The study included 2 845 353 Medicare beneficiaries from 1992 through 2010, contributing a total of 22 508 343 person-years with a median of 7 years per beneficiary (Table 1). Age and sex distributions were similar to the entire Medicare population but with fewer beneficiaries in western states and fewer Hispanic and Asian beneficiaries (data not shown). The mean age was 76.1 years in 1992 and 77.1 years in 2010.

A total of 281 317 incident cases of HZ occurred. Crude HZ incidence increased 40% from 10.2 per 1000 person-years in 1992 to 14.3 per 1000 persons in 2010; age- and sex-standardized HZ incidence increased 39% from 10.0 per 1000 person-years in 1992 to 13.9 per 1000 person-years in 2010 (Figure). Mean age at the time of HZ diagnosis remained stable at 76.9 years in 1992 and 77.9 years in 2010. The exponentiated parameter estimate for the interaction between calendar year and the indicator variable for the early varicella vaccine implementation period was 1.002 (95% CI, 0.984 to 1.020), indicating that the annual rate of increase in HZ incidence during this period (1996 to 1999) was 0.2% higher than that during the preimplementation period (1992 to 1995) but was not statistically significant. Similarly, the exponentiated parameter estimate for the interaction between calendar year and the indicator variable for full implementation period (2000 to 2010) was 1.008 (CI, 0.994 to 1.023).

Increases occurred among all age groups (Figure), racial or ethnic groups, and census divisions (data not shown) and both sexes; they remained apparent even in the selected 30% of beneficiaries without a potentially immunocompromising condition (Appendix Figure 1, available at www.annals.org). In addition, increases in HZ incidence were greater than those seen in 9 out of 10 analogous conditions that we investigated (Appendix Figure 2, available at www.annals.org). Table 2 shows data on the incidence of HZ among Medicare beneficiaries in 2010 to highlight the incidence of HZ in key subgroups.

We evaluated HZ incidence by sex, age group, and race or ethnicity (Table 3) before introduction of the universal varicella vaccination program (1992 to 1996). The RR for women was 1.21 (CI, 1.19 to 1.24) and increased for each 5-year age group up to the 80- to 84-year group. Compared with white persons, black persons (RR, 0.51 [CI, 0.48 to 0.53]) and Hispanic persons (RR, 0.76 [CI, 0.72 to 0.81]) had lower HZ incidence.

In the GEE model adjusted for sex, age, and calendar year, HZ incidence did not vary by state varicella vaccination coverage (RR, 0.9998 [CI, 0.9993 to 1.0003]). Varicella vaccination coverage lagged by 10 years was also not significant (RR, 1.0010 [CI, 0.9997 to 1.0022]). Incidence trends of HZ in states with consistently high varicella vaccination coverage from 1997 through 2010 (mean coverage, 82%) compared with states with low varicella vaccination coverage (mean coverage, 64%) overlapped with no pattern emerging (data not shown).

Varicella exposure has been postulated to boost VZV-specific immunity and reduce the risk for VZV reactivation. Concern has been expressed that routine childhood varicella vaccination, introduced in the United States in 1996, could thereby lead to an increase in HZ incidence by reducing opportunities for exposure to varicella. Our findings suggest that, although HZ incidence has increased in elderly persons, routine varicella vaccination has not influenced this increase.

First, the rise in HZ incidence clearly predates 1996 and HZ incidence did not accelerate after implementation of the varicella vaccination program, when varicella vaccination coverage reached 90% and varicella incidence decreased by 90% (30, 31). Second, state varicella vaccination coverage had no effect on HZ incidence concurrently or 10 years later. Finally, the mean age at the time of HZ diagnosis did not decrease during the study, which might occur if external boosting helped to maintain protection. Thus, although HZ incidence has increased in the elderly population, we did not find evidence to suggest that this increase has been influenced by the varicella vaccination program in the United States. This finding confirms the results of other studies before and after introduction of the varicella vaccine (19, 32–34).

The lack of evidence for an effect of the varicella vaccination program on HZ incidence might be explained in several ways. First, although some studies suggest that exposure to varicella protects against HZ (9–11), other findings contradict this (12, 13). Because latent VZV can reactivate subclinically and boost immunity (35–37), compensatory boosting might occur among persons who are not periodically exposed to varicella (8). Alternatively, external boosting might require a degree of varicella exposure that is uncommon among elderly persons, particularly because this population has less-frequent exposure to children than do younger adults (38).

Age- and sex-standardized HZ incidence among elderly persons increased 39% from 10.0 per 1000 person-years to 13.9 per 1000 person-years during the 19-year study. On the basis of our observed values, we calculate that 14% of men and 20% of women who reach age 65 years will have HZ during their remaining lifetime (39). The observed increase is robust, occurring among all age, sex, and race or ethnicity subgroups and all geographic regions. It does not seem to be an artifact of changing prevalence of immunocompromising conditions, because increases in age- and sex-standardized HZ incidence are seen even after restricting the analysis to a selected cohort of immunocompetent persons. Furthermore, increases in HZ incidence persisted when it was compared with 9 of 10 other acute outpatient conditions, suggesting that these increases were not due to underlying changes in health-seeking behavior or in the Medicare program over time.

