Impact of using the International Risk Scoring Tool on the cost-utility of palivizumab for preventing severe respiratory syncytial virus infection in Canadian moderate-to-late preterm infants

Abstract

Background and objective

To assess the cost-utility of palivizumab versus no prophylaxis in preventing severe respiratory syncytial virus (RSV) infection in Canadian moderate-to-late preterm (32–35 weeks’ gestational age) infants using an (i) International Risk Scoring Tool (IRST) and (ii) Canadian RST (CRST).

Methods

A decision tree was developed to assess cost-utility. Infants assessed at moderate- and high-risk of RSV-related hospitalization (RSVH) by the IRST or CRST received palivizumab or no prophylaxis and then progressed to either (i) RSVH; (ii) emergency room/outpatient medically attended RSV-infection (MARI) or (iii) were uninfected/non-medically attended. Infants admitted to intensive care could incur mortality (0.43%). Respiratory morbidity was accounted in all uninfected surviving infants for 6 years or 18 years (RSVH/MARI). Palivizumab efficacy (72.2% RSVH reduction) and hospital outcomes were from the Canadian CARESS, PICNIC and RSV-Quebec studies. Palivizumab costs (50 mg: CAN$752; 100 mg: $1,505) were calculated from Canadian birth statistics combined with a growth algorithm. Healthcare/payer and societal costs (May 2022; 1.5% discounting) were included.

Results

Cost per quality-adjusted life year (QALY) was $29,789 with the IRST (0.79 probability of being <$50,000) and $15,833 with the CRST (0.96 probability). The model was most sensitive to utility scores, long-term sequelae and palivizumab cost. Vial sharing improved the incremental cost-utility ratio (IRST: $22,319; CRST: $9,231).

Conclusions

Palivizumab was highly cost-effective (vs no prophylaxis) in Canadian moderate-to-late preterm infants using either the IRST or CRST. The IRST has fewer risk factors than the CRST (3 vs 7, respectively), captures more potential RSVHs (85% vs 54%) and provides another option to guide cost-effective RSV prophylaxis in Canada.

Graphical Abstract

Keywords:

• Canada
• cost-effectiveness
• cost-utility
• palivizumab
• respiratory syncytial virus
• RSV

JEL CLASSIFICATION CODES:

• I10
• I1
• I
• I18
• L65
• L6
• L

Introduction

Respiratory syncytial virus (RSV) is a major childhood pathogen estimated to cause approximately 12.9 million lower respiratory tract infections (LRTI) resulting in 2.2 million hospitalizations (RSVH) and >66,000 deaths annually in children <1 year worldwide¹. RSV-LRTI in premature infants and young children has a temporal association with long-term wheezing and asthma, which can persist throughout childhood and late adolescence^2–4, and may result in obstructive lung function in adulthood^5,6. It is well-established that infants born moderate-to-late preterm (32–35 weeks’ gestational age; wGA) are among those at the greatest risk of RSV-LRTI, with approximately 4% of these infants being hospitalized^7,8.

In Canadian infants born moderate-to-late preterm, recently published studies have estimated that the average direct healthcare cost of an RSVH to be CAD$16–28,500^9–11. In Quebec alone, the direct cost of RSVHs across two RSV seasons was estimated at $3.7 million¹⁰. Importantly, the true financial burden of RSV-LRTI is likely to be considerably greater, as these estimates do not account fully for medically-attended RSV infections (MARI) – those that require medical attention, but not hospital admission – and the costs of managing the long-term sequelae. RSV-LRTI has also been associated with a considerable financial (indirect costs) and quality of life (QoL) impact on the families of affected infants^12–14.

At present, immunoprophylaxis with the monoclonal antibody palivizumab (Synagis; AstraZeneca Canada) remains the only widely available intervention to prevent RSV-LRTI in high-risk infants, including those born at ≤35wGA. Prior to 2015, guidelines from the Canadian Paediatric Society (CPS) recommended palivizumab prophylaxis for those otherwise healthy infants born 32–35wGA assessed at moderate- (7.1%) or high- (18.7%) risk of RSVH by the Canadian Risk Scoring Tool (CRST)^15,16. This was based on a cost-utility analysis published in 2010 that reported an incremental cost-utility ratio (ICUR) versus no prophylaxis for high-risk infants to be $5,274/quality-adjusted life year (QALY) and $34,438/QALY for moderate-risk infants¹⁷. Since 2015, however, the CPS has recommended that palivizumab use be largely restricted to those otherwise healthy infants born ≤30wGA and <6 months old (reaffirmed in 2021)¹⁸. Guidelines from the Canadian National Advisory Committee on Immunization (NACI) are similar to those from the CPS although they state that palivizumab may also be considered for premature infants of 30–32wGA and age <3 months who are at high-risk for RSV infection from day care attendance or presence of another preschool child or children in the home¹⁹. A recent review has since reported marked differences in prophylaxis policies for moderate-to-late preterm infants across Canada, resulting in inequities in care²⁰. This variability in policies may reflect limitations in funding and human resources to deliver prophylaxis programs, differing interpretation of the evidence for palivizumab efficacy, variation in regional RSV epidemiology, and/or local pressures or advocacy²⁰.

Since the publication of the last Canadian cost-utility analysis of palivizumab, a new International RST (IRST) that can identify 32–35wGA infants at moderate- (3.3%) or high- (9.5%) risk of RSVH has been developed and validated²¹. There has also been a wealth of new data published on the epidemiology and burden of RSV-LRTI for inclusion in cost-analyses. The primary objective of this study was to undertake a new cost-utility analysis of palivizumab prophylaxis guided by the IRST in Canadian moderate-to-late preterm infants using the latest clinical data. A secondary objective was to provide an updated analysis of the cost-utility of palivizumab using the CRST.

Methods

This cost-utility analysis assessed whether palivizumab use in moderate-to-late preterm infants, defined as birth at 32–35wGA, who are at moderate- and high-risk of RSV-LRTI, as determined by use of the IRST²¹ at birth, is a cost-effective strategy versus no prophylaxis in Canada. Costs were considered from the societal perspective in the base-case and within all scenario analyses unless specifically stated otherwise. The cost-utility analysis is reported commensurate with the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) 2022²².

