Performance of four computer-coded verbal autopsy methods for cause of death assignment compared with physician coding on 24,000 deaths in low- and middle-income countries

Nikita Desai, Lukasz Aleksandrowicz, Pierre Miasnikof, Ying Lu, Jordana Leitao, Peter Byass, Stephen Tollman, Paul Mee, Dewan Alam, Suresh K. Rathi, Abhishek Singh, Rajesh Kumar, Faujdar Ram, Prabhat Jha

Research output: Contribution to journalArticle

Abstract

Background: Physician-coded verbal autopsy (PCVA) is the most widely used method to determine causes of death (CODs) in countries where medical certification of death is uncommon. Computer-coded verbal autopsy (CCVA) methods have been proposed as a faster and cheaper alternative to PCVA, though they have not been widely compared to PCVA or to each other.Methods: We compared the performance of open-source random forest, open-source tariff method, InterVA-4, and the King-Lu method to PCVA on five datasets comprising over 24,000 verbal autopsies from low- and middle-income countries. Metrics to assess performance were positive predictive value and partial chance-corrected concordance at the individual level, and cause-specific mortality fraction accuracy and cause-specific mortality fraction error at the population level.Results: The positive predictive value for the most probable COD predicted by the four CCVA methods averaged about 43% to 44% across the datasets. The average positive predictive value improved for the top three most probable CODs, with greater improvements for open-source random forest (69%) and open-source tariff method (68%) than for InterVA-4 (62%). The average partial chance-corrected concordance for the most probable COD predicted by the open-source random forest, open-source tariff method and InterVA-4 were 41%, 40% and 41%, respectively, with better results for the top three most probable CODs. Performance generally improved with larger datasets. At the population level, the King-Lu method had the highest average cause-specific mortality fraction accuracy across all five datasets (91%), followed by InterVA-4 (72% across three datasets), open-source random forest (71%) and open-source tariff method (54%).Conclusions: On an individual level, no single method was able to replicate the physician assignment of COD more than about half the time. At the population level, the King-Lu method was the best method to estimate cause-specific mortality fractions, though it does not assign individual CODs. Future testing should focus on combining different computer-coded verbal autopsy tools, paired with PCVA strengths. This includes using open-source tools applied to larger and varied datasets (especially those including a random sample of deaths drawn from the population), so as to establish the performance for age- and sex-specific CODs.

Original languageEnglish (US)
Article number20
JournalBMC Medicine
Volume12
Issue number1
DOIs
StatePublished - Feb 4 2014

Fingerprint

Cause of Death
Autopsy
Physicians
Mortality
Population
Certification
Datasets
Forests

Keywords

  • Causes of death
  • Computer-coded verbal autopsy (CCVA)
  • InterVA-4
  • King-Lu
  • Physician-certified verbal autopsy (PCVA)
  • Random forest
  • Tariff method
  • Validation
  • Verbal autopsy

ASJC Scopus subject areas

  • Medicine(all)

Cite this

Performance of four computer-coded verbal autopsy methods for cause of death assignment compared with physician coding on 24,000 deaths in low- and middle-income countries. / Desai, Nikita; Aleksandrowicz, Lukasz; Miasnikof, Pierre; Lu, Ying; Leitao, Jordana; Byass, Peter; Tollman, Stephen; Mee, Paul; Alam, Dewan; Rathi, Suresh K.; Singh, Abhishek; Kumar, Rajesh; Ram, Faujdar; Jha, Prabhat.

In: BMC Medicine, Vol. 12, No. 1, 20, 04.02.2014.

