Content uploaded by Bang Nguyen Pham
Author content
All content in this area was uploaded by Bang Nguyen Pham on Dec 22, 2022
Content may be subject to copyright.
Original article
Leading causes of deaths in the mortality
transition in Papua New Guinea: evidence from
the Comprehensive Health and Epidemiological
Surveillance System
Bang Nguyen Pham,
1
* Ronny Jorry,
1
Vinson D Silas,
1
Anthony D Okely,
2,3
Seri Maraga
1
and William Pomat
1
1
Population Health and Demography Unit, Papua New Guinea Institute of Medical Research, Goroka,
Papua New Guinea,
2
School of Health and Society and Early Start, University of Wollongong,
Wollongong, NSW, Australia and
3
Illawarra Health and Medical Research Institute, Wollongong, NSW,
Australia
*Corresponding author. Population Health and Demography Unit, PNG Institute of Medical Research, PO Box 60, Goroka
EHP 441, Papua New Guinea. E-mail: pnbang2001@yahoo.com
Received 12 July 2021; Editorial decision 10 November 2022; Accepted 10 December 2022
Abstract
Background: Changing causes of deaths in the mortality transition in Papua New Guinea
(PNG) are poorly understood. This study analysed community-level data to identify lead-
ing causes of death in the population and variations across age groups and sexes, urban-
rural sectors and provinces.
Method: Mortality surveillance data were collected from 2018–20 as part of the
Comprehensive Health and Epidemiological Surveillance System (CHESS), using the
World Health Organization 2016 verbal autopsy (VA) instrument. Data from 926 VA inter-
views were analysed, using the InterVA-5 cause of death analytical tool to assign specific
causes of death among children (0–14 years), those of working age (15–64 years) and the
elderly (65þyears).
Result: Nearly 50% of the total deaths were attributed to non-communicable diseases
(NCDs), followed by infectious and parasitic diseases (35%), injuries and external causes
(11%) and maternal and neonatal deaths (4%). Leading causes of death among children
were acute respiratory tract infections (ARTIs) and diarrhoeal diseases, each contributing
to 13% of total deaths. Among the working population, tuberculosis (TB) contributed to
12% of total deaths, followed by HIV/AIDS (11%). TB- and HIV/AIDS-attributed deaths
were highest in the age group 25–34 years, at 20% and 18%, respectively. These diseases
killed more females of working age (n¼79, 15%) than males (n¼52, 8%). Among the
elderly, the leading causes of death were ARTIs (13%) followed by digestive neoplasms
(10%) and acute cardiac diseases (9%).
V
CThe Author(s) 2022; all rights reserved. Published by Oxford University Press on behalf of the International Epidemiological Association 1
IEA
International Epidemiological Association
International Journal of Epidemiology, 2022, 1–20
https://doi.org/10.1093/ije/dyac232
Original article
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
Conclusion: The variations in leading causes of death across the populations in PNG sug-
gest diversity in mortality transition. This requires different strategies to address specific
causes of death in particular populations.
Key words: Papua New Guinea, mortality transition, Comprehensive Health and Epidemiological Surveillance
System, cause of death analysis, verbal autopsy interview, World Health Organization 2016 verbal autopsy instru-
ment, InterVA-5 cause of death analytical tool
Background
Approximately 65% of the world’s population lack quality
data that can be used for cause of death analysis,
1
particu-
larly in low- and middle-income countries (LMICs), where
most deaths are unregistered, unrecorded and unnoticed by
the health systems.
2
Inadequate data on mortality patterns
have impeded social planning and hindered the effective
development, monitoring and evaluation of health policies
in LMICs.
3
Papua New Guinea (PNG) is undergoing an epidemio-
logical transition.
4
Life expectancy of the PNG population
has improved over the past 50 years, from 45.5 years in
1970 to 59.2 years in 2000 and to 64.6 years in 2020,
according to the United Nations Report on World
Population Prospects.
5
The PNG National Census 2011 es-
timated a total population of approximately 8 million and
89 000 deaths in the preceding 12 months.
6
The projections
in PNG are based on the data from national censuses, con-
ducted every 10 years. However, data from the most recent
PNG National Census conducted in 2021 are not released
yet, with projections potentially introducing greater error
in life expectancy estimates.
There were variations of mortality between males and
females and across provinces. The Demographic and
Health Survey (DHS) 2016 provided estimates of adult
mortality rates that were 2.56 deaths per 1000 population
among women and 2.96 deaths per 1000 population
among men.
7
There was a gap of 13 years between
Sandaun Province—the province with the lowest life ex-
pectancy—and Port Moresby,the National Capital
City with the highest socioeconomic status (SES).
8
Previous studies on non-communicable diseases (NCDs)
suggested that a mortality transition was occurring in
the population as a result of recent socioeconomic
development.
9,10
The characteristics of the mortality transition in PNG
are not clear. In 2008, the National Department of Health
reported the first national data on morbidity and mortal-
ity, with a focus on the medical records from public health
facilities, mostly from the national health information sys-
tem (NHIS), which showed mortality data in the health
setting.
11
On the other hand, data from the 2016 Global
Burden of Disease (GBD) study showed that NCDs had
displaced infectious diseases as the major cause of death in
the adult population.
2,3
This study heavily relied on mor-
tality data from an urban population with access to tertiary
health care services and the modelling was based on
Polynesian patterns, where the mortality transition
Key Messages
•Variations in the epidemiological transition between the urban-rural sectors, four geographical regions (with data
from the major representing provinces), and sub-populations (children 0–14, working population 15–64 and the
elderly 65þyears) have been observed.
•Mortality data were collected for 2.5 years over 2018–20, continuing previous work where the mortality data were
collected from the integrated Health and Demographic Surveillance System (iHDSS) in the period 2011–15.
•For the first time in Papua New Guinea (PNG), the World Health Organization 2016 Verbal Autopsy (VA) instrument
was used as part of the routine surveillance activities to collect mortality data in the communities, and the InterVA-5
software was used to ascribe specific causes of death in the population, providing insights into the mortality
transition across sub-populations in PNG.
•The changing causes of death in the mortality transition in PNG highlight an urgent need for transforming the health
system, with focus on universal access to comprehensive primary health care services.
•Different trends in mortalities across sub-populations indicate that a range of approaches are needed to effectively
respond to the new demand of the population for health care services.
2International Journal of Epidemiology, 2022, Vol. 00, No. 00
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
reported was more advanced than in PNG, contributing to
incorrect modelling of the mortality transition in PNG.
In contrast, data from a cross-sectional survey of rural
areas of PNG from 2011–15 identified infectious diseases
as the main causes of death in the adult population.
4
The
discrepancy in the conclusions of leading causes of death in
PNG may be due to the sources of mortality data analysed
in these two studies. The proportion of deaths that oc-
curred in health facilities, mostly at the tertiary level, could
range from 15% among the urban populations in Eastern
Highlands, East New Britain, East Sepik and Central prov-
inces to about 20% in Port Moresby and Madang town,
12
meaning the majority of deaths occurred in the communi-
ties, particularly in rural areas. The 2016 GBD Study could
have overestimated the mortality transition in PNG to-
wards NCDs.
Understanding the variation in causes of death among
sub-populations, as well as the different patterns in mortal-
ity across urban-rural sectors, provinces and age groups, is
critical for policy makers, health planners and public
health researchers to be able to inform effective govern-
ment and health sector responses in PNG. Indicators for
monitoring and evaluation of the mortality transition have
been integrated into the country’s National Health Plan
2016–20 and Medium Term Development Plan 2018–
22.
12,13
These plans highlighted the importance of collect-
ing and reporting mortality data as a basis for the public
health sector to be able to respond to new demands for
health care services in the population.
The lack of mortality data from both urban and rural
communities, and the heterogeneity in health care service
delivery across provinces, urban-rural sectors and sub-
populations, make the projection of mortality trends and
variations in PNG challenging.
14
With new mortality data
available from the communities in both urban and rural
sectors, the Comprehensive Health and Epidemiological
Surveillance System (CHESS) could offer new insight into
the mortality transition in PNG. CHESS covers a surveil-
lance population of approximately 80 000 people and
15 000 households, equivalent to 1% of the PNG popula-
tion estimated at around 8.8 million for the period 2017–
22.
15
Whereas data from CHESS cannot be expected to
provide a comprehensive picture of mortality for the entire
population of PNG, it does provide a representative source
of information for the regions where the surveillance sites
were established.
CHESS is a new-generation surveillance system, devel-
oped based on the existing integrated Health and
Demographic Surveillance System (iHDSS) established in
the period 2011–17 (see Supplementary Table S1, available
as Supplementary data at IJE online). Two iHDSS sites,
Hiri in Central Province and Asaro in Eastern Highlands
Province (EHP), were retained in CHESS, with new sur-
veillance sites established in Port Moresby (POM), East
New Britain (ENB), East Sepik Province (ESP) and
Madang Province. Mortality data from CHESS are broader
than those from the iHDSS, with surveillance populations
living in relatively developed urban to semi-urban and ru-
ral areas in main provinces representing all four geographi-
cal regions of PNG, with different levels of socioeconomic
development.
15
In the iHDSS, verbal autopsy (VA) data
were collected by a separate mobile VA team through
cross-sectional household surveys. The VA team was sent
to the field to conduct interviews in a short period of time.
These VA interviewers were often not local people and had
limited knowledge of the local context and of the deceased.
The purpose of this study was to identify the distribu-
tion and variation of causes of death across the sub-
populations in CHESS surveillance sites. Specifically, there
were four research questions, as follows.
i. What is the distribution of causes of death among the
PNG population?
ii. What are the leading causes of death among sub-
populations such as children, people of working age
and the elderly?
iii. How do causes of death vary between male and
females, urban-rural sectors and provinces?
iv. Is there any evidence of shifting causes of death in the
PNG population over the past 50 years, and what are
the limitations in mortality data collection for moni-
toring and reporting mortality transition?
The populations living within the catchments of CHESS
are not representative of the entire PNG population. The
data source did not include people who live in further re-
mote areas, outside the catchment areas of CHESS. Hence,
the mortality data used in this study are not generalizable
to the total PNG population, but they are representative of
the urban-rural areas from which the surveillance sites
were selected and do provide a valuable source of informa-
tion on mortality in PNG, where there has been a lack of
available data.
Methods
Data source
This study used mortality data from the communities,
extracted from the CHESS database which was developed
by PNG Institute of Medical Research (PNGIMR) in the
period 2018–22. CHESS methodology has been published
elsewhere.
