Article, Cardiology

A clinical pathway for heart failure reduces admissions from the ED without increasing congestion in the ED

a b s t r a c t

Background: A Multidisciplinary team at a major academic medical center established an Acutely decompensated heart failure clinical pathway (ADHFCP) program to reduce inpatient readmission rates among patients with heart failure which, among several interventions, included an immediate consultation from a cardiologist familiar with an ADHFCP patient when the patient presented at the Emergency Department (ED). This study analyzed how that program impacted utilization of services in the ED and its subsequent effect on rates of admission from the ED and on disposition times.

Methods: ADHFCP inpatient visits were retrospectively risk stratified and matched with non-program inpatient visits to create a control group. A Cox survival model analyzed the ADHFCP’s impact on patients’ likelihood to visit the ED. Multivariable ANOVA evaluated the impact of the program on the patients’ likelihood of being admitted when pre- senting at the ED. The ADHFCP’s impact on bed-to-disposition time in the ED was evaluated by Wilcoxon’s rank-sum test, as were doses of diuretics administered in the ED.

Results: The survival analysis showed no impact of the ADHFCP on patients’ likelihood of visiting the ED, but ADHFCP patients presenting to the ED were 13.1 (95% CI: 3.6-22.6) percentage points less likely to be ad- mitted. There was no difference in bed-to-disposition times, but ADHFCP patients received diuretics more frequently and at higher doses.

Conclusions: Improved communication between cardiologists and ED physicians through the establishment of an explicit pathway to coordinate the care of heart failure patients may decrease that population’s like- lihood of admission without increasing ED disposition times.

(C) 2017

Introduction

The Hospital readmission Reduction Program (HRRP), instituted as a part of the Affordable Care Act of 2010, incentivized hospitals to reduce their readmission rates for a certain subset of inpatient di- agnoses by reducing reimbursements to hospitals with high read- mission rates [1]. One of the HRRP’s diagnoses of interest was congestive heart failure, which clearly merits focus for readmission reduction. Approximately 550,000 individuals are diagnosed with heart failure a year [2] and it has a national prevalence of over 5.8 million with a 5-year mortality rate approaching 50% [3]. A perfect

* Corresponding author at: The University of Chicago Pritzker School of Medicine, 924 E 57th St Suite 104, Chicago, IL 60637, United States.

E-mail address: [email protected] (T.B. Wassermann).

storm of high prevalence and high acuity, it is the most frequent cause of readmission within 30 days of discharge. Elixhauser and Steiner [4] estimate that in 2010 alone there were 209,017 total 30-day readmissions for heart failure. A review of medical records from 2008 and 2009 found that 74% of heart failure patients who used hospital-based care within 30 days of discharge were admitted as inpatients rather than being treated and released at the ED, giving heart failure the highest rate of readmission among the twenty listed discharge conditions [5].

Since a landmark article in 1995 initially recommended the use of Critical Pathways programs to improve efficiency and quality in high volume settings [6], Clinical Pathways programs have been widely adopted throughout Europe and North America [7]. While Clinical Pathways programs are generally associated with reduced in-hospital complications, reduced lengths of hospitalization, and

https://doi.org/10.1016/j.ajem.2017.12.012 0735-6757/(C) 2017

decreased Costs of care without increasing mortality or readmissions, the question of whether Clinical Pathways programs tend to shift costs or workload to other parts of the healthcare system remains large- ly unassessed [8]. In light of the demonstrated benefits of Clinical Path- ways programs, the multidisciplinary team at a major medical center

Table 1

Criteria for risk stratification.

Major risk criteria Minor risk criteria

Positive Booth Odds Ratio (N 0.5)a Two or more hospitalizations in the

past year

launched a risk-stratified, multidisciplinary Acutely Decompensated Heart Failure Clinical Pathway (ADHFCP) program in 2015 [9].

An analysis showed that the program reduced 30-day readmissions

Three or more hospitalizations in the past 6 months

Gait speed:

b0.5 m/s for age b 65

b0.4 m/s for age N 65

in a manner that was cost-effective given the HRRP’s incentives. The program is described in detail in the reference listed above, but one fea- ture of the ADHFCP program was the stipulation that as soon as an en- rolled patient arrived at the emergency department (ED), the ED provider would consult a cardiologist familiar with the case, and the cardiologist would provide recommendations that could be im- plemented in the ED. Informally, providers involved in the program postulated that the ED consultation was the most significant element of the program in reducing inpatient readmissions. However, this suggested an important question that was unaddressed in the initial analysis of the program: Did the decreased readmissions come at the expense of more congestion in the ED? In other words, did the pro- gram simply move patients’ treatment from Inpatient floors to the ED?

