increased risk of digoxin toxicity following hospitalization

pharmacoepidemiology and drug safety 2009; 18: 28–35ww.interscience.wiley.com) DOI: 10.1002/pds.1680
Published online 28 November 2008 in Wiley InterScience (w
ORIGINAL REPORT

Increased risk of digoxin toxicity following hospitalizationy

Kevin Haynes PharmD, MSCE1,2,3*, Sean Hennessy PharmD, PhD1,2,3, A. Russell Localio PhD1,2,3,Abigail Cohen PhD1,3, Charles E. Leonard PharmD1,2,3, Stephen E. Kimmel MD, MSCE1,3,Harold I. Feldman MD, MSCE1,3, Brian L. Strom MD, MPH1,2,3 and Joshua P. Metlay MD, PhD1,2,3,4

1Program for the Reduction in Medication Errors, Department of Biostatistics and Epidemiology and School of Medicine, University ofPennsylvania, Philadelphia, PA, USA2Center for Education and Research on Therapeutics, Department of Biostatistics and Epidemiology and School of Medicine, University ofPennsylvania, Philadelphia, PA, USA3Center for Clinical Epidemiology and Biostatistics, Department of Biostatistics and Epidemiology and School of Medicine, University ofPennsylvania, Philadelphia, PA, USA4Center for Health Equity Research and Promotion, Veterans Affairs Medical Center, Philadelphia, PA, USA

SUMMARY

Purpose Adverse drug events (ADEs) are an important cause of preventable hospitalizations among elderly individuals taking high-riskmedications. The objective of the study was to identify health care system factors that affect the risk of digoxin toxicity for older adults ondigoxin.Methods We conducted a prospective cohort study of older adults within the Pennsylvania Pharmaceutical Assistance Contract for theElderly (PACE) program, which provides comprehensive drug benefits for older adults with low income. Subjects were interviewed at the timeof enrollment regarding the management and coordination of their health care as well as medication comprehension. Hospitalizations wereidentified by linking patient identifiers to a state-wide registry. Trained abstractors reviewed discharge summaries of possible digoxin relatedADEs. Unadjusted and adjusted incidence rate ratios (IRR) were calculated based on person–months of exposure using Poisson regressionmodels, with variances adjusted for within subject repeated measures.Results We enrolled a total of 2030 adults on digoxin from May 2002 to June 2003. A total of 34 hospitalizations due to digoxin toxicityoccurred, equivalent to 1.12 hospitalizations per 1000 person–months of exposure. Adjusting for hospitalization in the past 2 months, age,total number physicians prescribing any medications in past 3 months, total number of pharmacies filling medications in past 3 months, andnumber of unique prescriptions filled in the past month had a 4.25-fold increased risk of subsequently experiencing digoxin toxicity (IRR95%CI 1.95, 9.27).Conclusions The risk of digoxin toxicity-related hospitalization, while low, is higher in the post-hospital period. Copyright # 2008 JohnWiley & Sons, Ltd.

key words—adverse effects; drug toxicity; physician–patient relationships

Received 11 April 2008; Revised 26 September 2008; Accepted 14 October 2008

INTRODUCTION

In the outpatient setting, patients and their caregiversplay a critical role in ensuring the safe use of drugs.Medication errors are the result of practitioner, patient,and system factors, some of which are potentially

*Correspondence to: K. Haynes, University of Pennsylvania School ofMedicine, 711 Blockley Hall, 423 Guardian Drive, Philadelphia, PA 19104-6021, USA. E-mail: [email protected] conflict of interest was declared.

Copyright # 2008 John Wiley & Sons, Ltd.

predictable and preventable.1 In studying medicationsafety, it is likely that strategies that reduce the risk ofadverse events might vary across drug classes. Drugswith a narrow therapeutic index are particularly pro-blematic because the dose that results in toxicity is onlyslightly higher than that required for efficacy.2 Amongnarrow therapeutic index drugs, digoxin is frequentlycited as a leading drug responsible for hospitalization-related adverse drug events (ADEs)3 and emergencyroom visits.4 In particular, the transition of care fromthe inpatient to the outpatient setting presents a

DIGOXIN TOXICITY 29

challenge due to dosage changes that require physician,pharmacist, and patient communication.The specific aim of this study was to identify

characteristics of health care coordination betweenphysicians, pharmacists, and patients that influence therisk of digoxin toxicity in older adults, with particularattention to the transition from inpatient to outpatientcare.