15,16
The sociodemographic characteristics of
CHESS surveillance sites are shown in Table 1. The mor-
tality data were collected from eight surveillance sites (four
International Journal of Epidemiology, 2022, Vol. 00, No. 00 3
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
Table 1 Overall description of socioeconomic demographic characteristics of surveillance sites, Papua New Guinea Institute of Medical Research’s Comprehensive Health and
Epidemiological Surveillance System, 2018–20
Province Central Port Moresby Eastern Highlands Madang East Sepik East New Britain
Region Southern Southern Highlands Momase Momase Islands
Surveillance site Hiri (rural) Hohola (urban) Goroka (urban),
Asaro (rural)
Newtown (urban) Maprik (rural) Kokopo (urban),
Baining (rural)
Year of setting up 2011 2017 2011 2018 2019 2018
Main industry Fishery, hunting,
LNG
Shipping, transportation Coffee, agriculture Fishery, services Vanilla, cocoa Fishery, cocoa,
tourism
Accessibility Road Road, airline Road, airline Road, airline Road, airline Sea, airline
Estimated provincial population size
54
269 756 364 125 579 825 493 906 450 530 328 369
Estimated district population size
17
44 000 in Hiri 18 000 in Hohola 47 000 in Asaro
97 300 in Goroka
112 000 in New town 12 400 33 000 in
Baining
10 000 in
Kokopo
Estimated surveillance population coverage
18
15 000
(þ/- 3000)
5000
(þ/- 1000)
15 000
(þ/- 3000)
5000
(þ/- 1000)
5000
(þ/- 1000)
10 000
(þ/- 2000)
Estimated number of household coverage
18
3000
(þ/-600)
1000
(þ/- 200)
3000
(þ/- 600)
1000
(þ/- 200)
1000
(þ/- 200)
2000
(þ/- 400)
Community health centre (CHC), hospital Three CHC One CHC Four CHC, one provincial hospitalOne CHC One CHC,
One district hospital
Two CHC
Laboratory – One laboratory One laboratory One laboratory – –
LNG: Liquified Natural Gas.
4International Journal of Epidemiology, 2022, Vol. 00, No. 00
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
urban vs four rural), established across six provinces:
Central, EHP, ESP, ENB, Madang and POM. Mortality
data from CHESS only represent the population who live
in the catchments areas of these surveillance sites.
The surveillance populations have high levels of sea-
sonal migrants (about 20%) moving in and out of the sites
throughout the year.
17,18
The estimated population sizes in
the surveillance site are rounded up to show the variation
of the surveillance populations across the sites. Each sur-
veillance site includes one or more primary health facilities
serving as sentinel site for the purpose of surveillance of
morbidities, particularly infectious diseases. Three surveil-
lance sites in Port Moresby, Goroka and Madang have lab-
oratories providing diagnostic and testing services,
including real-time polymerase chain reaction (PCR) and
GeneXpert. Hence, most of the surveillance populations
have access to health care services and but the results are
not generalizable to represent the entire PNG population,
particularly those who live in remote rural areas.
7
In con-
trast, mortality data collection was integrated into ongoing
routine surveillance activities of the CHESS and conducted
by the local staff. CHESS offered compatible mortality
data for comparisons with the iHDSS, providing further
insights into mortality transition in PNG.
World Health Organization 2016 Verbal Autopsy
data collection tool
The VA method uses interviews with relatives of the de-
ceased to collect mortality data to ascertain possible causes
of death at the population level.
19
VA methods are suitable
in routine settings, where civil registration and vital statis-
tics (CRVS) systems are not available and medical certifica-
tions of cause of death (COD) are poor.
20
The World
Health Organization (WHO) 2016 VA instrument used to
collect mortality surveillance data in CHESS is an elec-
tronic tablet-based data collection tool, developed based
on the consolidation and integration of existing VA
tools.
21
VA interviewees are asked questions about clinical
signs and personal and family history of the deceased per-
son, and whether medical records or a death certificate of
the deceased are available. The provided information is
recorded in the tablets and used for COD analysis. This
tool is designed as interactive and responsive to the pro-
vided information, involving multiple complex skip pat-
terns where interview sections not relevant to the specific
age and sex of the deceased are bypassed.
22
This approach
is more cost-effective and acceptable than paper-based VA
interview methods.
23
The WHO 2016 VA tool is suitable
for use in the current local context in PNG.
24
The evolu-
tion of causes-of-death analytical methods using VA data
have been discussed elsewhere.
23
Training, fieldwork and data management
The WHO 2016 VA instrument does not require inter-
viewers to have a health and medical background, unlike
the previous VA data collection tools. The principal inves-
tigator of CHESS was trained on VA and COD methods in
the period 2013–15. The first training on the WHO 2016
VA instrument for CHESS staff was held in April 2017 af-
ter the tool was adapted for optimal use in the PNG con-
text. The adapted tool was piloted over a 12-month period
in Asaro and Goroka sites in EHP. Following this, re-
fresher training was conducted in March 2018 for the
CHESS team, before scaling up the data collection across
the sites.
14
To ensure the VA interviews were consistent
and information on the deceased was accurately recorded
in the tablet, training focused on the design of the WHO
2016 VA tool, including the questionnaires, technical ter-
minologies and definitions. Training included practical sec-
tions on pre-test and post-test of the tool and uploading
and transferring the VA data from the tablets to the
CHESS database. Training also provided the CHESS team
with knowledge and procedures to identify potential par-
ticipants and techniques for use and maintenance of the
tablet, as well as the communication skills necessary to
successfully conduct VA interviews with relatives of the de-
ceased persons in the local context.
14
Mortality data were collected over a period of 2.5 years,
from March 2018 to September 2020. Only deaths that oc-
curred within 2 years prior to commencement and during
the data collection period were selected for VA interviews,
meaning the range of the actual dates of deaths was from
March 2016 to September 2020. The defined range of eli-
gible dates of death minimized recall biases and ensured no
overlap between mortality data from CHESS used in this
study and the data collected by the iHDSS in the period
2011–15.
4
Mortality data were collected from the communities
and no death records from health facilities were included
in this study. Deaths in the communities were identified by
village-based data reporters, who were local people living
in their villages and recruited under short-term or casual
employment contracts with PNGIMR to work for CHESS
in the communities. They conducted data collection for
reporting demographic changes at the household level.
Information on birth, death and migration was col-
lected through regular visits to households. Given the so-
cial and family networks, data reporters were easily aware
of deaths that occurred in their villages and had access to
the households at a convenient time to collect further de-
tailed information about morbidity and mortality history,
including the date of death. Data reporters were advised to
identify all deaths occurring in their villages, but
International Journal of Epidemiology, 2022, Vol. 00, No. 00 5
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
recommended only deaths that occurred within 2 years
prior to the commencement of the study and during the
data collection period to site managers for organizing VA
interviews.
VA interviews were conducted by national scientific
officers with informed consent. Data reporters pre-
arranged household visits at a time and location that suited
the respondent. For deaths that occurred during the data
collection period, VA interviews were usually scheduled in
the 2 weeks after the mourning period.
25
However, the or-
ganization of VA interviews may have taken up to several
weeks, due to logistical arrangements and the availability
of interviewees and means of transportation. Some VA
interviews required more than one visit to complete. Most
VA interviews were conducted in Tok-Pisin, the most com-
mon local language in PNG, and the Motu language was
used in Central Province.
Mortality data were extracted from the tablets in the
format of Excel spreadsheets, using ODK Briefcase. Data
were quality checked by site managers based in the field
offices before the electronic data files were sent to
PNGIMR main office in Goroka for processing. VA data-
sets were uploaded to the CHESS database by the database
manager and stored on the PNGIMR’s secure server. The
database manager checked completeness and consistency
and corrected missing data as needed. The raw VA dataset
was extracted from the CHESS database using SQL/
Process Maker and converted into InterVA-5 input format
(cvs.file). The clean VA dataset had 353 variables plus
identifiers was included in the InterVA-5 model for COD
analysis.
23
Quality control and quality assurance were
overseen by the CHESS principal investigator.
Analysing causes of death using InterVA-5
method
The interVA-5 software program was used to assign causes
of death. This method can identify 64 specific causes of
death and categories, following the International
Classification Diseases-Version 10 (ICD-10).
26
The VA
data of 926 deaths were eligible for inclusion in the
InterVA-5 models. The InterVA-5 can ascribe more than
one COD for each death, with a respective likelihood.
Only first-ascribed causes of death with the highest likeli-
hoods were presented in this study. For example, if the first
COD was pulmonary tuberculosis (TB) with likelihood of
55% and the second COD was HIV with likelihood of
45%, then pulmonary TB was listed as the COD. The term
‘undetermined’ was used if no specific COD was assigned
by the InterVA-5 model.
The distribution of causes of death was calculated for
the entire surveillance population and stratified by sex and
age groups: children (aged 0–14 years), people of working
age (15–64 years) and the elderly (aged 65þyears), rural-
urban location and province. This study reported the pro-
portions of deaths due to a specific COD or a group of
causes of death (%) among the total deaths recorded in a
particular population.
To show changes in mortality in PNG over the past
50 years, the specific causes of death in the adult popula-
tion were grouped into five main categories: (i) Emerging
infectious diseases including TB and HIV/AIDS; (ii)
Endemic infections including sepsis, acute respiratory tract
infections (ARTIs) and pneumonia, typhoid, malaria, diar-
rhoeal diseases, measles, meningitis and encephalitis, teta-
nus; (iii) Emerging NCDs including acute cardiac diseases
(ACDs), stroke and diabetes; (iv) Endemic NCDs including
neoplasms, chronic obstructive pulmonary disease,
asthma, gastrointestinal disorders, renal disorders, mental
and nervous system disorders, malnutrition and endocrine
disorders and maternal deaths; and (v) Injuries and exter-
nal causes including road traffic accidents, accidental falls,
drowning and submersion, exposure to smoke and fire,
venomous animals and plants, poisoning, noxious substan-
ces, intentional self-harm, assault, force of nature. The ra-
tionale for categorizing NCDs into Endemic and Emerging
has been documented and is consistent with definitions
used in the previous VA studies conducted in PNG in the
periods 1970s–2000s, 2007–10 and 2011–15.
4,8,27,28
All
frequencies, percentages and tables were produced in this
study using Statistical Package for Social Science (SPSS).
Results
To assess the completeness of the collected mortality data,
the reported crude death rate of 6.5 per 1000 population
in PNG for the period 2019–20 was used.
29
With a popula-
tion coverage of approximately 80 000 people and an esti-
mate of 20% of deaths that occurred in the health facilities
and were not included in the study, 1040 deaths might be
expected to have occurred in the communities within
the surveillance sites over the data collection period
(0.0065*80000*0.08*2.5). This is a good indicator of the
level of deaths expected in the surveillance populations
over the study period. However, the mortality data used in
this study also included deaths that occurred up to 2 years
prior to the study data collection; so the dataset of 1021
VA interviews conducted in this study is only a representa-
tive sample of all deaths expected in the surveillance sites.
