Resource utilization patterns using non-invasive ventilation in neonates with respiratory distress syndrome

Abstract

Objectives

To describe the frequency of non-invasive ventilation (NIV) and endotracheal intubation use in neonates diagnosed with respiratory distress syndrome (RDS); to describe resources utilization (length of stay (LOS), charges, costs) among NIV and intubated RDS groups.

Study design

Retrospective study from the national Kid’s Inpatient Database of the Healthcare Cost and Utilization Project, for the years 1997–2012. Propensity scoring and multivariate regression analysis used to describe differences.

Results

A total of 595,254 out of 42,912,090 cases were identified with RDS. There was an increase in NIV use from 6% in 1997 to 17% in 2012. After matching, patients receiving NIV only were associated with shorter LOS: (95%CI) 25 (25.3,25.7) vs. 35 (34.2,34.9) days, decreased costs: ($/1k) 46.1 (45.5,46.8) vs. 65.0 (64.1,66.0), decreased charges: 130.3 (128.6,132.1) vs. 192.1 (189.5,194.6) compared to intubated neonates.

Conclusion

There was a three-fold increase in NIV use within the 15-year study period. NIV use was associated with decreased LOS, charges and costs compared to intubated patients.

Background

Respiratory distress syndrome (RDS) is the eighth leading cause of death for infants in the United States, accounting for 13.4 per 100,000 live births in 2013 alone [1, 2]. RDS is due to a deficiency of pulmonary surfactant in premature neonates. The lack of surfactant leads to high surface tension, which can result in a host of sequelae including lung instability, decreased volume and compliance, lung inflammation and epithelial injury. If left untreated, RDS can progress in the first 48 h following birth and result in hypoxemia and inadequate gas exchange [3]. Antenatal steroid prophylaxis, delivery room and respiratory management with early nasal continuous positive airway pressure (NCPAP), surfactant replacement in the early phase of RDS, the INSURE (intubation, surfactant, extubation) procedure and the increased use of noninvasive ventilation (NIV) have been shown to improve respiratory outcomes [4–7].

Non-invasive ventilatory strategies in neonates with RDS include NCPAP, nasal intermittent positive pressure ventilation (NIPPV), synchronized NIPPV, and bilevel CPAP [8]. Advantages of NIV compared to invasive mechanical ventilation (MV) via the endotracheal tube include reduction of barotrama, biotrauma, and ventilator-induced lung injury [8–10]. Additionally, endotracheal intubation in neonates has been associated with adverse events, which are further increased by number of intubation attempts in emergent settings. Moreover, although the pathogenesis of bronchopulmonary dysplasia is multifactorial, ventilator-induced biotrauma and volutrauma are major factors. Hence multiple noninvasive strategies have gained favor [11].

Although there is increasing knowledge surrounding the clinical benefits of NIV usage in neonates with RDS, there is minimal literature focused on the association of both invasive and non-invasive ventilation with hospital resource utilization. Therefore, the objectives of this study are as follows:

1. To describe the frequency of NIV and endotracheal intubation in neonates diagnosed with RDS nationwide from 1997 to 2012;

2. To describe resource utilization in terms of length of stay (LOS) and total hospital charges and costs, among the NIV and endotracheal intubated groups.

Methods

Data source

This is a large population-based retrospective study approved by the Institutional Review Board at Children’s Hospital Los Angeles (CHLA-14-00479). Data was obtained from the Healthcare Cost and Utilization Project Kid’s Inpatient Database (HCUP-KID) of the Agency for Healthcare Research and Quality (AHRQ) triennially for the years 1997 to 2012. The HCUP-KID is a weighted database to create national estimates of participating community and non-rehabilitation inpatient stays for patients <21 years old in the United States. HCUP cost-to-charge ratio files were used to convert the amount that hospitals billed for services (charges) to the healthcare infrastructure payment infrastructure (costs) for the years 2003–2012. HCUP supplied weights were used to produce estimates. Costs and charges were subsequently adjusted for inflation relative to year 2012 utilizing Bureau of Labor Statistics Consumer Price Index Inflation Calculator. The HCUP-KID derives patient’s baseline demographics from administrative billing codes which included: sex, race, LOS in days, total hospital charges, and type of health insurance. Procedures and diagnoses were identified using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) discharge codes.

