|Home | About | Journals | Submit | Contact Us | Français|
Delays in care after abnormal cancer screening contribute to disparities in cancer outcomes. Women with psychiatric disorders are less likely to receive cancer screening and may also have delays in diagnostic resolution after an abnormal screening test.
To determine if depression and anxiety are associated with delays in resolution after abnormal mammograms and Pap tests in a vulnerable population of urban women.
We conducted retrospective chart reviews of electronic medical records to identify women who had a diagnosis of depression or anxiety in the year prior to the abnormal mammogram or Pap test. We used time-to-event analysis to analyze the outcome of time to resolution after abnormal cancer screening, and Cox proportional hazards regression modeling to control for confounding.
Women receiving care in six Boston-area community health centers 2004–2005: 523 with abnormal mammograms, 474 with abnormal Pap tests.
Of the women with abnormal mammogram and pap tests, 19% and 16%, respectively, had co-morbid depression. There was no difference in time to diagnostic resolution between depressed and not-depressed women for those with abnormal mammograms (aHR=0.9, 95 CI 0.7,1.1) or Pap tests (aHR=0.9, 95 CI 0.7,1.3).
An active diagnosis of depression and/or anxiety in the year prior to an abnormal mammogram or Pap test was not associated with a prolonged time to diagnostic resolution. Our findings imply that documented mood disorders do not identify an additional barrier to resolution after abnormal cancer screening in a vulnerable population of women.
Depression is associated with high overall mortality, 1 including increased breast cancer mortality 2. The underlying mechanism explaining the relationship between depression and increased breast cancer mortality is unknown. A possible explanation is underuse of screening mammography, given that screening mammography reduces mortality, 3,4 and women with psychiatric disorders are less likely to receive standard preventive care,5,6 including screening mammography.7 Another possible explanation is delayed follow-up after an abnormality is detected on a screening test, since delayed diagnosis can reduce survival if cancer is diagnosed at a later stage. Vulnerable populations of women, as defined by low income or with racial/ethnic minority status, are less likely to receive standard preventive healthcare and therefore experience worse breast and cervical cancer outcomes. Previous studies showed less than 25% of these vulnerable women received adequate follow-up care after an abnormal cancer screening test.8,9 In addition, the prevalence of depression is higher in vulnerable populations,10–16 and twofold higher in women than men. National annual prevalence estimates for women range between 4 – 14%.17,18
Our previous work examined factors associated with delays in cancer care among a diverse population of women who received screening at urban community health centers (CHC).19,20 We found the individual CHC site was a stronger predictor of timely resolution than race, ethnic group, language, insurance status, or age.19,20 We then sought to examine additional patient characteristics that may identify a group of women who are particularly vulnerable to delayed resolution of abnormal cancer screening tests and might benefit from tailored interventions such as patient navigation. We hypothesized that depression would contribute to delays in diagnostic resolution. In addition, since previous studies demonstrated that anxiety may result in more timely care,21,22 we hypothesized that anxiety may decrease time to resolution, and modify an association between depression and time to diagnostic resolution. Thus, our objective is to determine if depression and anxiety are independent predictors of delays in care after abnormal mammograms and Pap tests in a vulnerable population of urban women.
This is a secondary analysis of data collected at baseline of the Boston Patient Navigation Research Program (PNRP), before a patient navigation intervention was implemented.19 Boston PNRP partnered with six CHCs to collect retrospective medical record data for women who had abnormal mammograms or Pap tests. For this study, data were collected via chart review for psychiatric diagnoses, symptoms, and treatment for the 1-year period preceding the date of the abnormal cancer screening test.
Eligible subjects in the baseline PNRP included adult women with an abnormal screening Pap test or mammogram performed between January 1, 2004 and December 30, 2005 at the 6 CHCs.19 Eligible mammograms included Breast Imaging Reporting and Data System (BIRADS) scores indicating need for follow-up (BIRADS 0, 3, 4, and 5). Eligible Pap tests included cellular abnormalities indicating need for follow-up: atypical squamous cells of undetermined significance positive for human papillomavirus (ASCUS/HPV+), low-grade squamous intraepithelial lesion (LGSIL), and high-grade squamous intraepithelial lesion (HGSIL). All subjects with high-grade abnormalities were included (BIRADS 4, 5; HGSIL), and a random sample of low-grade abnormalities (BIRADS 0, 3; ASCUS/HPV+; LGSIL) to obtain approximately 100 women per site. At sites with fewer than 100 eligible cases, all eligible subjects were included. In order to prevent clustering of cases by CHC, a representative sample of equivalent numbers were selected from each center.
