gms | German Medical Science

Background: In Germany, acute stroke is the most frequent cause of permanent invalidity in adulthood. Apart from individual consequences, this also results in considerable health economic burdens. Some vascular risk factors of strokes are known from epidemiological research and can in principle be influenced by preventive measures. But their influence on the disease is unexplored in the synopsis among themselves or in connection with other diseases, or e.g. the prescription of medics.

Methods: We utilized pseudonymized longitudinal data of a large statutory health insurance. Approximately 2.1 million persons, living in the German federal state of Saxony from the years 2008 to 2014, were included. Various methods of machine learning were used to examine correlations between stroke first occurrence or recurrence and drug therapies as well as existing patient comorbidities.

Results: A total of 4,759 patients with hemorrhagic stroke (I61, I62), 29,382 patients with ischemic stroke (I63, I64), and 13,253 patients with TIA (G45) were identified. Of these, 716 patients had a recurrence following hemorrhagic insult (15.1%), 3,686 patients had a recurrence after ischemic insult (12.6%), and 1,095 patients a recurrence after TIA (8.3%). In hemorrhagic stroke recurrence men are affected somewhat more frequently (17% vs. 13%). However, the recurrence frequency of both sexes decreases strongly with age (45-49 – 25% to 85+ – below 6%). In ischemic stroke and TIA recurrences both sexes were affected equally, but recurrence frequency remains almost identical across all age groups studied. Only in the highest age group 85+ has the frequency of recurrence of ischemic stroke decreased, which in turn may be due to underdiagnosis of mild cases in very old people. Known risk factors for incident stroke and recurrence could be replicated in the data set with ML methods. The presence of hypertension had the strongest influence on the risk of recurrence in all ML models. In seven drugs we saw an influence on ischemic stroke recurrence. Four of them acting as preventive factors in the presence of hypertension, but only marginally reducing the risk of recurrence.

Discussion: The project provides new insights into the epidemiology of stroke and recurrence in Saxony. Regional differences putatively point to deficiencies in the health care system. Data mining works. However, further data sources on behaviour, on clinical factors, and on family anamnesis and biomarkers are obviously necessary to elucidate a larger part of the variance, especially in first occurrences. The method appears very promising and offers an innovative approach to transfer the concept of Precision Medicine to prevention.