Using Information Technology to Detect Ambulatory Adverse Events Related to Antidiabetic Drug Therapy Judy Wu, PharmD Duke University Hospital Co-Investigators:

Using Information Technology to Using Information Technology to Detect Ambulatory Adverse Events Detect Ambulatory Adverse Events

Related to Antidiabetic Drug TherapyRelated to Antidiabetic Drug Therapy

Judy Wu, PharmDDuke University Hospital

Co-Investigators: Heidi Cozart, RPh; Julie Whitehurst, PharmD;

Philip Rodgers, PharmD; Jennifer Mando, PharmD

Adverse Drug Events (ADEs)Adverse Drug Events (ADEs) Research primarily in the inpatient setting

3 – 6 ADEs per 100 admissions1-3

27% – 50% of ADEs are preventable1-3

Estimated cost: $ 3.5 billion (2006 dollars)4

ADE detection methods Chart review, patient surveys, computer event

monitoring, text scanning, voluntary reporting Multiple methods = more ADEs

Not well understood in other care settings

1. Bates DW et al. JAMA 1995;274(1):29-34.2. Classen DC et al. JAMA. 1997;277(4):301-6.3. Jha AK et al. J Am Med Inform Assoc 1998;5(3):305-14.4. Aspden P, IOM (U.S.). Preventing medication errors. Washington, DC: National Academies Press, 2007.

““Most data on medication error incidence rates Most data on medication error incidence rates come from the inpatient setting, butcome from the inpatient setting, but

- Institute of Medicine

the magnitude of the problem is likely to the magnitude of the problem is likely to be greater outside the hospital.”be greater outside the hospital.”

Aspden P, IOM (U.S.). Preventing medication errors. Washington, DC: National Academies Press, 2007.

MedicationMedicationErrorsErrors

Adverse Adverse Drug Drug

EventsEvents(ADEs)(ADEs)

DefinitionsDefinitions

Gandhi TK et al. International Journal for Quality in Health Care. 2000; 12:69–76.

Patient injury resulting from medical intervention

related to a drug

Bates DW et al. JAMA. 1995; 274:29–34.

Any error in any stage of the medication use process

(ordering, transcribing, dispensing, administering, or monitoring)

Bates DW et al. J Gen Intern Med 1995;10: 199-205.

Scope of the ProblemScope of the Problem Limited research in ambulatory care1,2

Baseline ADE incidence rate Identify strategies to decrease ADEs

Barriers to ambulatory care ADE research Inefficient Lack of accessible data Large patient population Most common medications resulting in ED

visits Insulin and warfarin3,4

1. Thomsen LA et al. Ann Pharmacother 2007;41(9):1411-26.2. Field T et al. Med Care 2005; 43: 1171-1176.3. Hafner J et al. Annals of Emergency Medicine. 2002; 30: 258-267.4. Budnitz DS et al. JAMA 2006;296(15):1858-66.

Research ObjectivesResearch Objectives Quantify hypoglycemia ambulatory ADEs Quantify hypoglycemia ambulatory ADEs

resulting in emergency department visits or resulting in emergency department visits or hospitalizationhospitalization

Characterize the population of subjects Characterize the population of subjects experiencing ADEs experiencing ADEs

Evaluate the utility of three different electronic Evaluate the utility of three different electronic adverse event detection methodsadverse event detection methods

Design a catalog of trigger words to detect Design a catalog of trigger words to detect possible ADEs through free-text searchingpossible ADEs through free-text searching

Study DesignStudy Design Retrospective, electronic chart review

Approved by Duke University Institutional Review Board Study site: Duke University Hospital Study period: January 1, 2007 to September 30, 2007

Inclusion criteria Subjects >18 years old experiencing possible antidiabetic

drug-induced hypoglycemia resulting in an emergency department visit or hospitalization

Exclusion criteria Subjects experiencing hypoglycemia not associated with

medication use Lack of objective evidence

Hypoglycemic ADE Blood glucose < 50 mg/dL while on antidiabetic

therapy ADE scoring

ADE = causality score ≥ 5 and a severity score ≥ 3 Causality - Naranjo algorithm1

Severity - Duke 7 point ADE severity score2

ADE group Comprehensive list of ADEs detected from any of the

3 tools

MeasurementsMeasurements

1. Naranjo CA et al. Clin Pharmacol Ther. 1981;30:239-245.2. Kilbridge PM et al. J Am Med Inform Assoc 2006; 13: 372-377.