The question of how a program designed to reduce readmissions im- pacts the ED is not entirely novel. It has been suggested that incentives to reduce readmissions could prompt a hospital to treat patients entire- ly in the ED and send them home [10]. Such a strategy would save money, but may not be in the best interest of patient care, clinician workloads, or Hospital operations. The ED has been targeted in interven- tions aimed at reducing hospital readmissions [11,12], including specif- ically for patients with heart failure [13]. Interventions aimed at reducing patients’ time in the ED have been tested to ensure they do not increase readmissions after ED discharge [14]. However, the impacts of programs aiming at reducing readmissions on utilization of ED ser- vices have been poorly studied.

A search of all articles indexed in PubMed with the MeSH terms “Emergency Service, Hospital” and “Patient Readmission” from Jan- uary, 2011 (prior to the implementation of the HRRP) to January, 2016 yielded only two articles describing interventions aimed to reduce hospital readmissions and also reporting their impact on ED utilization. One study reported that an in-home intervention addressing social and emotional issues that exacerbate chronic medical conditions reduced both readmission rates and ED visits among Elderly people of low socioeconomic status [15]. A program that aimed to reduce both hospital readmissions and ED visits among congestive heart failure patients did neither, though it did improve symptomatic measures [16]. Until now little research has been performed to assess how the need to minimize readmission may impact patient care and ED disposition times. Here we shall examine the impact of the ADHFCP program on patient utilization of services in the ED and discuss its subsequent effect on disposi- tion times.

Methods

Subjects

The ADHFCP program has previously been described in detail [9], but the program’s structure will be summarized briefly here. Investi- gators enrolled patients into the ADHFCP at a major academic medi- cal center during the year 2015. To be eligible, a patient had to have a primary diagnosis of Acutely Decompensated Heart Failure (ADHF), be admitted to the Cardiology or Hospitalist service, and be on a par- ticipating hospital floor. Patients were assigned to floors solely based on bed availability and not based on any clinical or demographic characteristics, and patients received care from the same clinical teams regardless of whether they were on a participating floor or a

Patient requires placement but refuses Creatinine N 3 and not on dialysis

Active psychiatric condition Active street drug use

All variables are binary, corresponding to 1 or 0. For example, if a patient is taking Digoxin, a1 would be used.

a Booth Odds Ratio was developed in-house for the ADHFCP program in conjunction

with the Booth School of Business at the University of Chicago. It was calculated as follows: (2.72)^(-2.05 – 1.41 * (Digoxin_on_admission) + 0.207 * (Medicaid) + 0.299 * (urgent or emergency hospitalization) + 0.693 * (Prescribed N 9 medications at admission) +

0.424 * (sodium_below_130) + 0.166 * (Hemoglobin_below_10) + 0.320 *

(systolic_below_125) + 0.0918 * (heart rate_not_between_60_100)).

non-participating floor. Patients with a left ventricular assist device , an orthotropic heart transplant (OHT), or patients planning to receive an LVAD or OHT during the study period were excluded. The diagnosis of ADHF was made clinically with no requisite labora- tory or radiographic findings. Risk Criteria were constructed to pre- dict a patient’s likelihood of inpatient readmission, and they are shown in Table 1. Patients meeting one or more Major Risk Criterion or at least three Minor Risk Criteria were stratified as High Risk, those meeting one or two Minor Risk Criteria were considered Intermedi- ate Risk, and those meeting no Risk Criteria were considered Low Risk.

Inpatients at the UCMC diagnosed with Acutely Decompensated Heart Failure without LVAD or OHT

Low Risk: Intermediate High Risk:

No Risk Criteria Risk: Any Major or

1-2 Minor Risk 3+ Minor Risk Criteria Criteria

As inpatients: patient education, consultations from pharmacist, dietician, social worker, Physical therapy

After discharge: discharge summary sent to next provider, follow-up call from RN at 3 days, cardiology appointment within 7 days, PCP appointment within 14 days, 24/7 phone access to cardiologist

If they present at ED: Immediate consultation given to ED docs by cardiologist familiar with patient’s case

Extra weekly phone call from RN for four weeks after discharge

Additional cardiology apt at 14-21 days

Fig. 1. ADHFCP program structure.

