Ethiopia missed opportunities for vaccination study: Cluster-randomized evaluation of 5-dose measles vaccine vials and a flexible open-vial policy, 2021–2022

Abstract

Introduction:

In 2024, an estimated 95,000 people died from measles globally, largely from suboptimal coverage with the measles-containing vaccine (MCV). Health workers may defer vaccinating eligible children to avoid wasting doses from 10-dose MCV vials, which must be discarded six hours after opening. These missed opportunities for vaccination (MOV) reduce coverage and timeliness. Ethiopia, which provides MCV at 9 and 15 months, considered switching to 5-dose vials.

Methods:

We conducted a 15-month, randomized controlled trial with a nested cross-sectional design in Ethiopia. Sixty woredas were randomized to: (1) policy-only, instructing health workers to open 10-dose vials for any number of eligible children, with additional stock for increased wastage; (2) 5-dose switch, combining this policy with replacement of 10-dose by 5-dose vials; or (3) control (routine 10-dose practice). Household and health-facility surveys at baseline and endline assessed effects on first-dose (MCV1) and second-dose (MCV2) coverage, MCV1 timeliness, and wastage. Generalized estimating equation models estimated net intervention effects versus control. We also estimated the government cost of a nationwide 5-dose switch.

Results:

MCV1 coverage was similar between policy-only and control (adjusted risk difference [ARD] = −1%, 95%CI: −17%, 15%), with no significant differences in MCV1 timeliness or MCV2 coverage. In the 5-dose switch group, MCV1 coverage changed little (ARD = 1%, 95%CI: −17%, 18%), but timely MCV1 at 9 months (ARD = 18%, 95%CI: 7%, 28%) and MCV2 coverage (ARD = 17%, 95%CI: 1%, 34%) rose significantly, and wastage fell (ARD = −7%, 95%CI: −14%, −1%). By endline, 29% of 5-dose health workers opened vials ≥10 times monthly (none in control); caregivers in both intervention groups reported 11% fewer measles-related MOVs. A nationwide 5-dose switch was estimated to save 17% in procurement cost per fully vaccinated (two-dose) child.

Conclusions:

The combined 5-dose intervention improved MCV1 timeliness and MCV2 coverage and reduced wastage, addressing a key operational barrier and potentially supporting measles elimination in Ethiopia.

1. Introduction

Measles is a highly infectious disease that caused an estimated 11 million cases and 95,000 deaths in 2024, despite the decades-long availability of a highly effective vaccine [1]. Two doses of measles-containing vaccine (MCV) with national coverage of 95% or more are needed to stop transmission [2], yet global MCV first-dose (MCV1) and second-dose (MCV2) coverage in 2024 was only 84% and 76%, respectively. Where measles transmission is high, the World Health Organization (WHO) recommends timely MCV1 at 9 months of age, when waning maternal antibodies provide substantially less protection against infection and infants are therefore vulnerable, and a second dose (MCV2) in the second year of life to improve protection [2].

In Ethiopia, where measles vaccination coverage has plateaued in recent years, outbreak investigations exploring reasons for non-vaccination found that health workers are hesitant to open MCV vials when few children are present [3]. In Ethiopia and most low- and middle-income countries, MCV is commonly supplied in 10-dose vials that, once reconstituted, must be discarded after 6 h; most other vaccines contain preservatives allowing re-use for several days. Under these conditions, health workers may turn children away and tell them to return on designated vaccination days, resulting in a missed opportunity for vaccination (MOV) [4–6]. MOVs may be more common in rural locations with fewer eligible children than in busier urban settings [6].

Health workers in Nigeria and Cambodia have reported turning children away for MCV when few are present because high wastage is viewed negatively for staff performance [4,5]; any dose discarded from an opened vial is counted as wastage. By contrast, WHO recommends vaccinating every eligible child at each opportunity, since a child may never return. However, many national policies do not clearly permit health workers to open vials regardless of the number of children present [7]. One option to address these concerns is a switch to smaller multi-dose presentations [8]. A 5-dose MCV vial is available, but its use has been limited by concerns about higher per-dose cost and supply-chain requirements, leaving little rigorous evidence on its impact relative to 10-dose vials [8].

This study evaluated two interventions in Ethiopia across three outcomes: MCV coverage, vaccination timeliness, and vaccine wastage. The first was a flexible open-vial policy allowing health workers to open 10-dose vials for any number of children at any time; the second combined that policy with a switch from 10-dose to 5-dose vials. We hypothesized that both interventions would increase coverage and timeliness, and that the 5-dose intervention would additionally reduce wastage and have greater overall impact. We also assessed health worker- and caregiver-reported MOV and related provider practices, intervention costs, and health worker perceptions of the 5-dose switch.

