OBJECTIVES: (1) To identify a resource use inflection point (RU-IP) beyond which patients in the NICU no longer received NICU-level care, (2) to quantify variability between hospitals in patient-days beyond the RU-IP, and (3) to describe risk factors associated with reaching an RU-IP.
METHODS: We evaluated infants admitted to any of the 43 NICUs over 6 years. We determined the day that each patient’s total daily standardized cost was <10% of the mean first-day NICU room cost and remained within this range through discharge (RU-IP). We compared days beyond an RU-IP, the total standardized cost of hospital days beyond the RU-IP, and the percentage of patients by hospital beyond the RU-IP.
RESULTS: Among 80 821 neonates, 80.6% reached an RU-IP. In total, there were 234 478 days after the RU-IP, representing 24.3% of the total NICU days and $483 281 268 in costs. Variability in the proportion of patients reaching an RU-IP was 33.1% to 98.7%. Extremely preterm and very preterm neonates, patients discharged with home health care services, or patients receiving mechanical ventilation, extracorporeal membrane oxygenation, or feeding support exhibited fewer days beyond the RU-IP. Conversely, receiving methadone was associated with increased days beyond the RU-IP.
CONCLUSIONS: Identification of an RU-IP may allow health care systems to identify readiness for discharge from the NICU earlier and thereby save significant NICU days and health care dollars. These data reveal the need to identify best practices in NICUs that consistently discharge infants more efficiently. Once these best practices are known, they can be disseminated to offer guidance in creating quality improvement projects to provide safer and more predictable care across hospitals for patients of all socioeconomic statuses.
- CCC —
- complex chronic condition
- CI —
- confidence interval
- ECMO —
- extracorporeal membrane oxygenation
- ED —
- emergency department
- IPC —
- inflection-point cohort
- IQR —
- interquartile range
- LOS —
- length of stay
- OR —
- odds ratio
- PHIS —
- Pediatric Health Information System
- RR —
- risk ratio
- RU-IP —
- resource use inflection point
What’s Known on This Subject:
Variation in health care is known to increase cost and decrease quality and safety. Inadequate anticipation of parental needs and education at discharge from the NICU, balanced against physician readiness to discharge infants, can lead to variation and longer length of stay.
What This Study Adds:
With our findings, we identify significant variation between freestanding children’s hospitals. Addressing this variation could provide a potential cost savings of >$483 million for the healthiest patients in the NICU among only 43 hospitals and far greater cost savings if generalized across NICUs nationally.
Neonatal intensive care is among the most expensive aspects of health care in the United States. A significant component of total NICU cost is length of stay (LOS), which is determined in part by patient-level medical characteristics: prematurity, ventilator dependence, need for constant monitoring, temperature control, and feeding dependence, among other factors.1–3 Some portion of NICU LOS, however, is likely related to nonmedical factors often guided by individual and institutional historical practice patterns rather than evidence. Patients and families or care providers must often balance the competing desire to make the transition out of the NICU versus the desire to remain in the NICU for fear that their infant is not yet medically stable enough to transition out safely.4 Health care providers must balance the parental perspective against their own clinical judgment to determine when it is safe to discharge. This balance between perceptions and comfort of the patient families and that of the health care providers is likely to lead to variability in determining readiness for discharge when NICUs do not have standardized discharge criteria.
Under the assumption that most people tend to err on the side of safer conservative decision-making when determining readiness for discharge from the NICU, we hypothesize that many children remain hospitalized longer than clinically necessary. If this is true, we expect the daily cost of patients in the NICU to taper along with the majority of high-cost critical interventions occurring early in the hospitalization. As patients stabilize and are safe for transition to a lower-level inpatient unit or to the outpatient environment, there would be fewer billable events over time. Our aim for this study was to identify a resource use inflection point (RU-IP) beyond which NICU patients received minimal NICU-level billable care. If infants were stratified into 2 statistically adjusted groups on the basis of whether the RU-IP was reached before discharge, the assumption could be made that those who never reached the RU-IP before discharge were being discharged more efficiently. To support this hypothesis, we also sought to quantify the variability between hospitals in terms of the number of days patients remain in the NICU beyond the RU-IP. Identification of an RU-IP may allow health care systems to identify readiness for discharge from the NICU earlier and thereby save hospital days and health care dollars.
