Cost of Initial Therapy in the Electrophysiological Study Versus ECG Monitoring Trial (ESVEM)
Background Patients randomized to either serial electrophysiological testing (EPS) or serial Holter monitoring (HM) to guide antiarrhythmic therapy for life-threatening ventricular arrhythmias had equivalent rates of mortality and arrhythmia recurrence in the ESVEM study. This report analyzes the effects of EPS, HM, and clinical factors on the charges for initial evaluation and management of patients with life-threatening ventricular arrhythmias.
Methods and Results Ten of 14 clinical centers participating in ESVEM provided bills from the initial hospitalization for randomized patients. Predictors of charges (1991 dollars) were analyzed by linear regression after logarithmic transformation. Initial hospital charge data were obtained for 286 patients randomized in ESVEM (88% of patients eligible for this substudy, 59% of all ESVEM patients). Patients with charge data were somewhat more likely to be older, to be female, and to have failed previous antiarrhythmic drug therapy at study entry and were less likely to have a drug predicted effective after randomization. Mean overall hospital charges were $35 986 (SD, $32 628) with a median of $24 532 (interquartile range, $16 126 to $43 593). Prerandomization patient characteristics generally had insignificant effects on charges, with the exception of presentation with resuscitated sudden death (28% increase in charges, P=.01) and heart failure (26% increase in charges, P=.02). Patients randomized to EPS had higher mean charges for evaluation ($42 002 versus $29 970, P=.0015) as well as more drug trials (3.0 versus 2.1, P=.0001) and a longer hospital stay (19.6 versus 13.9 days, P=.0007). In a multivariate regression model, failure to find an effective drug (P=.0001), the number of drug trials (P=.0001), and resuscitated sudden death as the presenting arrhythmia (P=.0001) were the only independent predictors of higher initial charges.
Conclusions (1) Initial hospital charges are significantly higher for EPS-guided than HM-guided therapy. (2) The higher charges for EPS-guided therapy were due to a greater number of drug trials and a lower probability of finding an effective drug. (3) Failure to find an effective drug, a larger number of drug trials, and a history of resuscitated sudden death independently predict higher charges.
Electrophysiological testing (EPS) and Holter monitoring (HM) with exercise testing have been used since the 1970s to prospectively identify drug therapy that might prevent recurrences of sustained ventricular arrhythmias. Although each method appears to predict treatment outcome, the superiority of either method has been uncertain. The question of whether invasive (EPS) or noninvasive (HM) evaluation is more accurate in predicting antiarrhythmic drug efficacy in patients with ventricular tachyarrhythmias was therefore addressed in the multicenter randomized Electrophysiologic Study Versus Electrocardiographic Monitoring (ESVEM) trial.1 2
ESVEM investigators have reported equivalent actuarial rates of death and recurrent arrhythmia for the EPS and HM groups during follow-up.3 Among patients with a drug efficacy prediction, the risk of arrhythmia recurrence was higher in patients who had previously failed antiarrhythmics or were discharged on drugs other than sotalol.4 Although determination of the effect of efficacy prediction method (EPS versus HM) on recurrent arrhythmias was the primary objective of ESVEM, an assessment of relative costs of therapy was an important secondary goal. The purpose of this report is to present an analysis of the factors that affected the initial hospital charges for the evaluation and treatment of life-threatening ventricular arrhythmias in ESVEM.
The design, methods, enrollment statistics, and main findings of ESVEM have been published in detail elsewhere.1 2 3 4 In brief, patients were screened for entry into the trial if they had either (1) sustained ventricular tachycardia or fibrillation, (2) resuscitated cardiac arrest, or (3) unmonitored syncope with subsequent inducible, sustained ventricular tachycardia. Since clinically eligible patients had to be suitable for serial drug testing by either EPS or HM, all patients were required to have frequent ventricular ectopy on Holter monitoring (at least 480 premature ventricular contractions during 48 hours) and reproducibly inducible, sustained ventricular tachyarrhythmia on EPS. Randomized patients then underwent serial testing of up to six of seven antiarrhythmic drugs (imipramine, mexiletine, pirmenol, procainamide, propafenone, quinidine, and sotalol) in random order until an efficacy prediction was achieved and a tolerable drug selected or until all drugs the patient was eligible to receive had been tried. Consenting patients who met inclusion criteria were randomized between October 1, 1985, and February 15, 1991, in 14 clinical sites and followed to detect arrhythmia recurrence and death until February 15, 1992.
