Author + information
- Received October 1, 2015
- Revision received January 22, 2016
- Accepted January 25, 2016
- Published online May 1, 2016.
- Paul J. Hauptman, MDa,∗ (, )
- Steven W. Blume, MSb,
- Eldrin F. Lewis, MD, MPHc and
- Suzanne Ward, PharmD, MBAd
- aDepartment of Medicine, Saint Louis University School of Medicine, Saint Louis, Missouri
- bEvidera, Bethesda, Maryland
- cDepartment of Medicine, Brigham & Women’s Hospital and Harvard Medical School, Boston, Massachusetts
- dBTG International Inc., Brentwood, Tennessee
- ↵∗Reprint requests and correspondence:
Dr. Paul J. Hauptman, Division of Cardiology, Saint Louis University Hospital, 3635 Vista Avenue, Saint Louis, Missouri 63110.
Objectives This study was developed to determine contemporary management of digoxin toxicity and clinical outcomes.
Background Although the use of digoxin in heart failure management has declined, toxicity remains a prevalent complication.
Methods The Premier Perspective Comparative Hospital Database (Premier Inc., Charlotte, North Carolina) was used to retrospectively identify patients diagnosed with digoxin toxicity and/or who received digoxin immune fab (DIF) over a 5-year period (2007 to 2011). DIF was evaluated using treatment date, number of vials administered, and total cost. Clinical outcomes included length of stay (total hospitalization; days after DIF), cost of hospitalization, and in-hospital mortality. Exploratory multivariate analyses were conducted to determine predictors of DIF and effect on length of stay, adjusting for patient characteristics and selection bias.
Results Digoxin toxicity diagnosis without DIF treatment accounted for 19,543 cases; 5,004 patients received DIF of whom 3086 had a diagnosis of toxicity. Most patients were >65 years old (88%). The predictors of DIF use were urgent/emergent admission, hyperkalemia, arrhythmia associated with digoxin toxicity, acute renal failure, or suicidal intent (odds ratios 1.7, 2.4, 3.6, 2.1, and 3.7, respectively; p < 0.0001 for all). The majority (78%) of DIF was administered on days 1 and 2 of the hospitalization; 10% received treatment after day 7. Digoxin was used after DIF administration in 14% of cases. Among patients who received DIF within 2 days of admission, there was no difference for in-hospital mortality or length of stay compared with patients not receiving DIF.
Conclusions Digoxin toxicity diagnoses are clustered in the elderly. One-fifth of cases receive treatment with DIF, most within 2 days of admission. Opportunities exist for improved diagnosis and post-DIF management. Prospective data may be required to assess the impact of DIF on length of stay.
Digoxin remains a therapeutic intervention in both atrial fibrillation and heart failure (HF), as described in current clinical practice guidelines (1,2). Although digoxin use has declined in HF and meta-analyses have raised questions about efficacy in atrial fibrillation, toxicity remains clinically relevant largely as a consequence of the drug’s narrow therapeutic window (3,4). The risk factors for digoxin toxicity have been amply described (5) and include advanced patient age (6), renal failure (7), metabolic disorders, and drug interactions (8). However, little is known about hospitalizations related to toxicity including contemporary management and resource utilization. Since 1986, antidotal therapy has been available; however, its use has only been indirectly quantified (9–11). Therefore, the primary objectives of the study were to describe patient characteristics, hospital utilization, and outcomes of patients with digoxin toxicity, and to compare patients treated with and without digoxin immune fab (DIF).
A retrospective cohort design was followed using the Premier Perspective Comparative Hospital Database (Premier Inc., Charlotte, North Carolina) for the 5-year period from 2007 to 2011 (excluding quarter 4). Premier collects data voluntarily submitted from more than 450 hospitals including detailed information about day-of-service resource use. Hospital region, patient, and payer distributions compare well to national statistics, although they tend to include larger hospitals (68% of discharges recorded in Premier are provided by hospitals with more than 300 beds, compared to 37% nationally; ). The database contains the data elements available in most hospital or payer datasets including patient demographics, marital status, gender, race, diagnosis and procedure codes, length of stay, total cost of inpatient care, and a date-stamped log of all billed items, including medications, laboratory, and diagnostic and therapeutic services at the individual patient level.
Member hospitals benchmark their clinical and financial performance against their peers. The underlying data undergo quality checks, and cost information is reconciled with the hospitals’ financial statements. The Premier data are subsequently de-identified and rendered Health Insurance Portability and Accountability Act (HIPAA) compliant to ensure patient confidentiality (12).
