Abstract
Abbreviations
Neonatal Intensive Care Unit Nationwide Children’s Hospital Birth Weight Gestational Age Length Of Stay Emergency Department Comprehensive Center for Bronchopulmonary Dysplasia Inter-Quartile Range After Visit Summary Quality Improvement
Key notes
We evaluated our current discharge process in our large Level IV NICU which provides us a unique opportunity to overcome the shortcomings of ‘silent’ NICU readmissions. Upon readmission, nearly half of all subsequent readmissions were for 3 or fewer days, indicating that the second admission tended to be for short-lived acute deteriorations in clinical status. We have further identified ‘at risk patient’ factors such as longer length of stay and Medicaid and hope to target further interventions towards this population, especially geared towards outpatient services
Introduction
Infants discharged from the Neonatal Intensive Care Unit (NICU) are a heterogeneous group of ‘high risk infants.’ They are often premature at birth, have special health care needs, are dependent on technology, have high risk family situations, are at risk for early death or have a combination of these factors [1, 2]. Because of their medical complexities, substantial efforts have been made to determine the optimal timing of hospital discharge, to improve discharge planning, and to address follow-up and home care needs [2–5].
The discharge process is intricate. Consequently, re-hospitalization is disruptive to families and has significant financial consequences [6–8]. The rate of re-hospitalization within 30 days after initial discharge is proposed as a benchmark of the quality of care provided in the NICU [9–11]. As such, 30 day readmission rates may be used as a balancing measure for length-of-stay (LOS) reduction efforts. However, many Children’s Hospitals are regional referral hospitals, so babies that are discharged may be readmitted to other hospitals or health-care systems closer to home and may not return to the index hospital. Readmissions may not be detected if patients are transferred to other patient-care units prior to discharge. Thus, both of these create ‘silent’ [12] readmissions which invalidates the metric [12–14]. Finally, drivers of readmission may be entirely unrelated to the initial hospitalization (post-discharge viral infections, for instance), and may not reasonably be improved with hospital-based interventions.
Objective
To evaluate our readmission data in order to determine readmission drivers to enhance a Quality Improvement project to reduce the readmission rate in our NICU.
Methods
Setting
The Nationwide Children’s Hospital (NCH) is a large 114 bed, Level IV NICU with an average of 1000 discharges per year. It is the primary referral NICU for 37 counties in Central and Southeast Ohio with an average of 41,809 live births per year from 2010–2014. In 1994, NCH partnered with one of the largest pediatric accountable care organizations in the United States, Partners for Kids. This model provides financial incentives to reduce the length of stay (LOS) and readmissions. Due to medical complexities, patients are rarely reverse transported to referral hospitals before discharge. 91% of patients are discharged to home from the NICU. In addition, NCH covers 78.9% of total discharges for children between 0 and 14 years of age. NCH is the sole source of pediatric beds within Franklin County and nearly all readmissions are referred back to our hospital. This provides us a unique opportunity and a complete dataset to review most of the readmissions. We used this dataset on 30-day readmissions after NICU discharge to identify drivers for readmission. The drivers determine characteristics of patients at high-risk for readmission that should be targeted with additional interventions in order to prevent readmissions.
Data source
We evaluated all infants discharged from our central level IV NICU and that were readmitted anywhere in our hospital within 30 days after NICU discharge during the period from 1/2011-12/2014. All data was extracted from our electronic medical records and electronic data warehouse.
Ethics
The study was screened and classified as Quality Improvement by the Institutional Review Board at The Nationwide Children’s Hospital and thus did not require full review.
Discharge process
Our center has multiple ongoing quality improvement interventions seeking to optimize the discharge process and management of high risk neonates. Several programs have been developed in order to provide the caregivers consistent information that is specific to their infant’s health and developmental needs (health literacy). This facilitates opportunities for developmentally appropriate caregiver-baby interactions, promotes caregivers’ confidence, coping skills and develops realistic expectations of their baby. Figure 1 shows the extensive process steps in our existing discharge process.
Clinical leaders from our NICU initiate a follow-up phone call with the caregivers of every discharged infant within 1 week to obtain feedback regarding their stay and also the utility of discharge teaching and management. We utilized our existing telephone follow-up process and administered a survey for 1 year, in 2014. The survey consisted of questionnaire with a Likert scale ranging from 1–10, to assess the parent/caregiver perception of discharge readiness, the quality of skills training to provide care for their infants and their infant’s medical readiness for discharge.
BPD population
At NCH NICU, we have developed the Comprehensive Center for Bronchopulmonary Dysplasia (CCBPD) with a dedicated physician group and a strong interdisciplinary team approach focused on respiratory support, nutrition, developmental milestones and family centered care. This program was developed using several principles from chronic care model [15]. Most of the infants diagnosed with BPD are transferred to this unit at 36 weeks corrected GA and may have a short stay in this unit before discharge. The discharge process in this unit is similar but the patients are provided comprehensive outpatient care within a medical home model following discharge.
Statistical analysis
All Statistical analysis was performed in Microsoft Excel, version 2010 Minitab version 17 and SPSS version 21. Data is presented as mean±SD, median and inter-quartile range (IQR) or percentages. T Test, Fisher’s exact test and multiple logistic regression analysis was performed as applicable. A p value <0.05 was considered significant.
Results
During the 4-year study period, 3181 patients were discharged to home from the NCH NICU. Table 1 shows the baseline characteristics of our population. Our average all-cause readmission rate within 30 days was 9.8% (n = 313) and a stable process as shown in the control chart (Fig. 2). However, the median and IQR have shown a decreasing trend over the 4 year period that have not yet reached statistical significance. The median was 9.1% vs 8.65% and IQR was 5.65% (7.4–13.05%) vs 3.73% (6.35–10.02%) in 2011 and 2014 respectively.
