Methods

We performed an evidence-based decision analysis to systematically combine empirical data on diagnostic accuracy, patient-relevant outcomes, and resource utilization from clinical trial results with unit costs from national databases. ^, We developed a decision-tree model ( Figure 1 ) to evaluate the clinical and cost outcomes of clinical assessment supported by NT-proBNP versus clinical assessment alone in the diagnosis and subsequent treatment of a cohort of dyspneic patients presenting to an ED.

Decision-tree model diagram with patient pathways after clinical assessment supported by N-terminal pro B-type Natriuretic peptide versus clinical assessment alone. AHF = acute heart failure; ED = emergency department. [+] all subtrees have the same structure and are therefore not shown for simplicity. Blue squares: decision nodes; green circles: chance nodes; red triangles: terminal (outcome) nodes. (Colored version of figure is available online.)

The primary clinical outcome of our decision analysis was serious adverse events occurring during the 6-month follow-up after initial discharge from the ED or hospital to home. This composite outcome consists of readmissions to ED, readmissions and late admissions to hospital, and urgent follow-up visits. Further secondary and disaggregated clinical outcomes included correct diagnoses and patient times in ED and hospital. The primary cost outcome was total direct medical cost. In addition, the disaggregated component costs for ED stay, hospitalization, and urgent follow-up are reported. Cost-effectiveness is reported as incremental cost-effectiveness ratio; that is, incremental costs divided by the incremental risk of serious adverse events (US dollars per serious adverse events avoided). If one of the strategies is dominant (i.e., improving clinical outcomes and saving costs), no ICER is calculated and incremental costs and clinical effects are reported separately.

The model was performed following the international guidelines of the Professional Society for Pharmacoeconomics and Outcomes Research (ISPOR) Society for Medical Decision Making (SMDM Joint Modeling Good Research Practices Task Force. We adopted the US Medicare perspective and chose a time horizon of 6 months, aligning with the duration of data collection of patients’ outcomes in ICON-RELOADED. In the NT-proBNP-supported strategy, the initial determination of the patient’s disease status was made based on an NT-proBNP test result, which could fall into1 of 3 categories: positive, negative, or inconclusive. Positive and negative diagnoses were respectively determined by rule-in (>450, >900, and >1,800 pg/ml for patients aged <50, 50–75, and >75) and rule-out thresholds (<300 pg/ml). Those thresholds were first identified in the ICON study and then prospectively validated in ICON-RELOADED. Test results that fell into the “gray zone” between the rule-out and rule-in thresholds were considered inconclusive; in our model, the diagnosis for those patients was based on clinical assessment. The clinical assessment alone strategy classified patients as positive or negative for AHF, without using NT-proBNP. In both strategies, patients with and without AHF were modeled separately based on the AHF prevalence.

Depending on the combination of patients’ true disease status (i.e., AHF or no AHF) and diagnostic assessment results, patients were stratified into 6 subgroups: true-positive, true-negative, false-positive, false-negative, inconclusive with AHF, and inconclusive without AHF. The last 2 categories apply only to the NT-proBNP-supported strategy and refer to the test result (inconclusive), not the ultimate diagnosis (positive or negative).

According to the medical resource utilization data collected in ICON-RELOADED, following the ED diagnosis, all patients could be hospitalized from the ED directly or discharged to home. Patients who were ruled out based on initial diagnostic assessment could still be hospitalized for non-AHF reasons (e.g., suspected acute pulmonary embolism). Patients who were initially hospitalized then discharged to home could be readmitted to ED or hospital directly, or not. Patients who were initially discharged to home could later be readmitted to ED or admitted to hospital directly, or not admitted at all. All patients could also undergo an urgent care or office visit requiring treatment. All patients were assumed to be at risk of all-cause death, which was not affected by the diagnostic strategy used.

AHF prevalence in dyspneic patients presenting to ED, as well as sensitivity and specificity of NT-proBNP with rule-in and rule-out thresholds, were obtained from ICON-RELOADED. However, diagnostic accuracy of clinical assessment alone was not available from this study. Therefore, a statistical prediction model was fitted to individual patient data from the study, using logistic regression with variables related to patients’ demographics, medical history, and clinical symptoms to estimate the accuracy of clinical assessment alone. (Additional details provided in Online Supplement Table S1.)

In the case of NT-proBNP, the probability of a true-positive was determined by the sensitivity of the rule-in test threshold; the probability of a false-negative by the inverse of the rule-out threshold sensitivity; the probability of a true-negative by the rule-out specificity; and the probability of a false-positive by the inverse of rule-in specificity. Patients not classified as positive or negative based on these probabilities were classified in the inconclusive category.

Probabilities of clinical events and resource utilization conditional on diagnostic category were based on data of US participants in the ICON-RELOADED study (n = 1,424). They were applied to NT-proBNP and clinical assessment equally, with a few exceptions to further stratify the inputs between the 2 strategies. These assumptions were based on the PRIDE study, reported in the Siebert et al model, and verified with clinical experts.

