Multiple cause-of-death data among people with AIDS in Italy: a nationwide cross-sectional study

The RCOD, managed by the Italian National Institute of Statistics (ISTAT), collects all death certificates issued in Italy. Individual death certificates consist of two sections: a medical section for certification of causes of death reported by a physician and a sociodemographic section filled in by local civil registration officers. Sociodemographic information refers to sex, age, place (municipality or foreign state) of birth, residence and place of death, citizenship, and education level. The cause-of-death certification includes the description of the sequence of conditions that directly lead to death and, where appropriate, other conditions that may have contributed but were not directly linked to death. The classification of causes of death and the selection of the UC is performed according to the rules and provisions included in the International Classification of Diseases, 10^th Revision (ICD-10) provided and updated by the World Health Organization (WHO). In recent years, ISTAT also disseminates MCOD data.

The RAIDS, managed by the National Institute of Health (ISS), collects information on PWA in Italy, according to the European AIDS case definition [16, 17]. The reporting of AIDS cases became mandatory in 1986. The RAIDS includes demographic information on date of AIDS diagnosis, sex, age at diagnosis, place of birth and residence at diagnosis; conversely, date of death is not mandatorily updated. The RAIDS has a national coverage; a previous study estimated a 6% underreporting by linking RAIDS data with the National mortality database [18].

In order to identify the death certificates referring to PWA, the RAIDS, updated to 31^st December 2010, was linked to the RCOD database for the period 2006–2010. The record-linkage was carried out in compliance with the national law regulating the use of data from PWA and according to the authorization of the Italian Data Protection Authority (provided in the National Statistical Plan).

The record-linkage was conducted using a semi-automated and validated procedure with an updated version of SALI software [19]. This is a deterministic matching procedure that relies on an automated multiple-step algorithm using blinded names, surnames, and date of birth as matching criteria. To optimize linkage sensitivity, SALI takes into account of most spelling errors; thus, subsequent manual comparison of other variables present in both databases (e.g., place of birth) is required to discard false-matched records. For the aims of the present study, two groups of deaths were identified through the record-linkage:

Furthermore, the following population subgroups were excluded from the analysis:

The present MCOD analysis focused on deaths occurred in the 35–54-year age class, where AIDS mortality rates were the highest, and which included 78% of total deaths among PWA in the period 2006–2010. The number of death certificates excluded from the comparison group due to the mention of B20-B24 or R75 codes was 1,905.

In the end, the study population included 2,515 records of AIDS deaths (among these, 486 death certificates had no mention of AIDS/HIV) and 123,224 records of non-AIDS/HIV deaths.

The conditions most frequently associated with PWA mortality were identified using an age-standardized relative risk indicator, which identifies the excess proportion of having a specific disease mentioned in the death certificate among PWA, as compared to those without AIDS/HIV.

For a specific cause of death c, the Age-Standardized Proportion Ratio (ASPR) was calculated as the ratio between the following estimated proportions:

The two estimated proportions were obtained through age standardization based on the 5-year age class distribution of total deaths aged 35–54 years, according to the following formulas:

where:

D _Ac observed number of deaths certificates mentioning the cause c among AIDS deaths

D _A observed number of AIDS deaths

D _Āc observed number of deaths certificates mentioning the cause c among non-AIDS/HIV deaths

D _Ā observed number of non-AIDS/HIV deaths.

_x w proportion of deaths at age x ( _x D) on the total number of deaths (D).

The 95% confidence intervals (CI) for the ASPR were calculated using the standard error formula for relative risks according to Altman [20]. ASPR significantly greater than unity indicates a higher frequency of a specific cause among AIDS deaths; conversely, values lower than unity indicate a higher frequency of the specific cause among non-AIDS/HIV deaths. Causes of death were analysed at a high level of specificity (ICD-10 category). Furthermore, the ASPR was calculated also for broad groups of causes (ICD-10 chapters).

Our results showed that although infectious diseases provided a small contribute to the UC among PWA, they represented the most common cause of death when considering the MCOD (Fig. 1); this result stressed the relevant role of these conditions (which are, in most cases, preventable or treatable) in the survival of HIV-infected individuals. The high proportion of infectious diseases among AIDS deaths has to be attributed in part to the immunodeficiency caused by HIV infection [21] and in part to behaviors that expose HIV-positive individuals to certain types of infections, such as viral hepatitis [22, 23]. In Italy, the HIV epidemic has been driven mainly by injecting drug users until a decade ago; some behaviors associated with drug use, such as alcohol abuse or non-sterile drug paraphernalia sharing, may have contributed to an increased risk of liver damage or parenteral acquisition of viral hepatitis infections [6, 15, 24]. Actually, chronic viral hepatitis was reported in about one-fourth of all death certificates of PWA. Of relevance, hepatitis C virus (HCV) was the most frequent causative agent of chronic viral hepatitis deaths, underscoring the need of an early management of HIV-HCV co-infected individuals [25].