Among the 1021 deaths identified, consents were
obtained for conducting 1003 VA interviews resulting in a
participation rate of 98%. All VA interview data were in-
cluded in COD analysis. The Inter VA-5 method success-
fully assigned specific causes of death for 926 deaths
6International Journal of Epidemiology, 2022, Vol. 00, No. 00
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
(92%), and but not for 77 other deaths (8%). These deaths
were excluded from the InterVA-5 modelling by the
InterVA-5 program because they were ineligible due to
poor data quality. No specific COD was assigned to these
deaths. There were 19 deaths having medical records, ac-
counting for 2% of total death records in the study, but
only one death reported an immediate COD during the in-
terview. The other 18 deaths had no specific COD, sug-
gesting the information on COD from the death
certificates and medical records had little impact on perfor-
mance of the InterVA-5.
Overall distribution of causes of death in the
surveillance population
As showed in Table 2, Inter VA-5 assigned specific causes
of death for 926 deaths, including 514 male deaths
(55.5%) and 412 female deaths (44.5%) (similar sex distri-
bution was observed in the 1021 VA interview data). The
proportions of interviews for male deaths were higher than
female deaths across provinces, except for ESP where the
male deaths accounted for 49% compared with 51% of fe-
male deaths. These data indicate more males than females
living in the surveillance sites, which is consistent with the
sex ratio in the surveillance population, 108 males per 100
females.
30
Most interviews were completed between 49 and
96 weeks after the date of death (42%), followed by
97–104 weeks (22%), with few interviews conducted
within 2–4 weeks and 5–12 weeks (4%). The longest dura-
tion from the date of death to the date of interview com-
pleted was 112 weeks. Lack of information on date of
death and/or date of interview (either day or month or
year) was treated as missing data in the InterVA-5 model.
Table 2 Sex and urban-rural residence of the deceased, and duration from date of death to date of completion of verbal autopsy
interview (in weeks) by province, Papua New Guinea Institute of Medical Research’s Comprehensive Health and
Epidemiological Surveillance System, 2018–20
n(%) Central Port
Moresby
Eastern
Highlands
Madang East
Sepik
East New
Britain
All provinces
Sector Urban – 30 62 67 – 69 228
– 100.0% 22.2% 88.2% – 59.5% 25.2%
Rural 288 – 217 9 116 47 677
100.020% – 77.8% 11.8% 100.0% 40.5% 74.8%
Total 288 30 279
a
76 116 116 905
100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0%
Sex of deceased Male 153 17 % 51 57 67 514
53.1% 56.7% 56.3% 67.1% 49.1% 57.8% 55.5%
Female 135 13 131 25 59 49 412
46.9% 43.3% 43.7% 32.9% 50.9% 42.2% 44.5%
Total 288 30 300 76 116 116 926
100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0%
Interview com-
pletion after
death
0–4 weeks 7 2 14 2 5 3 33
2.6% 7.4% 4.9% 2.7% 4.5% 2.7% 3.7%
5–12 weeks 8 – 12 1 9 8 38
2.9% – 4.2% 1.4% 8.0% 7.1% 4.3%
13–24 weeks 16 – 30 7 14 14 81
5.9% – 10.5% 9.6% 12.5% 12.5% 9.2%
25–48 weeks 31 6 59 16 31 25 168
11.4% 22.2% 20.7% 21.9% 27.7% 22.3% 19.1%
49–96 weeks 109 14 126 37 37 45 368
40.1% 51.9% 44.2% 50.7% 33.0% 40.2% 41.8%
97–104 weeks 28 – 10 7 7 6 58
10.3% – 3.5% 9.6% 6.3% 5.4% 6.6%
105–112 weeks 73 5 34 3 9 11 135
26.8% 18.5% 11.9% 4.1% 8.0% 9.8% 15.3%
Total 272 27 285 73 112 112 881
b
100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0%
a
21 deaths were missing information on urban-rural sector.
b
45 deaths were missing information on the date of death or/and date of interview, either day or/and month or/and year.
International Journal of Epidemiology, 2022, Vol. 00, No. 00 7
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
Table 3 InterVA-5 assigned causes of death in the population (all ages) according to International Classification of Diseases by province, Papua New Guinea Institute of Medical
Research’s Comprehensive Health and Epidemiological Surveillance System, 2018–20
Cause of death Central Port Moresby Eastern Highlands Madang East Sepik East New Britain All province
01.Infectious and parasitic disease 96 (33.1%) 6 (20.0%) 104 (34.6%) 28 (36.7%) 35 (30.2%) 48 (41.3%) 317 (34.2%)
01.01 Sepsis (non-obstetric) 1 (0.3%) – – – 2 (1.7%) 1 (0.9%) 4 (0.4%)
01.02 ARTI including pneumonia 24 (8.3%) 2 (6.7%) 26 (8.7%) 7 (9.2%) 6 (5.2%) 16 (13.8%) 81 (8.7%)
01.03 HIV/AIDS-related death 13 (4.5%) 2 (6.7%) 36 (12.0%) 8 (10.5%) 8 (6.9%) 8 (6.9%) 75 (8.1%)
01.04 Diarrhoeal diseases 3 (1.0%) – 6 (2.0%) 3 (3.9%) 7 (6.0%) 3 (2.6%) 22 (2.4%)
01.05 Malaria 1 (0.3%) – 1 (0.3%) 2 (2.6%) – 4 (3.4%) 8 (0.9%)
01.06 Measles – – – – – – –
01.07 Meningitis and encephalitis 5 (1.7%) 1 (3.3%) 1 (0.3%) – 1 (0.9%) 2 (1.7%) 10 (1.1%)
01.08 Tetanus – – – – – – –
01.09 Pulmonary tuberculosis 41 (14.2%) 1 (3.3%) 25 (8.3%) 8 (10.5%) 2 (1.7%) 7 (6.0%) 84 (9.1%)
01.10 Pertussis – – – – 1 (0.9%) 2 (1.7%) 3 (0.3%)
01.11 Haemorrhagic fever – – – – – – –
01.12 Dengue fever – – – – – – –
01.99 Other unspecified infect diseases 8 (2.8%) – 9 (3.0%) – 8 (6.9%) 5 (4.3%) 30 (3.2%)
02.Neoplasms 56 (19.5%) 8 (26.6%) 48 (16.0%) 10 (13.1%) 28 (24.1%) 11 (9.5%) 161 (17.5)
02.01 Oral neoplasms 3 (1.0%) – 3 (1.0%) – 1 (0.9%) – 7 (0.8%)
02.02 Digestive neoplasms 19 (6.6%) 1 (3.3%) 18 (6.0%) 3 (3.9%) 20 (17.2%) 8 (6.9%) 69 (7.5%)
02.03 Respiratory neoplasms 10 (3.5%) 6 (20.0%) 10 (3.3%) 4 (5.3%) 3 (2.6%) 2 (1.7%) 35 (3.8%)
02.04 Breast neoplasms – – 3 (1.0%) – – – 3 (0.3%)
02.05 Reproductive neoplasms M/F 10 (3.5%) 1 (3.3%) 9 (3.0%) 1 (1.3%) 2 (1.7%) – 23 (2.5%)
02.99 Other and unspecified neoplasms 14 (4.9%) – 5 (1.7%) 2 (2.6%) 2 (1.7%) 1 (0.9%) 24 (2.6%)
03.Nutritional endocrine disorders 11 (3.8%) 4 (13.3%) 12 (4.0%) 1 (1.3%) 1 (0.9%) – 29 (3.1%)
03.01 Severe anaemia – – – – – – –
03.02 Severe malnutrition – – 5 (1.7%) – – – 5 (0.5%)
03.03 Diabetes mellitus 11 (3.8%) 4 (13.3%) 7 (2.3%) 1 (1.3%) 1 (0.9%) – 24 (2.6%)
04.Cardiovascular diseases 55 (19.1%) 5 (16.7%) 57 (19.0%) 9 (11.8%) 27 (23.3%) 30 (25.9%) 183 (19.8%)
04.01 Acute cardiac disease 17 (5.9%) 2 (6.7%) 21 (7.0%) 5 (6.6%) 4 (3.4%) 14 (12.1%) 63 (6.8%)
04.02 Stroke 20 (6.9%) 1 (3.3%) 17 (5.7%) 3 (3.9%) 9 (7.8%) 8 (6.9%) 58 (6.3%)
04.03 Sickle cell with crisis – – – – – – –
04.99 Other unspecified cardiac diseases 18 (6.3%) 2 (6.7%) 19 (6.3%) 1 (1.3%) 14 (12.1%) 8 (6.9%) 62 (6.7%)
05.Chronic respiratory disorders 2 (0.7%) 1 (3.3%) 5 (1.7%) – – 3 (2.6%) 11 (1.2%)
05.01 Chronic obstructive pulmonary
diseases
2 (0.7%) 1 (3.3%) 5 (1.7%) – – 2 (1.7%) 10 (1.1%)
05.02 Asthma – – – – – 1 (0.9%) 1 (0.1%)
06.Gastrointestinal disorders 3 (1.0%) – 15 (5.0%) 3 (3.9%) 2 (1.7%) 5 (4.3%) 28 (3.0%)
06.01 Acute abdomen 1 (0.3%) – – 1 (1.3%) 2 (1.7%) 2 (1.7%) 6 (0.6%)
(Continued)
8International Journal of Epidemiology, 2022, Vol. 00, No. 00
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
Table 3 Continued
Cause of death Central Port Moresby Eastern Highlands Madang East Sepik East New Britain All province
06.02 Liver cirrhosis 2 (0.7%) – 15 (5.0%) 2 (2.6%) – 3 (2.6%) 22 (2.4%)
07.Renal disorders 2 (0.7%) – 1 (0.3%) 1 (1.3%) – 2 (1.7%) 6 (0.6%)
07.01 Renal failure 2 (0.7%) – 1 (0.3%) 1 (1.3%) – 2 (1.7%) 6 (0.6%)
08.Mental nervous system disorders 7 (2.4%) – 7 (2.4%) – 2 (1.7%) 2 (1.7%) 18 (2.0%)
08.01 Epilepsy 2 (0.7%) – 5 (1.7%) – 0 (0.0%) 0 (0.0%) 7 (0.8%)
08.99 Other unspecified NCDs 5 (1.7%) – 2 (0.7%) – 2 (1.7%) 2 (1.7%) 11 (1.2%)
09. Pregnancy, childbirth disorders 2 (0.6%) – 2 (0.7%) 3 (3.9%) 3 (2.7%) 1 (0.9%) 11 (1.1%)
09.02 Abortion-related death 1 (0.3%) – – 1 (1.3%) – – 2 (0.2%)
09.03 Pregnancy-induced hypertension – – – – 1 (0.9%) – 1 (0.1%)
09.04 Obstetric haemorrhage 1 (0.3%) – 2 (0.7%) 2 (2.6%) 1 (0.9%) – 6 (0.6%)
09.06 Pregnancy-related sepsis – – – – – 1 (0.9%) 1 (0.1%)
09.99 Other unspecified maternal causes – – – – 1 (0.9%) – 1 (0.1%)
10. Neonatal causes of death 12 (4.2%) – 1 (0.3%) – 4 (3.5%) 2 (1.8%) 19 (2.0%)
10.01 Prematurity – – – – 1 (0.9%) 1 (0.9%) 2 (0.2%)
10.02 Birth asphyxia – – 1 (0.3%) – 2 (1.7%) – 3 (0.3%)
10.03 Neonatal pneumonia – – – – 1 (0.9%) – 1 (0.1%)
10.04 Neonatal sepsis – – – – – – –
10.05Neonatal tetanus – – – – – – –
10.06 Congenital malformation 12 (4.2%) – – – – 1 (0.9%) 13 (1.4%)
10.99 Other unspecified peri-natal death – – – – – – –
11. Stillbirths – – 2 (0.7%) 1 (1.3%) 2 (1.7%) – 5 (0.5%)
11.01 Fresh stillbirth – – 2 (0.7%) 1 (1.3%) 2 (1.7%) – 5 (0.5%)
11.02 Macerated stillbirth – – – – – – –
12. External causes of death 34 (11.6%) 4 (13.3%) 34 (11.3%) 17 (22.4%) 9 (7.8%) 9 (8.0%) 107 (11.4%)
