Related Experiment Video
Updated: May 24, 2026

Inverse Probability of Treatment Weighting (Propensity Score) using the Military Health System Data Repository and National Death Index
Published on: January 8, 2020
Differences in the causes of death of HIV-positive patients in a cohort study by data sources and coding algorithms
Victoria Hernando1, Paz Sobrino-Vegas, M Carmen Burriel
1Red de Investigación en Sida, Centro Nacional de Epidemiología, Instituto de Salud Carlos III, Madrid, Spain. vhernando@isciii.es
Objectives:
To compare causes of death (CoDs) from two independent sources: National Basic Death File (NBDF) and deaths reported to the Spanish HIV Research cohort [Cohort de adultos con infección por VIH de la Red de Investigación en SIDA CoRIS)] and compare the two coding algorithms: International Classification of Diseases, 10th revision (ICD-10) and revised version of Coding Causes of Death in HIV (revised CoDe).
Methods:
Between 2004 and 2008, CoDs were obtained from the cohort records (free text, multiple causes) and also from NBDF (ICD-10). CoDs from CoRIS were coded according to ICD-10 and revised CoDe by a panel. Deaths were compared by 13 disease groups: HIV/AIDS, liver diseases, malignancies, infections, cardiovascular, blood disorders, pulmonary, central nervous system, drug use, external, suicide, other causes and ill defined.
Results:
There were 160 deaths. Concordance for the 13 groups was observed in 111 (69%) cases for the two sources and in 115 (72%) cases for the two coding algorithms. According to revised CoDe, the commonest CoDs were HIV/AIDS (53%), non-AIDS malignancies (11%) and liver related (9%), these percentages were similar, 57, 10 and 8%, respectively, for NBDF (coded as ICD-10). When using ICD-10 to code deaths in CoRIS, wherein HIV infection was known in everyone, the proportion of non-AIDS malignancies was 13%, liver-related accounted for 3%, while HIV/AIDS reached 70% due to liver-related, infections and ill-defined causes being coded as HIV/AIDS.
Conclusion:
There is substantial variation in CoDs in HIV-infected persons according to sources and algorithms. ICD-10 in patients known to be HIV-positive overestimates HIV/AIDS-related deaths at the expense of underestimating liver-related diseases, infections and ill defined causes. CoDe seems as the best option for cohort studies.
Related Concept Videos
Bias in Epidemiological Studies
Comparing the Survival Analysis of Two or More Groups
Confounding in Epidemiological Studies
Causality in Epidemiology
Censoring Survival Data
Cancer Survival Analysis