Background
Lack of valid and reliable data on malaria deaths, especially in endemic countries which house the greatest burden, continues to be a problem that plagues the global health community. In light of the recent adoption of the sustainable development goals (SDGs) and the World Health Organization’s
Global Technical Strategy for Malaria 2016–
2030 [
1,
2], it presents a challenge to the community on how best to capture malaria mortality data to assess progress on the ambitious goals and targets set within these agendas. Accurate data on malaria mortality at the national and subnational levels is essential for effective policy-making and programme planning; it will also be critical for countries as they move toward low transmission or pre-elimination status to monitor and evaluate progress and to adapt programmatic strategies as the burden declines.
Measuring malaria-specific mortality at the population level is challenging due to the lack of complete vital registration systems in most low- and middle-income countries, the difficulty in clinical assessment of malaria, and the fact that most malaria deaths occur outside of the formal health care system [
3‐
5]. To address this gap, the verbal autopsy (VA) method was developed to ascertain cause of death (COD) at the population level. Global and country specific data on malaria mortality is thus largely derived from incomplete vital registration data and supplemented with VA data, and in some instances inpatient mortality data, to produce estimates of the number of malaria deaths.
Verbal autopsy consists of an interview conducted with a family member or an individual familiar with the deceased using a structured questionnaire to gather information about the signs and symptoms, and their duration experienced by the deceased, and events leading up to the death. The information collected is use to determine the individual COD using the International Classification of Diseases, Tenth Edition (ICD-10) [
6]. The COD is assigned either directly by a trained physician or other automated methods. Verbal autopsy has been used as a valuable interim method to provide COD data, as countries work toward improving their civil and vital registration systems.
Despite the wide use of VA, there are many recognized limitations of VA tools and methods [
7‐
11]. The World Health Organization (WHO) in recent years has commissioned systematic reviews of VA tools and methods and held technical consultation meetings, in an effort to update and standardize VA methods and tools to address some of these limitations, including comparability of VA data across study sites [
10‐
13]. These reviews however, have had a more broad focus on how best to standardize tools and methods and have not thoroughly examined the specific limitations of VA methods and tools for measuring malaria mortality. Many VA studies have noted some of the specific limitations of VA tools and methods for measuring malaria mortality [
14‐
32]; however, to date no systematic review has been conducted to examine the challenges and limitations of VA for measuring malaria mortality and to determine how VA methods could be improved to provide more robust estimates of malaria mortality. A systematic review of the literature was conducted to document how VA tools and approaches have been used to measure malaria mortality and the key challenges and limitations of existing tools and methods.
Methods
The authors searched PubMed, the Cochrane Library, Popline, the WHOLIS, and Google Scholar, from 1 January 1990 to 15 January 2016. The search terms were “malaria” or “malaria mortality” and “cause specific mortality” and “verbal autopsy”/“post mortem interview”/“mortality surveillance”/“verbal post mortem.” We also searched available INDEPTH Network websites (27 websites from INDEPTH Network health and demographic surveillance system (HDSS) sites in Africa (21), Asia (5), and Oceana (1)) to identify additional programme reports, articles, and gray literature from the organization on malaria-specific mortality. References of included publications were also reviewed for other relevant studies. The inclusion criteria were publications that presented results from a VA study where malaria was an identified COD and/or publication that discussed limitations or challenges related to the measurement of malaria mortality through VA. The review was restricted to articles published in English.
Two authors independently searched the databases and websites. The titles and abstracts of the identified studies and reports were screened to determine if they met the inclusion criteria. Full texts of publications that passed the screening were reviewed to determine eligibility. A narrative synthesis of the publications that met the inclusion criteria was conducted. A narrative description was developed and information was extracted on key characteristics for each of the included publications using a standard matrix. Using this matrix, the reviewers carried out a thematic analysis of key challenges and limitations of measuring malaria mortality through VA.
Discussion
Over the past few decades VA has been increasingly used as a valuable interim measure to provide data on mortality rates and the main COD in low- and middle-income countries, where civil and vital registration systems are incomplete and lacking quality data on mortality. Thus, filling a large gap in providing essential information for effective health policy and programmatic planning, particularly in contexts where resources are very limited. VA has been widely used to measure malaria-specific mortality, particularly in sub-Saharan Africa (SSA) where the greatest burden of the disease exists and where there is the largest gap in COD data. Although it is widely used, it is generally recognized in VA studies and by the global malaria community that VA does not perform particularly well, regardless of the COD assignment methods used, for determining malaria mortality. Though, to date, no collaborative efforts have been made to thoroughly examine and address the main challenges and limitations of VA for measuring malaria mortality.