12.01 Road traffic accident 13 (4.5%) 2 (6.7%) 13 (4.3%) 10 (13.2%) 1 (0.9%) 3 (2.6%) 42 (4.5%)
12.02 Other transport accident 1 (0.3%) – 2 (0.7%) – – – 3 (0.3%)
12.03 Accident fall 5 (1.7%) – 5 (1.7%) – 1 (0.90%) – 11 (1.2%)
12.04 Accidental drowning, submersion 5 (1.7%) 1 (3.3%) 1 (0.3%) – – 1 (0.9%) 8 (0.9%)
12.05 Accidental exposure to smoke/fire 1 (0.3%) – – 1 (1.3%) – 1 (0.9%) 3 (0.3%)
12.06 Venomous animals, plants – – – – – 1 (0.9%) 1 (0.1%)
12.07 Accidental poisoning, noxious
substance
–– –––––
12.08 Intentional self-harm – – 2 (0.7%) – – – 2 (0.2%)
12.09 Assault 8 (2.8%) 1 (3.3%) 6 (2.0%) 4 (5.3%) 7 (6.0%) 1 (0.9%) 27 (2.9%)
12.10 Exposure to force of nature 1 (0.3%) – 4 (1.3%) – – 1 (0.9%) 6 (0.6%)
12.99 Other unspecified external causes – – 1 (0.3%) 2 (2.6%) – 1 (0.9%) 4 (0.4%)
99. Unknown 8 (2.8%) 2 (6.7%) 12 (4.0%) 3 (3.9%) 3 (2.6%) 3 (2.6%) 31 (3.3%)
99. Cause of death unknown 8 (2.8%) 2 (6.7%) 12 (4.0%) 3 (3.9%) 3 (2.6%) 3 (2.6%) 31 (3.3%)
Total 288 (100%) 30 (100%) 300 (100%) 76 (100%) 116 (100%) 116 (100%) 926 (100%)
InterVA-5 is an analytical tool for causes of death.
ARTI, acute respiratory tract infection; M/F, male and female; ACDs, acute cardiac diseases; NCD,: non-communicable diseases.
International Journal of Epidemiology, 2022, Vol. 00, No. 00 9
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
As a result, 45 deaths had missing data on the duration, ac-
counting 5% of the total deaths. These deaths were treated
as ‘missing data’ and excluded from the calculation of time
duration (see footnote in Table 2). Assuming data reporters
followed the VA interview selection criteria, and with their
knowledge of the local people, there were reasons to be-
lieve these deaths occurred within the study period and the
VA interviews of these deaths were completed within the
112 weeks from the date of death to the date of interview.
Table 3 shows the overall distribution of specific causes
of death assigned for 926 deceased, including 288 in
Central Province (24.6%), 30 in POM (3.2%), 300 in EHP
(32.4%), 76 in Madang (8.2%), 116 in ENB (12.5%) and
116 in ESP (12.5%). The highest proportion of deaths was
attributed to NCDs (47% of the total deaths at all ages in
the population), followed by infectious and parasitic dis-
eases (34%), injuries and external causes (11%) and ma-
ternal and neonatal deaths (4%). The leading causes of
death from NCDs were cardiovascular diseases, account-
ing for 20% of the total deaths, with the highest propor-
tion reported in ENB (25%) and lowest in Madang (12%).
Neoplasms were the second leading cause of death, which
accounted for nearly 18% of the total deaths, with the
highest reported in POM (27%) and lowest in ENB (about
10%).
Among infectious diseases, pulmonary TB accounted
for the highest proportion of deaths, 9% for all provinces,
ranging from 14% in Central to 2% in EHP. This was fol-
lowed by HIV/AIDS, which accounted for more than 8%
of the total deaths, with the highest proportion reported in
EHP (12%) and lowest in Central Province (5%).
Leading causes of death among children
Table 4 shows the analysis of the 93 child deaths. Specific
causes of death among the children were ranked from the
highest to the lowest proportions. Only the five leading
causes of death which accounted for the highest propor-
tions of child deaths were selected for presentation and
stratified by urban-rural sector, province, age group and
sex of the deceased. Infectious diseases were the most com-
mon cause of death among children. The leading causes of
death identified were ARTIs (including pneumonia) and
diarrhoeal diseases; each were responsible for 12 out of 93
child deaths (nearly 13%). Notably, ARTIs were attributed
to four out of 15 child deaths (27%) in urban areas and 8
out of 73 child deaths (11%) in rural areas. Diarrhoeal dis-
eases were assigned as COD for five out of 56 male child
deaths (9%) and seven out of 37 female child deaths
(19%), with similar proportion between urban-rural
sectors (13% and 12%, respectively). This disease
attributed to six out of 67 child deaths (9%) in the age
group 0–4 years. Given the fact that the numbers of obser-
vations are small and the figures represent only a fraction
of the child deaths that occurred in the communities within
these surveillance sites during the data collection period,
interpretation of the results should be cautious.
Leading causes of death among people of
working age
Table 5 shows the leading causes of death among the
working population aged 15–64 years. A total of 574
deaths were included in the analyses. Pulmonary TB was
the leading COD, responsible for 12% of total deaths in
this population. This disease killed more adults in rural
areas than in urban areas: 13% and 9%, respectively. The
number of people who died from pulmonary TB was high-
est in Central province (18%). Females were more likely to
die from pulmonary TB (15%) than male counterparts
(9%).
HIV/AIDS was the second leading COD in the working
population, contributing to 11% of the total deaths across
the surveillance sites, with the highest death proportion in
EHP (17%), followed by Madang (14%). HIV/AIDS con-
tributed 12% of adult deaths in rural areas and 8% in ur-
ban areas. The proportion of females who died from HIV/
AIDS was double that for males at 15% and 8%, respec-
tively. Notably, road traffic accidents contributed 6.5% of
deaths in the population of working age, and were respon-
sible for 10% of deaths in the urban areas and 20% of
deaths in Madang.
Leading causes of death among the elderly
Table 6 shows the leading causes of death among the el-
derly. Data from 256 elderly deaths were included in the
analyses. ARTIs were the leading cause of death, account-
ing for 13% of the total elderly deaths, with the highest
proportion reported in EHP and ENB (16%) and among
the elderly aged 85þyears (25%). The proportion of el-
derly deaths from ARTIs was slightly higher in rural areas
than in urban areas at 13% and 11%, respectively, but the
difference between males and females was not significant.
Two notable NCD-attributed causes of death were di-
gestive neoplasms and ACDs, each accounting for 10% of
the total elderly deaths. The proportion of the elderly who
died from digestive neoplasms was higher in rural areas
(13%) and particularly in ESP (24%). In contrast, deaths
attributed to ACDs were higher in urban areas (18%) and
in the ENB and Madang provinces at 16% and 14%,
respectively.
10 International Journal of Epidemiology, 2022, Vol. 00, No. 00
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
Changing causes of death in the mortality
transition in PNG
Table 7 compares the mortality data from previous studies
in PNG using different data sources, VA interview tools
and analytical methods. The summary data show changes
in distribution of the five main categories, highlighting the
trend in causes of death among the adult population over
the past 50 years. The classification of five main groups of
causes of death emerging and endemic infections, emerging
and endemic NCDs, and injuries and external causes was
used in previous studies in PNG to demonstrate the shift in
trends in mortalities over the time.
4,8,14,27,28
The proportion of deaths from endemic infectious dis-
eases declined from 34% in the 1970–2000s to 14% in the
period 2016–20. By contrast, the proportion of adult
deaths from emerging infections, including HIV/AIDS and
TB, increased from 3% in the 1970s–2000s to 12% in the
2010–15 and 19% in 2016–20. Notably, mortality from
emerging NCDs such as ACDs, stroke and diabetes in-
creased dramatically over the past 50 years, from a very
low level of 0.5% in the period 1970s–2000s to 15% in
the period 2010–15 and 16% in the period 2016–20. The
proportion of deaths due to injuries and external causes
also increased from 9% in the period 1970s–2000s to 11%
in the period 2016–20. It is noted that the data from
Population Health Metrics Research Consortium
(PHMRC), Medical Certification of Cause of Death
(MCCD) and CHESS were collected in almost the same pe-
riod, 2018–20.
Since this study was conducted during the COVID-19
outbreaks in PNG, we attempted to address the question
whether any COVID-19 deaths in the communities were
possibly captured in the mortality data. The first COVID-
19 case was officially reported in PNG on the 20 March
2020. As of 17 February 2021 when data collection for
this study was completed, nine COVID-19 deaths were
reported nationwide, all from the General Hospital in Port
Moresby; and 914 people tested positive with PCR.
