Cañas F, Tanasijevic MJ, Ma'luf N, Bates DW. Evaluating the Appropriateness of Digoxin Level Monitoring. Arch Intern Med. 1999;159(4):363-368. doi:10.1001/archinte.159.4.363
Copyright 1999 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.1999
Digoxin level determinations can be useful clinically in patients receiving digoxin therapy but are sometimes misused.
Explicit appropriateness criteria were adapted from previously published criteria and revised using local expert opinion. They were then used to evaluate the appropriateness of random samples of inpatient and outpatient serum digoxin levels. Overall agreement between reviewers regarding appropriateness was good (κ=0.65). Patients in the study included 162 inpatients in whom 224 digoxin levels were measured and 117 outpatients in whom 130 digoxin levels were measured during a 6-month period. The main outcome measure was the proportion of digoxin levels with an appropriate indication.
Among inpatient levels, only 16% (95% confidence intervals [CI], 11%-20%) were appropriate. Of the 189 digoxin levels considered inappropriate, only 26 (14%) had a result of 2.3 nmol/L or more (≥1.8 ng/mL). None of these levels resulted in an important change in therapy, and no patient had a toxic reaction to the therapy. Among inappropriate levels, daily routine monitoring accounted for 78%. Of the 130 outpatient levels, 52% (95% CI, 44%-61%) were appropriate. Of 62 inappropriate levels, only 4 (6%) had a result of 2.3 nmol/L or more (≥1.8 ng/mL). One result led to a change in therapy, but none of the patients were believed to experience a toxic reaction. Among the inappropriate levels, 87% of patients underwent early routine monitoring before a steady state was achieved.
A high proportion of digoxin levels were inappropriate, particularly among inpatients. In both groups, the primary reason tests were judged inappropriate was early routine monitoring. Few inappropriate tests resulted in important data. Interventions to improve the use of digoxin levels could potentially save substantial resources without missing important clinical results.
MANY FACTORS contribute to the rising costs of health care. These rising costs have led to attempts to rethink the ways services are provided and identify services that could be omitted, while maintaining high-quality care. While diagnostic tests are relatively inexpensive individually, they are costly in the aggregate and contribute significantly to the rising expenses of the health care system.1- 3 At our institution, digoxin level monitoring accounts for 21% of all therapeutic drug concentration measurements.
Digoxin level determinations can be clinically useful in patients receiving digoxin therapy for a variety of reasons, including evaluating compliance and verifying toxicity of the therapy.4- 6 While many studies have evaluated the use of digoxin therapeutic drug–monitoring services and sampling for drug monitoring of digoxin, few have evaluated the proportion of digoxin level determinations with appropriate indications.7- 10
Since we wanted to measure the appropriateness of serum digoxin levels in our institution, with the aim of developing interventions to improve it, we undertook a study with the following goals: (1) to develop explicit criteria that define appropriate indications for measuring digoxin levels in both inpatients and outpatients; (2) to use these criteria to evaluate the proportion of digoxin determinations with an inappropriate indication; (3) to describe the indications for ordering determinations of drug levels; and (4) to determine how often levels with an inappropriate indication generated clinically important results.
The study was conducted at Brigham and Women's Hospital in Boston, Mass, a tertiary care university teaching hospital with 726 beds and approximately 35,000 admissions per year. Resident physicians are the primary orderers of tests. The pharmacy does not provide routine pharmacokinetic consultations.
