Despite the widespread use of loop diuretics to treat acute decompensated heart failure (ADHF), robust data supporting their role and optimal dosing strategies are scarce. This analysis aimed to compare clinical outcomes of patients admitted with ADHF who received a diuretic dosing protocol with those who received the usual diuretic therapy. We performed an observational medical records review to compare the use of a nurse-driven diuretic dosing protocol with usual diuretic dosing for patients admitted with ADHF during a 1-year period. Using a propensity scoring model, comparisons were made between groups for total weight loss, length of stay (LOS), 30-day readmissions, in-hospital mortality, 30-day mortality, and acute kidney failure. Sixty-eight of the 596 patients admitted with ADHF during the study period received the diuretic protocol. Protocol use was associated with an additional 2.63-kg weight loss (P=.003) but a trend toward increased LOS compared with patients receiving usual care (P=.097). However, patients receiving the protocol had a significantly lower risk of 30-day readmission (odds ratio, 0.46, 95% confidence interval, 0.22–0.95). Protocol use was not associated with significant differences in kidney failure, inpatient mortality, or 30-day mortality. A diuretic dosing protocol for patients admitted with ADHF improves weight loss and may lower 30-day readmissions, at the cost of potentially increasing LOS.
More than 1 million hospitalizations for acute decompensated heart failure (ADHF) occur annually in the United States at a cost of more than $34 billion, and heart failure accounts for the largest Medicare expenditure for a single diagnosis-related group. Despite this significant use of resources, not all patients with ADHF receive the full spectrum of medications that improve survival.[2-5] Moreover, short-term readmissions, within 30 days of the index hospitalization, remain high because patients are often discharged with persistent symptoms and minimal weight loss.[1, 5] For more than 40 years, clinicians have used loop diuretics for volume removal in patients admitted with ADHF. Despite the widespread use of loop diuretics, robust data supporting their role are scarce and the optimal approach to the management of the patient with acute volume overload remains unclear.
Published studies have focused on comparing diuretic dose delivery mechanisms such as continuous infusions and bolus dosing.[7, 8] Although the method of dose delivery is an important issue in the management of ADHF, a number of critical factors must be considered to adequately remove volume. First, an appropriate dose of loop diuretic must be given to ensure sufficient and controlled diuresis. Second, the patient's urinary output in response to a given dose must be assessed at appropriate time intervals to determine subsequent diuretic dosing. Dosing intervals need to be based on pharmacokinetic knowledge of the various loop diuretics instead of arbitrary protocols and empiric physician decisions.
Furosemide and bumetanide have similar elimination half-lives of 1 to 2 hours and peak intravenous (IV) effects at 30 minutes.[10, 11] IV torsemide has a longer half-life of 3 to 4 hours.[10, 11] The quantity of urine output in response to a diuretic dose is best represented by a sigmoid curve. Above a specific dose threshold, further diuresis is not achieved. Based on the elimination half-life, dosing of furosemide or bumetanide should be every 4 to 8 hours in patients with volume overload and adequate blood pressure (BP). A patient's response to a diuretic dose is variable and is dependent on several factors including serum albumin level, renal and liver function, and diuretic resistance. Therefore, diuretic dosing should be individualized based on response.
Current guidelines published by the European Society of Cardiology, the American Heart Association (AHA), and the American College of Cardiology (ACC) do not include details about the optimal approach to using loop diuretics in patients with ADHF.[12, 13] To optimize treatment, we designed a loop diuretic protocol for patients admitted with ADHF that included frequent assessment and follow-up of dose response with appropriate titration. The objective of this study was to compare clinical outcomes of patients admitted with ADHF who received the diuretic protocol with those who received usual diuretic therapy.
We performed an observational medical records review to compare the use of a nurse-driven diuretic protocol with usual diuretic dosing for patients admitted with ADHF. Patients with ADHF admitted to a cardiac telemetry unit at Northwestern Memorial Hospital, an 894-bed academic cardiac transplant center, from September 1, 2010, to August 31, 2011, were included in the analysis. We compared clinical outcomes of study patients who were treated with the protocol with those who received usual diuretic treatment. The Northwestern University institutional review board approved this study.
