Civil registration and vital statistics systems are of critical importance in providing necessary information to assess the general health conditions of a population. Yet, in most developing countries, these systems are either inexistent or incomplete. Demographers and researchers in other fields have long been using different techniques to indirectly estimate either summary mortality indices such as probability of dying from birth to age 5 (5q0), or a full life table that contains various indices describing the dying out of a hypothetical population cohort.
In this paper, we propose a new model life table system that satisfies the following desirable attributes. First, it is parsimonious and requires only a few entry parameters to generate a full life table. Second, it could adequately capture the range of age patterns of mortality observed in real populations and yield high predictive validity, not just measured by summary indices such as life expectancy at birth, but more importantly by age specific mortality rates. Third, it provides satisfactory estimates of age specific mortality for countries with high levels of mortality, especially those plagued by the HIV/AIDS epidemic. Finally, it could generate age specific mortality with a plausible time trend, and the partial derivative of entry parameters such as 5q0 and 45q15 should be positive with respect to age specific mortality.
Our new model life table system is built upon a database with 7,294 empirical life tables, among which 1,174 (16.4%) are from populations with over 0.1% HIV seroprevalence among adults with ages 15 to 59. Preliminary results show that this new model exhibits both greater versatility and high predictive validity in estimating age specific mortality than various life table systems. In addition, our model's ability to estimate age specific mortality rates in the presence of significant HIV/AIDS epidemic is likely to be of considerable value in providing mortality estimates for countries plagued by the epidemic.