Systematic review

Prior to building our updated model, a systematic literature review was undertaken. The research question addressed was “How has the health economics of palivizumab in moderate-to-late preterm infants been modelled and assessed?”. The objective of the review was to inform our model design by identifying the key characteristics of previous cost-effectiveness and cost-utility models, including the varied modelling methodologies, model structures and parameters used, the most influential variables, and the use of novel approaches to handle important calculations such as infant weight at the time of palivizumab dosing. The review identified 20 cost-analyses of palivizumab use in infants born at 32–35wGA or a subset thereof (Supplementary Materials 1).

Model structure

A decision tree cost-utility model was employed wherein prophylaxed/non-prophylaxed infants have an RSV-LRTI resulting in either hospitalization, emergency room/outpatient attendance (MARI), or were uninfected/non-medically attended (Figure 1). Hospitalized infants could be subsequently admitted to the intensive care unit (ICU) or not, and those admitted to the ICU were assumed to either die or survive. For surviving infants, the possibility of a second identical cycle of RSVH was included. Any surviving infant, regardless of RSV infection status, could potentially experience some degree of childhood wheezing and/or asthma (up to 18 years in the base case for those with RSVH/MARI). A life-time horizon was used to adequately capture the effects of mortality and long-term respiratory sequelae associated with RSV. The model was built using the Canadian Agency for Drugs and Technologies in Health (CADTH) and International Society for Pharmacoeconomics and Outcomes Research (ISPOR) guidelines^23,24.

Figure 1. Decision tree describing the clinical pathway used in the model. Nodes represent points where more than one event is possible; the square node represents the decision addressed by the model. Branches represent the possible events that patients may experience. Triangles represent the decision tree endpoints. Abbreviations. ICU, intensive care unit; LTRM, long-term respiratory morbidity; MARI, medically-attended RSV infection; RSV, respiratory syncytial virus; RSVH, RSV-related hospitalization.

Model assumptions and inputs

IRST

The IRST was developed to predict the risk of RSVH in infants born 32–35wGA and thereby improve the targeting of RSV prevention²¹. The IRST was derived using logistic regression from a pooled dataset of six observational studies (including the Pediatric Investigators Collaborative Network on Infections in Canada [PICNIC] study⁷) which included risk factor data on 13,475 infants born 32–35wGA, 484 of whom had RSVH at ≤1 year (Table S1 in Supplementary Materials 2)²¹. The IRST included three risk factors: birth between three months before to two months after the start of the RSV season; household smoking and/or smoking while pregnant; and siblings (excluding those as part of a multiple birth with the subject) and/or day care attendance (Figure S1 in Supplementary Materials 2). Cut-off scores for RSVH were ≤19 for low- (RSVH rate: 1.0%), 20–45 for moderate- (3.3%), and 50–56 (9.5%) for high-risk infants (Table S2 in Supplementary Materials 2). The IRST had an area under the receiver operating characteristic curve (AUROC) of 0.773 (sensitivity 68.9%; specificity 73.0%)²¹, with a score of 1 representing perfect predictive accuracy.

Palivizumab efficacy

The overall efficacy of palivizumab was derived from two Canadian studies: CARESS (Canadian Registry of Synagis^25,26) and PICNIC⁷. CARESS collected data on 25,003 children receiving palivizumab between 2005 and 2017 and reported an RSVH rate of 1.0% (70/6,894) in 32–35wGA infants^26,27. The RSVH rate in 33–35wGA infants not receiving prophylaxis in the PICNIC study was 3.6% (66/1,832)⁷, equating to an overall efficacy rate for palivizumab of 72.2%. The impact of palivizumab on the incidence of respiratory morbidity was calculated using data from Simoes et al.²⁸, Blanken et al.²⁹ and Yoshihara et al.³⁰

RSVH

To reflect differences in the risk of hospitalization across the scoring tools and risk groups, the RSVH rates for infants not receiving palivizumab were drawn directly from the published data for each RST, since these populations excluded infants that received prophylaxis (Table 1)^16,21. The RSVH rates in those infants receiving prophylaxis was calculated by applying the efficacy of palivizumab (72.2% risk reduction) to the non-prophylaxed rates. In the base case, the RSVH rates for prophylaxed versus non-prophylaxed infants scored at moderate- or high-risk by the IRST were 1.8% and 6.3%, respectively.

Table 1. RSV-related hospitalization (RSVH) rates associated with risk scoring tools.

Risk band	No palivizumab	Palivizumab^a
International Risk Scoring Tool (IRST)^21
High	9.5%	2.6%
Moderate	3.3%	0.9%
Low^b	1.0%	0.3%
Moderate + high	6.3%	1.8%
Canadian Risk Scoring Tool (CRST)^16
High	18.7%	5.2%
Moderate	7.1%	2.0%
Low^b	1.7%	0.5%
Moderate + high	10.9%	3.0%
Standardized CRST^c
High	11.2%	3.1%
Moderate	4.3%	1.2%
Low^b	1.0%	0.3%
Moderate + high	6.5%	1.8%

Abbreviations. RSV, respiratory syncytial virus; RSVH, RSV-related hospitalization.

^aRSVH rate in the prophylaxed arm for each risk group was calculated by applying the relative risk reduction (72.2%) calculated from the baseline risk of RSVH in the palivizumab versus no-palivizumab arms (1.0%^26,27 vs 3.6%⁷, respectively).

^bIncluded for reference, no scenario considers prophylaxis in the low-risk group.

^cRSVH risk standardized with that in low-risk group (1%) for IRST.

For infants not receiving palivizumab the incidence of RSVH, ward length of stay (LOS) and ICU admission rate and LOS were not available in the PICNIC study and thus data were drawn from the studies by Papenburg et al.^10,31, which examined healthcare utilization and costs of RSV among 33–35wGA infants from Quebec (Table 2). For infants that received palivizumab and had RSVH, the equivalent information from CARESS²⁷ was used, except for ward LOS, which was again based on data from Papenburg et al.³¹. For any infants admitted to ICU, a mortality rate of 0.43% was applied^32,33.

Table 2. Input parameters associated with RSVH and MARI.