A multidisciplinary team of hospitalists, cardiologists, nurses, and pharmacists developed the diuretic protocol as part of a quality-improvement project charged with reducing length of stay (LOS) and short-term readmissions for patients admitted with ADHF. The protocol was designed to allow nursing staff to quickly titrate furosemide or bumetanide (bolus or continuous infusion) with a goal of 100 to 250 mL of urine output per hour. The protocol provided safeguards (BP and urine output parameters and twice-daily electrolyte measurements) that were set to prevent harmful rapid diuresis. The team based the protocol on a template used for observation unit patients admitted with ADHF. IV torsemide was not available at the study institution. The protocol was designed to enhance nurse to physician communication, standardize diuretic use, and facilitate electrolyte assessment to ensure appropriate monitoring of patients taking high doses of loop diuretics. Patients were encouraged to be involved in documenting urine output and weights. Initial diuretic doses were recommended based on the patient's creatinine clearance and outpatient doses. IV single doses were given in equivalent milligrams (not daily total) of outpatient oral dosages of furosemide or double the single dose equivalent (not daily total) of outpatient oral bumetanide. Furosemide or bumetanide were given by IV push every 8 hours, with a dosage not to exceed a maximum of 80 mg of furosemide or 4 mg of bumetanide per dose. In patients who were not receiving outpatient oral diuretics, furosemide was initiated at 40 mg IV push every 8 hours if the patient's serum creatinine was <2 mg/dL and at 80 mg IV push every 8 hours if the patient's creatinine was >2 mg/dL. If the patient was taking oral torsemide as an outpatient, their diuretic was changed to IV furosemide equivalents by doubling the dose of torsemide. Decisions to initiate or terminate the protocol were dependent on physician assessment of the patient. The attending physician was required to assess the patient and their electrolytes 2 times daily during the use of the protocol. Decisions to use continuous infusion were based on the patient's response to IV bolus dosing or per attending physician discretion. The final protocol for bolus and continuous infusion diuretic dosing can be found in the Figure 1.
The protocol was implemented in August 2010 on the designated telemetry unit after physicians, nurses, and patient care technicians underwent intense training on its use. Attending physician hospitalists used the protocol for treatment of ADHF at their discretion.
The Enterprise Data Warehouse (EDW) at Northwestern University was used to identify all patients admitted to the designated telemetry unit with an International Classification of Diseases, 9th Revision (ICD-9) code for heart failure (428) who also received 2 doses of IV diuretics in a 24-hour period. The EDW allows use of Structured Query Language syntax to obtain demographic and clinical information derived from the electronic medical record systems at Northwestern and the Social Security Death Index.
The EDW search included patient demographic and clinical variables such as age, sex, use of cardiac drugs (eg, digoxin, angiotensin-converting enzyme inhibitors, angiotensin 2 receptor blockers [ARBs], β-blockers, nitrate and hydralazine, and mineralocorticoid receptor blockers), use of inotropes (eg, milrinone, dobutamine, and dopamine), total days receiving an IV loop diuretic, intensive care unit (ICU) transfer, attending physician at discharge, and ICD-9 comorbid conditions. The search also included parameters on admission: body mass index (BMI), systolic BP (SBP, mm Hg), serum brain natriuretic peptide (BNP, pg/mL), sodium (mEq/L), serum urea nitrogen (BUN, mg/dL), and creatinine (mg/dL). Patient outcome variables (also extracted from the EDW) included total weight loss, LOS, 30-day readmissions, in-hospital mortality, 30-day mortality, and changes in creatinine from admission to discharge. Electronic medical records were also reviewed to confirm that incomplete data was accurately reported. Patients with incomplete data were excluded from further analysis. All patients' electronic records derived from the EDW search were reviewed to record the left ventricular ejection fraction (EF) from echocardiograms performed during the index hospitalization or within 1 year. BNP levels from the EDW that were reported as over the upper limit were included as the exact number (ie, a BNP of >5100 pg/mg was recorded at 5100 pg/mL).
One author, an advanced heart failure and transplant cardiologist (RG), adjudicated results of the EDW search by reviewing a random sample of 25% of the electronic medical records to verify the admission diagnosis of ADHF. This author was blind to protocol use status at the time of chart adjudication. The cardiologist used the European Society of Cardiology and American College of Cardiology/American Heart Association[12, 16] guidelines for diagnosis of ADHF.
Patients who received the diuretic protocol were compared with those who received standard diuretic therapy. Comparisons were made between groups for total weight loss, LOS, 30-day readmissions, in-hospital mortality, 30-day mortality, and changes in creatinine.
