Predictors of timely follow-up after abnormal cancer screening among women seeking care at urban community health centers

Authors

  • Tracy A. Battaglia MD, MPH,

    Corresponding author
    1. Women's Health Unit, Section of General Internal Medicine, Department of Medicine, and Women's Health Interdisciplinary Research Center, Boston University School of Medicine, Boston, Massachusetts
    • Assistant Professor of Medicine and Epidemiology, Boston University Schools of Medicine and Public Health, Women's Health Unit, 801 Massachusetts Avenue Suite 470, Boston, MA 02118
    Search for more papers by this author
    • Fax: (617) 638-8096

  • M. Christina Santana MPH,

    1. Women's Health Unit, Section of General Internal Medicine, Department of Medicine, and Women's Health Interdisciplinary Research Center, Boston University School of Medicine, Boston, Massachusetts
    Search for more papers by this author
  • Sharon Bak MPH,

    1. Women's Health Unit, Section of General Internal Medicine, Department of Medicine, and Women's Health Interdisciplinary Research Center, Boston University School of Medicine, Boston, Massachusetts
    Search for more papers by this author
  • Manjusha Gokhale MA,

    1. Data Coordinating Center, Boston University School of Public Health, Boston, Massachusetts
    Search for more papers by this author
  • Timothy L. Lash DSc,

    1. Department of Epidemiology, Boston University School of Public Health, Boston, Massachusetts
    Search for more papers by this author
  • Arlene S. Ash PhD,

    1. Healthcare Research Unit and Section of General Internal Medicine, Boston University School of Medicine, Boston, Massachusetts
    Search for more papers by this author
  • Richard Kalish MD, MPH,

    1. Department of Family Medicine, Boston University School of Medicine and South Boston Community Health Center, Boston, Massachusetts
    Search for more papers by this author
  • Stephen Tringale MD,

    1. Codman Square Health Center, Boston, Massachusetts
    Search for more papers by this author
  • James O. Taylor MD,

    1. East Boston Neighborhood Health Center, Boston, Massachusetts
    Search for more papers by this author
  • Karen M. Freund MD, MPH

    1. Women's Health Unit, Section of General Internal Medicine, Department of Medicine, and Women's Health Interdisciplinary Research Center, Boston University School of Medicine, Boston, Massachusetts
    Search for more papers by this author

Abstract

BACKGROUND:

We sought to measure time and identify predictors of timely follow-up among a cohort of racially/ethnically diverse inner city women with breast and cervical cancer screening abnormalities.

METHODS:

Eligible women had an abnormality detected on a mammogram or Papanicolaou (Pap) test between January 2004 and December 2005 in 1 of 6 community health centers in Boston, Massachusetts. Retrospective chart review allowed us to measure time to diagnostic resolution. We used Cox proportional hazards models to develop predictive models for timely resolution (defined as definitive diagnostic services completed within 180 days from index abnormality).

RESULTS:

Among 523 women with mammography abnormalities and 474 women with Pap test abnormalities, >90% achieved diagnostic resolution within 12 months. Median time to resolution was longer for Pap test than for mammography abnormalities (85 vs 27 days). Site of care, rather than any sociodemographic characteristic of individuals, including race/ethnicity, was the only significant predictor of timely follow-up for both mammogram and Pap test abnormalities.

CONCLUSIONS:

Site-specific community-based interventions may be the most effective interventions to reduce cancer health disparities when addressing the needs of underserved populations. Cancer 2010. © 2010 American Cancer Society.

Despite increasing gains in cancer care,1 disparities in cancer outcomes are well documented for racial/ethnic minorities and those of low socioeconomic status.2 In Massachusetts, non-Hispanic black women have higher mortality from breast cancer than their white counterparts (35.5 and 23.3 per 100,000, respectively).3 The age-adjusted cervical cancer incidence rates in Massachusetts are 5.8 per 100,000 for white women, 9.2 for black women, and 13.1 for Hispanic women.4 Differences in access to care along the entire cancer care continuum, from screening through diagnostic care to treatment and survivorship, and barriers to using otherwise accessible care, may contribute to these disparities.