Once a patient was enrolled in the ADHFCP, all future cardiology care was delivered according to the ADHFCP until the patient was no longer eligible. The program also included a cardiology appoint- ment seven days after discharge, a primary care provider appoint- ment 14 days after discharge, telephone follow-up at 3, 7, 14, and 21 days post-discharge, and additional visits at the discretion of the clinician. The care provided to ADHFCP patients is summarized in the algorithm in Fig. 1. As discussed above, the care algorithm includ- ed an immediate cardiology consultation once an ADHFCP patient presented in the ED. This consultation was provided by the cardiolo- gist overseeing the ADHFCP program, who discussed each patient’s case with the multidisciplinary team before discharge. Occasionally this cardiologist would consult with the patient’s primary cardiolo- gist, social work, Case management, or pharmacy if more detail was needed, but the staff cardiologist was the primary point of contact for the ED’s consultation.

To create a control group, each ADHFCP inpatient visit was ret- rospectively matched with an inpatient visit with a primary diag- nosis of ADHF and no LVAD or OHT in 2014 or 2015 that was not treated according the clinical pathway either because it was in a non-participating floor, outside the ADHFCP program timeframe, or on a different service. Visits were matched based on a variety of demographic and health-related factors (see Table 2), and the ED utilization of these ADHFCP and control patients was examined after their index inpatient stay. The control group used for the ini- tial analysis that demonstrated that the ADHFCP reduced inpatient readmissions was also used in this investigation of how the pro- gram impacted the ED.

Data sources

Data on patient encounters were obtained from the hospital’s elec- tronic medical record system, EPIC (Epic Systems Corporation, Verona, WI). Data were compiled and analyzed using Stata IC 11 (StataCorp LP, College Station, TX). This study and its handling of patient data were approved by the parent institution’s Institutional Review Board, approval number IRB16-0252.

Survival analysis

In order to assess the program’s impact on patients’ likelihood to visit the ED, Kaplan-Meier survival curves were constructed to document ED-free survival time. Patients were considered at risk for an event following their index admission or following any ED visit, an ED visit was counted as an event, and patients were no longer considered at risk for an event when they died or were scheduled for an LVAD or OHT. Patients were followed from their index admission to the end of that calendar year (either 12/31/ 2014 or 12/31/2015). An Andersen-Gill Cox model was used to perform survival analyses accounting for the possibility of multiple failure times [17].

Analysis of ED utilization

Utilization of ED services was considered only for visits in 2015 to minimize variation in providers and other exogenous factors. ED visits were considered from a patient’s index admission or from 1/1/2015

Table 2

Demographics of ADHFCP patients and control ADHF patients.

ADHF control patients (n = 342)

ADHF clinical pathway patients (n = 316)

Percentage

-or-

Mean Median (25th75th percentile)

Percentage

-or-

Mean Median (25th75th percentile)

Demographics

Age at index admission

65.0 64.7 (54.976.8)

65.5 66.4 (56.276.5)

Male

51.2%

50.9%

Black race

81.3%

82.6%

Health status

Hispanic ethnicity Household income

ED visits in previous year Inpatient visits in past year

Diabetes

3.5%

$44,541 $35,969 ($34,529$46,975)

1.7 1 (03)

1.3 1 (02)

32.2%

3.5%

$43,381 $35,969 ($33,059$46,975)

1.3 0 (02)

0.9 0 (01)

27.8%

Hypertension

80.1%

75.3%

Dyslipidemia

31.6%

26.3%

MI and/or CAD

38.9%

38.0%

Atrial fibrillation

32.2%

27.2%

COPD

13.2%

13.9%

Prescribed beta blocker

34.5%

37.3%

Prescribed Loop diuretic

68.1%

69.9%

Prescribed statin

30.1%

36.1%

Prescribed antiplatelet

32.5%

36.7%

Prescribed warfarin

19.0%

16.1%

Risk tier

Prescribed ACE or ARB Heart rate at discharge Systolic BP at discharge Diastolic BP at discharge Body mass index

eGFR at discharge LDL at discharge HDL at discharge

Total cholesterol at discharge Platelet count at discharge NT-proBNP at discharge Hemoglobin at discharge Left ventric. ejection fraction

Low risk

47.7%

81.4 80.5 (7289)

115.9 113 (101129)

65.3 64 (5573)

32.1 29.6 (25.536.0)

48.8 47.2 (27.966.6)

77.4 73 (61.2589)

46.3 43 (3753)

144.8 139 (126.25162.75)

227.4 211.5 (168270.75)

11,802 5028 (1892.2512,623.75)

11.2 10.9 (9.22512.7)

36.6 35 (3535)

22.8%

47.5%

81.3 81 (7191)

117.6 115 (102.75128.25)

