标题：Presenting the status of project SIFT – Research of the patient flows and structures in health and social services in three Finnish cities in 2009–2015
出版社：Utrecht University, Maastricht University, Groningen University
摘要：Introduction : Project “SIFT” aims at producing information for decision-makers involved in the fundamental reform of Finnish health and social services. SIFT focuses on heavy users of these services, and, more specifically on those appearing in three registers: primary care, secondary care and social services. Our data covers 8834 clients within three cities in Southern Finland (totally ca. 85 000 inhabitants), between January 2009 and June 2015. In this sub-study, we focus on 4270 clients who have at least once during the study period used mental health services. Could some improvements be done in the processes or integration of services? Could some predictions be done concerning service use at the end of the period? Methods : We created 14 classes based on 1) age, 2) total use of health and social care services, 3) percentage of secondary care, 4) maximum number of consecutive months without services and 5) likelihood of staying within mental health services. Classification was done with Self Organizing Map followed by hierarchical clustering. Use and timelines of services and distribution of ICD10-diagnoses between groups were studied. Prediction tests were run with Random Forest -algorithm. Results : Despite only one of the classification variables was related to mental health service use, combining it with four other relevant variables helped us differentiate various types of use of mental health services. Clear differences were seen between the classes in the distribution and stability of top diagnoses in secondary care over the 6.5 years. Related to typical diagnoses, also the character of returning to mental health services varied. In the heavy-use classes it was of course uncommon to change from secondary care to “no care”, but also uncommon to change to primary care within the 6 months following a visit in the secondary care. The children in our study are clients of child protective services or services for the disables, but for many adult clients, social care fell in the class of income support. Simple parameters describing the amounts of various services or diagnoses at main level (e.g. F) in 2009 did not succeed to predict the class in 2015. Discussion and conclusions : Especially in the case of heaviest service-use, there is a great potential for improvement of client experience by better service integration. We see that the level of integration should differ for different client types: from better information flow within a sub-branch into involvement of professionals from several branches. For predictive models to work, more specific input variables from a longer time period are needed and we are proceeding this way ahead. Limitations : We did not have access to private sector data. However, the heavy users of the health and social services mainly rely on the public sector in Finland. Furthermore, the time span is relatively short. Suggestions for future research : Plenty of work was required before ending up with a coherent data set. Integration of registers is a necessary requirement for this type of work to succeed.
关键词：heavy users of health and social services ; data fusion ; predictive model