Increases in HZ incidence in the elderly population have been reported in several studies in countries with and without varicella vaccination programs (3, 16, 17, 19, 20, 40, 41); however, other studies have not shown an increase (32, 33). We have no explanation for the trend and expect that it awaits a better understanding of why a substantial minority of persons have HZ during their lifetime, whereas most do not. In particular, chronic diseases prevalent among elderly persons do not explain most cases of HZ (42).

Data from the large, generalizable Medicare population provide an opportunity to more precisely characterize the well-recognized effect of age on HZ incidence. Although the association between HZ incidence and age seems to plateau after age 80 years, the risk for postherpetic neuralgia, hospitalization, and other complications continues to increase with age. We also provide precise data to confirm that older women are at greater risk for HZ than older men; most studies report increased HZ incidence among women in most age cohorts, including children, although the reason remains unknown (3, 17, 18, 20, 41, 43–47). Finally, we provide data that black and Hispanic persons are at lower risk for HZ than white persons. A few reports are available on the reduced risk for HZ among black persons (43, 45, 48), although data about HZ incidence in Hispanic persons are lacking (13). The RR for American Indians or Alaska Natives should be interpreted with caution because health services provided by the Indian Health Service were not fully captured in the Medicare data (49).

We followed elderly Medicare beneficiaries who are at high risk for HZ for a long period spanning the licensure of varicella vaccination. In 2006, the U.S. Advisory Committee on Immunization Practices recommended the HZ vaccine for adults aged 60 years or older for prevention of HZ and its sequelae (21). Uptake of the HZ vaccine has been slow, with only 1.9% coverage in adults aged 60 years or older in 2007 (50), increasing to 14.4% in 2010 (22).

Although we do not expect that such low coverage would substantially affect HZ incidence on a population level, these data provide a baseline for monitoring effects of the HZ vaccine program as coverage increases. A previous study of HZ incidence in all ages from 1993 to 2006 using MarketScan (Truven Health Analytics, Ann Arbor, Michigan), a database of health care claims from commercial insurers, had limitations in evaluating trends in adults aged 65 years or older because the size of the study population in this age group changed substantially over the observation period, allowing opportunities for unmeasured confounding (19). MarketScan did not incorporate Medicare data until 1996 and even then contained health care claims for only the selected subset of Medicare beneficiaries with managed care. The Centers for Medicare & Medicaid Services Chronic Conditions Data Warehouse used in this study contains claims for a stable and more representative population of beneficiaries in traditional fee-for-service Medicare during the study period.

The use of administrative data has limitations. The accuracy of findings from studies based on health care claims depends on high and consistent levels of health-seeking behavior and access and on accurate disease coding. These criteria are probably met for most cases of HZ because it is painful and the study population had Medicare Parts A and B coverage for health care costs; however, mild cases may not come to medical attention, resulting in an underestimate of true incidence of HZ. In a population-based survey of 141 adults aged 60 years or older who had had HZ, 95% sought medical attention (50).

Differences in health-seeking behavior or access may explain a portion of the variation in risk by sex or race or ethnicity. We could not directly assess the accuracy of HZ claims in our data through record reviews. However, previous studies of HZ have shown good concordance between coding and actual health care visits, with positive predictive values ranging from 85% to 100% (12, 18, 51). Furthermore, estimates from self-reports and health care settings of HZ incidence have been similar to ours, providing reassurance about the sensitivity of HZ coding (17).

Our study was also limited by the data available for analysis. Although more than one half of cases of HZ in the United States occur in persons younger than 65 years and therefore were missed by our Medicare analysis, HZ tends to be much more severe in older adults; thus, the Medicare population is of particular importance (3). We showed that HZ incidence was not affected by state-level variation in varicella vaccination coverage controlled for sex and age but were not able to consider other factors that vary across states and might affect HZ incidence, such as age at parenthood, family size, or living with young children.

Age-specific HZ incidence has been increasing in the U.S. population older than 65 years even before implementation of the childhood varicella vaccination program. Introduction and widespread use of this vaccine did not seem to affect this increase. This information is reassuring for countries considering universal varicella vaccination.

Our findings have several implications. First, in the absence of explanations for increasing HZ incidence, properly monitoring the effect of the HZ vaccination program or projecting future HZ incidence will be difficult. Furthermore, HZ poses a substantial burden of disease in the elderly population; continued increases in HZ incidence would be worrisome and need to be understood. Fortunately, an effective vaccine is now available that can help prevent this disabling disease.