Parameter	Point estimates		Source(s)
	Palivizumab	No palivizumab
RSVH^a
Overall rate^b	1.0%^27	3.6%^7	CARESS^27, PICNIC^7
ICU rate	18.9%^27	19.2%^31	CARESS^27, Papenburg et al.^31
Ward LOS, mean days	6.0^31	6.0^31	Papenburg et al.^31
ICU LOS, mean days	4.4^27	7.0^10	CARESS^27, Papenburg et al.^10
Utility whilst in hospital with RSV	0.60^13^,^41	0.60^13^,^41	Leidy et al.^13,Weiner et al.^41
Utility post discharge	0.88^42	0.88^42	Greenough et al.^42
No sequelae	0.79^43	0.79^43	Chiou et al.^43
Long-term sequelae	0.43^32^,^33	0.43^32^,^33	Wang et al.^32^,^33
Mortality^c
MARI
Rate outpatients only	2.48%^7^,^34^,^35	12.19%^7^,^34^,^35	PICNIC^7, Carbonell et al.^34, REPORT^35 Greenough et al.^42
Rate outpatients & ED	0.42%^7^,^34^,^35	2.08%^7^,^34^,^35
Rate ED only	0.05%^7^,^34^,^35	0.25%^7^,^34^,^35
Utility	0.95^42	0.95^42
MARI/no RSV^d
Utility	0.95^42	0.95^42	Greenough et al.^42
No sequelae or respiratory symptoms	0.79^43	0.79^43	Chiou et al.^43
Long-term sequelae or respiratory symptoms

Abbreviations. CARESS, Canadian Registry of Synagis; ED, emergency department; ICU, intensive care unit; LOS, length of stay; MARI, medically-attended RSV infection; PICNIC, Pediatric Investigators Collaborative Network on Infections in Canada study; REPORT, Respiratory Events Among Preterm Infants Outcomes and Risk Tracking study; RSV, respiratory syncytial virus; RSVH, RSV-related hospitalization.

^aFirst and subsequent RSVHs.

^bUsed to calculate the efficacy of palivizumab (72.2% relative risk reduction; N = 8,654).

^cApplied only to patients in ICU.

^dInfants without an RSV infection or one not requiring medical management.

Medically-attended RSV infection (MARI)

It was assumed that 2.95% of infants receiving prophylaxis would experience MARI, which was the rate among infants >32–35wGA in the palivizumab arm of the motavizumab trial³⁴. The proportion of these infants presenting to outpatients only, emergency department (ED) only, or both was calculated by applying the rates for these outcomes from the US REPORT (Respiratory Events Among Preterm Infants Outcomes and Risk Tracking) study³⁵ in infants born 32–35wGA who did not receive prophylaxis (Table 2). For non-prophylaxed infants, the rates of MARI observed in REPORT³⁵ were adjusted according to the relative difference in the RSVH rate in this study versus that observed in PICNIC⁷ (3.47% vs 3.6%, respectively).

Long-term respiratory morbidity

The algorithm by Sanchez-Luna et al.³⁶ was used to calculate the effect of palivizumab on long-term respiratory sequelae. Sanchez-Luna et al.³⁶ estimated the risk reduction associated with prophylaxis versus no prophylaxis for 32–35wGA infants up to six years old, using data from the Spanish SPRING³⁷ study combined with three palivizumab studies^28–30. The same approach was adopted for the current model for years 0–6, and extended to 18 years of age using prospectively collected data on rates of respiratory morbidity in Swedish infants published by Sigurs et al.^2,38–40. To adjust for differences in the baseline rates of respiratory morbidity, the Sigurs et al. data^38–40 was fitted to the lower rates observed in SPRING³⁷. This approach was chosen because SPRING³⁷ was specific to infants born 32–35wGA, whereas Sigurs et al.² considered all infants aged <1 year with a RSVH and therefore provided a less gestational age-specific estimate.

For infants with RSVH or MARI, long-term respiratory morbidity was assumed to persist until 18 years of age, albeit with lower rates for the latter (Table S3 in Supplementary Materials 2). Infants without an RSV infection or an RSV infection not requiring any medical management were assumed to have some background respiratory morbidity for up to six years, using the same rates as for MARI.

Utilities

To estimate the impact of RSVH on infants’ QoL, the utility value (0.60) used in the US analysis by Weiner et al.⁴¹ was adopted and applied only for the duration of the hospital stay (Table 2). This value was derived from a prospective study of 46 children born at ≤35wGA with RSVH and 45 chronologic age-matched controls from the same geographic area with a history of prematurity. The study used various measures to assess the QoL impact of hospitalization on the infant and parent¹³. Following discharge, a utility value of 0.88 was applied, which was taken from the UK study by Greenough et al.⁴², which followed premature infants (median 27wGA) with chronic lung disease and RSVH (and a control group without RSVH) until five years of age. For infants with MARI without long-term respiratory morbidity and those without an RSV-LRTI or an RSV infection not requiring medical management, the control group utility value (0.95) from Greenough et al.⁴² was used. The values from Greenough et al.⁴² were applied until five years of life, after which perfect health was assumed.

For the impact of long-term respiratory morbidity on QoL, no RSV-specific utility data were identified, so a value of 0.79, as reported in children with mild-and-moderate asthma symptoms, was utilized⁴³.

Costs

Costs were calculated on an individual infant level and reflected the acquisition and administration costs associated with palivizumab as well as the costs of RSVH, MARI, ICU admission, the ongoing costs associated with respiratory morbidity, and societal costs (Table 3). All costs were taken from publicly available Canadian sources^10,44–49 and were adjusted using the Consumer Price Index for Canada (CPI)⁵⁰ as of May 2022, with a 1.5% discount applied commensurate with the CADTH guidelines²³.

Table 3. Direct and indirect costs.

Parameter	Cost CAD$^a	Units	Source(s)
Healthcare/payer (direct) costs
Palivizumab
50 mg vial	752	Lowest combination of vials per infant weight	Canadian list price
100 mg vial	1,505		Canadian list price
Nurse administration	14.37	1 per injection	Papenburg et al.^10
Preadmission healthcare contact	214.16	1 per RSVH^b	Papenburg et al.^10
RSVH total stay (excluding ICU)	8,352.87	1 per RSVH^b	Papenburg et al.^10
ICU (per day)	5,747.00	Palivizumab: 4.4 days	CARESS^27
		No palivizumab: 7.0 days	Papenburg et al.^10
MARI
Outpatient visit	Initial: 175.40;	1 per affected infant^b	Ontario Ministry of Health^44
	Follow up: 91.35		Canadian Institute for Health Information^45
ED visit	336.56
Respiratory morbidity (per annum)	1,116.45	1 per affected infant^c	Ismaila et al.^46
Societal (indirect) costs
Palivizumab administration
Transport	74.86	1 per infant receiving palivizumab	Canada Revenue Agency ($/km)^47, assumes 30 km round trip per injection
Missed work	173.96		Statistics Canada (salary)^48, assumes 3 h missed per injection for 50% injections
RSVH
Missed work	1213.31	1 per infant with RSVH^b	Mitchell et al.^14 & Statistics Canada (salary)^48^d
Childcare	116.04		Mitchell et al.^14
Transport	124.86		Mitchell et al.^14
Other out of pockets	341.81		Mitchell et al.^14
MARI attendance		1 per infant with MARI^b	Canada Revenue Agency ($/km)^47, 30 km round trip
Transport	18.30		Statistics Canada (salary)^48
Missed work	42.52		3 h per MARI attendance
Loss of earnings following death	2,178,497	1 per infant suffering mortality^b	Statistics Canada (salary^48 and & employment^49), 49 years lost productivity