Detection MethodsDetection Methods

Computerized ADE Surveillance (ADE-S)Computerized ADE Surveillance (ADE-S)

Diagnosis (ICD-9) codesDiagnosis (ICD-9) codes

Free-text searchingFree-text searching

Detection Methods:Detection Methods:Computerized ADE Surveillance

Logic based rules Screens demographic and laboratory data,

medications, and other clinical results Alerts pharmacist about possible ADEs Review and scoring process Acute care setting vs emergency department Hypoglycemia rule

Dextrose 50% when BG < 50 mg/dL

Detection Methods:Detection Methods:Diagnosis (ICD-9) codes

Administrative dataAdministrative data International Classification of Diseases, 9International Classification of Diseases, 9thth edition edition Codes for diagnoses and proceduresCodes for diagnoses and procedures

E900 codes specific to adverse events due to E900 codes specific to adverse events due to drugs drugs E932.3 Adverse effect insulin/antidiabeticsE932.3 Adverse effect insulin/antidiabetics

Detection Methods:Detection Methods:Free-text searching

Electronic medical records Emergency department visits

Refinement of searching tool Identification of trigger words Elimination of negative and ambiguous terms

Final search strategy Include {DM or diabetes} AND {hypoglycemia or

hypoglycemic or low blood glucose or low BG or low glucose} AND exclude {(-)DM}

Results: Hypoglycemia Alerts DetectedResults: Hypoglycemia Alerts Detected

n = 138

n = 72

n = 212

# of unique alerts = 364

8 666626

12112

168

32

Results: Hypoglycemia ADEs DetectedResults: Hypoglycemia ADEs Detected

ComputerComputerSurveillanceSurveillance

ADEs = 154 (42%)ADEs = 154 (42%)

91 91 55 55

5757

6666

ADE PopulationADE PopulationCharacteristics

Number of events 154

Age in years (mean ± SD) 59 ± 16.6

GenderMale (%) 45

Number of comorbidities (mean ± SD) 6.8 ± 3.7

Number of medications (mean ± SD)

AntidiabeticTotal

1.7 ± 0.69.8 ± 5.0

Hospitalization (%) 49

ADE Distribution By RaceADE Distribution By Race

n = 154

ADE Distribution By AgeADE Distribution By Agen = 154

Number of Events

ADEs With Insulin InvolvementADEs With Insulin Involvementn = 154

Mean blood glucose value at time of hypoglycemic event:

32 mg/dL

Mean blood glucose value at time of hypoglycemic event:

32 mg/dL

Insulin +Insulin +SulfonylureaSulfonylurea

4.5%4.5%

Positive Predictive Value (PPV) of Positive Predictive Value (PPV) of ADE Detection ToolsADE Detection Tools

ComputerComputerSurveillanceSurveillance

43%43%40%40%

79%79% Overestimation

100%100%

91 ADEs 91 ADEs 55 ADEs 55 ADEs

57 ADEs57 ADEs

Sensitivity of ADE Detection ToolsSensitivity of ADE Detection Tools

0%

20%

40%

60%

80%

100%

ADE-S ICD-9 Free-text

36% 37%59%

n = 154

LimitationsLimitations Retrospective, chart reviewRetrospective, chart review Not generalizable to other ambulatory ADEsNot generalizable to other ambulatory ADEs Subjectivity in scoring ADEsSubjectivity in scoring ADEs Underestimation of hypoglycemic incidence rateUnderestimation of hypoglycemic incidence rate

Specific population Specific population Exclusion of symptomatic hypoglycemia with BG > 50Exclusion of symptomatic hypoglycemia with BG > 50 Undetected hypoglycemic ADEs?Undetected hypoglycemic ADEs?

Detection tool limitationsDetection tool limitations ADE-S, ICD-9, free-text searchADE-S, ICD-9, free-text search

ConclusionConclusion 17 hypoglycemia ADEs per month were 17 hypoglycemia ADEs per month were

detecteddetected 49% require hospitalization49% require hospitalization 71% of ADEs involved insulin use71% of ADEs involved insulin use

African American and older age present more African American and older age present more frequently with hypoglycemia ADEs frequently with hypoglycemia ADEs

Highest yield & sensitivity Highest yield & sensitivity free text search tool free text search tool Greatest PPV Greatest PPV ICD-9 coding ICD-9 coding Minimal overlap among toolsMinimal overlap among tools Combining methods increases ADE yieldCombining methods increases ADE yield

““The primary focus of research on medication errors in the The primary focus of research on medication errors in the next decade should be prevention strategies, recognizing next decade should be prevention strategies, recognizing that to plan an error prevention study, it is essential to be that to plan an error prevention study, it is essential to be

able to measure the baseline rate of errors.”able to measure the baseline rate of errors.”- - Institute of MedicineInstitute of Medicine

Future research:Future research: Expand into other populations and other Expand into other populations and other

ambulatory ADE areasambulatory ADE areas Tool refinementTool refinement Use of detection methods in outpatient clinicsUse of detection methods in outpatient clinics Prevention strategiesPrevention strategies

AcknowledgementsAcknowledgements

Heidi CozartHeidi Cozart Julie WhitehurstJulie Whitehurst DHTSDHTS Department of PharmacyDepartment of Pharmacy Residency Research CommitteeResidency Research Committee

QuestionsQuestions

Race DistributionRace Distribution

http://quickfacts.census.gov/qfd/states/37/37063.html

North Carolina Durham County

Duke

Race DistributionRace Distribution

http://quickfacts.census.gov/qfd/states/37/37063.html

Predicted Actual