We previously described the details of how the data were retrieved from the EMR and registration databases of the six CHCs.19 Socio-demographic and eligibility criteria were retrieved automatically, while clinical outcomes and psychiatric diagnosis and treatment data were abstracted manually by trained research staff.
We categorized a subject as actively “depressed” if either of the following criteria was documented in the EMR during the 12 months preceding the abnormal test: 1) a new diagnosis of depression [ICD-9 diagnoses 311.x, 309.28, 309.0-309.1, 300.4] entered into the active problem list1, or 2) depression diagnosis documented in the free-text "Assessment/Plan" section of office notes. By categorizing depression in this way, we ensured that depression was an active problem in the year prior to the abnormal cancer screening test. In order to further validate our definition of depression diagnosis, we abstracted depressive symptom data (depressed mood, sleep disorders, anxious feelings, fatigue, impaired concentration, appetite change, psychomotor change, social isolation, decreased interest, suicidality, hopelessness, other) from the EMR if documented as free-text in the “History of Present Illness” or "Review Of Systems" sections of provider notes, or listed in the Problem List during the specified time period. We categorized a subject as “anxious” if anxiety was documented either in the active problem list [ICD-9 diagnoses 300.0x, 300.2x] or in the free-text sections of office notes during the specified time period.
We obtained anti-depressant medication data from the medication list in the EMR, and evidence that the patient received psychotherapy through the following indicia: presence of an office note from a therapist or psychiatrist, mention of therapy in the free text Assessment/Plan section, or referral to behavioral health from the Orders section. We then categorized subjects into one of four mutually exclusive categories based on depression treatment: 1) anti-depressants only, 2) psychotherapy only, 3) both anti-depressants and psychotherapy, or 4) no treatment. Other psychiatric diagnoses were abstracted from the Problem Lists in the EMR, including bipolar, psychoses, and substance abuse. We collapsed the race/ethnicity EMR data into four mutually exclusive categories: White, Black/African American, Hispanic/Latina, or Asian/Other. Because the age distribution is very different for women obtaining breast and cervical cancer screening (older women in the breast cancer screening group), we categorized different age categories for the two screening populations and subdivided them into three age groups that had clinical significance. We categorized primary language as English, Spanish or other. We created the following three mutually exclusive categories from the primary and secondary insurance data in the EMR: no health insurance, publicly financed health insurance (Medicare and/or Medicaid), or private health insurance.
Our primary outcome of interest was time (number of days) from index screening abnormality to diagnostic resolution. Subjects were followed for a maximum of one year. We defined diagnostic resolution as either a definitive tissue diagnosis (biopsy with pathology report) or clinical evaluation (such as colposcopy) indicating no further need for evaluation. For subjects with LGSIL pathology, we considered the initial colposcopy with biopsy to be definitive, although surveillance is recommended. Similarly, for subjects with a recommended two-year surveillance for BIRADS three results, we considered diagnostic resolution as the next six-month mammogram, subtracting the six-month period for timely resolution in the analysis23. Due to the long times to resolution and most of the cases resolving within 6 months, we censored this outcome at a maximum of 365 days.
Women with abnormal mammograms were analyzed separately from women with abnormal Pap tests, because of marked differences between the groups in age, racial/ethnic distribution and characteristics of cancer screening test. Chi-square tests were used to compare the characteristics of depressed and not-depressed women and to compare the prevalence of depression across demographic subgroups. Next, we examined depressive symptom data in the depressed and not depressed groups to validate our definition of depression diagnosis. In the time-to-event analysis, we first compared time to resolution within 365 days between depressed and not-depressed women via the log-rank test. We then used Cox-proportional hazards modeling to predict timely resolution incorporating all of the covariates, and an interaction term for anxiety and depression. In the final models, we included only those categorical variables with a significant univariate Cox model p-value (< 0.05), excluding the interaction term because it was not significant. In secondary analyses, we modeled separate effects for depressed women with and without treatment, because treatment for depression may modify the effect of depression diagnosis. To determine if depression was associated with outcome for any subgroups of our population, we stratified our analyses by insurance, age and race/ethnicity. For women with abnormal Pap tests, the cumulative incidence curves with and without depression cross at about 120 days; therefore we modeled separate effects of depression for resolution before and after 120 days, using time-dependent indicators for depression in the Cox model. Analyses were conducted using SAS 9.1.24 A two-sided p-value <0.05 was considered statistically significant for reporting associations.
