Echocardiography in patients with suspected endocarditis: a cost-effectiveness analysis

CLINICAL STUDIES

Echocardiography in Patients with Suspected Endocarditis: A Cost-effectiveness Analysis Paul A. Heidenreich, MD, MS, Frederick A. Masoudi, MD, Brijeshwar Maini, MD, Tony M. Chou, MD, Elyse Foster, MD, Nelson B. Schiller, MD, Douglas K. Owens, MD, MS PURPOSE: We sought to determine the appropriate use of echocardiography for patients with suspected endocarditis. PATIENTS AND METHODS: We constructed a decision tree and Markov model using published data to simulate the outcomes and costs of care for patients with suspected endocarditis. RESULTS: Transesophageal imaging was optimal for patients who had a prior probability of endocarditis that is observed commonly in clinical practice (4% to 60%). In our base-case analysis (a 45-year-old man with a prior probability of endocarditis of 20%), use of transesophageal imaging improved quality-adjusted life expectancy (QALYs) by 9 days and reduced costs by $18 per person compared with the use of transthoracic echocardiography. Sequential test strategies that reserved the use of transesophageal echocardiography for patients who had an inadequate transthoracic

study provided similar QALYs compared with the use of transesophageal echocardiography alone, but cost $230 to $250 more. For patients with prior probabilities of endocarditis greater than 60%, the optimal strategy is to treat for endocarditis without reliance on echocardiography for diagnosis. Patients with a prior probability of less than 2% should receive treatment for bacteremia without imaging. Transthoracic imaging was optimal for only a narrow range of prior probabilities (2% or 3%) of endocarditis. CONCLUSION: The appropriate use of echocardiography depends on the prior probability of endocarditis. For patients whose prior probability of endocarditis is 4% to 60%, initial use of transesophageal echocardiography provides the greatest quality-adjusted survival at a cost that is within the range for commonly accepted health interventions. Am J Med. 1999;107:198 –208. 䉷1999 by Excerpta Medica, Inc.

I

gram be obtained? And second, should transesophageal or transthoracic echocardiography be used as the first test? Recent studies have demonstrated that, compared with transthoracic imaging, transesophageal echocardiography is more accurate for the diagnosis of endocarditis, particularly for aortic, mitral, and prosthetic valves (8 – 12). It is more costly, however, and carries a small risk of death (13). Several algorithms based on clinical experience suggest performing a transthoracic study first and, if the study is inadequate or if a high suspicion of endocarditis remains, proceeding to transesophageal study (14,15). No prospective study has compared the costs and benefits of the different strategies. In this study, we used published data to evaluate the cost-effectiveness of diagnostic strategies for endocarditis that use echocardiography. In particular, we sought to address the following questions: 1) What are the benefits and costs of performing a transesophageal study compared with other strategies using echocardiography? 2) At what pretest probability (suspicion) of endocarditis should testing be initiated? and 3) What patient, test, treatment, and disease characteristics affect the decision to use a particular test or combination of tests?

nfective endocarditis remains a serious disease associated with high mortality, morbidity, and cost (1,2). Although physicians traditionally have made the diagnosis of endocarditis clinically (3), the recently published Duke diagnostic criteria (4) include the use of echocardiography because of its ability to visualize valvular vegetations. Although it is now considered standard of care to use echocardiography to evaluate patients with suspected endocarditis (5–7), two questions remain. First, at what level of suspicion should an echocardio-

From the Veterans Affairs Palo Alto Health Care System, (PAH, DKO), Palo Alto, California; the Departments of Health Research and Policy, (PAH, DKO), Stanford University, Stanford, California; the Division of Cardiology (FAM), University of Colorado Health Sciences Center, Denver, Colorado and the Division of Cardiology (BM, TMC, EF, NBS), Department of Medicine, University of California at San Francisco, San Francisco, California. Drs. Owens and Heidenreich were supported by career development awards from the Veterans Affairs Health Services Research and Development Service. Requests for reprints should be addressed to Paul A. Heidenreich, MD, 111C Cardiology, Palo Alto VAMC, 3801 Miranda Avenue, Palo Alto, California 94034. Manuscript submitted January 7, 1999, and accepted in revised form May 6, 1999. 198

䉷1999 by Excerpta Medica, Inc. All rights reserved.

0002-9343/99/$–see front matter PII S0002-9343(99)00216-8

Echocardiography for Suspected Endocarditis/Heidenreich et al

Figure 1. The square node represents the decision to use one of six diagnostic strategies for a patient with endocarditis. The first strategy is treatment for bacteremia only, without echocardiography. The next two strategies use either a transthoracic study (TTE) or a transesophageal study (TEE) to determine whether treatment for endocarditis is needed. The fourth strategy starts with a TTE; a TEE is performed if the TTE is negative. The fifth strategy uses both the results and the quality of the initial TTE to determine whether a TEE is needed. The final strategy treats all patients for endocarditis without performing echocardiography. All patients then enter one of the subtrees shown in Figure 2.