METHODS

Weperformed a prospective cohort study ofADEs result-ing in hospitalization among older adults taking digoxin.The University of Pennsylvania institutional reviewboard approved the study protocol, and all participantsprovided written informed consent and authorizationfor access to medical records.

Study site

The study sample was drawn from the 200 200 mem-bers in the Pennsylvania Pharmaceutical AssistanceContract for the Elderly (PACE) and the 30 100 mem-bers of the PACE Needs Enhancement Tier (PACE-NET) Program. These state-run programs, referred tohereafter as PACE, offer comprehensive prescriptioncoverage to Pennsylvanians over the age of 65 whoare not on Medicaid and meet income eligibilityrequirements.

Subject enrollment

We have previously reported on the details regardingthe process of subject enrollment.5 Briefly, electronicclaims for digoxin transmitted to the PACE programallowed us to identify potentially eligible subjectswithin a few days of filling a claim for digoxin. Weattempted to sample an equal number of subjects whowere new users (no prior claims for digoxin within theprevious 12months) or chronic users (one or more priorclaims for digoxin within the previous 12 months). Thesampling algorithm was initially designed to recruit1000 subjects into each of the two categories over a 13-month sampling period (1 May 2002–30 June 2003).Potential subjects meeting inclusion criteria were

contacted by the PACE program. Subjects consentingto participate agreed to complete a baseline interviewand two annual follow-up interviews as well as provideaccess to pharmacy claims and future hospitalizationdata. Subjects were excluded if they resided in a skillednursing home or assisted living facility. Subjects couldonly be sampled once into the study.


Data collection

Telephone interviews. Interviews with participants occur-red at baseline, 12 and 24 months. A proxy interview wassought with a next of kin or caregiver for participantswho were unable to complete the interview (typicallydue to cognitive impairment). Standardized question-naires that were administered included (1) TheMedicalOutcomes Study 8 Item Short Form (SF-8),6 (2) the6 Item Cognitive Impairment Scale (6 CIT), which isan abbreviated version of the Blessed InformationMemory Concentration Scale,7,8 and (3) The GeriatricDepression Scale—Short Form (GDS-SF).9,10 Addi-tional questions regardingmedication-taking behaviorsand receipt of medication instructions were derivedthrough focused interviews conducted with olderadults visiting community centers.5

Pharmacy claims data. For each subject, exposure todigoxin was confirmed based on prescription claimsdata obtained from PACE which records for eachpharmaceutical claim the unique subject identifier,unique physician identifier, unique pharmacy identi-fier, date of prescription dispensing, National DrugCode (NDC), and name of drug, dose, and totalquantity dispensed.

Hospitalization data. Data obtained from the Penn-sylvania Healthcare Cost Containment Council (PHC4),an independent state agency which collects inpatientdischarge records from all Pennsylvania hospitals, wereused to identify hospitalizations of enrolled subjects.These records provided the investigators with access tohospital name, admission and discharge dates, andICD-9 CM discharge diagnoses. A contiguous hospi-talization, defined as an admission date being the sameas the discharge date for a prior inpatient stay, wasconsidered a single hospitalization.We developed lists of both broad diagnostic codes

potentially related to digoxin toxicity (available uponrequest) and specific (E942.1, 972.1) ICD-9 CM codesexpected to indicate ADE due to digoxin. A copy of thedischarge summary from the hospital for any hospi-talization with a discharge diagnosis (whether principalor not) matching any code on the list was requested.Wereceived discharge summaries from over 97% of ourrequests.Discharge summaries received were reviewed by

trained members of an abstraction team consisting ofnurses, nurses-in-training, a medical student, anda physician’s assistant. Abstraction forms were

Pharmacoepidemiology and Drug Safety, 2009; 18: 28–35DOI: 10.1002/pds

30 k. haynes ET AL.

