Study design and model
We used version 1.7.01 of the UNIVAC (Universal Vaccine) decision-support model. UNIVAC is a static cohort model with a finely disaggregated age structure (weeks of age < 5 years) and can be used to assess the impact, cost-effectiveness, and benefit-risk of a range of different vaccines. It features a standardised and user-friendly Excel-based interface with a standard set of input steps and outputs. The model has been widely used by low- and middle-income countries (LMICs) to support decision-making for new vaccine introductions, including RV vaccines.[18]
We ran two types of analysis:
-
cost-effectiveness and benefit-risk analysis of ROTARIX (1-dose vial) over a seven-year period (2018–2024) compared to no vaccination; and,
-
cost-effectiveness and benefit-risk analysis of ROTARIX (1-dose vial), ROTASIIL (1-dose vial), ROTASIIL (2-dose vial), and ROTAVAC (5-dose vial) over a ten-year period (2025–2034). ROTATEQ was excluded from the comparison as it is not covered under Gavi’s financial support. Each product was compared to no rotavirus vaccination and each other.
In both analyses, we estimated the numbers of RVGE cases, outpatient visits, admissions, and deaths with and without RV vaccination. We also estimated potential excess intussusception cases and deaths using previously described methods.[19] Vaccine programme costs and healthcare costs were estimated throughout the first five years of life, and disability-adjusted life-years (DALYs) were calculated over the lifetimes of the target birth cohorts. The primary outcome measure was the discounted cost per DALY averted (from government and societal perspectives) compared to no vaccination. The benefit-risk of vaccination was represented by the estimated number of RVGE deaths prevented per one excess intussusception death.
Since 2018, the estimated annual gross domestic product (GDP) per capita in Afghanistan has rarely exceeded US$ 500, so we compared our estimates of the cost per DALY averted to a range of potential cost-effectiveness thresholds (CETs) between US$ 0 and US$ 500.[20, 21] In compliance with WHO vaccine economic evaluation guidelines, we considered both government and societal perspectives, and all future costs and health benefits were presented at a discounted rate of 3% per year, and expressed in 2022 United States Dollars (US$).[22]
Data collection and consensus building
Demographic projections were pre-populated in the model from the United Nations Population (UNPOP-2022), providing population size by age/year, life expectancy at birth by age/year, and < 5 all-cause mortality by year. All other model inputs and their sources are shown in Tables 1 and 2.
Table 1
Input parameters for estimating the burden of rotavirus gastroenteritis (RVGE) in Afghanistan- 2018–2027
Parameter | Mid-point value | Low bound | High bound | Sources |
Annual rate per 100,000 < 5 children | | | |
RVGE non-severe cases | 8224 | 6990 | 9458 | [24] |
RVGE non-severe RVGE visits | 4367 | 3712 | 5022 | [24, 25] *53.1% diarrhoea disease health seeking < 5yrs |
RVGE severe cases | 1776 | 1510 | 2042 | [23, 4] |
RVGE severe visits | 943 | 802 | 1085 | [23, 25] * 53.1% diarrhoea disease health seeking |
RVGE hospital admissions | 444 | 377 | 511 | [26] inflated by DTP 1 coverage in 2022 (73% as proxy of getting healthcare) |
RVGE deaths | 26 | 22 | 30 | [27] |
Intussusception cases | 30 | 26 | 35 | [4,13,40{Anwari, 2024 #879}] inflated to account for those without access, using 2022 DTP 1 coverage (77%) |
Intussusception hospital admissions | 23 | 20 | 26 | [13, 32] |
Intussusception deaths | 7 | 6 | 8 | [13, 32] |