METHODS Decision Model We developed a decision model using EXCEL (Version 5.0, Microsoft Corporation, Redmond, Washington) and DATA (Version 3.0, TreeAge Software, Boston, Massachusetts) software. The model compares six strategies for the diagnosis of endocarditis (Figures 1 and 2). It consists of a decision tree that represents the immediate consequences for patients with suspected endocarditis, and a four-state Markov model (16,17) to evaluate long-term survival. In the first strategy (bacteremia treatment), no echocardiogram is performed, and the patients are treated for bacteremia only (10 to 14 days of antibiotics). In the second strategy (transthoracic echocardiography), the result of a transthoracic study is used to assign treatment. In the third strategy (transesophageal echocardiography), the result of a transesophageal study is used to assign treatment. In the fourth strategy (sequential, no quality check), a transthoracic study is performed; if it is negative, regardless of its quality, a transesophageal study is performed. In the fifth strategy (sequential with quality check), a transthoracic study is performed; if it is positive, the patients are treated; if it is negative, and of excellent quality, the patients are not treated; if it is negative, but not of excellent quality, then a transesophageal echocar-

diogram is performed and treatment is based on the results of the transesophageal study. In the sixth strategy (treat all), the patients are treated for endocarditis without relying on echocardiography for a diagnosis. The definition of a positive echocardiogram is the detection of a vegetation. All patients with a positive study are treated for endocarditis. We discounted quality-adjusted survival and costs using a 3% annual rate (18). Each diagnostic study leads to one of four outcomes: a true-positive, true-negative, false-positive, or false-negative diagnosis of endocarditis (Figure 2). We derived the probability of each test result by using Bayes’ theorem, the prior probability of disease (varied in sensitivity analysis between 0% and 100%) (19 –21), and the sensitivity and specificity of transthoracic and transesophageal echocardiography (Table 1) (9,11,12,22–25). We assumed that there was no immediate health risk associated with a transthoracic examination but that transesophageal echocardiography carried a 0.01% risk of death (13).

Health Outcomes We modeled three outcomes after treatment for endocarditis: death during hospitalization (in 15% of patients) (25,26), relapse (in 2.5%) (26 –28), and cure (Figure 2). For the patient who had undiagnosed endocarditis and

September 1999

THE AMERICAN JOURNAL OF MEDICINE威

Volume 107 199

Echocardiography for Suspected Endocarditis/Heidenreich et al

patients who survived hospitalization, we used a fourstate Markov model (16,17). The model determined the outcomes and costs during a 50-year period to simulate a cohort of patients who change in age from 45 to 95 years (Tables 1 and 2). Patients who survive endocarditis (recovered in Figure 2) can be in one of three states: postendocarditis without valve replacement, postendocarditis with valve replacement, and dead (not shown). We assumed that patients who recovered from endocarditis had an annual mortality that was 3.3-fold greater than patients in the general population, because of the risks of progressive valve dysfunction and heart failure (26,27). We assumed that patients who had bacteremia without endocarditis had a 1.1-fold greater mortality than the general population (21,31,32). We estimated age-specific death rates for causes other than endocarditis using a Gompertz survival function (16,33) that approximated US life-table data. We assumed a 20% valve replacement rate during the initial hospitalization for endocarditis (27,34,35). We assumed that after recovery from endocarditis, 4% of patients would require a valve replacement each year (26,27,36) during the first 10 years after endocarditis and that this rate would decrease to 0% by year 20.

Probability of Endocarditis Figure 2. These subtrees display the outcomes of the four possible diagnostic test results. For patients undergoing a transesophageal study (TEE), there is a small chance of death. Patients treated for endocarditis without echocardiography are placed in the true-positive branch (if endocarditis is present) or false-positive branch (if there is no endocarditis). Similarly, patients treated for bacteremia without echocardiography are placed in the true-negative branch (no endocarditis) or falsenegative branch (endocarditis present). Because all patients are treated for bacteremia, there is a small chance of endocarditis cure even if endocarditis treatment is not given after a falsenegative diagnosis. Patients who survive enter one of two Markov processes depending on whether they did (labeled “recovered”) or did not (labeled “well”) have endocarditis.

received 10 to 14 days of antibiotic treatment for bacteremia, we assumed, based on expert opinion, that this therapy would cure 20% of patients with endocarditis. For patients who were not cured, we assumed that they would either die or eventually be diagnosed and treated for endocarditis. We assumed, based on expert opinion, that a delay in diagnosis and treatment increased their mortality rate from endocarditis by 5%. After treatment of a relapse, the only possible outcomes were death and cure. We assumed a small risk (0.05%) of lethal adverse events for hospitalized patients (29,30) who were treated for endocarditis, regardless of the underlying condition. To determine quality-adjusted survival and costs for 200

September 1999

THE AMERICAN JOURNAL OF MEDICINE威

The prior probability for an individual patient depends on history, physical examination findings, and laboratory data (Table 3). The probability of endocarditis in patients with unexplained bacteremia is 4% to 40% (19 –21,37). Therefore, for our base-case analysis, we assumed that a patient with unexplained bacteremia had a prior probability of endocarditis of 20%. Patients with persistent bacteremia and a predisposing heart disease have a prior probability of endocarditis that is greater than 50%; patients with persistent bacteremia and a new regurgitant murmur have a prior probability of endocarditis near 100% (38,39). To account for these clinical scenarios, we evaluated prior probabilities of endocarditis between 0% and 100% in sensitivity analyses. Our base-case analysis evaluated the health and economic outcomes for a hypothetical cohort of 45-year-old patients with bacteremia, suspected endocarditis, and a prior probability of endocarditis of 20% (19 –21). In addition, we modeled several other clinical scenarios, including patients with suspected prosthetic valve endocarditis, injection drug users with fever, and patients in a setting in which transesophageal echocardiography was not available.

