Premium
Would you like to add a weight after this blood pressure, doctor? D iscovery of potentially actionable associations between the provision of multiple screens in primary care
Author(s) -
Kalia Sumeet,
Greiver Michelle,
Zhao Xu,
Meaney Christopher,
Moineddin Rahim,
Aliarzadeh Babak,
Grunfeld Eva,
Sullivan Frank
Publication year - 2018
Publication title -
journal of evaluation in clinical practice
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.737
H-Index - 73
eISSN - 1365-2753
pISSN - 1356-1294
DOI - 10.1111/jep.12877
Subject(s) - medicine , cronbach's alpha , waist , primary care , medical record , body mass index , lift (data mining) , family medicine , data mining , computer science , psychometrics , clinical psychology
Rationale, aims, and objective Guidelines recommend screening for risk factors associated with chronic diseases but current electronic prompts have limited effects. Our objective was to discover and rank associations between the presence of screens to plan more efficient prompts in primary care. Methods Risk factors with the greatest impact on chronic diseases are associated with blood pressure, body mass index, waist circumference, glycaemic and lipid levels, smoking, alcohol use, diet, and exercise. We looked for associations between the presence of screens for these in electronic medical records. We used association rule mining to describe relationships among items, factor analysis to find latent categories, and Cronbach α to quantify consistency within latent categories. Results Data from 92 140 patients in or around Toronto, Ontario, were included. We found positive correlations (lift >1) between the presence of all screens. The presence of any screen was associated with confidence greater than 80% that other data on items with high prevalence (blood pressure, glycaemic and lipid levels, or smoking) would also be present. A cluster of rules predicting the presence of blood pressure were ranked highest using measures of interestingness such as standardized lift. We found 3 latent categories using factor analysis; these were laboratory tests, vital signs, and lifestyle factors; Cronbach α ranged between .58 for lifestyle factors and .88 for laboratory tests. Conclusions Associations between the provision of important screens can be discovered and ranked. Rules with promising combinations of associated screens could be used to implement data driven alerts.