constructed to evaluate, on four levels, each hospital-ization’s likelihood of being due to a digoxin toxicityevent: (a) evidence of clinical signs and symptoms ofdigoxin toxicity on admission, (b) evidence of elevateddigoxin level (>2mcg/L) upon admission, (c) aphysician’s attribution of the hospitalization as dueto digoxin toxicity, and (d) evidence of holding dosesof digoxin or administration of a reversal agent,specifically digoxin immune fab, cholestyramine, oractivated charcoal. Any discharge summary with evi-dence for level ‘a’ and ‘b’ along with at least one of theother two levels of evidence was considered a probabledigoxin toxicity event and retained for further review.All hospitalizations deemed probable were then

reviewed by two members of a team of expert revie-wers blinded to the interview data. Upon reviewing thedischarge summaries, the experts independentlydetermined whether or not each hospitalization wasdefinitely due to an ADE. In instances when the tworeviewers arrived at opposite conclusions, a thirdmember of the team independently reviewed thesummary and their decision became final, via provisionof a majority opinion.

Data analysis

Defining exposure time. Based on the date of enroll-ment time at risk was divided for purposes of analysisinto 24–30-day period (hereafter, referred to asmonths). We examined the stream of PACE recordedprescriptions for digoxin from the month precedingenrollment through the subsequent 24-month periods.We assumed that each prescription recorded was a1-month supply based on the PACE program’sconvention of allowing prescriptions only for 30-dayperiods. Similarly, we assumed that any month inwhich two prescriptions were filled represented2 months of exposure to the study drug of interest.Thus, observed gaps of 1 month were ‘filled’ if theprevious month contained two prescriptions. Anymonth that did not contain a prescription or could notbe ‘filled’ in this manner was assumed to be a true gaprather than a month of exposure.

Defining risk factors. Potential digoxin-related toxi-city risk factors included self-reported factors from thebaseline and follow-up telephone interview, patterns ofprescription data from the PACE database, and patternsof hospitalization from the PHC4 database. For the6CIT dementia instrument, we categorized all scores ata cut point of 8 or greater as indicating moderate tosevere cognitive impairment.7,8 We counted the


number of non-toxicity-related hospitalizations foreach month and the prior month as a month-level time-varying measure of the risk of prior hospitalizations onpotential toxicity. We excluded hospitalizations thatoccurred after a toxicity-related hospitalization in anymonth. For the first study month, we also includedhospitalizations that occurred in the prior month. Thus,we assumed that the effect of a prior hospitalizationpersisted for up to 2 months.The PACE database provided the number of unique

prescription drugs filled per month. The total numberof unique prescription drugs, defined by NDC coding,has been validated as a measure of comorbidity andmedication complexity.11

We also estimated the total number of differentprescribing physicians for 3-month rolling windows ofexposure to assess the frequency of patient–physiciancontacts. The same 3-month rolling window applied tothe number of unique pharmacies that filled theprescriptions (as determined by the pharmacy identi-fiers in the PACE program).

Defining hospitalization outcomes. All dates of hos-pitalizations confirmed as resulting from digoxin-related toxicity based on expert review (see above)were compared to the dates of exposure as defined bythe prescribing history. We retained outcome events ifthey occurred during a month with a prescription fordigoxin or within 30 days of such a prescription month.All events that occurred more than 1 month after a drugexposure month, whether at the end of a sequence ofexposure months or during a ‘gap’ of known drugprescriptions, were deleted from the analysis becausewe had no way to measure exposure time across thecohort during these gap periods.In a qualitative analysis, we reviewed the discharge

summaries and discharge plans for all hospitalizationsthat occurred within 2 months prior to the toxicityhospitalization events confirmed in this study, theperiod for which we had hospital discharge summaries.Of the 34 digoxin toxicity hospitalizations, eightpatients had a hospitalization in the prior 60 days of theprecipitant digoxin toxicity. We requested thesehospitalization records to qualitatively determinedosage changes whichmight have led to the subsequentdigoxin toxicity hospitalization.