Study population

The study population included neonates, defined as being admitted ≤28 days of life, with diagnosis of ICD-9-CM: 769.0 ‘respiratory distress syndrome in newborn’, who were discharged from 1 January 1997–31 December 2012. Noninvasive ventilation group was defined as receiving NIV only identified by ICD-9-CM: 9390 ‘Non-Invasive Mechanical Ventilation’. Endotracheal intubation group may or may not have received NIV and was identified by ICD-9-CM: 96.04 ‘Insertion of Endotracheal Tube’.

Statistical analysis

Least squares linear regression was used to identify trends in NIV and intubation use. Pearson’s chi square was used to determine if associations existed among categorical variables by year. ANOVA test for trends was used to determine trends among continuous variables by year. Independent t-tests or Pearson’s chi square was used to determine if differences existed between unmatched NIV and endotracheal intubation group’s continuous and categorical variables. 1:1 matched propensity score analysis was used to model the likelihood of being intubated. Propensity score matching was estimated by a logistic regression model utilizing a greedy 8-to-1 digit matching algorithm [12] accounting for the covariates: sex, race, health insurance, mortality, year of discharge, prematurity defined as <37 weeks gestational age at birth, diagnosis of sepsis, patent ductus arteriosus, apnea, bradycardia, and other infections. Propensity score matching allows for matching of covariates that predict receiving an intervention (for instance endotracheal intubation) in a cohort study. Propensity score matching has been extensively used in previous neonatal studies. Generalized linear models were used to determine predictors of total hospital charges, costs and LOS. A priori covariates (race, sex, prematurity, and insurance status) were used for model adjustments, in lieu of missing severity scales. Appropriate model assumptions were made for the skewed resource utilization outcome data [13]. Continuous variables presented as mean ± SD. Categorical variables presented as frequency (percentage of the cohort). Statistical significance was considered a two-sided p-value < 0.05. Data was analyzed using SAS (Statistical Software Package v 9.4, Cary, NC).

Results

A total of 42,912,090 cases were identified from 1997 to 2012 after applying sampling weights; 42,316,837 persons were excluded for being >28 days old at admission or without diagnosis of RDS, leaving a target population of 595,254. Further details are outlined in Fig. 1.

Fig. 1.

Study flow chart of neonates with Respiratory Distress Syndrome (RDS). NIV: ICD-9-CM: 9390 ‘Non-Invasive Mechanical Ventilation’. RDS: ICD-9-CM: 769.0 ‘Respiratory Distress Syndrome in Newborn’. Endotracheal Intubation: ICD-9-CM: 96.04 ‘Insertion of Endotracheal Tube’

The majority of the RDS population were male (58.1%, n = 345,942), predominantly white (43%, n = 255,101) and used private insurance (48%, n = 179,771). NIV use occurred in 22.5% (n = 134,132) of the RDS population from 1997–2012. There were statistically significant yearly increases in total hospital charges (1997: $103,158–2012: $220,666), costs (2003: $53,506–2012: $62,996), and LOS (1997: 27.20–32.64 days) per patient discharge in the RDS population. The frequency of patients receiving NIV only has also statistically significantly increased yearly from 8.9% use in 1997 to 21.3% use in 2012. The trend of NIV use is modeled by the equation: y = 0.0196 × +0.046, indicating a positive trend. On the other hand, the use of endotracheal intubation was 50.7% in 1997 and 49.3% by 2012 of the RDS population. The trend of endotracheal intubation is modeled by the equation y = −0.0006 × +0.5098, indicating a negative trend. Table 1 and Fig. 2 provides further data on demographics and resource utilization trends in neonates diagnosed with RDS.

Table 1.

