Data mining refers to statistical methods used to identify patterns of associations and explore relationships in large datasets where the traditional hypothesis testing and “confirmatory” approaches are not be feasible or efficient. Broadly speaking, data mining also describes the exploratory analyses of large datasets to address secondary research questions from data originally collected for other purposes. Modern data mining is made possible by the tremendous advances in computing power, the availability of rapidly assembled electronic databases, and the development of sophisticated algorithms to explore and to detect complex relationships without the need of traditional “classical” statistical assumptions. Data mining methods typically approach the task of data exploration by classification (e.g., neural network), regression (e.g., tree regression), and clustering (e.g., nearest neighbor) methods. Such methods are applied to adverse event data and to other administrative healthcare data to monitor and evaluate the health and safety of pharmaceutical and medical device products, or to address other health care questions aimed at characterization of a potential patient population, addressing provider treatment patterns, or patient behaviors related to pharmaceutical products.
Exponent statisticians and epidemiologists have expertise in these analytical methods to evaluate potential relationships in available epidemiologic and health care data. Our research team also has extensive expertise in the data management and data quality issues involved with the handling of large administrative databases. Exponent health scientists and biostatisticians have consulted with pharmaceutical and medical device clients to assist them in:
- Identifying patient, provider, and institutional factors that affect the risk of post-surgical complications following hip replacement surgery
- Assessing disease prevalence, characteristics of selected subgroups of targeted potential patient populations to determine future needs for surgical treatment or pharmaceutical products
- Projections of the size of the future anticipated workforce of specialized healthcare providers needed (e.g., orthopedic surgeons) to meet growing needs from increasing number of elderly patients with joint disease
- Evaluating current treatment practice and to estimate potential savings in health care expenditures by using alternative or less-invasive treatment methods
Exponent health sciences, biostatistical and data programming staff apply statistical and data management procedures in conducting both exploratory techniques like data mining as well as traditional analyses involving confirmatory hypothesis testing and risk estimation. We work to assist clients in achieving the maximum information from available data to address a variety of research questions in pharmaceutical product safety, treatment compliance and behavior, and characterization of patient populations.