Prediction of future episodes of depression in primary medical care : development of a risk factor profile

Informazioni relative al progetto

PREDICT

ID dell’accordo di sovvenzione: QLG4-CT-2002-00638

Progetto chiuso

Data di avvio 1 Dicembre 2002

Data di completamento 31 Agosto 2006

Finanziato da

• Specific Programme for research, technological development and demonstration on "Quality of life and management of living resources", 1998-2002

Costo totale € 2 154 220,00

Contributo UE € 2 154 220,00

Coordinato da

UNIVERSITY OF LJUBLJANA
Slovenia

Risultati finali

Development of predictive models for risk of onset of depression

This study breaks new ground by quantifying the risk of future episodes of depression in primary care. The principal research objective was to develop a cross European multi-factor risk score for use by general practitioners to predict the onset of depression. This instrument was then validated in a non- European setting (Chile) to test its international applicability. We recruited primary care attendees from six European countries and Chile. 69% of the people approached took part and in total we recruited 10,048 people to the study. A diagnosis of DSMIV major depression was made using the Composite International Diagnostic Instrument and this assessment was repeated six and 12 months later. All participants were also assessed at recruitment to the study with a battery of known risk factors (39 in total) and six months later. Participation rates at the six and twelve months follow up interviews were 90% and 86% respectively. Only patient not depressed at baseline were included in the prospective analysis. The cumulative incidence of DSM IV depression over 12 months was 7.8%. A risk equation emerged from the European data based on 11 of the 39 risk factors measured at baseline. These were found to be significant independent predictors of new onset depression. The PREDICT risk equation that provides a probability of depression for each attendee over 12 months performed extremely well in terms of the Area Under the Receiver Operating characteristics Curve (AUROC). We calculated a shrinkage factor to adjust the prediction model for over fitting in the data from which it was developed (i.e. data from the six European countries). We also validated our prediction rules on our external non European (Chilean) dataset. This process is essential in the development of a valid and accurate prediction tool. However, the development of most clinical prediction tools seldom undertake external validation (Altman & Royston, 2000) and shrinkage for potential over fitting is a relatively new concept (Moons et al, 2004). In this study we created an innovative risk tool that compares favourably with a recent less rigorously evaluated European risk tool used to predict cardiovascular events over 10 years (Conroy et al, 2003). We have produced the first prediction rule for clinical depression in primary care. To our knowledge, there are no similar tools in psychiatry, which makes our instrument a first of its kind. Over the next three years we will disseminate our finding through publications in high impact international journals of general medicine, psychiatry and primary care. We will also present our findings at key international conferences on psychiatry, public health and primary care. Our presentations to date have already generated considerable interest. In the future, we will use the results of our research to develop a risk assessment intervention tool for possible prevention of depression in primary care. This instrument will be evaluated in a randomised trial and we will use the outcome of our trial to recommend the application of this instrument to clinical family practice. Although it difficult to assess the impact of such a technology on health at this stage, we expect that prevention of depression will lead to substantial health gains, reduction in days off work due to ill health, reduced disability and economic benefits to patients and society.