66.3 65 (5674)

32.2 30.3 (25.436.4)

46.1 44.0 (28.063.0)

79.8 77 (62.7593)

46.6 42 (34.7556)

148.0 145.5 (124167.25)

233.5 225 (172280)

13,136 6011 (2141.2513,854.5)

11.2 11.1 (9.412.825)

37.1 35 (3535)

24.7%

intermediate risk

37.1%

38.9%

High risk

40.1%

36.4%

(whichever was later) until 12/31/2015. ED visits with chief complaints likely unrelated to heart failure were disregarded.1 An Analysis of Vari- ance (ANOVA) was performed to determine the impact of the ADHFCP program on the likelihood of admission from the ED controlling for other quantifiable factors. Mortality of ADHFCP patients and control pa- tients within 30 and 90 days of an ED visit was analyzed using either the chi squared test or the Fisher’s exact test, both for all potentially heart failure-related ED visits and for all visits from which the patient was discharged rather than admitted. The time from a patient being placed in an ED bed to the patient receiving a disposition (admit, dis- charge, or transfer) was determined and, since the bed-to-disposition times were non-normally distributed according to the Shapiro-Wilk test, the times of ADHFCP patients and those of control patients were compared using the Mann-Whitney ranked sum test. The same test was used to compare the time from ED bed to receipt of diuretics, the time from ED bed to ordering of a chest X-ray, the time from ED bed to ordering a BNP, and the total doses of loop diuretics received by ADHFCP and con- trol patients who received those services in the ED. Results of these anal- yses are presented below.

The Bonferroni correction was applied to each category of analysis separately. For the four mortality comparisons, a p of 0.013 was consid- ered significant. For the two survival analyses, a p of 0.025 was considered significant. For three tests of likelihood to receive treatments in the ED and three tests of time to treatment in the ED, a p-value of 0.017 was con- sidered significant. The single ANOVA, the single test of time to disposi- tion, and the single test of diuretic dose were each measured at a p of 0.05.

Results

Recruitment

The study recruited 316 participants who had a combined total of 392 inpatient stays (1-5 inpatient stays per person) meeting the criteria for the ADHFCP program during 2015. As described above, each inpatient stay was matched with an inpatient stay with a primary diagnosis of ADHF in 2014 or 2015 that was not treated according to the clinical path- way. Matching was done on the basis of demographic factors and various markers of health status, all of which are displayed in Table 2. This yielded a control group of 342 patients with a combined total of 392 inpatient stays (1-3 inpatient stays per person). The patients in the ADHFCP pro- gram were risk stratified according to the algorithm described above and treated according to the Clinical Pathway described above. Character- istics of the two populations are shown below in Table 2.

ED-free survival

Two survival curves were constructed: one comparing ADHFCP pa- tients to control patients and another comparing patients in the three

1 Classified as potentially related to heart failure and included in the analysis: Abnormal test results, AICD problem, altered mental status, anasarca, anemia, asthma, blood pres- sure problem, body aches, cardiac arrest, chest pain, congestive heart failure, COPD exac- erbation, cough, coughing up blood, difficulty breathing, dizziness, fatigue, headache, leg pain, lethargic, Lower extremity edema, medical evaluation, multiple complaints, nausea, near syncope, not feeling well, other, palpitations, respiratory distress, shortness of breath, swelling, syncope, weakness, and wheezing.

Classified as likely unrelated to heart failure and excluded from the analysis: abnormal labs, abdominal pain, abscess, allergic reaction, assault and battery, ataxia, back pain, bleeding fistula, bleeding problem, blood in urine, Blood sugar problem, call back, cast problem, cellulitis, chills, constipation, decreased appetite, dialysis problem, diarrhea, dislocated shoulder, distended abdomen, drainage, dysphagia, EKG changes, epistaxis, ex- tremity pain, eye problem, facial droop, facial numbness, facial pain, facial swelling, fall, fe- ver, flank pain, flu like symptoms, foreign body in throat, hand pain, head injury, hiccups, hip pain, ingestion, inpatient admission, insomnia, itching, IV access, joint pain, knee pain, left sided numbness, malfunctioning dialysis access, medication request, Muscle spasms, neck pain, pain control, PICC line complication, possible stroke, rash, Rectal bleeding, rectal problem, rib injury, right sided weakness, seizure, stoma problem, suicidal ideation, su- ture/staple removal, toe pain, Vaginal bleeding, vaginal complaint, vomiting, vomiting blood, and wrist pain.