Characteristics of patients with respiratory distress syndrome (N = 595,254)

Description	1997	2000	2003	2006	2009	2012	Total
n (unweighted)	38,945	48,596	66,741	76,029	77,403	73,809	381,523
n (weighted)	87,606	93,114	102,102	112,355	105,806	94,271	595,254
Female (%)	41.6	41.3	41.5	42.2	42.2	42.3	41.9
Race (%)
White	40.4	46.3	38.8	40.0	45.0	47.3	43.0
Black	14.7	14.8	13.0	13.0	16.2	17.6	14.8
Hispanic	9.7	12.4	13.7	14.3	15.7	15.1	13.6
Asian/Pacific	1.5	1.7	1.8	1.9	2.4	3.0	2.0
Other	3.2	4.8	4.9	4.9	5.7	7.5	5.2
Unknown	30.5	20.0	27.9	26.0	15.1	9.4	21.5
Insurance (%)
Medicaid	40.2	39.9	42.4	46.0	49.6	51.1	45.0
Private	51.3	53.5	50.5	46.9	44.3	42.0	48.0
Self-pay	4.0	2.9	3.0	3.2	2.5	2.2	3.0
Other	4.5	3.7	4.1	3.8	3.6	4.7	4.0
Resource utilization
Charges($1/K)a	103.2 ± 252.1	113.0 ± 220.3	134.2 ± 229.1	145.8 ± 233.6	175.8 ± 280.5	220.7 ± 387.9	150.0 ± 283.3
Costs($1/K)a	–	–	53.5 ± 89.5	54.3 ± 87.0	58.4 ± 93.2	63.0 ± 107.3	57.3 ± 95.0
Length of Stay (days)	27.2 ± 48.7	29.3 ± 45.5	28.8 ± 41.4	29.3 ± 41.2	30.5 ± 40.1	32.6 ± 40.2	29.6 ± 42.3
Non invasive ventilation (%)	6.4	9.0	10.2	12.7	13.5	16.9	11.6
Intubation (%)	50.7	49.3	52.5	51.5	51.7	49.0	50.8

^aAdjusting for inflation relative to year 2012

Cost to charge ratio files only available for years 2003–2012

ANOVA test for trend was used to determine trends among continuous variables, Pearson’s chi square was used to determine associations among categorical variables

All variables were statistically significantly associated with year at p < 0.0001 significance

Fig. 2.

NIV and endotracheal intubation frequency

Comparing the intubated group (n = 302,811) to the NIV only group (n = 68,929), there were statistically significantly higher frequencies of prematurity (% of cohort) (NIV: 86.5%, Intubated group: 87.5%, p < 0.0001), sepsis (NIV: 17.8%, Intubated: 24.6%, p < 0.0001), patent ductus arteriosus (NIV: 11.5%, Intubated: 22.3%, p < 0.0001), other infections (NIV: 25.2%, Intubated: 36.7%, p < 0.0001) and death (NIV: 0.4%, Intubated: 7.3%, p < 0.0001) on unadjusted analysis. NIV group was associated with higher frequencies of apnea (NIV: 22.4%, Intubated: 19.8%, p < 0.0001) and bradycardia (NIV: 13.8%, Intubated: 11.6%, p < 0.0001). The intubated group had statistically significantly higher costs (NIV: $45.90k, Intubated: $72.32k, p < 0.0001), charges (NIV: $125.71k, Intubated: $196.48k, p < 0.0001) and length of stay (NIV: 25.23 days, Intubated: 37.18 days, p < 0.0001). These differences are highlighted in Fig. 3.

Fig. 3.

Unadjusted and Adjusted Resource Utilization Trends by Group. 1:1 Propensity score matching was estimated by a logistic regression greedy 8-to-1 digit matching algorithm accounting for the covariates: sex, race, health insurance, mortality, year of discharge, prematurity (<37 weeks gestational age at birth), diagnosis of sepsis, patent ductus arteriosus, apnea, bradycardia, and other infections for adjusted analysis. Costs and Charges adjusted for inflation relative to year 2012

After propensity score matching, these resource utilization associations persisted ((NIV costs: $46.12k, Intubated group costs: $65.03k, n = 29,836 pairs, p < 0.0001), (NIV total hospital charges: $130.30k, Intubated total hospital charges: $192.10k, n = 38,911 pairs, p < 0.0001), (NIV LOS: 25 days, Intubated LOS: 35 days, n = 38,911 pairs, p < 0.0001). On trend analysis, intubated group was associated with increased total hospital charges, costs, and LOS for all years. (Table 2, Fig. 3).

