Predictors of Endoscopy in Minority Women: RESULTS

Posted by James

Demographics
The two groups were demographically similar (Table 1). However, 32% of Hispanics completed the interview in English compared to 98% of the women in the African-American group (p<0.001). Additionally, African-American women were more likely to have an income >$ 10,000 (p<0.013).

Table 1. Demographic characteristics of the study population

Variable	African Americans	Hispanics	Sig
N	44	37
Age <65 >65	12 (27%) 32 (73%)	18     (49%) 19     (51%)	0.470
Language of Interview English Spanish	43 (98%) 1 (2%)	12 (32%) 25 (68%)	0.001***
Income <$ 10,000 >$ 10,000	21 (51%) 20 (49%)	26 (81%) 6 (19%)	0.013*
Marital Status Married or living with partner No married or living partner	6 (17%) 38 (83%)	7 (19%) 30 (81%)	0.520
Education < High school > High school	33 (75%) 11 (25%)	29 (78%) 8 (22%)	0.720
Insurance Federal Commercial None Missing (5) 6% * p<0.05; ** p<0.01; *•* p<0.001	33 (75%) 7 (15%) 1 (2%)	26 (70%) 5 (14%) 4(11%)	0.807

Univariate Analysis
Socioeconomic Factors and Endoscopy. We combined flexible sigmoidoscopy and colonoscopy into a single variable reflecting having had (or not) an endoscopic examination. Forty-nine percent of the participants reported having had either flexible sigmoidoscopy or colonoscopy (Table 2). This rate reflects participants having had endoscopy for screening as well as for diagnostic indications. African-American women were more likely than Hispanic women to have been recommended for or ever having had an endoscopy (p=0.002). Women age >65 were significantly more likely to have had endoscopy than those age <65 (p=0.012). There was a tendency for women who had completed high school or beyond to have had flexible sigmoidoscopy or colonoscopy (p=0.052). Income and marital status were not associated with endoscopy in this population. English-speaking Hispanics were significantly more likely to have had endoscopy compared to primarily Spanish-speaking Hispanics, 32% vs. 7%, respectively (p=0.009). There was no difference in endoscopy rates between women who had an annual income less than or greater than $10,000, married versus not married, or between federally insured and commercially insured patients.

Table 2. Socioeconomic factors and endoscopy rates

Variable	Endoscopy	No Endoscopy	Significance Level
Total	39 (49%)	42 (51%)
Age 50-64 >65	9 (23%) 30 (77%)	21 (50%) 21 (50%)	0.012*
Ethnicity African-American Hispanic	28 (72%) 11 (28%)	16 (38%) 26 (62%)	0.002**
Language of Interview English Spanish	32 (82%) 7 (10%)	23 (55%) 19 (45%)	0.009**
Education < High school > High school	13 (33%) 26 (66%)	23 (55%) 19 (45%)	0.052
Income <$ 10,000 >$ 10,000	22 (61%) 14 (39%)	25 (68%) 12 (32%)	0.565
Marital Status Married or living with partner Not married or living with partner	6 (15%) 33 (85%)	7 (17%) 34 (83%)	0.875
Insurance Federal Commercial None *p<0.05; **p<0.01; ***p<0.001	27 (71%) 9 (24%) 2 (5%)	32 (84%) 3 (8%) 3 (8%)	0.163

Medical Factors and Endoscopy. Physician recommendation for mammography (88%) and Pap tests (94%) was higher than for FOBT (83%) and endoscopy (53%). The study population reported higher rates of mammography and pap tests, 95% and 96%, respectively [Fatone et al., East Harlem Partnership for Cancer Awareness (EHPCA): collaborative cancer screening and prevention research in an urban minority community, submitted for publication 2004]. The most significant variable associated with endoscopy was physician recommendation (pO.OOOl) (Table 3). However, the increased rates for mammography and Pap test usage were not associated with increased rates for endoscopy. Whether the participant had a primary care physician or a personal or family history of cancer was not significantly different in patients who underwent endoscopy and those who did not.