Statistical analysis

Descriptive analyses determined the distribution ofexposures of interest. We calculated the overall rate ofobserved hospitalizations due to digoxin toxicityby dividing the number of digoxin toxicity-related


DIGOXIN TOXICITY 31

hospitalizations by the number of digoxin exposedperson–months.We used Poisson regression (with robustvariance estimates) for all analyses at the patient-monthlevel. The outcome was the number of hospitalizationsresulting from digoxin-related toxicity. This methodpermitted the analysis of time-invariant as well as thetime-varying exposures specifically, the number ofunique prescriptions in the previous 3 months, numberof providers seen in the previous 3 months, and thenumber of physicians seen in the last 3 months.We began with bivariable analyses to identify

opportunities for collapsing adjacent categories ofordered variable to avoid sparse data problems.Variables were included in a multivariable model ifthey were statistically significant in bivariable analysis(p< 0.1) or were pre-specified of clinical interestbased on patient risk factors identified in prior studiesspecifically, gender, age, race, marital status, edu-cation, living arrangement, lives alone, children live athome, urban setting, nurse visits home, proxy providedthe interview, prior digoxin use, visual impairment,hearing impairment, cognitive function, depression,medication instruction provider, number of physiciansseen in the last 3 months, number of providers seen inthe last 3 months, and the total number of uniqueprescriptions in the previous 3 months. We constructedfour models (1) a model with the variable for ahospitalization in the prior 2 months; (2) plus baselinecharacteristics; (3) plus markers of healthcare utiliz-ation; and (4) plus variables for where patient receivedinstructions.We performed multivariable Poisson regression with

robust variance estimates (using generalized estimat-ing equations (GEE)). This model allowed us toinclude repeated events for any subject. The impact onthe effective sample size of using robust variances(confidence intervals) was small. All quantitativeanalyses were conducted with SAS (version 9.1).

RESULTS

Of 3462 eligible patients who had filled digoxinprescriptions prior to the study and were contacted bythe PACE program, 2048 (59%) ultimately consentedto participate in the study and completed the baselineinterview. Of these, 1947 (95%) also completed a12-month follow-up interview and 1681 (82%)completed the final interview at 24 months. The majorreason for failure to complete a follow-up interviewwas death, 158 (8%) during months 1–12 and 201(10%) during months 13–24.When these participants were followed after enroll-

ment, 18 (0.9%) had no claims for digoxin during the


24-month study period. The remaining 2030 subjectscontributed a total of 30 296 months of exposure timeon digoxin. Based on the claims data, 533 (26% of thecohort) were considered new users of digoxin while1497 (74% of the cohort) were considered continuingusers of digoxin. While the study targeted balancedrecruitment between new users and chronic users, thispopulation contained a low proportion of new digoxinusers.

Characteristics of the cohort

The majority of study subjects were female (76%),white, non-Hispanic (95%) and greater than 75 years ofage (71%) which is consistent with the demographiccharacteristics of the PACE program. The majority ofsubjects were living alone (54%), without other familyin the home. Only 11% of the participants required aproxy to complete the baseline telephone interview;however an additional 18% of participants had CITscores in the cognitively impaired range. Thus, a totalof 26% of participants were considered cognitivelyimpaired (Table 1).Over the follow-up period, during any 3-month

observation period, the majority (94%) of participatingsubjects continued to fill digoxin prescriptions writtenby a single physician. However, considering all typesof medication prescriptions filled, the majority ofsubjects (66%) had prescriptions written by greaterthan one physician over 3-month observation period,and 15% of participants filled prescriptions written byfour or more physicians during a 3-month observationperiod.

Digoxin-related toxicity events

We identified a total of 34 hospitalizations that wererated as definite digoxin toxicity events based on expertreview. The overall risk of a hospitalization due to adigoxin toxicity event was 1.12 events per 1000person–months of exposure to digoxin (95%CI 0.78,1.57).

