Association of Home Monitoring and Unanticipated Interstage Readmissions in Infants With Hypoplastic Left Heart Syndrome

Abstract

Background

The impact of home monitoring on unanticipated interstage readmissions in infants with hypoplastic left heart syndrome has not been previously studied. We sought to examine the association of our institution's Infant Single Ventricle Management and Monitoring Program (ISVMP) with readmission frequency, cumulative readmission days, and readmission illness severity and to identify patient‐level risk factors for readmission.

Methods and Results

We performed a retrospective single‐center cohort study comparing infants with hypoplastic left heart syndrome enrolled in ISVMP (December 2010–December 2019) to historical controls (January 2007–November 2010). The primary outcome was number of readmissions per interstage days. Secondary outcomes were cumulative interstage readmission days and occurrence of severe readmissions. Inverse probability weighted and multivariable generalized linear models were used to examine the association between ISVMP and the outcomes. We compared 198 infants in the ISVMP to 128 historical controls. Infants in the ISVMP had more than double the risk of interstage readmission compared with controls (adjusted incidence rate ratio, 2.38 [95% CI, 1.50–3.78]; P=0.0003). There was no difference in cumulative interstage readmission days (adjusted incidence rate ratio, 1.02 [95% CI, 0.69–1.50]; P=0.90); however, infants in the ISVMP were less likely to have severe readmissions (adjusted odds ratio, 0.28 [95% CI, 0.11–0.68]; P=0.005). Other factors independently associated with number of readmissions included residing closer to our center, younger gestational age, genetic syndrome, and discharge on exclusive enteral feeds.

Conclusions

Infants in the ISVMP had more frequent readmissions but comparable readmission days and fewer severe unanticipated readmissions. These findings suggest that home monitoring can reduce interstage morbidity without increasing readmission days.

Nonstandard Abbreviations and Acronyms

CHOP

Children's Hospital of Philadelphia

HLHS

hypoplastic left heart syndrome

IRR

incidence rate ratio

ISVMP

Infant Single Ventricle Management and Monitoring Program

S1P

stage 1 Norwood palliation

S2P

stage 2 palliation

Clinical Perspective.

What Is New?

• In infants with hypoplastic left heart syndrome, enrollment in interstage home monitoring was associated with an increased frequency of readmissions; however, cumulative readmission days were equivalent and readmissions less severe.

What Are the Clinical Implications?

• Families of infants with hypoplastic left heart syndrome should be counseled that most infants will have at least 1 readmission during the interstage period and should be counseled on any relevant patient risk factors.

• Vigilant interstage home monitoring can eliminate severe presentations of coarctation of the neo‐aorta.

Despite medical and surgical advances over the past 50 years, hypoplastic left heart syndrome (HLHS) continues to be one of the most lethal forms of congenital heart disease. The interstage period, defined as the time from stage 1 Norwood palliation (S1P) hospitalization discharge to the time of stage 2 palliation (S2P), is a critical time for infants with HLHS who are vulnerable to fluctuations in systemic and pulmonary blood flow. ¹ Historically, interstage mortality rates were as high as 15%. ¹ , ² Home monitoring, a concept first introduced at the Children's Hospital of Wisconsin and now widely adopted, has resulted in a significant reduction in interstage mortality. ¹ , ³ , ⁴ , ⁵ The Infant Single Ventricle Management and Monitoring Program (ISVMP) was introduced at the Children's Hospital of Philadelphia (CHOP) in December 2010. We previously reported a 29% lower predicted probability of death in the interstage period for an infant with HLHS enrolled in the ISVMP compared with a historical control. ⁶ When evaluating all infants in the ISVMP (not only those with HLHS), we found a significant increase in the percentage of infants with at least 1 interstage readmission. ¹ , ⁶ Other studies have also reported frequent readmissions in the era of home monitoring. ¹ , ⁷ , ⁸

It has been hypothesized that the reduction in interstage mortality associated with home monitoring may be partially attributed to an increase in unanticipated interstage readmissions preventing adverse out‐of‐hospital events. ¹ , ⁶ However, the association of home monitoring with the frequency and severity of readmissions has yet to be investigated. Likewise, patient‐level risk factors for all unanticipated readmissions, not exclusively those prompted by changes in physiologic parameters referred to as “red flag” events, have yet to be identified. ⁷ , ⁹

In this context, the primary objective of this study was to evaluate the association of ISVMP with unanticipated interstage readmission frequency, cumulative readmission days, and illness severity in infants with HLHS. The secondary objective was to identify patient‐level risk factors for unanticipated interstage readmissions and severe illness on presentation.

