TABLE 4.

Two-part models: AORs and 95% CIs for depression and intercept and co-efficient for depression from OLS on 2009 log-transformed dollars in the Medical Expenditure Panel Survey, 2006–2009^a

Type of service	Logistic regression on Use, AOR [95% CI]	OLS on log-transformed $ in users, beta (SE)	% change
Inpatient	1.23 [0.97, 1.57]	0.131 (0.000)***	14.0
ED	1.46 [1.17, 1.82]***	0.097 (0.000)***	10.2
Other	1.26 [0.99, 1.61]	0.123 (0.018)***	13.1

^aBased on 4,766 adult cancer survivors older than 21 years. Asterisks denote statistical significance in parameter estimates from logistic regressions on use of services or ordinary least squares regressions on log-transformed 2009 dollars in users. Percentage change in expenditures by depression status were calculated by exponentiating regression co-efficient and subtracting one (eβ−1). All regression models adjusted for gender, age, race/ethnicity, metro status, education, poverty status, health insurance, perceived physical and mental health, cardiovascular disease, number of other conditions, anxiety, body mass index, smoking status, and year of observation.

AOR, adjusted odds ratio; CI, confidence interval; ED, emergency room; OLS, ordinary least squares

^***P < .001

^**.001 ≤ P < .01

^*.01 ≤ P < .05