2. Methods

2.1. Study design and interventions

We conducted a cluster-randomized controlled trial in 60 predominantly rural woredas (districts) across five zones in the Oromia and Amhara regions of Ethiopia (Supplemental Table 1), evaluating two intervention packages designed to reduce MOV by encouraging health workers to vaccinate all eligible children at each encounter. We hypothesized that both interventions would improve MCV1 and MCV2 coverage and MCV1 timeliness. Because the interventions required changes to vaccine supply—which is managed at the woreda level—randomization occurred at the woreda level (study cluster). Using the PROC PLAN statement in SAS 9.4, the 60 woredas were randomly assigned (20 each) to three groups:

• Policy-only: Health workers were trained on a flexible open-vial policy allowing them to open a 10-dose MCV vial for any number of children, regardless of day or time. Additional 10-dose vials were supplied to prevent stockouts.

• 5-dose switch: Health workers received the same training, and all 10-dose vials were replaced with 5-dose vials before the intervention began.

• Control: Routine measles vaccination continued using 10-dose vials per existing standard of care.

To measure outcomes, cross-sectional household and facility-based surveys were conducted in each group before (pre-intervention) and at the end (post-intervention) of the intervention period. The intervention period was 15 months, beginning January 2021, and was preceded by a three-month start-up period (October–December 2020) during which intervention-group health workers received orientation to the new policy and all 10-dose vials were replaced with 5-dose vials in the switch group. Throughout the intervention period, two independent study-team monitors remained in the study areas to provide ongoing oversight of measles vaccine supply and to reinforce messaging to health workers and supervisors on implementing the new vial-opening policy in intervention woredas during the first two months of the intervention phase.

2.2. Study population

The intervention targeted children 9–23 months of age (and their caregivers) eligible for routine measles vaccination in rural study catchment areas (Supplemental Table 1); in Ethiopia, MCV1 is scheduled at 9 months and MCV2 at 15 months. Surveys enrolled children aged 12–23 months. Rural settings were selected based on prior evidence that health worker reluctance to open a measles vial is more common in areas with fewer eligible children. Zonal-level data (zones are one administrative level above woredas) were reviewed to identify zones across the Oromia and Amhara regions with similar characteristics in terms of ruralness, population density, and measles incidence (Supplemental Table 1). The average annual woreda-level population of children 9–23 months was 3036 based on government administrative estimates (about 60,720 per study group). Using the standard Ethiopian vaccine forecasting calculation—estimated population size, 100% two-dose coverage, a 25% vial-stock buffer, the intervention duration, and a 55% estimated wastage rate—85,000 5-dose vials were procured for the switch-group health facilities.

2.3. Outcomes

Primary outcomes were MCV1 and MCV2 coverage among children aged 12–23 months. Secondary outcomes were: (1) MCV1 timeliness, defined as the proportion of children receiving MCV1 at the recommended age and measured through two indicators—the proportion vaccinated by a given age in months and the age (in days) at vaccination; (2) facility-level MCV wastage rates, from monthly data on doses administered and opened; (3) caregiver attitudes toward vaccination services, from a previously validated vaccine acceptance scale used to create a composite score in which a lower score indicated less hesitancy and more positive perceptions; (4) caregiver-reported MCV-specific MOV, from household interview responses about being turned away for measles vaccination; and (5) health worker-reported vaccination practices, including frequency of providing measles vaccination, reasons for turning children away, and knowledge of the open-vial policy.

2.4. Sampling and sample size

A three-stage cluster sampling design was used. At stage 1, the 60 woredas were randomized to the three study groups. At stage 2, within each woreda, a newly created enumeration area (EA) sampling frame was used to randomly sample EAs (small geographic units with a median of 106 households) with probability proportional to population size. At stage 3, within each EA, nine households were randomly sampled, with one child aged 12–23 months included per household. Because census-derived EA frames were deemed out of date with concerns about accuracy, we generated EAs from gridded WorldPop population data using the GridSample R package [9,10], setting a minimum EA population of 350 individuals (about 70 rural Ethiopian households) and a minimum area of 3 km2; the resulting frame had a median EA of 532 individuals (106 households). EAs were sampled by probability proportional to size for the baseline surveys and re-sampled by the same process at endline. EA maps were produced in QGIS for team navigation [11], and, as no recent household-level frames existed, survey teams first listed all households with children aged 12–23 months in each EA (for both the baseline and endline surveys) before team leads randomly selected nine.

Sample size and power were estimated using a difference-in-differences simulation programmed in SAS 9.4, which used a binomial distribution across four independent surveys (two per compared group) and was run 100 times across a range of inputs, varying power, children sampled per cluster, effect size, and clusters per group. Initial assumptions were baseline MCV1 coverage of 73% (from 2017 regional data), an intraclass correlation coefficient of 0.10, alpha of 0.05, and 80% power to detect a 19% or greater absolute increase in MCV1—a deliberately high threshold reflecting the programmatic intent to switch to 5-dose vials only if a strong effect were seen. We initially selected a scenario of 9 children per woreda across 20 woredas per group (240 per survey per group), holding other assumptions constant; additional funds available near the baseline survey allowed an increase to 24 children per woreda (480 per survey per group), improving precision.

