A questionnaire survey was carried out to all municipal authorities operating pneumococcal vaccination programs in 2008. At the time of the survey, 63 municipalities operated programs and the response rate was 100%. In the survey, following questions were asked: operating years of the program, definition and size of the target population, uptake population, price to get one vaccine shot, number of health facilities providing shots, and PR tools taken for promoting vaccination. The type of PR tools included:
To health nurses & care givers
“Distribution of brochures to their health facilities (PR11)”, “Hold information events (PR12)”, “Communication in regular meetings (PR13)”
These questions were asked on a Yes/No basis. With this municipal based data, we assume each municipality as a market for vaccination and examine the effects of PR tools implemented by the municipality and subsidy supported shot price on the vaccine uptake rate.
Descriptive statistics are reported in Table
1. Due to missing or inadequacy values for data analysis, 39 out of the 63 municipalities were chosen for the estimation. Inadequacy values included the price of a shot if it was changed many times during the operating years of the program, or if the municipality had set fixed subsidies to health care providers and had them decide their own retail shot prices that reflected discretional technical service fees. The “Uptake rate” is the uptake population divided by the target population for vaccination during the operating years of the program. A municipality with 0.991 in the maximum shows that vaccination program had run successfully, meaning that 99.1% of the target population had been vaccinated. The “Shot price” is individual cost burden for receiving one shot. The 0 in the minimum means that the individual shot price is fully covered by the subsidy, and on the other hand, 6.500 in the maximum implies that the price is not fully covered and an individual has to pay 6,500 Japanese yen (JPY) (62.5 USD; 1USD = 104JPY in 2008 annual average) for vaccination. The “Operating years” is the length of operating the program. Note the maximum is 5 years, meaning that the data taken is the first 5 years from the beginning of the vaccine program, therefore there is no individual who has taken the shot twice. This is because effectiveness of PPV shot is said to last for 5 years [
1] and revaccination was prohibited in Japan at the time of the survey. That is to say, the demand for vaccination would be decided by the shot price, PR, etc., and not by some sentiment from previous shot experience. The “Target population” is defined by the age criteria: ≥ 65, ≥ 70, or ≥ 75 years old, depending on the municipality. The difference of 0.217 in the minimum and 44.353 in the maximum is not because of age criteria chosen by municipalities, such as ≥ 75 in the minimum and ≥ 65 in the maximum, but of the population size: one with 0.217 shows a small village and the other with 44.353 shows a big city. The “Location intensity” is the number of health facilities providing vaccination per
km
2 in the municipality. The “Income” is the average income per capita by municipality obtained from System of Social and Demographic Statistics by Statistics Bureau [
17]. The demand function requires a budget constraint information such as income [
18], however, we cannot define the value from our municipal based questionnaire survey data. And there was no available income data on the targeted elderly but only of the entire citizen in the municipality. The “PR1 to 13” are binary or dummy variables which take 1 if the municipality implements the PR tool and take 0 otherwise. 79.5% of the municipalities take PR4, while only 7.7% take PR9 and PR12. “To target population”, “To physicians”, “To nurses & care givers” and “Total” show the number of PR tools implemented by the municipality during the operating years of the program; PR1 to PR7, PR8 to PR10, PR11 to PR13 and PR1 to PR13, respectively. 3.667 in the mean of “To target population” describes that municipalities implement on the average of 3 or 4 PR tools among PR1 to PR7.