The indications for digoxin levels for inpatients and outpatients were evaluated in an extensive search of the literature. English-language studies involving the appropriateness of digoxin level monitoring were identified by reviewing the following Medical Subject Heading (MeSH) key words in MEDLINE from 1966 through 1996: digoxin, clinical laboratory, therapeutic drug monitoring, appropriateness criteria, and consensus conference.4- 9,11- 14 In addition, the bibliographies of original and review articles were searched by hand and relevant references cross-checked with those identified through the computer search. Major texts in cardiology and clinical pharmacology were also searched, for chapters describing digoxin level monitoring.15,16 Articles providing explicit appropriateness guidelines were used with a focus on criteria pertinent to the treatment of both hospitalized and nonhospitalized patients.6,14 The appropriateness criteria were derived based on the information obtained from the literature search and were then reviewed and revised by an expert panel consisting of 2 cardiologists, an internist (D.W.B.), and a clinical pathologist (M.J.T.). The final criteria were derived based on input from the expert review. The aim was to establish simple operational appropriateness criteria that could be used as boundary rules.17
The digoxin levels were measured with an affinity column-mediated immunoassay using an ACA V Analyzer (Dade Behring, Wilmington, Del).18 Therapeutic ranges were 1.2 through 2.2 nmol/L (0.9-1.7 ng/mL).
Random samples of serum digoxin levels were drawn from drug levels obtained from adult inpatients and out-patients from October 1995 to March 1996 using a random-number generator. The sample sizes were 5% and 12% random samples for inpatient and outpatient levels, respectively. These sampling fractions were chosen to include more than 200 inpatient levels and more than 100 outpatient levels, to allow reasonably narrow confidence intervals (CIs) around the percentage appropriate, while providing a manageable number of charts for review. The unit of evaluation was the individual level determination rather than the patient, to allow assessment of the percentage of repeated tests and the clinical value of those tests.
Charts and computerized records were reviewed by one of us (F.C.) to obtain the following information: age, sex, and race of the patient, patient status, digoxin-level result, hospital service, daily drug dose with dosing interval, preceding digoxin level determination, and time elapsed since the preceding level was measured, indication for digoxin level determination, use of selected medications relevant for potential drug interactions (amiodarone, quinidine, spironolactone, nifedipine, and verapamil), use of a diuretic, creatinine level (to assess renal function), ejection fraction, and potassium level.
These data were used to categorize the digoxin level determination as "appropriate" or "inappropriate" in terms of indication according to the appropriateness criteria. To evaluate the reliability of chart review and interrater agreement for appropriateness, a random sample (10% inpatient, 15% outpatient) of digoxin level determinations was selected and reassessed by a blinded, independent reviewer using the same appropriateness criteria. Differences were resolved by having the reviewers meet and reach consensus. The overall interrater agreement19 for appropriateness was good at 0.65, although agreement was higher for inpatient levels (κ=0.80) than outpatient levels (κ=0.45).
Categorical comparisons were made using the χ2 statistic.20 A 2-sided P value of .05 was considered statistically significant. For normally distributed variables, means with 95% CIs are presented, while for nonnormal variables, we present median values with 25th to 75th percentiles. Ninety-five percent CIs were calculated around point estimates. Interrater reliability was assessed using the κ statistic.21 Data were analyzed using the SAS version 6.12 statistical program.22
The appropriateness criteria for serum digoxin levels measured in outpatients and inpatients were similar and were closely related to Michalko and Blain's criteria14 (Table 1). Digoxin level determinations were always considered to have an appropriate indication when a subtherapeutic response or previously undocumented toxic effects were suspected, when evaluating a high-risk patient, after initiation of digoxin therapy or dosage adjustment after steady state was reached, and as an admission level in inpatients and routine monitoring in outpatients every 10 months. No relevant data could be found regarding frequency of outpatient monitoring in stable, asymptomatic outpatients; the consensus of our experts was that yearly monitoring was reasonable. Ten days was chosen as appropriate timing to measure digoxin levels after initiation of therapy or dosage adjustment because steady state is reached in patients after 4 to 5 elimination half-lives of digoxin.13,16
A random sample of 224 digoxin levels measured in 162 adult inpatients was obtained from 4447 levels performed between October 1995 and March 1996. Among inpatients, 55% were male, mean age was 68 years (SD, 13 years), and 49% of patients were on the medicine service, with 51% on a surgery service (Table 2).
