Cause of Death in a Tertiary Referral Hospital in Maputo, Mozambique: A One-year Autopsy Study in an Adult Population

Background: Information on causes of death (CoD) is very poor in sub-Saharan Africa, although it provides vital information for planning health strategies. We aimed to analyze the CoD in a large series of autopsies conducted in Mozambique during one-year period. Methods: All adult autopsies conducted at the Maputo Central Hospital (Mozambique) from January,1st to December 31st 2013 (12 month-long period) were analyzed, according to HIV status. Maternal and traumatic deaths were excluded from the study. Results: Four hundred and forty-nine adult autopsies were performed (52% males, median age 42 years, range 15-93). HIV-positive patients comprised 37% (166), 19% (87) were HIV-negative, and in 44% (196) information on HIV status was unavailable. Infectious diseases were the leading CoD (249; 55%), followed by malignancies (76; 17%), cardiovascular (75;17%), other non-infectious diseases (46;10%), and non-conclusive CoD (3;1%). Infectious diseases were responsible for 64%, 51% and 50% of HIV-positive, -negative and -unknown patients, respectively. Tuberculosis was the leading infectious CoD, regardless of HIV status. Other common infections included pyogenic meningitis (42; 17%) and pneumonia (37;15%). Non-infectious cardiovascular diseases and malignancies comprised 33% of all CoD in HIV-negative, and 15% in HIV-positive patients. Conclusion: The burden of infectious diseases, particularly of tuberculosis, remains high in Mozambique, even in HIV-negative patients. Apart from tuberculosis, pyogenic infections of the central nervous system and respiratory tract are also frequent and thus require introduction of new point-of-care tests and improvement of treatment strategies. Non-communicable conditions, mainly cardiovascular and oncologic diseases are becoming prevalent in Mozambique. A Study


Introduction
Many low-and middle-income countries (LMIC) face important challenges in maintaining reliable information on cause of death (CoD) due to the poor quality of their vital statistics and death registration systems [1]. Most of the scant information available in these countries is based on verbal autopsy and clinical records, two methods with important drawbacks due to the overlapping symptoms of many conditions, particularly of infectious diseases [2,3]. In LMIC, and particularly in sub-Saharan Africa, the accuracy of the clinical diagnoses is limited due to the lack of access to laboratory diagnostic tests and imaging techniques; a common obstacle even in tertiary, referral hospitals [4,5]. In addition, although some of the sub-Saharan African countries have a very high HIV prevalence, reliable data on CoD in HIV-positive patients are particularly scarce. Unfortunately, the HIV-related in-hospital mortality remains high in sub-Saharan Africa despite the increased availability of antiretroviral therapy [6]. Indeed, many studies carried out in sub-Saharan Africa have shown a high burden of opportunistic infections associated with HIV [7,8], especially of active tuberculosis [9]. The clinical autopsy is the gold standard method for CoD determination, and an important tool for assessing the quality of clinical practice [10,11].
In many LMIC, where access to health system is limited and mortality surveillance is generally weak, autopsy data would be particularly important to complete the CoD picture which is key for health planning and prioritization [12,13]. In addition, clinical autopsy is a useful teaching tool for clinicians, vital to evaluate and improve the quality of care and the health outcomes. However, the high costs of the autopsy procedure, the lack of pathologists and of adequate pathology laboratories, as well as cultural concerns continue to be main obstacles for its expansion in LMIC [14]. In this prospective one-year study we aimed to explore the CoDs in a large series of adult deaths occurring in a tertiary referral hospital located in Maputo, the capital of Mozambique, using complete autopsy as the gold standard. Additionally, we analyzed the clinical and laboratory information on HIV status and compared the distribution of the CoDs between HIV-positive, and -negative and in patients with unknown HIV status.

Study Setting and Case Selection
This study was part of an observational study conducted at the Maputo Central Hospital (MCH), a 1500-bed government-funded quaternary care facility. The hospital is located in the capital of Mozambique and serves mainly the population of the Maputo urban and peri-urban area but is also the referral center for other hospitals in Southern Mozambique.