Ten of the 14 ESVEM centers participated in a study to document the hospital charges for initial evaluation and management. Hospital bills and professional fees were obtained from hospitalizations that included the date of randomization and date of the final efficacy prediction. Patients admitted to hospitals that do not bill patients for services were excluded. Copies of the hospital bills were checked for completeness and entered into computer storage with the dataease system (Software Solutions, Inc). The data were then transferred to PC-SAS datasets (SAS Inc) and subjected to range and consistency checks before further analysis.
Categorical data were presented as simple proportions. Continuous data were described by the mean±SD as well as median (interquartile range [IQR], ie, 25th to 75th percentile). Two sample statistical tests were performed with the Wilcoxon rank sum test. The null hypothesis was rejected when the probability value was P<.05. Hospital charges were statistically adjusted to 1991 dollars by application of the medical component of the consumer price index.5 The effect of clinical and technical factors on charges was examined by linear regression analysis after logarithmic transformation of charges, which were highly skewed.6 All regression analyses were adjusted first for cost differences among clinical centers before the effect of clinical and technical factors on charges was tested. All analyses were performed with sas version 6.04 for personal computers (SAS Inc).
In all, 486 patients were randomized in ESVEM: 242 to serial EPS and 244 to serial Holter monitoring. One hundred forty patients enrolled in hospitals that do not bill for services (137 from Department of Veterans Affairs Medical Centers and 3 from a Canadian center) were excluded from the cost analysis, as were an additional 22 patients who were enrolled in clinical sites that did not participate in the cost study. Hospital bill data were obtained for 286 of the remaining 324 patients (88% of the patients eligible for this cost substudy, 59% of all ESVEM patients). Patients with charge data were slightly older, less likely to be male, and more likely to have previously failed antiarrhythmic drug therapy (Table 1⇓). In addition, an effective drug was found in a higher proportion of patients with charge data. As expected, the clinical characteristics of patients randomized to HM were quite similar to those of patients randomized to EPS among patients with charge data.
Initial hospital charges for all patients (irrespective of method of evaluation) had a mean of $35 986 (SD, $32 628) and a median of $24 532 (IQR, $16 126 to $43 593). The largest contributors to charges (Table 2⇓) were room charges (39%), electrophysiology laboratory (22%), professional fees (13%), and clinical laboratory (5%). The interval from date of admission to randomization had a mean of 6.0 days (SD, 7.6 days) and a median of 5 days (IQR, 3 to 7 days). The interval from date of randomization to date of discharge had a mean of 16.8 days (SD, 13 days) and a median of 12 days (IQR, 8 to 21 days).
Total charges were significantly higher by Wilcoxon rank sum test in the patients randomized to EPS, with a mean of $42 002 (median, $29 096) versus a mean of $29 970 (median, $20 737) in the HM group (P=.0015). A breakdown of charges by hospital department (Table 2⇑) showed that EPS patients had higher charges for the hospital room (mean, $14 841 versus $10 277), electrophysiology laboratory ($10 132 versus $5207), and professional fees ($5711 versus $3602). The number of drug trials (mean, 3.0 [median, 3]) was higher for the EPS group than for the HM group (mean, 2.1 [median, 2]; P=.0001). Length of stay after randomization was also 23% longer for EPS than HM (mean, 19.6 days [median, 14 days] versus 13.9 days [median, 11 days]; P=.0007).
The ESVEM investigators previously reported that patients randomized to EPS were significantly less likely to achieve a drug efficacy prediction than patients randomized to HM.2 3 In patients with charge data available, 69 of the 142 patients randomized to EPS achieved an efficacy prediction (49%), whereas 118 of the 142 patients randomized to HM achieved an efficacy prediction (83%). The charges in the 69 patients randomized to EPS who achieved an efficacy prediction (mean, $24 693; median, $20 789) were quite similar to charges in the 118 patients randomized to HM who achieved an efficacy prediction (mean, $24 017; median, $19 578) (P=.49). Furthermore, the charges in the 73 patients who were randomized to EPS and who failed to achieve a drug efficacy prediction (mean, $58 362; median, $43 989) were elevated to a degree similar to the charges in the 24 patients randomized to HM who failed to achieve a drug efficacy prediction (mean, $59 240; median, $52 837) (P=.59).