For this retrospective study we used administrative data between 2010 and 2015 from hospitals contributing to the Pediatric Health Information System (PHIS) database. Children’s Hospital Association (Lenexa, KS) maintains the PHIS database, which includes daily clinical and resource use data (pharmacy, imaging, laboratory, room and board, and supplies) from inpatient, emergency, ambulatory surgery, and observation encounters at 48 tertiary care children’s hospitals in the United States. The data are subject to numerous validity and reliability checks before incorporation into the database, and data are deidentified at the time of submission to the PHIS. Data warehouse function is provided by Truven Health Analytics (Ann Arbor, MI). The study was reviewed and approved by the Institutional Review Board of Seattle Children’s Hospital.
Study Population and Setting
We included neonates <7 days of age admitted between January 2010 and December 2015 to any of the 48 freestanding children’s hospital NICUs (N = 167 337) abstracting data into the PHIS data set. All NICUs are classified as level III or IV given that each care for patients who are at <32 weeks’ gestation. Data from 5 hospitals were excluded before performing the analysis because of known data quality issues with detailed resource use in these specific hospitals (n = 16 102). Because our aim was to study characteristics at the time of transition from the NICU to either home or a lower level of care, we excluded neonates who died during the admission (n = 6377), had unknown discharge status (n = 1248), or were discharged to hospice care (n = 73). To minimize patient-related variability and confounding in the evaluation of LOS, we decided a priori to simplify the study population and exclude patients with complex chronic conditions (CCCs) (n = 62 716), defined by Feudtner et al5–8 as any medical condition that can be expected to last ∼12 months (unless death intervenes) and to involve either several different organ systems or 1 organ system enough to require specialty pediatric care and some potential period of hospitalization in a tertiary care center.
Daily Resource Use
We defined the RU-IP as the hospital day that a patient’s total daily standardized cost fell within a percentage of the mean first-day NICU room cost and remained within this range until the last 2 days of the hospitalization. Patients determined to have met this cost criterion were determined to have reached the RU-IP. To identify the most appropriate cost to be used to determine the RU-IP, we first graphed the percentage of patients who would reach the RU-IP against all percentages between 0% and 30% of the mean cost (Fig 1). This graph supported our prediction that as the percentage greater than the mean NICU cost increased, the percentage of patients reaching the RU-IP would increase. We identified a change in the slope between 10% and 15%. All authors subsequently independently reviewed summaries and specific types of billable events occurring on days after an RU-IP at 25%, 10%, and 5% greater than the mean NICU room cost (Supplemental Tables 5–7). As predicted, as the percentage decreased, the types of events that occurred were less likely to require NICU-level care to perform. We chose to move forward with the 10% cutoff because it was closest to the natural inflection point identified in Fig 1, but because it was to the left of the inflection, we would capture a smaller population and therefore improve the sensitivity of our findings while maintaining events considered by the authors not to require NICU-level care to perform (Supplemental Table 8). We therefore chose to continue our evaluation using only this 10% cohort, defining this as the inflection-point cohort (IPC) and those who did not reach an inflection point as the non-IPC. We chose to ignore costs incurred during the last 2 days of hospitalization after finding that many patients were receiving expensive treatments, such as immunizations and feeding-tube placements, on the last 2 days of hospitalization to prepare for NICU discharge. These days were included, however, in the overall LOS.
To eliminate unit-cost level variation, we standardized the cost of each of the ≥22 000 items in the PHIS data as the overall median hospital costs for that item.9 For example, the cost of a total bilirubin laboratory test ranged from $2 to $425 across the 43 hospitals, with a median cost of $22. We then recalculated every hospital bill using the standardized cost of each item (in this example, $22) and attributed any differences in cost across hospitals to differences in volume of use rather than differences in costing methodology. All currency is expressed in US dollars.
Outcome Measures and Covariates
Ultimately, patients were stratified into 2 groups: those who reached the RU-IP and those who did not. We examined the total number of days beyond the RU-IP. Covariates included level of prematurity, yearly case volume, race and/or ethnicity, payer type, and disposition. Medical covariates included mechanical ventilation, feeding support, extracorporeal membrane oxygenation (ECMO), and methadone use. Prematurity was considered carefully in our model because many patients remain in the NICU until they reach term status.10 NICU LOS was defined as LOS in the NICU (as opposed to total LOS). Readmission and emergency department (ED) visits were defined as returns to the index hospital.
