We obtained data from the 2007 National Health Interview Survey (NHIS),4
which uses a cross-sectional, multistage, stratified sampling design to question randomly selected households within the United States. The final sample included 23,393 respondents (response rate of 67.8%).4
We were interested in respondents who had used MBT (n=668) who were asked “In the past 12 months, did you use [a MBT] because it was recommended by a health care provider?”
Respondents answering “yes”
were classified as provider-referred users (P-MBT) and those answering “no”
were classified as self-referred users (S-MBT).
To account for NHIS’ complex sampling scheme, we conducted weighted analyses using SAS-callable SUDAAN (version 10.0). Potential correlates of P-MBT use were 1) sociodemographic characteristics (age, gender, race, education, income, region, marital status, and insurance status), 2) health status (current health, comparison to prior year, number of days in bed due to illness, and the Charlson Comorbidity Index), 3) pre-existing medical conditions (using the 16 most prevalent comorbidities in the United States), 4) health behaviors (smoking, alcohol use, exercise, and BMI), and 5) health utilization (office visits in the past year, ER visits in the past year, and encounters with general physicians, medical specialists and mental health professionals). Multivariable logistic regression modeling identified those factors independently associated with P-MBT use.