Charlson scores[17, 18] were calculated to assess severity of illness at the time of admission for ADHF. The Charlson score is based on 17 chronic disease comorbidities derived from ICD-9 codes and predicts 1-year mortality for hospitalized medical patients.[17, 18] The admission and discharge creatinine values were used to calculate a change in creatine during the hospitalization. Patients who were discharged with a change in creatinine >0.3 mg/dL (acute kidney injury) were identified and rates compared between the protocol group and the usual care group.
We used the chi-square test, t test, or Wilcoxon rank-sum test to compare the demographic and clinical characteristics of patients who received the protocol with those who received usual diuretic care. Logistic regression was used to model the likelihood of using the protocol with patient age, sex, BMI, admission SBP, serum sodium, BUN, creatinine, and BNP, EF, Charlson score, and the attending of record's propensity to care for a patient in the protocol group. Predicted probabilities derived from this protocol use model were then used as a propensity score to help control for selection bias in our analysis of outcomes.
Multiple linear regression was used to test the association of protocol use with weight loss and hospital LOS days. Analyses for LOS were conducted using both actual days and a log transformation because LOS was not normally distributed. Only actual LOS results are presented for ease of interpretation of the model coefficients. The weight loss and LOS regression models tested the effect of protocol use controlled for patient age, sex, BMI, SBP, serum sodium, BUN, creatinine, and BNP, EF, Charlson score, ICU admission, inotrope use, and protocol use propensity score. We also controlled for the total number of diuretic days in the weight loss model. Multiple logistic regression was used to test the significance of protocol use on 30-day readmissions, in-hospital mortality, 30-day mortality, and creatinine increase >0.3 mg/dL controlled for patient age, sex, BMI, SBP, serum sodium, BUN, creatinine, BNP, EF, inotrope use, Charlson score, ICU admission, inotrope use, total diuretic days, and protocol use propensity score. Cardiac medications did not significantly influence results when added as covariates in the regression models. Therefore, they were excluded from final models because of the large number of covariates in the regression equation.
All statistical analyses were performed using SPSS version 20 (SPSS Inc, IBM, Chicago, IL).
The EDW search initially identified 650 patients with ADHF admitted to the designated telemetry unit during the study period. The final study sample of 596 patients remained after removing 54 patients (8%) who did not have BNP measurements, of whom 7% (5 of 73) were in the protocol group. A random sample of 25% (164 of 650) of charts was reviewed by the cardiologist who confirmed that 95% (156 of 164) of patients were correctly labeled as having ADHF by the EDW search. For 8 patients, the cardiologist reviewer identified potential errors in diagnosis due to the admitting physicians' belief that these patients had ADHF when they did not meet guideline criteria.
Demographic and clinical characteristics of patients who received the diuretic protocol vs usual care are included in Table 1. Female sex, BMI, SBP, serum BUN, creatinine, EF, ARB and inotrope use, and total diuretic days were statistically different between protocol and usual care groups (Table 1). The diuretic protocol was used for a mean of 5.45 days (standard deviation [SD]=8.94) of the 7.87 days (SD=9.98) of total diuretic use in the protocol group. Higher BMI (odds ratio [OR], 1.06; 95% confidence interval [CI], 1.02–1.09) and attending of record (OR, 7517.94; 95% CI, 826.61–68374.95) were associated with a higher likelihood of protocol use in the propensity model. Only 28 of 112 attending physicians cared for patients who received the protocol. No other factors predicted protocol use in this model. The propensity model had excellent discrimination (c statistic=0.87).
Table 1. Baseline Data: Comparison of Demographic and Clinical Variables in the Protocol and Usual Care Groups
Usual Care (n=528)
Abbreviations: BNP, brain natriuretic peptide; BUN, serum urea nitrogen; ICU, intensive care unit; IQR, interquartile range; SD, standard deviation. aValue on day of admission. bStatistically significant at P<.05.