Categorical variables were summarized by using frequencies and percentages. Continuous variables were summarized by using medians and interquartile ranges (IQRs). Intergroup differences were tested for significance by using the χ2 or Wilcoxon rank test. We used generalized linear mixed-effects regression analyses with a negative binomial distribution to examine all relationships between the covariates and outcome measures using hospital as a random intercept to control for hospital clustering. Adjusted odds ratios (ORs) were calculated as well as 95% confidence intervals (CIs). Because of the large sample size, in all analyses, we used 2-tailed tests with an a priori significance level of P < .001.
We identified 80 821 neonates hospitalized in NICUs across 43 children’s hospitals between 2010 and 2015. Of these patients, 80.6% (n = 65 139) reached an RU-IP during their hospitalization (Table 1). In comparison with the non-IPC, the IPC was term (50.2% vs 54.4%), was non-Hispanic white (45.4% vs 50.7%), and had private-payer insurance (39.6% vs 42.1%) more frequently. LOS within the NICU did not differ significantly between the non-IPC and the IPC populations (5 days [IQR 3–9] vs 6 days [IQR 2–13]; P = .927). Standardized costs were found to be higher in the non-IPC, whether examining this as a cost per patient ($21 597 [IQR $12 308–$46 918] vs $16 555 [IQR $6798–$37 689]; P < .001) or as a cost per day ($2634 [IQR $2473–$2946] vs $2462 [IQR $2325–$2721]; P < .001). Although statistically significant due to the large sample size, patient disposition status was similar across both populations, with a lower proportion of IPC neonates discharged from the hospital to home (89.3% vs 87.8%) and, conversely, a higher proportion discharged to another care facility (4.8% vs 6.3%). When comparing inpatient, ED, or combined all-cause readmission rates at 7, 30, and 90 days, no significant differences were noted between the 2 groups (Table 2). When comparing the median days to inpatient readmission within 90 days, patients from the IPC returned significantly sooner (18 vs 10 days; P < .001).
The proportion of patients hospitalized beyond an RU-IP varied between institutions (33.1%–98.7%), and in the majority of hospitals (32 of 43;74.4%), >80% of patients reached the RU-IP (Fig 2). The variability between hospitals extended to the proportion of total days of hospitalization spent beyond the RU-IP (1.7%–56.7%) (Fig 3). Across all of the institutions examined, we observed a net of 234 478 total days beyond the RU-IP, representing 24.3% of the total hospital days across the cohort. This was calculated to represent $483 281 268 in standardized costs accumulated beyond the RU-IP in all 43 hospitals, or 18.9% of the total standardized cost of care (range: $128 228–$51 780 080; 0.5%–50.9% of individual hospital cost, respectively).
To examine factors contributing to neonates reaching the RU-IP, we modeled the outcome of reaching the RU-IP against several hospital- and patient-level characteristics (Table 3). Compared with neonates discharged from the hospital to home, neonates discharged to other care facilities were more likely to reach the RU-IP (OR: 1.29; 95% CI: 1.17–1.42; P < .001) during their NICU hospitalization. Conversely, neonates receiving mechanical ventilation (OR: 0.78; 95% CI: 0.74–0.83; P < .001) or feeding support (OR: 0.65; 95% CI: 0.6–0.68; P < .001) during their NICU hospitalization were less likely to reach the RU-IP.
We also examined factors associated with the number of days beyond the RU-IP (Table 4). Extremely preterm (risk ratio [RR]: 0.46; 95% CI: 0.44–0.48; P < .001) and very preterm (RR: 0.71; 95% CI: 0.69–0.73; P < .001) neonates were found to have fewer days beyond the RU-IP compared with term neonates, whereas moderate to late preterm neonates (RR: 1.08; 95% CI: 1.06–1.09; P < .001) were associated with slightly more post–RU-IP days. Patients discharged with home health care services exhibited fewer days beyond the RU-IP (RR: 0.89; 95% CI: 0.86–0.92; P < .001) compared with those patients discharged from the hospital without services. Receiving mechanical ventilation (RR: 0.79; 95% CI: 0.75–0.78; P < .001), feeding support (RR: 0.65; 95% CI: 0.64–0.66; P < .001), or ECMO (RR: 0.59; 95% CI: 0.48–0.71; P < .001) at any point during the hospitalization was associated with fewer post–RU-IP days. Conversely, receiving methadone during the hospitalization was associated with an increased number of days beyond the RU-IP (RR: 1.14; 95% CI: 1.06–1.24; P < .001).