Of the digoxin levels measured from inpatients, 16% (95% CI, 11%-20%) had an appropriate indication (Table 3). Of these levels, 89% were an admission level or measured for routine monitoring, 6% were in cases of high-risk patients, 3% because of suspected toxic reactions, and 3% as baseline or control levels after initiation of digoxin therapy or a change in regimen (Table 4). Eighty-four percent (95% CI, 80%-89%) of digoxin level determinations had no appropriate indication. Of these, 76% were due to early routine monitoring with serial determinations drawn less than 10 days apart, 9.5% were done on patients not receiving digoxin, 8.5% were due to determinations after initiation of digoxin therapy before pharmacological steady state had been achieved, 3% were due to determinations after adjustment of digoxin dosage before steady state had been reached, and 3% were done to follow levels in the toxic range that were measured before the level could be expected to decrease to therapeutic range. There was no difference in the distribution of appropriate and inappropriate digoxin levels by the service (medicine or surgery) ordering the digoxin level. The median time between the actual level and the preceding level was 24 hours, suggesting that routine once-a-day determinations were common (Table 3).
When the test results for inappropriate inpatient levels were analyzed, 31% had a subtherapeutic drug level less than 1.2 nmol/L (<0.9 ng/mL), 56% had a therapeutic drug level between 1.2 and 2.2 nmol/L (0.9 and 1.7 ng/mL), and 14% had a potentially toxic drug level of 2.3 nmol/L or more (≥1.8 ng/mL) (Table 5). No inappropriate level of 2.3 nmol/L (1.8 ng/mL) resulted in an important change in therapy. Of the 26 inappropriate levels of 2.3 nmol/L or more (≥1.8 ng/mL), 12 were elevated in patients not demonstrating any documented signs of toxic effects caused by digoxin, 6 were previously known to be elevated and were falling, 2 were measured the same day that oral loading doses of digoxin were administered, and 6 were measured 1 to 3 days after digoxin therapy was initiated or the dose of digoxin was adjusted in patients who were not demonstrating any documented signs of toxic effects due to digoxin.
A random sample of 130 digoxin levels measured in 117 adult outpatients was selected from the 810 outpatient levels obtained between October 1995 and March 1996. Among sample patients, 50% were male, the mean age was 68 years (SD, 16 years), 44% were emergency department patients, and 37% were evaluated in medical clinics, 15% in cardiology clinics, and 4% in surgical clinics (Table 2).
Of the levels measured in outpatients, 52% (95% CI, 44%-61%) had an appropriate indication: 26.5% in cases of high-risk patients, 26.5% because of routine monitoring, 21% because of suspected toxic reactions, 15% in cases of patients with worsening or not improving congestive heart failure or atrial fibrillation, 8% as baseline or control levels after initiation of digoxin therapy or a change in regimen, and 3% in cases of suspected noncompliance with digoxin therapy (Table 3 and Table 4). Of the digoxin levels measured in outpatients, 48% (95% CI, 39%-56%) had no appropriate indication. Of the inappropriate levels, 76% were due to early routine monitoring, 13% were performed on high-risk patients, 2% were performed on patients with worsening congestive heart failure or atrial fibrillation, 2% were after changing the digoxin regimen but before steady state had been reached, and 8% were due to other causes. Because appropriateness could vary depending on whether the indication for monitoring was atrial fibrillation or congestive heart failure, we compared the appropriateness rates in these 2 groups, but the results were similar. More of the appropriate levels were measured in the emergency department than in other services because outpatients came to the emergency department severely ill. The median time between the actual level and the preceding level was 56.5 days, suggesting that early routine monitoring was common (the guideline suggested only 1 level per 10 months for a stable, asymptomatic patient receiving a stable dose).
Of the inappropriate levels performed on outpatients, 35% had a drug level of less than 1.2 nmol/L (<0.9 ng/mL), 58% had a drug level between 1.2 and 2.2 nmol/L (0.9-1.7 ng/mL), and 6% had a drug level of 2.3 nmol/L or more (≥1.8 ng/mL) (Table 5). Of the 4 inappropriate levels of 2.3 nmol/L or more (≥1.8 ng/mL), all were elevated in patients not demonstrating any signs of toxic reactions to digoxin. In 1 patient, the elevated level led to a subsequent decrease in digoxin dose, while in the other patients there was no change in therapy.