Fig. 2. ED-free survival by treatment group and risk tiers.

risk tiers. There was no statistically significant difference between ADHFCP patients and control patients (p = 0.037, threshold for signifi- cance set at 0.025 by the Bonferroni correction). There was a statistically significant difference seen between patients in the three risk tiers (model p N 0.00001), with patients in the High Risk being likely to visit the ED before patients in the Intermediate Risk tier (p = 0.0001) and patients in the Intermediate Risk tier being likely to visit the ED be- fore patients in the Low Risk tier (p = 0.0049). These results are displayed in Fig. 2, and median ED-free survival is shown in Table 3.

Likelihood of admission from ED

All 2015 ED visits subsequent to at least one control or ADHFCP inpa- tient stay were examined. This yielded 757 total ED visits, including 486 ED visits from control patients and 271 ED visits from AHDFCP visits. As noted in the survival analysis, control patients were not more likely to visit the ED, but there were more control visits because most controls

Table 3

Days without visiting ED.

Median

95% conf. interval

ADHFCP

54.0

42.9

67.1

Control

43.8

36.1

50.5

Low risk

100.0

61.0

137.9

Intermed.

59.0

50.2

69.3

High risk

33.7

30.2

41.0

Total

47.2

42.1

53.2

time to disposition“>Table 4

ED visits in the study population.

ADHFCP

patients

Control patients

Total

only the ED visits from which a patient was discharged from the ED rather than admitted, again there was no statistically significant rela- tionship: 30-day mortality following these ED visits was zero of 70 for ADHFCP patients and one of 111 for control patients (p = 1.0 by Fisher’s

All ED visits Not admitted 121 209 330

Admitted 150 277 427

Total 271 486 757

Potentially ADHF-related ED visits Not admitted

81

124

205

Admitted

94

210

304

Total

175

334

509

(186 of 342) were recruited in 2014 and therefore had ED visits from all of 2015 considered, while the median AHDFCP patient was recruited on 8/18/2015, meaning that many of the ADHFCP patients had only a few months in which their ED visits were counted. Of these 757 total visits, 509 presented with chief complaints classified as potentially related to heart failure, and of those 509 visits, 304 resulted in inpatient admission while the remaining 205 did not. Table 4 summarizes these results.

Multivariable ANVOA determined that controlling for other factors, patients in the ADHFCP program were less likely to be admitted as inpa- tients than control patients after arriving at the ED with a chief com- plaint potentially related to heart failure. The model estimated that ADHFCP patients were 13.1 percentage points less likely to be admitted from the ED than control patients (p = 0.0069). Other variables deter- mined to be predictive included an Emergency Severity Index (the Emergency Severity Index is a five-level triage acuity rating system that provides stratification of patients into five groups from 1 (most ur- gent) to 5 (least urgent) on the basis of acuity and resource needs [18]) triage score of b 3 (36.7% increase in likelihood of admission, p b 0.0001) and frequent use of the ED (patients with more than three ED visits in the year preceding their enrollment in the program had a 28.3% de- crease in likelihood of admission from any given visit, p = 0.005). Other variables included in the model were not statistically significant predictors of the patient’s likelihood of being admitted from the ED. Table 5 summarizes these results.

The ADHFCP program was not associated with any changes in 30-

day or 90-day mortality, either among all ED visits with a chief com- plaint potentially related to heart failure or among all such ED visits that resulted in a discharge. For all heart-failure-related ED visits, six of 175 ADHFCP visits were associated with mortality within 30 days compared to 13 of 334 control patients (p = 0.80 by chi squared). Sim- ilarly, 90-day mortality showed no association with program participa- tion, with 10 of 175 ED visits by ADHFCP patients and 21 of 334 ED visits by control patients associated with mortality within 90-days. Looking at

exact test) and 90-day mortality was one of 70 ADHFCP patients and one of 111 ADHFCP patients (p = 1.0).

Time to disposition

Among patients who visited the ED with potentially ADHF-related chief complaints, there was no statistically significant difference in the time between the patient being placed in an ED bed and the patient re- ceiving a disposition (admit, discharge, or transfer). For patients in the ADHFCP program, the median time to disposition was 213.5 min, with an interquartile range of 140 to 312.5 min. For control patients, the me- dian time to disposition was 215 min, with an interquartile range of 149 to 323 min. These results are displayed in Fig. 3. The observed times were therefore slightly shorter in the ADHFCP group, but the difference was not statistically significant (p = 0.77).

