Economic Evaluation of Measles Eradication Study: Results for Six Countries and by Income Groups

Transcription

1 Economic Evaluation of Measles Eradication Study: Results for Six Countries and by Income Groups Prepared for: World Health Organization Ann Levin1, Colleen Burgess2, Louis Garrison3, Chris Bauch4, Joseph Babigumira3 August 17, Independent Consultant 2. MathEcology, LLC 3. Department of Pharmacy, University of Washington 4. Department of Mathematics and Statistics, Guelph University 1

2 Table of Contents Executive Summary Introduction: Background and Objectives Methodology Epidemiologic and Economic Transmission Model Natural History Age Classes and Aging Transmission Vaccination Model System Geographic Heterogeneity Probabilistic Uncertainty Analysis Transmission Model Scenarios Model Coding Model Validation Cost Estimation Calculation of Measles Immunization Costs Cost and Pricing Assumptions Cost Data Collection Probabilistic Uncertainty Analysis Integration of Cost and Transmission Models Methodology Global Extrapolation Methodology Extrapolation for Cost Estimates Economic Evaluation Methodology: Cost Effectiveness Analysis Results Country Level and Global Analyses Case Study Countries Global Analysis Cost and Cost Effectiveness Ratios for Case Study Countries Costs of Measles Vaccination in Six Countries Costs in Case Study Countries

3 6.2.3 Cost and Cost Effectiveness of Intermediate Measles Reduction Scenarios in Six Case Study Countries Cost Effectiveness of Achieving Measles Eradication Cost and Cost Effectiveness Ratios for Global Extrapolation Impact of Introduction of New Technology: Aerosol Device Financial Costs of Reaching Eradication by Consequences of Discontinuing SIAs due to Inadequate Funding Results of Sensitivity Analysis Discussion of Global Cost Effectiveness Results Conclusion References

4 Executive Summary The cost-effectiveness of five measles reduction scenarios were evaluated in comparison to the current strategy of 90% measles mortality reduction by The five scenarios were 1) 95% mortality reduction; 2) 98% mortality reduction; 3) eradication by 2020; 4) eradication by 2025; and 5) measles coverage if funding is longer available for SIAs in low-income countries. The most cost-effective scenario for reducing the number of measles cases and deaths was global eradication by 2020, followed by global eradication by There would be a significant reduction in measles cases and deaths since the last one would occur in The reduction in costs and in cases and deaths would be greater than would occur with a goal of eradication by 2025 since the benefits of eradication would be delayed. When the costs and impact were extrapolated to all countries, eradication by 2020 was found to be very cost-effective for all income groups. For countries that have not yet eliminated measles, eradication was very cost-effective in three of four income groups and cost in the fourth, lower-middle income, the group with the largest population. For countries that have already eliminated measles, all of the income groups had cost s. The scenario where measles eradication is achieved by 2025 would also be cost-effective. Similar to the eradication scenario by 2020, the more cost-effective strategy was the one where MCV1 and/or MCV2 are continued and SIAs are discontinued due to the lower costs. Among the three post-eradication strategies, if eradication is achieved in either 2020 or 2025, the more cost-effective are the ones that discontinue SIAs but maintain MCV1 and/or MCV2 since these strategies have the greatest cost s. However, any analysis of post-eradication strategies will need to consider the risk of re-introduction of the measles virus. If there is a risk of re-introduction, there would be more persons susceptible to measles if MCV2 and SIAs were discontinued. For example, the percentage of susceptible persons if eradication were achieved by 2020 would be 6.4% in 2050 with MCV1, MCV2 and SIAs, 9.5% if MCV1 and MCV2 were continued, and 10.3% if only MCV1 were continued. Achieving measles elimination in countries depends on having high quality SIAs, improvements in routine immunization, and good surveillance in place. In addition, an assumption was made that case importation would decrease. Thus, the incremental costs of achieving elimination were associated with the costs of improving the quality of SIAs, routine immunization and surveillance. The costs of increasing routine immunization coverage and finding harder-to-reach cases were assumed to be increasing and the rate of increase is greater at higher levels of coverage. In addition, costs per dose of SIAs are assumed to increase by approximately $0.01 per additional percentage of coverage. These increasing costs are not so high as to make eradication economically unattractive. The next most cost-effective scenario after global eradication is the intermediate strategy that reduced measles mortality by 98%. In this intermediate scenario, the combination of increasing 4

5 routine immunization coverage and improving surveillance and outbreak response with an assumption of a reduction in case importation results in a measles mortality reduction of 98%. The scenario is cost-effective or cost for four of six countries. The 95% reduction scenario also is cost-effective for five of the six countries although less so than the 98% scenario. It does not avert as many cases and deaths and is sometimes costlier than the 98% reduction scenario. One alternative technology was explored. Introducing aerosol vaccination could potentially bring cost s due to reductions in the cost of syringes and waste disposal. In addition, some potential s could be realized if the device required less time to administer measles vaccine. Since this device is still under development, the possibility of introducing it should be explored at a later date. The other scenario, maintaining current baseline levels of routine vaccination and discontinuing SIAs after 2010, is not cost-effective since the number of cases and deaths increase significantly, in spite of slightly decreasing costs. The results for this scenario suggest that it would be very preferable to continue with SIAs in the four non-elimination case study countries since the increases in costs are relatively low and many cases and deaths would be averted. 5

