Mortality data, when viewed comprehensively, can uncover changes in population health status over time and can be obtained through various methods to help understand these changes, including the ability to detect the initial cases of infectious diseases, monitor preventable diseases, implement public health response, raise awareness and provide insights to prevent further deaths. Mortality data help identify the causes and circumstances of death and thus provide essential information for an evidence-based response to changes in population health status.^1,2,3 Cause of death (CoD) is the official determination of the condition leading to death. Cause of death is currently classified using the International Statistical of Diseases and Related Health Problems (ICD) codes.⁴ International Statistical Classification of Diseases and Related Health Problems is the base used for mortality coding and classification for diseases, disorders, injuries and other health-related conditions with multiple uses such as on death certificates or use by healthcare centres for billing. Currently, routinely available mortality data are limited in most low- and middle-income countries (LMICs) since most deaths occur at home in these settings.⁵ Where implemented, the mortality data systems are ad-hoc or sub-national, and the CoD data are incomplete or unreliable.⁶ Data quality is low for 85 countries capturing CoD data; and 74 countries, predominately in sub-Saharan Africa, do not capture mortality data in a routine surveillance system.⁷

Mortality data can also assist in assessing the impact of human immunodeficiency viruses (HIV) prevention and antiretroviral therapy (ART) treatment programmes. Trend data on HIV-related deaths can help HIV programmes determine if deaths among people living with HIV are increasing, decreasing or remaining stable. This information, along with CoD information, can help countries design or improve their existing HIV programmes and improve patient outcomes, such as viral suppression.

Implementing mortality systems is also important as they, in combination with information on new infections, can assist in improving our understanding of HIV epidemic control.⁷ One of the definitions of epidemic control is the point at which the total number of new HIV infections falls below the total number of deaths from all causes among HIV-infected individuals while both exhibit a declining trend (additional conditions for using this metric include a minimum ART uptake of 70% and a falling trend of new HIV infections). As countries get closer to HIV epidemic control, it is critical to understand the remaining burden and find creative solutions to HIV prevention, care and treatment. The lack of accurate mortality data can potentially hinder countries documenting HIV epidemic control. Currently, mortality data among people living with HIV, especially in countries in sub-Saharan Africa, come from either reporting death from the next of kin to the healthcare facility (which is limited) or through the follow-up of patients that miss their clinical appointments. Many deaths are missed in both these cases, and CoD information is generally not available. Therefore, mortality estimates for many LMICs, including most countries in sub-Saharan Africa, are based on modelling (e.g. Spectrum model for HIV, which is used to generate annual HIV estimates in 170 countries).⁸

There is a dearth of published papers that provide information on methods to collect mortality data. Here we aim to describe various methods, listed further in the text, that can be utilised to gather mortality data.⁹ These methods provide mortality data, including overall and cause-specific mortality rates and trends, and CoD information including HIV as the CoD. Strengths and weaknesses, along with resource and implementation considerations, of each method are described in Table 1.

Civil registration is defined as the continuous, permanent, compulsory and universal recording of vital events (births and deaths) pertaining to the population, as provided through decrees or regulations in accordance with the legal requirements in a country.¹⁰ Establishing a legal requirement obliges all citizens to abide by said requirement so that burials cannot take place if the death is not registered.⁷ Civil Registration and Vital Statistics (CRVS) systems – when fully functional, with national coverage, and including CoD – are the gold standard for assessing mortality rates and trends in a country. A national CRVS system includes registration of all deaths, irrespective of location, along with medically certified CoD. Data captured by the CRVS system can generate continuous information on health indicators, help understand the prevalence, distribution and causes of mortality, as well as the health inequities affecting mortality to inform programmatic and policy decisions. Many LMICs record vital events including births and deaths, but they may not cover the entire population, generate data on a continuing basis, capture accurate CoD especially for those dying outside of health facilities or be linked electronically.^6,11 Therefore, these countries cannot compile data for accurate, detailed and ongoing analyses. Cause of death is a critical part of CRVS and should ideally come from medical certification. However, many countries fail to capture accurate CoD because of a lack of medical certification of deaths which can be because of multiple factors including lack of training for deaths occurring at health facilities and no clinical oversight for deaths occurring in the community. A high-functioning national CRVS system remains the gold standard for national mortality surveillance; however, achieving such a system is complicated by legal mandates, infrastructure and differing priorities of stakeholders.⁶

For HIV, CRVS systems can help assess the impact of HIV programmes and treatment on mortality in the total population. In many countries without a CRVS system, many deaths that occur outside of health facilities are not counted, thus leaving gaps in mortality data.^7,12 Therefore, maintaining comprehensive CRVS includes necessary infrastructure, community participation and medical certification of CoD. Countries could conduct national assessments of their CRVS system and develop strategies to address gaps in mortality data.