2.5. Data collection

Data sources were the cross-sectional household surveys, health facility immunization records, woreda health-office information-system records, health facility vaccine-supply records, and health worker surveys (Supplemental Table 2). Household surveys captured childhood vaccination status and caregiver MOV experiences; facility immunization records supplemented the survey-based vaccination status; woreda and facility vaccine stock logbooks (indicating vaccine doses received, doses administered and doses used on monthly basis) were sourced for the wastage calculations; and health worker surveys captured reported vaccination frequency and feedback on the interventions. A locally contracted firm experienced in survey-based data collection led field-work and provided a clean, de-identified dataset for analysis. Baseline surveys occurred in December 2019 and endline in May–June 2022, using the same design and questionnaire. At each selected household, teams interviewed caregivers, reviewed vaccination cards, and collected demographic and vaccination data; for each child, teams also abstracted vaccination records from the child’s health facility. At those facilities, teams abstracted six months of doses administered and opened to calculate wastage rates and interviewed all health workers who routinely provided vaccinations about practices, policy knowledge, and reasons for turning children away.

2.6. Insecurity adaptations

Late in the intervention period, insecurity in two of five zones prevented endline access to 24 woredas, distributed equally across groups (8 per group); baseline data from these woredas were excluded. To offset this loss, the endline household survey was expanded to 3–5 EAs per woreda (by probability proportional to size) with 9 children per EA.

2.7. Statistical analysis

Descriptive statistics summarized household- and child-level demographics and vaccination characteristics using SAS 9.4 survey procedures. Categorical variables were reported as unweighted counts and survey-weighted percentages, and continuous variables as survey-weighted means with 95% confidence intervals. In the weighting, the EA was the cluster and the woreda the stratum; survey weights equaled the inverse of each child’s selection probability, calculated as the population-based probability of an EA being selected within its woreda multiplied by the probability of the child being selected among eligible children in that EA.

The primary measure of intervention effect was an adjusted risk difference (ARD): the net change in pre- to post-intervention coverage in an intervention group relative to control [12], calculated as: ARD = (intervention endline coverage − intervention baseline coverage) − (control endline coverage − control baseline coverage).

Effects were estimated using generalized estimating equation (GEE) models via PROC GENMOD in SAS 9.4 to account for the group-randomized design and cluster surveys [13]. Each model included a dependent variable for the child-level outcome (e.g., measles vaccination status), a fixed binary variable for each intervention group, a fixed binary variable for survey time (baseline or endline), a survey-time-by-group interaction term providing the effect estimate for each intervention group relative to control, and a repeated-subject specification of EAs nested within woredas with an exchangeable correlation structure to account for correlated data within clusters. Coverage and timeliness models used a binomial distribution with an identity link to directly estimate risk differences and adjusted a priori for covariates known to be associated with vaccination status: mother’s education, mother’s religion, mother’s antenatal care status, and the child’s place of birth (home or facility), birth order, and sex.

Timeliness was assessed with the same model structure using two indicators: receipt of MCV1 by successive age thresholds (binary receipt by 9 months [270 days], 10 months [300 days], and so on) and age in days at MCV1. For caregiver attitudes, the validated vaccine acceptance score (0–100, where 0 is least hesitant and 100 most hesitant [14]) was rescaled to 0–1 and modeled with a binomial distribution and logit link; scores were also categorized into quartiles (0–25, 26–50, 51–75, 76–100) for description. Caregiver-reported MOV for measles and pentavalent vaccines were analyzed descriptively using the survey-weighted approach above. Health worker-reported vaccination frequency (categorized as 1–2, 3–9, or 10 or more times/month) was compared at endline across groups using survey-weighted logistic regression; pentavalent vaccine—usable up to 30 days once opened and given at 6, 10, and 14 weeks—served as a comparator with no equivalent open-vial restriction. MCV wastage rates were calculated as the number of opened doses not administered (doses opened minus children vaccinated in a month) divided by doses opened, averaged monthly across facilities in each group, and modeled with a GEE model (proportional outcome 0–100, exchangeable correlation) without individual-level covariates.

2.8. Costing

We built a scenario model estimating the financial implications of a nationwide switch to 5-dose vials from the health-sector perspective, combining trial-derived inputs (MCV1 effect, wastage effect, MCV2 effect) with programmatic assumptions for national extrapolation (Table 1). Key assumptions included procurement sufficient to reach 97% of the annual target population, 2021 UNICEF Supply Division vial costs, and 2021 WHO/UNICEF national coverage estimates. The model estimated total cost, cost per dose administered, and cost per fully vaccinated child, with one-way sensitivity analyses varying wastage and MCV2 coverage across their CI bounds. Values are in 2025 US dollars.

