Modeling the economic impact of different vial-opening thresholds for measles-containing vaccines

2.1. HERMES supply chain modeling software

HERMES is a discrete event simulation modeling software platform that can generate computer simulation models of supply chains. A HERMES-generated supply chain model is a virtual representation of all storage facilities (including refrigerators and freezers), transport routes (including shipping policies, vehicles, drivers, and per diems), healthcare personnel, vaccines, administration points (including vial opening policies and population demand), and includes the associated costs for each component. Previous publications have described HERMES models in detail [13], [14].

2.2. HERMES vaccine supply chain models: Benin, Mozambique and Niger

Our team used HERMES simulation models of the routine immunization supply chains of Benin [15], [16], [17], Mozambique [18], and Niger [19], [20]. The routine immunization programs in these countries are eligible for Gavi support [21] and their supply chains provide a range of different characteristics found in Gavi-supported vaccine supply chains [22]. The supply chain model of Benin is comprised of one national store, seven department stores, 80 communes, and 763 health posts and was developed using cold chain and population data from 2012. The supply chain model of Mozambique is comprised of one national store, ten provincial stores, 104 district stores, 1428 health facilities, and 254 mobile brigades and was developed using cold chain and population data from 2014. The supply chain model of Niger includes one central store, eight regional stores, 42 district stores, and 695 integrated health centers and was developed using cold chain data from 2011 and population data from 2015.

Benin’s routine immunization program provides one dose of measles (M) vaccine, while Mozambique and Niger’s routine programs provide two doses of measles-rubella (MR) vaccine. Each country has set a 40% maximum wastage rate for MCVs [23], [24], [25]. The vaccines modeled reflect the current EPI national immunization schedule for each country as described in Appendix A.

In HERMES, the expected demand per session – referred to as average session size – is determined by dividing the annual demand at a health facility (i.e. the total number of people seeking vaccination at a certain clinic for a certain vaccine over the course of one year) by session frequency (i.e. the number of vaccination sessions held annually), both of which are model inputs (see equation below). Data for these inputs was collected during construction of the model with direct input from country partners.

To better simulate reality, HERMES employs a stochastic model of consumer arrival. For each session, HERMES calculates the actual session size by randomly drawing from a Poisson distribution, whose mean parameter is the average session size.

2.3. Measles-containing vaccine (MCV) vial-opening threshold policies and presentations

In HERMES, a vaccine vial is opened for any number of people unless the user sets a policy dictating the number of people that need to be present in a vaccination session before a vaccine is opened for use. In each country, we modeled three vial-opening threshold policies to capture a range of existing policies and practices across different immunization programs.

• •

  Instituting no vial-opening threshold policy allowed MCVs to be opened for one or more children. This practice has high potential for MCV wastage, but is advocated for in literature and guidelines as a potential solution to improving MCV coverage and reducing missed opportunities to vaccinate [2], [5].

• •

  Instituting a 30% vial-opening threshold policy meant that MCVs would be opened only if the demand in a given session would use 30% of MCV doses, i.e. a 10-dose MCV would be opened if at least three children were present and a 5-dose MCV would be opened if at least two children were present. This threshold applies to each MCV vial needed such that if a vaccination session has 11 or 12 children, only one 10-dose vial of MCV would be opened. This policy is not referenced as current practice in any countries but serves as an intermediary scenario for the purpose of this study.

• •

  Instituting a 60% vial-opening threshold policy restricted the use of MCVs further, i.e. 10-dose MCV vials would be opened if at least six children were present and a 5-dose MCV would be opened if at least 3 children were present. This more restrictive practice is commonly identified as a way to meet MCV wastage targets, and is practiced by healthcare workers in a number of countries [1], [4].

In addition to simulating three different vial-opening thresholds, we considered the effects of these policies across different presentations of MCV. Currently, 10-dose MCV vials are used, although UNICEF is supplying 5-dose MCV vials as of 2018 [26]. As countries consider introducing 5-dose MCV, which has a higher volume- and price-per-dose, into their supply chain systems, we consider the effects of different vial-opening thresholds across three different MCV presentation scenarios:

• •

  Distributing 10-dose MCV vials to all vaccinating locations in a country (current practice)

• •

  Distributing 5-dose MCV vials to all vaccinating locations in a country

• •

  Distributing a mix of 10-dose and 5-dose MCV vials according to the average session size (see Section 2.2) of each vaccinating location, i.e. locations with average session sizes of 5 or fewer children receive 5-dose MCV, while locations with average session sizes of 6 or more children receive 10-dose MCV

Lastly, in addition to Benin, Niger, and Mozambique’s vaccine supply chains, we ran each of the above scenarios using an unconstrained Niger supply chain model. To do this, we ran a gap analysis in HERMES that calculated all of the necessary storage and transport space needed to achieve 100% vaccine availability and added this equipment to the Niger model. We then tested the effects of different MCV vial-opening thresholds and presentations on this hypothetical, unconstrained system.