Relative risk of intussusception | | | | |
After 1–7 days | | | | |
Dose 1 | 1.00 | 1.00 | 1.00 | [13] |
Dose 2 | 1.00 | 1.00 | 1.00 |
After 8–21 days | | | |
Dose 1 | 1.30 | 1.11 | 1.50 |
Dose 2 | 1.00 | 1.00 | 1.00 |
Age distribution of non-severe RVGE (%) | | | |
< 1 month | 1% | | | [28] (Pakistan RVGE visits age distribution) Burr shape1 (c) = 2.51 Burr shape2 (k) = 1.29 Burr scale (a) = 42.33 |
< 2 months | 2% | | |
< 3 months | 8% | | |
< 6 months | 28% | | |
< 12 months | 73% | | |
< 24 months | 95% | | |
< 36 months | 98% | | |
< 48 months | 99% | | |
< 60 months | 100% | | |
Age distribution of severe RVGE (%) | | | [28] Age distribution fitting: Burr shape1 (c) = 2.59 Burr shape2 (k) = 1 Burr scale (a) = 36.19 |
< 1 month | 0% | | |
< 2 months | 5% | | |
< 3 months | 8% | | |
< 6 months | 28% | | |
< 12 months | 73% | | |
< 24 months | 95% | | |
< 36 months | 98% | | |
< 48 months | 99% | | |
< 60 months | 100% | | |
Intussusception age distribution** | | | | |
< 1 month | 0% | | | [13] Age distribution fitting: Burr shape1 (c) = 7.19 Burr shape2 (k) = 0.15 Burr scale (a) = 19.52 |
< 2 months | 1% | | |
< 3 months | 2% | | |
< 6 months | 22% | | |
< 12 months | 75% | | |
< 24 months | 85% | | |
< 36 months | 90% | | |
< 48 months | 95% | | |
< 60 months | 100% | | |
DALY calculation | | | | |
Non-severe RVGE | | | | |
DALY weight | 0.19 | 0.19 | 0.26 | [29] (proxy: Moderate diarrhoea) |
Duration of illness (days) | 3 | 3 | 7 | [30] and EPI assumption |
Severe RVGE | | | | |
DALY weight | 0.25 | 0.25 | 0.36 | [29] (proxy: Severe diarrhoea) |
Duration of illness (days) | 7 | 6 | 8 | [30] and EPI assumption |
Intussusception | | | | |
DALY weight | 0.32 | 0.32 | 0.44 | [29] (Abdominopelvic problem, severe) |
Duration of illness (days) | 7 | 5 | 9 | [13] |
Vaccine effectiveness | | | | |
dose 1, initial VE | 100% | 45%* | 100% | [12] VE and 95%CI Assume efficacy of 3 doses equal to 2 doses Fitted the observed effectiveness of each dose and time of administration to the post-licensure surveillance data |
dose 2, initial VE | 100% | 45%* | 100% |
dose 3, initial VE | 100% | 45%* | 100% |
Mean duration of VE after each dose (in months) | 10 | 10 | 10 |
Parameter 2 (alpha or shape)/standard error | 3 | 3 | 3 |
Vaccine coverage | | | | |
Dose 1 | | | | |
2018 | 81% | 69% | 93% | [32], DTP1 as proxy for RV dose1 It was aligned with observed coverage in post-licensure surveillance. |
2019 | 75% | 64% | 86% |
2020 | 78% | 66% | 90% |
2021 | 74% | 63% | 85% |
2022 | 77% | 65% | 89% |
2023–2034 | 77% | 65% | 89% |
Dose 2 | | | | |
2018 | 77% | 65% | 88% | [32] Average DTP 1 and 3, proxy for RV dose 2 It was aligned with observed coverage in post-licensure surveillance. |
2019 | 74% | 62% | 85% |
2020 | 74% | 63% | 85% |
2021 | 70% | 60% | 81% |
2022 | 73% | 62% | 84% |
2023–2034 | 73% | 62% | 84% |
Dose 3 | | | |
2018 | 72% | 61% | 83% | [32] 2022, DTP3, proxy for RV dose 2 It was aligned with observed coverage in post-licensure surveillance. |
2019 | 72% | 61% | 83% |
2020 | 70% | 60% | 81% |
2021 | 66% | 56% | 76% |
2022 | 69% | 59% | 79% |
2023–2034 | 69% | 59% | 79% |
Coverage timeliness | | | | |
Median age at dose 1, in weeks (IQR) | 7 | 6 | 12 | [25, 33] |
Median age at dose 2, in weeks (IQR) | 20 | 16 | 25 |
Median age at dose 3, in weeks (IQR) | 29 | 19 | 43 |
Dose 1 | | | | |
< 1 month | 0% | 0% | 0% | [33] Scale: 12.89 (7.25, 22, 94) Shape: 1.91 |
< 2 months | 53% | 45% | 61% |
< 3 months | 63% | 54% | 72% |
< 6 months | 71% | 60% | 82% |
< 12 months | 75% | 64% | 86% |
< 24 months | 77% | 65% | 89% |
< 36 months | 77% | 65% | 89% |
< 48 months | 77% | 65% | 89% |
< 60 months | 77% | 65% | 89% |
Dose 2 | | | | |