Parity in receipt of cancer screening has been achieved in many settings, including Massachusetts.5 The Centers for Disease Control-funded National Breast and Cervical Cancer Early Detection Program provides millions of dollars annually to ensure that those most at risk have access to breast and cervical cancer screening.6 However, the prevention potential of cancer screening requires timely diagnostic follow-up once an abnormality has been detected. Delays in diagnosis and treatment as little as 3 months have been shown to increase recurrence7 and reduce survival rates.8, 9 The belief that these delays contribute to cancer disparities is evident in the emergence of innovative programs that aim to reduce delays in receipt of cancer care services. In 2005, the National Cancer Institute's Center to Reduce Cancer Health Disparities and the American Cancer Society funded 9 programs to participate in a Cooperative Group (the Patient Navigation Research Program)10-12 to evaluate patient navigation interventions to reduce time to diagnosis and treatment for at-risk underserved populations with abnormal cancer screening or newly diagnosed cancer.

The time it takes to complete diagnostic evaluation varies widely, with the uninsured or underinsured and racial/ethnic minorities often having the longest delays.13-18 Relevant literature is limited by a lack of consistency in reported outcomes. Most studies are small, limited to a single site of care, and include diverse socioeconomic strata. Therefore, we sought to describe delays in receipt of diagnostic services for abnormal mammography and Papanicolaou (Pap) test screening among inner city women seeking care at 6 community health centers in Boston, which serves as the baseline cohort for the Boston Patient Navigation Research Program. These centers serve a high proportion of the city's racial and ethnic minority populations and those of lower socioeconomic status.

MATERIALS AND METHODS

Study Design

This study was conducted to provide baseline estimates of time to diagnostic resolution for a prospective multisite intervention study, the Boston Patient Navigation Research Program. One of 9 groups in the national Patient Navigation Research Program Cooperative Group,10-12 the Boston Patient Navigation Research Program partnered with 6 independent community health centers (CHCs) in Boston to perform the study. Concurrent baseline data were collected via retrospective medical chart review to determine time to diagnostic resolution for women with mammogram and Pap test abnormalities. The Boston University Medical Center Institutional Review Board reviewed and approved this study.

Study Population

Eligible subjects for the baseline Patient Navigation Research Program cohort included adult women with an abnormality detected by a screening Pap test or mammogram performed between January 1, 2004 and December 30, 2005 at 1 of the 6 CHCs. Women were excluded if they were pregnant or younger than 18 years of age at the time of their abnormality. Eligible mammography results included any Breast Imaging Reporting and Data System (BIRADS) score indicating need for follow-up (BIRADS 0, 3, 4, and 5). Eligible Pap test results included any cellular abnormality indicating need for follow-up. These include atypical squamous cells of undetermined significance positive for human papillomavirus (ASCUS/HPV+), low-grade squamous intraepithelial lesion (LGSIL), high-grade squamous intraepithelial lesion (HGSIL), and carcinoma. 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) were used to reach a sample of approximately 100 screened-positive women per site. At sites with fewer than 100 eligible cases, all eligible subjects were included.

Data Collection

Chart abstraction began in July 2006. If an abnormality had not reached diagnostic resolution by the time of abstraction, the patient's chart was reviewed again, if necessary, at least 365 days after the index event to ascertain resolution. The majority of the abstraction was completed using the electronic medical record; occasionally, missing clinical data needed to be abstracted from the paper chart (approximately 4%). One of 2 authors (M.C.S., S.B.) then reviewed data abstraction forms for completeness, accuracy, and internal consistency, and then entered the data into a secure, password-protected study database.

Study Variables

Variables were selected based on both 1) availability in the medical record and 2) consistency with the data dictionary developed by the Design and Analysis Committee of the National Cancer Institute Patient Navigation Research Program, to enable future comparisons with data from other Patient Navigation Research Programs.12

Independent Variables

Race/Ethnicity was documented in the electronic medical record as mutually exclusive response values: White, Black/African American, Asian, Native Hawaiian/Pacific Islander, Native American/ Alaskan Native, Hispanic Latino, or Other. With the exception of White, Black, and Hispanic, the remaining race categories were collapsed into Other because there were too few subjects in the individual racial categories to yield meaningful analyses. For individuals in >1 category, only the first of Hispanic, Black, White, or Other (in that order) was used. Age was calculated from month and year of birth to the date of the screening test. Different age categories were used for the 2 screening populations; each screening population was separately categorized into 1 of 4 clinically relevant age groups. Primary Language was categorized as English, Spanish, or Other. Primary care status was determined by the presence of a named physician appearing in the electronic medical record. Primary and secondary insurance as documented in the electronic medical record were used to create the following 3 mutually exclusive categories: No Health Insurance, Publicly Financed Health Insurance Only (Medicare and/or Medicaid as sole insurers), or Some Form of Private Health Insurance.

