Web-based intervention for improving influenza vaccination in pregnant women: a cost-effectiveness analysis

ABSTRACT

A website with vaccine information and interactive social media was reported to improve maternal influenza vaccine uptake. This study aimed to evaluate cost-effectiveness of a web-based intervention on influenza vaccine uptake among pregnant women from the perspective of US healthcare providers. A one-year decision-analytic model estimated outcomes in a hypothetical cohort of pregnant women with: (1) website with vaccine information and interactive social media (intervention group), and (2) usual care (usual care group). Primary measures included influenza infection, influenza-related hospitalization, mortality, direct medical cost, and quality-adjusted life-year (QALY) loss. In base-case analysis, intervention group reduced cost (by USD28), infection (by 28 per 1,000 pregnant women), hospitalization (by 1.226 per 1,000 pregnant women), mortality (by 0.0036 per 1,000 pregnant women), and saved 0.000305 QALYs versus usual care group. Relative improvement of vaccine uptake by the intervention and number of pregnant women in the healthcare system were two influential factors identified in deterministic sensitivity analysis. The intervention was cost-effective in 99.5% of 10,000 Monte Carlo simulations (at willingness-to-pay threshold 50,000 USD/QALY). A website with vaccine information and interactive social media to promote influenza vaccination for pregnant women appears to reduce direct medical costs and gain QALYs from the perspective of US healthcare providers.

Introduction

Influenza is associated with severe disease among pregnant women [1,2]. The US public health service has recommended influenza vaccination of pregnant women since 1961, and currently the US Advisory Committee on Immunization Practices recommend influenza vaccination for persons who are pregnant or who might be pregnant during the influenza season [3,4]. Health economic findings have consistently shown influenza vaccination to be cost-effective for pregnant women during influenza season in the US [5–7]. During the 2021–2022 influenza season, vaccination coverage among pregnant women in the US was lower (49.6%) than the previous season (54.5%) [8]. Most pregnant women are unaware of the risk of severe complications associated with influenza infection. The healthcare provider has an important role to inform pregnant women about the effectiveness and safety of the influenza vaccine. Receiving a healthcare provider’s recommendation is the most frequently reported reason for vaccine acceptance among pregnant women [9,10].

A website with vaccine information and interactive social media components, provided by a healthcare provider, was found to be effective in improving maternal influenza vaccine uptake in a randomized controlled trial (n = 289 for influenza vaccination). The influenza vaccine uptake in the intervention group was significantly higher than the usual care group (57% versus 36%; p = 0.01) [11]. Despite the positive clinical findings of the web-based intervention on maternal vaccine uptake, the cost-effectiveness of the web-based intervention is yet to be examined. To inform healthcare providers on the decision-making process of resource allocation, this study aimed to evaluate the cost-effectiveness of a web-based intervention to increase influenza vaccine uptake among pregnant women from the perspective of US healthcare providers.

Methods

Model design

A decision tree model (Figure 1) was developed to estimate the clinical outcomes and costs in a hypothetical cohort of pregnant women with two interventions: (1) a website with vaccine information and interactive social media (intervention group), and (2) usual care (usual care group). The decision tree is a form of decision-analytical modeling in which hypothetical individuals proceed to different health states over the model timeframe. Decision-analytic modeling is a well-established and widely used tool to conduct health economic analysis of different interventions and programs in the healthcare system. It provides a computational framework to generate the health economic and clinical outcomes of health interventions by integrating findings from multiple sources (including published clinical trials and public data). The expected cost and clinical outcomes generated by decision-analytical modeling provide cost-effectiveness data to assist informed decision-making on allocation of healthcare resource [12,13]. In the present model, the model time horizon was one year. Primary model outcomes included influenza infection rate, influenza-related hospitalization rate, mortality rate, direct medical cost, and quality-adjusted life-year (QALY) loss.

Figure 1.

Simplified decision-analytical model for strategies to promote influenza vaccination in pregnant women. ICU: intensive care unit.

