Validity of InterVA model versus physician review of verbal autopsy for tracking tuberculosis-related mortality in Ethiopia

Abstract

Background
In most African countries where a legitimate vital registration system is lacking, physicians often review verbal autopsy (VA) data to determine the cause of death, while there are concerns about the routine practicality, accuracy, and reliability of this procedure. In Ethiopia where the burden of tuberculosis (TB) remains unacceptably high, reliable VA data are needed to guide intervention strategies. This study aimed to validate the InterVA model against the physician VA in tracking TB-related mortality in Ethiopia.

Methods
From a sample of deaths in Addis Ababa, Ethiopia, VAs were conducted on TB-related mortality, physician-certified verbal autopsy (PCVA) through multiple steps to ascertain the causes of death. InterVA model was used to interpret the causes of death. Estimates of TB-related deaths between physician reviews and the InterVA model were compared using Cohen’s Kappa (k), Receiver-operator characteristic (ROC) curve analysis, sensitivity, and specificity to compare agreement between PCVA and InterVA.

Results
A total of 8952 completed PCVA were used. The InterVA model had an optimal likelihood cut-off point sensitivity of 0.64 (95% CI: 59.0–69.0) and specificity of 0.95 (95% CI: 94.9–95.8). The area under the ROC curve was 0.79 (95% CI: 0.78–0.81). The level of agreement between physician reviews and the InterVA model to identifying TB-related mortality was moderate (k = 0.59, 95% CI: 0.57–0.61).

Conclusion
The InterVA model is a viable alternative to physician review for tracking TB-related causes of death in Ethiopia. From a public health perspective, InterVA helps to analyze the underlying causes of TB-related deaths cost-effectively using routine survey data and translate to policies and strategies in resource-constrained countries.

Full text

Bisratetal. BMC Infectious Diseases (2022) 22:200
https://doi.org/10.1186/s12879-022-07193-w
RESEARCH
Validity ofInterVA model versusphysician
review ofverbal autopsy fortracking
tuberculosis-related mortality inEthiopia
Haileleuel Bisrat1*, Tsegahun Manyazewal1, Hussen Mohammed1,2, Bilal Shikur1,3 and Getnet Yimer1,3
Abstract
Background: In most African countries where a legitimate vital registration system is lacking, physicians often review
verbal autopsy (VA) data to determine the cause of death, while there are concerns about the routine practicality,
accuracy, and reliability of this procedure. In Ethiopia where the burden of tuberculosis (TB) remains unacceptably
high, reliable VA data are needed to guide intervention strategies. This study aimed to validate the InterVA model
against the physician VA in tracking TB-related mortality in Ethiopia.
Methods: From a sample of deaths in Addis Ababa, Ethiopia, VAs were conducted on TB-related mortality, physician-
certified verbal autopsy (PCVA) through multiple steps to ascertain the causes of death. InterVA model was used to
interpret the causes of death. Estimates of TB-related deaths between physician reviews and the InterVA model were
compared using Cohen’s Kappa (k), Receiver-operator characteristic (ROC) curve analysis, sensitivity, and specificity to
compare agreement between PCVA and InterVA.
Results: A total of 8952 completed PCVA were used. The InterVA model had an optimal likelihood cut-off point
sensitivity of 0.64 (95% CI: 59.0–69.0) and specificity of 0.95 (95% CI: 94.9–95.8). The area under the ROC curve was 0.79
(95% CI: 0.78–0.81). The level of agreement between physician reviews and the InterVA model to identifying TB-related
mortality was moderate (k = 0.59, 95% CI: 0.57–0.61).
Conclusion: The InterVA model is a viable alternative to physician review for tracking TB-related causes of death in
Ethiopia. From a public health perspective, InterVA helps to analyze the underlying causes of TB-related deaths cost-
effectively using routine survey data and translate to policies and strategies in resource-constrained countries.
Keywords: Tuberculosis, Mortality, Verbal autopsy, InterVA, Cause of death, Ethiopia
© The Author(s) 2022. Open Access This ar ticle is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco
mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Background
Tuberculosis (TB) remains one of the major public health
threats worldwide [1]. According to the WHO 2020
annual TB report [2], an estimated 10 million people
fell ill with TB in 2019, and close to half a million peo-
ple developed rifampicin-resistant TB (RR-TB), of which
78% had multidrug-resistant TB (MDR-TB). In Ethiopia,
TB is still a major public health concern [3–6]. e coun-
try is among the 30 countries with the highest burden of
TB, TB/HIV, and multi-drug resistant TB. Although the
TB burden in Ethiopia has steadily declined, the esti-
mated incidence remains high at 151 per 100,000 popula-
tions and a death rate of 24 per 100,000 population.