Outcome Variables

Our primary outcome of interest was time from index screening abnormality to diagnostic resolution. Diagnostic resolution was defined as definitive tissue diagnosis (biopsy with pathology report) or clinical evaluation (such as colposcopy) indicating no further need for evaluation, in concordance with the Patient Navigation Research Program.12 Clinical evaluation was included to account for variation in clinical practices. Because of the variability in the number of days to resolution, and the likelihood of outliers that would result in skewed data, we censored this outcome at a maximum of 180 days for outcomes analyses. Because there are clinical implications to delays as short as 90 days, the authors felt that a 180-day cutoff for timely resolution has adequate clinical significance.19, 20 Subjects were categorized as having timely resolution if their diagnostic resolution occurred within 180 days from the index abnormality. For subjects eligible because of a BIRADS 3 result, the earliest date for resolution or “time 0” began 6 months from the date of the index abnormality, because clinical practice guidelines for follow-up call for a repeat mammogram 6 months after the index abnormality.21

Data Analysis

Subjects with abnormal mammogram and Pap tests were analyzed separately. In addition to having 2 different clinical screening programs, the 2 study populations differed markedly by age, racial/ethnic distribution, and proportion with a final diagnosis of cancer. Results are presented in parallel here to provide the opportunity to see whether particular CHCs are consistently better (or worse) than others on both types of cancer screening follow-up.

Descriptive statistics were performed to report the sociodemographic characteristics of the 2 study populations, and to determine median time to resolution and the rate of resolution at different time cutoffs. We calculated P values within CHC sites using analysis of variance for continuous variables and chi-square for categorical variables.

Univariate Cox proportional hazard ratios (HRs) were generated to test the association of each subject characteristic with “timely resolution,” such that larger hazards ratios are associated with shorter time to resolution. For our multivariate analysis, we predicted timely resolution using Cox-proportional hazards modeling. In the final models, we included only those categorical variables for which the group had a significant P value (<.05) under a univariate Cox model. The CHC site with the largest study enrollment was chosen as the referent group in both cohort regression models for ease of comparison. All analyses were conducted using SAS 9.1 (SAS Institute, Cary, NC). A 2-sided P value <.05 was considered statistically significant for reporting associations. We hypothesized that systems issues were extremely important, that is, CHC site was a key explanatory variable, not a confounder. To see if nesting with CHCs had confounded the relationship between patient factors (which differed substantially across CHCs) and timely resolution, we also performed regressions treating CHC as a hierarchical clustering variable. Because these analyses did not change any of our findings, they are not reported. Thus, we did not conduct analyses within strata as defined by CHC site.

RESULTS

A total of 997 women were included in the study (523 with mammogram and 474 with Pap test abnormality). Tables 1 and 2 display subject characteristics by CHC site for mammogram and Pap test subjects, respectively. The different age distributions for each cohort reflect recommended screening guidelines for that cancer site. For both screening groups, the majority were nonwhite, with 19% to 27% Hispanic, 33% to 34% black, and 11% to 14% other. Less than ⅓ had private health insurance. In each screening group, about ⅓ spoke a language other than English as their primary language; Spanish was the most common non-English language spoken (about 15%). The majority of screening abnormalities were low grade in their suspicion for cancer, including BIRADS 0 (67%) and BIRADS 3 (25%) for mammogram and LGSIL (87%) for Pap test subjects.

Table 1. Demographic Characteristics of Subjects With Abnormal Mammograms in the Boston PNRP Baseline Cohort by Study Site
CharacteristicsCommunity Health Center Site, No. (%)Pa
AllABCDEF
  • PNRP indicates Patient Navigation Research Program; BIRADS, Breast Imaging Reporting and Data System.