Risk factors

We report the results of the factors that were pre-specified as being important and found no additionalfactors that were confounders. Hospitalization in thepast 2 months was a significant risk factor forsubsequent digoxin-related toxicity resulting in hos-pitalization, with an incident rate ratio of 4.87 (95%CI2.34, 10.16) (Table 2). Hospitalization remained asignificant risk factor in a model that sequentiallyadjusted for baseline patient characteristics (i.e.,


Table 1. Characteristics of study subjects

Time invariant

Characteristic Category Number (%)

Patient characteristicsGender (N¼ 2029) Female 1544 (76.1%)Age categories (N¼ 2030) �70 years 187 (9.2%)

71–75 years 401 (19.8%)76–80 years 555 (27.3%)81–85 years 492 (24.2%)�86 years 395 (19.5%)

Race (N¼ 1983) White non-hispanic 1887 (95.2%)White hispanic 19 (1.0%)Non-white 77 (3.9%)

Marital Status (N¼ 2028) Married 384 (18.9%)Single 311 (15.3%)Widowed 1333 (65.7%)

Education (N¼ 2009) Grade school 357 (17.8%)Less than high school 476 (23.7%)High school 994 (49.5%)College 182 (9.1%)

Living arrangement (N¼ 2029) House 1361 (67.1%)Apartment 493 (24.3%)Assisted living 18 (0.9%)Retirement community 49 (2.4%)Mobile home/trailer 96 (4.7%)Religious 12 (0.6%)

Lives alone (N¼ 2028) No 939 (46.3%)Children live at home (N¼ 2027) No 1943 (95.9%)Urban (N¼ 2030) No 378 (18.6%)Nurse visits home (N¼ 2023) None 1837 (90.8%)

Nurse/Nurse’s aide 135 (6.7%)Companion 51 (2.5%)

Proxy interview (N¼ 2030) No 1815 (89.4%)Prior digoxin use (N¼ 2030) Chronic digoxin user 1497 (73.7%)

Clinical characteristicsVisual impairment (N¼ 2029) No 1403 (69.1%)Hearing impairment (N¼ 2026) No 1737 (85.7%)Cognitive impairment (CIT) (1782) Cit score 1–7 1465 (82.2%)Depression (N¼ 1771) No 772 (43.6%)

Medication instructionsPerson who gives instructions (N¼ 1839) Only MD 151 (7.6%)

Only pharmacist 602 (30.5%)Both MD and pharmacist 298 (15.1%)

Time varying

Characteristic Category Person–months (%)

Patient characteristicsNumber of physicians in last 3 months 1 physican in 3 months 10418 (34.4%)

2 or 3 physicans in 3 15478 (51.1%)>3 physicans in 3 month 4400 (14.5%)

Number of providers in last 3 months 1 provider in 3 months 27580 (91.0%)>1 providers in 3 months 2716 (9.0%)

Total number of unique prescriptions in the previous 3 months:1–3 unique Rxs 3995 (13.2%)4–8 unique Rxs 18313 (60.4%)>8 unique Rxs 7988 (26.4%)

32 k. haynes ET AL.

patient age and gender) (Model 2). Markers ofhealthcare utilization including, total number phys-icians prescribing any medications in past 3 months,total number of pharmacies filling medications in past3 months, and number of unique prescriptions filled inthe past 1 month adjusted the risk to 4.25 (95%CI 1.95,


9.27) (Model 3). Self-reported measures of whether thepatient received medication instructions (Model 4) didnot alter the estimate found in model three. Although,obtaining instructions only from the doctor was asignificant risk factor for toxicity (IRR¼ 3.56 95%CI1.41, 9.02).


