Page 324 - Contributed Paper Session (CPS) - Volume 7

CPS2113 Lucia Pereira B. et al.
developed by the Population Health Metrics Research Consortium (PHMRC),
(see Murray et al., 2011). The SmartVA software was developed by the Institute
for Health Metrics and Evaluation (see Serina et al., 2015).
The objectives of this study are to present the automatic method of
assigning the underlying cause of death based on a verbal autopsy instrument
and to show preliminary results.

Tariff Score
The Tariff-score was calculated based on information collected from
12,501 individuals in the University of Washington's Population Health Metrics
Research Consortium (PHMRC) in 6 cities in 4 countries - India (Andhra
Pradesh and Uttar Pradesh), the Philippines (Bohol), Mexico (Mexico) and
Tanzania (Dar es Salaam and Pemba Island).
The questionnaire contains open and closed questions about Closed
questions:
 Symptoms of terminal illness;
 Diagnosis of chronic diseases - records of previous health
services;
 Risk behavior (alcohol, tobacco);
 Details of any interaction with health services.
Open questions: open narrative by the relatives through which one can
identify words or groups of words associated with the cause of death. The
underlying cause of death was known based on medical records and the
authopsy made in the SVOC (gold standard). The input worksheet contains
values 0 and 1, being 1 when the observed symptom occurred for a specific
individual. With this, it is possible to calculate the endorsement rate, given by:
xij = fraction of VAs for which there is a positive response to deaths from
cause i for item j
The Tariff is the endorsement rate standardized by the median and
interquartile range of all causes of death, fixed a symptom. For cause i,
symptom j is given by

where median(xij) is the median fraction with a positive response for item
j across all causes, and IIQ(xij) is the interquartile range of positive response
rates averaged across causes. Note that Tariff can assume positive or negative
values. As a final step, Tariffs are rounded to the nearest 0.5 to avoid overfitting
and to improve predictive validity.
For the calculation of the Tariff-Score of case k, cause i, the 40 symptoms
with the highest absolute value of the Tariff are considered

311 | I S I W S C 2 0 1 9