< 1 month | 1% | 1% | 1% | [33] Scale: 16.96 (11.72, 24.55) Shape: 2.97 |
< 2 months | 3% | 3% | 3% |
< 3 months | 31% | 26% | 36% |
< 6 months | 61% | 52% | 70% |
< 12 months | 70% | 60% | 81% |
< 24 months | 73% | 62% | 84% |
< 36 months | 73% | 62% | 84% |
< 48 months | 73% | 62% | 84% |
< 60 months | 73% | 62% | 84% |
Dose 3 | | | |
< 1 month | 0% | 0% | 0% | [33] Scale: 28.93 (19.39, 443.15) Shape: 2.75 |
< 2 months | 0% | 0% | 0% |
< 3 months | 5% | 4% | 6% |
< 6 months | 53% | 45% | 61% |
< 12 months | 66% | 56% | 76% |
< 24 months | 69% | 59% | 79% |
< 36 months | 69% | 59% | 79% |
< 48 months | 69% | 59% | 79% |
< 60 months | 69% | 59% | 79% |
Vaccine price per dose (US$) with Gavi subsidy | | |
ROTARIX, 1-dose per vial, Liquid | $0.20 | | | [15] |
ROTASIIL, 1-dose per vial, liquid | $0.13 | | |
ROTASIIL, 2-dose per vial, liquid | $0.13 | | |
ROTAVAC, 5-dose per vial, liquid | $0.13 | | |
Vaccine price per dose (US$) without Gavi subsidy | | |
ROTARIX, 1-dose per vial, Liquid | $ 2.36 | | | |
ROTASIIL, 1-dose per vial, liquid | $ 1.05 | | |
ROTASIIL, 2-dose per vial, liquid | $ 0.80 | | |
ROTAVAC, 5-dose per vial, liquid | $ 1.15 | | |
International handling (% of vaccine price) | 3.00% | 2.55% | 4.50% | [36] |
International delivery (% of vaccine price) | 5.00% | 4.00% | 7.50% |
Vaccine wastage rate | | | | |
ROTARIX, 1-dose, Liquid | 5% | 4% | 8% | [35] and expert consensus |
ROTASIIL, 1-dose, liquid | 5% | 4% | 8% |
ROTASIIL, 2-dose, liquid | 9% | 8% | 14% |
ROTAVAC, 5-dose, Liquid | 15% | 13% | 23% |
Syringes wastage rate | 5% | 4% | 8% | [35] and expert consensus (the same as vaccine wastage rate) |
Health system delivery cost per dose (US$) for 2018 onward | 2.18 | 1.09 | 4.36 | [38] |
Additional health system cost per dose in the first year (2025) associated with switching from ROTARIX to ROTAVAC / ROTASIIL (US$)§ | 0.80 | 0.68 | 0.92 | [17, 38] Average of switching cost reported in Palestine and Ghana studies Lower bound: Ghana: US$0.68 per dose Higher bound: Palestine: US$0.92 per dose |
§ We used the mid-range between the switching costs reported by Palestine’s and Ghana’s studies divided by the number of doses to deliver in year 1. We used the low and high values for sensitivity analysis. |
Table 2
Input parameters for estimating health service costs in 2022 US$, Afghanistan
Parameter | Mid-point value | Low bound | High bound | Sources |
Healthcare costs of RVGE | | | | |
Government perspective | | | | |
Non-severe outpatient visit (US$ 2022) | $5.04 | $2.52 | $10.08 | [31] |
Severe outpatient visit (US$ 2022) | $5.04 | $2.52 | $10.08 |
Severe hospitalization (US$ 2022) | $17.56 | $8.78 | $35.12 |
Societal perspective | | | |
Non-Severe outpatient visit (US$ 2022) | $15.42 | $7.71 | $30.84 |
Severe outpatient visit (US$ 2022) | $15.42 | $7.71 | $26.46 |
Severe hospitalization (US$ 2022) | $38.05 | $19.03 | $76.10 |
Healthcare costs of intussusception (IS) | | | |
Government cost of IS hospitalization (US$ 2022) | $214.75 | 107.38 | $429.50 | [41,31] |
Societal cost of IS hospitalization (US$ 2022) | $234.38 | 117.19 | $468.76 |
To build consensus on model input parameters, we conducted consultation sessions with seven national experts between March and June 2024 during data collection and analysis. We engaged national experts in epidemiology, paediatric, immunization, health economics, and health system, representing stakeholders from the national expanded programme on immunization (EPI), WHO, UNICEF, and academia. This evaluation obtained approval from the Research Ethics Committee (ID: 29622, July 2023) of the London School of Hygiene and Tropical Medicine (LSHTM).