Methods

The study protocol was approved by the Institutional Review Board for the Protection of Human Subjects at CHOP (IRB 19–017193). The data, analytic methods, and study materials will not be made available to other researchers for purposes of reproducing the results or replicating the procedure. Informed consent was waived for this study.

Study Population

We conducted a retrospective cohort study of patients with HLHS and HLHS variants who underwent S1P from January 1, 2007, to December 31, 2019. The ISVMP cohort was composed of all infants who underwent S1P after December 1, 2010, and completed their interstage period before December 31, 2019. The historical control cohort was composed of all infants with HLHS who underwent an S1P between January 1, 2007, and November 30, 2010 (last date before the introduction of the ISVMP). Patients who underwent S1P in anticipation of a biventricular repair, those who died during the neonatal hospitalization, those who remained inpatient during the interstage period, and those who were discharged on palliative care were excluded. Given the variability in home monitoring programs, patients enrolled in outside hospital programs were also excluded.

Home Monitoring Program

All infants in the ISVMP cohort were discharged home with a pulse oximeter and weighing scale. The additional components of the ISVMP included (1) standardization of the neonatal discharge criteria and process; (2) standardization of parental education before neonatal discharge; (3) daily monitoring by parents, including daily oxygen saturation measurement, daily weight measurement, and feeding log; (4) home nursing visits documenting oxygen saturation, weight, mode of feeding, and formula type and amount; (5) weekly phone calls to parents by a dedicated ISVMP nurse practitioner; (6) involvement of a registered dietitian; (7) biweekly pediatrician visits; (8) biweekly cardiology visits with focused echocardiograms; (9) weekly review of patients by a dedicated ISVMP team; and (10) scheduling of standard S2P at 3 to 4 months of age. Standard criteria for remaining inpatient interstage included a continued need for inotropic support, respiratory support, or sedative medications and arrhythmias refractory to standard medical therapy. Other criteria were at the discretion of the ISVMP team. There was no dedicated ISVMP outpatient clinic; therefore, cardiology visits were scheduled with primary cardiologists within or outside our center. Before the introduction of the ISVMP, S1P hospitalization discharge criteria were determined by individual inpatient providers, and interstage monitoring was at the discretion of a patient's primary cardiologists.

Data Collection

Medical records were reviewed to abstract demographic and relevant medical and surgical variables from the S1P and S2P hospitalizations (Table S1). Readmission indications, diagnoses, and interventions, such as unplanned cardiac catheterizations and operations, were recorded. A health information exchange platform was queried for outside hospital readmissions. Echocardiographic variables were obtained by review of reports.

Outcome Measures

Primary Outcome

The primary outcome was number of hospital readmissions in the interstage period. The interstage period started on the date of discharge from the S1P admission and concluded on the date of S2P admission or death. Any admission in the interstage period except for a routine pre‐S2P catheterization or magnetic resonance imaging was characterized as an unanticipated interstage readmission.

Secondary Outcomes

Secondary outcomes included cumulative interstage readmission days and readmission illness severity.

Cumulative Interstage Readmission Days

Cumulative interstage readmission days was defined as the total number of days admitted to the hospital during the interstage period. The interstage period started on the date of discharge from the S1P admission and concluded on the date of S2P operation or death. Therefore, inpatient days between S2P admission and operation were also included in this analysis to capture total time at risk. Inpatient days attributed to routine pre‐S2P catheterization and magnetic resonance imaging were excluded.

Severe Readmission

A system was created to grade readmission illness severity as mild, moderate, or severe. Severity was determined by the interventions required during readmission in the context of our center's practice patterns (Table S2). Readmissions were assigned the severity level associated with the highest‐level intervention (Methods S1). Readmissions with no interventions (observational readmissions) were categorized as mild readmissions. For those who underwent S2P during readmission, interventions that occurred before but not after the operation were included. Interventions were abstracted from the medical record. Because patients could have >1 readmission, readmission severity was analyzed as a repeated‐measures outcome.

Statistical Analysis

Descriptive Statistics

Demographic and clinical characteristics for the ISVMP and control groups are summarized using median (interquartile range [IQR]) for continuous variables and frequency (percentage) for categorical variables. Patient characteristics were compared between the ISVMP and controls using the Wilcoxon–Mann–Whitney U test for continuous variables and χ² or Fisher's exact test for categorical variables.