Among 997 women, 523 had abnormal mammograms and 474 had abnormal Pap tests. Overall, 17% had depression and 8.5% had anxiety. Of the women with abnormal mammogram and Pap tests, 19% and 16%, respectively, had co-morbid depression. As expected, women with abnormal mammograms were older (96%>40 years old) compared to the women with abnormal Pap tests (70% between ages 18 – 30 years old). For both groups, depressed women were more likely to be on public insurance, and for women with abnormal mammograms, depressed women were less likely to be black. (Table 1) Depressed women in both study groups were more likely to have anxiety.
Depressive symptoms were statistically significantly more common among women defined as depressed compared to those who were not. The most common documented symptoms in the depressed group in decreasing frequency were: depressed mood (57% for both women with abnormal mammograms and Pap tests), sleep problems (41% and 38%), anxious feelings (33% and 15%), fatigue (32% and 18%), impaired concentration (21% and 14%), and changes in appetite (14% for either abnormal test). For depressed women with abnormal mammograms, 69% had some form of documented treatment (66% prescribed antidepressants), while 51% of depressed women with abnormal Pap tests had some treatment (45% prescribed antidepressants).
The median time to resolution was 27 days for women with abnormal mammograms, and 85 days for women with abnormal Pap tests. Figure 1 compares the time-to-diagnostic-resolution event analysis of women with and without depression diagnosis. In both of these screening groups, we note delays in diagnostic resolution, with less than half receiving a definitive diagnosis within 30 days after abnormal mammograms, and within 90 days after abnormal Pap tests. There was no difference in time to diagnostic resolution between those with and without depression diagnosis either for women with abnormal mammograms (Fig. 1a, p=0.22 from the log-rank test) or for women with abnormal Pap tests (Fig. 1b, p=0.53).
Tables 2a and b present the bivariate and multivariable findings from the Cox proportional hazard models predicting time to resolution for mammogram and Pap test abnormalities, respectively. In these analyses, a hazard ratio less than 1.0 indicates longer time to diagnostic resolution. When examining the whole study population, neither depression diagnosis nor anxiety was significantly associated with delayed diagnostic resolution for women with abnormal mammograms or abnormal Pap tests. Secondary analyses found no significant interaction between depression and anxiety for women with abnormal mammograms (p=0.64) or with abnormal Pap tests (p=0.16). The effects of depression remained non-significant when allowing for different effects in depressed women receiving treatment (aHR 0.8 (0.6, 1.1) for mammograms and aHR 1.0 (0.7, 1.4) for Pap tests), or not receiving treatment (aHR 1.0 (0.7, 1.5) for mammograms and aHR 0.8 (0.6 , 1.2) for Pap tests) for depression. Among women with abnormal Pap tests, there was no suggestion of a depression effect on time to resolution over the first 120 days of follow-up (aHR 1.0 (0.7, 1.4)); however, there was a non-significant trend toward slower resolution for depressed women after 120 days (aHR 0.6 (0.3, 1.1), p=0.096). We performed additional secondary stratified analyses in order to examine associations within subgroups of patient socio-demographics. Separate analyses by insurance status showed no depression effect for those with public or private insurance, but demonstrated a trend of delayed diagnostic resolution for those with no insurance (HR=0.3 (0.1 , 1.0) p=0.06 for mammograms and HR=0.5 (0.2 , 1.2) p=0.11 for Pap tests). There were no differences by race, ethnicity, or age.
To our knowledge, this is the first examination of the relationship between pre-existing depression with and without anxiety and time to diagnostic resolution of abnormal mammograms and Pap tests. In this vulnerable population seeking care at urban federally qualified community health centers, we found delays in diagnostic resolution after abnormal cancer screening. However, those with a depression diagnosis did not have increased delays compared to not-depressed women. In addition, there was no significant interaction between anxiety and depression.
Our findings suggest that for women with the lengthiest delays in diagnostic resolution (> 120 days), depression may contribute to diagnostic delays of abnormal Pap tests, but not abnormal mammograms. This may reflect differences in the populations themselves, as the populations are remarkably different in age, or differences in the perceived implications of delayed diagnostic resolution. For example, a delay of several months for an abnormal mammogram may be clinically more significant than several months of delayed resolution for abnormal Pap tests. However, our results contrast from studies of cancer screening tests reporting depression being associated with less screening mammography,2,25 but not with fewer Pap tests.