RVGE disease burden
All disease input parameters are shown in Table 1. We estimated the incidence of severe RVGE cases in children by multiplying the WHO Eastern Mediterranean region estimate of the rate of severe AGE (5,972 per 100,000 per year, aged < 5 years) by the rotavirus fraction among severe AGE cases (29.74%).[23] The rotavirus fraction was estimated using the mean of three estimates provided by Maternal and Child Epidemiology Estimation (MCEE), WHO/Centres for Disease Control and Prevention (WHO/CDC), and Global Burden of Disease (GBD).[4] The incidence of non-severe RVGE cases was estimated by subtracting the severe RVGE rate from the overall RVGE case rate (10,000 per 100,000 per year, aged < 5 years) estimated in a global systematic review and meta-analysis by Bilcke et al.[24] The rate of RVGE outpatient visits was calculated by multiplying the number of non-severe and severe RVGE cases by the proportion of children with diarrhoea seeking care (53.10%), as estimated from the Afghanistan Multiple Indicator Cluster Survey (MICS) 2022–2023.[25] The rate of RVGE hospital admissions was estimated using pre-vaccination RVGE surveillance in two regions of Afghanistan, Central and Western.[26] To estimate the RVGE mortality rate we used the mean of estimates by MCEE, WHO/CDC and GBD.[27]
The age distribution of severe RVGE (community cases, outpatient visits, admissions, and deaths) was estimated by week of age < 5 years using a parametric (Burr) distribution fitted to data from Afghanistan pre-vaccine surveillance (2013–2015). (Figure S1- panel a).[28] In the absence of national data on the age distribution of non-severe RVGE (community cases and outpatient visits), we used estimates for Pakistan from a study by Hasso-Agopsowicz and colleagues.[28] (Table 1)
For calculating DALYs, we used disability weights from Salomon et al.[29] and for duration of illness, we assumed 7 days for severe RVGE cases and 3 days for non-severe RVGE cases.[30]
RVGE healthcare cost
The cost per RVGE inpatient admission was taken from a systematic review of LMICs by Baral et al. (2020), which estimated the cost per gastroenteritis (GE) admission. The same source was used to estimate the cost per outpatient visit.[31] For the base case analysis, we included direct medical costs in the government perspective and the sum of direct medical, direct non-medical, and indirect costs in the societal perspective. We assumed that the cost per outpatient visit would be the same for severe and non-severe cases. Our analysis did not include costs for treatment given at home or through the informal sector. (Table 2)
Vaccination coverage, timeliness, and effectiveness
Observed RV vaccine coverage from post-licensure surveillance was very similar to the national administrative coverage reported in the WHO/UNICEF Estimates of National Immunization Coverage (WUENIC) for DTP1 and DTP3. Despite disruptions caused by the COVID-19 pandemic and political changes, the official administrative rotavirus vaccination coverage was reported to be stable from 2020 to 2022. We therefore used 2022 WUENIC estimates of DTP1 and DTP3 vaccine coverage as a proxy for RV1 and RV2 (average DTP1 and DTP3) vaccine coverage and assumed this would remain consistent throughout our evaluation period.[32] We included real-world vaccine delays and vaccine timeliness (coverage by week of age) using an analysis of data from a recent nationally representative survey, MICS 2022.[25, 33](Table 1)
We used estimates from a recent test-negative case-control study to approximate VE by time since dose administration using follow-up durations of 8 and 16 months for dose 1, and 7 and 15 months for dose 2.[12] We fitted a parametric gamma curve to each dose, assuming VE would be very high shortly after dose administration and then fall to very low levels after around 18 months of follow-up. (Figure S2) Estimates of VE and waning were very similar for doses 1 and 2, so we assumed the same waning rate for both.
Consistent with our previous analysis, we did not account for the indirect benefits of rotavirus vaccination, such as herd immunity, nor did we apply age restrictions to the vaccination schedule.
