Performance of the Tariff Method: validation of a simple additive algorithm for analysis of verbal autopsies.

Published August 31, 2011, in Population Health Metrics (opens in a new window)

The Tariff method, an easy-to-use tool developed by researchers at IHME for turning verbal autopsy (VA) results into meaningful cause of death data for health workers and policymakers, is capable of outperforming the more costly physician-certified verbal autopsy (PCVA) approach in most cases, according to new research by IHME as part of the Population Health Metrics Research Consortium (PHMRC).

The study, Performance of the Tariff Method: validation of a simple additive algorithm for analysis of verbal autopsies, shows that Tariff provides a reliable method for assigning cause of death data that is less expensive and more transparent than other VA tools and can be readily implemented in a wide variety of settings.

Research objective

VAs provide valuable information to determine what causes people to die in areas without reliable vital registration systems. But PCVA, the primary method currently used to assign cause of death, which relies on physicians to interpret VA results, is costly, sometimes inconsistent, and a questionable use of resources in rural areas with a shortage of physicians. Other tools that use statistics or computations to assign cause of death lack transparency and may be difficult for users to understand.
 
This study is part of ongoing work by IHME to develop the most accurate and efficient methods of predicting causes of death using VA. It compares the performance of the Tariff method, a computational method based on the premise that there are highly informative signs and symptoms for different causes of death, with the more complicated and labor-intensive PCVA approach using validated, gold standard deaths collected as part of the PHMRC gold standard VA validation study.

Research findings

At the individual level for chance-corrected concordance, or the agreement between predicted and observed cause of death, the Tariff method predicted the known cause of death 44.5% of the time in adults, 39% in children, and 23.9% in neonates. For adults, the Tariff method outperforms PCVA when the household’s recall of health care experience is excluded and is not significantly different than PCVA when health care experience is included. Health care experience includes any information the caretaker has about the patient's medical treatment, including medical records or documentation regarding cause of death. For children, PCVA outperforms the Tariff method both with and without health care experience. For neonates, the two methods are not directly comparable because PCVA cannot make assignments on the full list of neonate causes.
 
At the population level, the cause-specific mortality fraction  accuracy for the Tariff method was 0.745 for adults, 0.709 for children, and 0.679 for neonates. The Tariff method yields more accurate estimates of cause-specific mortality fraction than PCVA for adults and children, both with and without recall of health care experience. Again, it was not possible to directly compare the accuracy of PCVA and the Tariff method for neonates.

Analytical approach

The premise behind the Tariff method is to identify signs and symptoms identified during a VA that are highly indicative of a particular cause of death. A tariff is developed for each sign and symptom to reflect how predictive it is for each cause of death. A tariff score is calculated for each possible cause, which is used to assign the cause of death for a particular person. The researchers developed tariffs and tariff scores using household recall of health care experience and also excluding those variables, to estimate the method’s performance in places where access to health care is uncommon.
 
The study uses the PHMRC gold standard VA training dataset – a collection of 12,542 cases for which full VAs were collected and for which the true cause of death is known from diagnostic or pathological evidence – to develop tariffs and to assess the performance of the Tariff method compared to PCVA.

Policy implications

The Tariff method – a simple approach based on identifying items in a VA interview that are indicative of particular diseases – performs as well or better than PCVA in many instances. Using a simple matrix that will be available online, the Tariff method requires only multiplication and addition to make cause of death assignments using VA information. The steps are transparent for users to understand, and the matrix will be available on several different platforms, including cell phone operating systems.
 
Given that PCVA can be costly and time-consuming, the Tariff method provides an attractive way to collect cause of death data in areas of the world with minimal health information infrastructure. Since the Tariff method performs markedly better than InterVA, another automated VA tool that is widely available, users interested in rapid, low-cost, and easy-to-understand verbal autopsy methods should consider the Tariff method.
Read full article (opens in a new window)

Citation

James SL, Flaxman AD, Murray CJL, the Population Health Metrics Research Consortium (PHMRC). Performance of the Tariff Method: validation of a simple additive algorithm for analysis of verbal autopsies. Population Health Metrics. 2011; 9:31.

Related