Table 2.

Unadjusted and adjusted demographics and resource utilization of patients receiving NIV only vs endotracheal intubation

	Unmatched population			1:1 Matched population
	NIV	Endotracheal intubation	p-value	NIV	Endotracheal Intubation	p-value
n (unweighted)	46,076	193,773		38,911	38,911
n (weighted)	68,929	302,811		56,448	56,448
Female (%)	42.6	42.3	0.09	42.9	42.9	0.9457
Race (%)			<0.0001			0.0001
White	44.3	41.4		52.7	51.4
Black	15.3	16.3		18.1	18.5
Hispanic	13.9	14.1		16.6	17.4
Asian/Pacific	3.1	2.1		3.8	4.7
Other	7.4	5.0		8.8	8.9
Unknown	16.0	21.1
Insurance (%)			<0.0001			0.9218
Medicaid	42.2	45.5		42.7	42.6
Private	51.5	47.7		51.4	51.5
Self-pay	2.7	2.7		2.7	2.7
Other	3.5	4.0		3.2	3.2
Clinical comorbidities (%)
Prematurity	86.5	87.5	<0.0001	86.5	86.4	0.6429
Sepsis	17.8	24.6	<0.0001	18.7	19.3	0.0065
PDA	11.5	22.3	<0.0001	11.9	11.5	0.0225
Bradycardia	13.8	11.6	<0.0001	14.0	14.3	0.1153
Apnea	22.4	19.8	<0.0001	22.0	22.7	0.0121
Other Infections	25.2	36.7	<0.0001	26.2	25.3	0.0422
Mortality	0.4	7.3	<0.0001	0.4	0.4	0.6145
Resource utilization (mean ± SD)
Charges($1/K)a	125.7 ± 201.9	196.5 ± 320.8	<0.0001	130.3 ± 207.8	192.1 ± 297.3	<0.0001
Costs ($1/K)a	45.9 ± 66.0	72.3 ± 100.8	<0.0001	46.1 ± 67.5	65.0 ± 96.2	<0.0001
Length of Stay (days)	25.2 ± 28.5	37.2 ± 45.8	<0.0001	25.5 ± 28.6	34.6 ± 40.8	<0.0001

^aAdjusting for inflation relative to year 2012

Unmatched populations analyzed with independent t-tests or Pearson chi square for continuous and categorical data respectively

1:1 Propensity score matching was estimated by a logistic regression greedy 8-to-1 digit matching algorithm accounting for the covariates: sex, race, health insurance, mortality, year of discharge, prematurity (<37 weeks gestational age at birth), diagnosis of sepsis, patent ductus arteriosus, apnea, bradycardia, and other infections

On multivariate regression analysis, NIV was significantly associated with lower length of stay (20 days, p < 0.0001), total hospital charges ($76.54k, p < 0.0001), and costs ($29.70k, p < 0.0001) compared to intubated patients after adjustments. (Supplementary Table 1).

Discussion

From 1997–2012, there was a three-fold increase in NIV use among neonates with RDS. Non-invasive ventilation was significantly associated with decreased length of stay (9 days), total hospital charges ($61,800), and costs ($18,910) per discharge compared to neonates who were endotracheal intubated neonates with RDS.