6 1. Introduction: Background and Objectives Measles is one of the most infectious and severe diseases of childhood and remains an important cause of morbidity and mortality in children in developing countries. In recent years, with the support of WHO and UNICEF, countries have accelerated their efforts to reduce measles morbidity and mortality both through increasing routine measles coverage and conducting periodic supplementary immunization activities (campaigns). In the period , these accelerated measles activities led to a 74% reduction in estimated global measles mortality (90% in the Eastern Mediterranean and 89% in the African regions) (WHO 2007, Cutts 2008). In addition, high coverage of two doses of measles vaccine (delivered through routine programs with or without supplementary campaign strategies) has virtually eliminated measles from the western hemisphere since November The current goals in the six regions for measles are (1) elimination in the regions of the Americas (AMR), Eastern Mediterranean (EMR), Europe (EUR) and Western Pacific (WPR) and (2) mortality reduction in AFR and SEAR in line with the WHO-UNICEF strategic plan (WHO- UNICEF 2001). Due to the success of the measles mortality reduction and elimination efforts thus far through the Measles Initiative and related WHO-UNICEF efforts, WHO has raised the question of feasibility of possible new goals such as the eradication of measles or further significant reductions in measles mortality. Part of the rationale is that measles is a good candidate disease for eradication. As in the case of smallpox, humans are the only hosts for the measles virus, transmission occurs only when an infected person shows symptoms, lifelong immunity is provided following recovery from the disease, and the vaccine provides life-long protection from the disease (Griffin 2006). However, a few factors make measles more difficult to eradicate than smallpox: its greater infectiousness, and greater difficulties in conducting surveillance and in detecting infected individuals. The differences between measles eradication and mortality reduction can be summarized in terms of their programmatic implications. Measles eradication can be defined as the interruption of measles transmission worldwide as a result of deliberate efforts; intervention methods may no longer be needed. Eradication represents the sum of successful elimination efforts in all countries (WHO-UNICEF 2001 Strategic Plan). With eradication, no reintroduction of measles cases should occur unless the virus is re-introduced due to bioterrorism or laboratory accidents. In the case of significant (90% or 98%) mortality reduction (compared to the year 2000 baseline), measles incidence would be very low worldwide but there would remain a sufficient number of interconnected subpopulations with enough susceptible individuals to sustain unbroken chains of transmission. Interventions would still be needed to control periodic outbreaks of the disease and/or importation of cases into countries that have eliminated the disease. WHO would like to assess the appropriateness of measles eradication in terms of programmatic, biological and economic considerations. As part of this broad feasibility assessment, it issued a 6

7 request for proposals (RFP) to develop an economic analysis of measles eradication. In response to this WHO RFP, Ann Levin, PhD and her team (Colleen Burgess, MS; Louis Garrison, PhD; and Chris Bauch, PhD) the Measles Eradication Economic Team (MEET) proposed to build upon their existing measles vaccine cost-effectiveness analyses and dynamic measles transmission models to evaluate the economic impact of measles eradication and mortality reduction efforts. The two objectives were to: 1) Compare the cost and cost-effectiveness of global measles eradication to the cost and cost-effectiveness of achieving an intermediate goal of mortality reduction or 90% compared to 2000 levels; and to 2) Carry out the analysis for six countries using primary data obtained from countries representative of the six WHO regions. The specific scenarios addressed in the analysis were the following: 1) 90% reduction by 2013 (same as current coverage levels except India completes a catch-up) 2) Measles coverage if funding not available to poorest countries 3) 95% mortality reduction by ) 98% mortality reduction by ) Eradication of measles by ) Eradication of measles by