Abbreviations. CAD, Canadian dollars; CARESS, Canadian Registry of Synagis; ED, emergency department; ICU, intensive care unit; MARI, medically-attended RSV infection; RSV, respiratory syncytial virus; RSVH, RSV-related hospitalization.

^aCosts adjusted to 2022 levels using Consumer Price Index, May 2022⁵⁰.

^bSee Table 2 for rates.

^cSee Table S3 in Supplementary Materials 2 for rates.

^dCalculated by multiplying hours of work lost as reported directly by parents¹⁴ by the Canadian average hourly wage.

The cost of palivizumab was calculated using the Canadian list price (50 mg: $752; 100 mg: $1,505) and the lowest combination of vials per infant weight, assuming no vial sharing, and 100% compliance. Using the average birthweight of a Canadian 32–35wGA infant (2,308 g)⁵¹ and assuming an even spread of births across the year, infant weight at palivizumab administration was predicted using the growth algorithm described by Narayan et al.⁵². Predicated on this approach, the average cost of palivizumab in the base case was estimated to be $6,215/infant, when nurse administration was included (average number of injections required: 4.09).

Model outputs

Outcomes were expressed as the cost per QALY (incremental cost-utility ratio [ICUR]) for palivizumab versus no intervention.

Sensitivity analyses

As per best practice⁵³, uncertainty was assessed by way of probabilistic (PSA) and deterministic (DSA) sensitivity analysis. In the absence of measures of dispersion in the source materials, analyses employed limits of plus or minus 10% (PSA) and 20% (DSA) on the values of the tested variables. DSA results were used for purely qualitative purposes to rank and thereby identify the most influential variables. Within the PSA (10,000 Monte Carlo simulations), gamma distribution was used as the default for costs, beta distribution for utilities and hospitalization-associated rates, and normal distribution for discount rates and mortality (Table S4 in Supplementary Materials 2). In addition, probabilistic DSA (PDSA) was undertaken based upon the methodology of Vreman et al.⁵⁴ For the PDSA, the default number of probabilistic repetitions per plotted point was 1,000.

Scenario analyses

The main alternate scenario was to use the seven-variable CRST¹⁶ (AUROC: 0.762; sensitivity: 68.2%, sensitivity: 71.9%) to guide prophylaxis (Figure S2 in Supplementary Materials 2). The CRST has cut-off scores of 0–48, 49–64 and 65–100 for low- (RSVH rate: 1.7%), moderate- (7.1%) and high-risk (18.7%) infants, respectively (Table S2 in Supplementary Materials 2). The single most important risk factor in the CRST for determining an infant’s risk of RSVH is birth during the RSV season, which accounts for 25% of the overall risk score (this compares to 11% with the IRST)^16,21. This implies that almost all infants classified at moderate- or high-risk by the CRST are born during the RSV season (and therefore weigh less than low-risk infants); this has implications for the cost of palivizumab as it is dosed by weight. For the analyses using the CRST, infant weight was adjusted to reflect the low-, moderate- and high-risk categories (Figure S3 in Supplementary Materials 2). This resulted in an average palivizumab cost of $5,370 (average 4.01 injections) per moderate/high-risk infant. To further explore the relative benefits of the two RSTs, an updated version of the CRST was created in which the RSVH rate in the low-risk group was brought into line (standardized) with that of the IRST (changing from 1.7% to 1.0%), with the rates for the moderate- and high-risk groups being adjusted accordingly (Table 1).

Other scenario analyses included (Table S5 in Supplementary Materials 2): (i) assessment of high- and moderate-risk groups separately; (ii) no discounting; (iii) 3% discounting; (iv) exclusion of societal costs; (v) inclusion of vial sharing (assuming 5% wastage); (vi) alternative value for overall palivizumab efficacy (RSVH rate 1.8% vs 10.1% for placebo; risk reduction: 82.2%) taken from the IMpact^55,56 (registrational) study; (vii) alternative value for loss of productivity due to RSV mortality ($8,805,732⁵⁷); (viii) MARI rate from the REPORT³⁵ study utilized directly without adjustment for baseline RSVH rate in non-prophylaxed infants; (ix) MARI rate in non-prophylaxed infants calculated according to the rate in prophylaxed infants with the efficacy of palivizumab removed; (x) limiting the duration of respiratory sequelae to 13 or 6 years; (xi) reducing mortality rate to 0%; (xii) discounting the acquisition cost of palivizumab by 10%, 20% or 30%.

Model validation

The model structure, data and assumptions were first validated through consultation with globally recognized clinical experts in RSV (XCE and BP) and compared with other cost-analyses of palivizumab identified during the systematic review (Supplementary Materials 1). Model coding was validated using an alternative model built independently from the main model but based on the same structure, assumptions, and data. Finally, the calculation from the non-prophylaxed arm of the model was applied to the Spanish FLIP-2⁵⁸ (Risk Factors Linked to RSV Infection Requiring Hospitalization in Premature Infants) population of 5,441 infants born at 32–35wGA, and the resulting outcomes (predicted vs actual) compared: number of RSVHs; ICU admittance; and mortality rate.

Software

The model was developed using Microsoft Excel 365.

Results

Palivizumab prophylaxis increased treatment costs and indirect costs but led to a reduction in direct healthcare costs and improved QoL compared to no prophylaxis (Table 4). In the base case, prophylaxis in moderate- and high-risk infants, as guided by the IRST, resulted in an ICUR of $29,789. When considering high- and moderate-risk infants separately, the corresponding ICURs were $22,812 and $38,447, respectively.

Table 4. Costs and utilities for palivizumab versus no prophylaxis using IRST and CRST^a.