Quality of Life We adjusted for decrements in quality of life by assigning a utility to each state of health (1 for perfect health, 0 for death). We multiplied this utility by the number of years spent in each health state to calculate the total qualityadjusted life years (QALYs). To determine the quality ad-

Volume 107

Echocardiography for Suspected Endocarditis/Heidenreich et al

Table 1. Input Variables and Data Sources for Base Case and Range Tested in Sensitivity Analyses* Variable

Base Case

Range Tested

Test performance variables TEE sensitivity TEE specificity TTE sensitivity TTE specificity TEE sensitivity after a negative TTE

0.92 0.95 0.70 0.95 0.73

(0.75–0.95) (0.85–0.98) (0.60–0.85) (0.85–0.98) (0.5–0.92)

TEE specificity after a negative TTE

0.98

Probability that TTE quality equals TEE quality for detecting vegetations (%) Disease variables Endocarditis prevalence (%) Probability of in-hospital death from endocarditis (%) Probability of death from TEE (%) Probability of death from hospitalization for antibiotics (%) Probability of in-hospital death for bacteremic patients without endocarditis (%) Increased probability of death from a delayed diagnosis of endocarditis (%) Probability of relapse with standard therapy for endocarditis (%) Probability of cure of endocarditis when treatment is only for bacteremia (10–14 days) (%) Age (years)

Age-specific mortality

5

20 15 0.01 0.05

(5,9,11,12,22–25) (9,11,12,22,23) (5,9,11,12,22–25) (9,11,12,22,23) Assumes all vegetations observed on TTE are seen with TEE (9,22,23,54) (0.95–1.00) Assumes 2 of 3 false-positive TEE studies would be eliminated by screening first with TTE (0–20) Estimated

(0–100) (3–25) (0.005–0.1) (0.005–0.1)

5

(1–60)

(19,20) (25,26) (13,55) Estimated based on adverse events during hospitalization (29,30) (19,31)

5

(1–15)

Estimated

2.5

(0–10)

(26–28)

20

(0–50)

Estimated

45

(25–85)

1.1

(1–5)

Matched to average age in recent reports of endocarditis outcome (26,27) Based on Gompertz survival function of US life table data (16,33)† (21,31,32)

3.3

(1–5)

(26,27)

20

(0–40)

(26,27,36)

4

(0–10)

2

(0–10)

Declines to 0 by 20 years (27,34,35) (27,34,35)

Increases with age

Relative risk of death per year compared with the general population for survivors of uncomplicated bacteremia (without endocarditis)‡ Relative risk of death per year after endocarditis compared with the general population Probability of valve replacement during initial endocarditis episode (%) Probability of valve replacement each year after endocarditis (%) Probability of second valve replacement each year (%)

Data Source

* Base-case: 45-year-old patient with bacteremia of unclear origin. Values for echocardiography test characteristics are based on means of study results giving more weight to large studies and those most recently published. All probabilities and costs are annual unless otherwise indicated. † 1 ⫺ EXP (⫺.000185*EXP [0.07*(age ⫹ year)] where e is the base of the natural logarithm. ‡ The bacteremic patient is assumed to have a nonlethal underlying disorder. TEE ⫽ transesophageal echocardiography; TTE ⫽ transthoracic echocardiography.

justment for patients surviving endocarditis, we assumed a 15% prevalence of congestive heart failure in these patients (2) and used corresponding published utilities (40) obtained with the time-tradeoff technique (41). We assumed that the average quality of life decreased after valve replacement (to a utility of 0.9). We varied these qualityof-life assumptions in sensitivity analysis (range 0 to 1).

Costs Our analysis includes the direct costs of echocardiography, of endocarditis treatment, and of all medical treatments after discharge (Table 1). We adjusted all costs to 1996 dollars using the gross-domestic-product deflator (Bureau of Labor Statistics). We based the cost of endocarditis treatment on Medicare payments for DRG 403

September 1999

THE AMERICAN JOURNAL OF MEDICINE威

Volume 107 201

Echocardiography for Suspected Endocarditis/Heidenreich et al

Table 2. Costs (in 1996 US Dollars) and Quality of Life Used in Models Variable (units)

Base Case

Quality of life adjustment (utilities) Well, no endocarditis Survived endocarditis, no valve replacement

Survived endocarditis, valve replacement Death Toll for endocarditis (days)

Toll for inappropriate endocarditis treatment (days) Toll for valve replacement (days)

Range Tested*

1.0 0.95

(0.25–1.00) (0.25–1.00)

0.90 0 10

(0.25–1.00) 0 (0–300)

5

(0–150)

14

(0–300)

Toll for transesophageal echocardiography (hours) Time horizon (years) Costs Transesophageal echocardiography ($)

450

(400–1,000)

Transthoracic echocardiography ($)

250

(100–400)

Hospitalization, endocarditis ($)†

0.5

(0–4)

50

15,200

(6,500–40,000)

6,050

(2,500–10,000)

Hospitalization, false-positive diagnosis of endocarditis ($)

11,500

(6,500–40,000)

Hospitalization, valve replacement ($)

22,300

(10,000–35,000)

