COLORECTAL CANCER (CRC) is the third most common cancer in the United States, and a significant cause of cancer morbidity and mortality (Jemal et al., 2008; Ries et al., 2000). Screening is a cost-effective strategy for reducing cancer mortality through early detection of CRC and incidence through removal of adenomatous polyps (Levin et al., 2008). Despite the ability of colonoscopy and other screening modalities to decrease the risk of CRC, approximately 45% of the eligible American population remains unscreened (Klabunde et al., 2011; Mitka, 2008).

Depression is associated with lower rates of mammography (Carney & Jones, 2006; Egede et al., 2010), pap smears (Leiferman & Pheley, 2006; Lord et al., 2010), blood pressure monitoring, vaccinations, and cholesterol testing (Lord et al., 2010), suggesting that depression may also be a barrier to CRC screening. Few studies have evaluated the relationship between depression and CRC screening participation, yielding conflicting results. Identification of whether or not depression is a "risk factor" for nonparticipation in CRC screening may influence policymakers seeking targeted interventions to assist the needs of this group and help prioritize spending of resources in the most effective way possible.

The purpose of this study was to use a large database representative of the US population to determine the impact of depression on CRC screening participation, specifically colonoscopy. We hypothesized that individuals with depression have lower rates of up-to-date screening colonoscopy than individuals without depression.

MATERIALS AND METHODS

Data source

We used a cross-sectional sample of the Consolidated Household and Medical Conditions data files from the Household Component (HC) Survey of the 2009 Medical Expenditures Panel Survey (MEPS), a nationally representative survey of the US civilian, noninstitutionalized population. A detailed description of the MEPS-HC and additional information regarding the survey methodology and sampling are available elsewhere at http://www.meps.ahrq.gov/mepsweb/.

Study sample

Our sample included survey respondents aged 50 to 75 years without a history of colorectal cancer or inflammatory bowel disease who had complete outcome and covariate data (Figure).

Figure. Study sample diagram. CRC indicates colorectal cancer.

Measures

Outcome

The primary outcome was up-to-date CRC screening with colonoscopy (yes/no) as defined by responses to the question "When did you do your most recent colonoscopy?" Individuals who answered "within the past year," "within the past 2 years," "within the past 3 years," "within the past 5 years," or "within the past 10 years" were considered up-to-date. Those who answered "more than 10 years" or "never" were not considered up-to-date (Levin et al., 2008).

Main independent variable

Depression status, the main independent variable was determined by participant responses (yes/no) to the prompt "Now we are going to focus on health problems that have actually bothered you. Health problems include [...] mental or emotional health conditions, such as feeling sad, blue, or anxious about something. Include all conditions, accidents, or injuries regardless of whether you saw a medical provider, received treatment, or took medications."

Responses were then manually coded using the ICD9-CM (International Classification of Diseases, Ninth Revision, Clinical Modification) system and entered into the MEPS-HC Medical Conditions File. In our analyses, individuals with an ICD9-CM code of 311 were classified as having a self-reported history of depression.

Covariates

Covariates were selected because of their potential to confound results based on their association with CRC screening participation (Anderson, et al., 2011; Borum, 2009; Carcaise-Edinboro & Bradley, 2008; Ferrante et al., 2006; Guessous et al., 2010; Heo et al., 2004; Hughes et al., 2010; Jasek, 2011; Klabunde et al., 2009, 2011; Littman et al., 2011; Rich et al., 2011; Rosen & Schneider, 2004; Soni, 2007; Stimpson et al., 2012; Vaidya et al., 2012; Walker et al., 2010; Wilkinson & Culpepper, 2011). We included gender, marital status, race, education, born in the US, language spoken at home, US census region, age, health insurance status, income level, body mass index, diabetes, other comorbidities (hypertension, coronary heart disease, myocardial infarction, stroke, emphysema, chronic bronchitis, asthma, cancer, and arthritis), number of other comorbidities, presence of a usual care provider, and number of office-based provider visits per year.

Statistical analysis

Analyses were conducted in 2012 using SAS version 9.3 (SAS Institute Inc, Cary, North Carolina). We used proc survey commands to account for the complex survey design and maintain the clustering structure of the full sample when taking subsets of data for analysis. We used survey weights so that our results reflected the 2009 US civilian, noninstitutionalized population. P < 0.05 was considered significant.

We performed descriptive statistics on the study variables for the overall sample and by depression status. Frequencies and weighted percentages are reported for each of the categorical variables. Means and standard errors are reported for continuous variables. Sample characteristics were compared by depression status (yes/no) and between those missing and not missing outcome data, using Rao-Scott chi-square tests for the categorical variables and weighted regression models for the continuous variables. The prevalence of up-to-date screening colonoscopy was obtained for the overall sample and by depression status. Rao-Scott chi-square tests were used to test whether the prevalence of up-to-date screening colonoscopy differed by depression status.