2.4. Clinical and economic model for EPI vaccine-preventable diseases

In addition to calculating the effects of different MCV vial-opening threshold policies and presentations on supply chain operations and costs, we developed an economic and clinical outcomes model to calculate the incremental cost-effective ratio (ICER) of each scenario. This economic model takes a health systems perspective, which includes health care costs paid by third-party payers and out-of-pocket by patients [27]. We estimated the cost of healthcare delivery, assuming that patients would pay at-cost without significant profit margin. To this end, the model includes the cost of supply chain logistics for vaccines, vaccine procurement prices, and direct medical treatment costs associated with the vaccine-preventable infections of each modeled disease. We utilized disease incidence data from the Institute for Health Metric’s Global Burden of Disease (GBD) and the most recent authoritative meta-analysis from West Africa (see Appendix B for incidence data for each disease and relevant citations) to calculate vaccine-preventable infections (see Section 2.5 below).

We calculated total direct medical costs using a bottom-up approach, including the cost of visits to outpatient health posts, the daily cost of hospitalization, and pharmaceutical costs. We did not include the cost of diagnostics, complex procedures (that are often not available in the modeled countries), medical care to manage chronic disability, nor the indirect patient cost of time spent seeking care.

To avoid underestimating the cost of disease burden, we assumed that all individuals with acute symptoms would seek and receive the correct care for their condition, as defined by Medecins Sans Frontieres (MSF) guidelines for working in low-resource settings [28].

We calculated hospitalization costs by multiplying the probability of admission with the average length of stay [29] and the country-specific cost per day for inpatient care. Health post visit costs were calculated based on one visit per infection for those individuals not hospitalized. Unit costs per day and per visit were country-specific and based on the WHO’s CHOICE estimates for public facilities, updated to 2017 GDP and rural/urban population mix [30].

For all diseases except tuberculosis (TB), we based pharmaceutical costs on UNICEF’s published international indicative prices, assuming that public health providers would purchase supplies in reasonable bulk sizes when it affords lower prices. The price of immunoglobulin for tetanus treatment was taken from the WHO’s reported average cost [31]. We applied costs for symptom control drugs (i.e. fever reduction, pain management) to every infection and treatment drugs (i.e. antibiotics) to every hospitalization. We calculated dosing levels based on the amount required to care for a 20 kg 5-year old, since the largest portion of disease burden for most vaccine-preventable diseases is in young children.

For TB, we utilized pharmaceutical costs from the most recent cost catalog by the Stop TB program [32], a partner of the Global Fund to fight AIDS, Tuberculosis, and Malaria and is the largest funder of TB treatment globally. Dosages were calculated based on a 60 kg adult, since most TB burden is latent until later in life. We based our calculations on the WHO’s recommended regimen for new patients of daily administration for 2 months with HRZE and 3-times weekly administration for 4 months of HR in fixed dose combination formulations (FDC) [33].

2.5. Simulation scenarios for MCV vial-opening thresholds and presentations

For each country, we simulated nine scenarios pairing one of three MCV vial-opening thresholds with one of three MCV vial presentation scenarios (as described in Section 2.3). In the baseline scenario, 10-dose MCV vials were opened for any number of children (i.e. no vial-opening threshold). In each scenario, running the HERMES model determined how many vaccine-preventable infections occurred in the population:

The following formulas then determined the medical costs and disability-adjusted life years (DALYs) from the infections prevented:

We assessed the impact of uncertainty in the model parameters by using a Monte Carlo simulation to simultaneously vary the parameters presented in Appendix B. The simulation sampled values for the DALYs per case, direct medical cost per case, vaccine effectiveness, and disease incidence. Ranges for DALYs, vaccine effectiveness, and incidence were simulated as triangular distributions across their 95% confidence interval (CI) values. Uncertainty ranges for hospitalization day rates and outpatient visit unit costs were estimated based on previous CIs calculated for these costs in Mozambique [30]. Pharmaceutical prices are established on the international market, so we did not simulate uncertainty in their pricing. We generated 1000 stochastic results for each disease-country combination and then aggregated the results and calculated the simulated 95th percentile prediction intervals.

Additionally, we calculated the incremental cost-effectiveness ratio (ICER) of each experimental scenario compared to the scenario with the next-most DALYs (i.e. next best scenario). Scenarios were considered dominated if there was an alternative scenario which resulted in both lower costs and fewer DALYs.