Most clinical characteristics of patients had <10% effect on the charges for initial evaluation and treatment (Table 3⇓). Age, sex, cause of cardiac disease, prior myocardial infarction, frequency of ventricular tachycardia, left ventricular ejection fraction, prior antiarrhythmic drug failure, and cardiac enlargement on chest radiograph were not significantly correlated with initial charges (Table 3⇓). However, patients with evidence of heart failure had 26% higher charges (P=.02), and patients with resuscitated sudden death or documented ventricular fibrillation had 28% higher initial charges (P=.01) (Table 3⇓).
In contrast to the limited predictive value of clinical factors, technical aspects of the drug testing program were strongly predictive of initial hospital charges. As discussed in detail above, randomization to EPS led to an increase in charges by 32% (P=.0009 by regression analysis). In addition, the number of drug trials was strongly correlated with charges and length of hospital stay (P=.0001). Patients for whom no effective drug was found had 125% higher charges (P=.0001). Finally, clinical site of randomization was a significant predictor of charges (P=.002).
Patients for whom sotalol was predicted to be an effective drug had charges similar to those for patients for whom other drugs were predicted to be effective (mean, $21 696 versus $25 365; P=.44). Patients randomized to receive sotalol on the first drug trial also had charges similar to those for patients randomized to receive other drugs on the first drug trial (Table 4⇓). There was a complex interaction, however, among the first drug tested, the efficacy of that drug, and charges. If the first drug tested was effective, the charges of evaluation were low and equivalent between sotalol and other drugs. If sotalol failed as the first drug tested, however, there was a trend toward higher charges (mean, $53 870 versus $42 685; P=.14) than if other drugs failed as the first drug tested.
Clinical and technical factors that were predictive of charges on univariate testing (P<.05) were further evaluated by a multivariate stepwise linear regression with the logarithm of charges as the dependent variable (Table 5⇓). In all models, the effect of clinical site on charges was controlled statistically before other candidate predictive variables were examined. In the multivariate analysis of factors known before drug testing, patients who presented with resuscitated cardiac arrest or documented ventricular fibrillation had 27% higher hospital charges (P=.02), while patients with heart failure had 24% higher charges (P=.03). The primary randomization variable (method) was then added to the model, and those randomized to EPS had 34% higher charges after heart failure and presenting arrhythmia were controlled for (P=.0004). When the number of drug trials and an indicator of success in finding an effective drug were allowed to enter the model, randomization assignment was no longer significant, implying that these factors accounted for the higher charges for EPS. Failure to find an effective drug was the most powerful predictor of charges, associated with an 85% increase in cost, followed by presentation with resuscitated cardiac arrest (38% increase) and number of drug trials (15% increase per additional drug trial).
These results were essentially unchanged when the analyses were rerun after the data from the two patients with the highest charges and the two patients with the lowest charges were deleted. These results were also essentially unchanged by adjustment for the clinical factors (age, sex, and failure of prior antiarrhythmic therapy) that were significantly different between the ESVEM patients with and without available cost data (Table 1⇑).
The ESVEM trial compared serial EPS with serial HM and exercise testing as methods for evaluating the efficacy of antiarrhythmic drugs in the treatment of life-threatening ventricular arrhythmias. Randomized patients in the two groups had equivalent rates of mortality and arrhythmia recurrence over a follow-up of 6 years.3 The present investigation found that serial EPS increased, by 32%, the charges of the initial evaluation and management of patients with life-threatening ventricular arrhythmias. The higher charges for patients randomized to EPS were due to the greater number of drug trials required and the lower overall probability of finding an effective drug compared with the HM limb.