Vaccine price and delivery cost
According to Gavi’s eligibility and transition policy version 4 (effective from January 2023) Afghanistan is classified as being in the initial self-financing phase.[34] In our base-case scenario, we applied a co-financing contribution of US$ 0.20 per dose for ROTARIX (2-doses) and US$ 0.13 per dose for ROTAVAC (3-doses) and ROTASIIL (3-doses). In scenario analysis, we ran a separate analysis assuming the government would pay the full price for ROTARIX (US$2.36 per dose for the 1-dose vial presentation), ROTASIIL (US$ 0.80 for the 2-dose vial presentation, and US$ 1.05 for the 1-dose vial presentation) and ROTAVAC ($US 1.15 per dose for the 5-dose vial presentation).[15] We assumed prices would remain unchanged over the evaluation period. (Table 1 )
We applied wastage rates of 5%, 9%, and 15% for 1-dose, 2-dose and 5-dose vaccine presentations, respectively. These values were informed by reviewing the Comprehensive Multi-Year Plan (CMYP) 2021–2025. [[35]& Afghanistan EPI experts] A relatively high wastage rate of 15% for the 5-dose vaccine presentation has been derived from the current high wastage rates of the 1-dose and 2-dose vaccine presentations in the country. To calculate the cost of international handling and delivery, we applied 3% and 5% of the vaccine price, respectively.[35] The per dose cost of a safety bag with a capacity of 100 doses was estimated to be US$ 0.56.[35] No updated national data were available to estimate the health system delivery costs of rotavirus vaccination, so we used estimates for LMICs by Portnoy et, al (US$2.18 per dose in US$2022).[36] (Table 1) Due to limited data on the cost of switching products, we applied the average switching costs, which primarily cover the expenses of training, social mobilization, IEC materials, and stakeholder engagement, as reported by Palestine and Ghana, to the year of switching (2025).[17, 37] We calculated the mid-range cost between Palestine and Ghana and divided it by the number of doses to be delivered in the first year with the 3-dose products. (Table 1 )
Intussusception burden, risk, and costs
Intussusception (IS), invagination of a bowl segment causing blockage, is a rare but fatal medical condition if left untreated.[5, 38] In some (mostly high-income) settings rotavirus vaccination has been associated with a small excess risk of IS.[5] We estimated the potential number of excess IS cases, hospital admissions and deaths to account for the potential excess risk of IS in the 1–7 and 8–21 days following the first and second dose of RV vaccination. The post-licensure self-controlled case series (SCCS) study for ROTARIX in Afghanistan reported a small relative risk (RR = 1.3) in the 8–21 day period following administration of the first dose, so we included this risk in the base analysis.[13]
Our estimates of the background rate of IS cases, hospital admissions, and mortality were obtained from post-licensure surveillance of children aged < 12 months. This was the best available data source given that no IS surveillance existed pre-vaccine, and the RR of vaccination was very minor and restricted to specific age windows post-vaccination. The admission rate < 12 months was rescaled to age < 5 years using estimates from a systematic review of the proportion of < 5 years IS hospital admissions that occurred by age 1 year in the WHO Eastern Mediterranean region.[4, 13, 39] The incidence rate of IS cases was derived by inflating the IS hospital admission rate to account for the percentage of children without access to care, using WUENIC estimates of the coverage of DTP1 in the year 2022 as a proxy for the maximum possible access to healthcare for severe conditions.[13, 32] (Table 1) Estimates of the background rate of IS mortality were based on observed fatality rates among children < 1 year from post-licensure surveillance. These were rescaled to represent children < 5 years and inflated by DTP1 coverage to account for the percentage of children without access to care.[13, 19] The age distribution of all intussusception outcomes was calculated by week of age < 5 years using a parametric (Burr) distribution fitted to post-licensure surveillance data. (2018–2022).[13] (FigureS1- panel b)
Direct healthcare costs associated with intussusception treatment were included in the government perspective. These were estimated using the average cost per bed day in the National Hospitals Cost Analysis report, considering the average number of 5 days required for IS treatment.[40]. From the societal perspective, we also included direct non-medical cost and indirect cost estimates from Baral’s study without accounting for any allowance for the cost of surgery.[31] (Table 2)
Uncertainty analysis
We conducted deterministic scenario analyses to assess the sensitivity of the cost-effectiveness results to changes in different combinations of parameter values. We applied -/+15% to lower and higher bound of base values.[41] We considered the reported switch cost per dose of each country, Palestine and Ghana, to determine the lower and upper bounds of the vaccine delivery cost in the first year of the switch for the sensitivity analysis. We ran a scenario unfavourable to RV vaccination using the upper bound of the vaccine delivery cost per dose, a vaccine price without Gavi’s subsidy, the lower bounds of the disease burden rate, and the lower bounds of the healthcare costs. We also ran scenarios favourable to RV vaccination by assuming “on-time” vaccine administration and using the upper bounds of the disease burden rates and healthcare costs, and lower bound of vaccine delivery cost, and upper bounds of RVGE burden. Lastly, in one of the one-way sensitivity analyses, we assumed no vaccine impact on non-severe RVGE.
We ran a probabilistic sensitivity analysis (PSA) with 1000 Monte Carlo simulation runs. Within each run of the model, all parameters were simultaneously varied within their specified ranges. This was done to generate a 95% uncertainty interval around the central estimates of the cost per DALY averted. It was also done to determine the proportion of probabilistic runs with a cost-effectiveness ratio falling below CETs ranging from US$ 0 to US$ 500.
Role of the funding source
The funder of the study had no involvement in the study design, data collection, data analysis, data interpretation, or report writing. The authors had full access to all the data in the study and had final responsibility for the decision in submit for publication.