Hospitalization, bacteremia without endocarditis ($)

Additional annual cost of health care after endocarditis ($)

Additional annual cost for health care after valve replacement ($)

Other variables Cost-effectiveness threshold (dollars/QALY) ($) Annual discount rate (%)

600

(0–2,000)

$ 1,000

(0–4,000)

50,000 3

September 1999

THE AMERICAN JOURNAL OF MEDICINE威

Volume 107

Defined Estimated using time-tradeoff for congestive heart failure (40) assuming a 20% prevalence after endocarditis (2) Estimated Defined Estimated based on length of disability, accounts for loss of work because of symptoms and endocarditis treatment Estimated based on hospital stay, accounts for loss of work because of antibiotic treatment Estimated based on length of disability, accounts for loss of work because of surgery Estimated based on length of procedure Covers lifetime of 45-year-old Estimated based on data from northern California hospitals and Medicare reimbursement Estimated based on data from northern California hospitals and Medicare reimbursement Estimated based on Medicare reimbursement for DRG 403 including 20% valve replacement plus physician services Estimated based on Medicare reimbursement for DRG 420 plus physician services Estimated based on Medicare reimbursement for DRG 403 plus physician services Estimated based on Medicare payments for DRG 105 plus physician services Estimated based on 1 additional physician visit, 5% need for heart failure treatment, and periodic echocardiography (34) Estimated based on 4 physician visits, echocardiography, laboratory, and anticoagulation costs, and heart failure medications in 20% (56)

(25,000–500,000)

(44)

(0–5)

(18)

* Range tested in sensitivity analysis. † Cost for initial treatment. If relapse occurs, then additional hospitalization cost is incurred. DRG ⫽ diagnostic-related group. 202

Data Source

Echocardiography for Suspected Endocarditis/Heidenreich et al

Table 3. Patient Characteristics and the Probability of Endocarditis Patient Presentation Firm alternate diagnosis or resolution of endocarditis syndrome within 4 days (Duke criteria: “rejected endocarditis”) Gram-negative bacteremia with clear noncardiac source of infection Unexplained bacteremia Catheter-associated Staphylococcus aureus bacteremia Streptococcus viridans bacteremia Admission with fever and recent injection drug use Bacteremia and recent injection drug use Persistently positive blood cultures and predisposing heart disease (Von Reyn case definition of “possible endocarditis”) Persistently positive blood cultures and a new regurgitant murmur (Von Reyn criteria: “probable endocarditis”; Duke criteria: “definite endocarditis”)

Probability of Endocarditis (%)*

Source

⬍2

(4)

⬍2

(57)

5–40 6 (3–9) 14 (6–22) 13 (7–19) 31 (19–44) ⬎50

(19–21,58) (58) (57) (37) (37) (39)

⬎90

(39)

* Estimate (95% confidence interval) or range if the source is from more than one study.

(endocarditis treatment), DRG 104 (valve replacement with catheterization), DRG 105 (valve replacement without catheterization), and Medicare payments for physician services. Patients with a false-positive diagnosis were assigned the costs for DRG 403 and physician payments. We included age-adjusted health expenditures appropriate for the general population (42,43). In addition, patients who survived endocarditis had additional costs to account for extra physician visits, prophylaxis for dental procedures, follow-up echocardiography, and anticoagulation if needed. We determined costs for echocardiography using Medicare payments for two-dimensional transthoracic, transesophageal, and color Doppler echocardiography (CPT codes 93307, 93312, and 93325). We also evaluated echocardiography costs using the Transition Systems, Inc, (TSI; Boston, Massachusetts) cost accounting system. Because these estimates were different than Medicare reimbursements, we used echocardiography costs intermediate between Medicare and TSI for the base-case analysis. Relying on Medicare costs for the diagnostic strategies would bias the results in favor of the transesophageal echocardiography strategy.

Optimal Strategy Determination We used a threshold of cost-effectiveness of $50,000 per QALY gained to determine the optimal strategy (44). Because no single threshold is universally accepted, we tested thresholds between $25,000 and $500,000 per QALY in sensitivity analysis. We ordered each strategy by increasing QALYs and chose the strategy with the highest benefit that had an incremental cost-effectiveness that was less than the cost-effectiveness threshold ($50,000/ QALY) when compared with all other strategies.

RESULTS For our base-case analysis of a 45-year-old man with a 20% prior probability of endocarditis, use of transesophageal echocardiography, either alone or after transthoracic echocardiography, provided the greatest quality-adjusted life expectancy (Table 4). Both of these strategies provided 9 additional quality-adjusted days of life compared with using a transthoracic study alone to determine treatment. Compared with the use of transesophageal echocardiography alone, treatment of all patients for endocarditis decreased quality-adjusted life expectancy by 2 days; treatment only for bacteremia decreased qualityadjusted life expectancy by 41 days. If no adjustments were made for quality of life, treatment of all patients for endocarditis resulted in a gain of 2 days of life expectancy compared with the use of transesophageal echocardiography. The strategy that used transesophageal echocardiography alone was the least expensive ($18 per patient less than the transthoracic strategy). The sequential strategies cost $230 to $250 more per patient than the transthoracic strategy, and the strategies that treated everyone for endocarditis or bacteremia were considerably more expensive. Although the strategy that treated all patients for endocarditis provided an incremental gain in nonquality-adjusted life expectancy, this strategy cost $630,000 per life-year gained compared with the use of transesophageal echocardiography.