Post discharge survey
During 2014, a total of 394 parents/caregivers were called. There was a 58% response rate for our post-discharge questionnaire. We asked caregivers to retrospectively rate their readiness for discharge on a Likert scale ranging from 1–10 (10 being optimally ready). 79% of caregivers rated their readiness to care for the infant following discharge teaching in the hospital with scores from 7–10. 74% of caregivers rated their perception of their baby’s medical readiness for discharge with scores from 7–10. Of the caregivers who had high scores of 7–10, only 8% of the babies were readmitted.
Pre-discharge patient characteristics
We then determined patient characteristics that would identify risk factors for readmission. In univariate analysis, length-of-stay and having Medicaid were significant risk factors (Table 2). In multivariate logistic regression using LOS, GA, BW and Medicaid, each additional original admission day increased the readmission odds by 0.5% whereas having Medicaid instead of commercial insurance increased risk by 72.4%. Therefore, while LOS is a risk factor, specificity is relatively low as patients with excessive lengths of stay still have readmission rates of less than 20%.
Post-discharge characteristics
We then attempted to identify readmission pathways which might be amenable to intervention.
In our setting, the vast majority of patients who were readmitted presented through the emergency department (ED). We also examined the number of days patients had been home before readmission (through 30 days) and saw no appreciable trend as the number of day’s post-discharge increased (data not shown). Finally, we examined the length of stay for the second admission (readmission) (Fig. 3), and observed that nearly half of all subsequent readmissions were for 3 or fewer days, indicating that the second admission tended to be for short-lived acute deteriorations in clinical status.
Patients with BPD have extremely long initial LOS and are often clinically more fragile at initial discharge. As such, these patients should, according to ours and other models, be among the highest risk for readmission within 30 days of discharge. Table 3 shows that while LOS differs significantly between patients discharged from the CCBPD versus the remainder of the NICU, the difference in readmission rate is not statistically significant (10.8% vs 9%; p value 0.19).
Discussion
Our hospital has several QI efforts to optimize the discharge process. Here we describe the data of our existing extensive process and the post discharge survey. Our extensive and relatively complete set of data on patients discharged from our level IV NICU who were readmitted within 30 days of discharge indicates that our readmission rate has been a stable process in our population. We have identified possible factors that impact transition of care to home.
Previous studies have shown that infants discharged from any NICU are a high risk population with higher rates of re-hospitalization when compared to healthy term infants [16]. Efforts have been made to identify and stratify risk factors to focus planning and co-ordination of care after discharge. There is significant heterogeneity in the etiology and characteristics of NICU patients. Factors such as gestational age (GA) or birth weight have been not shown to be the sole risk factors for readmission [8, 17–21], a finding replicated in our data.
Studies have demonstrated a modest increase in risk of readmission with increasing LOS. In a stratified sample, Ambalavanan showed that infants with hospital stay of >120 days for pulmonary reasons had a 66% re-hospitalization rate compared with 42% without such a stay [17]. Our data corroborates a weak link between LOS and readmission rate in our total population, likely due to either decreased gestational age at birth or increased medical complexity of patients born near term, both of which will lead to increased fragility. In the first year of life patients with BPD have twice the rate of readmission as compared to non BPD populations (49% vs 23%, p = <0.0001) [22]. With the implementation of the CCBPD in 2004, our group has shown that the rate of readmission has decreased from 29% to 3.1% [15] and thereafter has remained at 10% as indicated in our current data.
By a large margin, the single most important risk factor for readmission in our population was having Medicaid as a payor, which is a marker for the socioeconomic status of the family. Our finding that public insurance is associated with higher likelihood of readmission as compared with private insurance is similar to evidence in other patient populations [23, 24]. With an optimal discharge process and short readmission LOS, we believe that, the primary driver of a NICU’s readmission rate may be the ability of a family with limited means to cope with inevitable mild to moderate setbacks in the condition of a patient who is medically fragile.
As highlighted in previous literature, we recognize that one of the limitations of analyzing the readmissions data is that, there may be a small number of patient population that may be readmitted to other hospitals [12–14].
The question then becomes “what can be done to prevent readmissions?” The efforts of our CCBPD program [15] have led to a reduced and lower-than-expected readmission rate in the population of patients with BPD [22]. Our CCBPD program provides robust, outpatient program services including an outpatient clinic and neonatal physicians and nurses on-call 24/7 to help caregivers and families. The closed loop communication amongst the team members provides easy access to the caregivers, continuity of care, and close follow-up of patients. Similar models of care for chronically ill children have been associated with a 23% reduction in the incidence of ED utilization [25], and have shown that the characteristics of outpatient facilities, not NICUs, are determinants of readmission rates [26]. In addition to preventing readmissions, comprehensive follow-up care by experienced caregivers has been highly effective in reducing life-threatening illness without increasing costs among high-risk inner-city infants [27], has reduced serious illness and total hospital/clinic costs [28].
The exact characteristics of these systems have yet to be defined. However, since our current processes have addressed the accuracy of the timings of discharge, discharge processes and caretaker readiness to care for the infant optimally, there could be benefit in emulating the CCPBD outpatient measures in the high risk population to further our QI project.
Conclusion
In our attempt to highlight the additional drivers related to our existing readmission QI project, and evaluate our process measures, we have identified that although readmission rates are a stable process, interventions targeted towards ‘at risk patient’ factors such as longer length of stay and Medicaid, interventions specifically targeted towards patients visiting the emergency department and developing an extensive outpatient program may be of value in addition to further optimizing the discharge process.