This is a retrospective study designed to characterize patients with digoxin toxicity and to compare patients with and without treatment with DIF. Inclusion criteria were:
1. Record of administration of DIF (brand names DigiFab [BTG International Inc., West Conshohocken, Pennsylvania] or DigiBind [formerly manufactured by GlaxoSmithKline, no longer commercially available]; procedure codes Current Procedural Terminology [CPT] J1162 or Q2006); and/or
2. International Statistical Classification of Diseases-Ninth Revision-Clinical Modification (ICD-9-CM) diagnosis of digoxin toxicity (listed as admitting, discharge, or secondary diagnosis) with code 972.1 (poisoning by cardiotonic glycosides and drugs of similar action—digitalis glycosides, digoxin, strophanthins) or code E942.1 (causing adverse effects in therapeutic use, cardiotonic glycosides, and drugs of similar action—digitalis glycosides, digoxin, strophanthins).
If an individual patient had more than 1 inpatient hospitalization fulfilling either criterion, the earliest hospitalization was selected. Exclusion criteria included patient visits for use of DIF for reasons other than digoxin toxicity, including any diagnosis indicating possible severe pre-eclampsia or a related diagnosis (ICD-9-CM code 642.x).
Patients were assigned to 1 of 2 cohorts, depending on their exposure to DIF. The day of exposure to DIF and the presence of a diagnosis of digoxin toxicity were also noted to define possible subgroup analyses or covariates.
The following variables were obtained: patient demographics (age, gender, race); evidence of acute ingestion (e.g., any diagnosis of suicidal intent); principal (discharge) diagnosis (1 per patient; e.g., digoxin toxicity, arrhythmias, HF, renal failure, hyperkalemia); admission diagnosis (reliably populated in 2009 and after); secondary diagnoses; admission type (emergency, urgent, elective); hospital characteristics (bed size, teaching status, rural vs. suburban/urban location and region); admitting physician specialty; and month/year of discharge. In addition, the APR-DRG group (All Patients Refined Disease-Related Group) (3M Health Information Systems, Salt Lake City, Utah) was recorded. APR-DRGs are similar to Medicare DRGs, classifying patients to predict intensity of resource use based on diagnosis and key surgical procedure codes but enhanced to define separate disease severity and mortality subclasses (1 through 4, with 4 being extreme).
DIF was evaluated using the date of administration, the number of vials administered, and cost. Data related to medications used for treating arrhythmias (potentially as a consequence of digoxin toxicity) were obtained, including anticholinergic agents (atropine sulfate) and lidocaine, as well as other potential therapies (magnesium, activated charcoal, and the placement of a temporary transvenous pacemaker). Use of digoxin after administration of DIF was also evaluated by day of administration.
Clinical outcomes of interest included length of stay (total and in intensive care) measured for both total hospitalization and for days after DIF; cost of hospitalization; and in-hospital mortality. Readmission and/or subsequent emergency department visits are limited to the same hospital over a 60-day period.
Descriptive tabular analyses were conducted presenting each of the preceding study measures for characteristics, treatments, and outcomes by cohort. Patients treated with DIF were further classified by whether a digoxin toxicity diagnosis was recorded. A simple nationwide projection of digoxin toxicity cases and DIF treatments using the Premier database, provided sampling weights was also calculated.
Continuous study measures were reported (mean, median, standard deviation, and range; percentiles where data appeared to be skewed). Categorical variables were reported using frequency distributions. Given the large sample sizes, most differences between cohorts on the basic 2-way analyses were statistically significant and p values were uninformative.
Given the constraints of this nonrandomized observational study design, exploratory multivariate analyses were conducted to determine predictors of treatment with DIF, as well as the treatment effect of DIF on length of stay, while attempting to adjust for patient characteristics and selection bias.
Fixed-effects linear regression models were used to estimate DIF treatment effects for patients within a given APR-DRG, severity level, and utilization of intensive care unit, thereby controlling for unobserved variables related to their condition (13,14); this was especially attractive in the current study because detailed chart data were not available for incorporation into the model. As a sensitivity analysis, a linear regression without fixed effects and with principal diagnosis, severity indices, and intensive care unit utilization as covariates was also estimated.
SAS version 9.2 (SAS Institute Inc., Cary, North Carolina) was the statistical software utilized.
Patient and provider characteristics
A total of 24,547 hospitalizations were identified (of a total of 28.5 million hospitalizations over nearly 5 years) for which a code was recorded for digoxin toxicity (n = 19,543), treatment with DIF (n = 1,918), or both (n = 3,086) (Figure 1).