Extremely high values of ASPR were observed for conditions included in the AIDS-defining disease list [16, 17]: toxoplasmosis, criptococcosis, infections due to other mycobacteria (other than tuberculosis and Hansen disease), pneumocystosis, cytomegaloviral disease, candidiasis, and respiratory tuberculosis. However, the contribution of each of these conditions was quite small in terms of proportion among infectious MCOD.

Neoplasms, in particular AIDS-defining ones, are among the most common underlying causes of death among people with HIV infection or AIDS [6, 8]. The use of MCOD, as compared to UC, clearly highlighted that the frequency of cancer diagnoses reported in the certificates of AIDS deaths is noteworthy (i.e., 29%), and largely variable according to cancer sites/types. In addition to AIDS-defining cancers (i.e., KS, NHL, and cervical carcinoma), other cancers known to be frequently incident among HIV-infected people turned out to be frequently reported also in death certificates of PWA. The ASPR values for these cancers ranged from about 300-fold for KS to about 2- or 3-fold for cervical carcinoma, Hodgkin lymphoma, anal, and liver carcinomas. Conversely, statistically significant reduced ASPR emerged for very common cancers in the general population (e.g., breast cancer in women, prostate cancer in men, colon-rectal, pancreatic cancers, and lung cancer). Findings regarding lung cancer, in particular, were apparently in contrast with the literature, since people with HIV/AIDS consistently reported higher incidence of lung cancer in Italy and elsewhere [26, 27]. Furthermore, in a previous UC analysis among Italian PWA, a 6.6-fold elevated risk of death for non-AIDS defining cancers was documented as compared to the general population, including a 5.9-fold higher death risk for lung cancer [28]. Nevertheless, the MCOD analysis allowed to explore the frequency of non-AIDS defining cancers when they were not UC. These findings underlined that a direct comparison of our findings – obtained by means of MCOD approach – with those from previous studies based on the risk of dying for specific UC may be misleading [6, 8]. The results of our analysis seemed thus to stress the strong overall burden of cancer among the causes of death, pointing to the importance of cancer prevention in people with HIV infection and AIDS to enhance their life expectancy.

Overall, 13% of death certificates of deceased PWA reported one or more diseases of the blood and/or immunological disorders – a proportion that was 3-fold higher than among deceased people without AIDS/HIV. Notably, anaplastic and hemorrhagic conditions turned out to be associated with highly positive ASPR, a potentially surrogate marker of HIV-immunosuppression. Although a comparison with similar investigations is made difficult, it seems that our findings are in agreement with the existing evidence indicating that anemia is a frequent complication of HIV infection associated with an elevated risk of death [29].

A negative association between unspecified diabetes mellitus and AIDS was found, due to a double proportion of this disease among non-AIDS deaths compared to AIDS deaths. The proportion of diabetes mellitus reported among non-AIDS deaths was consistent with the results from the PASSI Italian Surveillance System [30], suggesting that this condition could be underreported among AIDS deaths because deemed not relevant to death. Despite the high association reported in the literature between diabetes and HAART therapy [31], the low ASPR found for diabetes among AIDS deaths could also be due to a to a large number of late presenters without therapy.

The proportion of mental and behavioral disorders resulted similar between the two study groups, although a high value of ASPR for dementia was observed. Dementia is commonly associated to AIDS. It could be caused by specific characteristics of HIV that reaches the central nervous system, immediately after primary infection, and plays a central role in determining pathological alterations and neurocognitive deficits [32]. The prevalence of neurocognitive disorders associated to HIV remains high in the HIV population even in the HAART era, due primarily to the persistence, even among treated patient, of forms of HIV associated neurocognitive disorders (HAND) [33]. European data confirm the high prevalence of HAND in HIV treated population, which amounts to 69% of the HIV-positive population with suppression plasma viremia [34].

Mental and behavioral disorders due to psychoactive drugs also resulted highly associated with AIDS: historically, injecting drug users are indeed the first Italian group which acquired the HIV infection [17].

Among diseases of the nervous system, the highest value of the ASPR corresponded to encephalitis, myelitis and encephalomyelitis, indicating that these causes are about 14-fold more frequent in AIDS deaths than in non-AIDS/HIV deaths [35]. These conditions could be a consequence to many infectious AIDS related diseases such as herpes simplex virus, cytomegalovirus, and toxoplasmosis [16]. Furthermore, recent studies showed that HIV can cause direct damage to the brain [36].

Diseases of the circulatory system were more frequently reported among non-AIDS/HIV deaths than among AIDS deaths. However, they were mentioned in about 32% of death certificates of PWA. This remarkable proportion is consistent with the role of circulatory system diseases (especially cardiovascular diseases) as causes of death in people with AIDS or HIV infection described in previous studies [5, 6].

