Methods for projecting all-cause or cause-specific mortality

Population projections that forecast the future size and age-composition of a country are crucial tools for appropriately planning the future allocation of societal resources. A projection model for countries with generalized HIV epidemics should take into account the future trajectory of the epidemic given the severe effect a generalized epidemic can have on the mortality conditions and composition of a population. We present a model of age-specific mortality as a function of life expectancy, HIV prevalence, and anti-retroviral therapy coverage for the 39 countries of the world experiencing a generalized HIV epidemic. We perform an in-sample validation where results show slight errors for several mortality indicators. Combined with the outputs of existing epidemiological and demographic models, this model makes it possible to estimate future mortality profiles for countries with generalized HIV epidemics.

The level of maternal mortality in Burkina Faso presents huge differences according to the sources of estimate. While the last census estimated a maternal mortality ratio to be 307 in 2006, United Nation agencies and the Demographic and Health survey provided for the year 2010, the numbers of 300 and 341 respectively. These fluctuations and inconsistencies in the estimates create confusion among policy makers and authorities. This study went through the contradictions and divergences to establish the good quality of census data compare to the others in terms of maternal mortality estimates. However, findings also highlighted the crucial need of improving the adjustment method use during census 2006. Furthermore, this study provided projected levels of maternal mortality ratio for the period 2006 to 2050.

Assumptions on future mortality play a key role for pension, health care and long-term care policies: the understanding of mortality trends is therefore of paramount importance for projections purposes, especially for the forerunners countries. In general, improvement for adult mortality is understood as declining of the mortality curve. However, Ishii (2008) has showed how adult mortality improvement could be better modelled by the shift-type than by decline-type model, such as the Lee-Carter model (Lee and Carter 1992), and he has proposed the Linear Difference (LD) model. In this study, we apply the LD model to the adult mortality for several European countries to analyse the trends of mortality improvements. Through the comparison of the parameters for the LD model between countries, we try to elucidate the peculiar features of mortality in Europe, to be used for projections purposes. These characteristics will be also compared to those of Japan, a benchmark country in the field of mortality, and the projected trends compared to those from other sources. Given the variety of mortality patterns existing in Europe, this study is also an important test about the performance of the LD model.

Smoking has been the most important non-linear determinant of mortality in low-mortality countries. With changes in smoking behaviour, projections of smoking-attributable and of all-cause mortality including smoking become more important, especially for health care programs and insurance. However, these projections might be sensible to the indirect techniques to estimate smoking-attributable mortality being used.

We estimate future smoking-attributable and all-cause mortality and analyse its sensitivity to different indirect techniques for estimating smoking-attributable mortality.

Future smoking-attributable mortality is obtained by applying different indirect estimation methods to projected lung cancer mortality, e.g. Peto-Lopez and, Preston-Glei-Wilmoth methods. Lung cancer mortality rates are extrapolated using age-period-cohort analysis. Non-smoking related mortality is projected using the Lee-Carter model.

Smoking-attributable mortality will further decline for males and first increase but then decline for females. The different indirect estimation techniques have an effect on smoking-attributable mortality levels and its age structure. Furthermore, they will lead to higher differences in projected smoking-attributable and all-cause mortality for women because of their shorter history in smoking.

In terms of competing risks Mixture Long-term Survival Models are widely used for the analysis of individuals may never suffer the considered cause of failure. Under condition of a cured fraction, some individuals will be treated as immune to a specific cause of failure or be defined as long-term survivors. In case of multi- or bivariate cause-specific survival data different dependence structures between variables can be suited with different copula functions. There are two main methodical aspects for the marginal distributions need to account for: first the maximum of flexibility and second the application in case of masked causes. We proposed a bivariate mixture long-term model based on the Farlie-Gumbel-Morgenstern (FGM)copula. Data simulations will be provided with SEER Breast Cancer Data, and comparing the model with different types of copulas e.g. FGM, Positive Stable, Frank and Clayton Copula. Otherwise we will discuss optional ideas for this approach in a semi-competing risk setting.

In this paper, we develop a generalised GARCH-based stochastic mortality model with a view to incorporate conditional heteroskedasticity and conditional non-normality in stochastic mortality modelling. We provide an empirical analysis of the UK mortality rates from 1922 to 2009 and find that both features are long-term behaviour of mortality structures. These structures impact the valuation and hedging of longevity-related insurance products and have been largely overlooked in many existing literature except for a very recent work by Giacometti et. al (2012), where only conditional heteroskedasticity is considered. To describe conditional non-normality, we adopt a Double Exponential distribution, also capable of incorporating the conditional skewness and leptokurtosis in our dataset. For the practical implementation, as in Siu, Tong and Yang (2004), we propose a user-friendly two-stage estimation scheme. At the first stage, we employ the Quasi-Maximum Likelihood Estimation (QMLE) to estimate the GARCH structure whilst at the second stage we adopt the MLE to estimate the Double Exponential parameters using residuals as inputs. We also examine the forecasting performance of the proposed model and find that the Double Exponential GARCH model provides reasonably good forecasts for future mortality developments.