8 2. Methodology Epidemiologic and Economic The cost of measles eradication in a given country is defined to be the aggregate incremental cost that would be needed to achieve eradication, i.e., above the costs that would be incurred to achieve a lower level target (e.g., a 90% reduction in mortality by 2013, as compared to the year 2000 baseline). To make the additional effort to reach eradication presumably requires some increase in the level of MCV1 coverage, MCV2 coverage, more frequent or effective SIAs, and improvements in surveillance of measles cases. Estimates are available for the average cost of providing each of these in the more routine situation. However, accelerating the trajectory to achieve measles eradication is likely to result in a higher marginal (and average) cost per dose for each of these efforts. For example, expanding MCV1 coverage faster than planned as well as introducing SIAs in more remote, less populated areas may require more transport and social mobilization. More frequent campaigns clearly increase the average number of doses received per child for a given time horizon, which is more costly. Thus, the cost per dose provided will be higher for both routine immunization and SIAs. For each of the six countries, our approach was first to solve the eradication infectious disease problem selecting values for the tunable parameters of MCV1, MCV2, and SIA intensity that would result in meeting the target, and then apply appropriate average increasing annual costs to estimate the aggregate incremental cost. In this methodology section, we first describe the structure and estimation of the measles epidemiologic transmission model, and the methods used to estimate costs. 2.1 Transmission Model The measles transmission model breaks each country down into districts, with each district being represented by a compartmental model consisting of a system of ordinary differential equations and projecting cases and deaths by age and year under various scenarios for vaccination Natural History Individuals can be susceptible, exposed, infectious, or recovered/immune (SEIR model). Upon transmission susceptible individuals enter the exposed class where they remain for the duration of the latent period. During the latent period, individuals do not transmit measles, by definition. After the latent period ends, exposed individuals enter the infectious class, within which they are able to infect other susceptible individuals. Measles-related mortality is assumed only to occur within the infectious class. Infected individuals who do not die due to measles or background mortality recover after a period of time and enter the removed class, where they remain for the duration of their lives. 8

9 2.1.2 Age Classes and Aging Age classes are: birth 1 year, 1 2 years, 3 5 years, 6 15 years, and 16+ years. Individuals are assumed to age at a constant rate, i.e., a fixed proportion of individuals transitioning from one age group within the appropriate disease compartment to the next during each weekly time-step. Though for testing purposes, currently, demographic processes are assumed to be deterministic, within the final model, a Gaussian noise term was added to the birth rate to capture the effects of demographic stochasticity, which can lengthen the time between peaks in epidemic activity and hence may have implications for the 3-year certification process that certifies whether a country has eliminated measles. Birth and death rates are taken from historical and projected sources (e.g., UN population division, primary data from country visits, or a combination of both data sources) Transmission Transmission rates are assumed to be age-specific, with higher rates between similar age classes, particularly at younger ages. The transmission rate from age class j to age class i is a product of age-dependent transmission coefficients and the age-specific contact rate. The age-specific contact rates are based on a large diary based survey developed by the POLYMOD group, a consortium of infectious disease researchers in Europe working on transmission dynamics. While these data are for European populations, and thus may underestimate within-household transmission in other regions due to variations in family size and structure, no similar data are available for low-income populations, and empirical contact patterns are likely to be more accurate than WAIFW matrices that cannot usually account for parent-child transmission in a robust way, even if the contact patterns come from European populations. For country-specific scenarios we have made adjustments to the POLYMOD-derived values as shown below, in order to accommodate variations in family size between countries. The force of infection upon a susceptible individual of age class i is given by 5 λ i = β ij I j j=1 N j according to the standard incidence assumption. The transmission rate Β ij (defined as the rate at which an individual in age class i is infected by any and all individuals in age class j) is further broken down according to: B M N i wm m 1 n 1 ij = κcij = κ = = M i ( wm ) m = 1 D ij mn i where κ is a transmission rate coefficient that is fitted; w represents the statistical weight of the m m th individual of age class i in the POLYMOD contact data (and the sum from 1 to M represents 9

10 ij the sum over all diary respondents of age class i); and D mn reported contact of age class j of the m th respondent, and hence represents the duration of the n th N n= 1 D ij mn represents the contact intensity, i.e., the sum of the duration of all contacts reported by an individual of age class i with individuals of age class j. Furthermore, the differences in household sizes between Europe and each of the six countries were used to adjust the POLYMOD contact data as follows. For the case of Uganda, for instance, the contribution to the contact matrix of each contact in the POLYMOD contact data made in a household was multiplied by a factor of 6.9/2.44, reflecting the average recent household size in Uganda (6.9, Ref. 1 below) and the ten European countries that were part of the POLYMOD contact survey (2.44, Ref. 2 below). This adjustment makes the assumption that the number of contacts per day scales linearly with the average household size. The resulting values for daily contacts, C ij, for the age classes used in the model (<1, 1-2, 3-5, 6-15, 16+) are below. Age of Individual Age of Contact Infants 1-2 years 3-5 years 6-15 years 16+ years Infants years years years years For the sake of comparison, the values of C ij for equal-sized age classes 0-10, 11-20, 21-30, 31-40, and 40+ are: Age of Individual Age of Contact Infants 1-2 years 3-5 years 6-15 years 16+ years Infants years years years years Contact Matrices for Case-Study Countries Below are the resulting contact matches for each of six countries for the age classes used in the model (<1, 1-2, 3-5, 6-15, 16+): Bangladesh (average household size 4.8 in 2002)

