Study population, setting, and design

We conducted a retrospective population-based cohort study of Ontario residents aged ≥65 years between April 1, 2002-August 31, 2016. These individuals had universal access to physician services, hospital care, and prescription drug coverage. We selected the start date to align with the introduction of International Classification of Diseases version-10 (ICD-10) codes in Ontario and the end date to coincide with the introduction of the provincial publicly-funded HZ vaccination program for those aged 65–70 years. Our outcome was annual incidence (organized by April to March fiscal year [FY]) of HZ in outpatient, ambulatory, and hospital settings. We obtained ethics approval from Public Health Ontario’s Ethics Research Board.

Data sources and definitions

We used Ontario’s administrative health databases, which are securely linked using unique encoded identifiers and analyzed at ICES (www.ices.on.ca). Use of these databases is authorized under section 45 of Ontario’s Personal Health Information Protection Act, which does not require patient consent for health system evaluation and improvement. All data were fully anonymized before we accessed them.

Statistical analysis

We calculated the annual incidence of HZ per 10,000 older adults (i.e. 65 years and older) in Ontario by age group, sex, income quintile, frailty status, immunocompromised status, as well as age group stratified by frailty status. We calculated these incidences overall as well as by setting, providing separate results for outpatient as well as hospital-attended (hospital or ED) settings. We also calculated the incidence of HZ with or without acute complications. In our restricted cohort of individuals who had 365 days of follow-up, we calculated the percentage of HZ cases who developed PHN. In a sensitivity analysis, we repeated our analysis in those with 180 days follow-up to determine the proportion that developed PHN. We compared the incidence of HZ and PHN by income quintile in the periods before and after vaccine availability through private purchase (i.e., before and after September 1, 2009). We used negative binomial regression to derive relative risks (RR) and 95% confidence intervals (CI), adjusting for age, sex, and year.

Clinical factors

We estimated the incidence of HZ, acute complications, and PHN in several subgroups, aligning with an increasingly promoted life-course approach to vaccination, in which circumstances beyond age are necessary considerations in making vaccination decisions and in which vaccination to prevent disease is seen as a component in promoting healthy ageing [28, 29]. We found that the incidence of HZ in immunocompromised adults across all settings was twice that of those who were immunocompetent. This is in line with other published estimates but lower than in our previous published work which was limited to hospital-attended HZ [30, 31]; in this publication, we had hypothesized that there may be a greater difference when limited to more severe disease. We have now provided estimates for all medically attended HZ in those ≥65 years by immunocompromised status, which has important implications given that the vaccines available to prevent HZ have different contraindications.

We are not aware of any studies that have provided an estimate of risk of HZ and PHN in adults by frailty status, although it has been suggested that frail older adults may be disproportionately affected [9, 10]. Other research has examined the development of frailty following HZ infection (finding no effect) [32, 33], whereas we assessed the incidence of HZ within cohorts stratified by frailty status and found a higher incidence compared to older adults who are not frail. While aging correlates with frailty, the concept of frailty is distinct and may serve as a better proxy to immunosenescence [34, 35]. This has implications for vaccine promotion given the increase in susceptibility to infectious diseases as the immune response weakens and a goal to target vaccines to those most at risk, beyond considering age alone, but also may have implications for immune response to vaccine [10, 11, 29]. We compared the incidence of HZ and PHN by frailty within age groups in order to better understand the relative contribution of each factor. For both outcomes, we observed a higher incidence in frail older adults within each age group, but that the relative impact of frailty decreased with increasing age. For example, the incidence of PHN in frail 65–69 year old adults was almost twice that of non-frail adults of the same age group. Conversely, in those ≥80 years, the incidence of PHN was equal by frailty status. As Russell et al. noted, the effect of comorbidity was more pronounced in younger ages [4].

Socio-economic factors

In addition to clinical factors, we were interested in exploring disease incidence by SES in an effort to understand equity [36]. There is conflicting evidence regarding SES (i.e. income, deprivation) and HZ incidence, with some reporting no association, some reporting an increased risk in more deprived individuals and others reporting an inverse association [2, 37–39]. In comparison to lower income quintiles, we found similar or greater risk of HZ in higher income quintiles in the pre-vaccine era, but equal or lower risk of HZ in higher income quintiles in the private purchase era. We observed a larger effect when we repeated this analysis with PHN as an outcome. The risk of PHN was significantly lower in the pre-vaccine era in the highest income quintile, but the magnitude of difference between highest and lowest quintiles increased in the private-vaccine era.