31
COVID-19 deaths were not included in the InterVA-5 and
no COVID-19 testing was conducted in the communities
at that time. The CHESS clinical team conducted further
investigations of all 81 deaths attributed to ARTIs as the
primary COD, using the conventional medical review
method.
As a result, two COVID-19 suspected deaths were iden-
tified: one in Hiri, Central Province, and another in
Kokopo, ENB. Both the deceased were above 50 years of
age at death, with background conditions of lung cancer
and diabetes which were identified as the secondary cause
of death by InterVA-5. The deceased had developed flu-
like clinical signs in the weeks prior to death.
14
In this
study, deaths were primarily ascribed to ARTIs including
pneumonia. Since the number of observations on ARTI-
ascribed deaths was relatively low in a short period, it
might be inadequate for comment on the trend in ARTIs
deaths during the COVID-19 outbreaks in PNG.
The proportions of indeterminate causes of death have
been lower (5%) in more recent studies, likely reflecting
the improvements in COD studies conducted in the 2010s
when more advanced techniques were applied in VA stud-
ies in many LMICs, including PNG. New VA data collec-
tion tools, COD analytical methods and national capacity
in analysing and reporting causes of death have improved
in PNG, particularly in the period 2011–20.
14,32
The
WHO 2016 VA tool was adapted for optimal use in the
PNG local context, where new household and individual
identification coding systems were created and applied
consistently across surveillance sites. This facilitates link-
ages between mortality data and other data components
such as household socioeconomic status (SES), morbidity
and demographic change data, enhancing the scope of
COD analyses.
Discussion
The findings of this study have provided further insights
into the mortality transition in PNG, reaffirming the over-
all trend in declining mortality attributed to infectious dis-
eases but increased mortality from NCDs over the past
50 years. However the observed mortality trends were also
different within these broad COD categories, with an in-
crease in mortalities from emerging infections but a decline
in endemic infections, particularly over the past 20 years.
The current study has identified the leading causes of death
in different sub-populations, with ARTIs and diarrhoeal
diseases as leading causes of death among children (aged
0–14 years), TB and HIV/AIDS among the working popu-
lation (aged 15–64 years) and digestive neoplasms and
ACDs among the elderly.
Epidemiological characteristics of mortality
transition in PNG
Among infectious diseases, ARTIs were the leading cause
of death in the surveillance population and particularly
among children and the elderly (see Tables 3,4and 6).
These findings are supported by previous studies on mor-
bidity surveillance at primary health facilities in PNG.
33,34
ARTIs and pneumonia were previously reported as causing
high mortality in the highlands region, and the mortality
rate of ARTIs in the EHP was higher than the national
level in the 1990s.
35,36
In this study we found that ARTIs
were also the leading cause of death in the coastal areas
International Journal of Epidemiology, 2022, Vol. 00, No. 00 11
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
and the islands region. Effective treatment and manage-
ment of ARTIs and pneumonia are heavily reliant on the
primary health care services, particularly on the knowledge
of danger signs among child caregivers and access to effec-
tive antibiotics upon admission to health facilities.
37
The
proportion of deaths from endemic infectious diseases
which decreased across PNG regions is likely due to in-
creased access to primary health care services, among other
preventive measures, over the past 50 years.
38
The use of
immunization services, antenatal care and the availability
of preventive medicines like antibiotics, antimalarials, anti-
helminthics and antiprotozoals, have contributed to the
decline in morbidity associated with infectious, parasitic
and childhood communicable diseases.
39
More efforts are
needed to further improve access to and use of primary
health care services, to further reduce the mortality from
infectious diseases in the next decade. Public health educa-
tion to improve access to water, sanitation and hygiene is a
key intervention in all provinces and urban-rural areas to
reduce morbidity of infectious diseases.
The increased mortality from TB and HIV/AIDS was
the main driver underlying the current mortality transition
in PNG. These diseases were the leading causes of death in
the working population, contributing 23% of the total
deaths in this population (see Table 5). TB was first
reported in PNG in the late 19th century,
40
and the first
case of HIV infection in PNG was reported in 1987.
41
TB
was identified as the leading cause of death in the adult
population in the period 2010–15.
4
Co-infections of TB
and HIV/AIDS were reported in recent studies.
42–44
The re-
cent emergence of TB has been complicated with the resur-
gence of leprosy among young adults,
45
raising public
health concerns over multidrug resistance and the in-
creased burdens of disease on the health systems.
Managing infectious diseases is overwhelming the health
system and, bordering on the current health crisis
Table 4 Distribution of five leading causes of deaths among children (aged 0–14years) by urban-rural sector, province, age
group and sex of the deceased, Papua New Guinea Institute of Medical Research’s Comprehensive Health and Epidemiological
Surveillance System, 2018–20
Endemic infectious diseases All other causes
of death
Total all causes
of death
n(%) ARTIs Diarrhoea Meningitis Sepsis Pertussis
Sector
a
Urban 4 2 1 1 2 5 15
26.7% 13.3% 6.7% 6.7% 13.3% 33.3% 100.0%
Rural 8 9 5 2 1 48 73
11.0% 12.3% 6.8% 2.7% 1.4% 65.8% 100.0%
Province Central 3 2 3 1 – 22 31
9.7% 6.5% 9.7% 3.2% – 71.0% 100.0%
Port Moresby – – 1 – – – 1
– – 100.0% – – – 100.0%
Eastern
Highlands
13 1–– 18 23
4.3% 13.0% 4.3% – – 78.3% 100.0%
Madang 2 2 – – – 1 5
40.0% 40.0% – – – 20% 100.0%
East Sepik 2 3 1 1 1 10 18
11.1% 16.7% 5.6% 5.6% 5.6% 55.6% 100.0%
East New Britain 4 2 – 1 2 6 15
26.7% 13.3% – 6.7% 13.3% 40% 100.0%
Age group 0–4 8 6 5 3 3 42 67
11.9% 9.0% 7.5% 4.5% 4.5% 62.7% 100.0%
5–14 4 6 1 – – 15 26
15.4% 23.1% 3.8% – – 57.7% 100.0%
Sex Male 7 5 1 1 1 41 56
12.5% 8.9% 1.8% 1.8% 1.8% 73.2% 100.0%
Female 5 7 5 2 2 16 37
13.5% 18.9% 13.5% 5.4% 5.4% 43.2% 100.0%
Total 12 12 6 3 3 57 93
12.9% 12.9% 6.5% 3.2% 3.2% 61.3% 100.0%
a
Five child deaths had missing information on urban-rural sector.
12 International Journal of Epidemiology, 2022, Vol. 00, No. 00
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
Table 5 Distribution of seven leading causes of deaths among people of working age (15–64 years) by urban-rural, province, age group and sex of the deceased, Papua New
Guinea Institute of Medical Research’s Comprehensive Health and Epidemiological Surveillance System, 2018–20
Main groups of causes of death Emerging infections Endemic
infections
Endemic
NCDs
Emerging
NCDs
Injuries
Leading causes of death n(%) Pulmonary
tuberculosis
HIV/AIDS ARTIs Digestive neoplasms ACDs
Stroke
Road traffic accident Other causes
of death
Total all causes
of death
Sector Urban 14 12 13 7 15 5 15 68 149
9.4% 8.1% 8.7% 4.7% 10.1% 3.4% 10.1% 45.6% 100.0%
Rural 53 49 21 35 24 31 22 177 412
12.9% 11.9% 5.1% 8.5% 5.8% 7.5% 5.3% 43.0% 100.0%
Province Central 37 11 15 13 14 11 12 88 201
18.4% 5.5% 7.5% 6.5% 7.0% 5.5% 6.0% 43.8% 100.0%
Port Moresby 1 2 2 1 2 0 2 14 24
4.2% 8.3% 8.3% 4.2% 8.3% 0.0% 8.3% 58.3% 100.0%
Eastern
Highlands
15 30 9 10 9 15 10 75 173
8.7% 17.3% 5.2% 5.8% 5.2% 8.7% 5.8% 43.4% 100.0%
Madang 7 7 2 2 2 2 10 17 49
14.3% 14.3% 4.1% 4.1% 4.1% 4.1% 20.4% 34.7% 100.0%
East Sepik 2 5 2 10 3 5 1 29 57
3.5% 8.8% 3.5% 17.5% 5.3% 8.8% 1.8% 50.9% 100.0%
East New Britain 6 8 7 7 9 4 3 26 70
8.6% 11.4% 10.0% 10.0% 12.9% 5.7% 4.3% 37.1% 100.0%
Age group 15–24 7 6 4 3 1 5 8 28 62
11.3% 9.7% 6.5% 4.8% 1.6% 8.1% 12.9% 45.2% 100.0%
25–34 20 18 4 4 1 2 5 45 99
20.2% 18.2% 4.0% 4.0% 1.0% 2.0% 5.1% 45.5% 100.0%
35–44 17 11 11 6 3 1 8 37 94
18.1% 11.7% 11.7% 6.4% 3.2% 1.1% 8.5% 39.4% 100.0%
45–54 9 14 9 18 14 12 7 66 149
6.0% 9.4% 6.0% 12.1% 9.4% 8.1% 4.7% 44.4% 100.0%
55–64 15 14 9 12 20 17 10 73 170
8.8% 8.2% 5.3% 7.1% 11.8% 10.0% 5.9% 42.9% 100.0%
Sex Male 29 23 20 32 30 21 35 116 306
9.5% 7.5% 6.5% 10.5% 9.8% 6.9% 11.4% 37.9% 100.0%
Female 39 40 17 11 9 16 3 132 268
14.6% 14.9% 6.3% 4.1% 3.4% 6.0% 1.1% 49.3% 100.0%
Total 68 63 37 43 39 37 38 249 574
11.8% 11.0% 6.4% 7.5% 6.8% 6.4% 6.6% 43.4% 100.0%
ARTIs, acute respiratory tract infections; ACDs, acute cardiac diseases; NCD, non-communicable diseases.