Age, mean (SD), y
Female, No. (%)
Body mass index, mean (SD), kg/m2
Systolic blood pressure, mean (SD), mm Hga
Serum sodium, mean (SD), mEq/La
Serum BUN, median (IQR), No.a
Serum creatinine, median (IQR), mg/dLa
Serum BNP, median (IQR), pg/dLa
Ejection fraction, median (IQR),%
Ejection fraction <40%, No. (%)
Digoxin, No. (%)
Angiotensin-converting enzyme inhibitor, no. (%)
Angiotensin receptor blocker, No. (%)
β-Blocker, No. (%)
Nitrate and hydralazine, No. (%)
Mineralocorticoid receptor blocker
Inotrope use, No. (%)
Charlson comorbidity index, median (IQR), No.a
ICU admission, No. (%)
Total diuretic days, median (IQR)
Total weight loss was 7.8 kg (SD=10.7) in the protocol group vs 2.8 kg (SD=6.0) in the usual care group (P<.001). Patients who received the protocol remained in the hospital a mean of 9.9 days (SD, 12.5) compared with an LOS of 7.0 days (SD=6.9) for patients who received usual care (P=.005). Weight change and LOS model results are shown in Table 2. Protocol use was associated with a 2.6 kg additional weight lost (P=.003) after controlling for patient age, sex, BMI, SBP, serum sodium, BUN, creatinine, and BNP, EF, Charlson score, ICU admission, inotrope use, total diuretic days, and protocol use propensity score. Patients treated with the protocol trended toward 1.6 more days in the hospital compared with usual care (P=.097) after controlling for patient age, sex, BMI, SBP, serum sodium, BUN, creatinine, and BNP, EF, Charlson score, ICU admission, inotrope use, and protocol use propensity score. The log-transformed model decreased R2 from 0.29 to 0.24, and LOS still trended toward significance between groups (P=.068).
Table 2. Linear Regression of Protocol Use on Weight Loss and Length of Stay Controlled for Patient Factors
Change in Weight
Length of Stay
Abbreviations: BNP, brain natriuretic peptide; BUN, serum urea nitrogen; ICU, intensive care unit. aValue on day of admission. bStatistically significant at P<.05. Total diuretic days were used as a covariate in the weight loss model only.
Body mass index
Systolic blood pressure, mm Hga
Serum sodium, mEq/La
Serum BUN, mEq/La
Serum creatinine, mEq/La
Serum BNP, pg/dLa
Ejection fraction, %
Charlson comorbidity index, n
Total diuretic days
Protocol use propensity
Comparison of other outcomes can be found in Table 3. Lower EF (OR, 0.99; 95% CI, 0.97–0.998) predicted 30-day readmission in the regression model while lower EF (OR, 0.89; 95% CI, 0.81–0.98), higher age (OR, 1.14; 95% CI, 1.02–1.26), Charlson score (OR, 0.54; 95% CI, 0.30–0.95), and ICU transfer (OR, 44.47; 95% CI, 3.23–611.68) were significantly associated with increased inpatient mortality. Lower serum sodium was associated with increased 30-day mortality (OR, 0.89; 95% CI, 0.79–0.998). Female sex (OR, 0.57; 95% CI, 0.33–1.00), higher SBP (OR, 1.01; 95% CI, 1.00–1.02), admission creatinine (OR, 1.39; 95% CI, 1.06–1.82), Charlson score (OR, 1.17; 95% CI, 1.03–1.34), and total diuretic days (OR, 1.08; 95% CI, 1.02–1.14) were associated with a significant risk of developing kidney injury. Simple analysis showed a trend toward reduced readmissions at 30 days among patients who received the diuretic protocol (21.2%, 14 of 66) compared with those who received usual diuretic dosing (29.2%, 152 of 521; P=.18). After controlling for confounding variables, this association became significant (OR, 0.46; 95% CI, 0.22–0.95). Occurrence of inpatient mortality, 30-day mortality, or significant kidney failure did not differ between the protocol or control group.
Table 3. Patient Outcomes Using the Diuretic Protocol vs Usual Care Using Logistic Regression Controlling for Patient Factors
Usual Care (n=528)
Abbreviation: CI, confidence interval. aN=587 patients surviving to discharge. bStatistically significant at P<.05.