There is a statistically significant upward trend in the utilization of NIV in RDS neonates, indicated by the 12.4% increase in NIV among this population from 1997–2012. There are several studies that support our findings of increasing usage of NIV to treat RDS in neonates. The aforementioned research is diverse in geographic location and moreover, study design, ranging from observational studies to randomized, controlled trials [11]. Although the literature on NIV use in RDS infants is heterogeneous in methodology due to the various modes of NIV administration, a majority have found positive outcomes associated with NIV including; decreased incidence of BPD, reduced mortality and shorter duration of supplemental oxygen [14–17]. When considering the growing evidence of positive outcomes associated with NIV and evidence for detrimental outcomes associated with endotracheal intubation [9], it is logical that our study findings are consistent with the ongoing research on this subject.

An additional finding of interest between the two cohorts was the statistically significant discrepancy of nearly $20,000 and $62,000 in costs and hospital charges, respectively. The decreased cost and charges were both associated with the non-invasive ventilation cohort when compared to the endotracheal intubation cohort. Due to the limited literature related to this topic, it is difficult to ascertain whether this finding is indiscriminate or novel. One related study of interest is the implementation of protocols that standardizes respiratory care management amongst neonates in an effort to reduce rates of BPD. One such protocol involved the strict use of bCPAP when possible, with the rationale being the cost-effectiveness of bCPAP in comparison to mechanical ventilation. This study found that there were decreased equipment costs related to the change in protocol without any adverse consequences [18]. The same logic can be applied to the results of our analysis of resource utilization for the two cohorts. Should further studies provide similar data, non-invasive ventilation could emerge as a superior alternative to the more costly method of endotracheal intubation, proving benefits in both aspects of clinical care and resource utilization.

Another important finding was the shortened length of stay associated with the non-invasive ventilation cohort when compared to the intubated cohort. Although there are few published data on this topic as well, one retrospective study found a statistically significant decrease in the length of hospital stay in premature infants after shifting treatment towards NIV versus invasive mechanical ventilation [19]. Our study also found that the average length of stay among the NIV group was about 12 days less than that of the intubated group. The shorter length of stay in the cohort that received NIV could prove a contributing factor for the decreased cost and charges associated with the group as well. Although the endotracheal intubation group did have statistically significant higher rates of PDA, prematurity, sepsis and other infections, the increased length of stay in this group remained following propensity score matching. This correction for confounders suggests an association between endotracheal intubation for RDS and increased length of stay in our neonates.

There are several limitations associated with this study that warrant discussion. First, retrospective data is susceptible to the issues of temporality. For example, to distinguish which came first, intubation or non-invasive ventilation would not be possible. Furthermore, the study population was defined by a single ICD-9-CM code. Other respiratory diagnosis codes considered included; ICD-CM-9 77084 “respiratory failure of newborn” and ICD-CM-9 77089 “respiratory problems after birth”. However, including all three codes created duplicity and over-representation of diagnoses. HCUP-KID’s procedures and diagnoses (ICD-9-CM) are coded based on physician billing data and data integrity is dependent among individual hospitals. Another limitation was the lack of information regarding maternal demographics and diagnoses as well as the medications provided during inpatient stay. For example, there is no way to determine the associations between INSURE and resource utilization. Furthermore, HCUP-KID does not include medications. It is unknown whether prenatal steroids or surfactant use will bias these results towards or against the null. The lack of medications within this dataset, introduces a possible source of information bias. Gestational age coding was not available for 77% of the sample population while birth weight coding was not available for 60% of the population. To minimize bias, gestational age and birthweight were not used as covariates. As a solution for the possibility of confounding factors on study results, propensity score matching was utilized.

In conclusion, our analysis of resource utilization among RDS neonates using unadjusted methods, multivariate regression and matched propensity score methods indicate that NIV use was associated with shorter LOS, lower total hospital charges and costs when compared to endotracheal intubation and invasive mechanical ventilation. Additional prospective and retrospective research with more precise timed measurements of intubation and non-invasive ventilation, as well as medication data are needed to determine whether these associations persist. It is essential that further studies related to neonatal care include resource utilization in addition to clinical impact. Such research could be a motivating factor for the development and implementation of new protocols that could greatly benefit NICUs in the United States in the face of the evolving healthcare climate. In addition, these protocols would be extremely beneficial in resource-poor areas, where effective resource allocation is crucial.