Studies validating verbal autopsy methods: a systematic review BMC Public Health

The characteristics of the selected studies are presented in Supplementary File 2.

target and settings

Of the 66 studies, 59 included validation of vets as a primary objective, while in 7 studies it was either a secondary objective, comparator, or was conducted on a subsample [14,15,16,17,18,19,20]. 10 studies were performed to validate existing datasets [8, 9, 15, 21,22,23,24,25,26,27]. Other studies had a data collection component in different countries including Ethiopia, South Africa, India, China, the Philippines, Pakistan, Bangladesh, Georgia, Haiti, Ghana, Tanzania, Indonesia, Iran, Kenya, Zimbabwe, Liberia, Malawi, Malaysia, and Mexico. . Uganda and Vietnam.

risk of bias

Risks of assessment bias are summarized in Supplementary File 3. Most studies had low risks across the individual domains assessed, however, there were 9 studies [15, 28,29,30,31,32,33,34,35] A representative sample or covering the entire population (area 2), did not collect 17 studies [15, 17, 29, 30, 32, 35,36,37,38,39,40,41,42,43,44,45,46] had a high risk of nonresponse (domain 3), 15 studies [16, 17, 20, 31, 32, 40, 43, 46,47,48,49,50,51,52,53] You did not use a validated tool (area 7).

VA tools and data collectors

Except for the ten studies [8, 9, 15, 21,22,23,24,25,26,27] that used existing databases, and the most widely used victim assistance tool was the WHO questionnaire (n= 16) [29, 39, 41, 51, 54,55,56,57,58,59,60,61,62,63] and the Population Health Metrics Research Consortium (PHMRC) questionnaire (n= 5) [8, 42, 64,65,66]. in 7 studies [18, 40, 43, 48, 53, 67, 68], the tool is not mentioned. In other studies, locally designed VA instruments have been used. in 5 studies [30, 32, 49, 50, 67]The data collectors were nurses, in 3 others [20, 51, 59] They were doctors at 10 [14, 29, 38, 43, 60, 61, 64, 66, 69, 70] The type of data collectors is not mentioned. Other studies included several other groups such as field workers and non-medical graduates.

in the majority (n= 38) From the validation studies, data were collected in the family setting of the deceased. in 4 studies [30, 35, 58, 71], a range of settings were used. in 10 studies [28, 29, 40, 44, 45, 52, 64, 68,69,70], the data collection setting is not mentioned. Other data collection settings during Hajj, hospital, and cemetery were included. The time between death and interview varied from immediately after receiving the body to 42 months. in 20 studies [16,17,18,19, 28, 34, 35, 42, 45, 46, 48, 52, 53, 57, 59, 64, 67,68,69,70]The time between death and interview was not mentioned. Interviewees in 18 studies were not identified [20, 28, 29, 40, 43,44,45,46, 56, 58, 62, 64, 65, 67,68,69,70,71]but in all others clearly described as family members, relatives, or primary caregivers at the time of death.

Population study

In 29 studies, the study population included the general adult population and in 15 studies [14, 30, 31, 33, 38, 39, 52, 54,55,56, 60, 61, 63, 70, 72] General neonatal or pediatric deaths are included. Three studies included female mortality only [34, 47, 73] Others were conducted on selected groups such as deaths due to HIV/TB, pilgrims and respiratory tract infections. In general, studies with overall mortality in adults had approximately equal male-to-female ratios. In three studies there were fewer than 50 interviews [30, 53, 74]In other cases, it ranged from 100 to more than 26,000.

Analysis techniques and comparative data set

Physician’s testimony was used in 36 studies and computer coded methods were used in 14 studies [8, 15, 18, 23, 24, 27, 29, 32, 42, 44, 50, 64, 66, 75]. Both methods were used in 11 studies [21, 22, 28, 40, 47, 52, 58, 65, 70, 71, 76] The method was not clearly stated in 6 studies [9, 26, 38, 45, 67, 68]. Comparative data was the primary source in the majority of studies (n= 53), while it was in 13 secondary sources.

The quality rating of the comparative data set is shown in Table 1. Most studies (n= 37) a high-quality comparative data set (level 1 or 2) that included the results of the investigation.

Table 1. Quality rating of the comparative data set

Measures used for comparison included percentage agreement, chance-corrected concordance (CCC), CSMF accuracy, kappa coefficient, concordance correlation coefficient, sensitivity, specificity, and predictive value, and statistical tests to explore significant differences. Eleven studies [8, 21, 23,24,25, 27, 58, 64, 65, 71, 75] CCC used 14 studies [8, 21,22,23,24,25,26,27, 58, 64, 65, 68, 71, 75] CSMF accuracy used, 16 studies [8, 18,19,20, 23, 28, 34, 39, 55, 57,58,59,60, 66, 67, 73] Use of the kappa coefficient, 38 studies [9, 16, 17, 20, 28, 29, 31,32,33,34,35, 37, 39, 40, 43,44,45,46, 48,49,50, 52,53,54,55,56,57, 59, 60, 62, 63, 66, 67, 69, 72, 73, 75, 77] Sensitivity, specificity, or positive predictive value are used. In general, studies have shown that there are generally accepted conventions. It was observed that specificity is higher than sensitivity in most of the studies that measured these two parameters. In addition, the negative predictive value was higher than the positive predictive value.

Leave a Reply

%d bloggers like this: