Transcript Slide 1
Assessing Consumer Health Vocabulary Familiarity: An Exploratory Study
Alla Keselman 1,2 Tony Tse 1 , Jon Crowell 3 Allen Browne 1 Long Ngo 3 Qing Zeng 3 1 – US National Library of Medicine 2 – Aquilent, Inc.
3 – Harvard Medical School
Study Background
Consumers have difficulty with health texts
Study Background
Consumers have difficulty with health texts We would like to provide support – Authoring guidelines; tools; translators Need a way to evaluate readability – Readability formulas Health domain is unique – Familiar long words (diabetes); unfamiliar short words (apnea)
Term Familiarity Likelihood Regression Model
Computational (regression) model – Each term is assigned 0 – 1 score Algorithm basis – Empirical data – Term frequency counts from health text corpora Term score categories – 0.8 – 1.0 score – “likely to be familiar” – – 0.5 – 0.8 score – “somewhat likely to be familiar” 0.0 – 0.5 score – “not likely to be familiar” Source
:
Zeng Q, Kim E, Crowell J, Tse T. A text corpora-based estimation of the familiarity of health terminology. Proc ISBMDA 2005: 184-92.
Objectives
Validate regression model – Test with consumers Effect of demographic factors on familiarity – – Health literacy Education level Relate surface-level and conceptual familiarity – Term vs. concept
Hypotheses
I.
II.
III.
Significant effect of predicted familiarity likelihood 1. Surface-level familiarity 2. Conceptual familiarity Significant effect of demographic factors Surface level familiarity > conceptual
Survey Instrument
45 items – hypertension, back pain, GERD (gastroesophageal reflux) Random set of terms from MedlinePlus Two types of test items: – Surface-level – prominent association Surgery => knife – Concept level Surgery => removing or repairing a body part 45 surface questions; 15 concept questions (GERD)
Item Format
*Modeled on the Short Assessment of Health Literacy for Spanish-speaking Adults (SAHLSA) Lee S-YD, Bender DE, Ruiz RE, Cho YI. Development of an easy-to-use Spanish health literacy test. Health Serv Res. In press.
Participants
Procedure
Demographic survey Short Test of Functional Health Literacy in Adults (S-TOFHLA) Familiarity test
Decrease
Results
Results
Decrease
Results
Predictors of Surface-Level Familiarity
Regression I – DV: surface level term familiarity – IV: Predicted Familiarity Likelihood Level, Gender, English proficiency, Highest Education Level, Age, Race, Health Literacy Level Significant predictors – Predicted Familiarity Likelihood (P<.001) – – Health Literacy (P<.001) English Proficiency (P=.05)
Confirms Hypothesis I Confirms Hypothesis II
Predictors of GERD Concept Familiarity
Regression II (GERD) – DV: GERD concept familiarity – IV: Predicted Familiarity Likelihood Level, GERD surface level familiarity Gender, English proficiency, Highest Education Level, Age, Race, Health Literacy Level
Predictors of GERD Concept Familiarity
Regression II (GERD) – DV: GERD concept familiarity – IV: Predicted Familiarity Likelihood Level, GERD surface level familiarity Gender, English proficiency, Highest Education Level, Age, Race, Health Literacy Level Significant predictors – Predicted Familiarity Likelihood (P=.009) – GERD surface-level familiarity score (P<.001) Health Literacy (P.06) - trend
Confirms Hypothesis I
Predictors of GERD Concept Familiarity
Regression II (GERD) – DV: GERD concept familiarity – IV: Predicted Familiarity Likelihood Level, GERD surface level familiarity Gender, English proficiency, Highest Education Level, Age, Race, Health Literacy Level Significant predictors – Predicted Familiarity Likelihood (P=.009) – GERD surface-level familiarity score (P<.001) Health Literacy (P.06) - trend
Addresses Hypothesis III
Predictors of GERD Concept Familiarity
Regression II (GERD) – DV: GERD concept familiarity – IV: Predicted Familiarity Likelihood Level, GERD surface level familiarity Gender, English proficiency, Highest Education Level, Age, Race, Health Literacy Level Significant predictors – Predicted Familiarity Likelihood (P=.009) – GERD surface-level familiarity score (P<.001) Health Literacy (P.06) - trend
Trend for Hypothesis II
Relationship Between Surface Level and Concept Familiarity (GERD)
5 4.5
4 3.5
3 2.5
2 1.5
1 0.5
0 1
Familiarity type
Gap between surface and concept familiarity (P=.001) Size of gap greater for “likely” than for “unlikely” (P=.006) Trend for “somewhat likely” vs. “unlikely” (P=.07) 2 "Likely to be familiar" "Somewhat likely" "Unlikely"
Conclusions
Initial validity evidence for CHV familiarity model – Health readability utility Ways to improve the model – – Allow demographic corrections Distinguish between knowledge of terms / concepts Follow-up work – – – – Increase sample and term pool Education level?
Other predictors?
Work on integrated findings into health readability formula
Acknowledgements
Intramural Research Program of the US National Library of Medicine, US National Institutes of Health NIH grant R01 LM007222-05 Ilyse Rosenberg for medical expertise Cara Hefner for help with data collection