We built a de novo cohort-based Markov model to assess the cost-utility of the following three rotavirus vaccination strategies in Spain:
-
No vaccination;
-
Universal vaccination; and,
-
Targeted vaccination, solely for preterm babies born between 25 and 32 weeks of gestation.
We compared the cost-utility of these strategies, using Rotarix® or RotaTeq®, approached from both a societal and Spanish National Health System (SNHS) perspective. The model represents the most important clinical events conceivably linked to rotavirus infection. Figure 1 shows the decision tree of the No-Vaccination strategy. The other two strategies use the same decision tree but also include a second branch for subjects who, for different reasons, were finally not vaccinated under the strategy. The Targeted Vaccination strategy is divided into two branches: vaccinated and non-vaccinated babies, where the “vaccinated” are preterm babies, who comprise a high-risk population, and the “non-vaccinated” are non-premature babies, who comprise a low-risk population.
It is a Markov model of annual cycles that follows a hypothetical cohort of all new-borns in Spain until the end of their life course, set for study purposes at age 100 years. In the model, all new-borns start in the “no previous infection” state and can evolve annually through the tree, remaining in the same state or moving to other two states, such as “post-infection” or “death”.
Incidences and probabilities
As rotavirus is not an infection subject to compulsory surveillance, the most accurate information on rotavirus incidence in Spain comes from the Conjunto Mínimo Básico de Datos/CMBD (Minimum Basic Data Set)(9). This database does not furnish information on second or third infections, and as a consequence, the model incorporates all the epidemiological data as rotavirus cases but not as infected persons. The CMBD is an exhaustive registry of hospitalizations in Spain. It can be used to estimate the annual incidence of Rotavirus Acute Gastro-Enteritis (RV-AGE), which accounts for all RV-AGE cases, whether the main cause or non-primary cause of hospitalization.
In addition, other data sources were used to identify Spanish incidence of rotavirus infections and to calculate probabilities (Supplementary file 1). The incidences were transformed into probabilities, using the following formula:
1 – e –rate*time
Although most of the data required to populate the model were identified, some uncertainties remained. Expert advice was taken on the assumptions, whose impact on the model’s results were then evaluated in the sensitivity analysis. The following assumptions were made:
-
A 50% underdiagnosis of RV-AGE hospitalization. While the precise percentage of underdiagnosis is unknown some must assumed because the diarrhea causative agent is not identified in many hospitalized cases(10–12) .
-
The application of the Parashar model(13) to the Spanish setting(14) shows that rotavirus cases requiring a medical visit account for 20% of all cases. The remaining 80% do not require a healthcare visit, being home-cared.
-
Hospitalized cases have previously received primary or emergency care.
-
Mortality in non-infected population is that of the general population. Mortality due to rotavirus is so low that a possible overestimation of mortality in non-infected can be considered negligible.
-
A vaccine coverage of 94.7% was used in the base case, with changes in the sensitivity analysis. This was the observed vaccine coverage in Spain in 2019 for the following universal vaccinations: poliomyelitis; diphtheria; tetanus; pertussis; hepatitis B; and meningococcus B(15).
-
The vaccine coverage of the at-risk population was expected be higher. We used Bruijning et al’s estimate(4), which indicates a 2.3% higher vaccination coverage.
-
The proportion of preterm babies among all new-borns was estimated using data sourced from the Spanish National Statistics Institute (Instituto Nacional de Estadística/INE), which estimates that from 2017 through 2019, 0.9% of deliveries occurred before week 32(16).
-
Rotavirus hospitalizations and emergency visits were higher in the at-risk population (OR: 2.8)(17).
-
Rotavirus nosocomial infections were more likely in the at-risk population (OR: 2.6)(18).
The efficacy, effectiveness and safety of both vaccines were obtained through systematic reviews (Supplementary file 3). While the systematic review on the safety of both vaccines indicated that an increase in vaccine-related adverse events had not been demonstrated, some observational studies nevertheless reported varying information (Supplementary file 3). In light of this, we assessed the potential influence of the risk of intussusception in the sensitivity analysis. The highest value of the range considered was that observed in the technical file of the European Medicines Agency (EMA) (a higher risk of intussusception due to vaccination of 6 cases per 100,000 vaccinated babies)(19).
Costs
Systematic reviews described in Supplementary file 3 were helpful in identifying cost information. In addition, searches were extended to other Spanish cost-specific data sources, as described in Table 1. Indeed, all data on costs were drawn from Spanish sources. We identified RV-AGE-related healthcare costs, as well as non-healthcare and indirect costs.
