Herpes zoster, a common infectious disease related with varicella-herpes zoster virus, especially affects elderly people  -  . It causes a painful, blistering rash and can lead to postherpetic neuralgia, a persistent painful complication. However, in Japan, the Law Related to the Prevention of Infectious Diseases and Medical Care for Patients of Infections (The Infectious Diseases Control Law) has not included this disease in its surveillance. The official and nationwide epidemiology of herpes zoster therefore remains unclear. Some few local, small, but long-term studies have examined herpes zoster. They provide some valuable evidence related to herpes zoster  .
Recently, in Japan, all electronic medical claims nationwide (National Database of Electronic Medical Claims (NDBEMC)) have been disclosed as “Data of Medical Claims and Health Check-Ups for Metabolic Syndrome” from the Ministry of Health, Labor and Welfare. It covered 98.4% of all medical claims in 2015  . All doctors must record a diagnosis on medical claim. Therefore, it must necessarily constitute the most reliable data source. This paper presents examination of the epidemiology of herpes zoster using NEBEMC.
In October 2014, routine immunization for varicella was introduced in Japan. Earlier studies predicted and subsequently found that immunization for varicella raises the prevalence of herpes zoster because of diminished opportunities to boost the immunity of elderly people      . However, other studies found no significant effect    . One mathematical model implied that its effect depends on the situation. Some cases showed no effect  . Unfortunately, such an effect has been confirmed in Japan yet to date. Therefore, this study was conducted to estimate its effect on the epidemic of herpes zoster in the first half of the year using NDBEMC.
2. Materials and Methods
2.1. Data and Study Period
From electronic medical claims data, NDBEMC can count patients who have been diagnosed as having the herpes zoster, excluding suspected cases. In May 2015, the coverage of electronic medical claims was approximately 98.4% of all medical claims. NDBEMC data were those of April 2009 to March 2015. However, in 2009 and 2010, the coverage was, respectively only 52.1% and 81.4%. Conversely, in April 2011, it increased to 93.1%  . Therefore, we omit the low coverage period until March 2011. Then we adjust the number of patients by the coverage rate.
It is noteworthy that the number of patients in each month in NDBEMC should be regarded as the prevalence rate, not as the incidence, because the symptoms caused by herpes zoster typically continue for more than one month. For that reason, the same patients might appear in NDBEMC over several months. Herpes zoster is a common disease among elderly people. Therefore, we particularly focused 65 years old or older persons.
2.2. Statistical Analysis
We estimate the simple equation as
where, rt represents the prevalence rate of herpes zoster per 1000 population for people older than 65 years old divided by the coverage rate of NDBEMC, dt is a dummy variable for the initiation of the routine immunization for varicella: it is one from October 2014 and otherwise zero. Also, mit is a dummy variable that represents the month.
Moreover, we add a linear trend to explanatory variables in Equation (1) as
where Tt represents a linear time trend.
2.3. Ethical Considerations
This study used only anonymous data that had been de-linked from individual patient information. Therefore, ethical issues related to medical institutions and pharmacies are unrelated to this study. The use of NDBEMC data by MK for this study was approved by the Ministry of Health, Labor and Welfare of Japan on July 27, 2016 (Research project: Estimation of the number of patients of infectious diseases).
3. Results and Discussion
Figure 1 showed the prevalence of herpes zoster, underscoring the seasonality clearly: it is low in January and February and high in September and October. It also shows an apparently upward trend before introduction of routine immunization of varicella for children. The average of the prevalence was 6.621 per 1000 elderly people. The estimation result shows that the estimated coefficient for the
Figure 1. Prevalence of herpes zoster zoster in patients over 65 years old per 1000 population, divided by the NDBEMC coverage rate. Note: These numbers represent the total number of herpes zoster patients in NDBEMC, divided by the elderly population multiplied by 1000, and divided by the coverage rate of NDBEMC. Vertical line indicates October, 2014 when introduction of routine immunization for varicella in children.
initiation of the routine immunization for varicella was 0.5157. Its p-value was 0.001. The coefficient of determination was 0.6195.
If we add the time trend as an explanatory variable into the estimation equation, then the estimated coefficient of linear trend was 0.023. Its p-value was less than 0.0004. However, the estimated coefficient for the initiation of the routine immunization for varicella was −0.039. Its p-value was 0.384. The fitness is quite high. The coefficient of determination was 0.9654.
An earlier study conducted in Japan  found the incidence rate of herpes zoster in people older than 60 years old as around 7 - 8 per 1000 population in 1997-2011 in Miyazaki prefecture. The prevalence rate found in our study, 6.621, was comparable, but it was lower than the incidence in the US, which was 11.12 - 15.1 for elderly people older than 60 - 65 years old during 1992-2010   . In the US, routine varicella immunization for children had already been introduced in 1996. For that reason, those incidences might have been affected. If we adopt the varicella immunization effect on herpes zoster incidence in the US as 39%, referring to results from an earlier study  , and adjust those numbers, then the incidence of herpes zoster before the initiation of varicella immunization for children is expected to be 8.1 - 10.8. Despite such an adjustment, the incidence of herpes zoster in Japan by NDBEMC is apparently lower than that in the US.
Results show a higher value by 0.5157 after the introduction of routine immunization for varicella than that of the period before introduction. Its magnitude was 7.8% higher than in the before period, evaluated at its average. Therefore, we can infer that routine immunization for varicella raises the prevalence of herpes zoster significantly.
A similar earlier study  estimated its effect as 39%. Our result is apparently much smaller. This discrepancy might result from the evaluation timing for initiation. We set it earlier and with a shorter duration. The effect might be larger than ours if we extend the study period.
However, as shown in Figure 1, the prevalence has an apparently upward trend even in the period before. Incidentally, if we add the trend into the estimation equation as an explanatory variable, then the dummy variable for the routine immunization for varicella is not significant anymore. Therefore, our obtained significant effect of routine immunization of herpes zoster on herpes zoster prevalence without a trend was presumed to reflect such an upward trend.
The potential reason for this upward trend was the decline of contact between elderly people and children who sometimes are infected with varicella in a house or community. If routine immunization of varicella raises the prevalence of herpes zoster, then such a decline of contact also raises the prevalence of herpes zoster. In other words, this decline of contact might dominate the effect of initiation of routine immunization for varicella.
Though it may be out of our concern, we also see the situation of varicella incidence in the same period. It was shown in Figure 2. Apparently, the curve has
Figure 2. Incidence rate of varicella in patients per 1000 population, divided by the NDBEMC coverage rate. Note: These numbers represent the total number of herpes zoster patients in NDBEMC, divided by the population multiplied by 1000, and divided by the coverage rate of NDBEMC. Vertical line indicates October, 2014 when introduction of routine immunization for varicella in children.
downward trend before the introduction of the routine immunization. It might reflect the declining in birth rate. Even though the reason of such a downward trend in varicella incidence was not clear, it should associate with the upward trend in herpes zoster prevalence irrelevantly with the introduction of the routine immunization.
We examined the effect of routine immunization of varicella on the prevalence of herpes zoster in the first half of the year using the most reliable data source of illness. Although a significant effect of routine immunization of varicella was found, the upward trend started before initiation might dominate the effect. We must monitor a longer period to distinguish the effects of routine immunization of varicella other than an upward trend.
We gratefully acknowledge the Ministry of Health, Labor and Welfare for granting us permission to use NDBEMC.