RST	High-plus moderate-risk		High-risk		Moderate-risk
IRST
	No PVZ	PVZ	No PVZ	PVZ	No PVZ	PVZ
Treatment costs	$0.00	$6,169	$0.00	$6,169	$0.00	$6,169
Direct costs	$2,217	$536	$2,843	$662	$1,633	$419
Indirect costs	178.52	$296	$266	$320	$97	$273
Difference in costs	$4,605		$4,042		$5,130
Difference in QALYs	0.155		0.177		0.133
Cost per QALY (ICUR)	$29,789		$22,812		$38,447
Standardized CRST^b
	No PVZ	PVZ	No PVZ	PVZ	No PVZ	PVZ
Treatment costs	$0.00	$5,330	$0.00	$5,284	$0.00	$5,351
Direct costs	$2,259	$544	$3,177	$729	$1,827	$458
Indirect costs	$184	$293	$312	$328	$124	$276
Difference in costs	$3,723		$2,851		$4,133
Difference in QALYs	0.156		0.189		0.140
Cost per QALY (ICUR)	$23,841		$15,070		$29,418
CRST
	No PVZ	PVZ	No PVZ	PVZ	No PVZ	PVZ
Treatment costs	$0.00	$5,330	$0.00	$5,284	$0.00	$5,351
Direct costs	$3,110	$715	$4,675	$1,030	$2,381	$569
Indirect costs	$303	$325	$521	$385	$201	$297
Difference in costs	$2,957		$1,503		$3,635
Difference in QALYs	0.187		0.242		0.161
Cost per QALY (ICUR)	$15,833		$6,200		$22,645

Abbreviations. RST, risk scoring tool; CRST, Canadian RST; IRST, International RST; ICUR, incremental cost-utility ratio; PVZ, palivizumab; QALY, quality-adjusted life year; RSVH, respiratory syncytial virus-related hospitalization.

^a18-year respiratory morbidity, lifetime horizon, direct and indirect costs, 1.5% discounting (base case highlighted in grey).

^bRSVH rates standardized to 1% in low-risk group, as per IRST.

Sensitivity analyses

In the base case, a mean ICUR of $30,373 was predicted in the PSA (vs no-prophylaxis), with a probability of 0.79 to be below a $50,000 willingness to pay (WTP) threshold and 0.91 probability at a $75,000 WTP threshold (Figures 2 and 3). The DSA found the model to be most sensitive to utility scores, long-term morbidity rate, palivizumab cost, and the RSVH rate in non-prophylaxed infants (Figure S4 in Supplementary Materials 2).

Figure 2. Acceptability curve for palivizumab versus no prophylaxis in moderate- and high-risk infants born at 32–35wGA as guided by the IRST and CRST. Abbreviations. IRST, International Risk Scoring Tool; QALY, quality-adjusted life year; wGA, weeks’ gestational age.

Figure 3. Incremental cost-effectiveness plane for palivizumab prophylaxis in moderate- and high-risk 32–35wGA infants identified by (A) the IRST and (B) the CRST. Results are based on the probabilistic sensitivity analysis after 10,000 Monte Carlo simulations. The solid line represents an acceptability threshold of $50,000/QALY. Abbreviations. CRST, Canadian Risk Scoring Tool; IRST, International Risk Scoring Tool; PSA, probabilistic sensitivity analysis; QALY, quality-adjusted life year; wGA, weeks’ gestational age; WTP, willingness to pay.

The PDSA results indicated that the 10 most influential variables individually identified by the DSA broadly responded linearly, except for the variable “utility with significant respiratory sequelae”, which produced an exponential effect on the ICUR when increased ever closer to parity with long-term normal health (Figure S5 in Supplementary Materials 2). The results strongly suggest that the qualitative DSA observations on the relative impact of these variables are valid.

Scenario analyses

Using the CRST to identify moderate- and high-risk infants provided an ICUR of $15,833 (Table 4). The PSA resulted in incremental costs of $15,882/QALY with a 0.96 probability of palivizumab being cost-effective at the $50,000 WTP threshold and 0.99 at $75,000 (Figures 2 and 3). The point estimate for moderate- and high-risk infants using the standardized CRST was $23,841/QALY.

Vial sharing (N.B. palivizumab vials are single use) and using the IMpact trial⁵⁶ to derive palivizumab efficacy resulted in improved ICURs of $22,319 and $17,883, respectively, for moderate- and high-risk infants using the IRST (Table 5). Unsurprisingly, discounting the cost of palivizumab reduced the ICUR, with a 10% discount resulting in a value of $25,837. Conversely, reducing the period of long-term respiratory morbidity in the base case to 13 and 6 years increased the ICUR to $38,391 and $53,728, respectively. The remaining scenarios all had a relatively minor impact on the ICUR.

Table 5. Results of scenario analyses.

