InterVA versus Spectrum: how comparable are they in estimating AIDS mortality patterns in Nairobi's informal...

Reliable mortality data remain scarce in many regions, hindering effective public health planning. National estimates often rely on mathematical models when vital registration systems are absent. Spectrum serves as a standard tool for generating these national AIDS mortality figures. Meanwhile, verbal autopsy programs like InterVA-4 provide cause-specific data where medical certification is unavailable. No prior work had resolved how these two distinct approaches compare when applied to the same urban population. That uncertainty drove this investigation into mortality patterns within Nairobi's informal settlements. Researchers needed to determine if these tools yield consistent results for local health policy. This gap motivated a direct comparison of trends derived from both modeling systems.

This study aims to compare trends in adult AIDS-related mortality estimated by Spectrum with trends from the InterVA-4 programme. The researchers sought to evaluate the comparability of these two systems in settings where vital registration is lacking. They focused on data from a Health and Demographic Surveillance System in Nairobi, Kenya. The investigation addresses the challenge of estimating cause-specific mortality in informal urban settlements. By applying both methods to the same population, the team examined the consistency of their respective outputs. The authors intended to determine if these tools provide reliable mortality patterns for local health planning. This comparison highlights the potential biases inherent in different modeling approaches for infectious disease deaths. The work provides insight into the limitations of using probabilistic verbal autopsy tools for complex health outcomes.

The review approach involved a comparative analysis of mortality trends using two distinct computational frameworks. Investigators constructed a city-specific model for Nairobi by integrating HIV prevalence statistics and national antiretroviral therapy coverage. They also incorporated underlying mortality rates and migration assumptions into this primary simulation. The team analyzed verbal autopsy records from 1,799 adult deaths occurring between 2003 and 2010. These records originated from a specific Health and Demographic Surveillance System site. Researchers entered the autopsy information into the probabilistic program to assign causes of death. They calculated annualised trends for both AIDS-related and all-cause mortality across the observation window. Finally, the authors evaluated the consistency of the outputs by comparing the resulting mortality proportions.

Key findings from the literature indicate that Spectrum estimated a 7% HIV prevalence in Nairobi, whereas the surveillance site measured 12% in 2010. Spectrum projected higher levels of AIDS-related mortality despite the observed differences in prevalence data. The proportion of AIDS-related deaths decreased from 63% to 40% according to the Spectrum model. In contrast, the verbal autopsy tool showed a decline from 25% to 16% over the same timeframe. Net AIDS-related mortality in Spectrum aligned more closely with combined AIDS and tuberculosis rates from the verbal autopsy method. Overall trends remained similar across both systems throughout the eight-year study period. The values converged significantly when researchers included tuberculosis fatalities in the verbal autopsy calculations. These results demonstrate that the two methods produce divergent absolute values for AIDS mortality.

The authors report that both systems show a downward trend in AIDS-related mortality over the study period. Spectrum consistently produced higher estimates of AIDS deaths than the verbal autopsy approach. Discrepancies between the two tools appear linked to how each handles tuberculosis-related fatalities. The researchers propose that InterVA-4 might struggle to accurately differentiate between AIDS and tuberculosis as causes of death. Including tuberculosis deaths in the verbal autopsy analysis brought the final mortality rates closer to Spectrum figures. These findings suggest that the choice of estimation method significantly influences reported mortality proportions. The team highlights the necessity of interpreting these model outputs with caution in urban settings. Future efforts should focus on refining the diagnostic accuracy of verbal autopsy algorithms for complex infectious diseases.