Table 1
Healthcare, non-healthcare and indirect costs associated with rotavirus infection (figures are in € and updated in September 2021).
| Direct costs | Indirect costs | TOTAL |
---|
Healthcare costs. SNHS perspective | Healthcare costs. Societal perspective | Non-healthcare costs. Societal perspective | Societal perspective |
---|
Primary care | Primary value | 17 a | 10 a | 14 a | 125 a | 166 |
| Year | 2004 | 2004 | 2004 | 2004 | - |
| Updated value | 22.56 | 13.27 | 18.58 | 165.88 | 220.29 |
Emergency care | Primary value | 204 a | 16 a | 17 a | 172 a | 409 |
| Year | 2004 | 2004 | 2004 | 2004 | - |
| Updated value | 270.71 | 21.23 | 22.56 | 228.24 | 542.74 |
Hospitalization | Primary value | 2163.61 b | 14 a | 10 a | 279 a | 2466.61 |
| Year | 2013 | 2004 | 2004 | 2004 | - |
| Updated value | 2315.06 | 18.58 | 13.27 | 370.23 | 2717.14 |
Nosocomial | Primary value | 744 c | 14 a | 10 a | 152.8 c | 920.8 |
| Year | 2006 | 2004 | 2004 | 1999 | - |
| Updated value | 924.79 | 18.58 | 13.27 | 238.98 | 1195.62 |
Home-based care | Primary value | - | 5 a | 7 a | 23.8 c | 35.8 |
| Year | - | 2004 | 2004 | 1999 | - |
| Updated value | - | 6.635 | 9.29 | 37.22 | 53.145 |
(a) Giaquinto 2007 (36); (b) CMBD 2010–2015 (6); (c) Díez-Domingo 2010 (38) |
The healthcare costs included in the analysis from a societal perspective were those for oral serum and anti-diarrheal drugs for mild cases. Non-healthcare and indirect costs included were transportation, extra-diapers, caregivers at home, and parental productivity losses (Table 1).
Other costs incurred were vaccine-acquisition and vaccine-administration costs. Shown below are the current sales prices in Spain, which are the same for both private and public (governmental) procurement.
-
Rotarix® full vaccination, two doses €187.32.
-
RotaTeq® full vaccination, three doses €208.50.
Administration costs were taken from the analysis performed by the Spanish Ministry of Health, which estimated an average cost of €6 to administer a vaccination dose(20). As the two Rotarix® doses (months 2 and 4) coincide with other vaccines included in the Spanish routine vaccination schedule, we reduced the cost by half. In the case of RotaTeq®, the first two doses would be in the same months thus making the cost incurred €3, but the third dose would require a specific visit and so the cost incurred would be the full €6.
Utilities
The systematic reviews provided us with enough information on utility losses on each RV-AGE case included in the model. We estimated utility losses through Quality Adjusted Life Years (QALY), including infected children and two caregivers. This is a common methodologic approach in rotavirus infections because most cases are in children and QALY losses often affects both parents.
We used the estimates of Marlow et al.(21) for QALY losses among children and their caregivers for high and moderate severities. QALY losses among children with mild cases have been estimated by Aidelsburger et al.(22), who provide QALYs for children under and over 18 months old: the median between these two values was used. Since caregiver QALY losses related with mild cases were estimated by Hansen-Edwards et al.(5), without distinguishing between primary and secondary caregiver QALY losses, we used the same value in both cases (Table 2).
Table 2
Utility losses attributable to each of the rotavirus cases that were included in the model.
| Rotavirus case | QALY lost | Source |
---|
Children | Mild case | 0.0011 | Aidelsburger, 2014 (26) |
Moderate case | 0.0029 | Marlow 2015 (25) |
Severe case | 0.0034 |
Primary caregiver | Mild case | 0.0014 | Hansen-Edwards 2017 (27) |
Moderate case | 0.0014 | Marlow 2015 (25) |
Severe case | 0.004 |
Secondary caregiver | Mild case | 0.0014 | Hansen-Edwards 2017 (27) |
Moderate case | 0.0015 | Marlow 2015 (25) |
Severe case | 0.0028 |
Analysis
We performed a cost-utility analysis of a potential cohort of new-borns of 400,000, which is a proxy of the average of new-borns in the last years in Spain(23). The Incremental Cost-Utility Ratio (ICUR) was calculated to estimate the average cost per person needed to obtain an additional QALY for a given vaccination strategy as against others.
We have adopted the ICUR threshold used in the reports of the Spanish Network of Health Technologies Assessment Agencies (Red Española de Agencias de Evaluación de Tecnologías Sanitarias y Prestaciones del Sistema Nacional de Salud/RedETS), which is below €25,000/QALY(24). The discount rate was 3% for both costs and utility, as recommended by the Spanish Economic Health Technology Assessment Guide (30). All costs were updated to September 2021 and calculations included a half-cycle correction. For analysis purposes we used the TreeAgePro 2018® software program.
In addition, we also performed a univariate deterministic sensitivity analysis for all the variables of the model (52 parameters). Ranges used in the sensitivity analysis were mainly those found in the same literature as that selected for the base case (see Supplementary file 2). When data on ranges were not available, a reduction of 60% and an increase of 100% were used. Tables in show the results of the sensitivity analysis for the comparison between targeted and universal vaccination, but all variables were tested and those with the most relevant variations in the ICUR were selected to be shown in Tornado diagrams. Results of sensitivity analysis on including non-vaccination as a comparator are available upon request.
Lastly, we also calculated the budget impact of each of three vaccination strategies, considering the model’s results for the first five years of life, since the costs of an annual cohort of new-borns over five years are similar to the expenses of the whole population in a year. This is in line with the five-year calculation period recommended by budget-impact analysis guidelines(25,26). The average annual budget that the SNHS would need to implement for each of the strategies was likewise calculated.