11 Brazil (average household size 3.28 in 2008) Colombia (average household size 3.9 in 2009) Ethiopia (average household size 4.8 in 1998) Tajikistan (average household size 7.0 in 2002) Uganda (average household size 6.9 in 2001) Contact Matrix for Global Analysis The matrix below is based on the average household size of 5.11 of the six countries included in the analysis The age-specific transmission rates are modified with a sinusoidal function of period 1 year to capture the effects of seasonal variation, yielding 11

12 ( α ) β () t = 1+ sint B ij ij The outputs of the model in the presence of vaccination corresponding to historical immunization activities have been validated by computing the correlation coefficient with casereporting data from the vaccine era from the six countries based on data availability. Additionally, measles transmission in Uganda in the absence of vaccination has been simulated and validated against the expected behavior of the disease in a purely susceptible population Vaccination Susceptible individuals may also enter the removed class via vaccination through one of four possible routes routine first-dose (MCV1) immunization at 9 or 12 months of age, routine second-dose (MCV2) immunization at 15 or 18 months of age or at school entry (for countries already employing this vaccination timing), immunization during periodic campaigns (SIA), or immunization during outbreak response activities. Simulations for countries not already implementing routine second dose vaccination incorporate MCV2 for children in the second year of life once the national average MCV1 coverage has reached and been maintained at 80% for three years, as per WHO guidance. Campaign vaccination is simulated as follow-up campaigns only, with assumed coverage rates of 90% attained over a period of six weeks during campaign years. Simulated SIA frequency is assumed to remain stable based on current levels of funding, and varies based on country-specific MCV1 coverage and the time required reaching the mortality-reduction or eradication goal. Current WHO guidelines for SIA frequency based on MCV1 coverage are as follows: MCV1 Coverage Age MIN Age MAX Frequency > 80% 9 mo 59 mo 4 yr 60% - 79% 9 mo 47 mo 3 yr < 60% 9 mo 35 mo 2 yr Outbreak response vaccination within the model is implemented only after 2010, and is currently triggered when measles incidence in a given district exceeds 20 cases per million population during a single weekly time-step. Vaccination associated with the given outbreak occurs four weeks after its detection, and targets the age groups representing approximately 90% of cases at a coverage rate of 90%. Vaccine efficacy is assumed to be dependent upon age of administration, with 85% efficacy under 12 months of age and 95% efficacy over 12 months of age after a single dose that is, 15% of vaccinated infants under 12 months of age and 5% of vaccinated individuals over 12 months of age will still be fully susceptible in spite of their vaccination status. Efficaciously vaccinated individuals are assumed to remain protected for life. 12

13 For routine immunization, the probability of receiving a second opportunity will depend on whether the first opportunity was acquired, with a higher degree of dependence between first and second doses for when the first and second routine doses occur closer together in age. Based on the stepwise function D(s), the degree of dependence between MCV1 and MCV2 is 75% for MCV2 given in the second year of life, and 25% for MCV2 given at school entry that is, for MCV2 given in the second year of life, 25% of previously unvaccinated children will receive a dose of measles-containing vaccine at this opportunity. Similarly, for MCV2 given at school entry, 75% of previously unvaccinated children will receive a dose at this opportunity Model System The figure below is a schematic of the model compartments. The compartments are generalized to display transitions over the entire population; however the model as implemented accommodates age-structure within the population and provides all outputs (vaccinations, infections, immunity and deaths) by age group for each of the six case-study countries simulated and for the global analysis. The model equations for the above-described measles transmission system are given by: ds dt k de dt dik dt dr dt k k I j MCV 1 MCV 2 SIA OR = bω N S π() t + τ + β ( e< 1ρ + D( s) e> 1ρ + e> 1ρ + e> 1ρ ) S μs j N j I j = Sk π() t + τmn + βkj σek μek j N j k mn kj k k k k k k = σe γi μi μ I k k k Mk k ( ) = γi + S e ρ + D() s e ρ + e ρ + e ρ MCV1 MCV 2 SIA OR k k < 1 k > 1 k > 1 k > 1 k μr k where S k (respectively, E k, I k, R k ) is the number of susceptible persons (respectively, exposed, infectious, removed) in age group k; and parameters are described in the table below. Aging processes are not represented in these equations: aging is represented as occurring at a constant per capita rate of leaving age class k and entering age class k+1. 13

