Cost-Effectiveness of Nasal High Flow Versus CPAP for Newborn Infants in Special-Care Nurseries

The HUNTER trial was a multicenter, randomized, noninferiority trial, conducted in 9 Australian nontertiary special care nurseries in 2015–2017 (Supplemental Information 1).⁹  All participating centers managed at least 2000 births per year and had extensive experience using CPAP before the trial but did not routinely use nHF. Newborn infants with respiratory distress, born at ≥31 weeks’ gestation and with birth weight ≥1200 g, <24 hours old, and who required noninvasive respiratory support and/or supplemental oxygen for >1 hour were eligible. The primary outcome of the trial was treatment failure within 72 hours of randomization as determined by prespecified failure criteria available in the published protocol and results articles.^9,14 

The main characteristics of the trial design are summarized in Fig 1. Infants randomly assigned to nHF who met treatment failure criteria while receiving maximal therapy were changed to CPAP as a “rescue” therapy. Infants in whom nHF treatment failed, but who were then rescued by back-up CPAP, did not require escalation of care such as endotracheal intubation and/or transfer to a tertiary NICU. Infants in whom CPAP treatment failed had care escalated.

CPAP was delivered via a “bubble” CPAP system by using standard circuits and nasal prongs or masks (Teleflex Medical, Mascot, Australia; Fisher & Paykel Healthcare, Auckland, New Zealand). The “Optiflow Junior” (Fisher & Paykel Healthcare) system was used to deliver nHF.

This economic evaluation was prospectively planned alongside the HUNTER trial and was approved by the Royal Children’s Hospital Human Research and Ethics Committee, Melbourne, Australia, and by the participating sites.

The cost-effectiveness analysis was conducted from a hospital care perspective with a time horizon from birth to 12 months of age. The primary analyses were conducted by using the per-protocol HUNTER sample to maximize the match between resources used and outcomes achieved, with the intention-to-treat sample presented in a sensitivity analysis.

The costs of care included the inpatient costs at nontertiary special care nurseries, the inpatient costs at tertiary NICUs, and the costs of interhospital transfers including back transfers. Inpatient costs at the 9 enrolling special care nurseries were obtained from the administrative hospital costing units. Inpatient costs at other special care nurseries in the event of transfer were imputed on the basis of infant length of stay multiplied by the average inpatient cost per day obtained from cost of subsequent admissions at participating nurseries (cost of other special nursery stays for 44 infants were imputed). Costs of NICU stay were estimated by using the NICU length of stay multiplied by an estimated average cost per day based on previous research.¹³  Costs of infant transfer to tertiary NICUs and nontertiary nurseries were calculated on the basis of transfer mode (road, fixed wing aircraft, and rotary helicopter), locations of the 2 hospitals (metropolitan or regional), and transfer urgency (emergency or nonemergency).¹⁵ 

Because CPAP was the standard treatment in participating centers, a small proportion of infants allocated to nHF received a brief period of prerandomization CPAP. The cost of prerandomization consumable CPAP equipment for infants in the nHF group was excluded because the prerandomization CPAP treatment would be unnecessary outside the trial environment where patients would directly receive the treatment of choice. The unit prices of the CPAP consumable equipment were sourced from the local equipment representatives. No adjustment was made for the capital investment, because the bubble CPAP and nHF equipment used in the trial shared the same humidifier base, heater wire, and temperature probe. For the CPAP group, we assume that prerandomization CPAP does not affect the overall cost, because no change of circuits or interfaces were required.

The effectiveness outcomes considered most suitable for the economic evaluation were NICU transfer rate and endotracheal intubation rate. These 2 outcomes not only match with the overall resource use but also are important to patients, families, and clinicians.

We compared the cost-effectiveness of CPAP versus nHF therapy under 2 scenarios: (1) with rescue CPAP backup available as in the HUNTER trial (trial scenario), and (2) as sole primary support (hypothetical scenario). For the trial scenario, cost and effectiveness outcomes as obtained from the trial and hospital cost databases were used directly. For the hypothetical scenario, the cost and effectiveness outcomes for infants in whom nHF treatment failed and who were rescued by back-up CPAP were adjusted as if back-up CPAP were not available. More specifically, we assume that infants who were rescued by CPAP would receive intubation and/or transfer following the same pattern as observed for those in the CPAP group in whom CPAP treatment had failed (a random draw from the observed intubation and transfer distributions from infants in the CPAP group in whom CPAP treatment failed were used in which the distributions were constructed by using the mean and SE of the CPAP treatment failure outcomes). Costs were allocated accordingly if infants were instead intubated and/or transferred in the hypothetical scenario (a random draw from the observed distributions of costs of stay associated with intubation and/or transfer was used for individual infants).^16,17  The incremental cost-effectiveness ratio, defined as the difference in cost (CPAP versus nHF) divided by the difference in effectiveness (CPAP versus nHF), was to be calculated and interpreted as the average incremental cost associated with one additional intubation or transfer avoided.