  • a

    Each P value is from a chi-square test for independence between the distribution of the row categorical variable and the community health center site.

  • b

    Any race other than Hispanic, black, or white. Due to differences in race/ethnicity reporting, numbers for collapsed racial/ethnic categories are unavailable. Assignment is to 1 category only, with each category dominating the category below it.

  • c

    Any language other than English or Spanish. Includes Albanian, Arabic, French, Greek, Italian, Polish, and Portuguese.

Total52371 (14)107 (20)36 (7)106 (20)99 (19)104 (20) 
Age, y       ≤.001
 30-4023 (7)8 (11)10 (9)2 (6)7 (7)0 (0)8 (8) 
 41-50274 (52)27 (38)44 (41)15 (42)39 (37)99 (100)43 (41) 
 51-64161 (31)25 (35)40 (37)14 (39)33 (31)0 (0)40 (38) 
 65+65 (12)11 (15)13 (12)5 (14)27 (25)0 (0)13 (13) 
Race       ≤.001
 Hispanic99 (19)24 (34)16 (15)3 (8)4 (4)43 (44)9 (8) 
 Black171 (33)2 (3)25 (23)12 (33)3 (3)38 (38)91 (88) 
 White195 (37)38 (53)43 (40)10 (28)98 (92)3 (3)3 (3) 
 Otherb58 (11)7 (10)23 (22)11 (31)1 (1)15 (15)1 (1) 
Language       ≤.001
 Spanish75 (14)16 (23)11 (11)2 (6)3 (3)37 (38)6 (6) 
 English333 (64)47 (66)71 (66)25 (69)85 (80)23 (32)73 (70) 
 Otherc115 (22)8 (11)25 (23)9 (25)18 (17)30 (30)25 (24) 
Insurance       ≤.001
 No insurance221 (42)12 (17)63 (59)13 (36)44 (42)41 (42)48 (46) 
 Public188 (36)30 (42)31 (29)14 (39)52 (49)30 (30)31 (30) 
 Private114 (22)29 (41)13 (12)9 (25)10 (9)28 (28)25 (24) 
BIRADS score       ≤.001
 0352 (67)42 (59)91 (85)24 (67)77 (72)67 (68)51 (49) 
 3130 (25)16 (23)15 (14)8 (22)23 (22)21 (21)47 (45) 
 4, 541 (8)13 (18)1 (1)4 (11)6 (6)11 (11)6 (6) 
Table 2. Demographic Characteristics of Subjects With Abnormal Pap Tests in the Boston PNRP Baseline Cohort by Study Site
CharacteristicsCommunity Health Center Site, No. (%)Pa
AllABCDEF
  • Pap indicates Papanicolaou.

  • a

    Each P value is from a chi-square test for independence between the distribution of the row categorical variable and the community health center site.

  • b

    Any race other than Hispanic, black, or white. Due to differences in race/ethnicity reporting, numbers for collapsed racial/ethnic categories are unavailable. Assignment is to 1 category only, with each category dominating the category below it.

  • c

    Any language other than English or Spanish. Includes Arabic, Irish, Polish, and Portuguese.

  • d

    Includes Pap test results: atypical squamous cells of undetermined significance/human papillomavirus+ and low-grade squamous intraepithelial lesion.

  • e

    Includes Pap test results: high-grade squamous intraepithelial lesion, carcinoma.