At the entry of the model, the hypothetical pregnant women in the intervention group received the vaccine information website link from an e-mail sent by the healthcare provider. The components of the intervention adopted the website with vaccine information and interactive social media examined in clinical trials [11,14]. The website information included the US national vaccine recommendations on influenza during pregnancy, details on vaccine safety, descriptions of influenza infection, and answers to common vaccine concerns. The information was presented in short and easy-to-read sections. The website also provided access to a moderated interactive social media with a blog, discussion forum, chat room, and portal (for asking questions). New blog posts about the latest vaccine-related research or policy were displayed monthly and an online conversation with experts was hosted for all users. E-mail reminders of updated content were sent monthly. In the usual care group, pregnant women received routine prenatal care and did not have access to the vaccine information website.

In both study groups, the hypothetical pregnant women might receive the influenza vaccine. All pregnant women (vaccinated or unvaccinated) might be infected by influenza. The influenza-infected patients might have self-care. Those who sought medical care might receive outpatient or inpatient care (with or without intensive care unit (ICU) admission). Pregnant women who were hospitalized for influenza infection might survive or die.

Clinical inputs

All model inputs are listed in Table 1. A MEDLINE search was performed from 2000 to 2023 using the keywords such as ‘influenza infection,’ ‘influenza vaccine,’ ‘pregnant women,’ ‘influenza hospitalization,’ ‘influenza mortality,’ and ‘utility score.’ The selection criteria of clinical studies of influenza infection were: (1) Reports were written in English; (2) influenza infection was diagnosed; and (3) treatment outcomes of influenza infection were reported. A study was included if the data relevant to the model inputs were available. The weighted average was adopted as the base-case value if multiple sources were found suitable for model input. The high and low values of the variable reported in the literature formed the range in the sensitivity analysis.

Table 1.

Model inputs.

Parameters	Base-case value	Range for sensitivity analysis	Distribution	Reference
Clinical inputs
Vaccine uptake rate in usual care group	0.3636	0.249–0.494	Beta	[11]
Relative improvement of vaccine uptake by the intervention	1.57	1.0–2.0	Triangular	[11]
Influenza vaccine effectiveness against infection	0.54	0.23–0.73	Beta	[15]
Influenza vaccine effectiveness against hospitalization	0.40	0.12–0.59	Beta	[16]
Influenza attack rate	0.251	0.115–0.375	Beta	[15]
Probability of seeking medical care	0.37	0.35–0.40	Beta	[17]
Probability of hospitalization in unvaccinated pregnant women	0.0879	0.0768–0.1007	Beta	[18,19]
Probability of intensive care unit admission in pregnant women	0.0483	0.039–0.063	Beta	[20]
Mortality rate during hospitalization in pregnant women	0.0030	0.001–0.005	Beta	[20]
Utility inputs
Utility score of pregnant women	0.996	0.985–1	Beta	[21]
Utility score of self-care influenza	0.725	0.61–0.84	Beta	[22]
Utility score of outpatient influenza	0.65	0.49–0.81	Beta	[7]
Utility score of hospitalization	0.50	0.38–0.63	Beta	[7]
Utility score of intensive care unit admission	0.38	0.304–0.456	Beta	[23]
Age of pregnant women	30	18–44	Normal	[24]
Length of illness for outpatient/self-care influenza (days)	9.3	8.5–10.1	Normal	[25]
Length of hospitalization for influenza (days)	2.0	1–3	Normal	[20]
Length of intensive care unit care for influenza (days)	5.5	5.42–5.58	Normal	[26]
Cost inputs (USD)
Vaccine price (per dose)	14.24	13.41–18.07	Gamma	[27]
Cost of vaccine administration	15.59	10.57–20.61	Gamma	[7]
Cost of website maintenance (per year)	4,080	3,264–4,896	Gamma	[28]
Cost of reminder (per e-mail)	1.03	0.82–1.23	Gamma	[29]
Cost of over-the-counter products	8.72	6.54–10.90	Gamma	[30]
Cost per outpatient visit	541	358–724	Gamma	[31]
Cost per hospitalization	32,492	30,662–34,321	Gamma	[26]
Cost per intensive care unit admission	59,445	57,528–61,363	Gamma	[26]
Probability of using over-the-counter products in pregnant women	0.92	0.74–1	Beta	[25]
Number of clinic visits	1	1–3	Uniform	Assumption
Number of pregnant women in the healthcare system	5,000	50–5,000	Normal	[14]

