Background
Kiribati is an atoll country of 103,058 (2010 Census) situated in the central Pacific. Previous mortality estimates have been derived from demographic analyses of census data. This is the first mortality analysis based on reported deaths.

Methods
Recorded deaths were from the Ministry of Health and the Civil Registration Office for 2000–2009; populations were from the 2000, 2005, and 2010 censuses. Duplicate death records were removed by matching deaths within and between data sources using a combination of names, date of death, age, sex, island of residence, and cause of death. Probability of dying <5 years (5q0) and 15–59 years (45q15), and life expectancy (LE) at birth, were computed with 95 % confidence intervals. These data were compared with previous census analyses.

Results
There were 8,681 unique deaths reported over the decade 2000–2009 in Kiribati. The reconciled mortality data indicate 5q0 for both sexes of 64 per 1,000 live births in 2000–2004, and 51 for 2005–2009 (assuming no under-enumeration), compared with 69 and 59 for comparable periods from the 2005 and 2010 census analyses (children ever-born/children surviving method). Based on reconciled deaths, LE at birth (e0) for males was 54 years for 2000–2004 and 55 years in 2005–2009, five years lower than the 2005 and 2010 census estimates for comparable periods of 59 and 58 years. Female LE was 62 years for 2000–2004 and 63 years for 2005–2009, two-three years less than estimates for comparable periods of 63 and 66 years from the 2005 and 2010 census analyses. Adult mortality (45q15) was 47-48 % in males and 27-28 % in females from reconciled mortality over 2000–2009, higher than census estimates of 34-38 % in males and 21-26 % in females for the same periods. The reconciled data are very likely to be incomplete and actual mortality higher and life expectancy lower than reported here.

Conclusion
This analysis indicates higher mortality than indirect demographic methods from the 2005 and 2010 Censuses. Reported deaths are most likely under-reported; especially 5q0, as many early neonatal deaths are probably classified as stillbirths. These analyses suggest that the health situation in Kiribati is more serious and urgent than previously appreciated.