In this cohort of NICU patients without CCCs in 43 children’s hospitals over 5 years, we measured a potential excess NICU LOS of 234 478 days, representing 24.3% of the total NICU days and $483 281 268 in potential excess costs. A review of the top 25 resources used in this population during the post–RU-IP period of hospitalization (Supplemental Table 8) confirms that these resources are important but not necessarily indications for NICU-level care, such as laboratory draws (eg, bilirubin levels, but not lights for treatment), hepatitis vaccines, vitamins, and oximetry (but not oxygen).
There are of course many potential causes for prolonged LOS, and it is likely that we are identifying these associations within a complex and changing health care environment. Before the advent of standardization, timing for discharge from a NICU was a balance between several factors; the confidence of the health care team in an infant’s clinical stability, the confidence of the health care team in the family or guardian’s ability to care for the infant outside of the NICU, and the parent or guardian’s confidence to care for that infant at home.11 Each of these components can in turn be affected by several variables.
First, providers may not trust that a patient is medically stable for discharge, and therefore physicians and hospital systems in general may increase LOS to prevent home adverse events that lead to readmissions or ED visits.12,13 Foreseeing the argument that a more aggressive approach to discharge may result in increased revisits, we compared readmission and ED visit rates between the IPC and non-IPC and found that the IPC in fact had a significantly higher inpatient readmission rate at 7 days and that although there was not a difference at 30 or 90 days, those who did return within these time periods came in significantly sooner. It is important to emphasize that we did not identify any significant difference between the cohorts in 7- or 30-day all-cause returns to the system. Our data may therefore suggest that discharging infants from the NICU more efficiently likely does not result in increased readmissions or ED visits, and more importantly, there is no obvious protective effect of a longer LOS.
Counterintuitively, in our study, we identified that patients who were sicker were more likely to be among the non-IPC. Infants who were on ECMO, were mechanically ventilated, or required feeding support were all more likely to be discharged before reaching the RU-IP or to accrue fewer days after reaching it compared with their healthier counterparts. Unadjusted, the IPC was more frequently non-Hispanic white, term, privately insured, and discharged to facilities rather than home. Additionally, they returned to the hospital at a higher rate within the first week and in fewer days from discharge. Adjusted, the IPC remained more likely to be term and discharged to a facility rather than home and was less likely to be mechanically vented or to require feeding support. Longer LOS may simply allow more time to prepare for discharge and meet historical discharge criteria. Historically neonates remain in the NICU for the intensive nursing care that supports the growth and maturation of the newborn, including monitoring and correcting temperature, respiratory, and metabolic instability. Early consensus statements recommended discharge criteria that included both objective criteria, such as a weight of 2 kg, and subjective criteria, such as feeding stability.11,14 Evidence-based standardization fundamentally controls bias at the individual patient level. The variability between hospitals in the percentage of patients reaching the RU-IP (33%–99%) may be explained in part by which hospitals in our study have instituted standardized pathways. Standardized care pathways have in fact been shown to decrease LOS and minimize provider- and patient-level bias.15,16 These standardized quality improvement pathways are more likely to be developed for sicker patient populations, and this could contribute to their higher frequency in the non-IPC. Although one might argue that the longer LOS of the IPC is required to meet the discharge criteria referenced here, the significant variability among hospitals strongly suggests that other factors are influencing decision-making in terms of readiness for discharge.
One specific NICU population that has historically experienced long LOS is the group diagnosed with neonatal abstinence syndrome. We noted that patients in our study who received methadone were more likely to be in the IPC. These findings are not surprising given recent reports that have revealed the variation in care specifically among children with neonatal abstinence syndrome and that have similarly offered solutions to this problem through standardized treatment pathways that decrease LOS and diminish disparities in care.17–19
Second, physicians and nurses may not trust that a family is technically or emotionally ready for discharge, leading to prolonged LOS. Provider- and hospital-level practice patterns and perception of parent competence in caring for complex neonates at home have been shown to influence discharge planning.4 The confidence of the health care team in the infant’s stability and in the family’s ability to care for the infant outside of the NICU may be affected by the age and experience of the nurse or neonatologist, how long they have been on service (and therefore how familiar they are with the patient and family), and their particular clinical biases.