14 Table Transmission model parameter values. Model Description Baseline value Lower, Upper Reference Parameter Range b t Birth rate, year t Varies according to district -- Primary data collection, UN Population Division μ t Background death rate, year t Varies according to district -- Primary data collection, UN Population β ij Rate of transmission from an infectious person of age class j to a susceptible person of age class i See Section on transmission rate α Seasonal forcing amplitude See Section on +/- 50% transmission rate 1/σ Mean incubation period 9 days -- 1/γ Mean effective infectious 7 days -- period* μ Μ,k Case fatality rate in age class k Varies according to country e <1 % immune after 1 dose, <1 year e >1 % immune after 1 dose, > 1 year MCV1 ρ Vaccine coverage, MCV1, kt, in age class k at year t MCV 2 ρ Vaccine coverage, MCV2, kt, in age class k at year t SIA ρ Vaccine coverage, kt, SIA, in age class k at year t OR ρ Vaccine coverage, kt, outbreak response, in age class k at year t D(s) Degree of dependence between MCV1 and MCV2 T mn ω Μ Proportionality constant governing importation of cases from district m to district n Gaussian noise term for demographic stochasticity Parameter governing how quickly case imports decline as global eradication is approach -- 85% -- 95% -- Varies according to district Varies according to district Varies according to district Varies according to district 75% for MCV2 at 15-18mo, 25% for MCV2 at school entry Varies according to country Varies according to scenario Division +/- 50% Primary data collection, WHO guidance /- 15% WHO guidance * Effective infectious period is somewhat shorter than actual infectious period since symptomatic cases typically convalesce in presence of immune adults

15 Primary data has been collected separately in six countries in order to parameterize the transmission models, and combined with case reporting (obtained from the WHO-UNICEF Joint Reporting Form) and population projections (obtained from the UN Population Division, and extrapolated to extend the available data through 2050) to drive the dynamics of the models including attack rates, force of infection, and case fatality rates Geographic Heterogeneity Districts in a given country have been aggregated into subpopulations according to MCV1 coverage. Individuals in a given subpopulation i become exposed at a rate that is proportional to the number of infectious individuals in other subpopulation districts according to: τ ik m N k = T Infk () t N k = 1 where Infn () t is the total number of infected persons (summed over all age classes) in subpopulation n at time t, N k is the number of persons in subpopulation k, N is the total population of the country, T is a population proportionality constant common to all districts in the country. T is calibrated to produce qualitatively reasonable dynamics in the absence of vaccination (e.g., correlated outbreaks across districts). Within baseline and no-sia post-2010 scenarios case importation is modeled as a stochastic process, in which the probability of an infected individual immigrating into the country is held constant through For reduction in mortality and eradication scenarios, however, the probability of case importation from other countries declines linearly until the goal is reached, according to π () t = M ( t t) erad where π(t) is the rate at which a susceptible is infected due to case imports and where t erad is the time of eradication according to the scenario. The value of M was chosen simultaneously with the value of T to recover qualitatively reasonable dynamics in the absence of vaccination Probabilistic Uncertainty Analysis For the sensitivity analysis, intervals were defined around a number of baseline transmission and cost parameter values (see Table 2.1.1), from which samples were taken to determine the degree of sensitivity of the model to these inputs. Each of these parameters was varied individually, while the others were held constant at their baseline values, and the model was run for the sample country of Ethiopia to determine the overall impact of this variation on the cost per DALY ICER associated with eradication by 2020 with no SIAs post-eradication. Included in this analysis were the following parameters: 15

16 Transmission parameters: Transmission rate coefficient (κ) Measles case fatality rate (μ Μ,k ) Probability of case importation from outside the country Probability of case importation between districts Seasonal forcing amplitude (α) Degree of dependence between MCV1 and MCV2 (D(s)) Cost parameters: Initial cost per dose for routine vaccination Cost per percent increase in coverage for routine vaccination Initial cost per dose for campaign vaccination Cost per percent increase in coverage for campaign vaccination Cost to household Cost to treat a measles case Transmission Model Scenarios For each scenario of the country-level analysis, ten iterations of the stochastic model were performed and the results were then averaged to produce mean outputs which were then utilized to produce the cost analysis. To achieve the goal appropriate to the scenario by the target date, vaccination strategies were tuned in a sequential manner until the target is reached. Based on the starting point for the specific country under consideration, for each scenario: (1) MCV1 coverage is increased and if appropriate, age of MCV1 is increased from 9 months to 12 months once 95% reduction in mortality is achieved; (2) MCV2 is introduced and / or increased; (3) SIA coverage and / or frequency is increased; and (4) Outbreak response (OR) vaccination is implemented in sequence, until the reduction in mortality or eradication goal is attained. Reduction in mortality goals were assumed to be attained if the average percent reduction in mortality when compared with 2000 levels for post-goal years is within one percentage point of the target (for example, for 95% reduction in mortality by 2015 scenarios, the goal was assumed to be attained if average percent reduction in mortality over the period is between 94% and 96%). 16