Table 1.

Key inputs and results for a hypothetical nationwide switch from 10-dose to 5-dose measles-containing vaccine vials in Ethiopia using results from the Measles 5-dose switch study, 2022.

Inputs	New option: Switch to 5-dose MCV vials	Current option: Continued use of 10-dose MCV vials
Target population (surviving infants, 2021) [15]	3.25 million	3.25 million
Target MCV1/MCV2 coverage used to determine the number of annual doses to procure for country [16]	97%	97%
Measles vaccine wastage rate1	14%	21%
Cost per dose (Source: UNICEF supply division catalogue) [17]	US$0.48	US$0.39
Measles 1st dose coverage (WHO/UNICEF national estimate combined with intervention effect) [18]	61%	60%
Measles 2nd dose coverage (WHO/UNICEF national estimate combined with intervention effect) [18]	58%	41%

¹Input sourced from study results.

2.9. Ethical considerations

Ethical approval was granted by the Ethiopian Public Health Institute (SERO-044-12-2016). CDC reviewed the project (ID#2018–317) as a Category IIID activity not constituting engagement in human subjects research. Regional health directorates in Oromia and Amhara granted permission, and minimal risk was anticipated as participants were asked personal questions relating to their health status. The 5-dose vaccine was approved by the Ethiopian Food and Drug Administration based on the existing 10-dose approval, as it used the same formulation at half the volume. All participants provided informed consent, and data were de-identified before analysis.

3. Results

3.1. Descriptive statistics

3.1.1. Household sample characteristics

At baseline, 692 households of children aged 12–23 months were surveyed across the three groups; at endline, 1463. The measles vaccination status of most children (86%) was obtained from a vaccination card or health facility record, with the remainder based on parental recall. Demographic characteristics of caregivers and children were largely similar across groups (Table 2). At baseline, vaccination status was similar across groups for all vaccines, although the policy-only group had consistently higher coverage estimates across most vaccines. Most caregivers (over 80%) had very low hesitancy scores at both baseline and endline, indicating a high proportion with positive perceptions of vaccination.

Table 2.

Sociodemographic and vaccination characteristics of children aged 12–23 months across three study groups, Measles 5-dose switch study, Ethiopia, 2022, based on household surveys.

Characteristic	Policy-only Intervention Group n, (%)		5-dose Switch Intervention Group n (%)		Control Group n (%)
	Baseline (n = 232)	Endline (n = 506)	Baseline (n = 227)	Endline (n = 488)	Baseline (n = 233)	Endline (n = 469)
N study woredas5	12	12	12	12	12	12
N survey enumeration areas	30	44	30	44	30	44
average number of children sampled in survey enumeration area	8	11	8	11	8	11
Mother and household
Age of mother (mean, 95% CI)	29 (28,31)	28 (26, 28)	28 (25,29)	29 (28,30)	27 (25,30)	27 (26,28)
Mother’s highest education attended
No formal education	109 (51)	183 (41)	104 (52)	242 (54)	123 (64)	214 (51)
Primary	77 (30)	214 (38)	85 (35)	163 (30)	85 (30)	169 (31)
Secondary or higher	46 (19)	109 (21)	40 (13)	82 (16)	24 (6)	86 (18)
Family religion
Muslim	73 (25)	216 (31)	38 (12)	109 (19)	93 (31)	192 (24)
Orthodox tewahido	130 (67)	208 (57)	123 (66)	253 (63)	109 (61)	211 (65)
Protestant	23 (8)	82 (12)	66 (22)	126 (18)	31 (8)	66 (10)
Infant
Birth order
1st	63 (24)	159 (32)	75 (26)	137 (27)	64 (22)	125 (28)
2nd	45 (21)	136 (25)	49 (21)	109 (21)	49 (17)	104 (22)
3rd	34 (13)	79 (16)	31 (13)	86 (16)	35 (16)	87 (14)
4th	33 (15)	44 (9)	28 (14)	52 (11)	31 (14)	59 (14)
5th or more	57 (27)	88 (17)	44 (25)	104 (26)	54 (30)	94 (23)
Birth location
Home	83 (37)	124 (26)	72 (36)	162 (35)	101 (52)	149 (33)
Health facility	149 (63)	382 (75)	155 (64)	326 (65)	132 (48)	320 (67)
Vaccination source availability
Health card available	129 (56)	361 (78)	89 (42)	343 (74)	69 (39)	258 (63)
Facility record available1	186 (81)	139 (21)1	149 (69)	95 (18)1	149 (64)	152 (27)1
Health card or facility record available2	197 (85)	500 (99)	164 (76)	438 (92)	159 (68)	410 (90)
Vaccination coverage
Pentavalent 1st dose3	210 (92)	503 (99)	209 (90)	461 (95)	202 (87)	440 (95)
Pentavalent 3rd dose3	201 (88)	477 (95)	181 (79)	402 (85)	182 (77)	380 (85)
Parent vaccine hesitance score
Least hesitant group4	206 (89)	497 (98)	187 (82)	457 (94)	192 (86)	452 (97)
Most hesitant group4	0 (0)	0 (0)	0 (0)	0 (1)	0 (0)	0 (0)
Scale score (mean, 95% CI)4	11 (7, 15)	9 (7,10)	15 (11,18)	10 (8,12)	12 (8,16)	8 (6,9)

