Frans van Poppel

A LONGITUDINAL STUDY OF HEALTH SELECTION IN MARITAL TRANSITIONS

We examined whether differences in health were associated with different probabilities of marital transitions in a longitudinal study, using Cox proportional hazard analysis. Data on approximately 10,000 Dutch persons of the GLOBE study, aged 15-74 years, were used for this purpose. The study started in 1991 and study subjects have been followed for 4.5 years. Of the four marital transitions studied (marriage among never married and divorced persons, and divorce and bereavement among married persons), only divorce among married persons was associated with health status: married persons who reported four or more subjective health complaints or two or more chronic conditions were, respectively, 1.5 and two times more likely to become divorced during follow-up than persons without these health problems. Since hardly any other studies have examined the role of health selection in marital transition with longitudinal data, more research is required before firm conclusions can be drawn. It can be concluded, however, that the frequently made assumption that health selection contributes only little to the explanation of health differences between marital status groups, seems, at least for the divorced, not justified.

Long-term trends in marital status mortality differences in The Netherlands 1850-1970.

This article describes the long-term trends in marital status mortality differences in the Netherlands using a unique dataset relating to the period 1850-1970. Poisson regression analysis was applied to calculate relative mortality risks by marital status. For two periods, cause-of-death by marital status could be used. Clear differences in mortality by marital status were observed, with strongly increasing advantages for married men and women and a relative increase in the mortality of widowed compared with non-married people. Excess mortality among single and formerly married men and women was visible in many cause-of-death categories, and this became more widespread during the last decades of the nineteenth century. Hypotheses are formulated that might explain why married men and women underwent a stronger decrease in mortality up until the end of World War II.

MORTALITY DECLINE IN THE NETHERLANDS IN THE PERIOD 1850-1992: A TURNING POINT ANALYSIS

The aim of this paper is to give a detailed and fairly objective description of rapid mortality decline in The Netherlands between 1850 and 1992 with respect to the start, end, and phases of the decline. Turning points were estimated for the standardized mortality trend, and for age and sex-specific trends between 1850-1992. The technique used was derived from spline functions. The turning points divided the trends into phases with different paces of decline. Standardized mortality started to decline rapidly in The Netherlands around 1880. Four phases in the period of decline could be distinguished: 1880-1917 (1.2% annually), 1917-1955 (1.6%), 1955-1970 (0.4%), 1970-1992 (1.1%). For nearly all age groups, the most rapid decline occurred in a period comparable to 1917-1955. Causes of death which might have shaped the standardized mortality trend are, among others, respiratory tuberculosis (1917), heart disease (except ischemic) (1955), and ischemic heart disease (1970). Causes of death that shaped the mortality trend are related to trends of determinants of mortality decline. The technique used in this paper can also be applied to other trends e.g. fertility decline.

Parameterisation as a tool in analysing age, period and cohort effects on mortality: A case study of the Netherlands

Parameterisation models have been applied by demographers, medical scientists, and actuaries for about 250 years, being always a useful tool in the analysis of mortality. Their usefulness has been limited in the past to smoothing data, eliminating and/or reducing errors, constructing life tables, estimating missing data, facilitating comparisons of mortality, and forecasting (Keyfitz, 1982; Manton, Stallard, 1991). Recently, the role of parameterisation models has considerably broadened. Completely new areas of applications for parameterisation emerged, such as modelling disease processes (Manton, Stallard, 1984; 1988), solving specification problems of the traditional ageperiod-cohort (apc) models (Holford, Zhang, McKey, 1994), and contributing to a new type of symptomatic models for mortality (Lee, Carter, 1992). The symptomatic nature of parameterisation models is also subject of this chapter. First of all, however, some technical questions are discussed. In the technical part of this contribution, an attempt is made to systemise parameterisation functions for mortality and in particular to decide which representations of mortality are preferable as dependent variables in these functions. This goal is a natural consequence of the fact that the number of existing parameterisation models is rather high and some guidelines are needed to select a proper function. Secondly, we show how useful parameterisation can be in analysing age, period,and cohort effects on mortality. The motivation for the second goal originates from the opinion that parameterisation can be seen as an alternative to age, period, and cohort modelling.