International Journal of Epidemiology, 2022, Vol. 00, No. 00 13
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
Table 6 Distribution of six leading causes of death among the elderly (aged 65þyears) by urban-rural sector, province, age group and sex of the deceased, Papua New Guinea
Institute of Medical Research’s Comprehensive Health and Epidemiological Surveillance System, 2018–20
Main group of causes of death Endemic infections Endemic NCDs Emerging NCDs
Leading causes of death n(%) ARTIs Digestive neoplasms Respiratory neoplasms ACDs Stroke Diabetes mellitus Other causes
of death
Total all causes
of death
Sector Urban 7 2 5 11 4 2 31 62
11.3% 3.2% 8.1% 17.7% 6.5% 3.2% 50.0% 100.0%
Rural 25 24 11 12 15 6 98 191
13.1% 12.6% 5.8% 6.3% 7.9% 3.1% 51.3% 100.0%
Province Central 6 6 4 3 9 3 25 56
10.7% 10.7% 7.1% 5.4% 16.1% 5.4% 44.6% 100.0%
Port Moresby – – 2 – 1 – 2 5
– – 40.0% – 20.0% – 40.0% 100.0%
Eastern
Highlands
16 8 4 12 1 3 58 102
15.7% 7.8% 3.9% 11.8% 1.0% 2.9% 56.9% 100.0%
Madang 3 1 3 3 1 1 9 21
14.3% 4.8% 14.3% 14.3% 4.8% 4.8% 42.9% 100.0%
East Sepik 2 10 2 1 3 1 22 41
4.9% 24.4% 4.9% 2.4% 7.3% 2.4% 53.7% 100.0%
East New Britain 5 1 1 5 4 – 15 31
16.1% 3.2% 3.2% 16.1% 12.9% – 48.4% 100.0%
Age group 65–74 15 17 11 13 11 3 73 143
10.5% 11.9% 7.7% 9.1% 7.7% 2.1% 51% 100.0%
75–84 7 8 6 7 3 5 38 74
9.5% 10.8% 8.1% 9.5% 4.1% 6.8% 51.4% 100.0%
85þ10 1 – 4 5 – 19 39
25.6% 2.6% – 10.3% 12.8% – 48.7% 100.0%
Sex Male 19 17 10 15 8 5 76 150
12.7% 11.3% 6.7% 10.0% 5.3% 3.3% 50.7% 100.0%
Female 13 9 7 9 11 3 54 106
12.3% 8.5% 6.6% 8.5% 10.4% 2.8% 50.9% 100.0%
Total 32 26 17 24 19 8 130 256
12.5% 10.2% 6.6% 9.4% 7.4% 3.1% 50.8% 100.0%
ARTIs, acute respiratory tract infections; ACDs, acute cardiac diseases; NCDs, non-communicable diseases.
14 International Journal of Epidemiology, 2022, Vol. 00, No. 00
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
associated with COVID-19 outbreaks across the country,
posing a threat to the sustainable development of PNG.
Chronic infectious diseases such as TB, HIV and leprosy
are often treated as outpatients when the diseases are sta-
ble, with regular health checks. These patients are more
likely to die in the communities, as they are discharged
from hospitals in the later stage and the loss to follow-up is
high in PNG.
45,46
Pulmonary TB was not among the top six leading causes
of death in the elderly population. The InterVA-5 program
ascribed specific causes of death systematically, so the
potential for misclassifications of the two diseases: pulmo-
nary TB and chronic obstructive pulmonary disease among
the elderly was minimal. The possible explanation for this
observation could be that most TB patients might have
died before they reached the age of 65 years, resulting in a
low number of TB deaths in the elderly. PNG reported
a high TB incidence rate of 432 per 100 000 population in
the period 2008–16.
47
However, the TB notification rate
among the elderly was as low as 60 among males and 30
among females per 100 000 population (aged 65þyears)
in 2016.
42
This study showed that mortality from NCDs, including
emerging and endemic NCDs, accounted for 47% of total
deaths at all ages but more than 50% among adult deaths
(aged 15þyears), an increase of 10% over the past
50 years. This finding is consistent with the previous VA
study in the 2010s.
27,28,48
There was no record of
ischaemic heart disease in the 1960s, and little evidence of
stroke with a low prevalence of only 1% of hospital deaths
in the 1970s.
11
The PNG Cancer Registry in the period
1979–88 showed the most common cancers were digestive
neoplasms, including oral cancer, hepatoma, and stomach,
rectum, colon and bladder malignant tumours.
49
These dis-
eases remain high in the current data. The prevalence of
diabetes mellitus was at 1% of the adult participants in
Madang Province and 0% in EHP, according to the study
conducted in 2010–13,
9
but the current study found mor-
tality attributed to diabetes mellitus accounted for 3% of
the elderly deaths. This suggests prevalence of diabetes
mellitus could have increased among the adult population
in the 2010s.
As shown in Table 7, the mortality data from the
Discharge Health Information System (DHIS), which was
designed to reports deaths in health facilities including 20
tertiary hospitals and 635 health centres in PNG,
8
showed
that the composition of total deaths reported in the period
2007–10 was 45% from infectious diseases, which was
higher than the 33% from the current data. The opposite
was true for NCDs, with 44% reported from the DHIS
compared with 51% in this study. These figures show dif-
ferent mortalities from the two data sources. Whereas the
health facilities-based data showed a higher mortality from
infectious diseases and lower mortality from NCDs, the
communities-based data in the current study showed the
opposite, with lower mortality from infectious diseases
Table 7 Shifting causes of deaths among adult population in the mortality transition in PNG in the period 1970–20 by data
sources, Papua New Guinea Institute of Medical Research’s Comprehensive Health and Epidemiological Surveillance System,
2018–20
Main groups of
causes of death
1970s–2000s,
VA studies
4
2007–10,
DHIS
8
2011–15,
iHDSS
4
2018–19,
PHMRC
28
2017–20,
MCCD
27
2018–20,
CHESS
Emerging
infections
2.7% 12.0% 11.9% 8.3% 10.7% 18.7%
Endemic
infections
33.9% 33.0% 9.5% 18.7% 18.4% 14.3%
All infections 36.6% 45.0% 21.4% 27.0% 29.2% 33.0%
Emerging NCDs 0.5% 15.5% 14.7% 21.4% 14.4% 15.7%
Endemic NCDs 40.1% 28.0% 40.3% 29.4% 46.1% 35.3%
All NCDs 40.5% 43.5% 55.0% 50.8% 50.4% 51.0%
Injuries 8.8% 10.5% 19.10% 13.0% 16.3% 11.3%
Indeterminate 15.6% 1.0% 19.2% 9.2% 4.1% 4.7%
Total 100% 100% 100% 100% 100% 100.00%
Emerging infections include tuberculosis and HIV/AIDS. Endemic infections include sepsis, acute respiratory infections and pneumonia, typhoid, malaria, diar-
rhoeal diseases, measles, meningitis and encephalitis, tetanus. Emerging NCDs include acute cardiac diseases, stroke, diabetes. Endemic NCDs include neoplasms,
chronic obstructive pulmonary diseases, asthma, gastrointestinal disorders, renal disorders, mental and nervous system disorders, malnutrition and endocrine dis-
orders, maternal deaths. Injuries and external causes include road traffic accidents, accidental fall, drowning and submersion, exposure to smoke and fire, venom-
ous animals and plants, poisoning, noxious substance, intentional self-harm, assault, force of nature.
VA, verbal autopsy; DHIS, Discharge Health Information System; iHDSS, integrated Health and Demographic Surveillance System; PHMRC, Population
Health Metrics Research Consortium; MCCD, Medical Certification of Cause of Death; CHESS, Comprehensive Health and Epidemiological Surveillance
System; NCDs, non-communicable diseases.
International Journal of Epidemiology, 2022, Vol. 00, No. 00 15
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
and higher mortality from NCDs. Possible explanations
for this could be the access and use of health care services
in PNG. Tertiary health care services are only available in
large urban areas,
18
which is an important factor for utili-
zation of health care services in the population.
Compared with more recent mortality data from the
MCCD
27
and the PHMRC,
28
the findings of the current
study show a slightly higher proportion of deaths due to in-
fectious diseases (33.0%) and a similar proportion of
deaths due to NCDs (51.0%) (see Table 7). The consis-
tency in mortality data across these three data sources sug-
gest that the Global Burden of Disease Study (GBD) 2017
underestimated deaths from infectious diseases (16.5%)
but overestimated mortality from NCDs (70.5%) in
PNG.
2,3
For injuries and external causes of death, CHESS data
showed a lower proportion than the MCCD data, 11%
compared with 16%. The difference could be due to
CHESS data being from the communities whereas MCCD
data were mostly from tertiary provincial hospitals
27
(see
Table 7). The trend of increased mortality from road traffic
accidents among the working population is consistent with
the morbidity surveillance data,
34
requiring interventions
to prevent premature deaths from such accidents.
Although the cost for primary health care is relatively
low or free because of government subsidies, the cost for
emergency and intensive care could be high, particularly at
the tertiary health level.
30
This suggests that more wealthy
people are likely to use tertiary health services and possibly
die in a tertiary hospital than the poor. Serious and acute
health conditions were also more likely to die in hospitals
than mild and chronic diseases. The differences in mortality
patterns between the recent studies using three data sour-
ces—PHMRC, MCCD and CHESS—could reflect a chang-
ing composition of the mortalities, from infectious diseases
to NCDs in the PNG population in the 2010s. These data
sources complement each other to provide a more compre-
hensive picture of mortality in PNG. Comparing the results
from these studies for a similar period of time could pro-
vide more insights into the mortality situation in PNG, de-
spite the fact that none of these data sources is
representative of mortality in very remote areas of PNG.
CHESS data used in this study focused on deaths from
the communities that could better reflect the mortality of
the population than data from hospitals or mixed data
from hospitals and the communities. Given the difference
in access to health facilities in PNG, the proportions of
deaths in hospitals were lower than those in the communi-
ties, about 20% to 80%, respectively. Hospital data cap-
tured an even smaller proportion of deaths among the
rural population (about 15%).
14
With about 20% of
deaths occurring in health facilities, hospital deaths could
have different COD patterns compared with deaths occur-
ring in the community.
Mortality data from the community collected using the
WHO VA tools are a valid and important data source for
COD analysis, particularly in LMICs. On the other hand,
hospital death data are a separate data source and also
contribute to providing a complete picture of mortality.
Hypothetically, the combination of mortality data from
health facilities and from communities should be complete
as it would capture all deaths in the population. However,
previous studies in other LMICs found that a considerable
proportion of deaths were duplicated between the two
data sources.
20,50
This could also be the case in PNG, but
it is unlikely in this study as death certificates were avail-
able for only 19 deaths.
Understanding the trend and differential in mortality
could provide insights into the mortality transition in PNG
and its projection towards future trends. The current study
supports the previous observation that provinces that were
readily accessible by road had the highest estimated pro-
portion of deaths from emerging infectious diseases,
8
whereas provinces that were geographically isolated had
the highest proportion of deaths from endemic infectious,
maternal, neonatal and nutritional causes. By contrast, the
emergence of NCDs is evident in provinces of higher socio-
economic status, such as POM and Central Province (see
Tables 5 and 6).