Primary Outcome Analysis

To determine whether ISVMP was associated with a differential rate of interstage hospital readmissions compared with historical controls, we used weighted quasi‐Poisson regression models with an offset term for the duration of the interstage period. To account for differences in patient‐level characteristics between the ISVMP and control groups, Quasi‐Poisson regression models were weighted using stabilized inverse probability of treatment weights (Methods S2, Figure S1). Covariates used in the propensity score model were selected a priori on the basis of their hypothesized association as confounders (Table S3). We used robust standard errors to help mitigate any bias in the standard errors from using inverse probability weights in the analysis. Incidence rate ratios (IRRs) and 95% CIs are reported.

To verify that the distribution of each covariate was balanced across the ISVMP and control groups, we compared the covariate distribution before and after applying propensity score weights. Standardized mean differences were used to assess the balance of covariates in the pseudo‐population created using the weighted sample. We considered a standardized difference of <10% to be a negligible imbalance. ¹⁰

Secondary Outcome Analysis

Cumulative Interstage Readmission Days

To evaluate the relationship of ISVMP and cumulative interstage readmission days, we used multivariable negative binomial regression with the subset of the sample that had at least 1 readmission. We included an offset for the length of the interstage period in the model. This analysis controlled for confounding by including covariates (same as those in the inverse probability of treatment weights model) as fixed effects in the outcome model instead of using the inverse probability of treatment weights that were developed for the full sample. Adjusted IRR and 95% CIs are reported for the effect of ISVMP.

Severe Readmission

To compare the effect of ISVMP on the probability of severe readmission, we used readmission‐level data. Only patients with at least 1 readmission were included in this analysis. The adjusted effect of ISVMP on the likelihood of severe readmission was modeled using multivariable logistic regression with generalized estimating equations. Generalized estimating equation models were used with sandwich standard errors to account for the repeated admissions within some patients. Because there were only 29 severe readmissions, we adjusted for only a few key covariates associated with the outcome and could not control for all confounders. The covariates included in the model were presence of a genetic syndrome, stage 1 repeat surgical intervention, discharge from neonatal hospitalization on oxygen, and number of discharge medications. Odds ratios (ORs) and 95% CIs are reported for the effect of ISVMP program.

Sensitivity Analyses

We ran the primary and secondary outcome models with several patient exclusions to confirm that the conclusions reached from the main regression models were consistent across different conditions. The first sensitivity analysis excluded patients who were transferred to an outside hospital following S1P but before discharge home. For these patients, the interstage period was defined as the number of days between outside hospital discharge and the date of S2P admission or death. For patients with a missing outside hospital discharge date, the outside hospital length of stay was imputed using the median length of stay for those with known discharge dates. The second sensitivity analysis excluded those who died in the interstage period, and the third excluded both those who were transferred and those who died during the interstage period.

Additionally, to confirm the findings of the weighted analysis, we tested the relationship between ISVMP and rate of readmissions using an exact matching design (Methods S2).

Identification of Risk Factors for Rate of Readmission and Severe Readmission

We examined several a priori risk‐factor variables present at the time of S1P discharge to determine their association with the rate of unanticipated readmission. Controlling for enrollment in ISVMP, we ran a separate negative binomial regression for each patient characteristic to test the association of each risk factor with the rate of readmission. Risk factors with a P value <0.10 were included in a multivariable model to provide adjusted estimates of each risk factor on outcome. We also present patient‐level risk factors significantly associated with the rate of readmission identified using logistic regression with generalized estimating equations.

All statistical analyses used 2‐sided tests and were performed using α=0.05 (unless otherwise specified) for the threshold of statistical significance. SAS software version 9.4 (SAS Institute, Cary, NC) and R software version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria) were used.

Results

Patient Characteristics and Neonatal Hospitalization

There were 198 ISVMP infants and 128 historical controls (Figure 1). Overall, 60 ISVMP infants and 24 controls were excluded because they either died (32 ISVMP infants and 15 controls), were not discharged from the S1P hospitalization (12 ISVMP infants and 8 controls), were discharged on palliative care (1 ISVMP infant and 1 control), underwent biventricular repair (2 ISVMP infants), or were enrolled in an outside home monitoring program (13 ISVMP infants). One ISVMP infants and 21 control patients were transferred to an outside hospital before discharge home from the S1P admission.

Figure 1. Flow diagram for patient enrollment and outcomes.