The actual total charges for digoxin level determinations at our institution during the study year were $578,656. Compared with other laboratory assays, the digoxin assay is relatively costly and labor intensive since each is performed individually by a technician. The total number of digoxin levels was 11,128, consisting of 9760 levels performed in inpatients accounting for $507,520 in charges and 1368 levels performed in outpatients accounting for $71,136 in charges. If the level determinations with an inappropriate indication had not been performed (80% of all levels), this would have resulted in a decrease of charges of $460,462.
In this study, we developed criteria for determining the appropriateness of digoxin level monitoring. When we applied these criteria to digoxin level monitoring performed in our institution, we found that only 16% of levels measured in inpatients and 52% of levels measured in outpatients had an appropriate indication. Of the levels with an inappropriate indication, only 26 levels measured in inpatients and 4 levels measured in outpatients were markedly elevated. Only 1 of these levels measured in an outpatient led to an important change in therapy. This outpatient was stable from the cardiovascular standpoint and had no symptoms of toxic reactions to digoxin. Of the inappropriate levels measured in inpatients with results of 2.3 nmol/L or more (≥1.8 ng/mL), none led to an important clinical decision.
Previous studies have focused on the correct time for sampling blood for digoxin levels and on the role of a therapeutic monitoring service in use of digoxin levels.5,7- 11,13,14,23 The prospective study performed by Michalko and Blain14 compared the appropriateness of requests for serum digoxin levels with and without the contribution of a pharmacy-based clinical pharmacokinetic service in a tertiary care institution. In that study, 26.4% of requests for digoxin levels were found to be inappropriate with the intervention of the pharmacokinetic service vs 35.4% without such a service, although this difference was not statistically significant. Although Michalko and Blain's study defined appropriateness criteria for digoxin-level monitoring, its primary aim was to evaluate the role of the pharmacokinetic service in digoxin level monitoring. Also, the study was performed in Canada, which may account for some of the difference in appropriateness rates.
A study by Copeland et al24 evaluated the digoxin drug-monitoring service during a 7-week period at Leeds General Infirmary, Leeds, England, by assessing the appropriateness of the assay request and the action taken following receipt of the level result. Clinicians who ordered digoxin levels were interviewed by questionnaire to determine the appropriateness of the action taken after receipt of the result. Data were collected for 88 (78%) of the digoxin assays requested during the study interval. Of these, 48 (55%) were considered to have an appropriate indication and 22 (25%) were considered to have met all criteria for an appropriate level request and subsequent action. The study of Copeland et al concluded that the digoxin-monitoring service is used inappropriately at Leeds General Infirmary due to the high percentage (75%) of inappropriately obtained assays. This article offers no statistical analysis of the data, bringing into question the statistical significance of the results, and the study did not stratify the use of digoxin levels among inpatients and outpatients. However, these results support our finding of a large proportion of digoxin levels with no appropriate indication.
IN OUR STUDY, the primary reason for inappropriate digoxin levels was early routine monitoring. In inpatients, once-per-day levels without clinical justification were common. Similarly, in outpatients, several levels per year without clinical justification accounted for the majority of levels without an appropriate indication. The early monitoring may reflect test-ordering patterns in which levels are ordered daily in inpatients for as long as the patient stays in the hospital; we have found similar patterns for antiepileptic drug levels.25 Also, levels may be ordered in outpatients whenever the outpatient comes to the hospital for an ambulatory visit.