30-d readmission,a No. (%)
In-hospital mortality, No. (%)
30-d mortality,a No. (%)
Creatinine change >0.3 mg/dL, No. (%)
This study demonstrates that the use of a diuretic protocol for patients admitted to the hospital for ADHF increased volume removal as measured by weight loss and reduced 30-day readmissions at the potential cost of increased LOS without an increase in 30-day mortality rate. Additionally, a more aggressive approach to diuretic use with the protocol appeared to be safe because it did not increase inpatient mortality or the risk of acute renal failure at discharge. These findings are noteworthy given that those patients who received the protocol had a higher severity of illness based on Charlson score, greater inotrope use, higher BUN and creatinine levels, lower EF and BP, and greater diuretic days. Although retrospective and nonrandomized, we believe these data are provocative because we controlled for several known confounders of LOS, readmission, mortality, and acute renal failure including age, BP, serum sodium, BUN, creatinine, and BNP.[1, 5, 13, 21]
Recent studies indicate that the total amount of weight loss during admission for ADHF is not associated with improvements in recurrent heart failure or death.[22, 23] However, weight loss has been shown to improve shortness of breath and other signs of congestion on history and examination.[22, 23] We expected the more effective volume removal in the protocol group to result in lower LOS for patients. However, patients in the protocol group only received this approach for 70% of all diuretic days. Therefore, if these patients received the protocol during the entire duration of hospitalization, greater volume loss may have been achieved faster with a potential decrease in LOS. Additionally, patients in the protocol group were sicker (based on Charlson score) and may have been more volume overloaded, but we attempted to control for this finding in the regression model by adding diuretic days as a covariate.
The healthcare quality literature suggests that there may be a tradeoff in heart failure care between increased LOS and readmission rates and similarly between a reduction in readmission rates and an increase in 30-day mortality rates. A post hoc analysis of the Acute Study of Clinical Effectiveness of Nesiritide in Decompensated Heart Failure Trial (ASCEND-HF) study showed that for every extra day in the hospital, there was a significant drop in 30-day readmissions.[24, 25] This inverse relationship was also observed in a study of Medicare patients hospitalized with ADHF comparing 1993 with 2006. A study of patients hospitalized at US Veterans Affairs Hospitals also confirmed an inverse relationship between LOS and readmissions over time from 2002 to 2006. Many believe that the trade off between LOS and readmission rates is inevitable because the longer LOS allows care teams more time to educate, prepare, and treat the patient before the care transition. In the United States, the mean LOS is <6 days for patients admitted with ADHF while the 30-day readmission rate approaches 25%. The Initiation Management Pre-discharge Assessment of Carvedilol Heart Failure (IM-PACT HF) trial showed that >60% of patients admitted with ADHF were discharged without resolution of symptoms, resulting in persistent congestion, worsening symptoms, and high hospital readmission rates. Although LOS was higher than the national average in the protocol group, readmission rates were much lower. The longer LOS in the protocol group may have completely been accounted for from the “non-” protocol diuretic days (30%) or that the patients were sicker with higher Charlson scores. Impressively, patients in the protocol group were 56% less likely to be readmitted compared with the usual care group. Thirty-day readmissions rates in the usual care group (29.2%) were higher than the national average but consistent with the historical experience at our institution and served as the impetus to introduce this diuretic protocol. The absence of an increase in 30-day mortality rate is reassuring because recent analyses of data positioned on www.hospitalcompare.gov have repeatedly demonstrated higher mortality rates at those facilities with lower readmission rates and vice versa.[28, 30] This potential unintended consequence of a reduction in readmission rates is a growing concern.
If our diuretic protocol was applied to all patients in the usual care group, 41 fewer rehospitalizations would have occurred using the unadjusted relative risk of readmission demonstrated from the diuretic protocol (relative risk=0.27). Using the Centers for Medicare and Medicaid Services ADHF admission reimbursement of approximately $4500, this could result in significant savings, even accounting for the longer LOS. Our data supports the theory that more time spent removing fluid from patients with ADHF yields a decreased likelihood of 30-day readmission.[24, 26, 31]
Our study did not show a difference in mortality or increased renal failure. This is important because diuretic therapy augments vasoconstriction and elicits an exaggerated maladaptive neurohormonal response that may be harmful for patients. Studies show an association between high doses of diuretics and increased mortality and worsening renal function.[32, 33] Such findings may simply reflect the fact that patients who require high doses of diuretics have more advanced cardiac or renal disease. To the contrary, a recent analysis of the Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness (ESCAPE) trial data showed that higher doses of diuretics and increased hemoconcentration were associated with worsening renal function but an overall decrease in 180-day mortality. Although we did not directly evaluate the amount of diuretic given, the diuretic protocol was designed to give more aggressive but appropriate doses of diuretics than usual care and was based on the amount of volume removed. Further study is required to evaluate the effect of total daily dose of diuretics and clinical outcomes.