There are several reasons for why we may have observed this effect in the two eras. First, only LZV was available in Canada during the study period and its short-term efficacy against PHN has been shown to be higher than efficacy against HZ incidence [25, 38]; as such, the reduced incidence of PHN in particular may suggest that those in the highest income quintile were more likely to purchase vaccine out of pocket. Second, our capture of PHN included a combination of diagnostic codes and prescription data, with prescriptions being limited to those covered by the provincial drug plan. If those in a higher income quintile were being treated with medication not on the formulary, they would not have been identified with PHN and this could lead to a discrepancy in PHN capture between income quintiles. However, other studies have previously reported higher incidence of PHN in low SES prior to the availability of a vaccine [7], and given that we observed a further reduction in PHN in the private vaccine era, these results suggest PHN is lower in those with higher SES. Last, the likelihood of developing PHN may have been impacted by timely access to antivirals or supplemental therapies [40, 41], which could be influenced by SES. Although our PHN analyses were restricted to those with HZ and the majority of individuals were included using criteria that included an antiviral prescription (i.e., outpatients), we did not have data on other medications prescribed or those not publicly-funded.

We are unaware of other studies that have used this approach to identify and understand disparities. In Australia and the UK, HZ vaccine is publicly-funded for older adults. In Canada, Ontario is currently the only province to publicly fund HZ vaccination and does so using LZV for those 65–70 years; however, this program started after our study period [42]. While not comparing disease incidence directly by SES, several studies have reported discrepancies in vaccine uptake by sociodemographic characteristics. In Alberta, Liu et al. estimated vaccine uptake in those ≥60 years for the first four years following LZV licensure and found the odds of vaccination were 57% higher in those in the highest quintile relative to the lowest (OR = 1.57 [95% CI 1.45–1.69]) [43]. In the UK, Ward et al. estimated coverage in 2014–15 as 8.2% lower in the most deprived compared to the least deprived [44]. Also in the UK, Jain et al. reported a lower odds of vaccination in the most deprived compared to the least deprived (OR = 0.69 [95% CI 0.64–0.75]) [45]. Of note, both the UK studies were conducted in an era of public funding, indicating issues related to access may persist beyond cost alone.

Limitations and strengths

There are several limitations worth noting. First, our study was restricted to those who were ≥65 years. We were concerned that outpatient visits for HZ vaccination could be misclassified in our data as cases of HZ and we therefore added the criteria for an antiviral prescription in conjunction with a coded HZ visit in order to define a case of HZ. However, these prescription data are only available in Ontario for those ≥65 years and we were thus unable to explore incidence in younger age groups. We were also unable to assess those who received prescriptions related to HZ or PHN through private insurance (i.e., prescriptions not be covered by ODB). Further, due to the lack of HZ vaccine-specific billing codes, we could not identify whether an individual had received HZ vaccine. We also did not account for the introduction of varicella vaccine in our analyses, which is a hypothesized contributor to the changing epidemiology of HZ in adults given the reduction in opportunities for adults to be “boosted” due to lower immunosurveillance [46]; however, results supporting this hypothesis have been conflicting [4, 5, 47–50]. There also may be issues due to misclassification in our data. These include: missing some adults aged 65 who were on prior immunosuppressive therapy because the drug data is only available for those ≥65 years; misclassifying annual immunocompromising status as we defined it based on October 1 each year; and misclassifying ongoing PHN or recurrent zoster as a new case as our washout period was limited to one year.

Our study provides updated burden of disease data for adults ≥65 years in Ontario, and may be a helpful basis for evaluating future immunization programs. We estimated population-based incidence overall and in groups at increased risk of HZ during adulthood in an effort to apply a life-course approach to immunization that incorporates not just age group as an important component of immunization programs but also recognizes that providing protection should take into account other circumstances, such as frailty and immunocompromised [28, 29]. We also provided data on complications, including PHN, which may reduce quality of life and the impacts of which should be taken into consideration when discussing immunization as it relates to healthy ageing.

Conclusions

Our results suggest that additional prevention efforts should be targeted toward adults who are immunocompromised, frail, and those living in lower income neighbourhoods. Future work should assess the impact of the HZ vaccine program with a particular focus on equity in the publicly-funded era.