A latent class analysis of factors influencing preferences for infant respiratory syncytial virus (RSV) preventives among pregnant people in the United States

ABSTRACT

A maternal vaccine and long-acting monoclonal antibody (mAb) were recently approved to protect infants against respiratory syncytial virus (RSV). We identified subgroups of pregnant people with different preferences for RSV preventives and respondent characteristics associated with subgroup membership. An online survey, including a discrete choice experiment (DCE), was conducted among US pregnant people. RSV preventive attributes included effectiveness, duration of protection during RSV season, injection recipient/timing, preventive type (vaccine or mAb), and type of visit required to receive injection. In DCE choice tasks, pregnant people selected between two hypothetical preventive profiles with varying attribute-levels and a no-preventive option. Logistic regression, including latent class analysis (LCA), was used to analyze the data. Of 992 pregnant people (mean age: 30.0 years), 60.3% were expecting their second/later birth. LCA identified three preference subgroups: ‘Effectiveness’ (preventive choice mostly driven by increases in effectiveness; 51.4% class membership probability), ‘Season’ (preventive choice mostly driven by improvement in duration of protection during the RSV season; 39.2% class membership probability), and ‘No Preventive’ (frequently chose no-preventive option; 9.4% class membership probability). ‘Effectiveness’ and ‘Season’ preferred maternal vaccine over mAb; mAb was preferred by ‘No Preventive.’ Perceiving RSV as serious for infants, higher health literacy, and lower household income were associated with ‘Effectiveness.’ Perceiving RSV as serious for pregnant people was associated with ‘Season.’ Perceiving RSV to not be serious for pregnant people and not being employed were associated with ‘No Preventive.’ Subgroups of pregnant people vary in preferences for RSV preventives. Most pregnant people preferred a maternal vaccine, although some may be more willing to accept alternative preventive options.

KEYWORDS:

• Maternal vaccination
• monoclonal antibodies
• pregnant people
• respiratory syncytial virus
• preferences
• preference heterogeneity

Introduction

Respiratory syncytial virus (RSV) causes lower respiratory tract infections (LRTI). For most individuals, RSV infection causes only mild symptoms, similar to the common cold. Nevertheless, RSV can result in severe illness among infants, particularly those who are born pre-term or who have other risk factors (e.g., low birth weight, <6 months of age), as well as in older adults, people with underlying cardiovascular or respiratory disease, and immunocompromised individuals.¹ Annually, RSV leads to approximately 58,000 hospitalizations and 100–500 deaths among children <5 years old in the United States (US).² Recently, a surge of RSV infections in the US has challenged pediatric hospital capacity and resources,³ underscoring the need for preventive options to reduce the societal burden of RSV.

RSV prevention strategies for infants include maternal vaccination or administration of monoclonal antibodies (mAbs) to infants early in life.^4,5 Maternal vaccination, which confers immunity to infants via the transmission of maternal antibodies through the placenta, is indicated for use in pregnant people during the third trimester of pregnancy;⁶ with maternal vaccination, an infant is protected against RSV from the time of birth.⁵

In contrast, mAbs confer immunity to the infant from the time of injection. Among currently available mAbs, the use of palivizumab is limited due to cost and the need for monthly injections during the RSV season. Long-acting mAbs (e.g., nirsevimab), which have a longer half-life than palivizumab, do not require repeated administration.⁴ According to Advisory Committee on Immunization Practices (ACIP), nirsevimab is recommended for all infants aged <8 months who are born either during or just prior to their first RSV season, as well as for children aged 8–19 months who are entering their second RSV season if they are at higher risk of severe illness due to RSV.⁷

Despite the urgent need to protect infants from severe RSV illness, immunization coverage for vaccine-preventable infections is suboptimal, with < 70% of pregnant people reporting that they intend to receive recommended vaccines (influenza, Tdap, or COVID-19) during their pregnancy or to have their infant immunized.^8,9 Moreover, perceptions about vaccine safety and efficacy, knowledge and information about vaccines, and personal characteristics, such as race, ethnicity, household income, and number of children, impact pregnant people’s intentions to immunize themselves or their infant.^8–10 Hence, a key challenge for public health policymakers will be to overcome the many potential barriers to infant RSV immunization.

Research on preferences for RSV preventives is limited, given the very recent approval of a maternal vaccine and long-acting mAbs. A recent study found that nearly 9-in-10 pregnant people were willing to choose a preventive (maternal vaccination or infant mAb) over no preventive.¹¹ Effectiveness and duration of protection during the RSV season were the primary drivers of RSV preventive choice among pregnant people, irrespective of whether they were expecting their first or second/later birth.

Nevertheless, given the suboptimal immunization rates among pregnant people, further research is warranted to identify the characteristics of pregnant people with greater or lesser willingness to use currently available RSV preventives. Identifying preferences and priorities for infant RSV immunization among subgroups of pregnant people and understanding the characteristics of each subgroup will facilitate patient communication and the development of educational information targeted to what is important to different subgroups of pregnant people to support decisions on infant RSV prevention.

Therefore, the objective of this study was to identify subgroups of pregnant people with similar preferences and priorities for infant immunization and to identify the characteristics of each subgroup and how they vary across subgroups. To address this objective, we conducted a secondary analysis of data from Beusterien and colleagues.¹¹

Materials and methods

Study design and sample

A cross-sectional online survey, including a discrete choice experiment (DCE), was conducted between October 11 and November 11, 2022. We planned to recruit up to 1,000 pregnant people to participate in the study. Pregnant people, who were recruited via the LifePoints general panel, were included if they were aged ≥18 years, resided in the US, were currently pregnant at the time of the survey, were able to answer questions in English, and provided informed consent. To ensure diversity in key respondent characteristics, recruiting quotas were set to target approximately 400 pregnant people who were expecting their first birth and 400 expecting their second/later birth. Within each of these 2 quotas, we sought to recruit at least 100 pregnant people who were White, Black or African American, and Hispanic or Latina.

A formula used to determine sample size for aggregate level full-profile DCE modeling is nta/c > 1,000, where n is the number of respondents, t is the number of choice tasks, a is the number of alternatives (RSV preventive profiles) presented in each choice task, and c is equal to the largest number of levels (e.g., RSV preventive effectiveness could have four levels, 50%, 63%, 77%, and 90%) for any given attribute.¹² With a target sample size of 1,000 pregnant people, a maximum of four levels per attribute, two alternatives in each choice task, and 12 choice tasks, the formula result is 6,000, which is > 1,000, indicating that our target sample size was sufficient to estimate relatively precise preference weights from the DCE data.

Increasingly, DCEs have been used to quantify the preferences of patients, healthcare providers, and other decision-makers for health-related products, services, and outcomes, with the aim to facilitate stakeholder involvement in healthcare decision-making.¹³ Regarding childhood immunization specifically, DCE studies often focus on parents’ preferences for having their children vaccinated and most frequently assess vaccine uptake and willingness to pay for vaccines.¹⁴ The DCE approach stipulates that preventive options can be decomposed into features of interest, referred to as attributes.¹⁵ The attractiveness of a preventive product to an individual depends on their relative preferences for the attributes associated with that particular product, which is expressed by the frequency with which they choose RSV preventive profiles with the preferred features in a series of choice tasks. This method allows us to quantify how important each attribute is in influencing pregnant people’s choice of RSV preventive and the trade-offs that they are willing to make among the attributes of RSV preventive options.

RSV preventive attributes of potential importance to pregnant people that are likely to differ among preventive options were identified via literature search, confirmed during qualitative interviews, and included in the DCE. Table 1 presents the attributes and levels included in the DCE choice tasks. Prior to completing the DCE, pregnant people were presented with information about RSV and RSV preventive attributes.¹⁶ Pregnant people completed a series of 12 DCE choice tasks, each including two hypothetical RSV preventive profiles with varying levels of each attribute and a no-preventive opt-out option. An example DCE choice task is shown in Figure A1.

Table 1. Discrete choice experiment attributes and levels.

Attribute	Description	Levels	Source of Estimate
Effectiveness	[INSERT LEVEL]% effective. ([INSERT LEVEL] out of 100 babies will be protected against serious illness requiring medical attention because of RSV)	50%, 50 63%, 63 77%, 77 90%, 90	Etti et al. [NCT02624947]^Citation17 Griffin et al. [NCT02878330]^Citation18 Muňoz et al. [NCT02247726]^Citation19
Type of prevention		Vaccine Monoclonal antibody	World Health Organization^Citation20
Recipient of injection and timing of protection	Split into two rows in tasks Single injection into [INSERT LEVEL], which may cause injection-related pain in the [INSERT LEVEL] Protection begins [INSERT LEVEL]	Single injection into a pregnant woman’s arm before their baby is born, which may cause injection-related pain in the woman. Protection begins at birth Single injection into a baby’s leg within 14 days after birth, which may cause injection-related pain in the baby. Protection begins at the time of injection Single injection into a baby’s leg 2–4 months after birth, which may cause injection-related pain in the baby. Protection begins at the time of injection	Etti et al. [NCT02624947]^Citation17 Hammitt et al. [NCT03979313]^Citation4
Type of visit required	Administered during [INSERT LEVEL]	a routine checkup appointment an extra appointment (in addition to routine checkup appointments)	World Health Organization^Citation20
Protection corresponds to season (RSV is a risk for 6 months)	Protection covers [INSERT LEVEL]	1 month in the RSV season and 5 months outside the RSV season 3 months in the RSV season and 3 months outside the RSV season 4.5 months in the RSV season and 1.5 months outside RSV season 6 months in the RSV season and 0 months outside the RSV season	Schwarz et al. [NCT03674177]^Citation21 Griffin et al. [NCT02878330],^Citation17 Hammitt et al. [NCT03979313]^Citation4 Baral et al.^Citation22

RSV: respiratory syncytial virus.

The survey also included questions regarding respondent characteristics including: sociodemographic characteristics (e.g., age, income, education, race/ethnicity); perceptions about the seriousness of RSV infection for pregnant people and for infants (rated from 1 = not at all serious to 5 = extremely serious, or ‘don’t know’); social support (“How easy is it for you to get help from family, friends, or neighbors if you need it?,” rated from 1 = very difficult to 5 = very easy); health literacy (“How confident are you filling out medical forms by yourself?,” rated from: 1 = extremely to 5 = not at all);²³ and number of prior births (first birth vs. second/later birth).

Study participants provided informed consent electronically. The final study protocol and informed consent documents were submitted to Sterling IRB (Atlanta, Georgia, USA), and an exemption determination was granted on 30 June 2022 (IRB ID #: 10193-MMaculaitis).

Statistical analysis

The analysis involved three steps: 1) a latent class analysis (LCA) to identify different subgroups or classes within the population who may have different distributions of preferences; 2) estimation of the probability of choosing between one of two preventive profiles versus the no-preventive option by latent class; and 3) examination of the relationship between individual respondent characteristics and the probability of membership in each of the LCA classes by regressing class membership probability on respondent characteristics.²⁴

The LCA involved using the DCE data to assess potential RSV preventive product preference heterogeneity among pregnant people.^25–27 The procedure for LCA involves running the Expectation-Maximization (EM) algorithm, which uses conditional logistic regression.²⁸ The EM algorithm first uses conditional logistical regression to fit a latent class solution conditioned on a specified number of classes and yields preference weights for each attribute-level in the DCE for each class. Second, the EM algorithm assigns a probability to each individual of belonging to each latent class. In this study, the LCA evaluated solutions for 1 to 5 latent classes. Identification of the optimal number of latent classes was based on the Bayesian Information Criteria (BIC), where lower values, as well as larger decreases in value with each subsequent class, indicate a more optimal solution and the size and differences of the preference weights across classes in each solution.²⁹

Based on the attribute-level preference weights generated for each latent class from the LCA, attribute relative importance (RI) estimates were computed for each latent class by dividing the range of preference weights for each attribute (preference weight of the most preferred level minus preference weight of the least preferred level) by the sum of the ranges of all attributes and standardizing to a 0–100% scale, where higher values indicate greater importance to RSV preventive choice.

For each latent class, we estimated the probability of choosing two specific treatment profiles versus the no-preventive option using a preference share logit model. The profiles reflected a preventive profile that comprised all of the most favorable attribute-levels, based on the mean preference weights for the full sample (Most Favorable Preventive Profile), and a profile comprising all of the least favorable levels based on the mean preference weights for the full sample (Least Favorable Preventive Profile) (Figure A2).¹¹ Within each class, attribute-level preference weights from the LCA were used to compute the probability of selecting the most favorable or least favorable preventative profiles versus the ‘no-preventative’ option.

Respondent characteristics, both for the total sample of pregnant people and weighted by the probability of membership in each class, were reported via descriptive statistics. A one-way analysis of variance (ANOVA) test was used to compare across latent classes on age, with chi-square tests used to compare across latent classes on all other respondent characteristics.³⁰ P-values <0.05, 2-tailed, were considered statistically significant.

A separate multinomial logistic regression model was used to predict the probability of latent class membership among potential predictors.³¹ The potential set of independent predictor variables were prespecified and included sociodemographic characteristics, perceptions about the seriousness of RSV infection for pregnant people and for infants, social support, health literacy, and number of prior births. To select the independent variables, we first examined intercorrelations among the set of respondent characteristics (Spearman’s rho and phi coefficients). We then input the set of respondent characteristics into the multinomial regression model. All variables with p < .25 were retained in the final model, given that stricter thresholds for covariate selection may not identify all meaningful predictors.^32,33 For this analysis, adjusted odds ratios (ORs) and 95% confidence intervals (CIs) are reported. This analysis allowed us to examine the independent contribution of each patient characteristic to latent class membership, while controlling for the effects of all other variables in the model.

Results

Overall, 992 pregnant people completed the study survey and were included in the analyses. Table 2 reports respondent characteristics for the total sample and probability-weighted respondent characteristics by latent class. Nearly 40% of pregnant people in the total sample were expecting their first birth, with the majority (60.3%) expecting their second/later birth. A majority identified as White (69.9%), and approximately 1-in-5 pregnant people in the total sample identified as Black/African American.

Table 2. Respondent characteristics by latent class.

Variables	Total (N = 992)	Class 1: Effectiveness	Class 2: Season	Class 3: No Preventive
	Full Sample	51.4% Class Membership Probability	39.2% Class Membership Probability	9.4% Class Membership Probability	P-value***
Age (mean)^****	30.0	29.6	30.3	30.4	0.242*****
Number of births (%)					0.533
First birth	39.7	41.9	38.5	32.7
Second/later birth	60.3	58.1	61.5	67.3
Race (%)
African American	20.6	16.9	23.7	27.6	0.025
Asian	2.3	1.6	2.7	4.9	0.128
American Indian	2.5	3.2	1.9	1.1	0.310
Pacific Islander	0.7	0.6	0.7	1.1	0.844
White	69.9	74.2	66.3	61.0	0.211
Other race		8.9	6.6	9.6	0.426
Ethnicity (%)
Hispanic	27.2	27.0	27.2	28.2	0.982
Seriousness of RSV for infants (%)
Not at all	0.9	0.4	1.3	2.1	0.161
Slightly	4.4	2.5	7.3	3.4	0.003
Moderately	12.9	11.0	15.1	14.2	0.222
Very	37.3	39.9	34.2	35.8	0.378
Extremely	37.9	41.4	35.1	30.8	0.161
Don’t know	6.6	4.9	7.0	13.6	0.009
Seriousness of RSV for PP (%)
Not at all	1.7	1.3	1.4	5.3	0.019
Slightly	8.1	7.8	7.9	10.2	0.747
Moderately	24.4	25.1	23.2	25.3	0.825
Very	33.5	35.4	33.3	23.5	0.185
Extremely	21.6	19.8	25.0	17.1	0.151
Don’t know	10.8	10.6	9.2	18.5	0.048
US region of residence (%)
Northeast	14.3	14.3	13.6	17.5	0.663
Midwest	20.3	22.2	18.7	16.4	0.357
South	45.2	43.4	47.3	46.0	0.694
West	20.3	20.1	20.5	20.1	0.994
Education (%)
Some high school or less	4.8	3.5	6.9	3.2	0.055
High school graduate or equivalent	25.2	23.2	25.9	33.0	0.210
Completed some college	18.1	19.3	15.7	22.2	0.281
Associate degree	9.6	10.8	7.4	11.9	0.207
College graduate	23.1	24.3	23.2	16.1	0.323
Completed some graduate school	3.2	3.5	3.1	2.3	0.826
Completed graduate school	14.7	15.0	15.8	9.0	0.299
Prefer not to answer	1.2	0.4	2.0	2.3	0.063
Employment (%)
Employed full-time	48.4	47.2	53.3	34.5	0.055
Employed part-time	12.6	13.0	12.7	10.3	0.795
Employed, but on temporary leave	4.1	4.7	3.6	3.2	0.618
Not employed	27.1	25.7	25.3	42.4	0.012
Student	4.7	5.5	3.3	6.4	0.244
Other	3.0	3.9	1.8	3.3	0.208
Household income (%)
<$15,000	14.0	13.9	13.9	15.0	0.963
$15,001 to $30,000	13.7	13.3	13.3	17.6	0.563
$30,001 to $60,000	22.7	25.2	18.9	24.6	0.129
$60,001 to $90,000	20.1	19.2	20.3	24.1	0.623
$90,001 to $150,000	17.1	15.7	21.1	8.4	0.015
>$150,000	7.8	8.8	7.3	3.8	0.260
Prefer not to answer	4.6	3.9	5.2	6.5	0.436
Location of residence (%)
Major metropolitan area	16.6	16.3	16.9	17.2	0.971
Urban area	24.8	21.0	31.0	19.8	0.007
Suburb of a large city	27.4	28.7	24.4	32.7	0.280
Small city	15.1	16.5	12.3	19.4	0.155
Rural or small town	16.0	17.4	15.4	10.9	0.324
Insurance status (%)
Individual/family plan	50.0	54.8	46.0	38.3	0.046
Medicaid	41.2	38.3	44.8	42.6	0.316
Health literacy (%)^*
Extremely	71.4	75.4	66.6	69.0	0.291
Quite a bit	20.0	18.9	21.4	19.8	0.705
Somewhat	6.7	4.3	9.2	9.2	0.011
A little bit	0.9	0.5	1.3	1.9	0.252
Not at all	1.1	1.0	1.5	0.0	0.412
Social support (%)^**
Very difficult	9.0	7.3	11.7	6.2	0.063
Difficult	10.1	8.1	11.7	13.4	0.128
Possible	30.9	29.8	31.8	31.6	0.852
Easy	29.0	31.3	25.9	27.6	0.311
Very easy	21.0	23.5	18.9	14.8	0.137

Respondent characteristics were weighted using the probability of membership in each latent class. PP: pregnant people; RSV: respiratory syncytial virus; US: United States.

*Based on the following item: “How confident are you filling out medical forms by yourself?”

**Based on the following item: “How easy is it for you to get help from family, friends, or neighbors if you need it?”

***P-value for whether the percentages are the same between classes computed with a chi-square test. No correction for multiple testing or check for asymptotic normality was performed.

****Mean age in each class was estimated by assigning each respondent to the class for which they had the highest membership probability.

*****P-value computed with one-way ANOVA.

The LCA identified three classes with distinct preferences: Class 1 (‘Effectiveness Group’), Class 2 (‘Season Group’), and Class 3 (‘No Preventive Group’). The racial distribution differed across latent classes (p = .025), with the No Preventive Group having the highest proportion of Black/African American pregnant people (27.6%), followed by the Season Group (23.7%), and the Effectiveness Group (16.9%).

Additionally, the distributions for not employed (42.4%, 25.7%, and 25.3% for No Preventive Group, Effectiveness, and Season Group, respectively), household income of $90,001–$150,000 (21.1%, 15.7%, and 8.4% for Season Group, Effectiveness Group, and No Preventive Group, respectively), residing in an urban area (31.0%, 21.0%, and 19.8% for Season Group, Effectiveness Group, and No Preventive Group, respectively), and individual/family plan insurance (54.8%, 46.0%, and 38.3% for Effectiveness Group, Season Group, and No Preventive Group, respectively) differed across latent classes (all, p < .05). For items assessing health literacy and the perceived seriousness of RSV for infants and pregnant people, the distribution of some, but not all, response options significantly differed across latent classes (all, p < .05). No other differences in respondent characteristics were observed.

Preferences by latent class

Figure 1 presents the estimated preference weights for each DCE attribute level by latent class, and Figure 2 presents the attribute RI estimates by latent class. Overall, the biggest differences between latent classes were observed on effectiveness and duration of protection during RSV season. Although smaller in absolute terms, there were also differences between latent classes on injection recipient/timing and preventive type, with the levels of these two attributes varying in rank order by latent class. Specifically, vaccine was preferred to mAb in the Effectiveness Group and the Season Group, while mAb was preferred to vaccine in the No Preventive Group. In addition, earlier infant injection (≤14 days after birth) was preferred to later infant injection (2–4 months after birth) in the Effectiveness Group, while later infant injection was preferred in the No Preventive Group. There were no differences between latent classes in type of medical visit required to receive injection.

Figure 1. Attribute-level preference weights by latent class.

Attributes are shown in the order in which they were presented in the DCE choice tasks. Larger vertical distance between the preference weights across attribute-levels reflect a greater strength of preference for the respective change. For example, increasing effectiveness from 50% to 63% yielded a change in utility of approximately 1.5 (i.e., −0.5−[−2.0]) for the Effectiveness Group and approximately 0.8 (i.e., −0.2−[−1.0]) for the No Preventive Group. Increasing duration of protection during RSV season from 4.5 months to 6 months yielded a change in utility of approximately 0.2 (i.e., 0.6−0.4) for the Effectiveness Group and approximately 0.1 (i.e., 0.3−0.2) for the No Preventive Group. Therefore, an incremental increase in effectiveness from 50% to 63% is 7.5 times as important to those in the Effectiveness Group (i.e., 1.5 ÷ 0.2) and 8.0 times as important to those in the No Preventive Group (i.e., 0.8 ÷ 0.1) as an increase in duration of protection from 4.5 months to 6 months. The large negative estimate of the no-preventive option (−4.7) among the Effectiveness Group indicates that they are more likely to choose an RSV preventive option than the opt-out option across choice tasks; in contrast, the positive estimate for the no-preventive option (2.2) among the No Preventive Group indicates that they are more likely to choose the opt-out option than an RSV preventive option across choice tasks. PP: pregnant people; RSV: respiratory syncytial virus.

Figure 2. Attribute relative importance by latent class.

Relative importance is the importance of moving from the least preferred level of an attribute to the most preferred level of that attribute, relative to all the other attributes and conditional on the ranges of levels of all attributes. Relative importance estimates sum to 100% across attributes for each of the three latent classes. Attributes are presented in descending order of importance for the Effectiveness Group.

*Relative importance estimate is based on changing from ‘1 injection into infant’s leg ≤14 days after birth; protection at injection’ to ‘1 injection into pregnant person’s arm before baby’s birth; protection at birth’ for the Season Group and the No Preventive Group and changing from ‘1 injection into infant’s leg 2-4 months after birth; protection at injection’ to ‘1 injection into pregnant person’s arm before baby’s birth; protection at birth’ for the Effectiveness Group.

**Relative importance estimate is based on changing from ‘mAb’ to ‘vaccine’ for the Effectiveness Group and the Season Group and changing from ‘vaccine’ to ‘mAb’ for the No Preventive Group.

Class 1 (‘Effectiveness Group’), constituting 51.4% of overall class membership probability, prioritized greater effectiveness versus changes in all other DCE attributes (largest difference in preference weights [∆] was for a change from ‘50% effective’ to ‘63% effective:’ 1.95-[−0.51] = 1.44). An increase from ‘50% effectiveness’ to ‘90% effectiveness’ (RI = 61.8%) was most important to those in the Effectiveness Group; effectiveness was at least 2.4 times as important as each of the other attributes.

Class 2 (‘Season Group’), constituting 39.2% of overall class membership probability, prioritizes duration of protection during the RSV season over changes in the other DCE attributes (∆ for change from ‘1 month’ to ‘3 months:’ 0.07–[−0.39] = 0.46). An increase from ‘1 month’ to ‘6 months’ of protection during the RSV season (RI = 47.9%) was most influential to preventive choice among pregnant people in the Season Group and was at least 1.5 times as important as each of the other attributes. Additionally, changes in injection recipient/timing were more important to pregnant people in the Season Group, relative to those in the Effectiveness Group and No Preventive Group (RI = 16.0% vs. 8.9% and 5.3%, respectively). For all attributes, the preference weights for the Season Group were relatively flat; although the Season Group was most sensitive to improvements in duration of protection during the RSV season, they were still the least sensitive to this attribute across the three latent classes.

Class 3 (‘No Preventive Group’), constituting 9.4% of overall class membership probability, was most likely to choose the no-preventive option over a preventive option, regardless of the combination of attributes. The positive mean value of the preference weight for the no-preventive option (2.22) indicated that pregnant people in Class 3, on average, preferred the no-preventive option versus an RSV preventive. Nevertheless, improving effectiveness from ‘50% effective’ to ‘90% effective’ (RI = 71.6%) was of greatest importance to pregnant people in this group and was at least 3.3 times as important as each of the other attributes.

Opt-out versus preventive profile choice by latent class

The probability of choosing the no-preventive option over an RSV preventive profile for each latent class is shown in Figure A3. For the Effectiveness Group, 0.0% and 19.4% were estimated to choose the no-preventive option over the most and least favorable RSV preventive profiles, respectively. For the Season group, 9.0% and 26.6% were estimated to choose the no-preventive option over the most and least favorable RSV preventive profiles, respectively. It was estimated that 61.2% and 97.7% of those in the No Preventive Group would choose the no-preventive option over the most and least favorable preventive profiles, respectively.

Collectively, these results indicate that when an RSV preventive has highly preferred features, nearly all pregnant people in the Effectiveness Group and the Season Group would likely choose an RSV preventive. Furthermore, even with lower effectiveness, shorter duration of protection during RSV season, and other less preferred features, most pregnant people in the Effectiveness Group and the Season Group would still choose an RSV preventive. Most pregnant people in the No Preventive Group are anticipated to opt-out of immunization, regardless of the features of a given RSV preventive option. Yet, when an RSV preventive has highly preferred features, nearly 4-in-10 pregnant people in the No Preventive Group would potentially be willing to choose an RSV preventive.

Predictors of latent class membership

The respondent characteristics variables shown in Table 2 were input into the LCA model to evaluate their associations with latent class membership in terms of adjusted ORs; the Season Group was treated as the reference group for the outcome variable; therefore, the likelihood of being in either the Effectiveness Group or No Preventive Group is expressed as compared to the Season Group. Pregnant people who perceived RSV as ‘very’ or ‘extremely’ serious for infants were significantly more likely to be members of the Effectiveness Group (ORs = 1.39, 95% CI: 1.09–1.77 and 1.54, 95% CI: 1.19–1.99, respectively), whereas those who perceived RSV to be ‘extremely’ serious for pregnant people were significantly less likely (OR = 0.68, 95% CI: 0.51–0.90) to be members of the Effectiveness Group (Figure 3). Pregnant people who had high health literacy (‘extremely,’ OR = 1.22, 95% CI: 1.04–1.43) and household income <$60k (OR = 1.23, 95% CI: 1.04–1.47) were significantly more likely to be members of the Effectiveness Group. Those who perceived RSV as ‘not at all/slightly’ serious for pregnant people and pregnant people who were not employed were significantly more likely to be members of the No Preventive Group (ORs = 2.03, 95% CI: 1.30–3.19 and 1.42, 95% CI: 1.09–1.86, respectively).

Figure 3. Association of respondent characteristics and latent class membership for (a) effectiveness group versus the season group and (b) no preventive group versus the season group.

Odds ratios, adjusted for all other variables in the model, with 95% confidence intervals are shown. Variables are listed in the order in which they are shown in the figures. The Season Group was used as the reference group for the outcome variable in the model. Seriousness of RSV for infants used responses to the following item: “How serious do you think RSV infection can be for babies (12 months of age or younger)?” Seriousness of RSV for pregnant people used responses to the following item: “How serious do you think RSV infection can be for pregnant women?” Health literacy used responses to the following item: “How confident are you filling out medical forms by yourself?” Social support used responses to the following item: “How easy is it for you to get help from family, friends, or neighbors if you need it?”

*p < .05, 2-tailed.

†Includes response options of ‘not at all serious,’ ‘slightly serious,’ and ‘don’t know.

Discussion

The current study offers important insights on the factors underlying the willingness of pregnant people to immunize their infants against RSV. This is the first study to identify unique segments of pregnant people who differ in RSV preventive preferences. This analysis revealed three segments of pregnant people with different degrees of willingness to use RSV preventives. For one group (~51% of overall class membership probability), RSV preventive choice was mostly influenced by increases in effectiveness (‘Effectiveness Group’).

Conversely, there was a group of pregnant people (~39% of overall class membership probability) for whom RSV preventive choice was driven by improvement in duration of protection during the RSV season (‘Season Group’). Lastly, a smaller group of pregnant people (<10% of overall class membership probability), who frequently chose to opt-out of RSV immunization, was identified (‘No Preventive Group’). Collectively, these findings suggest that the choice to immunize infants against RSV is a function of both the attributes of the preventive options and the characteristics of the pregnant person. Notably, while the attributes of available RSV preventive options are fixed, the characteristics of pregnant people can potentially be used to inform targeted educational efforts to increase immunization coverage.

A recent study conducted by Gidengil and colleagues found that positive attitudes about vaccines, receiving/intending to receive certain immunizations (Tdap and influenza), and awareness that an infant will likely contract RSV in the first year of life were related to greater willingness among pregnant people to use an RSV preventive; living in the US South (vs. Northeast) was associated with less willingness to use an RSV preventive.³³ In contrast to Gidengil et al.,³⁴ our study relates pregnant person characteristics to both willingness to immunize and to the features of preventive options that drive willingness to immunize and preventive choice. Our results indicated that latent classes differed most on the importance of effectiveness and duration of protection during the RSV season; smaller differences were observed on injection recipient/timing and preventive type, with attribute-levels switching in rank order. Type of medical visit required to receive injection was least influential to preventive choice among all three latent classes.

Furthermore, we found that perceiving RSV as serious for infants was associated with membership in the Effectiveness Group, whereas perceiving RSV to be serious for pregnant people was associated with being a member of the Season Group. Having a high level of health literacy and lower household income were also correlated with membership in the Effectiveness Group. Perceiving RSV to not be serious for pregnant people and not being employed were associated with membership in the No Preventive Group. By determining these relationships, the present study not only identified additional characteristics than those reported by Gidengil et al.,³⁴ but also showed how those characteristics relate to the RSV preventive priorities of pregnant people.

Notably, a preventive with high effectiveness was strongly preferred by pregnant people in the No Preventive Group, explaining nearly three-quarters of the variation in their RSV preventive choice. These results suggest that, among the pregnant people who are less likely to choose a preventive, but who are willing to consider a preventive (as opposed to those who would never choose a preventive), efficacy is the primary driver of their decision. Likewise, efficacy was of greatest importance to preventive choice among the Effectiveness Group and was second most important among the Season Group, explaining 62% and 33% of the variation in preventive choice, respectively, for pregnant people in these two groups. These findings are consistent with previous DCE studies showing that vaccine effectiveness is highly influential to parents’ choice to vaccinate their children.^35,36

The pattern of attribute-level preference weights indicated a preference for maternal vaccine over mAb among pregnant people in the Effectiveness Group and the Season Group, which together comprise over 90% of overall class membership probability. In contrast, mAb was preferred over maternal vaccine by those in the No Preventive Group, which accounts for about 10% of overall class membership probability. As such, it is possible that those who are reluctant to vaccinate may not have the same aversion to a preventive that is not a vaccine. Nonetheless, among all three groups of pregnant people, preventive type had relatively little influence on preventive choice compared to the other attributes included in the study.

It has been projected that the adoption of RSV preventives will reduce RSV hospitalizations in infants by up to 50%.³⁷ Still, sufficient uptake will be paramount to realizing these potential public health benefits. The current study’s findings suggest that healthcare providers and public health officials should emphasize RSV preventive effectiveness when advising pregnant people about immunization, especially among those who might otherwise be hesitant to seek immunization, including individuals who are not employed and those who do not recognize the seriousness of RSV for pregnant people. For a subset of pregnant people, it will also be important to highlight the duration of protection for different preventive options. Maternal vaccine was preferred by most pregnant people in this study; yet, among some pregnant people who are reluctant to receive vaccination, other preventive options, including long-acting mAbs, may potentially be viewed as more acceptable.

Limitations

The DCE did not include all factors that could potentially influence the preferences of pregnant people regarding RSV preventives, such as preventive product safety. Nevertheless, prior DCE studies on the preferences of parents for vaccinating their children have found that safety-related attributes are among the least influential drivers of vaccine choice.^35,36 Further, vaccines must be demonstrably safe before they will be granted FDA approval in the US. Nonetheless, we cannot exclude the possibility that perceptions about safety could also be reflected in the preference of the No Preventive Group for mAbs over maternal vaccine to the extent that concerns about vaccine safety overshadow any concerns about the safety of mAbs.

