Analysing duration of episodes of pharmacological care: An example of antidepressant use in Danish general practice
Research output: Contribution to journal › Journal article › peer-review
Analysing duration of treatment episodes has become a standard task in many pharmacoepidemiological studies. However, such analyses are often carried out in a rather simplistic manner and more subtle issues are often ignored. In this paper, methods of analysing duration treatment episodes beyond simple analyses allowing investigation of the risk for certain events over time are demonstrated. In particular, the use of cumulative incidence functions, cause-specific hazard functions, hazard rate models and expected mortality in analysing duration of episodes is presented. We used these statistical techniques in analysing the early treatment history of patients who started a regular treatment with antidepressant drugs in the primary health care sector in Denmark. We have extracted some important features: The risk of discontinuing and switching treatment was very high around 10 weeks after starting treatment. After discontinuing the first treatment period, many patients soon started a second treatment period depending on the duration of the first treatment period with highest risk around 10 weeks. The mortality rate among the patients in treatment was about three times higher than the expected mortality. The risk of dying immediately after stopping treatment was about twice the expected mortality. The analysis suggests that: (1) there is a critical period for a first discontinuing, switching and restarting treatment around 10 weeks, (2) the GPs prescribing habits have more influence on the patterns than patient or drug characteristics, (3) over time Danish GPs tend to prolong the duration of first treatment period and avoid longer treatment breaks.
|Journal||Pharmacoepidemiology and Drug Safety|
|Number of pages||11|
|Publication status||Published - Mar 2006|
- Antidepressant, Cause specific hazard, Cumulative incidence function, Hazard ratio functions, Prescriptions data, Proportional hazards models, Treatment episodes