Third, parents may not feel ready for discharge. A family’s confidence in their ability to care for the infant may be related to their education level, health literacy, the health-related education that they have received before discharge, the support they have at home, their economic resources, and their life experiences, among many other factors. Authors have shown that parents of infants with greater medical complexity are more likely to feel prepared for discharge.20 On the flip side, others have shown that families that do not feel prepared are more likely to experience problems after discharge and have described a process of iterative process improvement through measurement of the family’s and the nurse’s assessments of the emotional and technical readiness for discharge.21
The significant variability we described among hospitals may suggest that there is an opportunity to identify practices in centers that result in more efficient discharge processes. For example, one opportunity may be to transfer patients who need continued observation, but do not require all of the resources of a level III or IV NICU, to a lower level of care within the same hospital or in another facility. Additional opportunities may be to standardize practices such as early parent engagement in care, and discharge based on physiologic stability instead of a simple weight criterion.14,20,22 Several clinical and administrative collaboratives exist that could identify such best practices, which could then be leveraged to identify best practices across hospitals. We may then be able to use a positive-deviance approach to standardize objective discharge criteria that would likely result in cost savings in addition to safer and higher-quality neonatal care.23 Although the variability and increased LOS we identified may be responsible for increased measurable costs of hospitalizations, as described by these data, they may also be responsible for additional immeasurable costs related to staffing and resource allocation. Minimizing variability has been shown to improve the overall quality and safety of care in many areas of health care and has the additional benefit of creating more predictable, if not shorter, LOS, thereby stabilizing a NICU’s census to a predictable model. The overall effect of creating shorter LOS and minimizing variability is to improve resource allocation and staffing and decrease cost, thereby increasing revenue.24 Our data support that shifting to fewer post–RU-IP days may lead to an increase in hospital revenue. The greater average daily and total hospitalization cost of the non-IPC ($2634) compared with the IPC ($2462) suggests that hospitals would replace discharged patients with a higher-acuity NICU population that, in turn, has higher daily billing.
There are several limitations to this study. First, reviews of large databases are subject to the risk of misclassification. In this study, however, we focused on billing events, which in general tend to be accurate. Second, the PHIS database represents tertiary referral pediatric hospitals often with level IV NICUs, leading to our potential inability to generalize the findings. These referral centers by definition, however, tend to care for more complex patients. If the findings that healthier children may in fact contribute a higher proportion of hospital days beyond the RU-IP, then the potential cost savings in level I, II, and III centers may be even greater. Third, there may be important events occurring during the hospitalization beyond the RU-IP that are critical and can only be performed within the inpatient context but cannot be identified through billable events. In our current system, however, it is hard to identify any such events that would require hospitalization but would not require a billable record. Fourth, the PHIS database only captures returns to the same hospital, and we may have therefore underestimated readmissions to other hospitals. Fifth, we were not able to identify hospitals that may have step-down units using billing data.
To our knowledge, no study has yet assessed the overall burden of excess NICU hospital days on the health care system. Our findings suggest that there is vast opportunity to decrease costs of care by improving efficiency and standardizing the NICU discharge process. Having been focused first on the NICU population without CCCs, these findings suggest beginning to investigate opportunities for cost savings with healthier patients, perhaps by identifying interventions that are focused on parent and health care provider perceptions. Qualitative research has revealed that some parents perceive delays in discharge processes or feel rushed at the time of discharge because of inadequate anticipation of parental educational needs.25,26 There may therefore be an opportunity to improve both efficiency and long-term health and patient safety by engaging and educating parents early in the hospitalization regarding ongoing needs of their child after discharge.27–29 Solutions for this potential excess cost and LOS will therefore need to address both the clinical and nonclinical drivers of this observation. Reducing variability in the discharge process from the NICU, however, may provide an opportunity to identify readiness for discharge from the hospital either to home or to step-down units earlier, increasing the value of the care we provide.
- Address correspondence to Adam B. Goldin, MD, MPH, Department of Surgery, School of Medicine, University of Washington and Seattle Children’s Hospital, 4800 Sand Point Way NE, Seattle, WA 98105. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: No external funding.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Copyright © 2020 by the American Academy of Pediatrics