Table 2. Adjusted incidence rate ratios for digoxin toxicity

Factors Model 1 95%CI Model 2 95%CI Model 3 95%CI Model 4 95%CI

Hospitalization in the previous 2 months 4.87 (2.34, 10.16) 4.95 (2.37, 10.31) 4.25 (1.95, 9.27) 4.27 (1.97, 9.26)Age: 71–75 � 0.85 (0.16, 4.63) 0.86 (0.16, 4.62) 0.84 (0.15, 4.59)Age: 76–80 � 1.53 (0.34, 6.87) 1.58 (0.35, 7.06) 1.56 (0.35, 7.00)Age: 81–85 � 1.42 (0.31, 6.50) 1.49 (0.33, 6.73) 1.56 (0.35, 7.01)Age: >85 � 2.41 (0.54, 10.78) 2.66 (0.62, 11.50) 2.48 (0.57, 10.85)Gender (Male) � 0.48 (0.17, 1.35) 0.49 (0.18, 1.35) 0.50 (0.18, 1.39)Total number of physicians seen in the previous 3 months: 2 or 3 � � 0.83 (0.39, 1.73) 0.83 (0.38, 1.80)Total number of physicians seen in the previous 3 months: >3 � � 0.86 (0.32, 2.27) 0.84 (0.32, 2.20)Total number of providers in the previous 3 months: >3 � � 0.95 (0.36, 2.49) 0.92 (0.35, 2.38)Total number of providers in the previous 3 months: >1 � � 1.72 (0.69, 4.31) 1.71 (0.67, 4.34)Total number of unique prescriptions in the previous 3 months: 4–8 � � 1.25 (0.38, 4.16) 1.27 (0.38, 4.22)Total number of unique prescriptions in the previous 3 months: >8 � � 2.16 (0.67, 6.99) 2.28 (0.71, 7.38)Instructions provided: only MD � � � 3.56 (1.41, 9.02)Instructions provided: only pharmacist � � � 1.05 (0.46, 2.41)Instructions provided: both MD and pharmacist � � � 0.44 (0.10, 1.96)Instructions provided: other person � � � 1.06 (0.23, 4.87)

DIGOXIN TOXICITY 33

Table 3 shows the results of the discharge summaryreview of eight patients who had one or morehospitalizations in the previous 60 days of a digoxintoxicity hospitalization. Patient E was requested tohave digoxin levels checked every week following aprevious discharge, but we cannot determine if thisinstruction was followed. Patients A, B, F, and H hadnotations in previous discharge summaries to reducethe digoxin dose. The PACE claims do not reveal thedirections to the patient printed on the prescription vialto indicate if these changes occurred. This raises thehypothesis that errors in communication betweenhospital discharge, outpatient physicians, pharmacies,and patients might lead to an increase in digoxintoxicity.

DISCUSSION

Patients who have been hospitalized during the last2 months are at significantly increased risk of experi-encing additional hospitalizations due to digoxin

Table 3. Digoxin patients with prior hospitalization within 60 days of digoxin

Patient Narrative

A Subject had an acute renal failure hospitalization 14 days beforedigoxin to 0.125mg 5 days per week.

B Subject had a cellulitis hospital visit 53 days before the event inC Subject had a pneumonia hospital visit 37 days before the eventD Subject had two CHF hospitalizations 14 days and 27 days beforE Subject had a CHF hospitalization 16 days before event and told

7–10 days. Subject also had a CHF hospitalization 41 days beforF Subject had a hypokalemia hospitalization 36 days before event

records indicate subject only ever filled 0.125mg prescriptionsG Subject had a stroke hospitalization 47 days before event with aH Subject had a CHF hospitalization 54 days before event with a d

0.125mg every 3 days.

CHF, congestive heart failure.


toxicity, even after adjusting for baseline character-istics (gender, age, race, marital status, education,living arrangement, lives alone, children at home,urban, nurse visit, proxy interview, prior digoxin use),measures of comorbidity (visual impairment, hearingimpairment, cognitive impairment, depression), com-plexity of health care (number of physicians seen in thelast 3 months, number of providers in the last 3 months,number of unique prescriptions in the previous3 months), and patient self-report of having receivedinstructions on medication usage.Previous studies of risk factors for digoxin toxicity

emphasize patient’s clinical12 and demographic fac-tors,13 along with co-administration of potentialinteracting medications, such as amiodarone, quini-dine, or verapamil.13,14 Our study did not adjust for anyco-administration of potential interacting medications.Other studies have suggested that poorly coordinatedcare can be a risk factor for medication errors.15 Forthis reason, the period after hospitalization might be aparticularly high risk period for medication errors,

toxicity

the event with a digoxin level of 2.1. Subject was instructed to decrease

structed to decrease digoxin to 0.125mg daily.and the digoxin level was 0.3.e the event.to check digoxin level every Thursday and return to primary care ine the event.and instructed to take digoxin 0.25mg every other day. Prescription

digoxin level of 2.3 and digoxin was held.igoxin level of 2.01. Subject instructed to change digoxin dose to


KEY POINTS

� Narrow therapeutic drugs continue to result inADEs particularly among the elderly. Medicationreconciliation between the inpatient and out-patient setting may disrupt therapy resulting inADEs.