Logistic regression models estimated the unadjusted and adjusted effect of depression and the study covariates on up-to-date screening colonoscopy. Because the covariates race, born in the US, and language spoken in the home were highly collinear, the final multivariable model included only race. Similarly, other comorbidities and number of other comorbidities were strongly correlated. Results were comparable when both variables were used, thus the final multivariable model only included other comorbidities (yes/no). Interactions between depression and each of the covariates were also examined. Odds ratios (ORs) and their 95% confidence intervals (CIs) were computed from these logistic regression models.

Sensitivity analyses were conducted to assess the robustness of the study findings. First, we expanded the outcome to include being up-to-date with CRC screening via any modality (fecal occult blood test [FOBT] within last 1 year, flexible sigmoidoscopy within last 5 years, or colonoscopy within last 10 years) and ever having participated in any modality of CRC screening. Second, we examined the relationship of CRC screening participation with other mental health disorders, specifically anxiety and any mental health diagnosis. Third, we included the short form (SF)-1, a global indicator of health status in our models. Finally, we examined the potential impact of missing data by conducting additional analyses assuming 2 extreme situations for cases missing outcomes (ie, all had screening and none had screening).

Study approval

This study was approved by the Boston University institutional review board.

RESULTS

The 2009 MEPS study included 36 855 respondents of which 6787 met the inclusion criteria (Figure). Among our sample, 836 (12%) had depression and 5951 (88%) did not. Table 1 presents descriptive data for the overall study sample and by depression status. Sixty percent of the sample reported having an up-to-date screening colonoscopy, with a slightly higher percentage of depressed individuals than nondepressed individuals meeting criteria for up-to-date screening colonoscopy (64% vs 58%, respectively, P = 0.01). The vast majority (89.8%) had a usual care provider. Those with depression differed from those without depression on all baseline characteristics except education level and US census region.

Table 1-a. Sociodemographic and Baseline Medical/Utilization Characteristics of the Study Sample in the 2009 Medical Expenditures Panel Survey by Depression Status

Table 2 presents findings from the regression models. The unadjusted odds ratio of having a current colonoscopy was 1.3 times higher for those individuals with depression compared with those without depression (OR = 1.3; 95% CI = 1.1-1.6). After controlling for the study covariates, those individuals with depression were still more likely to have an up-to-date screening colonoscopy than those individuals without depression (OR = 1.3; 95% CI = 1.1-1.7). All study covariates were associated with up-to-date colonoscopy, except gender and body mass index; diabetes was significantly related to up-to-date screening colonoscopy in the unadjusted model but insignificant in the adjusted model. None of the interactions between the covariates and depression were statistically significant (data not shown).

Table 1-b. Sociodemographic and Baseline Medical/Utilization Characteristics of the Study Sample in the 2009 Medical Expenditures Panel Survey by Depression Status

Table 2-a. Unadjusted and Adjusted Associations Between Depression and Up-to-date Colonoscopy (N = 6787) in the 2009 Medical Expenditures Panel Survey

Sensitivity analysis findings

Expanding the definition of CRC screening participation and including the SF-1 in our adjusted model did not alter the findings. Furthermore, the results from the analysis examining the association of anxiety and any mental health disorder with up-to-date screening colonoscopy were comparable to our main findings adjusted odds ratio (aOR)_anxiety = 1.09, 95% CI = 0.88-1.34); aOR_anyMH = 1.27, 95% CI = 1.09-1.49. Finally, whether or not cases with missing outcome data were assumed to be all up-to-date or all not up-to-date with screening colonoscopy did not change the original study results (Supplementary Tables A-C, available at http://links.lww.com/JACM/A26, http://links.lww.com/JACM/A27, and http://links.lww.com/JACM/A28, respectively).

DISCUSSION

In this study, we found that depression does not adversely affect participation in CRC screening, and it may be associated with marginally higher rates of screening. We theorize 2 explanations for this unexpected association. First, people with depression may be more engaged in medical care. While our model controlled for having a usual care provider and the number of annual visits, we could not capture aspects of care beyond utilization such as trust in the medical system, preventive health care orientation, and provider belief in the value of preventive health. More frequent medical visits also provide more opportunities for providers to discuss and promote CRC screening, and primary care provider recommendations are an important factor in patients attending a screening colonoscopy (Carcaise-Edinboro & Bradley, 2008; Guessous et al., 2010; Jasek, 2011; Klabunde et al., 2011; Rich et al., 2011).

Second, more than half of the patients presenting to a primary care provider have somatization as part of depression (Goldberg, 1979). Patients who somatize may have substantial physical symptoms, which may lead to increased utilization of health services, including general physical examinations (Steiner et al., 1998). Specific symptoms, such as anorexia, weight loss, and chronic pain may predispose patients to being more receptive to CRC screening, which indirectly alleviates patients' concern over their general well-being while simultaneously fulfilling their providers' responsibilities toward promoting health maintenance.