Note: unweighted n and survey sample-weighted proportions are shown. CI = confidence interval.

¹In the endline survey only, facility records were reviewed only for children who did not have cards available. In the baseline survey, facility records were reviewed for all surveyed children, regardless of card availability at the household visit.

²Total number/proportion of children who had either a health card or a facility record available. Some children had both cards and facility records.

³Pentavalent vaccine protects against diphtheria, tetanus, pertussis, hepatitis B and Haemophilus influenzae type b and is a 3-dose series recommended at 6, 10 and 14 weeks of age in Ethiopia.

⁴Vaccine hesitance scale score range between 0 and 100 with 0 being the least hesitant and 100 being the most hesitant. Scale score was divided into quartiles from least hesitant to most hesitant.

⁵Twenty woredas per study group were originally randomized and participated in the intervention; due to insecurity issues that affected endline survey accessibility in 24 woredas (8 per study group), only 12 per study group were retained in the final analysis.

3.1.2. Health facility sample characteristics

Baseline and endline facility surveys included 135 and 126 health facilities, respectively; health workers were interviewed at all of them, and 81 (64%) had complete wastage records (sourced from facility-kept vaccine supply & usage ledgers) at endline. Among surveyed facilities, 61% were health posts—the lowest level of service delivery, typically serving a small rural population and focused on preventive and maternal care—and 39% were health centers, which serve larger rural and peri-urban populations and provide more complex preventive care and treatment. The survey-weighted median annual target population was 109 (95% CI: 91, 128) for health posts and 864 (95% CI: 657, 1071) for health centers. All surveyed health posts and health centers had functional vaccine cold chain equipment for storing on-site vaccine.

3.2. Impact on measles vaccination coverage

No significant difference in MCV1 coverage was observed between either intervention group and control (Table 3): the adjusted risk difference (ARD) was −1% (95% CI: −17%, 15%) for the policy-only group and 1% (95% CI: −17%, 18%) for the 5-dose switch group. MCV2 coverage among children aged 18–23 months was significantly higher in the 5-dose switch group than control (ARD 17%; 95% CI: 1%, 34%), whereas the policy-only group showed a smaller, nonsignificant increase (ARD 4%; 95% CI: −16%, 24%) (Table 3).

Table 3.

Measles vaccination outcomes in Ethiopia Measles 5-dose switch study based on household surveys, weighted results.

Outcome	Policy-only Intervention Group % (95% CI)1			5-dose Switch Intervention Group % (95% CI)1			Control Group % (95% CI)1
Period	Baseline (n = 232)	Endline (n = 506)	Adjusted risk difference (95% CI)3,4	Baseline (n = 227)	Endline (n = 488)	Adjusted risk difference (95% CI)3,4	Baseline (n = 233)	Endline (n = 469)
Measles first dose	87 (81, 93)	82 (77, 88)	−0.01 (−0.17, 0.15)	78 (69, 87)	74 (67, 82)	0.01 (−0.17, 0.18)	81 (69, 93)	77 (70, 85)
Measles second dose	17 (10, 24)	46 (34, 57)	0.04 (−0.16, 0.24)	9 (4, 14)	51 (42, 59)	0.17 (0.01, 0.34)	21 (12,31)	46 (34, 58)
Measles first dose by 9 mo. of age	19 (12, 26)	23 (18, 28)	0.09 (−0.03, 0.21)	14 (8, 20)	26 (22,30)	0.18 (0.07, 0.28)	23 (17, 30)	18 (12, 24)
Measles first dose by 10 mo. of age	55 (47, 64)	41 (28, 53)	−0.03 (−0.22, 0.17)	49 (36, 61)	53 (44, 61)	0.10 (−0.12, 0.31)	50 (43, 57)	49 (38, 59)
Measles first dose by 11 mo. of age	69 (62, 77)	50 (36, 63)	−0.01 (−0.21, 0.18)	64 (51, 78)	63 (55, 70)	0.13 (−0.09, 0.36)	58 (50, 66)	59 (49, 70)
Measles first dose by 12 mo. of age	75 (68, 82)	52 (38, 66)	−0.02 (−0.21, 0.17)	71 (57, 85)	69 (62, 76)	0.17 (−0.05, 0.40)	65 (57, 73)	67 (57, 76)
Age at measles first dose (days) median (IQR)2	308 (297, 319)	308 (298, 319)	−2.4 (−24, 19)	328 (302, 354)	301 (294, 308)	−31 (−63, −1)	300 (290, 311)	304 (294, 313)

¹Survey sample weighted proportions and 95% confidence intervals.