Cost-effectiveness is defined as an ICER ratio <3× gross domestic product (GDP) per capita, while highly cost-effective is <1× GDP per capita. In the last reported year (2017) GDP per capita was $829.8 in Benin, $378.1 in Niger, and $415.7 in Mozambique.

3.1. Requiring HCWs to open MCVs for any number of children per session (No vial-opening threshold)

Across all three vaccine supply chains, instituting no vial-opening threshold for 10-dose MCV (baseline) resulted in the highest MCV wastage (33–47%) compared to all other scenarios. As shown in Fig. 1, opening 10-dose MCV vials for any number of children results in decreased MCV vaccine availability (i.e. successful immunizations administered to patients as a percentage of total immunizations needed) compared to other scenarios, except in Niger when all of the storage and transport constraints are removed. Given the existing cold chain and transport constraints in each system (Table 1), many vaccinating locations will not be able to maintain enough stock between shipments if an MCV vial is opened for even one or two children. Further, since each vaccinating location in these countries has an average session size of three or more children, using MCV for one or two children means that a stockout would likely occur during a larger session.

In terms of overall vaccine availability (Figs. 2 and 3), however, this policy outperforms many others, particularly in Mozambique and Niger, which both have highly constrained cold chain systems. When 10-dose vials are used everywhere, the total volume of MCV flowing through the system is smaller than when 5-dose MCV or a mix of 5-dose and 10-dose MCV are introduced. As such, this policy does not impede the flow of other vaccines. In Mozambique, in particular, this policy results in the second-fewest DALYs from all EPI vaccine-preventable diseases out of any scenario (see Fig. 4).

Conversely, switching to 5-dose vials everywhere and maintaining no vial-opening threshold only benefits Benin, which has the fewest cold chain constraints among these systems. Implementing no vial-opening threshold using 5-dose MCV vials reduces both total costs (-$146,094) and DALYs ( −380) in Benin compared to using 10-dose MCV vials without a threshold. Because a majority of locations in Benin (68%) have fewer than five children on average, matching the number of MCV doses-per-vial to session size substantially reduces MCV wastage (−21%). By decreasing the amount of extra MCV otherwise needed to account for wastage, the cold chain storage capacity can handle the increase in MCV volume associated with 5-dose vials. In both Mozambique and Niger, switching to 5-dose vials without a threshold exacerbates existing cold chain storage constraints, resulting in an increase in DALYs of 6178 and 7247, respectively.

Tailoring vaccine vials by session size and maintaining no vial-opening threshold improved total vaccine availability in Benin compared to using 10-dose vials (1.6%) and resulted in fewer DALYs than using 10-dose or 5-dose vials everywhere. In both Mozambique and Niger, tailoring MCV presentations reduced DALYs compared to using 5-dose vials everywhere, but did not improve the system compared to using 10-dose vials everywhere. In both of these supply chains, extensive cold chain constraints were increased when any number of 5-dose vials were introduced without a vial-opening threshold.

By completely removing cold chain constraints, as shown in the hypothetical unconstrained Niger scenario, maintaining no vial-opening threshold results in the highest vaccine availability out of all scenarios (Fig. 1, Fig. 2). In this system, even with high vaccine wastage, there is enough space available in refrigerators and vehicles to maintain stock levels and prevent stockouts from occurring. As such, opening a vial for even one child will result in vaccine availability of 99%. Any vial-opening threshold will increase missed opportunities and reduce vaccine availability.

3.2. Requiring HCWs to wait for 3+ children to open 10-dose MCV and 2+ children to open 5-dose MCV (30% vial-opening threshold)

Introducing a 30% vial-opening threshold policy for 10-dose vials everywhere reduces MCV open vial wastage by 6–14%. A 30% vial-opening threshold policy means that a healthcare worker must wait until three or more children are present before opening a 10-dose MCV vial or two or more children are present for a 5-dose MCV vial. As indicated in Table 1, only Mozambique has a large proportion of vaccinating locations with average MCV session sizes of one to two children (39%), meaning that both Benin and Niger will be able to reduce high MCV open vial wastage without substantially increasing missed opportunities to vaccinate. Even in Mozambique, however, preserving 10-dose vials for three or more children reduces stockouts of MCV for larger sessions and leads to an improvement in both MCV availability (+3%) and total vaccine availability (+0.4%). As shown in Fig. 2 and Table 3, this threshold policy results in the highest total vaccine availability and fewest DALYs in Mozambique.

In Benin and Niger, implementing a 30% vial-opening threshold with 10-dose MCV vials reduces MCV wastage and improves MCV availability (+3%) and total vaccine availability (0.1–0.3%). As both of these countries have few vaccinating locations with less than three children on average per MCV session (Table 1), switching to a 30% threshold reduces MCV wastage while preserving a finite number of MCV vials for larger sessions. Further, as fewer MCV vials are wasted, fewer MCV vials are procured, freeing up constrained transport space for other vaccines (see Table 2).