17 2.1.9 Model Coding The differential equation model described above was coded in the MATLAB programming language (m) and solved via fixed time-step fourth-order Runge-Kutta methods. Input data, including country-specific data such as yearly population numbers, birth rates and death rates, were read in from UN Population Division delimited text files. Country-specific historical vaccination data is also read in from delimited text files constructed from primary data collected from case-study countries and from WHO-UNICEF estimates or data derived from the GIVS (specifically historical SIA coverage rates). The model code was constructed to handle differences between districts with respect to demographics, but also with respect to case fatality rates, coverage rates for routine vaccination, and SIA frequency and coverage. Although for single-country simulations SIAs were assumed to occur in all districts simultaneously, for global simulations allowing countries to implement different interventions while the model runs on all countries simultaneously should allow for the exploration of various strategies to achieve mortality reduction and eradication goals. Scenario-specific parameters such as MCV1 starting and target coverage, MCV2 starting and target coverage, and SIA coverage and frequency were specified at simulation run time. WHO guidance with respect to timing and requirements for the introduction of MCV2, termination of SIAs, implementation of outbreak response immunization, and switching from M to MR and MMR vaccines are implemented as subroutines called from the main simulation. To accommodate the three-year certification period once a country achieves a particular goal, an internal timer is initiated and during this period pre-goal vaccination strategies continue. Once the timer terminates post-goal strategies are implemented, keeping whatever coverage levels are in place at the end of the certification period (as appropriate to the scenario). Cost calculations were incorporated into the MATLAB code as well in order to facilitate costeffectiveness analysis and probabilistic uncertainty analysis Model Validation As one method of validation of model functionality, additional scenarios for Uganda were run in which no measles vaccination routine or campaign was assumed to occur over the entire simulated time period from in order to confirm that transmission dynamics and seasonal oscillations fall in line with expectations. As shown in figures (a) and (b) below, large multi-year oscillations in country-level monthly incidence values correspond with larger outbreaks occurring every 2 years, with low-level endemic transmission continuing between larger outbreaks. 17

18 Figure (a). Country-level monthly incidence for 10-district simulation for Uganda, no vaccination, showing larger bi-annual outbreaks and multi-year periodicity. Figure (b). Country-level monthly incidence for 10-district simulation for Uganda, no vaccination, showing low-level endemic transmission occurring between outbreaks. 18

19 2.2 Cost Estimation Calculation of Measles Immunization Costs To estimate the costs of each specific strategy, the projected annual program costs were summed for the measles immunization activities for each strategy, country and year until measles eradication is reached. After eradication is reached, the costs of maintaining it were estimated for each country until at least 2030 and until 2050 if not cost-effective by The average cost per dose was estimated by dividing total annual costs by the number of doses administered. Conceptually, in basic economics, simple static cost curves for a firm do not have an explicit time dimension. Average cost varies only as a function of output level. By analogy here, for a fixed time horizon, say 2020, the output level would be the number of children with immunity. The target for eradication would be around 95% by the end of the period. Clearly, there is a time dimension here, since achieving 95% immunity at the end of the period would cost a different amount that achieving it earlier on. Computationally, we are estimating a dynamic cost function, i.e., one with a time dimension. For a given country in a given year, we estimate total annual costs based on the levels of MCV1, MCV2, and SIAs on the eradication trajectory, which is derived from our transmission modeling of the policy scenarios. For that country in that year, the average cost of MCV1, MCV2, and SIAs, respectively, could well be different than in the prior or in the following year. These average annual costs per dose delivered will increase over time as the trajectory approaches the eradication level. The costs of routine immunization and SIAs were estimated following an ingredients approach (WHO Guidelines 2008), to include the following: TC = cost pers + cost vacc + cost is + cost t + cost main +cost sm + cost ms +cost cc where pers = personnel, vacc = vaccines, is = injection supplies, t = transport, sm = social mobilization, ms = monitoring and surveillance, and cc = capital costs. When applicable, the cost of MR and MMR were substituted for measles vaccines. The cost of personnel was estimated by multiplying the average time that they spend on routine immunization and SIAs 1 by their annual salaries. The cost of vaccines and injection materials was calculated by multiplying the number of children in the target population multiplied by the strategy-specific wastage rates and coverage rates. The cost of other operational costs i.e., transport, maintenance and social mobilization were estimated by multiplying the value of the resources used for each of these by the amounts required (e.g., the number of litres of fuel used for transport was multiplied by the number required to transport the vaccines and other materials to the vaccination sites). 1 In the cmyp excel workbook, information on the amount of total annual time that health personnel spent on immunization activities is provided. 19

20 The costs of surveillance was estimated by calculating the value of the resources required for the following components: refresher training for surveillance focal persons, regional surveillance review meetings, printing of surveillance tools, transport of specimens, specimen collection supplies, measles laboratory reagents and supplies, monitoring and supervision of case-based surveillance, and printing of surveillance tools. The annualized amortized value of capital goods required for measles eradication (e.g., cold chain equipment, vehicles and laboratory equipment) was estimated when these are purchased for measles elimination and/or eradication. The medical s of not treating averted measles cases was also estimated using available data on the costs of treatment. These were estimated based on the treatment required for expected sequelae that would occur with measles. The cost s of each strategy also was also estimated using aerosol devices. This estimation involved substituting the cost of aerosol vaccines for liquid vaccines and syringes. As part of the calculation, potential cost s of using aerosols rather than liquid vaccines were estimated. The social costs of obtaining measles vaccinations both at routine immunization and SIAs were estimated if sufficient data were available. These costs include transport and travel time to the facility or outreach site and waiting time at the vaccination site. Other costs that were estimated include the costs of treating adverse events, technical assistance, and cost of managing outbreaks in developed countries. The additional and increasing costs of reaching measles eradication over a mortality reduction goal were estimated by costing the inputs that are needed to enhance measles immunization i.e., increase routine coverage (MCV1 and MCV2), including outreach activities, by reaching hard-to-reach population, improve surveillance in the country and conduct more frequent SIAs. In conflict countries, it is likely that transport costs will substantially increase due to the use of alternative forms of transportation such as air and boat. 20