²Survey sample weighed median and interquartile range (IQR).

³Risk difference compares the difference in proportion vaccinated in each intervention group compared with the proportion vaccinated in the control group at endline compared to this same difference between intervention and control groups at baseline.

⁴Adjusted for mother’s education, mother’s religion, mother’s antenatal care status, and the child’s place of birth (home or facility), birth order, and sex.

3.3. Impact on measles vaccination timeliness

Sufficient information on date of birth and date of measles vaccination was available for 597 (86%) baseline and 1360 (93%) endline interviews; missingness was due to illegible or absent dates (e.g., a checkmark rather than a date) in child records. Children in the 5-dose switch group were significantly more likely to receive MCV1 earlier than those in control: the ARD for receipt by 9 months of age was 18% higher (95% CI: 7%, 28%) (Table 3). Risk differences for receipt by 10, 11, or 12 months were 10–17% higher in the switch group but did not reach statistical significance. The median age at MCV1 in the switch group decreased significantly by 31 days (95% CI: −63, −1) relative to control (Table 3). The policy-only group showed minimal change in timeliness: the median age at MCV1 remained 308 days at baseline and endline, a 4-day net difference relative to control (95% CI: −24, 19).

3.4. Impact on MCV wastage rates

MCV wastage decreased by 7% in the 5-dose switch group compared with no change in control, a statistically significant net effect of −7% (95% CI: −14%, −1%) (Table 4). In the policy-only group, wastage rose by 5% from baseline to endline, a nonsignificant net effect of 5% (95% CI: −2%, 12%). Health centers consistently had lower wastage than health posts; in the switch group, endline wastage was lower in both facility types than at baseline compared to control, with a significant and more pronounced decrease at health posts.

Table 4.

Measles vaccine wastage rates and facility type details in Ethiopia Measles 5-dose switch study.

Outcome	Policy-only Intervention Group % (95% CI)			5-dose Switch intervention Group % (95% CI)			Control Group % (95% CI)
	Baseline (n = 50)	Endline (n = 42)	Risk difference (95% CI)2	Baseline (n = 47)	Endline (n = 42)	Risk difference (95% CI)2	Baseline (n = 38)	Endline (n = 42)
Measles vaccine wastage rate at health facility level1	24 (14, 34)	29 (21, 36)	5 (−2,12)	25 (18, 33)	18 (11, 24)	−7 (−14, −1)	22 (15, 28)	21 (13, 29)
Measles vaccine wastage rate in health centers (large facilities)	19 (12, 26)	28 (18, 38)	10 (−2,21)	17 (10, 24)	11 (4, 18)	−4 (−15, 7)	16 (9, 23)	19 (11, 27)
Measles vaccine wastage rate in health posts (small facilities)	36 (24, 38)	32 (20, 45)	3 (−6, 4)	35 (25,44)	21 (16, 27)	−10 (−18, −1)	26 (15, 37)	27 (17, 38)

¹Health facilities are defined as both types of health facilities surveyed in study e.g. health posts and health centers. Health posts serve small rural communities with an average catchment population size of 109 whereas health centers serve multiple peri urban communities, with an average catchment population size of 864.

²Risk difference compares the difference in proportion vaccinated in each intervention group compared with the proportion vaccinated in the control group at endline compared to this same difference between intervention and control groups at baseline.

3.5. Impact on caregiver attitudes about vaccination

Across surveys, 97% (95% CI: 95%, 98%) of caregivers scored in the lowest quartile (0–25) of the hesitancy scale, indicating consistently low hesitancy. Baseline mean scores ranged from 11% to 15% across groups, and all three groups showed minor decreases (2–5%) from baseline to endline (Table 2). No significant differences were observed between either intervention group and control (policy-only ARD 6 points, 95% CI: −4, 15; 5-dose switch ARD 1 point, 95% CI: −11, 13).