	Cost per QALY (ICUR)
RST	High-plus moderate-risk (original)	High-risk (original)
No discounting of costs or utilities
IRST	$26,450 ($29,789)	$19,915 ($22,812)
Standardized CRST^a	$21,009 ($23,841)	$12,821 ($15,070)
CRST	$13,524 ($15,833)	$4,625 ($6,200)
3% discount rate for costs and utilities
IRST	$32,948 ($29,789)	$25,497 ($22,812)
Standardized CRST^a	$26,503 ($23,841)	$17,115 ($15,070)
CRST	$17,937 ($15,833)	$7,548 ($6,200)
No societal/indirect costs (healthcare/payer perspective)
IRST	$29,031 ($29,789)	$22,507 ($22,812)
Standardized CRST^a	$23,150 ($23,841)	$14,990 ($15,070)
CRST	$15,713 ($15,833)	$6,761 ($6,200)
Vial sharing with 5% wastage
IRST	$22,319 ($29,789)	$16,290 ($22,812)
Standardized CRST^a	$15,945 ($23,841)	$8,541 ($15,070)
CRST	$9,231 ($15,833)	$1,105 ($6,200)
IMpact^56 efficacy rate for palivizumab
IRST	$17,883 ($29,789)	$13,748 ($22,812)
Standardized CRST^a	$13,886 ($23,841)	$8,484 ($15,070)
CRST	$8,991 ($15,833)	$2,476 ($6,200)
Treasury Board of Canada estimate of loss of productivity due to RSV mortality^57
IRST	$28,980 ($29,789)	$21,743 ($22,812)
Standardized CRST^a	$23,012 ($23,841)	$13,887 ($15,070)
CRST	$14,671 ($15,833)	$4,640 ($6,200)
MARI rate from REPORT^35
IRST	$30,120 ($29,789)	$23,044 ($22,812)
Standardized CRST^a	$24,115 ($23,841)	$15,230 ($15,070)
CRST	$16,000 ($15,833)	$6,273 ($6,200)
MARI rate for non-prophylaxed using prophylaxed rate removing palivizumab efficacy
IRST	$32,390 ($29,789)	$24,626 ($22,812)
Standardized CRST^a	$25,990 ($23,841)	$16,313 ($15,070)
CRST	$17,137 ($15,833)	$6,760 ($6,200)
6-year respiratory morbidity
IRST	$53,728 ($29,789)	$43,603 ($22,812)
Standardized CRST^a	$43,836 ($23,841)	$30,647 ($15,070)
CRST	$31,889 ($15,833)	$15,991 ($6,200)
13-year respiratory morbidity
IRST	$38,391 ($29,789)	$30,109 ($22,812)
Standardized CRST^a	$30,007 ($23,841)	$20,472 ($15,070)
CRST	$21,412 ($15,833)	$9,462 ($6,200)
0% mortality
IRST	$30,400 ($29,789)	$23,515 ($22,812)
Standardized CRST^a	$24,398 ($23,841)	$15,719 ($15,070)
CRST	$16,482 ($15,833)	$6,860 ($6,200)
Discounted price for palivizumab
IRST 10% discount	$25,837($29,789)	$19,363 ($22,812)
IRST 20% discount	$21,884 ($29,789)	$15,914 ($22,812)
IRST 30% discount	$17,932($29,789)	$12,465 ($22,812)
Standardized CRST 10% discount	$20,465 ($23,841)	$12,307 ($15,070)
Standardized CRST 20% discount	$17,088 ($23,841)	$9,544 ($15,070)
Standardized CRST 30% discount	$13,711 ($23,841)	$6,781 ($15,070)
CRST 10% discount	$13,010 ($15,833)	$4,044 ($6,200)
CRST 20% discount	$10,187 ($15,833)	$1,888 ($6,200)
CRST 30% discount	$7,364($15,833)	Dominant ($6,200)

Abbreviations. CRST: Canadian RST; ICUR, incremental cost-utility ratio; IRST, International RST; MARI, medically-attended RSV infection; QALY, quality-adjusted life year; REPORT, Respiratory Events Among Preterm Infants Outcomes and Risk Tracking study; RST, risk scoring tool; RSV, respiratory syncytial virus; RSVH, RSV-related hospitalization.

^aRSVH rates standardized to 1% in low-risk group, as per IRST.

Model validation

No marked differences between the model predictions and actual outcomes of FLIP-2⁵⁸ were found for number of RSVHs (196 vs 202, respectively), ICU admittance (19.0% vs 17.9%), and mortality rate (0.003% vs 0%).

Discussion

Moderate-to-late preterm infants (born 32–35wGA) comprise a substantial population at increased risk of severe RSV-LRTI^7,8,59. Identification of those 32–35wGA infants with additional risk factors and thus at greatest risk of severe RSV-LRTI can help guide the cost-effective use of palivizumab^17,60. In this first cost-utility analysis incorporating the IRST, palivizumab was found to be highly cost-effective when targeted to Canadian infants assessed at moderate- and high-risk of RSVH, with an ICUR of $29,789 (mean $30,373/QALY) versus no prophylaxis. While there is no strict per health outcome WTP threshold in Canada, a value of approximately $50,000/QALY is commonly cited⁶¹, with the PSA of the base case generating a 79% probability of being below this level. The ICUR also improved significantly when vial sharing (5% wastage) was considered in a scenario analysis ($22,319), which is a practice recommended in various provincial prophylaxis policies^62,63. The fact that palivizumab remained below the $50,000 threshold ($38,447) in moderate-risk infants alone (without vial sharing) further highlights the robustness of the base case result.

An updated assessment of CRST-guided prophylaxis also found this to be a highly cost-effective strategy, with an ICUR of $15,833 (mean $15,882; 96% probability of being below a $50,000 WTP threshold) in moderate- and high-risk infants. At face value, the CRST appears to be the more favourable of the two RSTs to use, with the ICUR approaching half that of the IRST ($15,833 vs $29,789, respectively). While the two RSTs have a similar level of good predictive accuracy (both AUROCs >0.75) and are well validated, a salient difference, aside from the number of included risk factors (IRST: 3 vs CRST: 7), is that the CRST ascribes a higher baseline RSVH rate (which was found to be a key driver of the ICUR in the DSA) to high- and moderate-risk infants than the IRST (10.9% vs 6.3%, respectively; Table 1)^16,21. This results in the IRST assigning approximately twice as many infants as being at moderate-risk compared to the CRST (19.9% vs 9.8%, respectively)^16,21. In practice, and taking into consideration the relative sensitivity and specificity of the tools, this means that approximately 30% more RSVHs would potentially be prevented by using the IRST than the CRST (85% vs 54%, respectively), but at a cost of a similar proportion of infants receiving prophylaxis unnecessarily (Table S6 in Supplementary Materials 2)⁶⁴. The impact of the difference in RSVH rates underlying the two RSTs was explored with the standardized CRST (where the RSVH rate in the low-risk group was set at 1.0% to match that of the IRST and the high and moderate rates adjusted accordingly), which produced an ICUR closer to that of the IRST ($23,841 in moderate- and high-risk infants). It should also be recognized that the way infant weight was adjusted to reflect risk category for the CRST (as birth in season was the preeminent risk factor) favours this tool over the IRST (where such an adjustment was not appropriate), as this ultimately led to a proportionally higher number of smaller infants being prophylaxed and, therefore, lower palivizumab cost. Subtle variation in the epidemiological characteristics of the data underpinning the two RSTs might also have influenced the results, such as the IRST encompassing infants born 32wGA while the CRST did not, as well as differences in follow-up duration (Table 6). Ultimately, the decision about which RST to use should take into consideration familiarity and ease of use (CRST used in Canada for over a decade vs fewer risk factors to score in the IRST) as well as available budget versus maximising outcomes (CRST more cost-effective vs potential for more RSVHs prevented with the IRST but with a proportion of infants receiving unnecessary prophylaxis). Maximizing the prevention of RSVHs is perhaps particularly relevant, when considering the stress on healthcare resources if there was a contemporaneous upsurge in COVID hospitalizations.