A moderate (30%) vial-opening threshold policy for 5-dose vials everywhere results in similar effects on the vaccine supply chain as using 5-dose vials with no vial opening threshold policy. Despite the decrease in MCV open vial wastage of 7–16% when switching from 10-dose vials to 5-dose vials under this policy, the larger vaccine volume in the system reduces MCV availability by 3–4% in the more constrained supply chains, Mozambique and Niger. In Benin, however, where constraints are fewest and no vaccinating locations have average session sizes of one or two children, using 5-dose MCV vials with a 30% threshold reduces both total costs (-$145,264) and DALYs ( −381) compared to the baseline policy of no threshold using 10-dose vials. While this isn’t the most cost-effective policy for Benin, it does improve the system.

The most cost-effective policy in Benin involves implementing a 30% vial-opening threshold using a mix of 10-dose and 5-dose MCV vials. In this scenario, 5-dose MCV is distributed to all vaccinating locations with fewer than five children per MCV session on average, and 10-dose MCV to all other locations. As shown in Table 1, the distribution of children per session is clearly divided between three to five children (69%) and six to ten children (31%). As such, tailoring vaccine vials with a 30% threshold prevents 671 DALYs compared to baseline at a cost/DALY averted of $471.10 (Table 3).

In Niger, tailoring vaccines and using a 30% policy improves both total and MCV availability compared to baseline, but does not match the benefits of simply using 10-dose MCV vials everywhere with a 30% vial-opening policy. And in Mozambique, introducing a large number of 5-dose vials into the system with existing constraints results in an increase in cold chain constraints and a decrease in both total and MCV availability compared to baseline.

3.3. Requiring HCWs to wait for 6+ children to open 10-dose MCV and 3+ children to open 5-dose MCV (60% vial-opening threshold)

Introducing a 60% vial-opening threshold policy across each supply chain system results in the largest decrease in open vial wastage compared to the other threshold policies, but overly restricts access to MCV for a number of vaccinating locations, particularly in Benin and Mozambique where a majority of all sessions have fewer than six children on average. In Niger, however, where the median session size is nine and a majority of locations have more than six children on average, setting this restrictive policy improves MCV availability across all MCV presentation scenarios, while improving total vaccine availability when either 10-dose MCV vials or a mix of 10- and 5-dose MCV vials are used.

In Benin, 69% of locations have fewer than six children per MCV session on average. While a 30% vial-opening threshold balances the tradeoffs between reducing wastage and preserving vaccines for larger sessions, the 60% threshold overly restricts MCV usage, particularly when 10-dose MCV vials are used everywhere. While MCV wastage decreases by 23%, MCV availability decreases by 18%, and the number of DALYs for all EPI vaccine-preventable diseases increases by 2888. When 5-dose vials are used with this policy, MCV availability improves (though not as much as 0% or 30% thresholds), but total vaccine availability decreases due to the increase in cold chain constraints. When a mix of MCV presentations are used, allowing 5-dose vials to be opened for three or more children and 10-dose vials to be opened for six or more children, both MCV availability and total availability increase compared to baseline. Yet, this policy does not outperform the 0% and 30% thresholds when using a mix of presentations.

In Mozambique, the 60% threshold is too restrictive when using 10-dose MCV everywhere, resulting in a 5% decrease in total vaccine availability and an increase of 4906 DALYS compared to no vial opening threshold. Additionally, as occurred with the other threshold policies when using 5-dose vials or a mix of presentations in Mozambique, the influx of 5-dose vials overly increased the strain on the system. While the 60% threshold improved MCV availability for 5-dose vials and a mix of vials compared to baseline, total vaccine availability decreased with the additional constraints, resulting in a net-negative outcome for the supply chain system.

In Niger, which has the largest median session size of the three modeled supply chains (nine children), the most cost-effective policy is instituting a 60% vial-opening threshold and tailoring MCV presentations. This policy results in a decrease in total costs of $79,281 and DALYs of 2897 at a cost/DALY averted of $16.05. This policy improves the availability of all vaccines by closely matching the doses per vial to session size and preserving vials for larger sessions. While not the most cost-effective overall, instituting a 60% vial-opening threshold while using 10-dose MCV vials everywhere produces nearly the same decrease in both DALYs (−2705) and total costs (-$66,134) compared to tailoring presentations and using a 60% policy. This policy does not have the additional advantage of distributing 5-dose MCV vials to those locations with five or fewer children on average but does reduce MCV wastage and preserve MCV vials for larger sessions. This not only improves MCV availability but frees up constrained cold chain space for other vaccines, resulting in a benefit to total vaccine availability.