Detailed flowchart to describe total neonates undergoing surgery, including which participants were included in analysis and outcomes. ISVMP indicates Infant Single‐Ventricle Management and Monitoring Program.

Before weighting, ISVMP infants had a greater gestational age and resided closer to CHOP compared with historical controls (Table 1). A greater proportion of ISVMP infants received a right ventricle to pulmonary artery conduit compared with a Blalock‐Thomas‐Taussig shunt (Table 2). At the time of discharge, ISVMP infants had a greater median weight‐for‐age Z score than controls. They were also more likely to be discharged home on digoxin and with a regimen of oral and enteral feeds compared with enteral feeds alone (Table 3).

Table 1.

Baseline and Demographic Characteristics

	ISVMP (n=198)	Historical (n=128)	P value
Male sex	123 (62.1)	83 (64.8)	0.65
Birth weight, kg	3.2 (2.9 to 3.6)	3.2 (2.9 to 3.5)	0.42
Birth WAZ	−0.15 (−0.96 to 0.61)	−0.14 (−0.94 to 0.45)	0.43
Gestational age, wks	39 (38 to 39)	38 (37 to 39)	0.007
Prematurity (<37 wks)	11 (5.6)	15 (11.7)	0.045
Prenatal diagnosis	189 (95.5)	117 (91.4)	0.14
Race
White	130 (65.7)	86 (67.2)	0.70
Black	27 (13.6)	17 (13.3)
Other*	19 (9.6)	9 (7.0)
Ethnicity
Hispanic	22 (11.1)	16 (12.5)	0.70
Primary English language	185 (93.4)	124 (96.9)	0.21
Distance from CHOP, miles	51.20 (18.6 to 83.7)	81.70 (37.9 to 379.6)	<0.0001
Genetic syndrome	8 (4.0)	8 (6.3)	0.37
Heterotaxy syndrome	5 (2.5)	2 (1.6)	0.71
Pulmonary vein anomalies	5 (2.5)	3 (2.3)	1.00
Anatomy
HLHS	154 (77.7)	87 (70.0)	0.18
Double‐outlet right ventricle	14 (7.1)	12 (9.4)
Unbalanced atrioventricular canal	14 (7.1)	10 (7.8)
Other	16 (8.1)	19 (14.8)

Characteristics are summarized using median (interquartile range [IQR]) for continuous variables and frequency (percentage) for categorical variables.

^*Includes race unknown.

CHOP indicates Children's Hospital of Philadelphia; HLHS, hypoplastic left heart syndrome; ISVMP, Infant Single‐Ventricle Management and Monitoring Program; and WAZ, weight‐for‐age Z score.

Table 2.

Hospital Characteristics at S1P

	ISVMP (n=198)	Historical (n=128)	P value
Age at S1P, d	4 (3–6)	4 (2–6)	0.11
Procedure before stage I operation	20 (10.1)	6 (4.7)	0.078
Norwood operation
BTT shunt	85 (43.0)	100 (78.1)	<0.0001
RV‐PA conduit	108 (54.5)	27 (21.1)
BTT shunt and RV‐PA conduit	5 (2.5)	1 (0.8)
Cardiopulmonary bypass time, min	85 (78–98)	85 (76–110)	0.5
Mechanical ventilation, d	2 (1–5)	3 (1–5)	0.91
CPR	17 (8.6)	14 (10.9)	0.48
ECMO	12 (6.1)	6 (4.7)	0.60
Arrhythmia	70 (35%)	58 (45.3)	0.072
Length of hospital stay, d	25 (19–39)	25.5 (17–40.5)	0.63

Characteristics are summarized using median (interquartile range [IQR]) for continuous variables and frequency (percentage) for categorical variables. BTT indicates Blalock–Thomas–Taussig shunt; CPR, cardiopulmonary resuscitation; ECMO, extracorporeal membrane oxygenation; ISVMP, Infant Single‐Ventricle Management and Monitoring Program; RV‐PA, right ventricle to pulmonary artery; and S1P, stage 1 Norwood palliation.

Table 3.