In the care of patients with a chronic disease, a long-term perspective on the cost of care is important, since a therapy that is more effective may cost more initially but avert long-term complications with their attendant costs. The present analysis included only the cost of initial evaluation and management, but the equivalence of long-term outcomes in the EPS and HM limbs of the ESVEM trial3 does not suggest that the initially higher cost of EPS will be recovered by cost savings over the long term. If the cost of treating a recurrence in HM patients were shown to be higher than the cost of treating a recurrence in EPS patients, however, the long-term cost of therapy in the two treatment strategies might prove equal over the course of illness. Long-term cost data from ESVEM are currently being collected retrospectively, and this issue will be examined directly.
Although cost has not generally been an outcome measure in randomized clinical trials, the one other randomized study of EPS and HM for evaluation of drug efficacy included an assessment of cost.7 8 In that trial, which was performed in Canada, the mean initial hospital costs were higher for EPS than for HM (mean, $CDN 13 164 versus $CDN 6869; P<.001), consistent with our finding of a higher initial cost of EPS. Since the 2-year actuarial probability of a recurrent arrhythmia in the Canadian trial was greater for HM than EPS (P=.02), the follow-up hospital costs were greater for HM (mean, $CDN 9204 versus $CDN 3784, P=.01). Long-term hospital costs were therefore comparable in the two limbs, because the higher initial cost of EPS in the Canadian trial was offset by the lower follow-up costs. The long-term cost data from ESVEM are not yet available, so direct comparison of the long-term cost findings in ESVEM with those in the Canadian trial is not possible. The Canadian trial differed in several important respects from ESVEM, however, including fewer patients (57 versus 486), fewer arrhythmia recurrences (18 versus 278), and briefer follow-up. The Canadian trial also used amiodarone in many of the patients (initially in 46% of the EPS group and in 0% of the HM group), which may have contributed to the significantly lower rate of arrhythmia recurrence in the EPS group. The Canadian trial differed so extensively from ESVEM that even when long-term cost data from ESVEM become available, the two trials will not be readily comparable.
Two clinical characteristics that affected charges in the present investigation were a presenting history of documented ventricular fibrillation or resuscitated cardiac arrest and the presence of congestive heart failure. Patients with a history of cardiac arrest probably had higher charges because of the need to treat associated comorbidity on admission, since they had a longer length of stay before randomization but similar mean length of stay after randomization, a similar number of drug trials, and probability of finding an effective drug. It is likely that patients with heart failure had higher charges because the probability of finding an effective drug was significantly lower in patients with a lower ejection fraction.3 These observations illustrate the interplay of technical factors and clinical factors in determining costs.
Sotalol was the most effective drug in preventing recurrent arrhythmic events in ESVEM, so its effect on cost is of particular interest.4 Overall initial hospital charges were insignificantly lower among patients who received sotalol (Table 4⇑). A potential explanation for this finding lies in the observation that since sotalol was the drug most likely to be effective,4 patients who failed other drugs still had the opportunity to receive sotalol and thus required fewer drug trials and incurred lower charges. Patients who failed sotalol, however, had received the most effective drug available and thus required more drug trials and incurred higher charges. The ESVEM protocol used serial drug testing of up to six drugs, and it is possible that other approaches to patients who failed initial therapy with sotalol would have reduced the cost of initial evaluation and management.
The approach to management of patients with life-threatening ventricular arrhythmias is now changing toward greater use of amiodarone,9 10 11 12 implantable defibrillators,13 14 15 16 and ablative therapy.17 It is essential to await randomized studies13 14 18 to judge whether outcome is improved by these new approaches, since ESVEM and CAST19 20 have both demonstrated that conventional wisdom about the superiority of diagnostic and therapeutic strategies may be wrong. It will be important to examine the costs as well as the clinical efficacy of these newer approaches.