Table 6. Comparison of key attributes of the IRST²¹ and the CRST¹⁶.

Parameter	Canadian analysis	Current analysis	Notes
Data sources underlying RSTs	Canadian Included infants born 33^0–35^6wGA during the RSV season with follow-up of one month post-RSV season See Table S1 for further details	Canadian, Spanish, Dutch, USA, Italian and multinational database covering Europe, Central America, Asia & Middle East Included infants born 32^0–35^6wGA with follow-up up to 1 year of age See Table S1 for further details	The CRST was designed to identify moderate-to-late preterm infants at high-risk in Canada, while the IRST sought to do the same for infants across the Northern Hemisphere (including Canada). The IRST included infants born 32wGA, while the CRST did not. Length of follow-up in the two underlying datasets also differed
Risk factors included in RSTs	7: small (<10th percentile) for gestational age; male gender; born during RSV season (Nov-Jan); family history without eczema; subject of siblings attending day care; >5 individuals in the home, including subject; >1 smoker in the household See Figure S1 for further details	3: birth between three months before to two months after the start of the RSV season start date; household smoking and/or smoking while pregnant; and siblings (excluding multiples) and/or day care attendance See Figure S2 for further details	Considerable overlap in the included risk factors (supporting that these are universal). IRST might be considered simpler than CRST due to fewer risk factors
Predictive accuracy of RSTs^a	0.762	0.773	Both RSTs are highly predictive, with a slight advantage to the IRST
Baseline hospitalization rate for moderate- + high-risk group	10.9% (no palivizumab) See Table 1 for full details	6.3% (no palivizumab) See Table 1 for full details	Infants ascribed moderate + high by the CRST have a higher risk of RSVH vs the IRST
Potential RSVHs covered in moderate- + high-risk group	54.4% See Table S6 for full details	84.8% See Table S6 for full details	Driven in large part by the differential in RSVH rate for moderate + high-risk infants, the IRST captures ∼30% more potential RSVHs vs the CRST

Abbreviations. CRST, Canadian Risk Scoring Tool; IRST, International Risk Scoring Tool; RST, risk scoring tool; RSVH, RSV-related hospitalization; wGA, weeks’ gestational age.

^aArea under the receiver operating characteristic curve, where 1 equals perfect predictive accuracy.

When compared to the previous cost-utility analysis of palivizumab using the CRST¹⁷, the ICURs in the high-risk group were broadly similar, although approximately 15% higher in the current analysis ($5,274 vs $6,200, respectively). The higher ICUR in the current analysis probably relates to the updated assumptions behind RSV-associated mortality and the palivizumab efficacy rate. In the previous analysis,^17,60 a mortality rate of 8.1% (from Sampalis 2003⁶⁵) was applied to all infants with RSVH, and the efficacy of palivizumab at preventing RSVH was 82% (from the IMpact Study^55,56). The current analysis used more conservative and arguably more representative values of 0.43%^32,33 for mortality, and applied this to infants in the ICU only, and a palivizumab efficacy rate of 72.2%^7,27. When the efficacy rate from IMpact⁵⁶ was used in a scenario analysis, the ICUR in the current model fell to $2,476 using the CRST, demonstrating the influence of this input on the ICUR. Similarly, the influence of the higher mortality rate used in the Canadian study is most evident in the high-risk infants. This results from the higher mortality rate and higher hospitalization rate combining to produce a large relative increase in the number of infants dying, and therefore incurring a loss of productivity, in the no palivizumab arm versus the palivizumab arm.

For moderate-risk infants, however, it was the current analysis that provided the more favourable ICUR ($22,645 vs $34,438 reported previously¹⁷) This almost certainly relates to how long-term respiratory morbidity was modelled in the two analyses. In the current analysis, long-term sequelae was modelled until 18 years (using data from SPRING³⁷ and Sigurs et al.^2,39,40) with palivizumab efficacy applied^28–30, an annual cost attached to its management, and all surviving infants assumed to be at some level of risk according to palivizumab use and RSVH status (Table S3 in Supplementary Material 2). By contrast, the earlier analysis⁶⁰ included respiratory morbidity for up to 16 years but extrapolated this from the study by Greenough et al.⁴² that measured a utility difference retrospectively and at one timepoint (5 years) between infants with and without RSVH. This provided a smaller disutility for respiratory morbidity (0.07) than in the current analysis (0.09–0.16 depending on RSVH status) applied for two years shorter duration and with costs and disutility applied according to RSVH status. The morbidity rates and utility values used in the current analysis reflect the substantial evidence supporting the association of RSVH in early childhood (including specifically in 32–35wGA infants in SPRING³⁷) with long-term respiratory morbidity that can persist into adulthood^2–4,66. Indeed, 12 of the 20 health economic models identified in the systematic review considered long-term respiratory sequelae, notably all those published in the last five years^36,52,67,68. Long-term respiratory morbidity and utility scores were identified as important contributors to the ICUR in the DSA. Importantly, however, even when this sequela was restricted to 13 or 6 years, palivizumab remained cost-effective ($38,391/QALY and $53,728/QALY, respectively). Table 7 summarises the differences in the key variables as discussed above.

Table 7. Comparison of key parameters in the Canadian cost-utility analysis¹⁷ with CRST versus the current analysis with IRST (base case analyses).

Parameter	Canadian analysis	Current analysis	Notes
Hospitalization rates	Derived from risk groups in CRST See Table 1 for full details	Derived from risk groups in IRST See Table 1 for full details	Base case included infants born moderate- + high-risk
Relative risk reduction with palivizumab	82.0% From IMpact study	72.2% From Canadian registry data	Current analysis uses Canada-specific data with more conservative estimate of efficacy vs Canadian analysis
Mortality	8.1% of RSV hospitalized infants	0.43% of infants admitted to intensive care	Considerably lower rate of mortality applied in current analysis vs Canadian analysis reflecting latest evidence
Costs attributable to loss of productivity	$25,572.72 per death	$58,964.36 per death	Current analysis assigns a higher cost to loss of productivity vs Canadian analysis, but should be balanced against much lower mortality rate
Long-term respiratory morbidity duration	16 years	18 years	2-year longer duration of morbidity in current analysis vs Canadian analysis reflecting latest evidence
Long-term respiratory morbidity disutility	0.07	0.09–0.16	Higher rate of disutility applied in current analysis, which varies depending on presence or absence and severity of RSV infection
Difference in incremental costs attributable to long term morbidity in full population (palivizumab vs no prophylaxis)^a	$1,918^a	$782	Higher costs applied to long-term morbidity in Canadian analysis vs current analysis (see footnote for caveat)

Abbreviations. CRST, Canadian Risk Scoring Tool; IRST, International Risk Scoring Tool; RSV, respiratory syncytial virus; wGA, weeks’ gestational age.