Table 1
Descriptive statistics
Uptake rate | | 0.268 | 0.007 | 0.991 | 0.223 |
Shot price (1,000 JPY) | | 3.556 | 0.000 | 6.500 | 1.310 |
Operating years (year) | | 2.128 | 1.000 | 5.000 | 1.341 |
Target population (1,000 persons) | | 3.625 | 0.217 | 44.353 | 7.139 |
Location intensity (per km2) | | 0.474 | 0.002 | 9.966 | 1.599 |
Income (1,000,000 JPY) | | 1.233 | 0.564 | 4.984 | 0.732 |
Implemented PR tools: | | | | | |
To target population (Count PR1-7) | | 3.667 | 1.000 | 7.000 | 1.284 |
PR1 | 0.564 | | | | |
PR2 | 0.718 | | | | |
PR3 | 0.538 | | | | |
PR4 | 0.795 | | | | |
PR5 | 0.513 | | | | |
PR6 | 0.205 | | | | |
PR7 | 0.333 | | | | |
To physicians (Count PR8-10) | | 0.897 | 0.000 | 3.000 | 0.968 |
PR8 | 0.487 | | | | |
PR9 | 0.077 | | | | |
PR10 | 0.333 | | | | |
To nurses & care givers (Count PR11-13) | | 0.385 | 0.000 | 2.000 | 0.747 |
PR11 | 0.179 | | | | |
PR12 | 0.077 | | | | |
PR13 | 0.128 | | | | |
Total (Count PR1-13) | | 4.949 | 1.000 | 11.000 | 2.176 |
The “Uptake rate” is the response variable in our demand function, and we have “Shot price” and “Income” as explanatory variables [
18]. As considered to affect the response variable, “Operating years”, “Target population”, “Location intensity” and “PR1 to 13”; “To target population”, “To physicians”, and “To nurses & care givers”, and “Total” are also included into the demand function. If the program has been operating for a long period of time, it would have become more common and high uptake rate is expected, in regards to “Operating years”. The “Target population” is regarded as nature of works which municipal authorities have to undertake to promote vaccination. It would be an extreme example, but making an effort to increase the uptake rate for a target population of one person is much easier than for a target population of 10,000 persons. The “Location intensity” is regarded as the non-cash price such as travel or time cost which has been proven to be significant in the demand for health care [
19] including vaccination [
20]. Investigating the effects of PR tools on the uptake rate is the aim of this study, therefore, we include “PR1 to 13” as dummy variables to examine which PR tools are effective to increase the uptake rate. In addition, the number of PR tools implemented by the municipality is considered as another PR scale, because it may reflect the intensity of PR within the municipality.
We assume a linier demand function, then the estimation form is expressed as [
21]:
(1)
where Y is the response variable “Uptake rate” in this study, are explanatory variables, are the constant and coefficients, ε is the error term, and n is the sample size. Three models are estimated: “Shot price”, “Operating years”, “Target population”, “Location intensity”, “Income” in all three models; “Total” is added in model 1; “To target population”, “To physicians” and “To nurses & care givers” are added in model 2; and “PR1 to 13” are added in model 3.
In order to examine the effects of explanatory variables, especially the effects of each PR tool on the “Uptake rate”, the best model is selected by c-AIC from model 3 as a full model, with a restriction to keep “Shot price” and “Income”. This is because “Shot price” and “Income” are conventional variables in the theory of demand [
18]. Then c-AIC is calculated for all possible regressions with combinations of 16 explanatory variables. ‘c-AIC’ is the Akaike Information Criterion (AIC) for small sample data: the smaller the value of c-AIC, the better the model [
22]. Sugiura (1987) suggested that AIC may perform poorly if there are too many parameters in relation to the size of the sample [
23]. Our sample size is 39 and small, while parameters are 20, quite large, in model 3. The best model is created as model 4.
The price elasticity calculated from the estimation results of model 1, model 2, model 3, or model 4 is written as [
18]:
(2)
where
x
1 is the “Shot price”,
is the estimated coefficient of the “Shot price”, and
is the expected value of the “Uptake rate” at
x
1. From equation (
1), model 1, model 2, model 3, or model 4 is called as the varying elasticity model since the price elasticity can vary in response to changes in
x
1 and
[
24]. Meanwhile, the constant elasticity model takes log(
x
1) and log(
Y), instead of taking
x
1 and
Y in the varying elasticity model [
24]. That is:
(3)
Then the constant price elasticity is expressed as:
(4)
We focus to the best model, which is model 4, and estimate the varying and constant price elasticities, using a command “margins” in the software STATA12 [
25]. The
is calculated with all explanatory variables except
x
1, fixed at their means by using a command “atmeans”. Model 5 is created as the constant elasticity version of model 4 for reference.
Regarding research ethics, this study is not an experimental research nor carried on humans. It also falls outside of the guidelines of health research ethics in Japan. The data used for this study is openly available, and we received permission to use this data by all respondents’ municipalities.