Sensitivity Analyses We tested the robustness of our base-case findings by varying each of the assumptions in Tables 1 and 2 over the ranges listed. The strategy of performing a transesophageal echocardiogram remained the optimal strategy

September 1999

THE AMERICAN JOURNAL OF MEDICINE威

Volume 107 203

204

September 1999

THE AMERICAN JOURNAL OF MEDICINE威

Volume 107

82 0 0

100 0 0

0

800

55

54

40

40

Excess Treatment for Endocarditis per 1,000 Patients

* Health and economic outcomes are discounted at a 3% annual rate. † Transesophageal echocardiography provides more life-years at a lower cost. TEE ⫽ transesophageal echocardiography; TTE ⫽ transthoracic echocardiography.

78

100

TTE; if negative and not high quality, then TEE TTE; if negative, then TEE Endocarditis treatment, no echocardiography Bacteremia treatment, no echocardiography

0

100

100

0

Percent Who Undergo TEE

TTE

TEE

Strategy

Percent Who Undergo TTE

200

0

16

15

60

16

Endocarditis Cases Missed per 1,000 Patients

Lifetime Costs

⬍0.1 days ⬍0.1 days 2 days ⫺47 days

⬍⫺0.1 days ⫺2 days ⫺41 days

⫺11 days

$550

$3,400

$252

$231

$18

18.1 years $50,030 Incremental change relative to TEE strategy

Life Expectancy (No Quality Adjustment)

⬍⫺0.1 days

⫺9 days

18.0 years

Quality-adjusted Life Expectancy

Table 4. Outcomes and Costs after Different Diagnostic Strategies for the Base-Case 45-Year-Old Patient with Bacteremia and 20% Prior Probability of Endocarditis*

TEE dominates†

$630,000

⬎$1,000,000

TEE dominates† ⬎$1,000,000

—

Incremental Cost per Life-year Gained (vs TEE)

Echocardiography for Suspected Endocarditis/Heidenreich et al

Echocardiography for Suspected Endocarditis/Heidenreich et al

Figure 3. This figure displays the relations between the prior probability of endocarditis, the increase in sensitivity with transesophageal echocardiography, and the appropriate diagnostic strategy. At low probabilities of endocarditis, a strategy of not using echocardiography and treating only for bacteremia (black line below 2%) is optimal (defined as providing the greatest QALYs at an incremental cost-effectiveness of less than $50,000 per QALY gained compared with all other strategies). Above this range, using a transesophageal study alone (TEE) is optimal up to moderate to high probabilities of disease (60%). At greater probabilities of disease, treatment should be provided regardless of the echocardiographic findings (Treat). If there is only a small gain in sensitivity with TEE (less than 10%), transthoracic imaging (TTE) is optimal if the prior probability is greater than 2% but less than 15% to 20%. See Table 3 for a description of the Duke and Von Reyn criteria. The base-case sensitivity of TTE for endocarditis is 70%.

throughout the range of every variable tested except as noted below. Prior probability of endocarditis. The transesophageal strategy was optimal for the range (4% to 60%) commonly observed in clinical practice (Figure 3). At low probabilities of endocarditis (less than 2%), treating all patients for bacteremia without performing echocardiography is optimal. This range would likely include patients with a fever and a clear source of infection (Table 3). For probabilities of endocarditis between 2% and 3%, the transthoracic strategy provided benefit over the bacteremia treatment strategy at a cost less than $50,000 per QALY gained. For high probabilities of endocarditis (greater than 60%), a strategy of treating all patients for endocarditis without relying on echocardiography for diagnosis is optimal. Sensitivity, specificity, and safety of echocardiography. The sensitivity, specificity, and safety of echocardiography may not be as good as that reported from specialized centers. Therefore, we evaluated a scenario in which the rates of detection of endocarditis were significantly lower both for transesophageal (sensitivity ⫽ 0.80) and transthoracic (sensitivity ⫽ 0.50) echocardiography, at the same specificity (0.95) as in our base case. In this scenario, transesophageal imaging remained the strategy of choice, but the threshold to treat patients without relying on echocardiography was lower (40%). Only if the test sensitivity was decreased, and the risk of death from a transesophageal study was much greater (1 in 1,000) than we

estimated in our base case (1 in 10,000), was the transesophageal strategy no longer optimal. Serial transthoracic studies have been suggested as an alternative to transesophageal echocardiography in patients with suspected endocarditis (45). We found that even if the sensitivity of two serial transthoracic studies was 90% (with a specificity of 95%), initial use of transesophageal echocardiogram was less expensive and provided more QALYs than did use of serial transthoracic studies. If transesophageal echocardiography is not available, transthoracic echocardiography should be used for patients with a prior probability of endocarditis of 2% to 35%. For greater probabilities of disease, treatment of endocarditis without echocardiography is optimal. Alternative clinical scenarios. In patients with prosthetic heart valves, the sensitivity of transthoracic echocardiography (35%) is markedly lower than that of transesophageal echocardiography (85%) for the detection of vegetations (23,24). These patients require repeat valve replacement more frequently (40%) and have a greater mortality rate (25%) than patients with native valve endocarditis (46). For analyses of patients with prosthetic heart valves, we made these changes to our model but kept all remaining values from the base case. Our results suggest that transesophageal echocardiography should be used for patients with suspected prosthetic valve endocarditis if the prior probability of endocarditis is 3% to 48%. If the prior probability of endocarditis is greater than 48%, then treatment for endocarditis should be pro-