Our findings are consistent with 3 previously published studies. In a retrospective single-center study, Stecker et al. (2007) found that women older than 50 years with depression were more likely to receive colonoscopy within the past 5 years compared with women with hypertension. A retrospective study at a single Veterans Administration (VA) center found that female veterans with mental health diagnoses were as likely as those without a mental health diagnosis to receive any CRC screening and were more likely to be up-to-date on CRC screening, though neither were statistically significant likely due to underpowered results (Yee et al., 2011). A separate survey of elderly persons in North Carolina found that depression was not associated with receipt of FOBT (Heflin et al., 2002).

In contrast, 2 studies at the VA have shown a negative association between depression and CRC screening participation. In a national study, Druss et al. (2002) found rates of CRC screening with FOBT, flexible sigmoidoscopy, or colonoscopy within the previous 5 years were marginally lower among those with psychiatric conditions compared with those without any psychiatric comorbidities. The weaknesses of this study include lack of adjustment or stratification by number of visits and use of a 5-year window for the outcome measure, despite guidelines recommending yearly FOBT.

A study at the VA in Minnesota found that the relationship between mental health diagnoses and up-to-date CRC screening by FOBT, flexible sigmoidoscopy, or colonoscopy was greatly influenced by the number of outpatient visits (Kodl et al., 2010). In unadjusted models, those with any mental health disorder appeared more likely to be screened (57% vs 47%, P <0.01), with a similar but nonsignificant trend for depression. After accounting for usual care visits, however, the directionality of this association changed, supporting the inclusion of this variable in our model. This study was limited in generalizability because of a comorbid VA population and related selection effects. Those with mental disorders had far greater medical visits compared to those without (16.8 vs 5.9) and high burden of concurrent mental health diagnoses (on average 5.8 per patient).

Differences between the MEPS sample and the VA cohorts may help explain the discrepancy between our findings and those within the VA. Medical Expenditures Panel Survey is a sample of individuals designed to be nationally representative of the US population. On average, the VA population tends to be sicker with an increased number of comorbidities and concurrent mental illness, including a large proportion of substance abuse and posttraumatic stress disorder. These particular diagnoses may influence screening behavior differently than depression does. Perhaps those with depression in MEPS are less severe and more functional than those with depression at the VA, where issues of self-efficacy, navigating logistics, impaired psychomotor ability, and lack of family support become important barriers to screening. Addressing severity of depression and its influence on CRC screening is an opportunity for future research.

Our study has several strengths. First, we used a large sample of US households within the rigorous study design of MEPS and to our knowledge, this is the first study conducted on a nationally representative sample. Second, we used self-reported mental health disorders to capture patients who may not have sought health care for their symptoms or perhaps were underdiagnosed or misdiagnosed by their providers. Third, we evaluated 3 different CRC screening outcomes with the primary being colonoscopy to reflect temporal trends in clinical practice and account for all acceptable CRC screening modalities by current guidelines. Fourth, we included having a usual care provider and number of provider visits as a proxy for engagement in care, which was previously shown to be relevant to models describing CRC screening participation (Kodl et al., 2010). Finally, we confirmed findings in the literature with regard to characteristics influencing CRC screening and sociodemographic factors associated with depression, lending further validity to our findings.

Despite these strengths, we acknowledge certain limitations. First, the cross-sectional design allows evaluation of associations but does not provide direct data on causality or any plausible pathway of explanation. Second, the temporal relationship between depression and utilization of CRC screening remains unknown, but again, we only attempt to show an association, not causality. Third, we rely on self-reported depression and CRC screening in lieu of documentation within medical charts or billing codes. However, studies show that self-reported mental health is the third most accurately reported condition in MEPS, with a sensitivity rate of 88% (Machlin et al., 2009; Mawani & Gilmour, 2010). Self-reported colonoscopy is also dependable, with sensitivity and specificity rates of 97% and 72%, respectively (Khoja et al., 2007; Partin et al., 2008). Fourth, we do not know disease severity, utilization of antidepressants or other medications, or functional impairment as a result of the mental health conditions, all of which would be useful to evaluate in future studies. Finally, those missing data or lost to follow-up were slightly different than the sample with complete data. To address this, we performed sensitivity analysis in which our results remained relatively unchanged and robust.

CONCLUSION

Our study shows that depression does not negatively impact CRC screening rates. These results have implications for clinical practice and research. In the general US population, depression, while important to consider in overall management, may not be a risk factor for underutilization of CRC screening and equal effort should be expended to encourage CRC participation regardless of depression status. Future research should consider the age of onset, duration, and severity of depression, as well as the role of somatization with regard to CRC screening to identify subgroups that may be particularly vulnerable.

REFERENCES