Characteristics of non-DIF and DIF-treated cohorts are shown in Table 1. Patients in both cohorts were generally older (31% ≥85 and 88% ≥65 years of age) and admitted as emergencies (78%). Race and sex were similar across cohorts. Patients in the DIF cohort were more likely than non-DIF patients to be rated as “extreme” for disease severity (37% vs. 20%), and risk of mortality (33% vs. 17%; on a scale of mild, moderate, major, extreme).
Overall, arrhythmia diagnoses accounted for 19% of principal discharge diagnoses (4.8% with atrial fibrillation or flutter). Congestive heart failure and acute renal failure each accounted for 11% of patients, and electrolyte disorders another 4%. Diastolic heart failure ICD-9-CM code 428.3x accounted for 15% of the HF diagnoses. In general, diagnoses did not differ by cohort, except for a higher portion of arrhythmia diagnoses among DIF patients than non-DIF patients (27% vs. 17%). Admission diagnoses followed a similar pattern (data not shown). Secondary diagnoses included arrhythmias in more than 90%, HF in 53%, digoxin toxicity in 90%, and chronic kidney disease in 61% of patients.
Three quarters of the admitting physicians of record in the database were internists, hospitalists, or family practice physicians (Table 1). Admitting physicians for the DIF cohort were somewhat more likely to be specialists than for the non-DIF cohort, mostly due to the higher proportion of cardiologists (13% vs. 9%). Attending physician type followed a similar pattern (data not shown). Hospital size and geographic distributions differed little between cohorts.
DIF was administered in 5,004 (20%) of patients. After multivariable adjustment, the largest predictors of DIF use were emergency admission (odds ratio [OR]: 1.7; p < 0.0001) and coding for hyperkalemia, arrhythmias associated with digoxin toxicity, other arrhythmias, acute renal failure, or suicidal intent (ORs 2.4, 3.6, 2.0, 2.1, and 3.7, respectively; p < 0.0001 for all) (Table 2). Physician generalists were less likely to treat with DIF than cardiovascular specialists (OR: 0.6 to 0.8; p < 0.05). Urban and teaching hospitals were slightly less likely than community nonurban hospitals to use DIF (OR: 0.88, p = 0.01; and OR: 0.87, p = 0.04; respectively). Age, sex, and region were not important factors.
The majority (78%) of DIF was administered on days 1 and 2 of the hospitalization; 10% received DIF after day 7. Only 6% of patients treated with DIF were coded as intentional overdose (4% in patients not given DIF). The median number of vials used per toxicity episode was 3 with 90% of administrations 5 or fewer vials. Patients receiving DIF were also more likely to receive other therapies (activated charcoal, atropine, lidocaine, or temporary pacemaker) compared with patients not receiving DIF (Table 3).
Digoxin was administered during the hospitalization to a large portion of patients including 31% of patients receiving DIF and 14% after receipt of DIF.
The mean length of stay was longer for patients administered DIF (8.9 days) than for patients who did not receive DIF (6.6 days); in the former, the mean cost of DIF was $1,540. Overall costs ($22,328 vs. $12,032) and in-hospital mortality (14% vs. 6%) were greater for DIF-treated patients, but 60-day readmission rates (limited to the same hospital) were equivalent (22% and 23%, respectively) (Table 3) and 180-day readmission rates were somewhat lower for DIF-treated patients (33% vs. 37%, p < 0.0001). A sensitivity analysis that restricted the cohort to those patients who received DIF within 2 days did not show a difference for in-hospital mortality or length of stay compared with the non-DIF patients. Among patients treated with DIF, those with a digoxin toxicity diagnosis had a much shorter length of stay than those who did not (7.2 days vs. 11.6 days). Regression models adjusting for disease and severity variables, and limiting analysis to patients with a coded diagnosis of digoxin toxicity, estimated that the effect of DIF on length of stay was a mean reduction of 0.3 to 0.7 days. Of those patients readmitted, 7% were diagnosed with toxicity or treated with DIF, regardless of whether they were treated with DIF on their index admission.
Nationwide projection of in-patient cases and treatment
A nationwide estimate of the number of digoxin toxicity cases and treatments by year is shown in Table 4, based on our sample and Premier-provided sampling weights. Estimated cases decreased from 41,000 in 2007 to 35,000 in 2011, while the proportion treated with DIF was essentially unchanged at 21% to 22%.