September 1999

THE AMERICAN JOURNAL OF MEDICINE威

Volume 107 205

Echocardiography for Suspected Endocarditis/Heidenreich et al

vided without relying on echocardiography for diagnosis. If the prior probability of endocarditis is less than 2%, then a strategy of treating for bacteremia (without echocardiography) is preferred. Many injection drug users have tricuspid valve endocarditis, and it is unclear whether transesophageal echocardiography is more sensitive than transthoracic echocardiography for this condition (47). To address this scenario, we assumed that the test characteristics of transesophageal echocardiography for right-sided endocarditis equaled the base-case transthoracic values. Our analyses indicate that if the aortic or mitral valve is involved in at least 20% of patients with endocarditis, as suggested by previous studies (37,48), then the transesophageal strategy remains the optimal diagnostic strategy. Costs. Our results were insensitive to the costs of echocardiography. The transesophageal strategy remained cost-effective when compared with all other strategies, even when transesophageal costs were as great as $1,100 or transthoracic echocardiography costs were as low as $50. Our findings were also insensitive to the initial treatment and follow-up costs of bacteremia, endocarditis, valve replacement, and other medical care.

DISCUSSION We evaluated the cost-effectiveness of several diagnostic strategies for a patient with suspected endocarditis. Our base-case analysis (a 45-year-old patient with unexplained bacteremia) indicates that the use of transesophageal echocardiography alone provides the greatest QALYs at the lowest cost when compared with the use of a transthoracic study alone, use of a transthoracic study first followed by a transesophageal study if the initial study is negative, or treatment for endocarditis without echocardiography. This finding was largely insensitive to a wide range of model assumptions. Our analysis shows that the prior probability of endocarditis (based on history, physical examination, and laboratory data) is the most important factor in determining which strategy to choose (Figure 3). This finding is in agreement with previous work (15,49,50) and confirms that the diagnosis of endocarditis should not be made by echocardiography alone. Our findings suggest that at low probabilities of endocarditis (less than 2%), a strategy of treating only for bacteremia without echocardiography is appropriate. Because treatment for endocarditis is not without risk of adverse events (25,26), a limited treatment strategy provides the greatest benefit in these patients. Conversely, if the probability of endocarditis is high (greater than 60%, as in a patient with persistent bacteremia, embolic phenomenon, or a new regurgitant murmur) 206

September 1999

THE AMERICAN JOURNAL OF MEDICINE威

(39), then the patient should be treated for endocarditis regardless of the results of echocardiography. For patients with a small to moderate probability of endocarditis, a strategy of using transesophageal echocardiography provides the greatest QALYs at a cost that is within the range for commonly accepted health interventions (less than $50,000 per QALY gained) (44). Accepted interventions in the United States include renal dialysis ($35,000 per year of life saved, in 1990 US dollars) (51) and screening for hypertension ($12,000 to $43,000 per year of life saved, in 1993 US dollars) (52), although no cost-effectiveness threshold is accepted universally. Our algorithm for the use of echocardiography is complementary to the Duke (4) and Von Reyn (3) diagnostic criteria. Although these case definitions allow standardized reporting and accurate comparisons of treatment outcome (39), in many patients it is not possible to exclude or definitively diagnose endocarditis. For example, 37% of patients in the Duke series were classified as possible endocarditis. For these patients, the clinician must still decide if treatment is worthwhile. Furthermore, the Duke criteria, which incorporate echocardiography, do not state which form of echocardiography to use or when echocardiography is not needed. Our study addresses these issues and shows that transesophageal echocardiography should be used in many patients. Previous studies of the diagnostic accuracy of echocardiography have been criticized for being nonrepresentative of routine clinical practice (53). We therefore evaluated a scenario in which the sensitivity and specificity of transesophageal and transthoracic echocardiography were inferior to the published data, and in which the risk of a transesophageal study was 10-fold greater than our base-case estimate. Our results were most sensitive to the risk of death from a transesophageal study. Therefore, if a clinician has concerns regarding the safety of a transesophageal procedure, use of a sequential strategy or reliance on a transthoracic study is appropriate when the prior probability of endocarditis is moderate (2% to 35%). If there is a strong clinical suspicion (greater than 35% prior probability), then empiric treatment for endocarditis is optimal. In summary, our study found that the initial use of transesophageal echocardiography is the optimal diagnostic strategy for most patients with suspected endocarditis. Patients with low probabilities of endocarditis (less than 2%) should not be evaluated with echocardiography, whereas those with high probabilities (greater than 60%) should be treated for endocarditis regardless of echocardiographic results.

REFERENCES 1. Sande M, Kaye D, Root R. Endocarditis. New York: Churchill Livingstone, 1984. 2. Weinstein L, Brusch J. Infective Endocarditis. New York: Oxford University Press, 1996.