The exceptions mentioned above follow: (1) All false-positive patients were assumed to be hospitalized if the diagnosis was made by clinical assessment alone, as reported by Siebert et al. Specifically, if patients without AHF were falsely diagnosed as positive based on clinical assessment alone, they would not be discharged to home from the ED. In contrast, based on observation in ICON-RELOADED, 80% of false-positive patients were hospitalized if the diagnosis was based on NT-proBNP, consistent with PRIDE study. (2) The probability of initial hospitalization as well as hospitalization at a later time for true-negative patients diagnosed by clinical assessment alone was a third higher than for patients with diagnoses supported by NT-proBNP, reflecting confidence in NT-proBNP for ruling out AHF. We applied the difference between the strategies observed in PRIDE to the estimate obtained by ICON-RELOADED to derive the probability for hospitalization for true-negative patients diagnosed by clinical assessment alone. (3) Patients with false-positive diagnoses based on the NT-proBNP-supported strategy had an increased probability of receiving echocardiography (by relative 29%), as observed in PRIDE and reflective of clinical practice. (4) Patients with true-negative diagnosis by the NT-proBNP-supported strategy were assumed not to receive echocardiography, reflecting confidence in NT-proBNP for ruling out AHF. The impact of each assumption was tested in scenario analyses (See Online Supplement, Table S2).

No data to inform different risks of death depending on diagnostic status could be identified; no deaths were recorded in the ICON-RELOADED study. Therefore, the same risk of death, based on Siebert et al was applied for all patients regardless of the diagnostic method used or its result.

All costs are expressed in 2019 US dollars ($). Unit costs were based on the Medicare perspective and adjusted for inflation to 2019 prices when needed, utilizing adjustment factors from the annual average of the Bureau of Labor Statistics’ Consumer Price Index for medical care. Cost of ED care was based on the Healthcare Cost and Utilization Project (HCUP) Nationwide Emergency Department Sample database. Cost of hospitalization was sourced from the HCUP Hospital Inpatient National Statistics database. The cost associated with urgent care visits came from the Physicians’ Fee & Coding Guide. When patients were admitted to hospital directly from ED, it was assumed that their ED stays would be considered part of their hospitalization and only hospitalization costs would be incurred. The key inputs are presented in Tables 1 and 2 .

Table 2

Model inputs: parameters dependent on diagnostic category

Input	Value estimated within each diagnostic category						Source
	True-positive	False-negative	False-positive	True-negative	Inconclusive-acute heart failure	Inconclusive non-acute heart failure
Inputs related to clinical pathway
Probability of hospitalization after index emergency department admission in patients diagnosed with N-terminal pro B-type natriuretic peptide supported strategy	0.93	0.82	0.80	0.41	0.90	0.74	ICON-RELOADED
Probability of hospitalization after index emergency department admission in patients diagnosed with clinical assessment	0.93	0.82	1.00	0.55	0.90	0.74	ICON-RELOADED, Siebert et al.
Probability of late or repeated hospitalization in patients diagnosed with N-terminal pro B-type natriuretic peptide	0.40	0.35	0.31	0.14	0.23	0.28	ICON-RELOADED
Probability of late or repeated hospitalization in patients diagnosed with clinical assessment only	0.40	0.35	0.31	0.19	0.23	0.28	ICON-RELOADED, Siebert et al.
Probability of readmission to emergency department	0.38	0.38	0.41	0.31	0.33	0.31	ICON-RELOADED
Probability of urgent follow-up visit	0.39	0.50	0.30	0.23	0.46	0.38	ICON-RELOADED
Resource use
Average number of emergency department readmissions per readmitted patient	1.59	2.33	1.58	1.96	1.38	1.69	ICON-RELOADED
Average number of late/repeated hospital admissions per patient rehospitalized or admitted after discharge to home from emergency department	2.04	2.67	2.10	1.81	2.00	1.53	ICON-RELOADED
Proportion of late/repeated hospital admissions directly from emergency department	0.54	0.53	0.33	0.67	0.61	0.47	ICON-RELOADED
Proportion of patients receiving echocardiogram during index emergency department admission in patients diagnosed with N-terminal pro B-type natriuretic peptide-supported strategy	0.12	0.06	0.08	0.03	0.05	0.04	ICON-RELOADED, Siebert et al.
Proportion of patients of receiving echocardiogram during index emergency department admission in patients diagnosed with clinical assessment only	0.12	0.06	0.06	0.03	0.05	0.04	ICON-RELOADED
Proportion of patients of receiving echocardiogram during hospitalization from index emergency department admission	0.52	0.43	0.54	0.00	0.66	0.30	ICON-RELOADED
Probability of intensive care unit admission after index emergency department admission	0.44	0.36	0.30	0.11	0.37	0.17	ICON-RELOADED
Proportion of patients admitted to cardiology ward after index emergency department admission *	0.48	0.38	0.36	0.13	0.43	0.19	ICON-RELOADED
Average length of index stay in emergency department [hours]	12.32	13.26	9.75	10.31	13.09	11.07	ICON-RELOADED
Average length of emergency department stay on readmission [hours]	10.23	20.07	12.50	7.36	12.81	11.63	ICON-RELOADED
Average length of hospitalization from index emergency department admission [hours]	118.33	67.65	109.87	60.90	107.54	103.51	ICON-RELOADED
Average length of hospitalization on late or repeated admission [hours]	132.23	86.93	126.13	97.76	75.06	93.36	ICON-RELOADED

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

1. Comparison of Metoprolol Versus Carvedilol After Acute Myocardial Infarction
2. Frequency of Atrial Arrhythmia in Hospitalized Patients With COVID-19
3. Usefulness of Rhythm Monitoring Following Acute Ischemic Stroke
4. Sex-Specific Differences in Etiology and Prognosis in Patients With Significant Tricuspid Regurgitation

Stay updated, free articles. Join our Telegram channel

Join