Hospital Characteristics at Discharge

	ISVMP (n=198)	Historical (n=128)	P value
Weight at discharge, kg	3.53 (3.16 to 3.94)	3.42 (3.13 to 3.72)	0.02
WHO WAZ	−1.21 (−1.77 to −0.55)	−1.46 (−2.26 to −0.89)	0.0038
Number of prescribed medications	4 (3 to 6)	4 (3 to 6)	0.72
Prescribed digoxin	107 (54.0)	39 (29.7)	<0.0001
Feeding mechanism
PO	49 (24.8)	36 (28.1)	0.019
PO+NGT	123 (62.1)	61 (47.7)
Exclusive enteral feeds	26 (13.1)	31 (24.2)
Ventricular function*
Normal to mildly diminished	191 (96.5)	123 (96.1)	1.00
Moderate to severely diminished	7 (3.5)	5 (3.9)
Atrioventricular regurgitation*
No to mild regurgitation	160 (80.8)	110 (85.9)	0.23
Moderate to severe regurgitation	38 (19.2)	18 (14.1)

Characteristics are summarized using median (interquartile range [IQR]) for continuous variables and frequency (percentage) for categorical variables. ISVMP indicates Infant Single‐Ventricle Management and Monitoring Program; NGT, nasogastric tube; PO, per os; WHO, World Health Organization; and WAZ, weight‐for‐age Z score.

^*On echocardiogram.

In the weighted sample, the ISVMP and historical control groups were well balanced, with nonsignificant standard differences (<10%) for all patient characteristics included in the propensity score model (Data S1).

Primary Outcome Analysis

Not accounting for differences in patient characteristics between groups, the number of interstage readmissions per 100 interstage days was greater in the ISVMP group compared with historical controls (median, 1.09 [IQR=0–2.30] versus 0.61 [IQR=0–1.35]; P=0.0002). The majority of ISVMP (71%) infants had ≥1 readmission (Table 4). Using an inverse probability weighted quasi‐Poisson regression model, infants in the ISVMP had double the rate of readmission during the interstage period compared with controls (weighted IRR 2.38 [95% CI, 1.50–3.78]; P<0.0003; Figure 2). Sensitivity analyses as described above yielded comparable results (Figure 2, Data S2).

Table 4.

Interstage Readmission Characteristics, 2 Group Comparisons

	ISVMP (n=198)	Historical (n=128)	P value
Median number of readmissions per 100 interstage days	1.09 (0–2.30)	0.61 (0–1.35)	0.0002
≥1 interstage readmission	140 (70.7)	67 (52.3)	0.0007
Median readmission duration, d	3 (1–9)	4 (2–17)	0.025
Median number of cumulative interstage readmissions days per 100 interstage days	7.81 (2.63–16.47)	6.72 (2.09–14.44)	0.36
Readmission severity*
Severe	12 (4.3)	17 (15.2)	0.0011
Moderate	108 (38.6)	33 (29.5)	0.45
Mild	159 (56.8)	59 (52.7)	Reference
Severe readmission diagnosis
Coarctation of the neo‐aorta	0 (0)	8 (47)	0.009
BTT shunt narrowing	3 (25)	1 (6)
RV‐ PA conduit narrowing	2 (17)	0 (0)
Cardiogenic shock	1 (8)	1 (6)
Arrhythmia	0 (0)	1 (6)
Pulmonary vein stenosis	1 (8)	0 (0)
Viral URI
With heart failure	0 (0)	2 (12)
Without heart failure	1 (8)	1 (6)
Bacterial infection	0 (0)	2 (12)
Other	4 (34)	1 (6)

Characteristics are summarized using median (interquartile range [IQR]) for continuous variables and frequency (percentage) for categorical variables. BTT indicates Blalock–Thomas–Taussig shunt; ISVMP infant single‐ventricle management and monitoring program; RV‐PA, right ventricle to pulmonary artery; URI, upper respiratory infection.

^*4 patients with missing data.

Figure 2. Primary and secondary outcome analysis comparing infants in the ISVMP versus historical controls.

Forest plot demonstrating adjusted effect estimates for primary and secondary outcomes and sensitivity analyses. Infants in the ISVMP had more than double the rate of readmission during the interstage period compared with controls (weighted IRR, 2.38 [95% CI, 1.59–3.78]; P=0.0003). There was no difference in cumulative interstage readmission days between the ISVMP and control groups (adjusted IRR, 1.02 [95% CI, 0.69–1.50]; P=0.90), and infants in the ISVMP were 4 times less likely to have a severe readmission (adjusted OR, 0.28 [95% CI, 0.11–0.68]; P=0.005). Sensitivity analyses yielded comparable results. IRR indicates incidence rate ratio; ISVMP, Infant Single‐Ventricle Management and Monitoring Program; and OR, odds ratio.