Autopsy Procedure
In all cases a full autopsy was performed as soon as possible after consent from the next-of-kin was obtained. The autopsy was carried out by a pathologist assisted by a technician. During the autopsy, all the thoraco-abdominal organs, as well as the central nervous system were eviscerated and dissected for detailed gross examination. A sample was obtained for histological analysis from the main organs (both lungs, liver, brain, heart, kidneys, spleen and bone marrow). The sampling was directed to any grossly evident lesion, if detected, or from normally appearing tissues if not identified. In addition, samples of any other abnormality identified in other areas (e.g. lymph nodes, bowel, skin, etc.) were obtained.

Methods
Tissue samples for histological analysis were fixed in 10% neutral buffered formalin for 24-48 hours and routinely embedded in paraffin. Four-micron thick sections were stained with H&E and evaluated using an Olympus BX41 light microscope. In all the H&E slides a thorough search of hemozoin in macrophages and of malaria parasites was conducted. When deemed necessary, ancillary histochemical (e.g. Zieh-Neelsen, Gram, Periodic-Acid Schiff, Grocott stains, etc.) and/or immunohistochemical (e.g., Cytomegalovirus, Herpes, etc.) stains were performed on selected paraffin blocks.

Review of the Clinical Charts
The available clinical information was reviewed to collect diagnostic evidence for each case by one local investigator (FF). In all cases, the HIV status of the deceased patients, antiretroviral treatment, CD4 count data and the time from admission to death were abstracted from clinical charts.

Determination of the CoD
The main cause of death, all associated conditions and any underlying diseases, were codified following the International Classification of Diseases, tenth revision (ICD-10). The CoD were classified into five major groups: Kaposi's sarcoma and lymphoid neoplasms were always classified as disseminated. The non-infectious cardiovascular diseases included cardiovascular and cerebrovascular diseases. Other non-infectious diseases were classified by system (respiratory, central nervous system, digestive system, genitourinary, etc.) or as nonclassifiable. Finally, when no CoD was identified after a thorough exam, the case was classified as non-conclusive.

