Follow-Up and Timeliness After an Abnormal Cancer Screening Among Underserved, Urban Women in a Patient Navigation Program

Study design

The Chicago Cancer Navigation Project (CCNP) was conducted from October 2005 through March 2010 and is one of the 10 NCI-supported PNRP projects developed in 9 sites across the country (10). The CCNP used a nonrandomized, controlled design and targeted low-income, minority women seeking medical care in the safety net system. Overall, navigated and control subjects originated from 20 different primary care delivery sites. Of these, 19 were part of a single federally qualified health center (FQHC) network and 1 was a hospital-based ambulatory care center.

Similar numbers of navigated and control patients were recruited from the single, hospital-based ambulatory site using individual nonrandom assignment. Because the hospital is a major tertiary referral center in the region, all control subjects and 80% of navigated subjects entered the study with a cancer diagnosis.

Within the FQHC network, 5 navigation sites were chosen because they served a majority African American or Latina population and had logistic ease of administering the navigation intervention. Patients seeking care within the FQHC network, who became navigated subjects, were patients of a clinic that had an on-site navigator. The 14 medical record-based control sites were selected because they had similar characteristics to the navigated sites according to patient volume, racial/ethnic composition, and payor mix. Control subjects remained controls even if they changed clinics to receive follow-up care at a navigation site.

Study participants

Adult women whose breast or cervical cancer screening test was abnormal were eligible to participate in this study. A breast cancer screening test was defined as “abnormal” if there was (i) confirmed breast mass or other abnormality suspicious for cancer as a result of clinical breast exam (CBE) or (ii) suspicious or incomplete mammogram, ultrasound, or magnetic resonance imaging results [American College of Radiology Breast Imaging and Reporting Data Systems (BI-RADS)]. A cervical abnormality was defined as a (i) visible or suspicious lesion on cervix or (ii) either low- or high-grade abnormal Pap test. The index screening served as the point of identification and enrollment of women into the study. All eligible patients were approached and clinics refusals varied from 3% to 5% (N = 1,048).

Women were ineligible if they were younger than 18 years, being treated (or had been treated) for another cancer, or pregnant. Because the outcome of interest in this study is diagnostic resolution, patients who entered the study with a breast cancer diagnosis (n = 81) or a cervical cancer diagnosis (n = 66) were omitted. Those with concurrent breast and cervical screening abnormalities (n = 4) also were excluded. A total of 897 women (breast = 352 and cervical = 545) participated in this study. (Fig. 1)

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Study sample of nonrandomized, controlled trial of patient navigation among 897 women with an abnormal breast or cervical cancer screening test (2005–2010).

Patient navigation intervention

In line with PNRP's definition of navigation and navigator role (10), CCNP was guided by the principles of case management (15). CCNP had 4 major components including identifying and recruiting patients, identifying individual barriers to receiving care, developing and implementing an individualized plan to address the barriers, and tracking patients through problem resolution. CCNP replaced and extended the regular standard of care protocol at the clinics. The standard of care protocol included up to 2 phone calls from the clinic asking women to return for test results and a third and final attempt via certified letter when an abnormal result was recorded.

The CCNP team consisted of two master's-prepared licensed clinical social workers and 2 lay patient navigators. The social worker navigators were part of a larger American Cancer Society PN program and had access to statewide network of resources. The lay navigators were high school graduates and were racially/ethically concordant with the patients. One of the lay navigators was bilingual in Spanish and English. Navigators participated in annual national training sessions with a standardized curriculum developed by experts from the national PNRP program (16).

Navigators were assigned to specific clinics based on patient population characteristics, patients' needs, and workload. The PN team collectively sought to identify and resolve barriers to receiving timely follow-up cancer care. The social worker navigators assumed the role of team leaders and provided support for lay navigators in dealing with patients with complex barriers. For each patient, the intensity of resources required for successful navigation was tailored based on identified needs. In addition to resolving barriers to receiving timely follow-up cancer care, navigation services included helping enroll patients in public assistance programs and linking patients to community resources. Women in the control group received the regular standard of care.