To account for sampling uncertainty, probabilistic sensitivity analysis was performed by using a bootstrapping method with 1000 replications drawing from cost and outcomes at patient level. The bootstrapped results were presented by using cost-effectiveness planes, and each of the 1000 dots represents 1 mean cost and effectiveness outcome from a bootstrapped sample with replacement of the original sample.¹⁸  One-way sensitivity analysis was performed to test the methodologic assumptions and sampling variability for each of the following independent scenarios with the results presented using cost-effectiveness acceptability curves, including (1) using the intention-to-treat sample; (2) family cost incorporated including travel, hospital parking, and accommodation (detailed assumptions documented in Supplemental Table 6); (3) infants whose main reason for transfer to a NICU was recorded as “nonrespiratory” were excluded (CPAP 10, nHF 9); (4) infants randomly assigned to nHF who received prerandomization CPAP were excluded (nHF 51); (5) infants who do not have 12 months cost data capture because of hospitals providing varying lengths of costing data depending on local financial year accounting policies were excluded (CPAP 17, nHF 14); (6) time horizon limited to 6 months; (7) time horizon limited to the trial duration (initial hospital admission until death or first discharge from the hospital). The acceptability curves were constructed by counting the proportion of bootstrap replicates that are acceptable under various willingness-to-pay levels.¹⁹  The same one-way sensitivity analyses were conducted for the trial scenario and the hypothetical scenario, respectively. In addition, 2 post hoc subgroup analyses were performed including (1) only the centers where the nearest NICUs were ≥70 km away (3 hospitals), and (2) only the centers where there were NICUs within 30 km (6 hospitals).

To adjust for the differences between timing of valuations, inpatient costs incurred before 2019 were adjusted to 2019 prices by using the consumer price indices published by the Australian Bureau of Statistics. Costs are presented in Australian dollars (1US$ = 1.472A$, 2019 Purchasing Power Parity published by the Organization for Economic Co-operation and Development Statistics). Analyses were conducted and reported after the cost-effectiveness analysis alongside clinical trials good practice guideline¹⁹  and the Consolidated Health Economic Evaluation Reporting Standards reporting guideline.²⁰  The Stata statistical software package (version 14.0) (Stat Corp, College Station, TX) was used.

There were 754 infants included in the HUNTER trial. One infant did not have administrative costing data available. The linked sample contains 753 infants with inpatient records available from birth to 6 months of age, and 719 (95%) infants have costs available over the first 12 months from the participating hospitals.

In the primary per-protocol analysis, 676 infants were included, and there were no statistically significant (P < .05) differences in the baseline characteristics between the 2 treatment groups (Table 1). For the 31 (5%) infants who were not yet 12 months of age in the data capture period, all available costing records were used. The trial outcomes that are most relevant to this study are also provided in Table 1. Table 2 reports cost per inpatient day and other unit cost inputs.

The cost outcomes of the trial are presented in Table 3, disaggregated by type of care. Overall, the cost differences were neither statistically significant nor economically meaningful when there is back-up CPAP available for nHF.

The cost-effectiveness outcomes for the trial and the hypothetical scenarios are summarized in Table 4. When back-up CPAP is available as implemented in the trial, the differences in cost between the groups were neither statistically significant nor economically meaningful. Similarly, there are no statistically significant or clinically meaningful differences in the effectiveness outcomes. Thus, when back-up CPAP is available, there is no reason to make a decision between the 2 treatment algorithms. In the hypothetical scenario where the 2 therapies were the sole primary supports, CPAP was more effective and on average cheaper, and thus was the preferred strategy, and calculating an incremental cost-effectiveness ratio is therefore redundant.¹⁸ 

The probabilistic sensitivity analysis is presented in Fig 2, and the results are consistent with the primary analysis that the 2 treatment algorithms are comparable when back-up CPAP is available; however as sole primary support, CPAP is better and on average cheaper. The one-way sensitivity analysis is presented in Supplemental Information 2), and the results are consistent with the primary analysis except for the intention-to-treat scenario. In the intention-to-treat scenario, CPAP has a higher probability of being cost-effective even when nHF has back-up CPAP available. It is worth noting that the cost and effectiveness outcomes do not match well under the intention-to-treat scenario compared with the primary per-protocol scenario.