Research output: Contribution to journalArticle

Desai, N, Aleksandrowicz, L, Miasnikof, P, Lu, Y, Leitao, J, Byass, P, Tollman, S, Mee, P, Alam, D, Rathi, SK, Singh, A, Kumar, R, Ram, F & Jha, P 2014, 'Performance of four computer-coded verbal autopsy methods for cause of death assignment compared with physician coding on 24,000 deaths in low- and middle-income countries', BMC Medicine, vol. 12, no. 1, 20. https://doi.org/10.1186/1741-7015-12-20
Desai, Nikita ; Aleksandrowicz, Lukasz ; Miasnikof, Pierre ; Lu, Ying ; Leitao, Jordana ; Byass, Peter ; Tollman, Stephen ; Mee, Paul ; Alam, Dewan ; Rathi, Suresh K. ; Singh, Abhishek ; Kumar, Rajesh ; Ram, Faujdar ; Jha, Prabhat. / Performance of four computer-coded verbal autopsy methods for cause of death assignment compared with physician coding on 24,000 deaths in low- and middle-income countries. In: BMC Medicine. 2014 ; Vol. 12, No. 1.
@article{4fd6b4c4c1ac43fb8db52f3ea25b21cb,
title = "Performance of four computer-coded verbal autopsy methods for cause of death assignment compared with physician coding on 24,000 deaths in low- and middle-income countries",
abstract = "Background: Physician-coded verbal autopsy (PCVA) is the most widely used method to determine causes of death (CODs) in countries where medical certification of death is uncommon. Computer-coded verbal autopsy (CCVA) methods have been proposed as a faster and cheaper alternative to PCVA, though they have not been widely compared to PCVA or to each other.Methods: We compared the performance of open-source random forest, open-source tariff method, InterVA-4, and the King-Lu method to PCVA on five datasets comprising over 24,000 verbal autopsies from low- and middle-income countries. Metrics to assess performance were positive predictive value and partial chance-corrected concordance at the individual level, and cause-specific mortality fraction accuracy and cause-specific mortality fraction error at the population level.Results: The positive predictive value for the most probable COD predicted by the four CCVA methods averaged about 43{\%} to 44{\%} across the datasets. The average positive predictive value improved for the top three most probable CODs, with greater improvements for open-source random forest (69{\%}) and open-source tariff method (68{\%}) than for InterVA-4 (62{\%}). The average partial chance-corrected concordance for the most probable COD predicted by the open-source random forest, open-source tariff method and InterVA-4 were 41{\%}, 40{\%} and 41{\%}, respectively, with better results for the top three most probable CODs. Performance generally improved with larger datasets. At the population level, the King-Lu method had the highest average cause-specific mortality fraction accuracy across all five datasets (91{\%}), followed by InterVA-4 (72{\%} across three datasets), open-source random forest (71{\%}) and open-source tariff method (54{\%}).Conclusions: On an individual level, no single method was able to replicate the physician assignment of COD more than about half the time. At the population level, the King-Lu method was the best method to estimate cause-specific mortality fractions, though it does not assign individual CODs. Future testing should focus on combining different computer-coded verbal autopsy tools, paired with PCVA strengths. This includes using open-source tools applied to larger and varied datasets (especially those including a random sample of deaths drawn from the population), so as to establish the performance for age- and sex-specific CODs.",
keywords = "Causes of death, Computer-coded verbal autopsy (CCVA), InterVA-4, King-Lu, Physician-certified verbal autopsy (PCVA), Random forest, Tariff method, Validation, Verbal autopsy",
author = "Nikita Desai and Lukasz Aleksandrowicz and Pierre Miasnikof and Ying Lu and Jordana Leitao and Peter Byass and Stephen Tollman and Paul Mee and Dewan Alam and Rathi, {Suresh K.} and Abhishek Singh and Rajesh Kumar and Faujdar Ram and Prabhat Jha",
year = "2014",
month = "2",
day = "4",
doi = "10.1186/1741-7015-12-20",
language = "English (US)",
volume = "12",
journal = "BMC Medicine",
issn = "1741-7015",
publisher = "BioMed Central",
number = "1",

}

TY - JOUR

T1 - Performance of four computer-coded verbal autopsy methods for cause of death assignment compared with physician coding on 24,000 deaths in low- and middle-income countries

AU - Desai, Nikita

AU - Aleksandrowicz, Lukasz

AU - Miasnikof, Pierre

AU - Lu, Ying

AU - Leitao, Jordana

AU - Byass, Peter

AU - Tollman, Stephen

AU - Mee, Paul

AU - Alam, Dewan

AU - Rathi, Suresh K.