Mortality surveillance can be conducted with engagement of staff at burial sites. Each death brought to the burial site is registered in the surveillance system and basic demographic information is collected as part of registration. Cause of death is recorded if the cadaver has been brought from a healthcare facility or a mortuary (in case an autopsy is conducted at the mortuary). For cadavers brought from the community, verbal autopsy (VA) may be conducted by trained personnel to determine the CoD. Burial systems surveillance can also be used to calculate mortality trends using local population size.^13,14

In countries without medical certification of CoD or CRVS, mortality questions added to household surveys may provide a framework to understand the leading CoD in a population.^7,15 Although VA modules are the most commonly used method to collect information on CoD in household surveys, supplemental methods may include a follow-up survey or a post-census survey.⁷ Follow-up surveys typically take place within a year of a death identified through the initial survey.¹⁶ Some countries may include additional modules such as the social autopsy, which identifies situations that may have affected mortality risk. Results from additional questionnaires help inform future survey methodology and strategies to improve mortality data capture.¹⁷ Post-census surveys are similar to follow-up surveys with the key difference being that in post-census surveys, deaths are identified through a census, and a sample of households with recorded deaths are followed up with VA interviews.

Routinely collected hospital data are a main source of cause-specific mortality statistics. Hospital data on cause-specific mortality are guided by the following HIV mortality principles: (1) recording deaths individually by patient age, sex, CoD and location of the hospital, (2) recording deaths in line with ICD principles when more than one cause is reported, (3) maintaining confidentiality and (4) sharing local mortality statistics with certifying clinicians.⁷ Clinical staff are trained on proper reporting of diagnosis and CoD data. Decision tables may also help staff accurately identify complicated underlying CoD.¹⁸ Hospitals can consider creating a local ICD coding index categorising the most frequent diagnoses and including comorbidities and sequences of events leading to death.^19,20,21

Loss to follow-up (LTFU) tracking is a method specific to HIV treatment programmes, although it can be utilised for other CoD including tuberculosis (TB). Many deaths among people living with HIV which occur outside the healthcare system go unreported and are incorrectly deemed ‘lost to follow-up’ in HIV programme data; this can lead to incorrect mortality estimates among people living with HIV receiving HIV services.^22,23 A few methods have been proposed to correct mortality estimates.²³ In settings with active death registration systems, notified deaths can be linked to the patient’s ART records, allowing documentation of the outcome of the LTFU tracking.^6,24 In settings without adequate civil registration records or with incomplete, low-quality CoD data, tracing of individuals LTFU can be used to determine the current status and reason(s) for interruptions of treatment by surveillance and/or clinic staff via phone calls, home visits or social networks.^23,25 If the death of an individual who has been LTFU is confirmed, the civil record can be updated by submitting a death report form to the national death registry, and mortality estimates can be adjusted accordingly. In addition to helping improve HIV mortality estimates, LTFU tracking is crucial to people living with HIV monitoring and improving their clinical outcomes.

For HIV-specific mortality surveillance, mortality status, along with CoD, can be included as a part of existing HIV case surveillance (CS) systems. Human immunodeficiency viruses case surveillance consists of routine reporting of longitudinal individual-level data on all people living with HIV in a population from the point of diagnosis and through the cascade of sentinel events including treatment, viral suppression and death. Routine CS facilitates HIV programmes to monitor and address gaps in HIV service delivery and health outcomes, including mortality, among people living with HIV. However, CS systems are currently limited in scale in most LMICs, rarely collect mortality data, including CoD and face similar challenges as LTFU. When CS systems are limited or do not collect mortality data systematically, data from other mortality systems can be linked to CS to inform mortality data or bridge gaps in mortality data in CS. Designing an HIV CS system depends on the country context. Availability of well-established patient management systems facilitates reporting of data from sentinel events, including death, in the design of CS systems. Mortality data in HIV CS can assist in understanding the factors that contribute to mortality in HIV-positive individuals.^11,26

An autopsy is a surgical examination of a body to determine the cause and manner of death or to evaluate any disease or injury. Although histological examination related to autopsy is the most reliable method of determining CoD, an autopsy is not commonly used especially in LMICs, as autopsies are time-consuming and are conducted by trained medical professionals.²⁷ In many LMICs, autopsies are only performed for medico-legal cases or upon request of the family of the deceased.