In developing countries and Africa in particular, accu-
rate and reliable data on causes of death is essentially
lacking [7–9]. Ethiopia is one of those countries with
an impaired vital registration system [9–11]. Mortality
Open Access
*Correspondence: haylishb2@gmail.com
1 Center for Innovative Drug Development and Therapeutic Trials
for Africa (CDT-Africa), College of Health Sciences, Addis Ababa University,
P.O. Box 9086, Addis Ababa, Ethiopia
Full list of author information is available at the end of the article
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Page 2 of 9
Bisratetal. BMC Infectious Diseases (2022) 22:200
estimates, so far, are derived poorly from demographic
and health surveys, surveillance systems, and math-
ematical models. e Addis Ababa Mortality Surveil-
lance Program (AAMSP) is a unique undertaking, which
monitors citywide mortality based on burial surveillance
in all cemeteries within the city boundary [12]. e pro-
gram collects further information on causes of death and
other socio-demographic characteristics through Verbal
Autopsy (VA) from cases sampled out of burial registries.
e verbal autopsy expert algorithm (InterVA) is a
compter-driven model developed and used to interpret
VA data into probable causes of death [13]. It calculates
the probability of a set of causes of death, given the pres-
ence of indicators (circumstances, signs, and symptoms)
reported in VA interviews [14]. It is faster and cheaper,
although statistical modeling of this sort may not reflect
the subjective subtleties of physicians’ review [15], it is
advantageous in terms of efficiency, consistency, and
standardization. As a result, there is an increasing trend
to shift from physician’s based review to InterVA model;
however, the reliability of this model across different set-
tings has not been studied well [15].
In most of the African countries, including Ethio-
pia, where a legitimate vital registration system is lack-
ing, physicians often review VA data to determine the
cause of death, while there are concerns on the routine
practicality, accuracy, and reliability of this procedure.
In Ethiopia where the burden of TB remains unaccept-
ably high, reliable VA data are needed to guide decision-
making strategies. e country needs reliable data on
the level and causes of TB-related mortality for effective
TB program implementation and realization of the End
TB strategy by 2035. is study aimed to validate the
InterVA model against the physician VA in tracking TB-
related mortality in Ethiopia.
Method
is study used a completed verbal autopsy between
2007 and 2017. e cause of death assigned by physicians
and InterVA was compared at the individual level.
Source ofdata
e source of data was from VA and physician review
records of the AAMSP database. e AAMSP used to
register all deaths that happen in Addis Ababa on regu-
lar basis. e primary source of AAMSP data was death
registration followed by verbal autopsy interviews. Infor-
mation gathered from a standardized interview about
the circumstances of death from relatives or friends or
close caregivers of the deceased was reviewed by physi-
cians to assign causes of death. e AAMSP was estab-
lished in 2001 in Addis Ababa, the capital of Ethiopia
with the main objective of monitoring cause-specific
mortality with a special focus on HIV/AIDS. From a
total of 89 cemeteries, the program collected information
about decedents’ background information and lay the
cause of death from family members[13, 16].
Data collection
Burial surveillance
A burial surveillance form has been used by cemetery
clerks who trained about death registrations in training
workshops. In each of the cemeteries, one or two ceme-
tery-based clerks were assigned to register deaths using
the burial surveillance form prepared for this purpose.
e variables registered by the cemetery clerks include
the name of the deceased, date of burial, age, sex, birth
region, marital status, ethnicity, religion, specific address,
and lay cause of death. More than 15,000 deaths have
been registered annually from all the cemeteries. e
data that was collected from all cemeteries of Addis
Ababa was entered into computer software to serve as a
sampling frame for VA.
Verbal autopsy
A random sample of about ten percent of burial records
from all cemeteries except ‘Baytewar’ cemetery was
selected for VA interviews. ‘Baytewar’ is an Amharic
word used to refer to a stranger or someone who is
socially isolated. In Baytewar cemetery, bodies with no
close relatives or friends to facilitate a funeral were bur-
ied. e Baytewar cemetery alone accommodated around
15% of the total number of burials. Most of these were
infant bodies delivered by the obstetric wards of hospitals
and remain unidentified cemetery and those with com-
plete addresses were not eligible for VA interviews.
Verbal autopsies were conducted by trained inter-
viewers after 2 to 3months of the mourning period. e
interviewers were trained on how to contact respondents
when to interview and complete the questionnaires. Of
the total records that were sampled for VA interviews,
about 7.6% were not completed with the main reasons
either respondents refused the interview or the house-
holds were not found.
VA employs a standardized questionnaire to produce
information about the causes of death. rough this pro-
cess, the information on the sign, symptoms, medical
history, and circumstances preceding deaths were elic-
ited by interviewing the next kin or caregiver. e cause
or sequence of causes that led to death were assigned
based on data collected from the questionnaire and any
health records and narrative section that has been avail-
able. e VA questionnaire was adapted from one used
at the INDEPTH Network site [17] and included a set of
questions previously used in Ethiopia [18–20] and modi-
fied on the WHO 2012 VA instrument developed by the
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Page 3 of 9
Bisratetal. BMC Infectious Diseases (2022) 22:200
WHO, Health Metrics Network (HMN), and the IN-
DEPTH-Network [21].