21 2.2.2 Cost and Pricing Assumptions Table presents the data requirements, data sources and assumptions for the cost analysis. Table Parameters, Data requirements, Sources and Assumptions for Cost Analysis Parameters Data Data Source Assumption made Recurrent Cost Components Capital Costs Personnel Vaccines Injection Supplies Operational Costs Social Mobilization Monitoring and Evaluation Surveillance Cold Chain Equipment Vehicles and other Transport Data collection in six countries, cmyps, published studies Data collection in six countries, cmyps, published studies Percentage of immunization costs allocated to measles based on number of doses of vaccines given ; prices assumed to increase at same rate Only included if equipment purchased for measles eradication Medical Savings Societal Costs Discount Rate Period of Time Existing and New Technologies Scaling-Up % with sequelae from measles, cost of treatment Client Costs for Travel and Waiting 3% for costs and effects (sensitivity analysis from 1-7%) (will also check whether cost-effective at ) Aerosol presentation for measles and MR Rate of increase in costs with increase in coverage WHO Choice, data collection Data collection in six countries if possible, published studies Data collection in six countries and PAHO countries Cost per dose of RI and SIAs increase with coverage The prices of vaccines and injection supplies were based on current UNICEF prices and wastage rates were taken from country comprehensive multi-year plans. Vaccine wastage rates were taken from country data. Shared costs for routine immunization were allocated by assuming 10% are for measles vaccination. Capital costs were only included if additional cold chain equipment, vehicles and laboratory equipment would be purchased for measles eradication activities. All prices and improved efficiencies are assumed to increase at the same rate (Gold et al., 1996), except for improved efficiencies due to the introduction of aerosol vaccine. 21

22 Discounting of both costs and benefits employed a rate of 3% (Gold et al., 1996) Cost Data Collection The study team collected cost data from various sources: WHO HQ, visits to six countries, and regional offices. The six countries where data were collected include Uganda, Ethiopia, Bangladesh, Colombia, Tajikistan and Brazil. During the six country visits, data on costs of conducting routine immunization, SIAs and surveillance were collected from country immunization program offices and WHO offices. The team also conducted interviews with key stakeholders regarding the resources required in their countries to increase coverage of routine immunization, SIAs and surveillance to achieve measles eradication. In addition, during the country visits, the study team traveled to a sample of districts/regions to collect information on local resources used for measles immunization activities. The study team also used information collected from regional WHO offices to ascertain the costs of enhancing measles vaccination activities and surveillance so that the various goals can be achieved Probabilistic Uncertainty Analysis As described above, intervals were defined around each baseline parameter value, from which samples were taken for probabilistic uncertainty analysis Integration of Cost and Transmission Models A cost module was written in MATLAB for incorporation into the code that runs the transmission model. Hence, for each infected case and vaccinated person, a cost was attached and the MATLAB code outputs the resulting costs as well as health outcomes. 22

23 3. Methodology Global Extrapolation of Cases, Deaths, and Vaccinations For the global analysis, the existing country-level dynamic transmission model was expanded to represent the world as a single Mega-Country with 180 districts, each representing an individual country. Utilizing the same model structure as described above, country-level values were defined for the following parameters: Annual age-specific population numbers (historical and projected); Annual birth rates (historical and projected); Annual background death rates (historical and projected); Annual age-specific life expectancy (historical and projected); Disease-specific death rates; Historical vaccination activities and coverage levels; and Target vaccination coverage levels. Global values were employed in the contact matrix, as defined in the sections above. A minimum of three iterations were performed for each scenario, due to the massive CPU time required for each iteration, and resulting outputs were averaged for use in the global cost analysis. The mean outputs for cases, deaths and vaccinations were then distributed among low, lower-middle, upper-middle and high income groups for countries that have and have not eliminated measles transmission on a national scale, based on the historical percentage of global cases and vaccinations associated with the countries included in each of these classifications. For each of these classifications, total and incremental costs and ICERs were derived based on group-specific estimates for cost per dose for routine and SIA vaccination; cost to household; and cost of case treatment. 4. Methodology Extrapolation for Cost Estimates For each of the six countries, estimates of the higher average cost per dose of measles vaccine were developed for each year until the countries reach measles eradication. For routine immunizations (MCV1 and MCV2) and SIAs, the cost of reaching hard-to-reach populations will rise due to lower productivity rates (i.e., fewer children seen per provider per day at fixed sites and outreach sessions), higher wastage rates, higher transport costs, improved measles surveillance etc. Based on the local data that we obtained as well as country-specific plan data, we developed a cost curve specification for the six countries that predicts the rising marginal and average cost per dose delivered as a function of the routine coverage proportion. How this was used to extrapolate to the routine costs in other developed countries depended on the patterns that we observe across the six countries. Data on the average increasing costs of measles immunization was taken from cmyps, other cost studies, regional measles eradication plans and information gathered from the six countries on strategies required to increase coverage. 23