Despite reports of decreasing use of digoxin as a treatment for heart failure (3) and uncertainty about survival benefit in both HF and atrial fibrillation (15–17), toxicity remains an important clinical entity. Contemporaneous data about the incidence of toxicity cases are conflicting (3,18,19). In one report, the incidence of digoxin toxicity did not decline over time, and the percentage of patients receiving DIF for digoxin toxicity appeared to have increased between 1996 and 2003 (3). However, Haynes et al. reported that the rate of digoxin toxicity hospital admissions had steadily declined between 1990 and 2004 to 8 per 100,000 population, based on National Hospital Discharge Survey data (18). Consistent with these observations, the national projection data in our study show a continued decrease in the annual rate of digoxin toxicity admissions between 2007 and 2011. Recent data on the incidence of toxicity in relation to the number of users of digoxin has been more difficult for researchers to obtain. Referencing 5 sources from 1990 to 2000, Haynes et al. (18) reported that the annual incidence of digoxin toxicity is between 4% and 5% per year per user of digoxin. The challenge in this and other reports has been to find commensurate data for both numerator and denominator. For example, although Haynes et al. (18) could estimate the number of hospitalizations for toxicity, the measure of digoxin use was based only on the number of office visits during which time digoxin was mentioned and the number of digoxin scripts per 100,000 Medicaid beneficiaries.
In the current report, we estimated 35,000 toxicity or DIF inpatient discharges in 2011. This number is 2.4% of the total of 1 million inpatient discharges with HF as the primary diagnosis and the 480,000 discharges with atrial fibrillation as primary (based on statistics from the American Heart Association ). We previously reported, using ADHERE (Acute Decompensated Heart Failure National Registry) data, that 23.5% of those hospitalized for acute decompensated HF in 2004 were using digoxin upon admission (3). More recently, See et al. (19) found that digoxin toxicity accounted for 1.0% of emergency department visits for adverse drug events among patients aged ≥40 years and 5.9% of hospitalizations for all adverse drug events among patients ≥85 years of age.
It is possible that under-reporting of digoxin toxicity occurs, potentially as a result of lack of recognition; another contributing factor may be overreliance on laboratory reporting, especially when the therapeutic range is inappropriately wide (21). In the DIG (Digitalis Investigation Group) study, toxicity was suspected in 11.9% of cases during follow-up, but it led to hospitalization in only 2% (15). However, it is also noteworthy that at least in HF therapeutics, there is increasing recognition that the digoxin’s clinical benefit may be limited to serum concentrations between 0.5 ng/ml and 0.9 ng/ml (22); as a consequence, lower dosing may reduce risk of toxicity, especially in patients who are at risk such as the elderly and patients with renal failure. The recent studies suggesting an increase in mortality with digoxin use in atrial fibrillation with and without HF indicate a need to reassess the role of both digoxin and digoxin levels across multiple indications.
Using a large hospital database, we demonstrated that the cohort with toxicity is clinically diverse in terms of coded diagnoses and predominantly elderly, with nearly one-third of patients 85 years of age or older. DIF tends to be prescribed by specialists as opposed to generalists and used in sicker patients as gauged by a number of variables. The use of other supportive interventions, although low, was also greater in patients receiving DIF. Previous studies of treatment with DIF have been limited to relatively small series (9–11); low heart rate, acute overdose, and hyperkalemia as well as DIF availability were reported as predictors of use.
The fact that 10% of patients who received DIF did so after day 7 suggests that some toxicity cases failed usual care, were not recognized earlier in the hospital course, or developed during the hospitalization. Indeed, the mean length of stay, regardless of whether DIF was administered, was long in contrast to a prior report suggesting a mean of only 3.3 days (23).
Of some concern, digoxin is not infrequently prescribed after DIF administration. It is unlikely that its use is based on treating worsening of HF after antibody administration, although this remains a theoretical explanation. It is possible that quality initiatives will need to be designed and implemented to limit or prevent the reintroduction of digoxin during the index hospitalization with or for digoxin toxicity. Further, we noted that digoxin levels are often measured after DIF, which may not be appropriate unless the assay measures free digoxin alone.
It is uncertain whether early use of DIF leads to better outcomes. As an observational study, the ability to clarify factors explaining the difference in length of stay or cost is limited. As expected, disease severity was an important contributing factor to the length of stay and associated cost. As such, a range of multivariable analyses were conducted to attempt to correct for bias. The estimated DIF effect on length of stay was a mean reduction of 0.3 days to 0.7 days if it is assumed that only patients with a coded diagnosis of digoxin toxicity should be compared; otherwise, there was no effect.
Approximately 38% of patients receiving DIF did not have a concomitant digoxin toxicity code; these patients had more comorbidities than others receiving DIF, even after multivariable adjustment. One possible explanation is that the sicker the patient, the greater the importance placed on coding other principal and secondary diagnoses, perhaps for reimbursement and documentation purposes.