Secondary Outcome Analyses

The median cumulative readmission days per 100 interstage days were comparable between the 2 groups (median ISVMP, 7.81 [IQR=2.63–16.5] versus control, 6.72 [IQR=2.09–14.4]; P=0.36), despite the increased readmission frequency. ISVMP readmissions were shorter in duration compared with controls (3 versus 4 days; P=0.025; Table 4). Within the sample of patients who had at least 1 readmission (n=207), there was no difference in cumulative interstage readmission days between the ISVMP and control groups (adjusted IRR, 1.02 [95% CI, 0.69–1.50]; P=0.90; Figure 2). Sensitivity analyses did not significantly alter these results (Figure 2).

The frequency of interventions by readmission severity level are listed in Table S4. Infants in the ISVMP with mild readmissions were more likely to require nasal cannula and feed titration than their historical control counterparts (Table S4). Of the 392 patient readmission observations, there were 29 severe readmissions (12 ISVMP and 17 controls) in 27 individual patients. Compared with a historical control, an infant in the ISVMP was significantly less likely to have a severe readmission (adjusted OR, 0.28 [95% CI, 0.11–0.68]; P=0.005; Figure 2). Sensitivity analyses yielded similar results (Figure 2). The overall frequency of cardiopulmonary resuscitation (P=0.02), mechanical or positive pressure ventilation (P=0.0006), and vasoactive support (P=0.0063) was lower in the ISVMP cohort. There was no significant difference between the cohorts in the cumulative frequency of coarctation (ISVMP, 15.0%; and controls, 18.3%; Table S4). However, nearly half (47%) of severe readmissions in the historical controls were due to coarctation of the neo‐aorta; in contrast, there were no severe admissions related to coarctation in the ISVMP cohort (P<0.009; Table 4). The severe readmission mortality rate was not significantly different between the 2 groups (ISVMP, 33.3% versus control, 23.5%; P=0.56).

Risk Factors for Readmission

Using a separate negative binomial regression model for each characteristic and controlling for enrollment in ISVMP, we identified 16 patient factors that were associated with the rate of interstage readmission (Table 5). In the final multivariable model, 4 variables were significantly associated with interstage readmission: younger gestational age (adjusted IRR, 1.12 [95% CI, 1.02–1.23]; P=0.023), presence of a genetic syndrome (adjusted IRR, 1.98 [95% CI, 1.20–3.26]; P=0.008), discharge from S1P admission on exclusive enteral feeds (adjusted IRR, 1.50 [95% CI, 1.08–2.08]; P=0.014), and residing a shorter distance in miles from CHOP (adjusted IRR, 1.01 per 25 miles [95% CI, 1.01–1.03]; P=0.025).

Table 5.

Risk Factors for Rate of Interstage Readmission (n=326)

	Univariable analysis		Multivariable analysis
Risk factor	IRR (95% CI)	P value	IRR (95% CI)	P value
Closer distance to CHOP, per 25 miles	1.02 (1.01–1.03)	0.016	1.01 (1.01–1.03)	0.025
Younger gestational age, wks	1.14 (1.04–1.25)	0.004	1.12 (1.02–1.23)	0.023
Presence of genetic syndrome	2.03 (1.19–3.47)	0.009	1.98 (1.20–3.26)	0.0075
Noncardiac abnormalities	1.52 (0.93–2.47)	0.095	1.24 (0.76–2.00)	0.39
Postnatal diagnosis	1.63 (1.00–2.70)	0.051	1.59 (0.98–2.56)	0.056
Anatomy–Other	0.57 (0.37–0.90)	0.015	0.67 (0.44–1.03)	0.069
Stage 1 CPR	1.49 (1.01–2.18)	0.043	1.15 (0.78–1.69)	0.48
Stage 1 cardiac catheterization	1.33 (1.03–1.73)	0.032	1.24 (0.95–1.61)	0.12
Stage 1 repeat surgical intervention	1.40 (0.97–2.02)	0.076	1.27 (0.89–1.80)	0.18
Discharge WAZ	0.88 (0.78–0.99)	0.030	1.02 (0.90–1.16)	0.79
Discharge oxygen requirement	1.53 (0.92–2.55)	0.010	1.18 (0.72–1.95)	0.51
Discharge on oral feeds	0.73 (0.55–0.98)	0.051	1.00 (0.74–1.35)	0.99
Discharge on enteral feeds	1.92 (1.42–2.60)	<0.001	1.50 (1.08–2.08)	0.014
Number prescribed of discharge medications	1.10 (1.02–1.18)	0.014	1.02 (0.94–1.10)	0.71

Univariable and multivariable negative binomial regression models control for enrollment in ISVMP. CHOP indicates Children's Hospital of Philadelphia; CPR, cardiopulmonary resuscitation; IRR, incidence rate ratio; ISVMP, Infant Single‐Ventricle Management and Monitoring Program; and WAZ, weight‐for‐age Z score.