This synthesis of the literature revealed the main limitation of VA for malaria mortality measurement to be its overall low and varying sensitivity and specificity, the reasons for which are multifaceted. The nonspecific symptoms of malaria make it difficult to distinguish malaria deaths from other common illnesses, most notably acute febrile illnesses such as ARI and meningitis, thus introducing misclassification bias. Due to this, in a few VA studies reviewed, malaria deaths were lumped under the category of ‘fever’ or ‘acute febrile illness’ death [
51,
74‐
76]. The underlying COD profile also influences the sensitivity and specificity. A few studies for example noted the difficulty of distinguishing malaria from HIV deaths in high HIV prevalence areas due to the overlapping symptoms, suggesting that in these contexts malaria deaths are likely to be overestimated, while HIV-related deaths are underestimated [
15,
41,
53]. Due to the complex aetiology of malaria, it can also be difficult to identify whether malaria was the direct or an underlying COD, or an indirect COD. A few studies noted the difficulty in distinguishing between a malaria and anaemia death [
60,
61]; in the study by Murray et al. the authors note that they redistributed a proportion of anaemia deaths to malaria deaths due to this reason. Malaria infection has also been shown to associated with an increased risk for potentially fatal invasive bacterial infections, including non-typhoidal Salmonellae [
77]; in these cases, malaria will not be recorded as the COD despite its role in indirectly influencing the death. The malaria epidemiological context, including areas where malaria is highly seasonal, also influences the sensitivity and specificity of VA for malaria mortality, with potential bias introduced by physicians’ backgrounds and experience [
39,
42,
45,
47‐
49,
52,
56‐
58,
61,
62]; resulting in either an under or over estimation of the malaria mortality burden.
The lack of standardization of tools and methods used in VA studies is another key challenge, as it makes comparability of malaria mortality findings across sites and over time difficult [
11,
14,
36,
56,
64,
78]. This challenge is further exacerbated by the lack of detailed information provided in published studies on the tools and methods used, which was evident in this review. While the lack of standardized tools and methods is a recognized overarching issue in general for VA studies, it further complicates being able to assess from the literature what the best practices are for measuring malaria mortality through VA. For example, most VA studies in the review did not provide information on the criteria used to assign a malaria COD and for the studies that did include this information, the criteria used ranged widely across studies; thus providing little insight on the most accurate cause of death assignment for a malaria death. On the other hand, VA can still be a valuable tool for monitoring and evaluating trends in malaria mortality in specific settings over time when the same methods are applied consistently, as they are affected by the same set of biases over time [
56,
79].
A number of the studies also discussed the challenge of not having a ‘true’ gold standard by which to test the performance of VA and, therefore, caution in the interpretation of validation study findings [
16,
28,
34,
36,
42,
47,
58,
65,
68,
69,
71]. For malaria specifically, it is not just a challenge of incomplete and poor quality records, but in many validation sites the coverage of parasitological testing is incomplete and therefore, the malaria COD diagnosis is made without a confirmed laboratory test. The population health metrics research consortium (PHMRC) gold standard VA validation study initiated in 2005 is helping to address this issue through the development and use of stringent diagnostic criteria to identify gold standard deaths, thus providing a better understanding for VA performance for malaria mortality measurement across different COD assignment methods and guidance for future validation studies [
65].
To provide more robust malaria mortality estimates moving forward, it is pertinent that the global malaria community come together to review the limitations of malaria mortality data sources and methods and define a strategy for how these methods could be improved upon. Given the significant contribution of VA data in informing global malaria mortality estimates, it will be important for the strategy to include revisiting current VA tools and methods for measuring malaria mortality to determine if updates could provide improved estimates. This could be an opportunity for the field to better refine and improve the criteria used for assigning malaria deaths in VA studies and ultimately improve the sensitivity and specificity of current VA tools. There is also a strong need for better collaboration across VA experts and transparency of methods used in VA studies to ensure better standardization of VA methods and to allow for greater comparability across study findings. Further, in view of the increasing coverage of parasitological testing in malaria endemic countries, it is possible, as shown in the PMHRC validation study, to use more stringent diagnostic criteria for assigning malaria deaths.
Lessons learned from this study should be used to inform future VA validation studies. It is also important we continue to explore new strategies. Very recently, minimally invasive tissue sampling (MITS) for autopsy has emerged as a potential new method for determining COD in developing countries where full autopsies are not possible [
80‐
82]. This technique offers the potential for improved diagnostic accuracy of COD, and could potentially in the long term obviate the need for VA studies. Exploration and development of these new strategies should happen alongside longer-term efforts to improve civil and vital registration systems in low- and middle-income countries.
Authors’ contributions
SH and YY conceived the study. SH, YE, GA, KAN and KPA conducted the literature review and synthesis of the articles. SH wrote the first draft of the manuscript. YY provided guidance throughout development of the manuscript. All authors contributed to reviewing the manuscript. All authors read and approved the final manuscript.