A cross-sectional population-based study was conducted via a retrospective database review, which provided a large sample size to assess disease burden. The adopted study design allowed for the estimation of actual healthcare utilization since patients were managed according to routine clinical practice in a real-world setting. In addition, the use of a database provided the large amount of data required to identify patterns and potential associations between HZ and various factors.

Data was sourced from the SingHealth-IHiS Electronic Health Intelligence System (eHints), a repository that integrates information from administrative, clinical, and ancillary healthcare transactional systems. eHints was established in 2009 and contains data from the SingHealth cluster, one of three public healthcare clusters in Singapore, that covers the Eastern part of the country. All Singapore residents have the option to visit any public primary care clinic regardless of their residential address. These clinics offer subsidized care to citizens and permanent residents, with higher subsidies for children and the older adults. Data was collected from all eight public primary care clinics within the SingHealth cluster.

Data on demographic characteristics, clinical diagnoses, prescriptions, clinic attendances, and referrals were extracted. Clinical diagnoses were coded using the ICD-10 classification. Clinic attendances occurred exclusively in the form of in-person visits. Telephone and home visits were not available at public primary care clinics during the study period. All referrals were made electronically. From the “Clinical Document (Referrals Only)” section of the eHints database, the following fields were extracted: visit number, referral destination, and referral indication. Medical certificates were issued electronically. Data regarding the visit number and duration of medical leave given each visit was extracted from the “Medical Certificate” section of the eHints database.

A trusted third party was appointed by the healthcare cluster to perform data de-identification. The de-identified data was passed to the study team via secure file transfer protocol. As secondary data without personal identifiers was used, informed consent was not required, and the study was exempted from approval by the SingHealth Centralised Institutional Review Board (Reference number: 2021/2709).

The de-identified data was checked for missing and out-of-range values, duplication as well as mislabeling prior to merging. Six individuals were excluded for missing values and another two individuals were excluded for outlying values. After the data was cleaned and merged, further checks on data consistency were conducted.

The study population comprised of all individuals, aged 21 years and above, who attended a cluster of eight public primary care clinics in eastern Singapore from 1 January 2018 to 31 December 2020. Cases were defined as individuals who had one or more visits with a diagnosis code of HZ between 1 January 2018 to 31 December 2020.

The factors of interest were age, gender, ethnicity, and chronic medical conditions that are prevalent in the primary care population. These factors were studied to identify high-risk groups to facilitate the development of targeted prevention and treatment strategies. Increased age, female gender, and chronic medical conditions have been reported to increase the risk of HZ (24). Compared to Caucasian individuals, African-American individuals were shown to have a lower risk of contracting HZ, while Asian Americans had a similar risk (24). The three major ethnic communities in Singapore were examined—Chinese, Malay, and Indian.

Chronic conditions identified from the literature as potential risk factors included asthma (24–26), chronic kidney disease (27, 28), chronic obstruction pulmonary disease (29, 30), depression (31, 32), diabetes mellitus (33, 34), hyperlipidemia (32), hypertension (35), hypothyroidism (36), ischemia heart disease (37), osteoarthritis (32), and rheumatoid arthritis (38). Including the former conditions, we selected 21 diseases covered under the Chronic Disease Management Programme (CDMP), a healthcare financing scheme that allows the use of compulsory medical savings (Medisave) for outpatient treatment of common chronic conditions. The 21 chronic diseases were anxiety, asthma, benign prostatic hyperplasia (BPH), bipolar disorder, chronic kidney disease (CKD), chronic obstructive pulmonary disease (COPD), dementia, depression, diabetes mellitus, epilepsy, hyperlipidemia, hypertension, ischemic heart disease (IHD), osteoarthritis, osteoporosis, Parkinson’s disease, prediabetes, psoriasis, rheumatoid arthritis, schizophrenia, and stroke. Hypothyroidism, although not part of CDMP, was also selected since it was identified as a potential risk factor from the literature (36). When examining the association between each chronic condition and HZ, cases were included only if there were any visit(s) for each chronic condition that occurred before the earliest recorded HZ visit. A look-back period from 1 January 2015 to 31 December 2020 was used to check for visits with diagnosis codes of the studied chronic conditions. The diagnosis codes used to ascertain HZ and the studied chronic conditions are specified in the Supplementary material.

The yearly prevalence of HZ for years 2018, 2019, and 2020 was calculated as the number of cases that occurred per 100,000 persons who visited the eight public primary care clinics. The calculation was based on the formula for period prevalence, defined as the proportion of people in a population having disease over a period of time (39). Individuals who had visits with a diagnosis code of HZ in 2018, 2019, and/or 2020 would be counted as cases for the respective years.

Given HZ occurs more commonly in older adults (5), the prevalence by age was examined to determine the burden of HZ in older age groups. To determine the number of unique cases across each age group, each case was identified by the index HZ visit, which was the earliest recorded visit with a diagnosis code for HZ. The date of the index visit was considered the date of diagnosis of HZ.