The only two diseases with a higher frequency among AIDS deaths, i.e., acute and subacute endocarditis and other pulmonary heart diseases, are very common cardiac complications due to injecting drug use [37].

Values of ASPR for ischemic heart diseases and heart failure, in both cases lower than unity, are in agreement with the results obtained by Whiteside and colleagues (2015) in comparing mortality rates of people with HIV infection to those of the general population in the US, by means of MCOD data.

With regard to the frequency of respiratory conditions, it is worth noting that one out of three AIDS deaths had a respiratory disease listed in the death certificate – a frequency magnitude concealed when the UC approach is used. Pneumonia (organism unspecified) wasthe most common listed respiratory condition, in agreement with what was observed by Schwarz (2014) in the same period.

Hepatic diseases accounted for the largest proportion of causes of death and showed significantly high proportions among AIDS deaths. Liver diseases are one of the most frequent non-AIDS-defining causes of death in people living with HIV for a number of reasons: the high prevalence of hepatotropic viral co-infections, the hepatotoxicity of antiretroviral drugs, and the emergence of new liver conditions, including non-alcoholic fatty liver disease and non-cirrhotic portal hypertension [38]. Considering the high association between HIV and hepatic viral infections [39], it can be hypothesized that death certificates reporting chronic viral hepatitis in the group of infectious diseases would also frequently report a liver-associated condition coded in the group of diseases of the digestive system. The contribution of concurrent liver-related causes of death would suggest that hepatic viral co-infections and chronic liver damage also constitute a relevant cause of death among Italian PWA.

Genitourinary diseases were rarely mentioned as UC, but they appeared to be listed in about 11% of death certificates of PWA when the MCOD approach was used. Our finding is in agreement with previous evidence showing that up to 3% of HIV-infected patients may suffer of chronic kidney diseases in the HAART era [40]. The MCOD approach seems, thus, to offer an important opportunity to quantify the prevalence of renal failure, in particular, and to stress the importance of screening for renal disorders as a clinical challenge in the management of people with HIV/AIDS. Furthermore, genitourinary diseases were more frequently mentioned in the death certificates of PWA than among those of non-AIDS/HIV people, in agreement with other studies showing that people with AIDS or HIV infection are at higher risk of death for renal diseases [6, 15].

Symptoms, signs, and ill-defined causes were reported in about 40% of certificates for AIDS deaths: the condition most frequently mentioned was cachexia, typically linked to the wasting syndrome, as reported in other international studies on AIDS deaths [41].

Among external causes of deaths, only “accidental poisoning by and exposure to narcotics psychodysleptics” showed positive association with AIDS due to overdose among injecting drug users, that constituted the majority of Italian AIDS cases reported from the beginning of the epidemic [17].

The main strengths of this study were the national coverage of the population-based registries used to identify people with and without AIDS, the AIDS case definition based on official diagnoses mandatorily reported to the RAIDS, and the use of MCOD data.

The exclusion from the analysis of death certificates from the Trento and Bolzano provinces is expected to not affect the results, since the number of cases from these areas is expected to be very small (overall UC deaths due to AIDS account for 0.8% of UC deaths due to AIDS in Italy in the period 2006–2010).

The choice to focus on deaths aged 35–54 years enabled feasible (robust) MCOD analysis by detailed cause-of-death category. Furthermore, it prevented possible bias due to misdiagnosis of HIV/AIDS in people at older ages, which could result in undiagnosed AIDS cases among the comparison group. However, future developments of the study will include the analysis of deaths of PWA aged 55 years and over, in order to investigate also the mortality pattern of older AIDS patients.

A quota of death certificates were not matched to the RAIDS mentioned codes B20-B24 or R75 (1,905) and were then excluded from the comparison group. These cases are currently under investigation to establish how many of them referred to AIDS deaths or, more plausibly, to deaths of people reported with HIV infection. This in-depth analysis requires to take into account both multiple cause codes and the text reported by the certifier.

Only a really small number of death certificates of PWA reported codes B20-B24 or R75 as single cause of death (8 out of 2,515), without any impact on the validity of MCOD analysis.

The ASPR compares proportions of deaths and needs only mortality data without the risk exposed population. The indicator highlights the conditions most frequently mentioned in AIDS versus non-AIDS/HIV deaths. Therefore, the ASPR does not measure the overall mortality risk of PWA compared to the general population, as in other analyses [15]. Briefly, it is a measure of the propensity of a condition to act as a cause of death in AIDS patients. In addition, it is useful in detecting the conditions reported in the chain of events leading to death.

It is well known that mortality data derived from death certificates have some limitations. Among these, the death certificate completion is entrusted to the medical knowledge of the certifying physician, and it can suffer from a lack of specificity and/or underreporting [42]. On the other hand, the death certificates are the most relevant tool for understanding which causes of death are the most prevalent in a given population, and this information is central to resource allocation for prevention and treatment.