Digoxin levels are most useful when ordered in the context of a patient's symptoms and clinical condition. Levels should be performed to answer a particular clinical question or to monitor a stable patient's condition at reasonable time points.7,9,11,23 In addition, levels should generally be determined after digoxin has reached steady-state concentration (after 4 -5 half-lives) after changing the dose or starting digoxin therapy.13
Adhering to the simple criteria outlined in this study has the potential of improving the use of digoxin levels without compromising useful clinical information. However, some of the indications considered inappropriate according to the guidelines validated by our expert panel are debatable. Specifically, the intervals for monitoring stable, asymptomatic outpatients; the optimal interval after a dose change; the use of peak levels; and measuring falling levels after an initial toxic level may be controversial and deserve further study. Regarding frequency of monitoring for stable, asymptomatic outpatients, few data are available. Our experts felt annual monitoring was reasonable, but this issue in particular should be evaluated further. How long to wait after a dose change is also somewhat problematic because the half-life varies between individuals. Peak levels may on rare occasions be helpful, but troughs are almost always sufficient. After an initial toxic level, it may be useful to confirm the result or to determine that the level has fallen to a degree sufficient to restart administration. Our guidelines suggested that clinicians wait until the level would have been expected to fall into the therapeutic range based on pharmacokinetics.
Assuming a system that could prevent all inappropriate tests from being performed and no adverse impact on patient care, yearly charges would decline by about $426,317 and $34,145 in inpatients and outpatients, respectively. It is important to note that the resource utilization figures we use represent charges, not costs. For laboratory tests in particular, charges are much higher than costs, so that the cost savings would be smaller than the charge estimates. Moreover, incremental cost declines with reductions in volume are not proportional with small test volume charges.
One strategy with great potential for reducing the proportion of levels with inappropriate indications involves implementing the criteria in a computerized order-entry system, in which physicians write orders directly on computers. Redundant tests should be particularly easy to eliminate using a computerized system to check for tests performed earlier than a specific interval.26 Previous studies have demonstrated that computerized medical information systems and educational interventions can decrease the number of digoxin assays ordered by physicians, suggesting that interventions using order-entry systems may also lead to a decrease in levels ordered.12,27
This study has a number of limitations. The criteria do not address specific clinical situations, such as patients with a recent myocardial infarction and thyroid disease, and may not pertain to particular patient populations, such as children and pregnant women. However, it has been suggested that routine monitoring of digoxin levels in children may be unnecessary unless digoxin toxicity is suspected since there is a substantial overlap between toxic and therapeutic concentrations.28 Another limitation is that our study only included certain laboratory parameters as markers for patient function. For example, we used creatinine levels as a marker for renal function without considering serum urea nitrogen levels and we did not investigate the calcium levels of the patients, which may be important in patients in whom digoxin toxicity might have been enhanced due to a decrease in serum potassium levels with a concomitant increase in serum calcium levels. An additional limitation is that the appropriateness rates may not be generalizable to other settings because data used in this study were collected in a single university teaching hospital that does not have a drug monitoring service, and the inpatients receiving digoxin were relatively ill as reflected by the long median length of stay. However, the appropriateness criteria are likely to be useful in other settings. A further limitation is that appropriateness was assessed based on chart review and computerized records, which may not always reveal reasons for ordering a digoxin level. Patient symptoms and suspected noncompliance may not always have been recorded as reasons to order a level. Therefore, our study may have underestimated the proportion of levels with an appropriate indication. Also, there was some disagreement between reviewers about which levels were inappropriate—more in the outpatient setting than in the inpatient setting. However, few inappropriate levels resulted in important clinical consequences and, if appropriate levels were frequently being classified as inappropriate, more frequent important results would be expected. Another issue is that inpatients were not stratified by their indication for digoxin treatment, atrial fibrillation vs heart failure, and in atrial fibrillation there is a physiologic marker (heart rate) of the effect of digoxin. However, among outpatients the inappropriateness rate did not differ by indication for digoxin therapy, and clinically the indications for performing levels are the same in both conditions.
We conclude that only 16% of levels measured in inpatients and 52% of levels measured in outpatients had an appropriate indication. These data suggest that measures to improve the ordering of digoxin levels are needed, and may result in significant reductions in the numbers of tests done, without loss of clinically useful information.
Accepted for publication May 26, 1998.
This study was supported in part by grant R01 HS08927 from the Agency for Health Care Policy and Research, Rockville, Md.
Reprints: David W. Bates, MD, MSc, Division of General Medicine and Primary Care, Brigham and Women's Hospital, 75 Francis St, Boston, MA 02115 (e-mail: email@example.com).