For healthcare utilization, the number of clinic attendances per episode of HZ, referrals made to tertiary care for further management of HZ, medications prescribed for HZ, and days of medical leave taken per episode of HZ were evaluated using frequency distributions. The denominator used for all the above was the total number of cases, except for days of medical leave where the denominator was the number of cases requiring medical leave. For each case, all visits with a diagnostic code of HZ were assumed to belong to a single HZ episode.

In determining associated factors of HZ, the independent variables used were age, gender, ethnicity, and the 22 chronic diseases. The dependent variable was HZ occurrence. Age was analyzed as a dichotomous variable (aged below 50 years, aged 50 years and above). The cut-off of 50 years old was chosen as it is known that HZ risk increases sharply onwards from this age (3). Pearson’s chi-square test of independence was used to test for the presence of associations between the independent variables and HZ. For factors that demonstrated an association with HZ, logistic regression was performed to determine the strength of the association.

The frequencies of each chronic condition were determined among the cases, non-cases, and total population. Chronic conditions with fewer than 10 HZ cases were excluded from the analysis—Parkinson’s disease, epilepsy, bipolar disorder, schizophrenia, rheumatoid arthritis, and psoriasis. Univariate analysis was conducted between each of the remaining conditions and HZ. Factors that showed a statistically significant association with HZ (p < 0.05) were added to a multivariate model. Chronic conditions shown to be associated with HZ occurrence in literature were added to the multivariate model regardless of whether they demonstrated a significant relationship with HZ in univariate analysis. Multivariate logistic regression was used to calculate adjusted odds ratios (aOR) with 95% confidence intervals (95% CI). Checks for multicollinearity were undertaken by assessing variance inflation factors (VIF) for independent variables and performing correlation matrix analysis. VIF for all independent variables was <2.2, and pairwise correlation coefficients were < 0.7, suggesting that multicollinearity was not a concern, using thresholds of 5 and 0.8, respectively (40).

Statistical analysis was performed using SPSS version 28.0 (IBM SPSS Statistics for Windows, Version 28.0. Armonk, NY: IBM Corp.).

The reduction in HZ prevalence in 2020 may be related to a decrease in respiratory infections resulting from Covid-19 control measures and a shift in health-seeking behavior. A Singaporean study reported that mandatory mask-wearing, social distancing, and lockdowns led to a reduced prevalence of multiple respiratory viruses in 2020 (41). Physical stress is a risk factor for HZ and coinfections may contribute to varicella zoster virus (VZV) reactivation (24). Hence, the dip in HZ prevalence could have mirrored that of respiratory infections by reducing reactivation triggers. An English study presented a 27% decrease in HZ cases during the peak pandemic period in 2020 compared to the same period in 2019 (42). The authors postulated that the incidence of non-Covid-19 conditions reduced with pandemic control measures, and health-seeking behavior changed due to fears of contracting Covid-19in healthcare settings.

Similar age-related patterns of HZ prevalence were demonstrated in the literature. A study on the global epidemiology of HZ reported the mean age of diagnosis as 59.4 years (1). A systematic review of Asia-Pacific countries established that HZ incidence peaked between 70 and 79 years, decreasing in those aged 80 years and above (5). While age-specific risks are a key factor in decisions regarding HZ vaccination, other factors such as age-specific vaccine efficacy and complication rates require careful consideration (43).

The number of visits for HZ was low, with most individuals consulting only once. In comparison, the mean number of outpatient visits per HZ episode ranged from 1 to 5.7 in Japan, South Korea, and Thailand (13–15, 18). In Japan, the higher number of visits was explained by differences in the healthcare system and the use of outpatient neurotropin injections for the treatment of post-herpetic neuralgia (15). The low number of visits observed may be explained by early presentation and treatment with oral antivirals. Prompt treatment of HZ with oral acyclovir reduces the time to pain resolution by more than 50% (44), and the time to 50% lesion healing by about 2.5 days (45). With appropriate early treatment, complication rates and the need for further healthcare visits may be reduced.

Similarly, reduction of complications via early effective antiviral therapy could explain the low referral rate in this study. The main reasons for referrals were to evaluate and manage complications, control zoster pain, or closely monitor high-risk patients. However, individuals with HZ may have consulted other healthcare providers for referrals, which would not be captured in the results.

Topical acyclovir was prescribed for almost half of individuals with HZ, but there is a lack of evidence to support this practice. Guidelines on the management of HZ have stated that topical antiviral therapy lacks efficacy in patients with HZ and is not recommended (46). A randomized controlled trial also demonstrated that topical acyclovir did not improve rash progression, pain severity, or the incidence of post-herpetic neuralgia (47). One possible reason for this practice is that patients sought topical acyclovir treatment after learning about it from sources other than their healthcare providers. Physician-prescribing decisions are substantially affected by patient requests (48). Physicians’ lack of knowledge of the limited evidence for topical acyclovir use in HZ may be another contributing factor. Future studies should explore the reasons for, benefits, and harms of this prescribing pattern.

The analgesia prescribed in this study was appropriate, but the use of analgesia combinations can be explored further. First-line treatment options for HZ-related pain are gabapentinoids and TCAs, while opioids and tramadol are second- or third-line options (49). Gabapentin reduced pain severity by 66% compared to 33% with placebo (50). Recent evidence has shown that combining gabapentin with a lidocaine patch has better analgesic effect than gabapentin alone, and a lower incidence of adverse events (51). However, only 2.9% of individuals with HZ were prescribed topical lidocaine. Primary care physicians may consider the use of combination therapy for patients who do not respond adequately to gabapentin alone.

The detection of associations may generate hypotheses for further studies to identify high-risk individuals for HZ vaccination. A preponderance of HZ cases was identified among individuals of Chinese and Malay ethnicities. Little is known about ethnic differences in HZ occurrence, which may be explained by differences in genetic risk and healthseeking behaviors. A Malaysian study showed that Malays were more likely to consult a doctor for health issues than Indians (52).

The relationship demonstrated between asthma and HZ is consistent with other studies. In a meta-analysis, the pooled relative risk (RR) of HZ in a patient with asthma was 1.25 (95% CI 1.13–1.39) (24). Asthma, a Th2-predominant condition, results in reciprocal suppression of Th1 cell-mediated immunity (25), which may lead to VZV reactivation (53). The risk of HZ is even higher if asthmatic individuals are using inhaled or oral corticosteroids (26, 28). Systemic steroid use increases the risk of HZ by 59% (54), reflecting the importance of good asthma control.

Similar findings between hypertension and HZ have been reported. A Japanese study established a hazard ratio of 1.93 (95% CI 1.66–2.26) for HZ among hypertensive patients (35). A database review also demonstrated that hypertension was independently associated with HZ among patients with rheumatoid arthritis (55). T cell activation underlies the pathogenesis of hypertension (56). Although the mechanism underlying the observed association is unclear, it is suspected to be a decline in cell-mediated immunity (35).

Individuals with osteoarthritis had higher odds of HZ. A case–control study revealed a similar association between osteoarthritis and HZ (OR 1.24, 95% CI 1.20–1.28). A prospective study reported a higher incidence of HZ among individuals with non-inflammatory musculoskeletal disorders (57). Cell-mediated immunity is generally impaired in patients with chronic illnesses (58). A study comparing healthy and “almost healthy” persons defined by the lack of exercise or by the use of medications for chronic conditions such as hypertension or osteoarthritis demonstrated lower levels of interleukin 2 and higher levels of interleukin 6 in the latter (59).

CKD and diabetes were negatively associated with HZ. A prospective cohort study demonstrated that patients with diabetes were less likely to develop HZ (HR 0.7, 95% CI 0.5–1.0) (57), for reasons unknown. A systematic review established that evidence of diabetes as a risk factor for HZ was inconclusive (5). A possible reason for the negative association with HZ is that individuals with CKD and diabetes may have regular follow-ups with other healthcare providers and increased opportunities for HZ vaccination, thereby reducing the risk of contracting HZ. Another postulation for the observed relationship between CKD and HZ is effect modification by CKD subgroups. CKD is diagnosed via markers of kidney damage (urinary indices, imaging) or reduced glomerular filtration rate for more than 3 months, as per the Kidney Disease: Improving Global Outcomes 2012 Clinical Practice Guideline (60). Literature is scarce regarding the risk of HZ in different CKD stages or diagnosis groups. However, subgroup analysis was not performed due to limited data availability regarding how CKD diagnoses were established.

This study is one of few population-based studies that have examined the burden of HZ and healthcare utilization in a primary care setting. There has been a limited number of epidemiological studies on HZ implemented in primary care, which is significant as vaccinations are mostly administered in primary care settings.

Since patient visits to public primary care clinics in other clusters, as well as private clinics, could not be captured in the eHints database, the prevalence of HZ, healthcare utilization rates, and associated factors such as chronic conditions may be underestimated. Individuals who chose to self-medicate instead of seeking medical attention would contribute to underreporting of cases.

In examining the association between factors of interest and HZ, the varying clinical profiles of patients who sought care from different healthcare providers may have confounded results. The differences in patient profiles and practices of public and private care limit the generalizability of findings to the private arm of primary care.

The duration from the onset of HZ symptoms to presentation at the first outpatient visit was unknown. This duration would have influenced whether oral antiviral therapy was prescribed (i.e., the proportion of cases prescribed oral antiviral therapy), the effectiveness of oral antiviral therapy, and the subsequent likelihood of complications and referral rate.

Due to limited data availability, all HZ visits for each case were assumed to belong to a single HZ episode. Therefore, recurrent HZ episodes would be missed. Nevertheless, recurrence often occurs many years after the initial episode (61). Less than 1% of cases in this study had HZ visits spaced three or more months apart, suggesting that the assumption above was less likely to affect the results.