Navigating veterans with an abnormal prostate cancer screening test: a quasi-experimental study

The study took place at a Chicago VA, an equal access system in an urban medical district. The tertiary care facility and its four community-based outpatient clinics provide care to approximately 58,000 veterans in Chicago, the county (Cook) in which Chicago resides, and six counties in northwestern Indiana. A large proportion of veterans in Illinois are African American, and nearly half in Cook County are over age 65 [28].

Our protocol has been described in detail previously [29]. Briefly, the study used a quasi-experimental design. Eligible patients were identified through electronic medical records. Patients were selected to receive navigation if they met inclusion criteria and presented to the urology clinic on a clinic day designated for the navigation intervention (n = 245). All patients who presented on a clinic day designated for non-navigation and met inclusion criteria were identified as controls. Control subjects received usual care and comprised a consecutive and concurrent records-based sample with a one-to-one match goal (n = 245). The clinic was staffed with rotating resident physicians and stable supervising attending physicians and nurses who remained constant across navigated and non-navigated clinic days.

Our study enrolled patients from November 2006 through April 2010. All patients who presented to the VA urology clinic for follow-up of an abnormal prostate cancer screening test (prostate specific antigen (PSA) ≥ 4.0 ng/mL, PSA < 4.0 ng/mL with an abnormal PSA velocity defined by a physician, or abnormal DRE) and who were recommended for a biopsy were eligible. We excluded patients if they were less than age 18, institutionalized, cognitively impaired, previously navigated for an abnormal cancer finding, or treated for any cancer except non-melanoma cancer within the past five years. Our study team approached patients for participation following provider referral on the first follow-up visit in the urology clinic. We obtained written informed consent from navigated patients, and the Institutional Review Board (IRB) granted a waiver of authorization for our control sample. The Northwestern University IRB, Jesse Brown VA Medical Center, and the Collaborative IRB at the University of Illinois at Chicago approved the study.

The patient navigation team comprised two full-time male navigators: one social worker and one lay health worker [29]. Navigators participated in local training, monthly state-level ACS trainings, and national annual PNRP training conferences relative to navigator roles, health disparities, cancer knowledge, cultural diversity, and communication [30].

Upon enrollment, navigators conducted face-to-face interviews with patients to identify potential barriers to care, and follow-up assistance occurred in person or over the phone until diagnostic resolution for those without prostate cancer, primary therapy completion for those with cancer, or until the end of the study. The number of contacts varied according to need, but at minimum, each patient received appointment reminder calls one month and ten days prior to his scheduled biopsy appointment to ensure compliance with preparatory instructions [29]. Additional navigation activities included social support, facilitation of patient-provider communication, transportation coordination, and patient education. Navigators also often accompanied patients with prostate cancer to treatment appointments and connected each with a cancer survivor through an established ACS program.

We collected self-reported demographic data at baseline in the navigated arm and obtained them through chart audit in the control arm. Navigators documented each encounter, barrier, and action taken on standardized tracking logs developed based on the needs assessment of our targeted population. Our needs assessment involved interviews with key hospital staff and patients who delayed abnormal screen follow-up and focused on identification of personal and system barriers to care [29]. Using the VA's comprehensive computerized patient record system, trained research assistants performed chart audits for clinical outcomes data following national PNRP guidelines [13]. Teams of researchers (N.N., P.B., T.L., and E.R.) independently abstracted data on a subset of the entire sample and iteratively compared results until they achieved consistent agreement; clinical co-investigators (M.S. and J.M.) adjudicated discrepancies. The project and data managers thoroughly reviewed data completeness and cleaned entry errors on a quarterly basis.

The primary outcome for the prostate screening group was time from abnormal screening test to diagnostic resolution (Time 1). For the prostate cancer diagnosis group, the primary outcome was time from diagnostic resolution to treatment initiation (Time 2). These events had to occur before 365 days from abnormal screen for Time 1 or 365 days from diagnosis for Time 2 to be considered events in the statistical analysis. Otherwise, they were censored at 365 days. Abnormal screening was defined as the date of the abnormal prostate screening test (PSA or DRE) prompting referral for follow-up, and definitive diagnosis was defined as the date of the diagnostic test (transrectal ultrasound (TRUS)-guided biopsy or follow-up PSA test) resolving the screening test into a cancer diagnosis or non-cancer diagnosis (negative biopsy on pathology or normal PSA level/velocity for those who resolved with a follow-up PSA test). Treatment initiation was defined as the first date of treatment receipt with single or combination therapy (surgery, radiation, hormonal therapy, cryotherapy, or decision to watch and wait).

Based on preliminary data and our group’s clinical experience with veterans with prostate cancer, we estimated that the mean time from abnormal screen to diagnostic resolution (Time 1) in the control group would be 103 days, and the mean time from cancer diagnosis to treatment initiation (Time 2) in cancer patients would be 128 days. The initial power calculation indicated that for Time 1, 300 patients per group would have 80% power to detect a mean Time 1 of 86 days in the navigated group, and that 120 cancer patients per group would have 80% power to detect a mean Time 2 of 90 days in the navigated group. Since only 245 instead of 300 patients were enrolled per group for Time 1, the actual effect size for Time 1 that was detectable with 80% power was 103 days versus 84 days. Since only 85 instead of 120 patients were enrolled per group for Time 2, the actual effect size for Time 2 that was detectable with 80% power was 128 days versus 82 days. Two-tailed tests and a Type I error rate of 5% were assumed.

All statistical analyses were conducted using SAS version 9.2 (SAS Institute, Inc., Cary, North Carolina) [31]. For categorical variables, frequencies and percentages were calculated by group and for the whole sample. Comparisons between groups were made using Fisher’s exact test. Unadjusted and adjusted time-to-event analyses were done for Time 1 and Time 2. For unadjusted analyses, Kaplan-Meier curves were calculated, median times (25th percentile for one analysis since the median was not attained) within groups were determined from these curves, and curves were compared between groups using the log-rank test. For adjusted analyses, proportional hazards regression was used. The proportional hazards assumption was assessed by inclusion of a group by log-time interaction term. For Time 1, age (< 65, ≥ 65), and race (African American, non-African American) were used as covariates. Baseline absolute PSA value was not included in the model, as nearly one-fifth of patients were eligible based on an abnormal DRE or PSA velocity. For Time 2, age (< 65, ≥ 65), race (African American, non-African American), and total Gleason score (≤ 6, > 6) were used as covariates. Adjusted hazard ratios (HR) and 95% confidence intervals are reported, together with a p-value testing whether the HR was 1.0. A HR greater than 1 indicated that the outcome occurred earlier in the navigated group compared with the control group. A HR less than 1 indicated that the outcome occurred later in the navigated group compared with the control group. Kaplan-Meier curves showed a crossover time point, an indication that the proportional hazards assumption might be violated; thus, time to event was analyzed separately before the cross-over period in each analysis. A significant group by log-time interaction (p = 0.0009) indicated the proportional hazards assumption did not hold for Time 1. For Time 1, navigated and control groups were then compared (a) using all participants in the analysis and censoring everyone with a follow-up time greater than 80 days and (b) using only those participants with follow-up greater than 80 days. The proportional hazards assumption did hold for Time 2. For Time 2, navigated and control groups were compared within subgroups determined by: (a) Gleason score (≤ 6 vs. > 6); (b) treatment type; and (c) receipt of consultation beyond a urologist, given that multiple specialist visits could delay treatment initiation.

Figure 1 presents a flow chart of patient progression through the study. Of the 490 study participants, 245 were enrolled in each of the control and patient navigation intervention groups. For both Time 1 and Time 2, reasons for censoring are tabulated for each study group. Of the 490 patients, 414 reached diagnostic resolution and are considered events in the time-to-event analysis of Time 1. Of the 170 patients diagnosed with cancer, 154 initiated treatment and are considered events in the analysis of Time 2.

Baseline participant characteristics are reported in Table 1. Almost all patients (92%) were eligible based on an abnormal PSA test, 47% were at least 65 years old, two-thirds were African American, and 86% had a low Charlson Comorbidity Index [32] score (0-2). There were significant differences between control and navigated participants in age (p < 0.05), race/ethnicity (p < 0.0001), and baseline PSA (p = 0.008). Among the 170 patients diagnosed with cancer, half were at least 65 years old, 69% were African American, and 57% had a Gleason score greater than 3 + 3 (Table 2). A total of 213 patients refused to participate in navigation, many citing lack of time as the reason.

Among our total sample, 85% reached a diagnostic resolution (50% with a negative resolution based on a negative biopsy or repeat PSA test; 35% with a positive diagnosis) (Table 1). Seventy-six (16%) were unresolved (Figure 1). Resolution rates were comparable between the study groups.

With respect to Time 1, navigated patients had a diagnostic test completion rate of 211 events among 245 patients versus 203 events among 245 control patients. Navigation did not significantly affect time to diagnostic resolution compared to controls (median days of 97 versus 111; adj. HR 1.17, 95% CI, 0.96 – 1.43, p = 0.12). Kaplan-Meier (Figure 2) curves show a crossover time point; thus, time to event was analyzed separately before 80 days (using the total sample size and censoring follow-up times greater than 80 days) and after the 80 days (using those with follow-up greater than 80 days). Prior to 80 days, the 25th percentile of the time to diagnostic resolution was not significantly different between the navigated and control groups (25th percentile of 68 days versus 55 days, p = 0.45, Table 3). The adjusted HR was 0.91 (p = 0.53). Beyond 80 days, median time to diagnostic resolution was significantly shorter among the navigated group compared to the control group (median of 151 days versus 190 days, p = 0.015). The adjusted HR was 1.41 (p = 0.01).

In the analysis of Time 1 by race (African American, White), outcomes were similar to overall results. Prior to 80 days, the 25th percentile of the time to diagnostic resolution was similar in the navigated and control groups within each racial group (data not shown). Beyond 80 days, the median time to diagnostic resolution was shorter among the navigated group within each racial group, and significant among White participants (median of 122 days versus 179 days, p = 0.047).

Among patients diagnosed with cancer, 16 (9%) did not initiate treatment after diagnosis (Table 2 and Figure 1). With respect to Time 2, navigated patients had a treatment initiation rate of 75 events among 84 patients versus 79 events among 86 control patients (Table 3). Median time to treatment initiation was 93 days in the navigated group and 87 days in the control group, p = 0.36. The adjusted HR was 1.15 (p = 0.41). Kaplan-Meier curves are illustrated in Figure 3.

In the analysis of Time 2 by race (African American, White), results were similar to the overall results: no navigated-control differences in time to treatment initiation were observed within either race group. In addition, we found no difference in navigated versus control group median days to treatment initiation with respect to Gleason score or consultation beyond a urologist (data not shown). With respect to treatment types, no difference was observed except among men whose primary therapy was radiation, where time to treatment initiation was significantly lower for navigated versus control group participants (median of 95 days versus 120 days, p = 0.05).