The relative improvement of vaccine uptake (1.57) by the intervention group versus the usual care group was derived from the uptake rates of the intervention group (57.14%) and the usual care group (36.36%), retrieved from a randomized controlled trial (n = 289 for influenza vaccination) of the website with vaccine information and interactive social media components to increase vaccine uptake in pregnant women [11]. Influenza vaccine effectiveness (IVE) against infection varies by season depending on the antigenic match of viruses in the vaccine and the circulating subtype viruses [32]. The IVE against infection among pregnant women had been reported to be similar to the IVE of non-pregnant individuals [33]. The model input of IVE against infection therefore adopted the interim estimates of IVE against influenza A (54%; range 23%-73%) among persons aged 6 months–64 years in the 2022–23 season of the US [15]. The influenza attack rate among unvaccinated pregnant women also adopted the attack rate in 2022–23 seasonal influenza as base-case value, ranging from the lowest and highest attack rates reported from 2010–11 to 2022–23 (25.1%; 11.5%-37.5%) [15,34]. The incidence of influenza infection in vaccinated pregnant women was estimated by the following formula: Influenza attack rate*(1- IVE against infection).

The probability of pregnant women seeking medical care was assumed to be the same as the probability of adults aged 18–49 (37%), reported in a US epidemiology study (n = 216,431 adults) of self-reported influenza-like illness data during the 2009 influenza pandemic [17]. The proportion of hospital admission among those who sought medical care (1.5%) was estimated from the CDC-reported numbers of influenza-related hospitalizations and medical visits [18]. Pregnant women were at higher risk of influenza-associated hospitalization when compared to non-pregnant individuals (RR 5.86; 95% CI 5.12–6.71) [19]. The proportion of hospitalization among infected unvaccinated pregnant women who had sought medical care (8.79%; range 7.68%-10.07%) was therefore estimated by the corresponding proportion of hospitalization among influenza-infected adults aged 18–49 years (who had sought medical care) and the relative risk of influenza-related hospitalization in pregnant women. A retrospective study (n = 19,450 hospitalizations) on influenza vaccine in preventing influenza-related hospitalizations (with acute respiratory or febrile illness) during pregnancy reported that the influenza vaccine offered protection against influenza-associated hospitalizations during pregnancy with an estimated IVE of 40% against influenza-related hospitalizations [16]. The proportion of hospitalization among infected vaccinated pregnant women was approximated by the corresponding hospitalization among infected unvaccinated pregnant women and the IVE against hospitalization. The occurrence of ICU admission (4.83%) and death (0.30%) during hospitalization were sourced from a population-based US surveillance with repeated cross-sectional study on the outcomes of hospitalized pregnant women with influenza (n = 9,652) from 2010 to 2019 [20].

Utility inputs

The expected influenza-related QALY loss in each study group was calculated using the difference between the utility of an influenza-relate health event and the utility of pregnancy, and the patient-time spent in the corresponding event. Events included in the model were: (1) Self-care influenza; (2) outpatient care; (3) hospitalization without ICU admission; and (4) ICU care. The utility values of influenza-related health events were estimated from findings of health-related quality of life studies [21] and cost-effectiveness analysis studies in maternal influenza vaccinations [7,22,23]. Life-long QALY loss due to influenza-related death was estimated using the age-specific expectation of life from the US life table 2020 [35] and the age-specific health utilities [36], discounted to the current year with an annual rate of 3%. The base-case age of pregnant women was 30 years old in the model (ranging from 18–44 years), adopting the median age of pregnant women in the US reported by the US Census Bureau [24].

Costs inputs

The present study was performed from the perspective of US healthcare providers, and direct medical costs were included in the cost analysis. Cost items included intervention delivery, vaccination, and influenza treatments. Since the model horizon was one-year, discounting costs was therefore unnecessary. All costs were adjusted to the year 2023 using the medical Consumer Price Index provided by the US Bureau of Labor Statistics [37].

The costs of sending e-mail reminders to pregnant women and website maintenance in the intervention group adopted the estimated cost of similar interventions reported by clinical trials on web-based interventions [28,29]. The set-up costs of the website were not included in the model, as the intervention was delivered via the existing online platform of the healthcare provider. Website maintenance costs included technical and content maintenance costs associated with the online information update, checking comments, and answering questions. Annual website maintenance costs were divided by the total number of pregnant women enrolled in the care of the healthcare system to estimate the cost per head. The base-case number of pregnant women was 5,000, adopting the estimated annual number of pregnant women served by a healthcare system in the US [14]. Vaccination costs included the costs of vaccine and vaccine administration. The vaccine cost was obtained from the 2022–23 adult influenza vaccine price listed by CDC [27]. Cost items for influenza treatment were self-care, outpatient care, and hospitalization (with and without ICU care). The cost of self-care was calculated using the costs of over-the-counter (OTC) medications estimated from an economic burden research [30] and the proportion of pregnant women who used OTC products (92%) [25]. Outpatient care cost was retrieved from an online dataset (Medical Expenditure Panel Survey Household Component) from the Agency for Healthcare Research and Quality (AHRQ) [31]. The number of clinical visits for outpatient was assumed to be one time (range 1–3 visits). Hospitalization cost was retrieved from the AHRQ’s Healthcare Cost and Utilization Project (inpatient) dataset [26].

Cost-effectiveness analysis and sensitivity analysis

All analyses were conducted in TreeAge Pro 2023 (TreeAge Software, Inc., Williamstown, MA, U.S.A.) and Microsoft Excel 2019 (Microsoft Corporation, Redmond, WA, U.S.A.). The intervention group was considered cost-effective if it resulted in (1) QALY gain and cost-saving, or (2) QALY gain at a higher cost and the incremental cost-effectiveness ratio (ICER =Incremental Cost/QALY gain) was lower than the willingness-to-pay (WTP) threshold per QALY gained. A WTP threshold of 50,000 USD/QALY gained was adopted in the present analysis. If the increment cost was negative value (cost of intervention group < cost of usual care group), the incremental net monetary benefits (INMB = QALY gain*WTP- incremental Cost) would be calculated. The intervention group was accepted as cost-effective when the INMB was a positive value, meaning that the monetary benefits of the web-based intervention outweighed the incremental cost incurred by the intervention [38]. The magnitude of cost-effectiveness is directly proportional to the (positive) value of INMB.

The sensitivity analysis was performed to test the robustness of the model results. Deterministic sensitivity analysis on all parameters was conducted to identify influential factors. All the parameters were varied over the upper and lower limits, if available. Otherwise, a 95% confidence interval (CI) or a range of variation by ±20% of the base-case value was used. A probabilistic sensitivity analysis was performed using Monte Carlo simulation to evaluate the impact of uncertainty in all variables simultaneously. The direct cost and QALYs of each group were recalculated 10,000 times by randomly drawing each of the model inputs from the parameter-specific distribution. The range and distribution of each parameter are listed in Table 1.

Results

Base-case analysis

The expected influenza infection rate, influenza-associated hospitalization rate, mortality rate, medical direct cost, and QALY loss of each study group in the base-case analysis are shown in Table 2. When compared to the usual care group, the intervention group reduced the infection rate (by 28 per 1,000 pregnant women), hospitalization rate (by 1.226 per 1,000 pregnant women), mortality rate (by 0.0036 per 1,000 pregnant women), direct cost (by USD28 per pregnant woman), and gained 0.000305 QALY per pregnant woman. The estimated INMB was 43. The intervention group was therefore the preferred cost-effective option to promote maternal influenza vaccination from the perspective of the healthcare providers in the US.

Table 2.

Base-case analysis results.

Strategy	Influenza infection ratea	Influenza-associated hospitalization ratea	Influenza-associated mortality ratea	Cost (USD)	Incremental cost (USD)	QALY loss	QALY gain	INMB (USD) at WTP = 50,000 USD/QALY
Intervention	173	4.780	0.0142	227	−28	0.00168	0.0003	43
Usual care	201	6.007	0.0179	255	-	0.00198	-	-

Sensitivity analysis

The tornado diagram (Figure 2) shows the one-way sensitivity analysis on the INMB of the intervention group. Two influential factors with threshold values were identified: Relative improvement of vaccine uptake by the intervention and the number of pregnant women in the healthcare system. The INMB of the intervention group became a negative value (indicating not cost-effective) when the relative improvement of vaccine uptake by the intervention was lower than 1.13 (base-case value 1.57), or when the number of pregnant women in the healthcare system was less than 93 (base-case value 5,000).

Figure 2.

Tornado diagram of variation of INMB (intervention versus usual care) against all model inputs in one-way sensitivity analysis. QALY: quality-adjusted life year; INMB: incremental net monetary benefits.

The variation of costs and QALY loss against the relative improvement of vaccine uptake by the intervention were shown in Figure 3. Compared to the usual care group, the intervention group was less costly and gained QALYs when the relative improvement of vaccine uptake by the intervention was higher than 1.18. When the relative improvement ranged between 1.13 and 1.18, the intervention remained to be the cost-effective option with ICER below the WTP threshold.

Figure 3.

(A) costs and (b) QALY loss of the intervention versus usual care against the relative improvement of vaccine uptake by the intervention. QALY: quality-adjusted life year.

A two-way sensitivity analysis was further performed on the two influential factors to explore their impact on the cost-effectiveness of the intervention (Figure 4). The combination of relative improvement of vaccine uptake by the intervention and number of pregnant women in the light gray area indicated that the intervention to be cost-effective, and the combination of variables in the dark gray area indicated that the usual care became the cost-effective option.

Figure 4.

Two-way sensitivity analysis of relative improvement of vaccine uptake by the intervention and the number of pregnant women in the healthcare system on cost-effectiveness of the web-based intervention.

Probabilistic sensitivity analysis was conducted by 10,000 Monte Carlo simulations of the cost and QALY loss in both groups. The incremental cost and QALY gained by the intervention group versus the usual care group are shown in the scatter plot (Figure 5). When comparing to the usual care group, the intervention group saved cost by USD27.87 (95% CI USD27.57-USD28.17, p < 0.01) and gained QALY by 0.000305 (95% CI 0.000302–0.000307, p < 0.01). The intervention group saved QALYs and cost in 100% and 99.39% of the 10,000 simulations, respectively. The probability of the intervention being accepted as the cost-effective option was 99.5% at the WTP threshold of 50,000 USD/QALY.

Figure 5.

Scatter plot of incremental cost versus incremental QALYs gained by the intervention versus usual care in 10,000 Monte Carlo simulations. QALY: quality-adjusted life year; WTP: willingness-to-pay.

QALY: quality-adjusted life year; INMB: incremental net monetary benefits = (QALY gain*WTP- incremental Cost); WTP= willingness-to-pay

a: Event rate per 1,000 pregnant women

Discussion

The present cost-effectiveness analysis examined the outcomes of a website intervention with vaccine information and interactive social media for promoting influenza vaccination during pregnancy in the US, measured as influenza infection rate, influenza-associated hospitalization rate, mortality rate, direct medical cost, and QALY loss. The base-case analysis found that the intervention was cost-effective in saving both QALY (by 0.00031 per pregnant woman) and costs (USD28 per pregnant woman) from the perspective of healthcare providers. The relative improvement of vaccine uptake by the web-based intervention (>1.13) and the number of pregnant women in the healthcare system (>93) were two influential parameters. The probabilistic sensitivity analysis further supported the cost-effectiveness of the intervention group in 99.5% of 10,000 Monte Carlo simulations at WTP 50,000 USD/QALY. The present findings showed that, with the improvement of vaccine uptake rate benefited from the web-based intervention, the consequential reduction in infection, hospitalization and mortality rates of pregnant women collectively contributed to cost savings (in influenza-associated treatment) and QALY gain (in reduced infection, hospital/ICU admission and death).

This was the first study to evaluate the cost-effectiveness of using a website with vaccine information and interactive social media to promote maternal influenza vaccination in the US. The cost-effectiveness of the television campaign was reported to be USD17.79 per additional vaccinated elderly person in the US, though the influenza-related treatment cost was not considered in the cost estimation [39]. In the present study, the cost of the web-based intervention was offset by the cost reduction in influenza-associated treatment cost. The total cost of the intervention was therefore lower than the total cost of usual care. The prior study on television campaigns might likely found the cost per vaccinated elderly person further lowered (or even cost-saving) if the total cost of influenza-related events were considered.

The relative improvement of vaccine uptake via this web-based intervention represented the effectiveness of the intervention. The relative improvement by the web-based intervention is needed to reach the threshold (>1.13 as indicated by the one-way sensitivity analysis) to generate adequate savings in cost and QALY to meet the cost-effectiveness requirements per WTP. The website maintenance cost per head was directly related to the number of pregnant women in the healthcare system, and the number of pregnant women was a critical factor in the cost of intervention per pregnant woman. The cost per individual became lower when the web-based intervention had a large volume of participants [40]. It is therefore essential for healthcare providers to build a well-established feedback system to continuously improve the function of the website and maintain a high subscription volume.

The one-way sensitivity analysis of our study further found that the influenza attack rate, despite no threshold value identified, was also a main cost-effectiveness driver of the web-based intervention (as indicated by the tornado diagram of INMB). Due to the seasonal variation, the influenza season severity can be divided into low, moderate, and high depending on the combined consideration of the proportion of influenza-like illness, hospitalization rate, and mortality rate [41]. The positive impact of vaccination in terms of influenza-related cost reduction and QALY gained was further enhanced during high severity season. The web-based intervention to promote influenza vaccination among pregnant women was therefore expected to be cost-effective, especially in a high-severity season. The circulation of influenza was decreased during the COVID pandemic due to the measures such as masking, social distancing, and working from home. With the removal of these anti-COVID restrictions worldwide, the severity of influenza had resumed (or even exceeded) pre-pandemic levels [42–44]. Cost-effective interventions to promote influenza vaccination among high-risk groups are essential in the post-pandemic era. With the high utilization of digital technology for daily activities during the COVID pandemic among the general public, the web-based intervention to promote influenza vaccination in pregnant women in the post-pandemic era is a timely and cost-effective strategy. Future studies on the utilization of web-based interventions to promote vaccination among high-risk groups of different demographics are warranted.

There were some limitations in this study. Decision-analytical modeling is in general subject to the uncertainty of model inputs and simplification of the disease complexity. The IVE model input was adapted from the CDC findings on vaccine effectiveness against influenza for individuals aged 6 months to 64 years in season 2022–2023. The CDC data found only influenza A samples and no influenza B virus-positive sample was received during the reporting period. The model input for IVE was therefore limited to non-age-specific effectiveness against influenza A. The probability of pregnant women seeking medical care was adapted from the epidemiological findings of the 2009 influenza pandemic, and this model input was limited by the lack of more recently published findings. Rigorous sensitivity analysis was therefore performed on all model inputs to evaluate the impact of model input uncertainty on the base-case cost-effectiveness results. In addition, the model design simplified the influenza-related events in pregnant women and did not include vaccine protection for infants of a vaccinated mother or the impact of herd immunity generated by a higher vaccine coverage rate. The present findings might therefore underestimate the benefits of the web-based intervention to promote influenza vaccination among pregnant women.

Conclusion

A website with vaccine information and interactive social media to promote influenza vaccination for pregnant women appears to reduce influenza-related total direct medical costs and gain QALYs from the perspective of US healthcare providers. The cost-effectiveness acceptance of the web-based intervention is subject to the relative improvement in vaccine uptake generated by the intervention (when compared to usual care) and the number of pregnant women in the healthcare system.

Funding Statement

The author(s) reported work featured in this article was funded by Direct Grant for Research, The Chinese University of Hong Kong (grant number 2022.079).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

All the data related to this study are available in the manuscript.