Data collection

Navigators recruited eligible women for the navigation intervention by phone and in person generally within one week of the documentation of the abnormal screening test results. Written consent was obtained for all navigated subjects. Consent from control subjects was obtained through the primary care delivery sites' standard consent process. Clinical data were collected from medical charts. The study protocols were approved by Institutional Review Board of The University of Illinois at Chicago (Chicago, IL).

Study variables

The main outcome measures were having reached diagnostic resolution by 60 days and time to diagnostic resolution. Diagnostic resolution was achieved when a patient received all tests necessary to make a final determination of cancer or no cancer. Information on receipt of diagnostic resolution was dichotomized as to whether or not diagnostic resolution was achieved by 60 days (yes/no). The time period of 60 days corresponds to the National Breast and Cervical Cancer Early Detection program (NBCCEDP) guidelines for timely follow-up for women with abnormal screening results (17). Time to diagnostic resolution was calculated as the number of days between the initial abnormal screening test and confirmation of a definitive cancer diagnosis in the medical chart (malignant or benign). Patients who did not reach diagnostic resolution within the predefined study period of 365 days were censored at 365 days. To ensure similar treatment of navigated and controls, when navigator notes indicated that they were unable to reach a patient by phone or mail, time to diagnostic resolution was censored at the end of the study period (365 days).

The PNRP design and analysis committee required all sites to use standard patient covariates in all analyses to allow for comparisons of findings across sites. Covariates included race/ethnicity (black, white, Hispanic, other), age at initial abnormal screening, marital status, insurance status (uninsured, public, private), and language (English, Spanish, other). In our study, language was dropped from the analysis because of multicollinearity with race. In addition, we controlled for relevant clinical information; these include eligibility due to abnormal CBE (yes/no), BI-RADS score (0, 3, 4/5/other) for breast cancer subjects and eligibility due to abnormal Pap smear finding (low- or high-grade) for cervical cancer subjects (18–21). Low-grade signified atypical cells of undetermined significance (ASCUS) with positive high-risk serotype, atypical glandular cells, low-grade squamous intraepithelial lesion (LGSIL), and ASCUS with no reflex human papillomavirus (HPV) conducted. High-grade signified high-grade squamous intraepithelial lesion (HGSIL) or microinvasion.

Statistical analysis

Descriptive statistics were used to provide overall characteristics of the study population. χ² tests for categorical variables and t tests for continuous variables were conducted to assess differences between the study groups. All reported P values are 2-sided, and statistical significance was defined as α = 0.05.

A mixed-effects logistic model was used to estimate the odds of receiving diagnostic resolution by 60 days in the navigated group compared with the control group, controlling for demographic and clinical covariates. Separate analyses were carried out for breast and cervical cancer groups. Site was treated as a random effect and given its own intercept in the model. Survival analysis was used to compare time from initial abnormal screening test to date of diagnostic resolution for the navigated and control groups. Kaplan–Meier curves were developed to plot survival time for success in receiving a diagnostic resolution while addressing censoring of unresolved cases. The Cox proportional hazards model with shared frailty was used to obtain hazard ratios (HRs) with 95% confidence interval (CI). Models controlled for demographic and clinical covariates. Gamma-shared frailty was used to model the within-sites correlation. Higher HRs signify a shorter time interval between initial abnormal screening and receipt of a definitive diagnosis in the navigation group relative to the controls.

By testing the interaction of navigation with time in the breast and cervical models, the proportional hazards assumption was assessed and met for breast cancer and violated for cervical cancer, as the coefficient of the interaction of navigation with time was significant in the cervical model. Stratification would have been an option for cervical cancer analysis. However, stratified Cox regression is primarily suitable for nuisance variables because it is not possible to obtain estimation and inference for the effects of the stratifying variable. Therefore, for the cervical cancer analysis, because the survival curves for the navigated and controls crossed at approximately 30 days, we included an interaction term between study type (navigation vs. control) and time allowing differential effects of PN over days 0 to 30 and days 31 to 365.

Logistic regression analyses and survival analyses were repeated with and without Whites and Others, who comprised 2% of subjects. The inclusion or exclusion of Whites and Others did not meaningfully change the results and therefore these subjects were dropped from the analyses. All statistical analyses were conducted using STATA version 11.2 (StataCorp).