Treatment in the ED

Once patients reached the ED with complaints potentially related to heart failure, treatment of ADHFCP patients and control patient differed slightly. Though control patients and ADHFCP patients were equally likely to get chest X-rays (p = 0.94) and equally likely to have their B-type Natriuretic Peptide levels measured in the lab (p = 0.023, with the threshold for significance at p = 0.017 by the Bonferroni correction), ADHFCP patients were more likely to have loop diuretics administered (p b 0.0001), as displayed in Table 6. Among patients who received chest X-rays, BNPs, and loop diuretics, there was no significant difference between ADHFCP pa- tients and control patients in the time that elapsed from the patient’s callback in the ED to the time the test was ordered (p = 0.35, p = 0.83, and p = 0.15, respectively). Among patients who received di- uretics in the ED, ADHFCP patients were likely to receive higher total doses according to the Wilcoxon rank-sum test (p = 0.019). Among control paitents, the highest total dose was 180 mg of equiv- alent IV furosemide and the highest dose of pathway patients was 360 mg of equivalent IV furosemide. These results are displayed in Fig. 4. So, while the time to initiate treatment and the total time to disposition did not change, the nature of the treatment changed slightly, with AHDFCP patients being more likely to receive diuretics and receive larger doses of diuretics than control patients.

Table 5

Predictors of likelihood of inpatient admission from an ED visit with a potentially heart-failure-related chief complaint in the study population. Results of ANOVA model predicting likelihood of admission from ED visit

Coeff. Std. error T p N t 95% confidence interval

ADHFCP program participant

-0.131

0.048

2.71

0.007

-0.226

-0.036

BMI category

Overweight

-0.078

0.065

-1.2

0.229

-0.205

0.049

Obese

-0.022

0.059

-0.38

0.706

-0.138

0.093

Black race

-0.032

0.098

-0.32

0.748

-0.225

0.162

Hispanic ethnicity

0.091

0.200

0.45

0.651

-0.302

0.484

Male gender

0.035

0.045

0.78

0.434

-0.053

0.123

Household income b $33,534

-0.077

0.055

-1.4

0.163

-0.186

0.031

Admissions in year prior to program enrollment

1-2

0.110

0.090

1.22

0.221

-0.066

0.286

3+

0.098

0.100

0.98

0.328

-0.099

0.294

ED visits in year prior to program enrollment

1-3

-0.052

0.089

-0.59

0.557

-0.227

0.122

4+

-0.283

0.099

-2.85

0.005

-0.479

-0.088

Risk stratification

Intermed. risk

0.058

0.095

0.61

0.544

-0.129

0.244

High risk

0.156

0.100

1.55

0.121

-0.041

0.352

Emergency Severity Index: 1-2

0.367

0.072

5.06

0.000

0.225

0.510

Age N 65

0.013

0.046

0.28

0.779

-0.077

0.103

Constant

0.312

0.152

2.05

0.040

0.014

0.611

Fig. 3. Time from ED bed to disposition for study visits with a potentially heart-failure- related chief complaint.

* For discharged patients, diuretics given between disposition and ED exit are included

** 2 mg PO furosemide = 1 mg IV furosemide. 1 mg bumetanide = 40 mg IV furosemide

Fig. 4. Total diuretics given in the ED.

Discussion

This study confirms various aspects of the previously published anal- ysis of the ADHFCP program [9] and builds on these previous findings in a few important ways. As previously described, this analysis demon- strated that the ADHFCP did not effectively reduce patients’ visits to the ED, but it did reduce their likelihood of being admitted from the ED once they presented there. This analysis adds to previous conclu- sions by showing that patients’ risk stratification did significantly pre- dict their likelihood of visiting the ED. Once in the ED, this analysis showed that program patients did not spend more time than control pa- tients in the ED, but they were more likely to receive diuretic and they received diuretics at higher doses.

The ADHFCP program included many elements aimed at keeping heart failure patients healthier: additional patient education prior to discharge, a follow-up phone call from a nurse to make sure that the pa- tient had been able to get the proper medication, additional outpatient appointments, and direct phone access to a cardiologist all were aimed at ensuring that patients could stay well enough to avoid returning to the hospital. The survival analysis showed no evidence that the likeli- hood of a patient with ADHF returning to the ED was changed by the ADHFCP program. In contrast, a patient’s likelihood of visiting the ED was significantly impacted by the risk tier of the patient’s previous inpa- tient visit, which was assigned based on variables related to the patient’s health status and demographics. Taken together, this indicates that each patient’s likelihood of returning to the hospital to seek care was primarily determined by their prior health status and not signifi- cantly changed by the ADHFCP program.

