Introduction

For a few decades now, the development in mental health care has been guided by two main principles, which are based on humanitarian as well as financial factors. The first principle is that patients should be kept and treated in the community as much as possible. The second is that ‘continuity of care’ should be provided as a flexible response to variations in need of care in a patient career. The realisation of these goals necessitates restructuring of mental health care, in which three processes can be distinguished: deinstitutionalisation (decreasing the number of beds in mental hospitals, increasing dayand outpatient care); sectorisation (the establishment of mental health regions) and integration of in-, day- and outpatient services.

To evaluate these processes properly we should take into account that:

1. the whole system of mental health care should be studied and not just specific programmes or new facilities, because changes in one part of the system will often affect others;

2. to evaluate the development of the mental health care system long-term longitudinal studies are needed, because these processes evolve very slowly; and

3. the functioning of mental health care is reflected in patterns of service utilisation. This is the care that has come to function within a given period of time.

Ultimately, planning is directed towards altering these patterns. Therefore, we can evaluate restructuring processes aimed at reaching certain goals by studying patterns of service utilisation. In order to obtain unbiased data about service utilisation the use of a psychiatric case register is indispensable. Until now, each patient's career in care tends to be described in register studies in terms of one summary statistic (Sytema et ai, 1989). Yet, a patient's career in modern mental health care is a dynamic process in which several ‘transitions’ to and from various types of care may occur. Moreover, new statistical techniques, known as hazard models, have become available allowing the study of'change’ in patients’ careers. In this chapter we will outline a simple approach to a dynamic analysis of patterns of care. First the basic concepts will be described, then the Cox's proportional hazards model will be introduced, and finally an example, using the data of a cohort of schizophrenic patients, will be given. In the Appendix the reader will find an example of how to organise the data.