult inpatient location (panel A).

Table Change in Test Ordering Over Time After Intervention by Site and Test

Location Test Immediate Change* Change in Rate†

Adult inpatient Culture −1.40‡ −0.0038‡Ova and parasites −1.47‡ −0.0042‡Clostridium difficile 0.37 −0.0046‡

Pediatric Culture −0.52 −0.0086‡Ova and parasites 0.096 −0.0071‡Clostridium difficile 0.81 −0.0036‡

Emergency department Culture 0.13 −0.0032‡Ova and parasites −0.15 −0.0019‡Clostridium difficile 0.19 0.00033

Outpatient Culture −0.070 0.00044‡Ova and parasites 0.0089 0.00065‡Clostridium difficile −0.041 0.00032

*All values are per 1000 patient days for the pediatric and adult inpatient locations and per 1000 patient visits for the emergency department andoutpatient locations.

†All values are per 1000 patient days per month for the pediatric and adult inpatient locations and per 1000 patient visits per month for the emer-gency department and outpatient locations.

‡Denotes statistically significant result with P < .05.

197Tewell et al Reducing Inappropriate Diagnostic Testing Using a Computerized Algorithm

Cost AnalysisAn estimated $21,931 was saved in laboratory costs annu-ally based on the decrease in stool culture and ova and parasitestesting at the adult inpatient location after implementation ofthe diarrhea testing advisor.

DISCUSSIONThe use of a computerized testing advisor to guide clini-cians on the appropriate testing for hospitalized patients withdiarrhea led to a significant and immediate reduction in therate of bacterial stool culture and stool ova and parasites testingin adult inpatients. This impact was sustained following im-plementation. It is difficult to know with certainty what partof the algorithm is responsible for the majority of the impact;however, we found that the “three-day rule”-based restric-tions limited 43% of attempted orders, with the remaining57% being filtered into testing by other aspects of the advisor.Interestingly, a positive collateral impact occurred in otherareas of the medical center where the advisor was not acti-vated. Namely, there was a significant decrease in testing ratesat the emergency department and pediatric inpatient loca-tions. This effect may have been related to the presence ofclinicians who work in both the intervention and control units,raising awareness of testing appropriateness that affected habitsin other locations. Specifically, in the adult emergency de-partment, a substantial proportion of patients are admitted andawaiting an available bed in an inpatient unit. These pa-tients are cared for by clinicians from these locations whowould be aware of the testing algorithm and have receivededucation from the tool on appropriate testing. In addition,emergency residents rotate on adult inpatient clinical ser-vices, where they would have been exposed to the advisor.The pediatric inpatient units have less cross-over of clini-cians, but adult emergency department clinicians, medicine-pediatric residents, and some surgical housestaff also rotatein the pediatric units, which may have led to more aware-ness of recommended testing strategies. Supporting thehypothesis that clinician crossover led to the collateral impactin nonintervention units is the fact that many clinicians whowould perform diarrhea testing in the outpatient setting (eg,primary care and internal medicine providers) either do notcare for patients once they have been admitted (due to a veryrobust hospitalist service in the adult inpatient units) or donot perform the actual test ordering in the adult inpatient units(ie, this task is performed by the housestaff on the team). Thesignificant increase in testing rates among outpatients com-pared with the other areas is also interesting. Multiple factorsmay have contributed to this effect, but one of the strongestmay have been the large number of outpatient clinicians whodo not order tests on adult inpatients, which suggest a di-minished opportunity for the collateral impact noted in theother areas. There was also a decrease in C. difficile testingin some areas despite this test not being restricted by theadvisor. This is likely due to increased provider awarenessof reducing unnecessary testing from our education efforts

during the rollout of our intervention and from continuing re-inforcement of awareness of overuse of diarrhea testing frominteraction with the advisor. However, the decrease was modestand it is possible that there was simply a reduction in thenumber of cases of diarrhea seen in these locations.