� Using data from the PACE, a prospective cohortstudy of ADEs resulting in hospitalization amongolder adults taking digoxin was conducted.

� Patients who have been hospitalized are atsignificantly increased risk of experiencinghospitalization due to digoxin toxicity for up to2 months after the date of discharge.

34 k. haynes ET AL.

given the frequent changes in medication regimenoccurring during hospitalization, which might bepoorly communicated to outpatient providers, andthe general frailty of patients after hospitalization.16

Our study is the first to link this high risk period to aspecific type of medication error, suggesting thepotential for targeted interventions.The prospective cohort design which allowed us to

assess patient reported behaviors and knowledge priorto the occurrence of any adverse events is a keystrength of this study. However, digoxin toxicityoccurred in the background of many other health-related events, including deaths, emphasizing that thepopulation was quite frail and at risk for manycompeting adverse clinical events. Our estimate of theincidence of digoxin toxicity is less than 1%,considerably less than studies conducted over the pasttwo decades which suggest that the annual incidence isbetween 4 and 5% per year per user.17,18 However, ourestimate is consistent with recent evidence that digoxintoxicity has declined.19

The transition of care between the inpatient settingand the outpatient setting appears to be particularlyproblematic for narrow therapeutic drugs. Our group20

and others16 have suggested that warfarin therapy inthe frail elderly are especially problematic in thetransition between inpatient and outpatient care.Additionally, our study did not show a great effectregarding the instruction patients received affectingtoxicity. Kripalani et al. 21 recently reviewed the keyissues for hospitalists in the effective transitions of careat the hospital discharge; the key point is communi-cation. Thus, healthcare systems need better interven-tions specifically designed to improve communicationand reduce ADEs of high risk medications, includingdigoxin, especially in the setting of transition of care.One limitation of this study is the potential for

selection bias. Subject refusals occurred at both theinitial recruitment stage by PACE and the enrollmentstage by our staff. Our limited comparisons ofcharacteristics indicate that non-participants wereolder, more likely to be living alone, and in the lowerincome categories, which based on our analysis shouldnot affect the results found regarding transition of care.Additionally, this survey was conducted among apopulation of community-dwelling older adults whomet income eligibility for a state-wide pharmaceuticalassistance program and filled prescriptions for a selectgroup of high risk medications. A large proportion ofthe study subjects reside in rural settings and most arewidowed, living alone. On one hand, this populationrepresents a vulnerable population for whom digoxintoxicity rates might be elevated, but found to be low.


On the other hand, participation in a state-widepharmaceutical benefit program with a well-estab-lished real time drug utilization review program wouldbe expected to provide some protection against homebased medication errors. Thus, these risk factors mightbe even more prominent in other community-dwellingpopulations on digoxin.

CONCLUSIONS

In summary, the risk of adverse events on high-riskmedications, specifically digoxin, is influenced by thequality and complexity of inpatient–outpatient medicalcare hand off as evidenced by an increase in digoxintoxicity following recent hospitalization. Improvedinformation systems may be able to coordinateinpatient and outpatient care to reduce ADEs causedby changes to drug regimens following hospitalization.Further research will need to provide evidence forimprovements in clinical care following implementa-tion of information systems.

ACKNOWLEDGEMENTS

The authors acknowledge Jen Tjia, M.D. for expert review ofcases.This study was suported by Grant P01-HS11530 from the

Agency for Healthcare Research and Quality. Dr Metlay issupported by an Advanced Research Career DevelopmentAward from the Health Services Research and DevelopmentService of the Department of Veterans Affairs.

REFERENCES

1. Hennessy S. Potentially remediable features of the medication-useenvironment in the United States. Am J Health Syst Pharm 2000;57: 543–548.