AU - Singh, Abhishek

AU - Kumar, Rajesh

AU - Ram, Faujdar

AU - Jha, Prabhat

PY - 2014/2/4

Y1 - 2014/2/4

N2 - Background: Physician-coded verbal autopsy (PCVA) is the most widely used method to determine causes of death (CODs) in countries where medical certification of death is uncommon. Computer-coded verbal autopsy (CCVA) methods have been proposed as a faster and cheaper alternative to PCVA, though they have not been widely compared to PCVA or to each other.Methods: We compared the performance of open-source random forest, open-source tariff method, InterVA-4, and the King-Lu method to PCVA on five datasets comprising over 24,000 verbal autopsies from low- and middle-income countries. Metrics to assess performance were positive predictive value and partial chance-corrected concordance at the individual level, and cause-specific mortality fraction accuracy and cause-specific mortality fraction error at the population level.Results: The positive predictive value for the most probable COD predicted by the four CCVA methods averaged about 43% to 44% across the datasets. The average positive predictive value improved for the top three most probable CODs, with greater improvements for open-source random forest (69%) and open-source tariff method (68%) than for InterVA-4 (62%). The average partial chance-corrected concordance for the most probable COD predicted by the open-source random forest, open-source tariff method and InterVA-4 were 41%, 40% and 41%, respectively, with better results for the top three most probable CODs. Performance generally improved with larger datasets. At the population level, the King-Lu method had the highest average cause-specific mortality fraction accuracy across all five datasets (91%), followed by InterVA-4 (72% across three datasets), open-source random forest (71%) and open-source tariff method (54%).Conclusions: On an individual level, no single method was able to replicate the physician assignment of COD more than about half the time. At the population level, the King-Lu method was the best method to estimate cause-specific mortality fractions, though it does not assign individual CODs. Future testing should focus on combining different computer-coded verbal autopsy tools, paired with PCVA strengths. This includes using open-source tools applied to larger and varied datasets (especially those including a random sample of deaths drawn from the population), so as to establish the performance for age- and sex-specific CODs.

AB - Background: Physician-coded verbal autopsy (PCVA) is the most widely used method to determine causes of death (CODs) in countries where medical certification of death is uncommon. Computer-coded verbal autopsy (CCVA) methods have been proposed as a faster and cheaper alternative to PCVA, though they have not been widely compared to PCVA or to each other.Methods: We compared the performance of open-source random forest, open-source tariff method, InterVA-4, and the King-Lu method to PCVA on five datasets comprising over 24,000 verbal autopsies from low- and middle-income countries. Metrics to assess performance were positive predictive value and partial chance-corrected concordance at the individual level, and cause-specific mortality fraction accuracy and cause-specific mortality fraction error at the population level.Results: The positive predictive value for the most probable COD predicted by the four CCVA methods averaged about 43% to 44% across the datasets. The average positive predictive value improved for the top three most probable CODs, with greater improvements for open-source random forest (69%) and open-source tariff method (68%) than for InterVA-4 (62%). The average partial chance-corrected concordance for the most probable COD predicted by the open-source random forest, open-source tariff method and InterVA-4 were 41%, 40% and 41%, respectively, with better results for the top three most probable CODs. Performance generally improved with larger datasets. At the population level, the King-Lu method had the highest average cause-specific mortality fraction accuracy across all five datasets (91%), followed by InterVA-4 (72% across three datasets), open-source random forest (71%) and open-source tariff method (54%).Conclusions: On an individual level, no single method was able to replicate the physician assignment of COD more than about half the time. At the population level, the King-Lu method was the best method to estimate cause-specific mortality fractions, though it does not assign individual CODs. Future testing should focus on combining different computer-coded verbal autopsy tools, paired with PCVA strengths. This includes using open-source tools applied to larger and varied datasets (especially those including a random sample of deaths drawn from the population), so as to establish the performance for age- and sex-specific CODs.

KW - Causes of death

KW - Computer-coded verbal autopsy (CCVA)

KW - InterVA-4

KW - King-Lu

KW - Physician-certified verbal autopsy (PCVA)

KW - Random forest

KW - Tariff method

KW - Validation

KW - Verbal autopsy

UR - http://www.scopus.com/inward/record.url?scp=84893353935&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84893353935&partnerID=8YFLogxK

U2 - 10.1186/1741-7015-12-20

DO - 10.1186/1741-7015-12-20

M3 - Article

C2 - 24495855

AN - SCOPUS:84893353935

VL - 12

JO - BMC Medicine

JF - BMC Medicine

SN - 1741-7015

IS - 1

M1 - 20

ER -