Physician review
Physicians reviewed the completed VA questionnaire to
ascertain the causes of deaths which were done in mul-
tiple steps. Initially, the cause of death was assigned
independently by two physicians after they review the
completed VA questionnaire. en, the two-physicians
diagnosis was checked by the surveillance team mem-
bers. If the two-physician diagnosis (for the assigned
cause of death) contradicts, a third physician reviewed
the case, and the final assignment was made based on
the agreement between any two of the three physicians.
However, if the assigned cause of death by the third phy-
sician was not in agreement with any of the two-physi-
cian diagnoses, the cause for death was assigned on a
panel discussion between them. In the situation where
the three physicians could not reach an agreement, the
cause of death was assigned as undetermined. e phy-
sicians were participating to review the process which
was second- or third-year internal medicine residents of
Addis Ababa University recruited to join the university
after serving two or more years as a General Practitioner
(GP) in any of the public hospitals. We provided them
training and annual refreshments on the standard verbal
autopsy method.
InterVA
e physician and the InterVA-3.2 model independently
assessed the same basic data from the VA questionnaire.
e study tested the concordance of assigning any kind
of TB as a cause of death between physician review and
the interval model. e model’s input data include signs,
symptoms, medical history, and situations collected from
the VA questionnaires’ close-ended questions. Compiling
the same VA data into an input file for the InterVA model
and processing it into the cause of death data, data adap-
tations were made to match the model. e model addi-
tionally demands a "high" or "low" input to describe the
local prevalence of two specific causes, which can vary by
order of magnitude between settings. e WHO verbal
autopsy tool does not have data on some InterVA mark-
ers, thus they were left blank.
A STATA do file used to validate the model in 2003 was
revised and applied to produce parameters essential for
the model. Kappa statistics, ROC curve, sensitivity, and
specificity were applied to compare agreement between
PCVA and InterVA. Causes of death are assigned
to a predefined matrix of evaluated probabilities of
occurrence.
Interpretation oftheInterVA model
e model relates a range of input indicators, like sex,
age, physical signs and symptoms, medical record, and
therefore the circumstances of death to likely CODs
using Bayesian probabilities [22]. e model leads to up
to 3 likely causes per case when possible; each related
to a quantified likelihood. To assign an estimate of the
certainty for that patient, the model gives the common
likelihood for a maximum of three CODs [23]. during
this study, a high prevalence of Malaria and HIV/AIDS
were used as basic epidemiological parameters for the
model as their prevalence varies from place to put. Data
were entered case-by-case into Microsoft visual FoxPro
window of the InterVA version 3.2 to assign the pos-
sible COD responsible for the death of every individual.
Distribution ofN subjects byphysician review andInterVA
model category
e formula we used to determine the distribution of
subjects by physician review and InterVA model cat-
egory was:
where Po = the relative observed agreement among
raters.
Pe = the hypothetical probability of chance agreement
K = Kappa statis
Physician review
Category 1
(Yes) Category 2
(No) Total
InterVA
Model Cate-
gory 1
(Yes)
a B a + b P1. = (a + b)/N
Cat-
egory
2 (No)
c D c + d P2. = (c + d)/N
Total a + c b + d N
P
0.1 = (a + c)/N P
0.2 = (b + d)/N
where a = the total number of instances that both raters
said were correct. e raters are in agreement.
b = the total number of instances that rater 2 said
was incorrect, but rater 1 said were correct. is is also
disagreement.
K
=
P
o
−P
e
1
−
Pe
=
1−1−P
o
1
−
Pe
Po
=
a
+
d
N
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Page 4 of 9
Bisratetal. BMC Infectious Diseases (2022) 22:200
c = the total number of instances that rater 1 said
was incorrect, but rater 2 said were correct. is is also
disagreement.
d = the total number of instances that both raters said
was incorrect. Raters are in agreement.
In Cohen’s kappa, the chance agreement is defined as
the sum of the products of marginal distributions, i.e.
Receiver operator characteristics (ROC) curve
For both PCVA and InterVA, the area under the receiver
operator characteristics (ROC) curve was used to assess
overall diagnostic performance (properly diagnosing all
diseases). e area under the curve (AUC) of a procedure
should be near one for it to be highly sensitive and spe-
cific. e approach is more accurate if the curve closely
aligns the left-hand border and the top border of the
ROC space. If the area under the ROC curve was more
than 0.75, we evaluated our methods to be appropriate.
Validity measures: sensitivity andspecicity
Sensitivity and specificity with their 95% confidence
intercal (CI) were compared for PCVA and InterVA
model. e formula for the calculation were defined as:
Where: TP = true positive; FP = false positive;
TN = true negative; FN = false negative.