Using univariable logistic regression with generalized estimating equations, we identified 4 risk factors that were significantly associated with severe readmission (Table 6): presence of genetic syndrome (OR, 4.42 [95% CI, 1.63–12.0]; P=0.004), need for repeat surgery during S1P admission (OR, 2.59 [95% CI, 1.00–6.68]; P=0.049), discharge home on supplemental oxygen (OR, 3.19 [95% CI, 1.07–9.46]; P=0.036), and a greater number of discharge medications (OR, 1.30 [95% CI, 1.00–1.69]; P=0.045).

Table 6.

Risk Factors for Severe Readmission Controlling for ISVMP (n=392 Patient Readmission Observations)

	Univariable analysis		Multivariable analysis
Risk factor	OR (95% CI)	P value	OR (95% CI)	P value
Presence of genetic syndrome	4.42 (1.63–12.00)	0.0036	3.43 (1.00–11.71)	0.049
Stage 1 repeat surgical intervention	2.59 (1.00–6.68)	0.049	2.16 (0.91–5.13)	0.081
Discharge oxygen requirement	3.19 (1.07–9.46)	0.036	4.47 (1.44–12.94)	0.0058
Number of discharge medications	1.30 (1.00–1.69)	0.045	1.13 (0.88–1.44)	0.35

Univariable and multivariable logistic regression models control for enrollment in ISVMP. ISVMP indicates Infant Single‐Ventricle Management and Monitoring Program; and OR, odds ratio.

Discussion

In this study, we investigated the association of home monitoring with the frequency, cumulative length, and severity of unanticipated interstage readmissions at a high‐volume center. We found that although ISVMP was associated with an increased frequency of readmissions, cumulative readmission days were equivalent and readmissions less severe, independent of patient factors. The decline in readmission severity was seemingly driven by the elimination of severe presentations of coarctation of the neo‐aorta. To our knowledge, this study is the first to report the association of home monitoring with unanticipated interstage readmissions.

We found that infants in the ISVMP had double the rate of readmissions compared with historical controls, with ≈70% of infants in the ISVMP having at least 1 unanticipated interstage readmission. This is consistent with other studies, including a multicenter report from the National Pediatric Cardiology—Quality Improvement Collaborative, which found that 66% of patients experienced at least 1 interstage readmission. ⁶ , ⁷ , ¹¹ The independent association of home monitoring with increased interstage readmission rates is not unexpected. Through standardization of care and close communication, the goal of home monitoring is to detect physiologic variances that precede clinical deterioration. When these changes are detected, infants require hospitalization for evaluation, monitoring, and treatment. This results in greater readmission frequency but may be protective against interstage mortality. Infants in the ISVMP with mild readmissions were more likely to require noninvasive interventions, including supplemental oxygen via nasal cannula and feed titrations. We attribute this to the daily home oxygen saturation and weight gain monitoring implemented by the ISVMP.

Despite the increase in unanticipated readmissions, the ISVMP was not associated with an increase in cumulative interstage readmission days, as readmissions were shorter compared with those of historical controls. This can likely be attributed to the 4‐fold decline in severe readmissions associated with the ISVMP. This reduction was largely driven by the decline in severely ill infants presenting with coarctation of the neo‐aorta. While nearly half of severe readmissions in the historical control group were attributed to coarctation, there were none in the ISVMP group. As the overall rates of coarctation were not significantly different between the ISVMP and historical control groups, we believe this improvement is the result of early detection through vigilant interstage surveillance.

Controlling for enrollment in the ISVMP, we identified 4 independent patient‐level risk factors for readmission: younger gestational age, presence of a genetic syndrome, discharge from S1P admission on exclusive enteral feeds, and living a shorter distance to the hospital. Given the hypothesized relationship between readmissions and mortality, it is not surprising that risk factors for interstage mortality—younger gestational age, presence of a genetic syndrome, and exclusive enteral feeds—were also identified as risk factors for readmission. ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ Prior studies have demonstrated that distance to a cardiac (or tertiary) care center is not a risk factor for mortality in infants with congenital heart disease, including those with HLHS. ⁶ , ¹⁷ , ¹⁸ We found that further distance from CHOP was associated with fewer readmissions. We hypothesize that patients who reside further from CHOP may be less likely to be admitted for more “minor” concerns as the ISVMP team is aware that distance may place a hardship on families. These patients are also more likely to be followed by a cardiologist outside our institution who may be less likely to refer patients for readmission compared with a cardiologist employed at our institution.