Total47486 (18)50 (11)49 (10)92 (19)98 (21)99 (21) 
Age, y       .10
 18-2192 (19)14 (16)11 (22)11 (22)16 (17)17 (17)23 (23) 
 22-25151 (32)25 (29)21 (42)13 (27)35 (38)32 (33)25 (25) 
 26-35130 (27)27 (31)15 (30)7 (14)25 (27)28 (29)28 (28) 
 36+101 (21)20 (23)3 (6)18 (37)16 (17)21 (21)23 (23) 
Race       ≤.001
 Hispanic129 (27)40 (46)8 (16)2 (4)9 (10)60 (61)10 (10) 
 Black160 (34)4 (5)21 (42)25 (51)4 (4)30 (31)76 (77) 
 White121 (26)23 (27)12 (24)6 (12)75 (82)1 (1)4 (4) 
 Otherb64 (14)19 (22)9 (18)16 (33)4 (4)7 (7)9 (9) 
Language       ≤.001
 Spanish71 (15)31 (36)5 (10)1 (2)4 (4)24 (25)6 (6) 
 English317 (67)44 (51)34 (68)36 (73)73 (80)51 (52)79 (80) 
 Otherc86 (18)11 (13)11 (22)12 (25)15 (16)23 (23)14 (14) 
Insurance       ≤.001
 No insurance210 (44)6 (7)29 (58)22 (45)36 (39)71 (73)46 (47) 
 Public129 (27)24 (28)9 (18)18 (37)46 (50)15 (15)17 (17) 
 Private135 (28)56 (65)12 (24)9 (18)10 (11)12 (12)36 (36) 
Cervical abnormality       .43
 Low graded414 (87)71 (83)46 (92)45 (92)80 (87)83 (85)89 (90) 
 High gradee60 (13)15 (17)4 (8)4 (8)12 (13)15 (15)10 (10) 

Subjects across CHCs differed significantly on all demographic characteristics, reflecting the singular populations specific to the communities they serve. For example, the proportion of black subjects ranged from 3% to 88% across the 6 CHCs. Those sites with largely white populations, as demonstrated by Health Center D, which had 92% white subjects, also had the lowest rate of private health insurance (9%), demonstrating a socioeconomically disadvantaged group. Although the data are not shown here, we know that this CHC serves a largely immigrant, Albanian population.

During the 1-year of follow-up, 20 breast and 4 gynecological cancers were diagnosed from the abnormal screening tests. Most cancers occurred in patients whose index abnormality was high grade, including BIRADS 4 or 5 on mammography (11 breast cancers), and carcinoma on Pap test (all 4 gynecologic cancers). The remaining breast cancers occurred in women with a low-grade index mammogram result, including BIRADS 0 (8 cancers) and BIRADS 3 (1 cancer).

Time to diagnostic resolution by screening abnormality is displayed in Table 3. Overall, median time to resolution was shorter for mammogram abnormalities compared with Pap test abnormalities (median Day 27 vs Day 85, respectively). Ninety-two percent of all mammogram abnormalities achieved diagnostic resolution by 180 days, compared with only 65% of Pap test abnormalities. However, almost all abnormalities were resolved within 12 months (97% of mammogram and 93% of Pap test abnormalities). Time to resolution differed by screening abnormality; subjects with high-grade mammogram lesion (BIRADS 4,5) had the longest median time to resolution (36 days), followed by BIRADS 0 (28 days). This same pattern was not observed for high-grade Pap test abnormalities; HGSIL had the shortest time to resolution (median Day 56) compared with the lower grade Pap test lesions ASCUS/HPV+ and LGSIL (median Day 89 and Day 84, respectively). We found little variability across CHCs in the time to diagnostic resolution for a breast abnormality (data not shown). Median time to resolution ranged from 24 to 30 days across CHCs. Minimal increase in proportion of resolved abnormalities was noted beyond 6 months for abnormal mammography screening; increases in diagnostic resolution rates from 6 to 12 months across CHCs ranged from 0% to 8%. In contrast, resolution of Pap test abnormalities was more variable across CHCs, with a median time to resolution range of 59 to 181 days, with as many as an additional third of diagnostic resolution completed between 181 and 365 days after the index abnormality. Examination of differences in race showed greater variation by CHC than by racial category (data not shown).

Table 3. Uncensored Time to Resolution by Type of Screening Abnormality in the Boston PNRP Baseline Cohort
AbnormalityNo.Median Days to Resolution (Q1, Q3)% Resolved at 6 Months% Resolved at 12 Months
  1. PNRP indicates Patient Navigation Research Program; Q, quartile; BIRADS, Breast Imaging Reporting and Data System.