Prior research found conflicting effects between the relationship of cancer screening behavior and depression. This may be due to depression frequently being confounded with anxiety, and the potential for anxiety having opposite effects on health outcomes. For example, a meta-analysis found that depressed patients were three times less likely to be adherent to treatment recommendations for a variety of illnesses, including breast cancer, but anxiety had little effect6. Patten and colleagues surveyed a large population, finding that a major depressive episode was not associated with receiving a screening mammogram in the subsequent year; however, co-existing anxiety was not accounted for in that analysis.26 Additional analytical complexities are introduced by differential reporting and documentation of depressive and anxious symptoms. Frequently, depression and anxiety symptoms are under-documented.27 Even if symptoms are documented, there remains an analytic challenge of determining if these symptoms reflect a stable psychological “trait” versus a temporary “state,” and determining which is more important in determining future health behavior and outcomes. This distinction is of particular importance when studying the outcomes of follow-up care after abnormal cancer screening, as in this study, for the negative psychological states associated with receipt of an abnormal test could adversely affect the outcome of follow-up care.
In addition to the analytical difficulties, the relationship of depression and cancer screening behavior could vary by populations of women. For example, in our study population of ethnically diverse, economically disadvantaged women, who already have other socioeconomic barriers to reaching timely resolution after abnormal cancer screening,28,29 our findings suggest that depression does not contribute to delays in diagnostic care, except perhaps for those who both are depressed and lack insurance. It is possible that depression does delay diagnostic resolution in women not socio-economically disadvantaged. Conversely, the women in our study were already engaged in the health care system, as having an abnormal cancer screening test was a mandatory inclusion criterion. If depression is a barrier to receiving screening mammography and/or Pap tests in a sub-population of women, they would not have been included in our study.
This study is limited by the relatively small sample size, limiting our power to detect statistical differences between the depressed and not-depressed groups. Using retrospective chart review to classify patients as depressed has the potential for misclassification bias in two ways, each biasing the results toward the null. First, it is possible that patients classified as “not-depressed” were truly depressed. In addition to known under-diagnosis of depression in primary care,30 which may be differential in vulnerable populations31 and those without insurance,32 there may be under-documentation of it as well, since primary care providers are frequently not reimbursed if depression is the primary ICD-9 diagnosis. The accuracy of using administrative data or the EMR to determine depression diagnosis ranges from 58 – 83% depending on the contents of the EMR.33,34 However, our 12 month depression prevalence was slightly higher than that previously reported for depression using survey tools, suggesting the possible extent to which under-diagnosis and under-reporting of depression were limited in our study.15,35,36 Second, patients adequately treated for depression may be classified as “depressed” even though they are no longer experiencing symptoms. However, since almost all of our depressed patients had corresponding documented depressive symptoms, it is unlikely that a patient classified as depressed did not have depressive symptoms during the year preceding the abnormal cancer screening test. Although depression diagnosis documented in administrative data is more likely to correspond to severe and recurrent depression in patients,37 we were not able to account for the clinical severity of depression or whether it was adequately treated. Finally, as our data was collected via retroactive chart review, there may be some residual confounding that we were not able to account for in our analyses.
This is the first examination of the relationship between pre-existing depression and time to diagnostic resolution of abnormal mammograms and Pap tests.9 Previous studies focused on screening mammography and Pap tests. Our findings did not support the notion that pre-existing mood disorders documented in the EMR delays the time to resolution of abnormal cancer screening tests for vulnerable populations, although we did find a trend of delayed care in depressed women who were uninsured or took longer to resolve their Pap abnormalities. It is possible that the timing of depressive symptoms in relation to the time of being informed of abnormal cancer screening is a more informative predictor. Future studies should validate our results by using survey data to classify patients’ depressive symptoms, their severity, and whether they reflect a stable psychological “trait” or temporary “state.”
Our current findings are consistent with our previous work, in that individual characteristics of patients may be less important than system characteristics in determining the time to diagnostic resolution of abnormal mammograms or Pap tests in vulnerable women19,20. However, identifying which patients are most vulnerable may allow for more efficient allocation of resources, as patient-centered-medical-home models emerge and incorporate more elements of case management. Our findings imply that pre-screening the EMR for mood disorders may not be the most reliable approach to identify a group of patients at higher risk of delayed diagnostic resolution of abnormal cancer screening tests in a vulnerable population.
This work was presented at the Society for General Internal Medicine National Meeting, in Miami Beach, FL, on 14 May 2009.The authors thank Ignacio De La Cruz, John Pagliaro, and Cynthia Schoettler for their support with manuscript preparation. Dr. Kronman is supported by a Cancer Control Career Development Award for Primary Care Physicians from the American Cancer Society (CCCDA-09-217-01 Kronman). This work is also supported by the National Cancer Institute (CA116892 Freund).