Inclusion andexclusion criteria
e inclusion criteria were all records where causes of
death were assigned and those with all the information
like identification of decedents, sex, age, house address,
and date of birth.
e exclusion criteria were records with missing in
address, name, inconsistency in the address including
place of burial. Causes of death assigned by the InterVA
model as “Indeterminate” were also excluded from the
analysis.
Data management andanalysis
Data analysis was conducted using STATA 14 software.
Individual decedents had a unique ID and all datasets
were merged using this ID before analysis. Two sepa-
rate variables (TB_inerva and TB_pr) were generated
for comparison purposes and documenting the trend of
Pe(k)
=
P.1
∗
P1.
+
P.2
∗
P2.
Sensitivity
=
TP
(TP
+
FN)
Specificity
=
TN
(FP
+
TN)
TB-related death. Coding and recoding, labeling, and
analysis were done.
Estimates of TB-related deaths between physician
reviews and the InterVA model were compared using
Cohen’s Kappa (k) and ROC curve analysis. A Kappa
value of < 0 indicates no agreement and 0–0.20 as slight,
0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as
substantial, and 0.81–1 as almost perfect agreement.
In this paper, sensitivity refers to the ability of the
InterVA model to correctly identify TB-related deaths
as assigned by the physician method; whereas specificity
refers to the proportion of Death from other causes that
are correctly identified as non TB. ese two measures
were closely related to type I and type II errors. Both sen-
sitivity and specificity were calculated.
Ethical considerations
e study has been reviewed and approved by the Ethio-
pian National Research Ethics Review Committee of the
Ethiopian Ministry of Science and Technology, and the
Institutional Review Board of the College of Health Sci-
ences, Addis Ababa University. Informed consent was
obtained from caregivers or another eligible adult in the
family during the VA interview. Access to the AAMSP
data was obtained from the Addis Ababa University Mor-
tality surveillance team that manages the project. e
participants remained anonymous and were not identi-
fied in the research report or in other means used to dis-
seminate findings.
Results
Background characteristics
A total of 8952 VA was completed with physician diagno-
sis in the period September 2007 to 2017. Of these, 4618
(51.6%) were male and 4086 (45.6%) were from the age
group 65years and above. Regarding educational status,
3836 (42.8%) had completed elementary and/or second-
ary school, while 3721 (41.6%) were illiterates. About
3936 (44%) and 2471 (27.6%) of decedents were married
and widowed, respectively (Table1).
TB‑related cause ofdeath
From the total 8952 PCVA conducted, 972 (10.9%) were
assigned as TB-related deaths by physicians. Of these,
313 VA were excluded from the InterVA analysis due to
important missing variables. Of the 8,639 VAs able to be
analyzed using InterVA, 975 (11.3%) were assigned as
TB-related deaths (Table2).
Of those classified by physicians as TB-deaths, 533
(54.8%) were female, and from the InterVA model, 520
(53.3%) were female. e proportion of male TB-related
death by Inter VA model was comparable with physi-
cian review, 65 (47.1%) and 66 (48.5%), for the period
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Page 5 of 9
Bisratetal. BMC Infectious Diseases (2022) 22:200
2009. In the same period, the proportion of female TB-
related death by InterVA and physician review were 73
(52.9%) and 70 (51.5%), respectively. However, male
and female TB-related death varies from year to year
for both InterVA model and physician review (Fig.1).
Agreement indiagnosis betweenphysician andInterVA
Model
According to the analysis, the results of the level of agree-
ment was 92.9%, with a kappa value of 0.59 (95% CI:
0.57–0.61), indicating that there is a moderate agreement
(Table3).
e observed agreement was defined as:
From this analysis, the level of agreement in the Kappa
value was 0.59, which indicates that there was a moderate
agreement.
ROC analysis
Figure 2 presented the ROC analysis in a one-to-one
square. e area under the curve captures the relation-
ship between the sensitivity and specificity of the Inter
VA method and is indicative of how the method per-
formed. e curve follows the left-hand border and then
the top border of the ROC space indicating an acceptable
level of accuracy.
In this study, the ROC curve indicated that the InterVA
model predicted the cause of TB-related death with an
area under curve or probability of 0.79 when compared
Po
=
622
+
7314
8639
=
0.918
Pe(k)
=
(0.112
∗
0.112)
+
(0.887
∗
0.887)
Pe(k)
=
0.799
k
=
0.918
−
0.799
(1
−
0.799)
=
0.59
Table 1 Background characteristics of decedents 2007–2017
Addis Ababa, Ethiopia
Background characteristics Number Percent
Sex
Male 4618 51.59
Female 4334 48.41
Age
15–24 382 4.27
25–34 1019 11.38
35–44 1124 12.56
45–54 1071 11.96
55–64 1270 14.19
65+ 4086 45.64
Education
No formal school 3721 41.6
Primary 2025 22.6
Secondary 1811 20.2
Tertiary + 833 9.3
Unknown 303 3.4
Marital status
Single 1648 18.41
Married 3936 43.97
Separated/divorced 852 9.3
Widowed 2471 27.60
Unknown 45 0.50
Table 2 Comparison of all causes of death by Physician and InterVA model
InterVA model
Year Physician review Cause of death with non‑TB Cause of death with TB Cause of death
with non‑TB
2007 874 (76.5) 221 (20.1) 875 (79.8)
2008 224 (19.9) 902 (80.1) 184 (16.9) 908 (87.5)
2009 136 (11.9) 1009 (88.1) 138 (12.5) 966 (12.7)
2010 98 (9.2) 966 (90.8) 102 (9.9) 930 (90.1)
2011 86 (8.1) 978 (91.9) 96 (9.3) 934 (90.7)
2012 101 (9.3) 984 (90.7) 114 (10.9) 931 (89.1)
2015 26 (3.8) 664 (96.2) 51 (7.7) 612 (92.3)
2016 20 (2.3) 840 (97.7) 45 (5.5) 778 (94.4)
2017 13 (1.7) 763 (98.3) 24 (3.2) 730 (96.8)
Total 972 (10.9) 7980 (89.1) 975 (11.3) 7664 (88.7)
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Page 6 of 9
Bisratetal. BMC Infectious Diseases (2022) 22:200
with physician review which indicates the good diag-
nostic performance of the method. e sensitivity and
specificity of the InterVA model were 0.64 and 0.95,
respectively.
Discussion
is study compared the cause of TB-related death
agreement between the InterVA model and physician
review methods. e study showed that level of agree-
ment between physician review and the InterVA model
was 92.9% with the kappa value of 0.59, which indicated
a moderate agreement. e finding was similar to a pre-
vious study in the country [24], Kappa 0.58% (95% CI:
0.50–0.65), but higher when compared with another
study[25], kappa = 0.50, (95% CI: 0.4–0.6). is indicated
the potential of the IntraVA model to be used to establish
TB-related death data.
When it comes to assigning cause-specific mortalities
using VA data at the population level, the probabilistic
InterVA model produced substantially similar results as
the physicians in this study, which was in line with previ-
ous studies [14, 26]. e InterVA model’s cause-specific
mortality data was consistent with existing knowledge
Fig. 1 Proportion of TB-related deaths based on gender in physician review and InterVA model in the years 2007 to 2017
Table 3 Distribution of N subjects by physician review and InterVA model category
Physician review
TB‑related death Death from other cause Total
InterVA Model TB-related death 622 (a) 353 (b) 975
Death from other cause 350 (c) 7314 (d) 7664
Total 972 7667 8639 (N)
0.00 0.25 0.50 0.75 1.
00
Sensitivity
0.00 0.25 0.50 0.75 1.00
1 - Specificity
Area under ROC curve = 0.7955
Fig. 2 Receiver operating characteristic curve of sensitivity and
specificity
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Page 7 of 9
Bisratetal. BMC Infectious Diseases (2022) 22:200
about the burden of diseases in the sub-Saharan Africa
context [27, 28], suggesting that the model performed
well ingenerating cause-specific mortality data from VA.
e study finding showed that the overall cause of TB-
related death both by physicians and the model were low
when compared with previous studies from Ethiopia
which has given a kappa value of 31%, with TB-related
death to be 36.0% and 23.0% by the InterVA model and
the physicians respectively [24]. is variation might be
due to the study population and time of the study.
e issue of how to obtain a true gold standard in VA
validation research arises repeatedly. CODs based on
hospital diagnoses have been considered the gold stand-
ard in many studies [29, 30]. Hospital diagnoses, on the
other hand, have limitations as a gold standard since the
composition and distribution of hospital CODs may not
be indicative of community mortality. Furthermore, in
resource-constrained healthcare settings, when hospital
diagnoses are available, they are of poor quality and are
limited by insufficient clinical data and record keeping
[31–33]. Moreover, hospital users and residential users
may have varied abilities to notice, recall, and report
indicators of sickness. Physician review was used as a
reference standard in this study to investigate InterVA.
For this study’s population, physician review was the
only option for COD assessment. is option, however,
has drawbacks. Physicians’ experiences, perceptions,
and interpretations of local epidemiology might con-
tribute to variations in COD data, making it difficult to
make valid temporal and regional comparisons. Further-
more, they frequently make decisions based on open his-
tory and may not account for all indicators consistently.
Our report showed that the use of the physician review
has helped to find the most relevant facts relating to the
cause of death by tuberculosis, but the choice had limita-
tions. e physicians have considered the detailed infor-
mation by going through the questionnaire, using their
clinical skills and experiences in determining the cause
of death. ey might however be influenced by their own
biases.