The additional cost of DIF needs to be weighed against costs associated with nontreatment. In a statistical simulation, DiDomenico et al. demonstrated that DIF is associated with cost savings over usual supportive care when the digoxin level is >3.5 ng/ml or creatinine clearance is <22 ml/min; similarly, length of stay is reduced when digoxin levels are >2.3 ng/ml for patients with non–life-threatening toxicity (24). Finally, the use of DIF appears to depend at least in part on the specialty of the attending physician (25) and whether or not the ingestion was acute as part of an intentional overdose; in the current database, the number of such cases was low. Other variables such as admission status of the patient and clinical indicators of severe toxicity (such as arrhythmias and hyperkalemia) were also predictors.
Our data are limited by lack of long-term outcomes including rehospitalization (in hospitals other than the initial hospital) and 30-day and longer-term mortality. Therefore, we cannot assess prevalence of repeat toxicity requiring hospitalization. Not all digoxin toxicity is clinically recognized as such. Further, digoxin levels, dosing, outpatient medication use, and renal function are not available in the Premier database, which would otherwise provide additional insight into the selection and timing of DIF. We cannot account for patients who may have experienced digoxin toxicity but were not coded as such; further, subclinical toxicity would not have been identified. We also cannot determine digoxin exposure and the risk of toxicity on that basis. We were, however, able to make an estimate of the number of digoxin toxicity cases nationwide.
We have shown that digoxin toxicity was often associated with complex hospitalizations and a minority of patients (20%) was treated with DIF. Digoxin was also prescribed in 14% of patients following administration of DIF, suggesting that opportunities exist for improved management. Use of electronic medical record data may facilitate analyses that take advantage of laboratory data and as such may allow us to further examine the management of digoxin toxicity and, in particular, the relationship between digoxin levels and specific treatments. Further research is also needed to understand best practices, appropriate patient selection, and timing for administration of DIF, with prospective data collection to assess whether early identification and treatment can lead to shorter lengths of stay and improved outcomes.
COMPETENCY IN MEDICAL KNOWLEDGE: Digoxin toxicity remains a prevalent complication. The use of DIF is often associated with complex presentations. A significant proportion of cases of digoxin toxicity are not treated with DIF. Late administration of the antidote and reinitiation of digoxin following its use represent potential targets for quality improvement.
TRANSLATIONAL OUTLOOK: DIF was developed more than 40 years ago for the treatment of digoxin toxicity and remains the gold-standard approach. Additional study is needed to assess best practices, including evaluation of the impact of early identification and treatment.
BTG purchased the Premier Perspective hospital database and commissioned Evidera for the data analysis. Mr. Blume was an employee of Evidera at the time of this study, which provides research services to pharmaceutical and device manufacturers; in his salaried position, he worked with a variety of organizations and was precluded from receiving payment or honoraria directly from these organizations. Ms. Ward is Scientific Director for BTG International Inc. Drs. Hauptman and Lewis were unpaid consultants and have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- All Patients Refined Disease-Related Group
- digoxin immune fab
- coded diagnosis of digoxin toxicity
- International Classification of Diseases-Ninth Revision-Clinical Modification
- heart failure
- odds ratio
- Received October 1, 2015.
- Revision received January 22, 2016.
- Accepted January 25, 2016.
- American College of Cardiology Foundation
- Yancy C.W.,
- Jessup M.,
- Bozkurt B.,
- et al.
- January C.T.,
- Wann L.S.,
- Alpert J.S.,
- et al.
- Ziff O.J.,
- Lane D.A.,
- Samra M.,
- et al.
- Chan K.E.,
- Lazarus J.M.,
- Hakim R.M.
- ↵Premier healthcare informatics. Premier capabilities and hospital characteristics. Charlotte, NC: Premier, Inc., 2009.
- Allison P.D.
- Woolridge J.
- Vamos M.,
- Erath J.W.,
- Hohnloser S.H.
- Haynes K.,
- Heitjan D.,
- Kanetsky P.,
- Hennessy S.
- See I.,
- Shehab N.,
- Kegler S.R.,
- Laskar S.R.,
- Budnitz D.S.
- Mozaffarian D.,
- Benjamin E.J.,
- Go A.S.,
- et al.
- Hauptman P.J.,
- McCann P.,
- Ramirez Romero J.M.,
- Mayo M.
- Adams K.F. Jr..,
- Gheorghiade M.,
- Uretsky B.F.,
- Patterson J.H.,
- Schwartz T.A.,
- Young J.B.
- DiDomenico R.J.,
- Walton S.M.,
- Sanoski C.A.,
- Bauman J.L.
- Kirrane B.M.,
- Olmedo R.E.,
- Nelson L.S.,
- Mercurio-Zappala M.,
- Howland M.A.,
- Hoffman R.S.