2. Bochner F, Tonkin A. The clinician and therapeutic drug monitoring inthe 1990s. Med J Aust 1993; 158: 422–426.


DIGOXIN TOXICITY 35

3. Hussain Z, Swindle J, Hauptman PJ. Digoxin use and digoxin toxicityin the post-DIG trial era. J Card Fail 2006; 12: 343–346.

4. Budnitz DS, Shehab N, Kegler SR, et al. Medication use leading toemergency department visits for adverse drug events in older adults.Ann Intern Med 2007; 147: 755–765.

5. Metlay JP, Cohen A, Polsky D, et al. Medication safety in older adults:home-based practice patterns. J Am Geriatr Soc 2005; 53: 976–982.

6. Ware JE, KosinskiM, Dewey JE, Gandek B.How to Score and InterpretSingle-Item Health Measures: A Manual for Users of the SF-8 HealthSurvey. Lincoln(RI): Quality Metric Inc: 2001.

7. Blessed G, Tomlinson BE, Roth M. The association between quanti-tative measures of dementia and of senile change in the cerebral greymatter of elderly subjects. Br J Psychiatry 1968; 114: 797–811.

8. Brooke P, Bullock R. Validation of a 6 item cognitive impairment testwith a view to primary care usage. Int J Geriatr Psychiatry 1999; 14:936–940.

9. Burke WJ, Roccaforte WH, Wengel SP, et al. The reliability andvalidity of the Geriatric Depression Rating Scale administered bytelephone. J Am Geriatr Soc 1995; 43: 674–679.

10. Sharp LK, Lipsky MS. Screening for depression across the lifespan: areview of measures for use in primary care settings. Am Fam Physician2002; 66: 1001–1008.

11. Schneeweiss S, Seeger JD, Maclure M, et al. Performance of comor-bidity scores to control for confounding in epidemiologic studies usingclaims data. Am J Epidemiol 2001; 154: 854–864.

12. Rea TD, Siscovick DS, Psaty BM, et al. Digoxin therapy and the risk ofprimary cardiac arrest in patients with congestive heart failure: effect ofmild-moderate renal impairment. J Clin Epidemiol 2003; 56: 646–650.


13. Pahor M, Guralnik JM, Gambassi G, et al. The impact of age on riskof adverse drug reactions to digoxin. For The Gruppo Italiano diFarmacovigilanza nell’ Anziano. J Clin Epidemiol 1993; 46: 1305–1314.

14. Onder G, Pedone C, Landi F, et al. Adverse drug reactions as cause ofhospital admissions: results from the Italian Group of Pharmacoepi-demiology in the Elderly (GIFA). J Am Geriatr Soc 2002; 50: 1962–1968.

15. Forster AJ, Murff HJ, Peterson JF, et al. Adverse drug eventsoccurring following hospital discharge. J Gen Intern Med 2005; 20:317–323.

16. Hanlon JT, Pieper CF, Hajjar ER, et al. Incidence and predictors of alland preventable adverse drug reactions in frail elderly persons afterhospital stay. J Gerontol A Biol Sci Med Sci 2006; 61: 511–515.

17. Mahdyoon H, Battilana G, Rosman H, et al. The evolving pattern ofdigoxin intoxication: observations at a large urban hospital from 1980to 1988. Am Heart J 1990; 120: 1189–1194.

18. Williamson KM, Thrasher KA, Fulton KB, et al. Digoxin toxicity: anevaluation in current clinical practice. Arch Intern Med 1998; 158:2444–2449.

19. Haynes K, Heitjan D, Kanetsky P, et al. Declining public health burdenof digoxin toxicity from 1991 to 2004. Clin Pharmacol Ther 2007;DOI: 10.1038/sj.clpt.6100458

20. Hanlon JT, Schmader KE, Koronkowski MJ, et al. Adverse drug eventsin high risk older outpatients. J Am Geriatr Soc 1997; 45: 945–948.

21. Kripalani S, Jackson AT, Schnipper JL, et al. Promoting effectivetransitions of care at hospital discharge: a review of key issues forhospitalists. J Hosp Med 2007; 2: 314–323.


increased risk of digoxin toxicity following hospitalization

Documents