This analysis did note a small but statistically significantdecrease in the number of positive stool culture tests per monthfollowing implementation of the advisor, amounting to ~1fewer positive stool culture a month. It is unclear if this wasdue to the advisor, as a similar decrease in the number of pos-itive stool cultures was also seen at the nonintervention sites.If this was an effect of the intervention, the clinical impactis unclear but likely to be small for several reasons. First, thesickest patients (such as immunocompromised persons or thosewith fever) are recommended for infectious disease consul-tation in the algorithm, so it is unlikely that clinically significantpositive tests in this group would have been missed. For pa-tients without concerning symptoms, it is likely that even witha positive culture, treatment would not have been indicatedor would have been of minimal benefit.17-19

The financial impact of reduced stool testing, while modestat $21,931 a year, is likely underestimated, given that reducednursing workload, decreased need for collection supplies, anda reduction in other indirect costs anticipated as a result ofless testing were not included in the analysis. To put this valuein context, a CPOE-based intervention targeted at reducingredundant B-type natriuretic peptide orders was estimated tosave $92,000 yearly, and another intervention designed to limitrepetitive C. difficile testing saved an estimated $21,650annually.12,20 Our cost savings estimate included only the in-tervention site and may have been greater if the effects seenat the other sites were included. Finally, the impact in the re-duction of unnecessary antibiotic use and subsequent risk ofC. difficile infection and contribution to the development ofmulti-drug-resistant organisms is important to acknowl-edge, even if this is not able to be quantified.

This intervention and analysis does have some limita-tions. First, it encompasses patients and clinicians at a singleacademic medical center, where testing rates and the under-lying patient characteristics may differ from that of other acutecare facilities. Second, the advisor guidance relies on the en-tering of accurate information about the patient’s clinicalpresentation. Clinicians could enter false data in order to obtaina desired test. To assess this, investigators reviewed a subsetof cases to assess whether responses on the duration-of-symptoms question were accurate when compared with theclinical documentation. This question was completed by alladvisor respondents, making it a uniform assessment of datavalidity. In a review of 67 patients (12.6% of the total), 76%of responses reported a duration of symptoms that matchedthe clinical documentation. This relatively low rate is not par-ticularly surprising, as it has been noted in studies of physicianself-report of sepsis bundle compliance and documentationof clinic visits that inaccuracies are common.21,22 If this hadbeen closer to 100% accurate, the impact on testing may havebeen even more pronounced. Perhaps most concerning is thatsome providers would enter the advisor multiple times for

198 The American Journal of Medicine, Vol 131, No 2, February 2018

the same patient and provide different answers, likely to obtaintheir desired test and bypass the recommendations. This issueshould be considered in any future testing algorithms that aredesigned. Third, our test positivity rate analysis assumes thatthe test characteristics remained stable over time. However,it is possible that if the test performed differently over time(ie, a higher or lower sensitivity/specificity), this was respon-sible for the changes observed in the number of positive stoolcultures monthly, as opposed to the effect of the advisor.

Another limitation of this study is that we were unable tocollect data on noninfectious etiologies of diarrhea. Finally,this study does not address the ongoing transition from con-ventional culture and microscopy-based methods of pathogenidentification to newer molecular techniques. We believe thatthe concepts included in this study, namely, the restriction ofdiagnostic testing, provide a basis for the application of similaralgorithms, including multipathogen molecular testing. Wewould expect similar high rates of inappropriate testing withthese platforms if implemented without restriction; al-though, their role in the evaluation of hospitalized patientswith diarrhea is not precisely defined at this time.

In conclusion, a point-of-order testing advisor algorithmfor hospitalized adults with diarrhea was effective in reduc-ing bacterial stool cultures and ova and parasites testing,leading to substantial cost and time savings. This effect oc-curred in areas where the advisor was directly employed, aswell as in other clinical areas that did not have access to thetool. The method of test restriction used in this study is dif-ferent from those described previously, as it directly impactsthe provider’s ability to order the studies rather than relyingon rejection of already collected specimens that do not meettesting criteria. In an era of medicine where the reduction ofunnecessary treatment and testing is recognized as an im-portant goal, projects such as this are crucial. Future effortsto expand this work to other health care systems and for theevaluation of other disease states should be considered.

References1. Mead PS, Slutsker L, Dietz V, et al. Food-related illness and death in

the United States. Emerg Infect Dis. 1999;5(5):607-625.2. Garthright WE, Archer DL, Kvenberg JE. Estimates of incidence and

costs of intestinal infectious diseases in the United States. Public HealthRep. 1988;103(2):107-115.

3. Lucado J, Mohamoud S, Zhao L, Elixhauser A. Infectious Enteritis andFoodborne Illness in the United States, 2010. HCUP Statistical Brief#250. Agency for Healthcare Research and Quality, 2013.

4. Guerrant RL, Van Gilder T, Steiner TS, et al. Practice guidelines forthe management of infectious diarrhea. Clin Infect Dis. 2001;32:331-348.

