Use of association rule mining to study smoking interventions in primary care
Dr. Yue Huang (UKCTCS)
12.15, Tuesday 9 July. The Clubhouse, level 4
Association rule mining (ARM), a form of data mining, is used for discovering new, unexpected and potential useful relationships between variables in a dataset. ARM offers an alternative approach to the analysis of medical activity in large healthcare databases, and in particular to assessing equality of health care provision.
The ARM method involves identifying strong rules, ‘association rules’, among different variables. The concept of ARM was introduced to discover patterns of purchasing in supermarket data (the market basket analysis problem). For example, the rule X={Bread, Butter} à Y={Milk} means if people buy bread and butter, they will also buy milk, and such rules would be used in decision making about marketing activities. ARM has been widely applied in many areas of business analysis to identify patterns or combinations of events which occur together, but has been little used in public health and epidemiology, yet it has huge potential in this context as it provides a structured way of exploring patterns in data that are not hypothesis-driven.
We have used association rule mining to investigate the patterns of prescribing of smoking cessation medications in an electronic primary care dataset, and to identify the characteristics of numerically important groups of patients who typically do, or do not, receive cessation therapy. We found that prescribing is still underused among younger smokers and those with co-morbidity, particularly dementia, high alcohol intake, atrial fibrillation and chronic renal disease.
This novel approach identified sizeable and easily definable groups of patients who are systematically failing to receive support for smoking cessation in primary care. Association rule mining is a powerful means of identifying those at high and low risk of receiving a public health or other healthcare intervention and hence potentially improving healthcare delivery.
The ARM method involves identifying strong rules, ‘association rules’, among different variables. The concept of ARM was introduced to discover patterns of purchasing in supermarket data (the market basket analysis problem). For example, the rule X={Bread, Butter} à Y={Milk} means if people buy bread and butter, they will also buy milk, and such rules would be used in decision making about marketing activities. ARM has been widely applied in many areas of business analysis to identify patterns or combinations of events which occur together, but has been little used in public health and epidemiology, yet it has huge potential in this context as it provides a structured way of exploring patterns in data that are not hypothesis-driven.
We have used association rule mining to investigate the patterns of prescribing of smoking cessation medications in an electronic primary care dataset, and to identify the characteristics of numerically important groups of patients who typically do, or do not, receive cessation therapy. We found that prescribing is still underused among younger smokers and those with co-morbidity, particularly dementia, high alcohol intake, atrial fibrillation and chronic renal disease.
This novel approach identified sizeable and easily definable groups of patients who are systematically failing to receive support for smoking cessation in primary care. Association rule mining is a powerful means of identifying those at high and low risk of receiving a public health or other healthcare intervention and hence potentially improving healthcare delivery.