^aSmart et al.¹⁷ do not provide a value for the cost of long-term respiratory morbidity, and differences in how this value is applied in the respective models would make direct comparison of raw values uninformative. The tabulated Canadian value is calculated from the incremental costs per QALY published in Smart et al.¹⁷ combined with the QALY information published in the original publication of the model by Lanctot et al.^60,73

A clinically relevant component of this new cost-utility analysis was the inclusion of MARI, which is now widely recognized as a significant contributor to the overall burden of RSV¹. None of the 20 analyses identified by the systematic review included formal assessment of MARI. A key challenge to including MARI (which may explain its absence from earlier analyses) is the limited data on rates in 32–35wGA infants and the efficacy of palivizumab in its prevention. For this reason, MARI was explored in two scenarios, both of which found palivizumab to remain highly cost-effective versus no prophylaxis in moderate- and high-risk infants identified by the IRST ($30,120/QALY and $32,390/QALY). It is also reassuring that the calculated rate of MARI used for non-prophylaxed infants of 12.2% was not too dissimilar from the rate of 9.5% in the placebo arm of the nirsevimab preterm (29–34 completed wGA) study⁶⁹.

Extensive sensitivity analyses including DSA, PDSA and PSA were conducted to deal with uncertainty associated with key model inputs. Data limitations included the availability of short- and long-term QoL outcomes for infants born 32–35wGA, with the DSA and PDSA identifying long-term utility scores as a key driver of the ICUR. The Greenough et al.⁴² data are almost 20 years old but arguably remain the best available source for utilities following RSVH in premature infants and are used as the basis of many published assessments^17,36,52. More recent estimates of the impact of RSV on QoL are available; however, they are less applicable to our population of interest than the older data. For example, two recent studies by Hodgson et al.⁷⁰ and Diez-Gandia et al.⁷¹ were not restricted to infants (rather children less than five and two years, respectively), included term as well as preterm children, and, importantly, did not provide a disutility for RSV hospitalized infants, rather the QoL of all RSV-positive children regardless of disease severity. The estimate of the disutility during hospitalization is based on data from 2005; a recent US systematic review of studies that derived utility lost per RSV episode identified only 7 such studies⁷², none of which provided more relevant data than that used herein. It is also important to note that this disutility is applied only for the duration of the hospital stay, which is for an average of six days. Therefore, it does not impact the outputs of the model to any significant degree but does reflect the clinical situation. Utility data related to long-term morbidity following RSV-LRTI are also lacking, so a surrogate value was used from a follow-up of children with asthma⁴³, which was the same approach taken in the cost-utility analysis by Sanchez-Luna et al.³⁶. Another limitation is that while gestational age-specific data are available for long-term respiratory morbidity (SPRING³⁷) and the efficacy of palivizumab^28–30 in its prevention for up to six years, beyond this time point more general sources (Sigurs et al.^2,39,40) and data extrapolation had to be employed. However, given the increasingly recognized temporal association between RSV LRTI and long-term morbidity^2–4,37,66 and the documented beneficial effects of palivizumab in this regard up to age three years^29–31, we believe that our approach is clinically justified. It would also have been preferable if there were available one Canadian source of data for palivizumab efficacy at preventing RSVH and hospital resource use, although the three key sources used (PICNIC⁷, CARESS²⁷, and Quebec^10,28) were all large, well conducted studies in the relevant population. That the base case was well below $50,000/QALY (point estimate: $29,789; mean from PSA: $30,373) and that the model was validated against outcomes from FLIP-2⁵⁸, afford confidence in the reliability of the results. Moreover, the base case considered the list price of palivizumab, and local discounts will inevitably reduce costs and improve the ICUR associated with the use of palivizumab, particularly in combination with local policies recommending vial sharing (Table 5) ^62,63. Further benefits of prophylaxis not captured in this cost-utility analysis include avoiding the mental health impact on families/carers of having a child with severe RSV-LRTI and the additional costs associated with managing patients from indigenous and remote communities^{12–14,72,73}.

The current model can readily be adapted for cost analyses of palivizumab in other countries, particularly those with a similar economic status and RSV seasonality; it would principally involve updating to local costs and including country-specific data on RSVH rates, as available. There are several new preventive interventions for RSV on the near horizon, including the extended half-life monoclonal antibodies nirsevimab (which is already licenced in Europe) and clesrovimab and maternal and childhood vaccines⁷⁴. These newer interventions have the potential to provide protection to all infants, although differences in their immunization coverage as well as financial, logistical and manufacturing constraints, and, importantly, parental choice, will likely mean that a variety of options will be needed to fully address the burden of RSV^75–77. The overall structure and decision tree of the current model could be adapted for economic evaluations of these newer therapies.

Conclusions

This current analysis demonstrates that selectively targeting palivizumab to those Canadian 32–35wGA infants at moderate- and high-risk of severe RSV-LRTI using the IRST (or CRST) is a highly cost-effective strategy versus no prophylaxis. Adoption of a RST within prophylaxis policies for 32–35wGA infants will help to ensure the more equitable use of palivizumab across Canada.

Transparency

Declaration of financial/other relationships

XCE and BP have received research funding and/or compensation as advisor/lecturer from AstraZeneca and Sanofi. BRG, IK and JF employer has received payment from AstraZeneca for work on various projects. JET has nothing to disclose. The funding body had no role in the design and development of the cost-utility model, the interpretation of the data or writing the manuscript. Peer reviewers on this manuscript have received an honorarium from JME for their review work but have no other relevant financial relationships to disclose.

Author contributions

All authors contributed to the study conception and design, the development of the publication, maintained control over the final content, and approved publication.

Previous presentations

Information on the model and its results have been previously presented as abstracts and posters at Scientific and Medical meetings.

Acknowledgements

The authors would like to thank Krista Lanctôt (Medical Outcomes and Research in Economics Research Group, Sunnybrook Health Sciences Centre, University of Toronto, Toronto, Ontario, Canada) for organising additional analyses of the CARESS database.

Data availability statement

All data generated or analysed during this study are included in this published article [and its supplementary information files].