Deaths during TB treatment signify gaps in the accu-
rate implementation of TB programs [34, 35] and this has
been higher in countries like Ethiopia where the health-
care system is hindered by infrastructure and human
resource constraints [36–39]. e current study sheds
light on the role and feasibility of using the InterVA
model alternative to physician review of verbal autopsy
that competes for the scarce health human resources.
e limitation of this study was that it used second-
ary data for the primary analysis which is because the
questionnaire is developed for all causes of death not
specifically for TB-related death identification. Another
drawback of using physician review as a gold standard
is that physicians may misinterpret some of the VA data,
leading to a potentially incorrect cause of death result.
e experience, observation, and interpretation of the
physicians might also influence interpretations and reach
a biased decision of cause of death. Otherwise, the study
was conducted carefully and the data management and
analysis were conducted in line with the standard and
appropriate procedures and statistics methods.
Conclusion
e InterVA model is a viable alternative to physician
review for tracking TB-related causes of death in Ethio-
pia. From a public health perspective, InterVA helps to
analyze the underlying causes of TB-related deaths cost-
effectively using routine survey data and translate to poli-
cies and strategies in resource-constrained countries.
Abbreviations
VA: Verbal autopsy ; TB: Tuberculosis; PCVA: Physician-certified verbal autopsy;
k: Cohen’s Kappa; MDR-TB: Multidrug-resistant TB; RR-TB: Rifampicin-resistant
TB; AAMSP: Ababa Mortality Surveillance Program; GP: General practitioner; PR:
Physicians review; WHO: World Health Organization; ROC: Receiver operator
characteristics; InterVA: Verbal Autopsy Expert Algorithm.
Acknowledgements
The authors are thankful to the Center for Innovative Drug Development and
Therapeutic Trials for Africa (CDT-Africa), Addis Ababa University, for the suc-
cessful coordination of the work. The authors are grateful for their permission
to use the needed data for the Addis Ababa Mortality surveillance program.
We are grateful to Professor Peter Byass, who trained the application of InterVA
to the Addis Ababa Mortality surveillance research team.
Authors’ contributions
HB conceived the idea, analyzed the data, and write the first draft; TM, HM, GY,
and BS revised the draft; and all authors approved the final version for publica-
tion. All authors have read and approved the final manuscript.
Funding
This work was supported in part by the European and Developing Countries
Clinical Trials Partnership (EDCTP2) program supported by the European
Union under grant number CSA2016S-1608. TM was supported by the Fogarty
International Center and National Institute of Allergy and Infectious Diseases
of the U.S. National Institutes of Health under Award Number D43TW009127.
Availability of data and materials
The dataset supporting the conclusions of this article is included in the article.
Declarations
Ethics approval and consent to participate
The study was approved by the Institutional Review Board (IRB) of the College
of Health Sciences, Addis Ababa University (Ref. No. AAUMF 03-008) and The
Federal Democratic Republic of Ethiopia Ministry of Science and Technology
(Ref. No. 310/616/06). Letters of permission were obtained from respective
facilities. After enough time is given by responding to any question raised by
study participants, interviews and procedures were performed. All procedures
were performed as per Helsinki Declaration and national ethical standard.
Informed consent for participation was obtained from caregivers or another
eligible adult in the family. It was performed in a private space. The English
language consent or parental consent and/or assent form was translated to
Amharic back-translated to English for validation.
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Page 8 of 9
Bisratetal. BMC Infectious Diseases (2022) 22:200
Consent for publication
Not applicable.
Competing of interests
The authors declare that they have no conflict of interest.
Author details
1 Center for Innovative Drug Development and Therapeutic Trials for Africa
(CDT-Africa), College of Health Sciences, Addis Ababa University, P.O. Box 9086,
Addis Ababa, Ethiopia. 2 Department of Public Health, College of Medicine
and Health Science, Dire Dawa University, Dire Dawa, Ethiopia. 3 Department
of Public Health, College of Health Sciences, Addis Ababa University, Addis
Ababa, Ethiopia.
Received: 18 August 2021 Accepted: 17 February 2022
References
1. El Bcheraoui C, Mimche H, Miangotar Y, Krish VS, Ziegeweid F, Krohn
KJ. Burden of disease in francophone Africa, 1990–2017: a systematic
analysis for the Global Burden of Disease Study 2017. Lancet Glob Health.
2020;8(3):e341–51.
2. Global tuberculosis report 2020. https:// www. who. int/ publi catio ns-
detail- redir ect/ 97892 40013 131. Accessed 5 Dec 2021.
3. Mohammed H, Oljira L, Roba KT, Ngadaya E, Ajeme T, Haile T. Burden of
tuberculosis and challenges related to screening and diagnosis in Ethio-
pia. J Clin Tuberc Other Mycobact Dis. 2020;19: 100158.