Volume 107

Echocardiography for Suspected Endocarditis/Heidenreich et al 3. Von Reyn CF, Levy BS, Arbeit RD, Friedland G, Crumpacker CS. Infective endocarditis: an analysis based on strict case definitions. Ann Intern Med. 1981;94:505–518. 4. Durack DT, Lukes AS, Bright DK. New criteria for diagnosis of infective endocarditis: utilization of specific echocardiographic findings. Duke Endocarditis Service. Am J Med. 1994;96:200 –209. 5. Dittrich HC, DeMaria AN. Is transesophageal echocardiography necessary in all patients with suspected endocarditis? Cardiol Rev. 1993;1:228 –238. 6. Durack DT. Prevention of infective endocarditis. NEJM. 1995;332: 38 – 44. 7. Schiller N. Clinical decision making in patients with endocarditis: the role of echocardiography. In: Otto C, ed. The Practice of Clinical Echocardiography. Philadelphia: WB Saunders, 1997:389 – 403. 8. Pedersen WR, Walker M, Olson JD, et al. Value of transesophageal echocardiography as an adjunct to transthoracic echocardiography in evaluation of native and prosthetic valve endocarditis. Chest. 1991;100:351–356. 9. Birmingham GD, Rahko PS, Ballantyne FD. Improved detection of infective endocarditis with transesophageal echocardiography. Am Heart J. 1992;123:774 –781. 10. Sochowski RA, Chan KL. Implication of negative results on a monoplane transesophageal echocardiographic study in patients with suspected infective endocarditis. J Am Coll Cardiol. 1993;21: 216 –221. 11. Shapiro SM, Young E, De GS, et al. Transesophageal echocardiography in diagnosis of infective endocarditis. Chest. 1994;105:377– 382. 12. Shively BK, Gurule FT, Roldan CA, Leggett JH, Schiller NB. Diagnostic value of transesophageal compared with transthoracic echocardiography in infective endocarditis. J Am Coll Cardiol 1991;18: 391–397. 13. Daniel WG, Erbel R, Kasper W, et al. Safety of transesophageal echocardiography. A multicenter survey of 10,419 examinations. Circulation. 1991;83:817– 821. 14. Khandheria G, Freeman W, Sinak L. Infective endocarditis. In: Freeman W, Seward J, Khandheria B, et al, eds. Transesophageal Echocardiography. Boston: Little, Brown and Company, 1990:307– 335. 15. Lindner JR, Case RA, Dent JM, Abbott RD, Scheld WM, Kaul S. Diagnostic value of echocardiography in suspected endocarditis. An evaluation based on the pretest probability of disease. Circulation. 1996;93:730 –736. 16. Beck JR, Pauker SG. The Markov process in medical prognosis. Med Decis Making. 1983;3:419 – 458. 17. Sonnenberg FA, Beck JR. Markov models in medical decision making: a practical guide. Med Decis Making. 1993;13:322–338. 18. Lipscomb J, Weinstein MC, Torrance GW. Time preference. In: Gold MR, Siegel JE, Russell LB, et al, eds. Cost-Effectiveness in Health and Medicine, vol 1. New York: Oxford University Press, 1996:214 –235. 19. Amit M, Pitlik SD, Samra Z, Konisberger H, Drucker M, Leibovici L. Bacteremia in patients without known underlying disorders. Scand J Infect Dis. 1994;26:605– 609. 20. Lautenschlager S, Herzog C, Zimmerli W. Course and outcome of bacteremia due to Staphylococcus aureus: evaluation of different clinical case definitions. Clin Infect Dis. 1993;16:567–573. 21. Rayner BL, Willcox PA. Community-acquired bacteraemia; a prospective survey of 239 cases. Q J Med. 1988;69:907–919. 22. Erbel R, Rohmann S, Drexler M, et al. Improved diagnostic value of echocardiography in patients with infective endocarditis by transoesophageal approach. A prospective study. Eur Heart J. 1988;9:43– 53. 23. Lowry RW, Zoghbi WA, Baker WB, Wray RA, Quinones MA. Clinical impact of transesophageal echocardiography in the diagnosis

24.

25.

26. 27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39. 40.

41. 42. 43. 44.

45.