5. Siegel DL, Edelstein PH, Nachamkin I. Inappropriate testing for diar-rheal diseases in the hospital. JAMA. 1990;263(7):979-982.

6. Hines J, Nachamkin I. Effective use of the clinical microbiology lab-oratory for diagnosing diarrheal diseases. Clin Infect Dis. 1996;23:1292-1301.

7. Morris AJ, Wilson ML, Reller LB. Application of rejection criteria forstool ovum and parasite examinations. J Clin Microbiol.1992;30(12):3213-3216.

8. Seyler L, Lalvani A, Collins L, Goddard L, Bowler IC. Safety and costsavings of an improved three-day rule for stool culture in hospitalisedchildren and adults. J Hosp Infect. 2007;67(2):121-126.

9. Chitkara YK, McCasland KA, Kenefic L. Development and implemen-tation of cost-effective guidelines in the laboratory investigation of diarrheain a community hospital. Arch Intern Med. 1996;156(13):1445-1448.

10. May TA, Clancy M, Critchfield J, et al. Reducing unnecessary inpa-tient laboratory testing in a teaching hospital. Am J Clin Pathol.2006;126(2):200-206.

11. Vegting IL, van Beneden M, Kramer MH, Thijs A, Kostense PJ,Nanayakkara PW. How to save costs by reducing unnecessary testing:lean thinking in clinical practice. Eur J Intern Med. 2012;23(1):70-75.

12. Levick D, Stern G, Meyerhoefer CD, Levick A, Pucklavage D. Reduc-ing unnecessary testing in a CPOE system through implementation ofa targeted CDS intervention. BMC Med Inform Decis Mak. 2013;13:43.

13. Krasowski MD, Chudzik D, Dolezal A, et al. Promoting improved uti-lization of laboratory testing through changes in an electronic medicalrecord: experience at an academic medical center. BMC Med InformDecis Mak. 2015;15(1):11.

14. Bates DW, Kuperman GJ, Rittenberg E, et al. A randomized trial of acomputer-based intervention to reduce utilization of redundant labora-tory tests. Am J Med. 1999;106(2):144-150.

15. Otto CC, Shuptar SL, Milord P, et al. Reducing unecessary and dupli-cate ordering for ovum and parasite examinations and clostridium difficilePCR in immunocompromised patients by using an alert at the time ofrequest in the order management system. J Clin Microbiol.2015;53(8):2745-2748.

16. Schubart JR, Fowler CE, Donowitz GR, Connors AF. Algorithm-based decision rules to safely reduce laboratory test ordering. Stud HealthTechnol Inform. 2001;84:523-527.

17. Christopher PRH, David KV, Sankarapandian V. Antibiotic therapy forShigella dysentery. Cochrane Database Syst Rev. 2010;(8):CD006784.

18. Onwuezobe IA, Oshun PO, Odigwe CC. Antimicrobials for treatingsymptomatic non-typhoidal Salmonella infection. Cochrane DatabaseSyst Rev. 2012;(11):CD001167.

19. Proulx F, Turgeon JP, Delage G, Lafleur L, Chicoine L. Randomized,controlled trial of antibiotic therapy for Escherichia coli O157:H7 en-teritis. J Pediatr. 1992;121(2):299-303.

20. Luo R, Spradley S, Banael N. Alerting physicians during electronic orderentry effectively reduces unecessary repeat PCR testing for Clos-tridium difficile. J Clin Microbiol. 2013;51(11):3872-3874.

21. Dresselhaus TR, Luck J, Peadbody JW. The ethical problem of falsepositives: a prospective evaluation of physician reporting in the medicalrecord. J Med Ethics. 2002;28(5):291-294.

22. Brunkhorst FM, Engel C, Ragaller M, et al. Practice and perception-anationwide survey of therapy habits in sepsis. Crit Care Med.2008;36(10):2719-2725.

SUPPLEMENTARY DATASupplementary data accompanying this article can be foundin the online version at doi:10.1016/j.amjmed.2017.10.006.

199Tewell et al Reducing Inappropriate Diagnostic Testing Using a Computerized Algorithm

APPENDIX

Supplementary Figure 1 Interrupted time series analysis of C. difficile testing rates in relation to implementation of the diarrhea advisoralgorithm, which restricted only stool culture and O & P testing at the adult inpatient location.

199.e1 The American Journal of Medicine, Vol 131, No 2, February 2018