3.6. Health worker-reported frequency of measles vaccination

At endline, no control-group health workers reported providing MCV 10 or more times/month (0%; 95% CI: 0%, 0%), and most (85%; 95% CI: 76%, 94%) reported providing it 1–2 times/month (Fig. 1, Supplemental Table 3). By contrast, 29% (95% CI: 17%, 40%) of switch-group health workers reported providing MCV 10 or more times/month and 50% (95% CI: 37%, 63%) reported 1–2 times/month; policy-only proportions were similar to the switch group. For pentavalent vaccine—which lacks MCV’s open-vial restrictions and can be used up to 30 days once opened—the proportion providing it 10 or more times/month was similar across groups (33%, 26%, and 29% for policy-only, switch, and control). At endline, most control-group health workers (84%; 95% CI: 73%, 95%) reported turning children away for measles vaccination on non-designated days or when too few children were present, compared with only 35% (95% CI: 22%, 48%) and 44% (95% CI: 35%, 53%) in the policy-only and switch groups. Among intervention health workers who reported turning children away, 98% (95% CI: 92%, 100%) were unfamiliar with the vial-opening policy, having started after the initial orientation; excluding them, under 1% of remaining intervention health workers reported turning children away.

Fig. 1.

Proportion of health workers reporting how often they conduct measles and pentavalent vaccination sessions per month, by study group at endline, Measles 5-dose switch study, Ethiopia, 2022. Notes: Pentavalent vaccine: protects against five diseases including Diphtheria, Pertussis, Tetanus, Hepatitis B and Haemophilus influenzae type b.

3.7. Caregiver-reported missed opportunities for measles vaccination

Across all groups, caregiver-reported MOV was 2–3 times higher for MCV than for pentavalent vaccine. At baseline, roughly 10–12% of caregivers reported a pentavalent MOV versus 18–26% for MCV. By endline, MCV-specific MOV was lower in both intervention groups than control, with a net difference of 11% lower in the policy-only group (95% CI: 0%, 20%) and 11% lower in the switch group (95% CI: 0%, 30%).

3.8. Financial costing

Fewer total doses would need to be procured under the 5-dose option due to lower wastage, despite more doses administered from higher MCV2 coverage, compared with no switch (Table 5). Total procurement cost per two-dose fully vaccinated child fell by USD $0.38 (17%) under the switch, reflecting both improved MCV2 coverage and reduced wastage (Table 5). One-way sensitivity analyses showed the cost advantage of the 5-dose switch was robust to a higher assumed measles vaccine wastage rate (Supplemental Table 5) but was sensitive to lower MCV2 coverage: at an MCV2 coverage of 48%, the cost per two-dose fully vaccinated child under the switch ($2.19) equaled that of continued 10-dose use ($2.20) (Supplemental Table 4).

Table 5.

Key financial costing results for a hypothetical scenario involving a nationwide switch from 10-dose to 5-dose measles-containing vaccine in Ethiopia, Measles 5-dose switch study, 2022.

Outcome	New option: Switch to 5-dose MCV	Current option: Continued use of 10-dose MCV	Difference between options: 5-dose use versus 10-dose use
Total doses procured for target 97% 2-dose coverage (inclusive of wastage rate)	7.19 million	7.63 million	−0.44 million
Total vials procured for target 97% 2-dose coverage (inclusive of wastage rate)	1.44 million	0.76 million	0.68 million
Total procurement cost for vials purchased to reach target 97% 2-dose coverage	US$3.4 million	US$2.9 million	US$0.50 million
Total MCV two-dose vaccinated children	1.89 million	1.33 million	0.56 million
Total MCV doses administered	3.85 million	3.28 million	0.57 million
Total procurement cost / per two-dose vaccinated child	US$1.82	US$2.20	−US$0.38

4. Discussion

This study provides evidence that, in this setting, supporting the use of a 5-dose MCV vial combined with a flexible open-vial policy improves measles vaccination performance. The intervention had significant positive effects on MCV2 coverage and MCV1 timeliness, key indicators for effective measles control. In both intervention groups, a greater proportion of health workers reported offering MCV more frequently, and caregiver-reported measles-related MOV declined relative to the standard of care. The significant reduction in wastage achieved with the 5-dose intervention is also expected to improve vaccine utilization. Under our modeled nationwide switch to 5-dose vials, these efficiency gains may translate into lower procurement costs per fully vaccinated child—savings driven primarily by reduced wastage and higher MCV2 coverage, which together lower the doses required per fully vaccinated child.

While the policy-only intervention showed no significant effect on coverage or timeliness, it produced slightly higher wastage than control, likely because continuing to use 10-dose vials under a permissive opening policy increases discarded doses. This suggests the importance of pairing flexible open-vial policies with smaller vial presentations to mitigate wastage and maximize impact on coverage and timeliness. Notably, health worker-reported practices and caregiver-reported experiences with being turned away both improved in the policy-only group, even though these changes did not immediately translate into gains in MCV1 coverage or timeliness.