Table 6 Frequency of relevant procedures among ED visits with a chief complaint potentially relat- ed to heart failure.

ADHFCP

patients

Control patients

Total

B-type Natriuretic Peptide

Not tested

45

119

164

Tested

130

215

345

Pearson’s Chi2 = 5.17, p = 0.023

Chest X-ray

Not performed

34

64

98

Performed

141

270

411

Pearson’s Chi2 = 0.005, p = 0.94

Loop diuretics

Not administered

81

231

312

Administered

94

102

196

Pearson’s Chi2 = 25.80, p b 0.0001

Totals 175 334 509

In contrast, the intervention changed the likelihood of patients being admitted after they presented at the ED with a chief complaint poten- tially related to heart failure. Controlling for factors related to patient demographics and health status, participation in the program resulted in a 13.1 percentage point decrease. Given that nearly 60% of overall ED visits with relevant chief complaints resulted in inpatient admission and that mean ED-free survival was only 47 days without a visit to the ED, this reduction implies a meaningful decrease in the number of inpa- tients admitted for repeat visits. Other factors found to be predictive are unsurprising: patients with more severe triage scores were more likely to be admitted and patients who present at the ED frequently (perhaps those who use the ED for primary care) were less likely to be admitted from any given visit. The fact that there was no increase in 30-day or 90- day mortality among discharged patients indicates that the program did not prompt Emergency doctors to discharge patients who were danger- ously ill.

Contrary to the expectations of the ED practitioners prior to the ini- tiation of the ADHFCP program, ADHFCP patients did not spend longer in the ED. While the difference between the time spent between ADHFCP patients and control patients was not statistically significant, patients in the program actually spent slightly less time than controls, suggesting that the program may have even saved time in the ED. The ADHFCP program required ED doctors to get an immediate consultation from a cardiologist familiar with the patient’s case and that consultation appears to have changed care. ADHFCP patients were more likely to re- ceive diuretics and, among patients who received diuretics, ADHFCP pa- tients received more diuretics with fewer receiving low doses and the maximum dose received being double the maximum dose received by control patients. These differences in care may have helped reduce the need to admit these patients.

An additional factor that may have reduced the likelihood of admis- sion for ADHFCP program patients was the direct communication be- tween the ED provider and the staff cardiologist frequently resulted in a guaranteed short-term clinic follow-up visit. Knowing these patients would be re-evaluated shortly after discharge increased the comfort- level of the ED providers in not admitting these patients. This discussion and follow-up planning was not available for control group patients. The combination of the differences in care outlined above and the avail- ability of quick cardiology follow-up suggest a mechanism for the re- duction in the likelihood of admission from the ED that was described in the previous analysis of the ADHFCP program [9].

This study had multiple limitations. Controls were matched based on similarity to ADHFCP inpatient stays rather than prospectively random- ized. Some controls were enrolled in 2014 and ED visits were examined

during 2015, which meant that some control ED visits happened over a year after the patient’s Initial diagnosis; though this was a significant limitation, it was seen as preferable to comparing control patients’ 2014 ED visits to ADHFCP patients’ 2015 ED visits given changes in ED staff and operations over the intervening year. This study is of a single readmission reduction program at a single medical center which might limit the study’s generalizability to other readmission reduction programs. Also, patient data were only gathered from a single hospital, which means that patient ED visits or readmissions at other medical centers would not be captured by this analysis.

Ultimately, there are multiple insights that can be gained from this study. First, EDs can be useful partners as hospitals devise strategies to reduce preventable readmissions. As the point of entry where patients seek care, EDs have a primary role in determining which patients will be admitted. Second, the ways that readmission reduction programs impact the workflow in the ED merit scholastic attention. These impacts on the ED affect both the hospital and the patient, and the effects ob- served are not always those initially expected. Third, in the context of a structured clinical pathway, closer coordination between ED medical staff and other specialists can reduce patients’ likelihood of readmission without increasing the time those patients spend in the ED.

Conclusions

A clinical pathway program for heart failure patients that included an immediate consultation with a cardiologist when a patient presented at the ED reduced readmissions. The reduction was primarily achieved by an increased likelihood for patients to be treated and discharged from the ED rather than by decreasing the frequency with which pa- tients sought care. This increased propensity to discharge from the ED did not come at the expense of increased patient mortality or increased time spent in the ED. These changes may have been mediated by in- creased use of diuretics in the ED and by the confidence that a quick follow-up with cardiology was available. Taken together, these results suggest a means by which EDs can help reduce readmissions without negative consequences to their own workflow or to the patients they serve.