The post hoc subgroup analysis grouping hospitals based on the distance to NICUs were provided in Table 5. The results revealed a difference in the observed nHF efficacies based on distance from a NICU. More specifically, nHF with back-up CPAP appeared to be nonstatistically cheaper in the subgroup of centers <30 km away from the nearest NICU. In centers where the nearest NICU was at least 70 km away, CPAP was cheaper and more effective.

Neonatal respiratory distress syndrome was the most costly in-hospital pediatric condition in the United States between 2004 and 2009.²¹  In assisting resource allocation and clinical practice, evidence taking into account both the cost and the effectiveness of treatment alternatives is needed. In this study, we compared the cost-effectiveness of CPAP and nHF and found that as sole primary support, CPAP is more effective and on average cheaper, and thus is the dominant strategy. In special care nurseries with staff experienced in using CPAP and who wish to choose only 1 respiratory support mode, CPAP would be the recommended strategy.

Nevertheless, the results may not preclude a role for nHF in treatment algorithms. Nasal HF treatment was successful in >80% of infants, although none of the 9 participating special care nurseries were experienced with its use before the trial. This observation is consistent with the perception that nHF is easy to use. Because using nHF with CPAP backup produces almost equal cost and effectiveness results, when both CPAP and nHF are available, using nHF as first-line therapy may be acceptable.

The post hoc subgroup analysis by geographic distance from NICUs revealed a difference in the observed nHF efficacies based on distance from a NICU. In the subgroup of centers <30 km away from the nearest NICU, nHF with back-up CPAP appeared to be nonstatistically cheaper. This is in line with the economic evaluation results for the HIPSTER trial in tertiary NICU settings.¹³  This could be due to infants in the nHF group being “less sick” when nHF treatment failure occurred or when nHF was weaned down compared with infants in the CPAP group. The observed nonsignificant difference was not surprising given that the trial was not powered to detect statistically significant differences in cost. However, the average cost difference is economically meaningful and approximates 1 inpatient day and may be worth further investigation. In centers where the nearest NICU was at least 70 km away, CPAP was both cheaper and more effective and, thus, the preferred strategy. One speculation is that this could reflect a difference in clinician perception of risk, with centers further away from a NICU potentially more risk averse and more likely to change therapy from nHF to back-up CPAP and/or escalation of care if a patient is deteriorating when using nHF. This signals that the cost of adopting nHF might be different depending on the thresholds for transfer in different units. We have not detected any meaningful differences in infant or mother characteristics, meaning that the observed geographical distance results are unlikely to reflect a difference in patient characteristics. Although the subgroup results may be vital to decision-makers, we caution that the analysis was post hoc and there is a probability of finding statistically significant difference due to random variation.²⁰ 

Several limitations are identified. The trial was conducted in centers with >2000 deliveries each year, allowing maintenance of skills required to provide CPAP safely. The results should be interpreted with care in resource-limited settings where the level of medical and nursing expertise with CPAP is different.²²  We have adopted a hospital care perspective and have not surveyed family experience and cost. Nevertheless, we incorporated a limited set of family costs using realistic assumptions in sensitivity analysis, and the results were unchanged. We have relied on hospital costing records from the participating centers after infants’ first discharge from the hospital until 12 months of age. Admission to other hospitals is possible; however, readmissions to tertiary hospitals were rare (observed to be 1 out of 274 preterm infants),¹³  and the participating hospitals are most likely to be the hospitals providing both maternal and follow-up care for catchment reasons.

We conclude that CPAP is more effective and on average cheaper than nHF when used as sole primary support for newborn infants with respiratory distress cared for in Australian nontertiary special care nurseries, and thus it is the dominant strategy. Special care nurseries choosing to use only 1 device may apply CPAP as primary respiratory support. However, using nHF as first-line therapy with back-up CPAP produced equivalent cost and effectiveness results and may be acceptable for some units with both devices available.

We thank the costing managers from the respective institutions that extracted the hospital cost data, including but not limited to Melissa Blake, Kelvin Heard, Paul Hougham, Kim Lim, Isaac Marshall, and Andrew McDonell.

CPAP

continuous positive airway pressure

nHF

nasal high-flow