246585 Association Rule Mining on National Survey Data: An Analysis of Timely Receipt of Preventive Care or Screening Examinations

Tuesday, November 1, 2011: 10:50 AM

Yoon-Ho Seol, PhD , Department of Health Informatics, Georgia Health Sciences University, Augusta, GA
Genny Carrillo-Zuniga, MD, MPH, ScD , School of Rural Public Health South Texas Center, Texas A&M, Health Science Center, McAllen, TX
Miguel A. Zuniga, MD, DrPH , Department of Health Policy and Management, Texas A&M Health Science Center, McAllen, TX
Given the current health care environment, there is a growing importance of practicing preventive care to enhance the effectiveness and quality of public health. Applying data mining methods on national survey data provides a practical approach to investigate risk factors that are related to the timely receipt of recommended preventive care. This study is to explore the potential of association rule mining in identifying relevant demographic and other factors, and to assess disparities in preventive care among population subgroups using national estimates.

In our experiments, data were extracted from the cross-sectional, nationally representative 2007 Medical Expenditure Panel Survey (MEPS) Household Component. The target population was the U.S. older adults aged 50 years or over. In developing association rules in the form of if-then rules, we selected 4 preventive care or screening examinations as target variables (consequent; the “then” part): dental checkup, prostate specific antigen (PSA) test, mammogram, and routine health checkup. For antecedent variables (the “if” part) of the rule, we used demographic and other variables including health status variables and access to care variables. For each consequent variable, we first used antecedent variables from each category (e.g., demographics and socioeconomic status), and then significant variables from each category were combined to develop potentially useful association rules. The strength of discovered association rules was evaluated by their support and confidence measures. We also tested varying support and confidence thresholds to detect rare association rules which have low support and high confidence measures. We then used antecedent variables of each selected rule to derive national estimates of the corresponding population subgroup. The findings of our experiments provided useful insights into risk factors that contributed to the timely receipt of preventive care, and also showed interesting population subgroups that demonstrated preventive care disparities among the U.S older adults.

Learning Areas:
Communication and informatics
Public health or related research

Learning Objectives:
1) Identify the characteristics of the U.S. older population that influence the receipt of preventive or screening examinations. 2) Recognize the potential and limitations of association rule analysis in assessing health care disparities using national health survey data.

Presenting author's disclosure statement:

Qualified on the content I am responsible for because: I am currently teaching a data mining course in the MPH program at Georgia Health Sciences University. In this study, I prepared and analyzed the data.
Any relevant financial relationships? No

I agree to comply with the American Public Health Association Conflict of Interest and Commercial Support Guidelines, and to disclose to the participants any off-label or experimental uses of a commercial product or service discussed in my presentation.