Acknowledgements

No grant funding for this work was provided by any governmental, corporate, or outside not-for-profit institutions. The program was funded by an Innovation Grant to CET from the University of Chicago Medicine Center for healthcare delivery Science and Innovation (692797). TBW received funding from the Pritzker Summer Research

Program at the University of Chicago. The authors have no competing interests to declare.

References

  1. MedPAC. The hospital readmission penalty: how well is it working? [Internet]. Washington, DC: Medicare Payment Advisory Commission; 2015 March 24[Cited 2017 February 22]. Available from http://www.medpac.gov/blog/march-2015/ march-2015/2015/03/24/the-hospital-readmission-penalty-how-well-is-it- working.
  2. Aghababian RV. Acutely decompensated heart failure: opportunities to improve care and outcomes in the emergency department. Rev Cardiovasc Med 2001 Dec;3:S3-9.
  3. Roger VL. Epidemiology of heart failure. Circ Res 2013 Aug 30;113(6):646-59.
  4. Elixhauser A, Steiner C. Readmissions to U.S. hospitals by diagnosis, 2010. HCUP Sta- tistical Brief #153. Rockville, MD: Agency for Healthcare Research and Quality; April 2013 [Cited 2016 February 22]. Available from http://www.hcup-us.ahrq.gov/ reports/statbriefs/sb153.pdf.
  5. Vashi AA, Fox JP, Carr BG, D’Onofrio G, Pines JM, Ross JS, et al. Use of hospital-based acute care among patients recently discharged from the hospital. JAMA 2013 Jan 23; 309(4):364-71.
  6. Pearson SD, Goulart-Fisher D, Lee TH. Critical pathways as a strategy for improving care: problems and potential. Ann Intern Med 1995 Dec 15;123(12):941-8.
  7. Vanhaecht K, Bollmann M, Bower K, Gallagher C, Gardini A, Guezo J, et al. Prevalence and use of clinical pathways in 23 countries–an international survey by the European Pathway Association. J Integr Pathw 2006 Apr 1;10(1):28-34.
  8. Rotter T, Kinsman L, James E, Machotta A, Gothe H, Willis J. Clinical pathways: effects on professional practice, patient outcomes, length of stay and hospital costs. Cochrane Database Syst Rev 2010 Jun;3(3).
  9. Tabit CE, Coplan MJ, Spencer KT, Alcain CF, Spiegel T, Vohra AS, et al. Cardiology con- sultation in the emergency department reduces re-hospitalizations for low- socioeconomic patients with acute decompensated heart failure. Am J Med 2017 Sep;130(9):1112.e17-31.
  10. Joynt KE, Jha AKA. Path forward on Medicare readmissions. N Engl J Med 2013 Mar 28;368(13):1175-7.
  11. Ota KS, Lazkani M, Orme GJ. An automated alert system in a transitional care pro- gram to improve continuity of care in the emergency department: a strategy for re- ducing rehospitalizations. J Emerg Med 2014 Jan 8;47(2):213-4.
  12. Kocher KE, Kehoe B. ED may be key to cutting readmissions. Hosp Health Netw 2014 Jan;88(1):28-9.
  13. Hadi A, Hellman Y, Malik AS, Caccamo M, Gradus-Pizlo I, Kingery J. Reducing heart failure readmission rates with an emergency department early readmission alert. Am J Med 2014 Jan 11;127(11):1060-2.
  14. Cardin S, Afilalo M, Lang E, Collet JP, Colacone A, Tselios C, et al. Intervention to de- crease emergency department crowding: does it have an effect on return visits and hospital readmissions? Ann Emerg Med 2003 Feb 28;41(2):173-85.
  15. Prior MK, Bahret BA, Allen RI, Pasupuleti S. The efficacy of a senior outreach program in the reduction of hospital readmissions and emergency department visits among chronically ill seniors. Soc Work Health Care 2012 Apr 1;51(4):345-60.
  16. Wang SP, Lin LC, Lee CM, Wu SC. Effectiveness of a self-care program in improving symptom distress and quality of life in congestive heart failure patients: a prelimi- nary study. J Nurs Res 2011 Dec 1;19(4):257-66.
  17. Andersen PK, Gill RD. Cox’s regression model for counting processes: a large sample study. Ann Stat 1982 Dec 1:1100-20.
  18. Tanabe P, Gilboy N, Travers DA. Emergency Severity Index version 4: clarifying com- mon questions. J Emerg Nurs 2007 Apr;33(2):182-5.

Leave a Reply

Your email address will not be published. Required fields are marked *