Table 3. Medical factors and endoscopy rates

Variable	Endoscopy	No Endoscopy	Significance Level
Primary Care Physician Yes No	33 (47.1%) 6 (60%)	37 (52.9%) 4 (40%)	0.447
MD Referral for Endoscopy Yes No	36 (92%) 3 (8%)	7 (17%) 35 (83%)	0.000***
Personal Cancer History Yes No	3 (50%) 36 (50%)	3 (50%) 36 (50%)	0.610
Family History Yes No	26 (52%) 12 (41.4%)	24 (48%) 17 (58.6%)	0.362
Breast Cancer (Arimidex tablet is used to treat breast cancer)Yes No	36 (92%) 2 (8%)	41 (100%) 0	0.137
Cervical Cancer Screening Yes No *** pO.001	38 (97%) 1 (3%)	40 (93%) 0	0.308

Psychosocial Factors and Endoscopy. The only psychosocial factor that was associated with endoscopy was cancer cons (p=0.029), with a tenden cy toward cancer pros (p=0.053) (Table 4). Participants with a higher degree of cancer (Generic Hydrea reducing the number of painful episodes and blood transfusions needed by adults) cons were less likely to have had endoscopy. However, participants who had higher pros scores tended toward endoscopy.

Table 4. Psychosocial factors and endoscopy rates

Variable	Endoscopy (Mean Score)	Standard Deviation	No Endoscopy (Mean Score)	Standard Deviation	Significance Level
Cancer pros	26.43	3.46	24.58	4.60	0.053
Cancer cons	32.73	11.41	38.53	11.49	0.029*
Cancer worry	5.86	3.55	5.78	3.03	0.901
Acculturation
African-American	1.97	0.368	2.19	0.311	0.058
Hispanic	1.76	0.335	1.67	0.440	0.547
Medical mistrust	28.19	9.19	30.54	10.26	0.308
Fatalism	6.69	3.58	7.48	3.75	0.983
Temporal Orientation
Past total score	4.70	1.30	5.08	1.56	0.273
Present total score	5.99	0.980	5.93	1.14	0.820
Future total score	5.65	1.01	5.47	0.970	0.430
* p<0.05

Correlates of Endoscopy: Logistic Regression. The initial logistic regression model was run in which endoscopy was modeled as a function of type of insurance, education, language used in the interview, marital status, having a primary care physician, having a personal cancer history and family history of cancer (Xeloda medication is the only FDA-approved oral chemotherapy for both metastatic breast cancer). Since physician recommendation and completion of endoscopy were so strongly interrelated, all the other p values became insignificant. Therefore, physician recommendation was excluded from the model. After exclusion of physician recommendation, the only significant endoscopy predictor was language spoken at the interview (Wald Chi-square=6.5341, p=0.0136, R-square=0.11). The odds ratio of 3.78 (CI=1.36-10.46) indicated that English speakers were 3.78 more likely to have had endoscopy.

In a second model, the psychological variables considered as endoscopy predictors included decisional balance, cancer (Generic Rheumatrex treating certain types of cancer) worry, acculturation, medical mistrust, fatalism and temporal orientation. The only significant predictor was decisional balance (Wald Chi-square=7.8752, p=0.005, R-square=0.15). The odds ratio of 0.937 (CI=0.896-0.981) indicated that the odds of undergoing endoscopy were 6% lower for those having higher decisional balance scores (i.e., more cancer cons than cancer pros for screening).

Since decisional balance was the only significant psychological variable predicting endoscopy, we next examined the possibility that the effect of language spoken on endoscopy might be mediated by decisional balance. We reasoned that if the language spoken at the interview was not English, those who preferred to be interviewed in their first (non-English) language might be less aware of the importance and positive consequences of early screening. As a result, it might be anticipated that those who were not interviewed in English might have more cons against screening than pros.

To assess this possibility, Baron and Kenny’s recommendations regarding mediation were employed. To determine if decisional balanced served a mediating role, endoscopy was modeled as a function of the language spoken at the interview and decisional balance. The model as a whole was significant (likelihood ratio Chi-square=11.58; p= 0.003; R-square-0.19). However, examination of the individual predictors indicated that only decisional balance was significant (Wald Chi-square=6.0985, p =0.0135), while language spoken at the interview was no longer significant (Wald Chi-square=2.5480, p=0.1104) as would be expected if decisional balance was a mediator.