Substudy 2 Executive Summary
Care Trajectories: The Natural History of Clients Moving Through the Continuing Care System
by Marcus J. Hollander
It was initially expected that there would be four to six common trajectories or patterns of movement which would account for a significant portion of all patterns. It was also assumed that people would progress through the system of care in a certain logical manner, entering at a low level of care in the community, moving up one or two care levels, then moving to a facility where they would move up through one or two care levels and then pass away.
The findings from this study indicate that, unlike our initial expectations, there was a wide variety of care trajectories, none with a large percentage of the clients. Within the ten years of data available to us, the most common pattern was for clients to enter the system at a given level and type of care and die without any changes in the level and type of care. The implications of these two results must be considered by clinicians, planners and policy makers.
For those who entered care in the community at level one, none of the top ten patterns of care included a move to a facility. However, at higher starting levels of care, there was a general trend toward an increasing proportion of clients having patterns which involved moves to a facility. In the vast majority of cases, those who entered care in a facility remained in facility care until they passed away or until the end of the ten years of data available to us.
Some 15 Markovian and non-Markovian forecasting methods were tested on nine years of actual data on the states of health of 6384 clients of the BC Continuing Care System. Markovian methods were shown to be more robust both in times of rapid and modest change. The forecasting errors from Markovian methods were small and were in general smaller than the errors for the non-Markovian methods (moving average, exponential smoothing, linear regression) against which they were compared. Further testing should be undertaken to determine how well Markovian methods work on longer range forecasts and on smaller populations.