Statistical Methods
Data were analyzed using the SPSS program version 20.0. The proportions were compared by chi-square test. The ANOVA test was used to compare the means between three different categories.
p-values <0.05 were considered as significant.
The distribution of main CoD categories was markedly different in HIV-positive compared with HIV-negative or -unknown patients.
Similar differences were observed in our previous study conducted at the MCH [15] and in another autopsy series in Uganda [7].
Interestingly, the burden of infectious CoD was also high in patients with HIV-negative or unknown status (51 and 50%, respectively).
Another autopsy study conducted in Zambia reported even higher prevalence of infection in seronegative cases (79%) [9]. In contrast, our previous postmortem study in Mozambique [15,16] and an Ugandan series have reported lower number of infectious diseases in HIV-negative patients (27%) [7]. As expected, and similarly to other post-mortem series in sub-Saharan Africa [7][8][9]15,18,19], India [20] and Brazilian Amazon [3], tuberculosis was the leading infectious CoD in HIV-positive patients, contributing to onethird of all CoD in this group. These findings are clearly different from studies conducted in other geographic areas that reported Cytomegalovirus as the leading CoD in HIV-positive patients, whereas tuberculosis accounted for a small percentage of cases (<3%) [21]. In our series, disseminated (miliary) tuberculosis was the predominant form of the disease, which is in keeping with the findings of other reports [7,9,20]. This high proportion of disseminated tuberculosis probably reflects the high degree of immunosupression of HIV-positive patients included in the study [22]. However, tuberculosis was also a frequent CoD in HIV-negative and -unknown patients, comprising up to 15% of all deaths in these groups. These findings are in contrast with other autopsy studies in sub-Saharan Africa, in which tuberculosis was rare in HIV-negative and -unknown patients [7,9,16,19,23]. These results highlight the need of expanding the use of the point-of-care tests for tuberculosis, such as the Xpert MTB/RIF and Xpert MTB/RIF ultra [24,25], which, although already introduced in Mozambique [26] and other LMIC [27], are probably underutilized. As in many other autopsy series, other frequent infectious CoD included pyogenic pneumonia and central nervous system infections [7,8].
Notably, pyogenic pneumonia was more frequent in HIVnegative and -unknown patients, in keeping with other studies in sub-Saharan Africa [19,23]. These results highlight the need of continued introduction of rapid point-of-care tests [28] for nontuberculous bacterial infections in LMIC. Point-of-care tests already introduced in Mozambique for some specific conditions (Xpert MTB/RIF and Xpert MTB/RIF ultra for tuberculosis, cryptococcal antigen assays, birth HIV testing [26]) could serve as successful examples to encourage implementation of a new sensitive and specific molecular automated next generation tests for other infectious diseases. Surprisingly, severe malaria was low but much more frequent than expected in an adult population from an endemic country (2%). Most patients who died of malaria were middle-aged (mean age, 40), which is in contrast with the older age of patients dying of malaria in adulthood reported in other series [29]. Although the low immunity acquired by the population of Maputo City and Province, an area that has successfully reduced the burden of malaria [30], can be the cause of this unexpected mortality, further studies on the true burden of severe malaria in this particular age group are warranted.
Another curious finding in this series is that we did not identify any case of Cytomegalovirus infection. This infection has been reported in Ugandan, Kenian and Zambian series, but with a low frequency [7,9,23]. However, the unavailability of microbiological testing or immunohistochemistry may explain, at least in part, the absence of any identified case, because the viral inclusions in H&E stained slides are usually scarce and unevenly distributed and, consequently, difficult to detect [31]. Similarly to other LMIC, Mozambique shows a dual burden of malignant diseases. Indeed, a high incidence of malignancies of infectious etiology was identified, including the typically HIV-associated neoplasms (Kaposi's sarcoma and malignant lymphomas), hepatocellular carcinoma, associated with hepatitis B virus and bladder cancer, related to Schistosoma infection. Kaposi's sarcoma was the most frequent malignancy in this series, which is in accordance with the cancer registry reports from Beira [32], the second biggest city in Mozambique [33], and with other autopsy studies conducted in sub-Saharan Africa [7,9].
Liver carcinoma was the second most frequent malignancy in our series and was predominantly identified among non-HIV-infected male patients. Despite the reduction in its frequency, recent data has shown that liver cancer is the third most frequent cancer in Maputo in males [34]. Interestingly, in the first half of 20 th century, Mozambique showed the highest incidence of this cancer among other sub-Saharan countries, especially in young black African males living in the capital [35,36]. The carcinogens responsible for this high incidence included mainly chronic hepatitis B virus infection, but also hepatocarcinogenic toxins and dietary iron overload [35]. However, in addition to these infection-related cancers characteristic of LMIC, an increasing incidence of cancers typical of western populations, like colorectal or breast carcinoma [37] was observed, which is likely due to the westernization of the lifestyles of the population living in the Maputo urban area, a phenomenon occurring in other African countries [38,39].
Cardiovascular diseases accounted for a quarter of the CoD in HIV-negative and -unknown patients. The World Health Organization estimated that by 2020, 80% of the global cardiovascular disease burden will occur in LMIC countries [40][41][42][43][44][45][46]. The urbanization and westernization of the lifestyles in Mozambique might be responsible for the increasing incidence of cardiovascular diseases [47], in particular increasing alcohol use, together with highly caloric diet with high salt intake [48] and low fruit and vegetable consumption in urban areas [49,50].
Furthermore, the high burden of non-infectious digestive deaths in our study reminisces the overall high prevalence of this digestive fatal conditions in high-income countries [51]. Notably, many LMIC are not prepared to manage this new burden [52,53]; lack of clinical protocols, algorithms for risk stratification and poor access to essential care have been recently identified as critical gaps in the management of hypertension in Mozambique [54].
Other studies have recently highlighted the increasing prevalence of hypertension and cerebrovascular diseases in African LMIC [40], including Mozambique [41,42], especially in urban areas [43,44].
Interestingly, the deaths caused by heart failure were mostly due to hypertensive heart disease, and not to ischemic cardiopathy, which is in line with the accumulated evidence that the burden of coronary artery disease remains low in the black Africans [45].
The main limitation of this study is the complete absence of any microbiological testing in blood, cerebrospinal fluid and tissues.

Conflicts of Interests
The authors have no conflicts of interest to declare.