and management of infective endocarditis. Am J Cardiol. 1994;73: 1089 –1091. Schulz R, Werner GS, Fuchs JB, et al. Clinical outcome and echocardiographic findings of native and prosthetic valve endocarditis in the 1990’s. Eur Heart J. 1996;17:281–288. Werner GS, Schulz R, Fuchs JB, et al. Infective endocarditis in the elderly in the era of transesophageal echocardiography: clinical features and prognosis compared with younger patients. Am J Med. 1996;100:90 –97. Delahaye F, Ecochard R, de Gevigney G, et al. The long term prognosis of infective endocarditis. Eur Heart J. 1995;16:48 –53. Tornos MP, Permanyer-Miralda G, Olona M, et al. Long-term complications of native valve infective endocarditis in non-addicts. A 15-year follow-up study. Ann Intern Med. 1992;117:567–572. Karchmer AW, Moellering R Jr, Maki DG, Swartz MN. Single-antibiotic therapy for streptococcal endocarditis. JAMA. 1979;241: 1801–1806. Brennan TA, Leape LL, Laird NM, et al. Incidence of adverse events and negligence in hospitalized patients. Results of the Harvard Medical Practice Study I. NEJM. 1991;324:370 –376. Leape LL, Brennan TA, Laird N, et al. The nature of adverse events in hospitalized patients. Results of the Harvard Medical Practice Study II. NEJM. 1991;324:377–384. Bates DW, Pruess KE, Lee TH. How bad are bacteremia and sepsis? Outcomes in a cohort with suspected bacteremia. Arch Intern Med. 1995;155:593–598. Leibovici L, Konisberger H, Pitlik SD, Samra Z, Drucker M. Bacteremia and fungemia of unknown origin in adults. Clin Infect Dis. 1992;14:436 – 443. Wilson DL. The analysis of survival (mortality) data: fitting Gompertz, Weibull, and logistic functions. Mech Ageing Dev. 1994;74:15– 33. Malquarti V, Saradarian W, Etienne J, Milon H, Delahaye JP. Prognosis of native valve infective endocarditis: a review of 253 cases. Eur Heart J. 1984;5:11–20. Verheul HA, van den Brink RB, van Vreeland T, Moulijn AC, Duren DR, Dunning AJ. Effects of changes in management of active infective endocarditis on outcome in a 25-year period. Am J Cardiol. 1993;72:682– 687. Hecht SR, Berger M. Right-sided endocarditis in intravenous drug users. Prognostic features in 102 episodes. Ann Intern Med. 1992; 117:560 –566. Weisse AB, Heller DR, Schimenti RJ, Montgomery RL, Kapila R. The febrile parenteral drug user: a prospective study in 121 patients. Am J Med. 1993;94:274 –280. Pelletier L, Petersdorf R. Infective endocarditis: a review of 125 cases from the University of Washington Hospitals, 1963–72. Medicine. 1977;56:287–313. Von Reyn CF, Arbeit RD. Case definitions for infective endocarditis. Am J Med. 1994;96:220 –222. Fryback DG, Dasbach EJ, Klein R, et al. The Beaver Dam Health Outcomes Study: initial catalog of health-state quality factors. Med Decis Making. 1993;13:89 –102. Sox H, Blatt M, Higgins M, Marton K. Medical Decision Making. Boston: Butterworth, 1988. Census US Bot. Statistical Abstract of the United States: 1996, 116 ed. Washington, DC, 1996. Hodgson T. Cigarette smoking and lifetime medical expenditures. Millbank Quarterly. 1992;70:81–125. Laupacis A, Feeny D, Detsky AS, Tugwell PX. Tentative guidelines for using clinical and economic evaluations revisited. Can Med Assoc J. 1993;148:927–979. Weisse AB, Campi JS, Zoneraich N. Limiting transesophageal echocardiography in valve endocarditis. N J Med. 1996;93:31–36.

September 1999

THE AMERICAN JOURNAL OF MEDICINE威

Volume 107 207

Echocardiography for Suspected Endocarditis/Heidenreich et al 46. Calderwood SB, Swinski LA, Karchmer AW, Waternaux CM, Buckley MJ. Prosthetic valve endocarditis. Analysis of factors affecting outcome of therapy. J Thorac Cardiovasc Surg. 1986;92:776 –783. 47. San Roman JA, Vilacosta I, Zamorano JL, Almeria C, Sanchez-Harguindey L. Transesophageal echocardiography in right-sided endocarditis. J Am Coll Cardiol. 1993;21:1226 –1230. 48. Levine DP, Crane LR, Zervos MJ. Bacteremia in narcotic addicts at the Detroit Medical Center. II. Infectious endocarditis: a prospective comparative study. Rev Infect Dis. 1986;8:374 –396. 49. Yvorchuk KJ, Chan KL. Application of transthoracic and transesophageal echocardiography in the diagnosis and management of infective endocarditis. J Am Soc Echocardiogr. 1994;7:294 –308. 50. Otto C, Pearlman A. Textbook of Clinical Echocardiography. Philadelphia: WB Saunders Company, 1995. 51. Roberts SD, Maxwell DR, Gross TL. Cost-effective care of end-stage renal disease: a billion dollar question. Ann Intern Med. 1980;243– 248. 52. Littenberg B, Garber AM, Sox HJ. Screening for hypertension. Ann Intern Med. 1990;112:192–202.

208

September 1999

THE AMERICAN JOURNAL OF MEDICINE威

53. Chambers HF. Transesophageal echocardiography in endocarditis. Chest. 1994;105:333–334. 54. Irani WN, Grayburn PA, Afridi I. A negative transthoracic echocardiogram obviates the need for transesophageal echocardiography in patients with suspected native valve active infective endocarditis. Am J Cardiol. 1996;78:101–103. 55. Stoddard MF, Longaker RA. The safety of transesophageal echocardiography in the elderly. Am Heart J. 1993;125:1358 –1362. 56. Suryapranata H, Roelandt J, Haalebos M, Degener J, Box E, Hugenholtz PG. Early cardiac valve replacement in infective endocarditis: a 10-year experience. Eur Heart J. 1987;8:464 – 470. 57. Weinstein MP, Towns ML, Quartey SM, et al. The clinical significance of positive blood cultures in the 1990s: a prospective comprehensive evaluation of the microbiology, epidemiology, and outcome of bacteremia and fungemia in adults. Clin Infect Dis. 1997; 24:584 – 602. 58. Ing M, Baddour L, Bayer A. Bacteremia and infective endocarditis: pathogenesis, diagnosis, and complications. In: Crossley K, Archer G, eds. The Staphylococci in Human Disease. New York: Churchill Livingstone, 1997:341.

Volume 107