24 The rising cost curve estimated from the six countries was applied to the other countries to make the global extrapolation. In countries that have already eliminated measles except for occasional outbreaks, the cost of handling measles outbreaks was estimated. 5. Economic Evaluation Methodology: Cost Effectiveness Analysis The costs estimated for each scenario were compared with outcome measures obtained from the dynamic modeling. These measures included measles cases, deaths averted, and disabilityadjusted life years (DALYs) averted for each strategy. Sensitivity analysis was conducted on key epidemiological and cost parameters (e.g. discount rate, case fatality rates, vaccine wastage rates and vaccine prices) that have uncertainty. The cost-effectiveness of the various strategies was judged against the Commission on Macroeconomics and Health threshold that the cost per DALY be less than three times the per capita national GDP for each country. The cost-effectiveness of measles eradication/mortality reduction strategies was also compared to other public health interventions using published data. 6. Results Country Level and Global Analyses Case Study Countries Measles morbidity and mortality for each of the six case-study countries were generated by the transmission model described above for reduction in mortality and eradication scenarios over the time-periods from and In Uganda (current average MCV1 = 68%), Ethiopia (current average MCV1 = 72%), Bangladesh (current average MCV1 = 88%), and Tajikistan (current average MCV1 = 86%, MCV2 = 83%) vaccinations were assumed to ramp up under mortality reduction and eradication scenarios, as illustrated for Uganda and Ethiopia in figures and 6.1.2, respectively (additional details provided in Appendices). In these two countries, significant ramp-up of vaccination, in terms of both routine vaccination and outbreak response, is required to move from 95% reduction in mortality to 98%. 24

25 Figure Monthly measles incidence and percent of population under 5 years protected by vaccination, Uganda. 25

26 Figure Monthly measles incidence and percent of population under 5 years protected by vaccination, Ethiopia. For the two elimination countries (Brazil and Colombia) vaccination strategies were assumed to remain at current levels, with the exception of the various post-eradication scenarios within which dropping vaccination opportunities is explored. Discounted cases for baseline, 95% and 98% reduction in mortality and eradication by 2020 with discontinued SIAs post-eradication scenarios are presented in table Note that, since Brazil and Colombia have already eliminated measles transmission nationally, they are not pursuing the reduction in mortality goals, but rather are experiencing fewer cases due to lower case importation as a result of reduced incidence globally as other countries reach these targets. 26

27 Table Discounted total cases by country and scenario. Country Baseline 95% RM 98% RM E2020 Bangladesh 17,638,000 9,368,000 9,198,000 2,353,000 Brazil 2,000 1,000* 1,000* 500 Colombia 4,000 3,000* 3,500* 900 Ethiopia 6,390,000 3,490,000 1,850, ,000 Tajikistan 69,000 26,000 23,000 8,000 Uganda 413, , ,000 27,000 In non-elimination countries, campaigns (both SIAs and outbreak response) are more effective than ramping up routine vaccination at reducing mortality quickly, though funding shortfalls may make such activities to sustain on the long-term. For Ethiopia, Bangladesh and Uganda, the scenario in which regular SIAs are discontinued after 2010 was evaluated, and results indicate significant increases in mortality over the simulation period in the absence of this intervention (see Table 6.1.2) Table Discounted total deaths by country with discontinuation of SIAs post Country Baseline No SIA post-2010 % Increase Ethiopia 79, ,000 39% Bangladesh 118, ,000 48% Uganda 11,000 93, % Global Analysis Measles morbidity and mortality for the world as a mega-country consisting of 180 districts were generated by the transmission model described above for reduction in mortality and eradication scenarios over the time-periods from and Global simulations yielded estimated mortality for 2000 at 853,000 (743, ,000, within WHO estimated mortality range of 530, ,000), and mortality reduction of 90%, 96% and 98% for baseline, 95% and 98% reduction in mortality scenarios, respectively. In countries with low current average routine vaccination coverage levels (average MCV1 less than 70% - 80%, depending on the scenario) vaccinations were assumed to ramp up under mortality reduction and eradication scenarios to achieve the goal by the target date. Global monthly incidence per 1,000 population for reduction in mortality and eradication scenarios is 27