Discussion
The results from this review show that HZ-associated mortality rate is generally low and, that despite the data being heterogeneous, in all studies mortality increased with age. Globally HZ-associated mortality is very low (<0.1/100 000) in those aged under 70 to 75 years and increases in older people, reaching between 1.2 to 7.2/100 000 in those aged ≥85 years. Not all studies reported trends in HZ mortality incidence over time, but one study in England and Wales reported a reduction in HZ-associated mortality over the 90's. The overall EU data from the Detailed Mortality Database (DMDB) showed variations over time for those aged 50–69 years and those aged ≥70 years. However, the HZ mortality rate did not appear to increase or decrease from 1994 to 2012.
Similar to the EU data reported here, data from the US, for 1979 to 2007, showed that the HZ mortality rate was higher in older patients; an average of 134 deaths were recorded with HZ as the underlying cause, and 45% of these occurred in those aged ≥85 years. The overall age-adjusted HZ-mortality was reported to decrease by 42% from 7.8 to 4.5 per 100 000. In another study in the US, from 1986 to 1995 the overall HZ-HFR was 5.3%; in those with underlying conditions the HFR was 8.7%, compared with 3.7% in those without. In Australia, it was also reported to be higher in older people and women; 219 of the 238 (92%) deaths with HZ as the primary cause recorded from 1971 to 1993; (219) were in people aged >65 years and 65% were women. Also the overall crude HZ mortality incidence was stable over this period (0.068 per 100 000); 0.043 and 0.092 per 100 000 in men and women, respectively. In another study in Australia, from July 1998 to June 1999 the overall HZ-HFR was 4% (186/4718). The HFR was higher in those with HZ as a secondary diagnosis compared with those with HZ as a primary diagnosis: 6% vs. 1%.
Mortality rates were higher when HZ was reported to be an associated or secondary cause of death suggesting that the HZ episode could have an indirect role in mortality. Underlying comorbidities can increase the severity of the HZ episode and thus the risk of mortality. It has been estimated that the risk of dying was five-times higher within three years of HZ reactivation than in age-matched controls that had not had HZ. Diagnosis of HZ infection can increase the risk of subsequent diagnosis of diseases such as cancer, cerebrovascular disease or myocardial infarction and HIV. In addition, hospitalisation of elderly people for acute diseases can result in functional decline and loss of autonomy.
It is important to be cautious when making comparisons between countries given the differences in study methodologies and coding practices in each country. Also, the data are presented for differing age groups, some of which are very large, making comparisons between studies difficult. In particular, some data are given for patients aged ≥65 years when we know that HZ-associated mortality rates are very different between those aged 65 to 75 years and those aged ≥85 years.
The differences in case definitions used could account for a substantial part of the heterogeneity. Most studies used clinical diagnosis for HZ, without requiring laboratory confirmation. In an elderly population, with many patients presenting underlying conditions or other comorbidities, it is difficult for a physician to be certain about the causal relationship between HZ and death, particularly if the certifying physician does not have access to all the patient's medical records and autopsy results.
It is possible that the variations in the HZ-HFR are due to differences in the numerator since HZ hospitalisation rates may vary between countries due to different healthcare pathways, healthcare seeking behaviours, or perception of symptoms. In addition, differences in coding practices could contribute to this heterogeneity. Although only four studies in this review reported CFRs, there was less heterogeneity observed; it has been reported that HZ incidence rates do not vary between European countries. Other factors such as differences in the prevalence of underlying risk factors for severe illness, differences in comorbidity rates or some other unidentified factors in different populations may also contribute to the observed heterogeneity.
Although most studies used HZ codes on death certificates for the primary cause of death, some used the codes for either primary or associated causes of death. Some of the limitations of using death certificates for estimating mortality incidence include incompleteness and errors. In the US, an assessment of the hospital records for decedents, whose death certificate recorded HZ as the underlying cause of death, showed that HZ was the underlying cause in only 52.5% (21/40) of cases and a contributing cause in another 12.5% (5/40). Other studies have reported that use of electronic or paper death certificates can lead to underestimations or overestimates of the true mortality rate due to other infectious and non-infectious diseases.
The strengths of our review include the fact that many studies and databases provide mortality data from national databases thus providing data for the whole population. However, there are some limitations. For example, the data are not from patient-level longitudinal prospective studies and, therefore, the retrospective design of the studies could reduce the accuracy of the data and provide an underestimation of the burden of HZ mortality. Some studies only reported mortality with HZ as the primary cause, whereas it has been suggested that the mortality is higher when HZ is an associated or contributory cause of mortality. In addition, there were not much data available and there was a range of methods used for the estimations.
Six European Union countries have national surveillance systems and five have sentinel systems for HZ. In the future, access to these surveillance databases should improve the quality of data concerning the incidence of HZ in Europe. In addition, the recent initiatives to harmonise coding for death certificates within Europe should play an important role in future HZ surveillance programmes, in particular by improving the comparability of data sources. The initiatives are important for the evaluation of the impact of infant varicella vaccination and adult HZ vaccination on HZ incidence and mortality rates.