The increase in mortality from the NCDs identified in
this study is important for the PNG Government and the
health sector, to support specific policies and interventions
targeting NCDs to reduce morbidity among the working
population and protect them from premature mortality. It
is noteworthy that the emergence of NCDs varied across
urban-rural sectors and provinces, suggesting the uneven
distribution in these diseases and that the mortality transi-
tion progresses at different rates across geographical
regions. This requires different approaches and strategies
to address the high mortality from these diseases. A study
in Latin America, India and China suggested that an early
health education intervention significantly reduced the
prevalence of NCD mortality among the elderly popula-
tion.
51
Noticeably, PNG does not have a health plan for
the elderly. The PNG Government consider developing
new policies with prevention interventions targeting hyper-
tension, diabetes mellitus and ischaemic heart disease
among the elderly, as the population of PNG ages over the
coming decades.
17
Limitations
It is difficult to comment on the mortality rate in PNG
over time because of the lack of reliable measures to
16 International Journal of Epidemiology, 2022, Vol. 00, No. 00
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
estimate the population size of the surveillance sites due to
the large volume of migration flows in and out. The esti-
mated population coverage in CHESS surveillance sites
cannot be used as denominators for calculating mortality
rate for particular diseases in each site. As such, calculation
of mortality rates for specific causes of death in the surveil-
lance population over the data collection period was
precluded.
Because each surveillance site has different developmen-
tal history in terms of timeline for establishment, scope of
activities and number of staff, the proportions of deaths
captured among the surveillance populations were not con-
sistent across the sites. Seasonal migration in and out of
the sites was high, up to 20%.
17
Hence, the estimation of
crude death rates would not be reliable and is beyond the
scope of this study. The estimation of birth rates and child
mortalities was covered in other papers.
52,5353
A comprehensive assessment of the completeness of
mortality data was not conducted in this study. With the
total population of approximately 8.8 million people in
PNG for the period 2019–20, there would be expected to
be 57 200 deaths per year (8.8 million*0.065). The total
death count identified by the data reporters in the study
was 1021 over a period of 2.5 years, meaning 408 deaths
were identified per year on average, which is less than 1%
of the total deaths expected in PNG. CHESS does not aim
provide representative samples for a population either at
the national or the provincial level. CHESS is designed to
provide mortality data from the populations living within
the specific catchment areas. The surveillance sites were se-
lected based on the PNG Government agencies’ interest in
monitoring and reporting socioeconomic indicators and
the impact of socioeconomic development programmes on
the health of the population, and after consultation with
stakeholders and partners at national and local levels.
Hence, the data used in this study are not country-
representative and maybe do not represent the COD pro-
file of each province and mortality pattern of the entire
surveillance population. Although deaths in the surveil-
lance sites were recommended by the village-based data
reporters for VA interviews, there was no way of ensuring
that all the deaths had been identified and reported within
the time frame.
Accurate diagnosis of causes of death largely relied on
the quality of mortality data. Recall biases are considerably
high in VA interviews. As such, data from deaths that oc-
curred more than 2 years prior to the VA interviews are
not recommended in COD analyses.
25
In this study, about
15% of VA interviews were not completed in the due
course of 2 years (or 104 weeks) after the date of death,
with the highest proportions of delayed completions
reported in Central Province and POM, 26.8% and
18.5%, respectively. The VA interview process was pro-
longed in these sites because these provinces were in lock-
down for several periods and CHESS staff members were
infected with COVID-19 during the COVID-19 outbreaks
in 2020.
30
Previous study that analysed the optimal recall period
for VA interviews recommended that data collection
should be conducted within 3 months of the death, to im-
prove the quality of the data: WHO recommended that VA
data collection be conducted within 12 months of the
death.
23
In this study, the duration from the date of death
to the date of completion of VA interviews was calculated
(see Table 2). Many VA interviews required more than one
visit to the household for clarification of the information
provided and correction of the information recorded. The
duration for completing VA interview and data collection
was lengthened for those interviews conducted in 2020
when the COVID19 pandemic broke out in PNG, resulting
in cancellation and postponement in the fieldwork.
The numbers of child deaths reported in POM and
Madang were particularly low and not proportionate to the
scope of the surveillance population or the size of the entire
population in these provinces, because the two sites were set
up more recently in 2017–18. The current surveillance sites
were established in urban areas where primary health care
services were available and the majority of the surveillance
population in these provinces reported having access to ter-
tiary children’s hospitals. In this context, most child deaths
would have occurred in health facilities whereas the current
study focused on deaths that occurred in communities. This
could have introduced bias in the reported child deaths in
POM and Madang. The incompleteness of reported mortal-
ity data among children in these provinces could most sig-
nificantly impact on the interpretation of the study results
since a higher proportion of child deaths could have oc-
curred in tertiary health facilities. Hence, interpretation of
the results should be cautious.
The incompleteness of reported mortality data among
children in these provinces could most significantly impact
on the interpretation of the study results, since a much
higher proportion of child deaths could have occurred in
tertiary health facilities. Previous COD studies in PNG fo-
cused on adolescent and adult deaths, with few data on
child deaths available to show changes in causes of death
among children. There is still a large gap in valid data on
mortality among children representative of the PNG popu-
lation. The continuity of the mortality surveillance compo-
nent of CHESS is crucial to providing more complete data
on COD in children.
The elderly deaths recorded in POM were relatively low
because the site was established in 2017, later than other
sites. Three in five elderly deaths reported in POM were
International Journal of Epidemiology, 2022, Vol. 00, No. 00 17
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
attributed to respiratory neoplasms, accounting for 60%
compared with only 7% of deaths in this population across
the sites. This suggests that the respiratory neoplasms data
did not well represent the elderly living in POM, and the
data should be interpreted with caution. The POM surveil-
lance site has been recently scaled up to provide more rep-
resentative data for the population in the future.
The scaling up of mortality surveillance activities using
CHESS research infrastructure could be possible, but the
application of the WHO VA instrument for collecting bet-
ter mortality data from the communities in the very remote
isolated areas of PNG is challenging due to logistics and
technical issues.
Conclusion
This study provided data on COD among sub-populations
in PNG. Evidence of an epidemiological shift in causes of
death in the population has confirmed the mortality transi-
tion currently occurring in PNG. The contemporary mor-
tality transition shows four epidemiological characteristics:
(i) the dominance of infectious diseases, particularly the
persistence of ARTIs among children; (ii) the increase in
TB and HIV/AIDS mortality among the working age popu-
lation, particularly young adults; (iii) the emergence of car-
diovascular diseases, particularly ACD and stroke, among
the elderly; and (iv) the increasing mortality from injuries
and external causes in the population. The mixed patterns
of mortality suggest considerable changes in lifestyle occur-
ring across the populations in PNG over the past 50 years.
CHESS is a reliable source of mortality data from the
communities, providing further insights into the mortality
modality in contemporary PNG. The introduction of rou-
tine surveillance of mortality in the communities as a result
of CHESS improves the availability and quality of mortal-
ity data to inform health policy and social planning, and to
better control infections and NCDs in PNG. New VA data
collection instruments such as WHO 2016, powered by
cause of death analytical software such as InterVA-5, have
generated timely and more reliable data for monitoring
and reporting leading causes of death in the population.
These data provide important evidence for the health sec-
tor to guide further reform and develop interventions to ef-
fectively respond to the increasing demand for new health
care services in the population. This calls for more studies
on mortality to assist the PNG Government in developing
strategies to effectively address multiple complex public
health issues in such a critical transitional period in the
country. Given the variation in leading causes of death
among children, the working population and the elderly,
the health system should cater to all conditions. Universal
access to comprehensive primary health care services is
essential. The trends in mortalities across populations need
to be closely monitored in the next decades.
Ethics approval
CHESS has ethics approvals from Internal Review Board of PNG
Institute of Medical Research (IRB’s Approval no. 18.05) and the
Medical Research Advisory Committee of Papua New Guinea
(MRAC’s Approval no. 18.06). These approvals covered all the data
components under CHESS, including mortality data, which were
used in this manuscript. Informed consent was sought from a rela-
tive of the deceased. They were informed about their right to with-
draw from the study at any stage.
Data availability
The datasets used in this study are available from the corresponding
author on reasonable request. The corresponding author has full ac-
cess to all the data used in this study and had final responsibility for
the decision to submit the study for publication.
Supplementary data
Supplementary data are available at IJE online.
Author contributions
B.N.P. .designed the CHESS, conceptualized the manuscript, ana-
lysed and interpreted the data and drafted, revised, finalized and
submitted the manuscript. R.J., V.D.S., S.M. supervised the field-
work, collected and analysed the data and provided inputs for the
manuscript. A.D.O. commented on and edited the manuscript. W.P.
provided oversight from PNGIMR and approved submission of the
manuscript. All authors contributed to the article and approved the
submitted version for publication.
Funding
CHESS is financially supported by the PNG Government through
the Department of National Planning and Monitoring (PIP No.
02704). A.D.O. is supported by an NHMRC Investigator Grant
(APP1176858). The funders had no role in study design, data collec-
tion and analysis or writing of the manuscript.
Acknowledgements
We acknowledge the following individuals and organizations: com-
munity leaders, councillors and religious leaders, community mem-
bers in the surveillance sites, the Provincial Health Authorities in
Central, Eastern Highlands, East New Britain, East Sepik, Madang
provinces and Port Moresby and collaborators in the CHESS pro-
gramme: Salvation Army, Evangelical Church PNG, Lutheran
Health Services, Evangelical Brotherhood Church Health Services
and Catholic Health Services.
Conflict of interest
None declared.
18 International Journal of Epidemiology, 2022, Vol. 00, No. 00
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
References
1. United Nations. Sustainable Development Goals: SDGs
Indicators. 2015. https://unstats.un.org/sdgs/metadata?Text=&
Goal=3&Target=3.2 (26 November 2022, date last accessed).
2. Naghavi M, Abajobir AA, Abbafati C. Global, regional, and na-
tional age-sex specific mortality for 264 causes of death, 1980-
2016: a systematic analysis for the Global Burden of Disease
Study 2016. Lancet 2017;390:1151–210.
3. Wang H, Naghavi M, Allen C et al. Global, regional, and na-
tional life expectancy, all-cause mortality, and cause-specific
mortality for 249 causes of death, 1980-2015: a systematic
analysis for the Global Burden of Disease Study 2015. Lancet
2016;388:1459–544.
4. Gouda HN, Hazard RH, Maraga S et al. The epidemiological
transition in Papua New Guinea: new evidence from verbal au-
topsy studies. Int J Epidemiol 2019;48:966–77.
5. Macrotrends. Papua New Guinea Life Expectancy 1950-2021.
2021. https://www.macrotrends.net/countries/PNG/papua-new-
guinea/life-expectancy (26 November 2022, date last accessed).
