Cost-effectiveness of measles control during elimination in Ontario, Canada, 2015

Background Given that measles is eliminated in Canada and measles immunisation coverage in Ontario is high, it has been questioned whether Ontario’s measles outbreak response is worthwhile. Aim Our objective was to determine cost-effectiveness of measles containment protocols in Ontario from the healthcare payer perspective. Methods We developed a decision-analysis model comparing Ontario’s measles containment strategy (based on actual 2015 outbreak data) with a hypothetical ‘modified response’. The modified scenario assumed 10% response costs with reduced case and contact tracing and no outbreak-associated vaccinations; it was based on local and provincial administrative and laboratory data and parameters from peer-reviewed literature. Short- and long-term health outcomes, quality-adjusted life years (QALYs) and costs discounted at 1.5%, were estimated. We conducted one- and two-way sensitivity analyses. Results The 2015 outbreak in Ontario comprised 16 measles cases and an estimated 3,369 contacts. Predictive modelling suggested that the outbreak response prevented 16 outbreak-associated cases at a cost of CAD 1,213,491 (EUR 861,579). The incremental cost-effectiveness ratio was CAD 739,063 (EUR 524,735) per QALY gained for the outbreak response vs modified response. To meet the commonly accepted cost-effectiveness threshold of CAD 50,000 (EUR 35,500) per QALY gained, the outbreak response would have to prevent 94 measles cases. In sensitivity analyses, the findings were robust. Conclusions Ontario’s measles outbreak response exceeds generally accepted cost-effectiveness thresholds and may not be the most efficient use of public health resources from a healthcare payer perspective. These findings should be balanced against benefits of increased vaccine coverage and maintaining elimination status.

as well as the initial number of susceptible people.
6. The infection transmission rate is a product of contact rate and infectivity. Infectivity is taken to be the same over all infectious groups; 7. To approximate contact rates between different age groups we use a survey given in [1]. Data relevant to the Netherlands (Table S1) was chosen. Note that the grouping in this matrix is coarser than the susceptibility data provided by PHO. We will deal with this in two ways.
We will develop a model where the size of contact matrix will fit the length of the susceptibility vector, by making subgroups in each age group of the contact matrix have the same contact rates (for example, we will break group 0-4 years into 6 groups with similar contacts. The second approach involves grouping age groups in the PHO provided data into groups of people that correspond to the matrix in [1].
8. There are no disease induced deaths.
9. The length of infectious period is 7 days.
The following model is designed for a population divided into n groups. This division is with respect to age. The dynamics of the system can be described by the following set of equations: with i = 1 . . . n. Xi denotes a person in age group i. S denotes susceptible, I infectious, R recovered and N is the total population. Observe that if we "stretch" the contact matrix in Table S1 to match susceptibility data, then n = 46 and the correspondence between indexing of groups and the actual age is i = 1 → 6 − 12 month, i = 2 → 1 year, i = 3 → 2 years,. . . i = 46 → 45 years. Finally, note that, we divide each entry of Table S1 by 5 to obtain the contact rates between the individuals in 46 groups. If, on the other hand, the susceptibility data is "compressed", then n = 10, where i = 1 → 0 − 4 years, . . . i = 9 → 40 − 44 years, i = 10 → 45 years.
The initial contact rate matrix for Netherlands is given by Table S1. We "stretch" Table S1 to obtain Table S2. (Note that this is a very rough approximation of the contact matrix over five age groups in each category of Table S1. The meaning and value of the rest of parameters is as follows: • γ is the rate of leaving infectious compartment, where γ = 1/7 days. • βij = p × cij , where p the infectivity and c = [cij ] is the contact rate between people in group i with people in group j. We use Netherlands data for contacts. We calculate p as follows. If we assume that the contact number between each group is the same average number c and therefore β = c×p and R0 = β . Therefore, p = 18γ . We find the average amount of contacts a γ c person in each age group makes per day. Then, the average contact rate is a weighted average (by the age group size). To calculate an average number of contacts we only consider these groups that according to PHO data have susceptible people in them, i.e. people younger than age of 46 years and not including babies aged 0-6 month. The contact data for Netherlands is given by

"Stretched Contact Matrix Approach"
We have seeded the infection one compartment at a time, using an increasing number of index cases. Table S3 summarizes our results for the model.

"Clustered Contact Matrix Approach"
As already mentioned in the model description, we have grouped the age classes into 10 different groups.
As before, we have seeded the infection one compartment at a time, with a different number of infected individuals. Table S4 lists the final size of the epidemic as the number of initial infecteds, and the age of the initial infecteds varies.  Table S3: Final size of the epidemic given a single infected in each age group. The final size ranges from 1 to 5 cases. If more infecteds are introduced into the population, the final size will grow. Table S4: Variations in susceptible population as the placement of 1.5 index cases changes. The final size ranges from 1 to 5 cases of measles. The final size ranges from 1.72 cases to almost 7 cases when only 1 initial infected is introduced into the population. It ranges from 7 to 33 cases if 5 infected individuals are introduced into the population.

Conclusions
The age group location of the initial case, and the number of initial cases affect the final size (as expected). In the models that we considered, initial infecteds aged 5-9 resulted in the largest epidemic final size. If one initial infected aged 5-9 years old, the expected final size of the epidemic is almost 7 cases. An expected final size of 13-14 cases requires two initial infected individuals aged 5-9, that are independent of each other. A stochastic model, if developed, would provide the variability around these expected values.
A more refined contact matrix, and one that is applicable to the Canadian population is needed to provide better estimates.