4. Manyazewal T, Woldeamanuel Y, Holland DP, Fekadu A, Blumberg HM,
Marconi VC. Electronic pillbox-enabled self-administered therapy
versus standard directly observed therapy for tuberculosis medication
adherence and treatment outcomes in Ethiopia (SELFTB): protocol for a
multicenter randomized controlled trial. Trials. 2020;21(1):383.
5. Mohammed H, Oljira L, Roba KT, Yimer G, Fekadu A, Manyazewal T. Con-
tainment of COVID-19 in Ethiopia and implications for tuberculosis care
and research. Infect Dis Poverty. 2020;9(1):131.
6. Manyazewal T, Woldeamanuel Y, Blumberg HM, Fekadu A, Marconi VC.
The fight to end tuberculosis must not be forgotten in the COVID-19
outbreak. Nat Med. 2020;26(6):811–2.
7. Uneke CJ, Uro-Chukwu HC, Chukwu OE. Validation of verbal autopsy
methods for assessment of child mortality in sub-Saharan Africa and the
policy implication: a rapid review. Pan Afr Med J. 2019;33:318.
8. Karat AS, Maraba N, Tlali M, Charalambous S, Chihota VN, Churchyard GJ.
Performance of verbal autopsy methods in estimating HIV-associated
mortality among adults in South Africa. BMJ Glob Health. 2018;3(4):
e000833.
9. Gebremedhin S. Development of a new model for estimating maternal
mortality ratio at national and sub-national levels and its application for
describing sub-national variations of maternal death in Ethiopia. PLoS
ONE. 2018;13(8): e0201990.
10. Silva R, Amouzou A, Munos M, Marsh A, Hazel E, Victora C, et al. RMM
Working Group. Can community health workers report accurately on
births and deaths? Results of field assessments in Ethiopia, Malawi and
Mali. PLoS One. 2016;11(1):e0144662.
11. Weldearegawi B, Spigt M, Berhane Y, Dinant G. Mortality level and predic-
tors in a rural Ethiopian population: community based longitudinal study.
PLoS ONE. 2014;9(3): e93099.
12. Araya T, Tensou B, Davey G, Berhane Y. Burial surveillance detected signifi-
cant reduction in HIV–related deaths in Addis Ababa, Ethiopia. Trop Med
Int Health. 2011;16(12):1483–9.
13. Araya T, Reniers G, Schaap A, Kebede D, Kumie A, Nagelkerke N, et al.
Lay diagnosis of causes of death for monitoring AIDS mortality in Addis
Ababa, Ethiopia. Trop Med Int Health. 2004;9(1):178–86.
14. Oti SO, Kyobutungi C. Verbal autopsy interpretation: a comparative analy-
sis of the InterVA model versus physician review in determining causes of
death in the Nairobi DSS. Popul Health Metr. 2010;8:21.
15. Fottrell E, Byass P. Verbal autopsy: methods in transition. Epidemiol Rev.
2010;32:38–55.
16. Misganaw A, Mariam DH, Araya T, Aneneh A. Validity of verbal autopsy
method to determine causes of death among adults in the urban setting
of Ethiopia. BMC Med Res Methodol. 2012;12:130.
17. INDEPTH Network. Accra, Ghana. http:// www. indep th- netwo rk. org/ resou
rces/ tools.
18. Weldearegawi B, Ashebir Y, Gebeye E, Gebregziabiher T, Yohannes M,
Mussa S, et al. Emerging chronic non-communicable diseases in rural
communities of Northern Ethiopia: evidence using population-based ver-
bal autopsy method in Kilite Awlaelo surveillance site. Health Policy Plan.
2013;28(8):891–8.
19. Tadesse S, Tadesse T. Evaluating the performance of interpreting Verbal
Autopsy 3.2 model for establishing pulmonary tuberculosis as a cause of
death in Ethiopia: a population-based cross-sectional study. BMC Public
Health. 2012;12:1039.
20. Lulu K, Berhane Y. The use of simplified verbal autopsy in identifying
causes of adult death in a predominantly rural population in Ethiopia.
BMC Public Health. 2005;5:58.
21. Byass P, Chandramohan D, Clark SJ, D’Ambruoso L, Fottrell E, Graham WJ,
et al. Strengthening standardised interpretation of verbal autopsy data:
the new InterVA-4 tool. Glob Health Action. 2012;5:1–8.
22. Byass P, Fottrell E, Huong DL, Berhane Y, Corrah T, Kahn K, et al. Refining a
probabilistic model for interpreting verbal autopsy data. Scand J Public
Health. 2006;34(1):26–31.
23. Inter VA 3.2 model. http:// www. inter va. net. Accessed 2012 Feb 12.
24. Tensou B, Araya T, Take DS, Byass P, Berhane Y, Kebebew T. Evaluat-
ing the InterVA model for determining AIDS mortality from verbal
autopsies in the adult population of Addis Ababa. Trop Med Int Health.