Moreover, pregnant people may prefer to choose to opt-out of immunization because they are concerned about the safety of both types of preventives, either for themselves and/or their infant. Should this be the case, we would expect that a preventive with an unfavorable safety profile could increase the probability of opting out, particularly among those in the No Preventive Group, although this will need to be verified in future research specifically focusing on RSV preventives in the context of maternal and infant immunization.

Factors associated with preterm birth may influence the RSV preventive preferences of pregnant people, including having a previous preterm birth, a newborn with health issues, or problems with the present pregnancy that increase the risk of preterm birth. Given preterm infants are at greater risk of severe illness from RSV, discussing RSV prevention options with all pregnant people, but particularly those at increased risk for preterm birth, may be especially important. Future research will be needed to understand how these factors influence the RSV preventive preferences of pregnant people who are at higher risk for preterm birth to further inform public health education and communication efforts.

Hypothetical preventive profiles may differ from the clinical profiles of currently available preventives, although the profiles presented in the DCE aligned with the data published on those products at the time the study was conducted. Of note, hypothetical preventive profiles did not include an attribute to represent availability or accessibility of preventives, which could influence real-world preventive choice, given the limited supplies of nirsevimab for the 2023–2024 RSV season.³⁸ It is also possible that the implementation of an alternative approach to evaluate preference heterogeneity (e.g., random-parameters logit models) may have yielded a somewhat different pattern of correlations between preferences and respondent characteristics,²⁵ although the approach applied in the current study (LCA) was appropriate for addressing the research questions of interest with the observed data. At the same time, there is no reason to believe that the overall results and conclusions from an alternative approach would materially differ from those found in the current study. While the use of a convenience sample may potentially limit generalizability of results to the broader population of pregnant people in the US, the sample of pregnant people in the current study was generally similar to US national data in terms of age, number of prior births, education, and race.^39,40

Conclusions

The current study provided evidence of naturally occurring subgroups of pregnant people who have similar RSV preventive preferences. Specifically, three segments of pregnant people who varied in their receptivity to RSV preventives were identified. RSV preventive effectiveness was important to preventive choice for all groups, although for some pregnant people, duration of protection during RSV season was the foremost consideration. While nearly all pregnant people preferred a maternal vaccine, a subset who did not prefer vaccination may potentially be more willing to accept alternative preventive options. This study not only identified the target groups who may be less willing to immunize but also facilitated a better understanding of the barriers to immunization that could be influenced by targeted information or education. Our findings suggest that patient educational efforts should focus on improving knowledge and awareness about the seriousness of RSV among pregnant people who are not currently employed to potentially increase the willingness to accept an RSV preventive option among the subset of pregnant people who are most likely to opt-out of immunization. Ultimately, these findings can help to inform the development of targeted public health initiatives, which appeal to subpopulations of pregnant people by differentially emphasizing effectiveness and duration of protection during RSV season, to increase uptake of RSV preventives and to ensure that immunization decision-making is in accord with pregnant people’s values and preferences.

Author contribution statement

All authors contributed to the conception and design of the study, interpretation of the data, and revising the manuscript critically for intellectual content. OW was responsible for data analysis; MCM was responsible for drafting the initial version of the manuscript. All authors approve of the version to be published and agree to be accountable for all aspects of the work.

Acknowledgments

The authors acknowledge Kari Yacisin and Jessica Atwell (employees of Pfizer Inc.) for their contributions to the various phases of the project.

Disclosure statement

MCM, KMB, OW, and LK are employees of Oracle Life Sciences, which received funds from Pfizer for study conduct and manuscript development. BH, AWL, JTV, JCC, JRC, and SP are employees of Pfizer and may own Pfizer stock; KMS was an employee of Pfizer at the time the study was conducted and may own Pfizer stock.

Additional information

Funding

The study was funded by Pfizer.

Appendix

A

Figure A1. Example discrete choice experiment choice task.

Options for effectiveness: 50%, 63%, 77%, 90%; injection recipient/timing: pregnant person’s arm before their baby is born, injection into infant’s leg 14 days after birth, injection into infant’s leg 2-4 months after birth; duration of protection (months during the RSV season): 1, 3, 4.5, 6. Injection recipient and timing of protection were presented as separate attributes for ease of comprehension, but the attributes were linked in how they were presented to preclude illogical combinations (e.g., “Protection begins at birth” never appeared with “Single injection into a baby’s leg 2-4 months after birth”); they were treated as one attribute in analyses. FDA: Food and Drug Administration; RSV: respiratory syncytial virus.

Figure A2. Most and least favorable RSV preventive profiles.

RSV: respiratory syncytial virus.

Figure A3. Probability of choosing the no-preventive option versus the most and least favorable RSV preventive profiles by latent class.