Verbal autopsy is an interview-based method, used to determine the probable CoD, in which an interview is conducted with the next of kin or caregivers after a death occurs outside of a healthcare setting.²⁸ The interview, normally conducted by a trained field worker, is guided by a structured survey and gathers information on signs and symptoms related to illnesses exhibited by the deceased prior to death.²⁹ For HIV, the VA questionnaire asks if the deceased ever had a positive HIV test or was diagnosed with AIDS, among other questions. Data are then analysed by physicians or automated computerised models to assign probable CoD. Many studies have been conducted to validate various VA approaches in the absence of CRVS, ranging from physician review of VA to automated and data-derived algorithms. For comparability, there are standaridsed VA guidance and data-collection instruments.^28,30,31,32 Resources for training, evaluation and careful ethical consideration when discussing death in various cultures and populations are important in creation and maintenance of a successful VA system.³³

Sample vital registration with verbal autopsy (SAAVY), although not a long-term replacement for CRVS, can provide birth, death and CoD data needed for policy and programmes. Sample vital registration with verbal autopsy is a combination of CRVS and VA where vital event registration is completed in a sample of surveillance areas. Verbal autopsy is used to determine the CoD for deaths in a sample of surveillance sites, selected through probability sampling (as opposed to national coverage for CRVS). Following the selection of sites, a census is conducted at each site to gather baseline information about the population. Key informants in each site routinely monitor and report vital events and a VA is conducted for every death. The key difference between SAVVY and CRVS is that CRVS systems should have national coverage and be routine and ongoing, whereas SAVVY is done in a sample of surveillance sites.

Full autopsies are considered the gold standard for determining CoD; however, they are cumbersome, require expertise and resources and have low acceptability. Autopsies, and associated testing, can help identify HIV in undiagnosed individuals and determine HIV-related CoD. In place of full autopsies, biomarker-based sentinel surveillance can provide trends in cause-specific mortality and can be conducted in two ways.

While LMICs work to implement CRVS, other methods can supplement or fill gaps in mortality data. The following factors could be considered when deciding which method is most feasible in the local context: methods of data collection, monitoring and storage; availability of electronic data systems versus reliance on paper-based systems; availability of laboratory infrastructure and skills; availability of human resources and technical skills of staff and availability of funding. Countries could also assess their existing infrastructure, resource availability and local mortality data need to determine the most feasible methods for implementation. In addition, cultural sensitivities around HIV and death in different communities should be considered, as acceptability of each method may vary across cultures. In LMICs, prioritising low-cost methods such as verbal autopsies or burial surveillance, combined with targeted training for local staff could help to address immediate mortality data gaps while laying the groundwork for more long-term CRVS development. Future research could potentially explore CRVS implementation challenges in LMICs, as a deeper understanding of these barriers may support more targeted and effective advocacy efforts while strengthening this gold standard system.

All these methods provide information on mortality generally, which includes mortality among people living with HIV and HIV as the CoD. Loss to follow-up tracking and HIV case surveillance are specific to mortality among people living with HIV and may provide information on other CoD among people living with HIV. With increasing numbers of people living with HIV in treatment and care, HIV-specific mortality data may be helpful to monitor and address gaps in epidemic control. This need has underscored the lack of consistent and routinely available mortality data. Although many of the existing mortality systems in LMICs are ad-hoc, and thus data from these systems often cannot be systematically and routinely used to inform national mortality estimates, they are still an important effort, as they provide extremely useful data that are not available otherwise. In the long run, mortality data-collection systems could be implemented in an ongoing or periodic manner so mortality data, including CoD data, are available over time. Data from these mortality methods can help prevent deaths by tailoring interventions to a particular region, disease or population.