Comparison of physician-certified verbal autopsy with computer-coded verbal autopsy for cause of death assignment in hospitalized patients in low- and middle-income countries: systematic review

Table 1 Mean, ranges and number of reviewed studies for sensitivity and specificity estimates of physician-certified verbal autopsy for selected causes of death, among hospital-based deaths

	Sensitivity			Specificity
	Mean (%)	Range	Number of studies	Mean (%)	Range	Number of studies
Infections and parasitic diseases and maternal deaths
Tuberculosis	39	18 to 62	3	97	93 to 99	2
HIV/AIDS	59^a	0 to 61	3	90^a	0 to 96	3
Diarrheal diseases^b	38^a	0 to 75	2	96	94 to 99	2
Malaria^b	60^a	0 to 67	2	89^a	0 to 100	2
Pneumonia	42	18 to 75	4	93	84 to 99	3
Maternal deaths	63	-	1	100	-	1
Neonatal conditions
Prematurity or low birth weight	48	-	1	95	-	1
Birth asphyxia or birth trauma	43	-	1	90	-	1
Neonatal infections	31	11 to 50	2	100	99 to 100	2
Noncommunicable diseases
Nutritional conditions	33^a	0 to 58	2	94	87 to 99	2
Digestive cancers	82^a	56 to 96	3	100	99 to 100	2
Respiratory cancers	84	80 to 89	2	99	-	1
Other cancers	61^a	27 to 95	5	99	98 to 99	2
Heart disease	39	16 to 64	3	98	98 to 99	2
Stroke	71	63 to 82	3	95	94 to 97	2
Chronic respiratory diseases	61	60 to 62	2	98	-	1
Cirrhosis of the liver	58	45 to 71	2	98	-	1
Other digestive diseases	36	21 to 52	2	99	-	1
Renal or endocrine diseases	32	13 to 54	3	99	99 to 99.4	2
Injuries
Road traffic accidents	97	97 to 98	2	100	-	1
Other injuries	57	35.3 to 74	3	100	99 to 100	2

^aMedian was used instead of means due to outlier values in the ranges of estimates. However, the medians and means yielded similar results (that is, the pooled studies gave PCVA a sensitivity mean of 80.1% and a median of 81.9% for ascertaining digestive cancers). ^bAlthough two studies were used to generate results for these CODs, the studies provided results for several population sub-samples.

ISSN: 1741-7015