The randomized clinical trial is accepted as the most reliable clinical research design because random assignment helps ensure that the treatment groups will be equivalent within the play of chance on all factors, both measured and unmeasured. Randomization ensures the internal validity of the trial results, and this feature elevates the randomized trial above investigations that use either historical, literature, or concurrent nonrandomized comparison groups. Despite the strengths of their design, the generalizability of randomized trials to wider groups of patients may be questioned, as well as their continued relevance as scientific knowledge and management options evolve. We have attempted to analyze cost data from ESVEM to ensure maximum relevance to the “moving target” of contemporary arrhythmia management strategies. Our findings that patients with congestive heart failure or prior cardiac arrest require more costly resources to manage probably reflect the generally higher costs in patients with a greater severity of illness. We found that serial EPS studies were more expensive than serial HM studies, but we also showed that this difference was completely explained by the greater number of drug trials and the lower overall probability of ultimately finding an effective drug rather than the cost of the studies themselves. The analyses reported here suggest, quite plausibly, that costs will be minimized by use of approaches that shorten the time to find an effective therapy. These basic insights about the determinants of cost in the ESVEM trial should be generalizable to most forms of evaluation and management of life-threatening arrhythmias.
Cost analysis in clinical studies is a relatively recent development, and several methodological aspects of this study merit discussion. It is well known that hospital charges are not equal to hospital costs,21 so the validity of using charges as a proxy for costs can be questioned. In a recent investigation, we examined four different methods of calculating the cost of coronary angioplasty and coronary bypass surgery and found that the absolute difference in cost between the two procedures varied from $1935 to $10 087, depending on the cost-accounting method used.22 Importantly, we also found that the different cost-accounting methods affected cost estimation of the two procedures in a roughly parallel fashion, so that the ratio of costs between angioplasty and surgery was relatively insensitive to the method of cost accounting. Furthermore, we found that the ratio of charges for coronary angioplasty to the charges for coronary bypass surgery (0.51) was a very good proxy for the ratio of costs calculated by the four alternative cost methods (0.58, 0.47, 0.48, 0.49). The present study used the ratio of charges between groups of patients as the primary mode of analysis, which our previous work demonstrates is a much more reliable reflection of underlying costs. Thus, we believe that the present analysis provides an accurate estimate of the relative costs in ESVEM.
We found statistically significant heterogeneity in hospital charges among the 10 ESVEM centers participating in the cost analysis. Charges by hospitals may vary because of differences in either practice style or underlying cost structures. Variation in practice style among different regions of the country has received considerable attention recently but is unlikely to have had a major effect on costs in ESVEM because clinical management was dictated by a standardized protocol. Differences in the operating costs and costs of the physical plant of the hospitals undoubtedly affected the charges in different hospitals. We performed all statistical analyses of the effect of clinical factors on charges both with and without adjustments for the cost variations among centers. The results of the two sets of analyses were virtually identical, and we report only the adjusted analyses. We believe inclusion of a diversity of hospitals makes the cost analysis presented here more widely generalizable than if the study had been performed in a single institution.
This study has a number of limitations. First, although 88% of eligible patient bills were available, the study patients composed 59% of all randomized ESVEM patients, largely because of the large number of patients enrolled from Veterans Affairs Medical Centers. Although patients with charge data were older, more often female, and more likely to have failed prior antiarrhythmic therapy (Table 1⇑), none of these factors were significantly correlated with cost (Table 3⇑). The multivariable analysis (Table 5⇑) was not affected by adjustment for the clinical factors that differed between patients with and without hospital charge data. Thus, the main conclusions appear to be generalizable to patients who meet eligibility criteria for ESVEM. Second, although dates of randomization were used to compute study lengths of stay, total charges refer to the entire length of stay during the index hospitalization, introducing additional variations into the analysis. These variations were balanced by random assignment, however, and simply make the calculated levels of statistical significance more conservative. Third, the study reports only initial charges and should therefore be regarded as an interim analysis. Follow-up cost data are currently being collected retrospectively. Finally, hospital charges were used in these analyses rather than costs. As discussed above, the ratio of charges in different clinical subgroups has been shown to be an accurate reflection of the ratio of costs, so that our conclusions are unlikely to have been affected.22
The present study suggests that arriving at a management plan expeditiously for patients with arrhythmias leads to substantially lower costs and also suggests that evaluation approaches that yield an accurate prediction of drug efficacy or identify drug failure quickly may have a cost advantage over more time-consuming approaches. Documentation of costs in long-term follow-up will be necessary to evaluate the relative cost-effectiveness of alternative management strategies for life-threatening ventricular arrhythmias.
This study was supported by a grant from Bristol-Meyers Squibb Co, Princeton, NJ, and grant HL-34071 from the National Heart, Lung, and Blood Institute, Bethesda, Md; with additional support from Knoll Pharmaceuticals, Whippany, NJ; Boehringer-Ingleheim Corp, Ridgefield, Conn; Parke-Davis, Morris Plains, NJ; CIBA-Geigy Corp, Summit, NJ; and grant RR-00064 from the National Center for Research Resources, Bethesda, Md. We thank Kathryn Gelman for assistance in programming, Joel Alderman and Mary Harrison for assistance in data entry, and Judy Davis, Carrie Bryson, and Elaine Steel for typing the manuscript.
Reprint requests to Mark A. Hlatky, MD, Stanford University School of Medicine, HRP Redwood Building, Room 265, Stanford, CA 94305-5092.
- Received August 12, 1994.
- Revision received August 25, 1994.
- Accepted September 23, 1994.
- Copyright © 1995 by American Heart Association
ESVEM Investigators. The ESVEM trial: electrophysiologic study versus electrocardiographic monitoring for selection of antiarrhythmic therapy of ventricular arrhythmias. Circulation. 1989;79:1354-1360.
ESVEM Investigators. Determinants of predicted efficacy of antiarrhythmic drugs in the electrophysiologic study versus electrocardiographic monitoring trial. Circulation. 1993;87:323-329.
Statistical Abstract of the United States. Washington, DC: US Government Printing Office; 1993.
Ceremuzynski L, Kleczar E, Krzeminska-Pakula M, Kuch J, Nartowicz E, Smielak-Korombel J, Dyduszynski A, Maciejewicz J, Zaleska T, Lazarczyk-Kedzia E, Motyka J, Paczkowska B, Sczaniecka O, Yusuf S. Effect of amiodarone on mortality after myo-cardial infarction: a double-blind, placebo-controlled, pilot study. J Am Coll Cardiol. 1992;20:1056-1062.
Cairns JA, Connolly SJ, Gent M, Roberts R. Post-myocardial infarction mortality in patients with ventricular premature depolarizations: Canadian Amiodarone Myocardial Infarction Pilot Study. Circulation. 1991;84:550-557.
Pfisterer ME, Kiowski W, Brunner H, Burckhardt D, Burkart F. Long-term benefit of 1-year amiodarone treatment for persistent complex ventricular arrhythmias after myocardial infarction. Circulation. 1993;87:309-311.
Saksena S, Camm AJ. Implantable defibrillators for prevention of sudden death: technology at a medical and economic crossroad. Circulation. 1992;85:2316-2321.
Adler SW, Remole S, Benditt DG. Impact of implantable cardioverter-defibrillators on prognosis of cardiac arrest survivors: a continuing controversy. Circulation. 1993;88:1348-1350.
Scheinman MM, Laks MM, DiMarco J, Plumb V. Current role of catheter ablative procedures in patients with cardiac arrhythmias. Circulation. 1991;83:2146-2153.
Echt DS, Liebson PR, Mitchell LB, Peters RW, Obias-Manno D, Barker AH, Arensberg D, Baker A, Friedman L, Greene HL, Huther ML, Richardson DW. Mortality and morbidity in patients receiving encainide, flecainide, or placebo: the Cardiac Arrhythmia Suppression Trial. N Engl J Med. 1991;324:781-78.
Epstein AE, Hallstrom AP, Rogers WJ, Liebson PR, Seals AA, Anderson JL, Cohen JD, Capone RJ, Wyse DG. Mortality following ventricular arrhythmia suppression by encainide, flecainide, and moricizine after myocardial infarction: the original design concept of the Cardiac Arrhythmia Suppression Trial (CAST). JAMA. 1993;270:2451-2455.
Finkler SA. The distinction between costs and charges. Ann Intern Med. 1982;96:102-110.
Hlatky MA, Lipscomb J, Nelson C, Califf RM, Pryor D, Wallace AG, Mark DB. Resource use and cost of initial coronary revascularization. Circulation. 1990;82(suppl IV):IV-208-IV-213.