The second intervention—combining orientation to the vial-opening policy with the introduction of 5-dose vials—had little effect on overall MCV1 coverage but significantly improved MCV2 coverage and the proportion receiving MCV1 on time. Earlier MCV1 vaccination is consistent with children being turned away less frequently, and the reported provider practices and caregiver experiences both aligned with this likely change. Wastage improved particularly at health posts, which serve very small target populations and thus see few children per session. Of note, a high proportion of health workers (>50%) in the 5-dose switch group were still providing MCV 1–2 times/month (the common practice in Ethiopia) at the endline survey, indicating the potential difficulty in changing health worker behavior to open vaccines and hold vaccination sessions more frequently, even with more permissive open-vial policies and a smaller dose MCV vial. High health worker turnover may have limited the full impact of both interventions: a substantial proportion of intervention health workers reported being unfamiliar with the open-vial policy because they started after the initial training. This underscores the need for ongoing training, supportive supervision, and strong communication to ensure consistent implementation and maximize effectiveness. Ensuring that any nationwide rollout of the use of 5-dose MCV vials incorporates these aspects of ongoing reinforcement of the vial opening policy with health workers appears critical for maximizing the benefits of 5-dose MCV vial use.

Our findings that a switch to the 5-dose MCV vial can potentially reduce costs through reduced wastage rates and improvements in MCV2 coverage has important implications for decision-making, especially for a country such as Ethiopia which can often face serious health sector resource constraints. Whether countries are choosing to introduce a new vaccine or switching vaccine presentations, understanding the economic impacts are critical to the decision-making process. Historically, the decision about using the 5-dose MCV vial versus a 10-dose MCV vial has often centered on whether it will be more expensive to use the 5-dose MCV vial even if it can potentially have a positive impact on coverage and/or timely vaccination, so our results showing the financial implications for a country like Ethiopia if scaled to nationwide use is an important datapoint for other countries considering a switch to 5-dose MCV vials. Additionally, our study provides important data to help with more accurately estimating annual vaccine procurement needs, particularly for Ethiopia as it begins a switch to 5-dose MCV vial use.

Our findings are consistent with a study of 5-dose MCV vials in Zambia, with some key differences [19]. The Zambia study reported significant increases in both MCV1 and MCV2 coverage and lower wastage in the switch group, whereas our study found significant effects only for MCV2 coverage and wastage. Our study also examined timeliness, providing further insight into the broader impacts of a switch from 10-dose to 5-dose vials. Variations in country context—such as baseline coverage and standard immunization practices—alongside methodological differences such as sample size may have contributed to the divergent findings.

In the 5-dose group, the improvement in MCV1 timeliness and MCV2 coverage but not MCV1 coverage initially appears inconsistent. We hypothesize that after the introduction of 5-dose vials, health workers in the switch group turned children away much less frequently than those in control, allowing more children to receive MCV1 on their first attempt, whereas a higher proportion of control-group children were initially turned away. However, because measles is a very well-known and visible disease in Ethiopia, control-group parents may have been more likely to return repeatedly; many initially turned-away children were eventually vaccinated, but at older ages and after multiple attempts. This pattern would produce similar MCV1 coverage across groups while still yielding earlier vaccination in the switch group. Higher MCV2 coverage in the switch group likely followed from earlier MCV1 receipt and less frequent turning away. Additionally, we surveyed children only up to 23 months of age—then the upper age limit for MCV2 under national policy—so a higher age limit might have shown MCV2 coverage converging between groups beyond 23 months.

Several limitations exist. First, insecurity affecting access to baseline woredas reduced our sample size and the precision of our estimates, yielding point estimates with wide confidence intervals. Second, MCV1 coverage decreased across all groups from baseline to endline, likely because baseline surveys preceded the COVID-19 pandemic and endline surveys followed it, given the pandemic’s major negative impact on routine vaccination worldwide. Third, the pandemic delayed the intervention start for several months after the December 2019 baseline; even with randomization, pandemic effects may have differed across groups during this gap. Fourth, higher-than-expected health worker turnover, including many staff not present during initial training, likely attenuated intervention effects. Fifth, in the switch group, independent monitors tracked vaccine supply needs, and their presence may have enhanced the intervention’s effects.

Despite these limitations, this study provides important evidence supporting a policy shift toward adoption of the 5-dose MCV presentation. The approach can help address an often-cited barrier to timely measles vaccination—health worker hesitancy to open MCV vials—thereby increasing opportunities for earlier vaccination and earlier protection. The observed reductions in wastage suggest more efficient vaccine use, with modeled estimates indicating potential cost savings under a nationwide implementation scenario.

5. Conclusions

This study demonstrates the significant positive impact of switching to a 5-dose MCV vial combined with a flexible open-vial policy on measles vaccination coverage, timeliness, and wastage in Ethiopia. It adds to the growing evidence supporting smaller vial presentations to overcome barriers to vaccination and improve program effectiveness. Adoption of the 5-dose MCV vial has the potential to meaningfully advance measles elimination goals in Ethiopia and other settings facing similar challenges. Based on these results, the Ethiopian Government decided to adopt the 5-dose measles vial and began implementing the switch in 2026.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.vaccine.2026.129065.

Data availability

The data that has been used is confidential.