6. National Statistics Office. Population 2011 Census Summary
Figures National Statistics Office. 2011.https://www.nso.gov.
pg/statistics/population/ (26 November 2022, date last
accessed).
7. National Statistics Office. Papua New Guinea Demographic and
Health Survey 2016-18: Key Indicators Report. Port Moresby:
National Statistics Office, 2019.
8. Kitur U, Adair T, Riley I, Lopez AD. Estimating the pattern of
causes of death in Papua New Guinea. BMC Public Health
2019;19:1322.
9. Rarau P, Vengiau G, Gouda HN et al. Prevalence of non-
communicable disease risk factors in three sites across Papua
New Guinea: a cross-sectional study. BMJ Glob Health 2017;2:
e000221.
10. Pham NB, Jorry R, Abori N, Silas VD, Okely AD, Pomat W.
Non-communicable diseases attributed mortality and associated
sociodemographic factors in Papua New Guinea: Evidence from
the Comprehensive Health and Epidemiological Surveillance
System. PLoS Glob Public Health 2022;2:e0000118.
11. Riley I. Demography and the epidemiology of disease in Papua
New Guinea. PNG Med J 2009;52:83–95.
12. National Department of Health, World Health Organization.
Papua New Guinea - WHO Country Cooperation Strategy
2016–2020. Geneva: WHO, 2016.
13. Department of National Planning and Monitoring. Medium
Term Development Plan III 2018–2022. Implementation
Framework and Investment Plan. Port Moresby: Department of
National Planning and Monitoring, 2018.
14. Pham NB, Jorry R, Abori N et al. Comprehensive Health and
Epidemiological Surveillance System Technical Report:
Mortality Surveilance in Communities. Goroka: PNG Institute
of Medical Research, 2020.
15. Pham NB, Whittaker M, Pomat W, Siba P. CHESS: a new gener-
ation of population health surveillance for sustainable develop-
ment of Papua New Guinea. PNG Med J 2017;60:154–72.
16. Sankoh O; INDEPTH Network. CHESS: an innovative concept
for a new generation of population surveillance. Lancet Glob
Health 2015;3:e742. Vol
17. Pham NB, Maraga S, Boli R et al. Partnership in Health
Programme: March 2018 Edition on Population Census and
Demographic Changes in the PNG IMR’s Health and
Epidemiological Surveillance Sites in the Period 2015–2017.
Goroka: PNG Institute of Medical Research, 2018.
18. Pham NB, Maraga S, Boli R et al. Comprehensive Health and
Epidemiological Surveillance System: September 2018 Edition
on Household Socioeconomic and Demographic Characteristics.
Goroka: PNG Institute of Medical Research, 2018.
19. Murray CJ, Lopez AD, Black R et al. Population Health Metrics
Research Consortium gold standard verbal autopsy validation
study: design, implementation, and development of analysis
datasets. Popul Health Metr 2011;9:27.
20. Byass P, Kahn K, Fottrell E, Collinson MA, Tollman SM.
Moving from data on deaths to public health policy in
Agincourt, South Africa: approaches to analysing and under-
standing verbal autopsy findings. PLoS Med 2010;7:e1000325.
21. WHO. Verbal Autopsy Standards: Ascertaining and Attributing
Causes of Death. 2021. https://apps.who.int/iris/handle/10665/
43764 (26 November 2022, date last accessed).
22. Nichols E, Byass P, Chandramohan D et al.; WHO Verbal
Autopsy Working Group. The WHO 2016 verbal autopsy in-
strument: an international standard suitable for automated
analysis by InterVA, InSilicoVA, and Tariff 2.0. PLoS Med
2018;15:e1002486.
23. Byass P, Hussain-Alkhateeb L, D’Ambruoso L et al. An inte-
grated approach to processing WHO-2016 verbal autopsy data:
the InterVA-5 model. BMC Med 2019;17:102.
24. Jorry R, Pham NB, Pomat W. Piloting the New WHO 2016
Verbal Autopsy Tablet Based Data Collection Method in Asaro/
Goroka PNG Medical Symposium. Port Moresby; 2018.
25. Serina P, Riley I, Hernandez B et al. What is the optimal recall
period for verbal autopsies? Validation study based on repeat
interviews in three populations. Popul Health Metr 2016;14:
40–48.
26. National Center for Health Statistics, Centers for Diseases
Control and Prevention. International Classification of Diseases,
Tenth Revision (ICD-10). 2021. https://www.cdc.gov/nchs/icd/
icd10.htm(26 November 2022, date last accessed).
27. Hart JD, Mahesh P, Kwa V et al. Diversity of epidemiological
transition in the Pacific: Findings from the application of verbal
autopsy in Papua New Guinea and the Solomon Islands. Lancet
Reg Health West Pac 2021;11:100150.
28. Hart JD, Kwa V, Dakulala P et al. How advanced is the epidemi-
ological transition in Papua New Guinea? New evidence from
verbal autopsy. Int J Epidemiol 2022;50:2058–69.
29. United Nations. Data Portal Population Division: Population
Division. 2022. https://population.un.org/dataportal/data/indica
tors/59,61/locations/598/start/1990/end/2022/table/pivotbyloca
tion (26 November 2022, date last accessed).
30. Pham BN, Jorry R, Maraga S et al. CHESS September 2021
Technical Report: Household Socioeconomic Status and
Implication for National Response to COVID-19 2021.
Goroka: Papua New Guinea Institute of Medical Research.
31. Joint Agency Task Force. Media Statement: PNG COVID-19
National Pandemic Response. Port Moresby: Office of the
Controller, 2021.
International Journal of Epidemiology, 2022, Vol. 00, No. 00 19
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
32. Riley I, Lehmann D. The demography of Papua New Guinea: mi-
gration, fertility, and mortality patterns. In: Attenborough R,
Alpers M (eds). Human Biology in Papua New Guinea: The
Small Cosmos. Oxford, UK: Oxford University Press, 1992,
pp. 67–92.
33. Pham NB, Maraga S, Aga T et al. Comprehensive Health and
Epidemiological Surveillance System: March 2019 Edition
on Child Health Goroka2019.https://www.researchgate.net/
publication/333678257_PNG_IMR’s_CHESS_Technical_Report_
March_2019_Edition_Child_Health (26 November 2022, date
last accessed).
34. PNG Institute of Medical Research. Partnership in Health
Project Report: Morbidity Surveillance at Primary Health
Facilities (Reporting Period: January-June 2017). Goroka: PNG
Institute of Medical Research, 2017
35. Kakazo M, Lehmann D, Coakley K et al. Mortality rates and the
utilization of health services during terminal illness in the Asaro
Valley, Eastern Highlands Province, Papua New Guinea. PNG
Med J 1999;42:12–26.
36. Lehmann D. Demography and causes of death among the Huli in
the Tari Basin. PNG Med J 2002;45:51–62.
37. Riley I. Pneumonia in Papua New Guinea: Account of the
Effects of Western Medicine upon Disease in a Developing
Country. Sydney, NSW: University of Sydney, 1979.
38. Boli R, Pham NB, Siba P. Assessing the changing burden of dis-
eases at the primary health care level in rural Papua New
Guinea. PNG Med J 2017;60: 122–38.
39. Fenner F. Malaria control in Papua New Guinea in the Second
World War: from disaster to successful prophylaxis and the
dawn of DDT. Parasitologia 1998;40:55–63.
40. Ley SD, Riley I, Beck HP. Tuberculosis in Papua New Guinea:
from yesterday until today. Microbes Infect 2014;16:607–14.
41. Vallely A, Page A, Dias S et al. The prevalence of sexually trans-
mitted infections in Papua New Guinea: a systematic review and
meta-analysis. PLoS One 2010;5:e15586.
42. Aia P, Wangchuk L, Morishita F et al. Epidemiology of tubercu-
losis in Papua New Guinea: analysis of case notification and
treatment-outcome data, 2008-2016. Western Pac Surveill
Response J 2018;9:9–19.
43. Vakadem K, Anota A, Sa’avu M et al. A mortality review of
adult inpatients with tuberculosis in Mendi, Papua New Guinea.
Public Health Action 2019;9: S62–67.
44. Pham NB, Abori N, Silas VD et al. Tuberculosis and HIV/AIDS-
attributed mortalities and associated sociodemographic factors
in Papua New Guinea: evidence from the comprehensive health
and epidemiological surveillance system. BMJ Open 2022;12:
e058962.
45. Pham NB, Appo J, Gende G, Rao C, Okely AD, Pomat W.
Leprosy-tuberculosis co-infection: a case report in Papua New
Guinea. J Med Clin Res Rev 2020;4:1–3.
46. Kelly-Hanku A, Nightingale CE, Pham MD et al. Loss to follow
up of pregnant women with HIV and infant HIV outcomes in the
prevention of maternal to child transmission of HIV programme
in two high-burden provinces in Papua New Guinea: a retrospec-
tive clinical audit. BMJ Open 2020;10:e038311.
47. WHO. HIV/AIDS Fact Sheet. Geneva: World Health
Organization, 2020. https://www.who.int/news-room/fact-
sheets/detail/hiv-aids (26 November 2022, date last accessed).
48. Rarau P, Pulford J, Gouda H et al. Socio-economic status and
behavioural and cardiovascular risk factors in Papua New
Guinea: A cross-sectional survey. PLoS One 2019;14:e0211068.
49. Martin W, Sengupta S, Murthy D, Barua D. The spectrum of
cancer in Papua New Guinea: an analysis based on the cancer
registry 1979-1988. Cancer 1992;70:2942–50.
50. Adair T, Lopez AD. Estimating the completeness of death regis-
tration: An empirical method. PLoS One 2018;13:e0197047.
51. Ferri CP, Acosta D, Guerra M et al. socioeconomic factors and
all cause and cause-specific mortality among older people in
Latin America, India, and China: a population-based cohort
study. PLoS Med 2012;9:e1001179.
52. Pham NB, Whittaker M, Okely AD, Pomat W. Measuring unmet
need of contraception among women in rural areas of Papua
New Guinea. Sex Reprod Health Matters 2020;28:1–17.
53. Pham NB, Emori RB, Ha T, Parrish A-M, Okely AD. Estimating
child mortality at the sub-national level in Papua New Guinea:
evidence from the integrated health and demographic surveil-
lance system. Front Public Health 2021;9:723252–14.
54. National Statistics Office. National Population and Housing
Census 2011. Port Moresby: NSO, 2013.
20 International Journal of Epidemiology, 2022, Vol. 00, No. 00
Downloaded from https://academic.oup.com/ije/advance-article/doi/10.1093/ije/dyac232/6955640 by guest on 21 December 2022