We believe our protocol showed favorable clinical outcomes for several reasons. Although not specifically studied, we believe its development and use drove culture change among physicians, nurses, and patients. The nurses and patients using the protocol were more involved in collection of urine and daily weights and thus forced to focus on this aspect of heart failure care. Greater patient involvement also allowed more opportunities for education surrounding the importance of monitoring input and output and daily weights. Additionally, the protocol improved teamwork and collaboration between nurses and physicians, which has been associated with reduced adverse drug events in hospitalized patients. Our approach of using an algorithm for treatment that directs titration of medication, monitors response, and clearly outlines communication channels to adjust doses allows for more effective medication administration is consistent with best practices for administration of medications with potentially dangerous adverse effects.
Many clinicians who treat hospitalized patients with ADHF prescribe a fixed daily diuretic dose and evaluate the natriuretic response based on the 24-hour urine output and weight loss. Our protocol was unique because it used continuation and titration of diuretics based on urine output, which may have allowed better volume removal within a narrow therapeutic window. In the Diuretic Optimization Strategies Evaluation (DOSE) trial, bolus dosing of loop diuretics was set at only twice daily for up to 48 hours before clinicians were allowed to adjust doses. Not surprisingly, more doses were changed at 48 hours in this strategy compared with continuous infusion. This twice-daily dosing strategy is not consistent with loop diuretic pharmacokinetics yet likely reflects a “real world” scenario. With optimal protocol dosing for loop diuretics, continuous infusion may not be necessary. In one study, Peacock and colleagues followed a multidisciplinary care protocol that included a diuretic protocol to treat patients admitted to an observation unit with ADHF. Using this protocol, 90-day HF readmission rates decreased by 64% (P=.007) with a trend toward decreased 90-day mortality. Although the multidisciplinary approach may have been the major contributor to these outcomes, the diuretic protocol allowed rapid achievement of “euvolemia” in an observation unit patient population with ADHF and importantly was not associated with harm.
Our analyses have several limitations. First, our data are nonrandomized and gleaned from retrospective review of medical records and thus our findings are not meant to be definitive but rather to serve as impetus for future study. We were unable to control for unknown factors that may have confounded our clinical outcomes, a known limitation of observational studies. Specifically, we do not know whether one study cohort had more immediate follow-up after discharge that may have affected readmission differences between groups. We were also unable to control for fluid restriction or intake that may have affected weight loss between groups. It is also possible that some patients admitted with ADHF were missed in the EDW search. Second, we used the discharge attending in the analysis. Many patients were in the hospital for longer than a week and patients may have had multiple attending physicians. Nevertheless, hospital quality metrics often attribute care to the attending physician at the time of discharge. We believe that the presence of multiple attendings did not affect the results because the overall service caring for the patient did not change and the attending of record was added as a confounder through the propensity modeling. Third, we excluded 8% of patients who did not have a BNP measurement. BNP has an important association with the outcomes measured and patients who did not have a BNP measured may not have had the same severity of illness as those who did. Fourth, this study was limited to a single center of attending physicians and patients, thus components of selection bias cannot be ruled out. Finally, we did not evaluate the specific aspects of the diuretic protocol that may have affected patient outcomes. Use of the diuretic protocol per se may have been an indicator of a broader evidence-based quality focus system of care exercised by the physicians using this protocol. A randomized trial using the diuretic protocol is needed to confirm the results of these data.
To date, clinicians have no clear evidence-based strategies for safely and rapidly improving congestion in patients with ADHF. Our data suggest how a diuretic protocol based on actual volume removal may improve weight loss and lower 30-day readmissions without the cost of increased mortality or acute renal failure. Given the importance of 30-day readmission rates, a more rigorous randomized prospective study is imminently needed to identify the best strategy for volume removal in patients with ADHF. Until more effective strategies for volume removal are available, appropriate diuretic administration is a major component in improving the care of patients with ADHF.
We acknowledge Dr Douglas Vaughan for his support and encouragement of this work. We thank Amanda Creden and Yael Simons for their assistance with chart review.
The authors report no specific funding in relation to this research. The authors have no conflicts of interest to disclose.