2010;15(5):547–53.
25. Tadesse S. Validating the InterVA model to estimate the burden of mortal-
ity from verbal autopsy data: a population-based cross-sectional study.
PLoS ONE. 2013;8(9): e73463.
26. Fantahun M, Fottrell E, Berhane Y, Wall S, Högberg U, Byass P. Assessing a
new approach to verbal autopsy interpretation in a rural Ethiopian com-
munity: the InterVA model. Bull World Health Organ. 2006;84(3):204–10.
27. Jamison DT, Feachem RG, Makgoba MW, Bos ER, Baingana FK, Hofman
KJ, et al. Disease and Mortality in Sub-Saharan Africa. 2nd ed. Washington
(DC): The International Bank for Reconstruction and Development / The
World Bank; 2006.
28. de-Graft Aikins A, Unwin N, Agyemang C, Allotey P, Campbell C, Arhinful
D. Tackling Africa’s chronic disease burden: from the local to the global.
Global Health. 2010;6:5.
29. Setel PW, Macfarlane SB, Szreter S, Mikkelsen L, Jha P, Stout S, et al. Moni-
toring of vital events. A scandal of invisibility: making everyone count by
counting everyone. Lancet. 2007;370(9598):1569–77.
30. Bauni E, Ndila C, Mochamah G, Nyutu G, Matata L, Ondieki C, et al.
Validating physician-certified verbal autopsy and probabilistic modeling
(InterVA) approaches to verbal autopsy interpretation using hospital
causes of adult deaths. Popul Health Metr. 2011;9:49.
31. Mohammed H, Oljira L, Roba KT, Ngadaya E, Tesfaye D, Manyazewal T,
et al. Impact of early chest radiography on delay in pulmonary tuberculo-
sis case notification in Ethiopia. Int J Mycobacteriol. 2021;10(4):364–72.
32. Mohammed H, Oljira L, Teji Roba K, Ngadaya E, Mehari R, Manyazewal T,
et al. Who to involve and where to start integrating tuberculosis screen-
ing into routine healthcare services: positive cough of any duration as the
first step for screening tuberculosis in Ethiopia. Risk Manag Healthc Policy.
2021;14:4749–56.
33. Said B, Charlie L, Getachew E, Said B, Wanjiru CL, Abebe M, Manyaze-
wal T. Molecular bacterial load assay versus culture for monitoring
treatment response in adults with tuberculosis. SAGE Open Med.
2021;9:20503121211033470.
34. Naik PR, Moonan PK, Nirgude AS, Shewade HD, Satyanarayana S, Raghu-
veer P, et al. Use of verbal autopsy to determine underlying cause of
death during treatment of multidrug-resistant tuberculosis, India. Emerg
Infect Dis. 2018;24(3):478–84.
35. Charlie L, Saidi B, Getachew E, Wanjiru CL, Abebe M, Tesfahunei HA, et al.
Programmatic challenges in managing multidrug-resistant tuberculosis
in Malawi. Int J Mycobacteriol. 2021;10(3):255–9.
36. Mussie KM, Gradmann C, Yimer SA, Manyazewal T. Pragmatic manage-
ment of drug-resistant tuberculosis: a qualitative analysis of human
resource constraints in a resource-limited country context-Ethiopia. Int J
Public Health. 2021;66: 633917.
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Page 9 of 9
Bisratetal. BMC Infectious Diseases (2022) 22:200
•
fast, convenient online submission
•
thorough peer review by experienced researchers in your field
•
rapid publication on acceptance
•
support for research data, including large and complex data types
•
gold Open Access which fosters wider collaboration and increased citations
maximum visibility for your research: over 100M website views per year
•
At BMC, research is always in progress.
Learn more biomedcentral.com/submissions
Ready to submit your research
Ready to submit your research
? Choose BMC and benefit from:
? Choose BMC and benefit from:
37. Temesgen E, Belete Y, Haile K, Ali S. Prevalence of active tuberculosis and
associated factors among people with chronic psychotic disorders at St.
Amanuel Mental Specialized Hospital and Gergesenon Mental Rehabilita-
tion center, Addis Ababa, Ethiopia. BMC Infect Dis. 2021;21(1):1100.
38. Manyazewal T, Woldeamanuel Y, Blumberg HM, Fekadu A, Marconi VC.
The potential use of digital health technologies in the African con-
text: a systematic review of evidence from Ethiopia. NPJ Digit Med.
2021;4(1):125.
39. Mussie KM, Yimer SA, Manyazewal T, Gradmann C. Exploring local realities:
perceptions and experiences of healthcare workers on the management
and control of drug-resistant tuberculosis in Addis Ababa, Ethiopia. PLoS
ONE. 2019;14(11): e0224277.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in pub-
lished maps and institutional affiliations.
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com