Our study is the second to report on risk factors for interstage readmission. Results from the National Pediatric Cardiology Quality Improvement Collaborative identified risk factors for readmission: presence of a genetic syndrome, preoperative ventricular dysfunction, tricuspid regurgitation, increased duration of circulatory arrest, and undergoing a hybrid procedure compared with an S1P with right ventricle to pulmonary artery conduit. ⁷ , ⁹ While the National Pediatric Cardiology Quality Improvement Collaborative study was multicenter, there were 2 notable limitations. The risk factor analysis excluded observational readmissions, and the interstage mortality rate for the cohort was relatively high at 9.5%. ⁷ This contrasts with an interstage mortality rate of 2.5% in our study population. In contrast with the National Pediatric Cardiology Quality Improvement Collaborative study, we did not find that tricuspid regurgitation or cardiopulmonary bypass times were risk factors for readmission. We did not evaluate preoperative ventricular dysfunction as a risk factor and excluded those who solely underwent a hybrid procedure as they are infrequently performed as the definitive first stage at our institution.

Our findings suggest that while home monitoring has resulted in more frequent readmissions, this may not be at the expense of increased in‐hospital resource utilization. Not only were cumulative interstage readmission days not significantly different between the 2 groups, but ISVMP readmissions were less severe, with significantly fewer patients requiring intensive support. While only surrogates for resource use and not direct measures of cost, these results suggest that in‐hospital interstage resource use in the era of home monitoring is likely not higher and may in fact be lower compared with controls. Families should be counseled that most infants with HLHS will have at least 1 unanticipated readmission during their interstage period and should be notified of any relevant patient risk factors.

Limitations

We acknowledge several limitations to our study. Our study design is retrospective and quasi‐experimental, with a “before versus after” analysis. While we used inverse probability of treatment weights to control for differences in patient‐level characteristics between the ISVMP and historical control groups, it is difficult to fully account for era bias or some degree of misappropriation of the impact of instituting the ISVMP. As such, our results reflect an association with introduction of the ISVMP, with findings not directly attributed to causation.

There is also the potential for misclassification bias due to incomplete capturing of outside hospital readmissions. Due to changes in our referral patterns over the past decade, historical control infants resided further from CHOP than infants in the ISVMP. As those who reside further from CHOP are more likely to be admitted to an outside hospital, misclassification would bias our results away from the null. To reduce this possible bias, distance to CHOP was included in the propensity score model. In addition, we were reassured by the sensitivity analysis excluding those transferred to an outside hospital, as these patients were presumably at the highest risk for outside hospital readmission.

Finally, this is a single‐center study at a large academic medical center. While the study population is diverse and likely representative of other high‐volume academic centers, our results may be less generalizable to smaller‐volume programs or those with different models of home monitoring. The readmission severity schema was designed on the basis of our institution's typical clinical practice. Notably, our center often uses milrinone monotherapy in those with compensated ventricular dysfunction and reserves vasoactive infusions for those with hemodynamic instability and end‐organ dysfunction. We recognize that this may be different than other programs.

Conclusion

In this single‐center study, we report that despite a significant increase in unanticipated interstage readmission frequency, the ISVMP was not associated with an increase in cumulative interstage readmission days and resulted in a reduction in severe readmissions. These findings demonstrate that home monitoring can reduce interstage morbidity and mortality without significantly increasing interstage readmission days. Future studies will focus on cost analyses and measure the resource use associated with readmissions and home monitoring.

Sources of Funding

This work was supported by the Cardiac Center Clinical Research Core at the Children's Hospital of Philadelphia. Dr Shustak received support from the National Institutes of Health's National Heart, Lung, and Blood Institute, grant T32 HL007915. Dr Mercer‐Rosa received support from grant K01HL125521 from the National Institutes of Health's National Heart, Lung, and Blood Institute.

Disclosures

None.

Supporting information

Data S1

Data S2

Methods S1

Methods S2

Tables S1–S4

Figure S1

Reference ¹⁹