Mammogram 27  
 All52327 (15, 52)9297
 BIRADS 035228 (20, 56)9297
 BIRADS 313011 (2, 37)8994
 BIRADS 4, 54136 (10, 57)95100
Pap test 85  
 All47482 (45, 174 )6593
 Low grade41485 (47, 174)6593
 High grade6056 (34, 175)6592

Tables 4 and 5 present the univariate and multivariate findings from the Cox proportional hazard models predicting timely resolution for mammogram and Pap test abnormalities, respectively. Univariate analysis of subjects with abnormal mammograms found that CHC and BIRADS designation were significantly associated with timely resolution, but neither composite variable was statistically significant in the multivariate (Table 4). However, there were still intergroup differences, with CHC C (HR, 1.56; confidence interval [CI], 1.02-2.38) and CHC F (HR, 1.41; CI, 1.02-1.95) more likely to have timely resolution compared with referent group CHC A, and BIRADS 0 abnormalities less likely to have timely resolution (HR, 0.79; CI, 0.64-0.98) in comparison to BIRADS 3 abnormalities.

Table 4. Predictors of Timely Resolution of Mammography Abnormalitya in the Boston PNRP Baseline Cohort
CharacteristicUnivariate, HR (95% CI)Multivariate, HR (95% CI)Pb
  • PNRP indicates Patient Navigation Research Program; HR, hazard ratio; CI, confidence interval; ref, reference; BIRADS, Breast Imaging Reporting and Data System; CHC, community health center.

  • a

    Larger hazards ratios are associated with shorter time to resolution.

  • b

    P value is from a chi-square test for model fit between the row categorical variable and the outcome.

  • c

    Statistically significant finding.

Age, y   
 30-401.05 (0.70, 1.59)  
 41-500.78 (0.59, 1.02)  
 51-640.79 (0.58, 1.06)  
 65+ (ref)  
Race   
 Hispanic1.03 (0.28, 1.33)  
 Black1.11 (0.89, 1.38)  
 White (ref)  
 Other1.22 (0.90, 1.66)  
Language   
 Spanish0.90 (0.79, 1.17)  
 English (ref)  
 Other0.86 (0.69, 1.08)  
Insurance   
 No insurance0.88 (0.68, 1.12)  
 Public0.98 (0.80, 1.20)  
 Private (ref)  
BIRADS  .10
 00.79 (0.64, 0.98)c0.79 (0.64, 0.98)c 
 3 (ref) 
 4, 50.86 (0.60, 1.32)0.94 (0.65, 1.36) 
CHC site  .11
 A (ref) 
 B1.27 (0.92,1.74)1.34 (0.97, 1.86) 
 C1.50 (0.99,2.28)1.56 (1.02, 2.38)c 
 D1.25 (0.91, 1.72)1.28 (0.93, 1.77) 
 E1.04 (0.75, 1.43)1.05 (0.76, 1.46) 
 F1.42 (1.04, 1.96)c1.41 (1.02, 1.95)c 
Table 5. Predictors of Timely Resolution of Pap Test Abnormalitya in the Boston PNRP Baseline Cohort
CharacteristicUnivariate, HR (95% CI)Multivariate, HR (95% CI)Pb
  • PNRP indicates Patient Navigation Research Program; HR, hazard ratio; CI, confidence interval; ref, reference; CHC, community health center.

  • a

    Larger hazards ratios are associated with shorter time to resolution.

  • b

    P value is from a chi-square test for model fit between the row categorical variable and the outcome.

  • c

    Statistically significant finding.

Age, y   
 18-210.79 (0.55, 1.12)  
 22-250.88 (0.64, 1.20)  
 26-351.05 (0.77, 1.44)  
 36+ (ref)  
Race/ethnicity  .35
 Hispanic1.20 (0.90, 1.61)1.36 (0.87, 2.11) 
 Black0.62 (0.46, 0.84)c1.10 (0.74, 1.65) 
 White (ref) 
 Other0.90 (0.63, 1.31)  
Language  .24
 Spanish1.82 (1.35, 2.45)c1.35 (0.88, 2.08) 
 English (ref) 
 Other1.18 (0.88, 1.59)1.22 (0.89, 1.67) 
Insurance  .09
 No insurance0.84 (0.62, 1.12)0.71 (0.51, 0.98)c 
 Public insurance0.71 (0.54, 0.93)c0.77 (0.58, 1.04) 
 Private (ref) 
Pap test   
 Low grade  
 High grade1.2 (0.83, 1.63)  
CHC site  <.001
 A (ref) 
 B0.62 (0.41, 0.93)c0.70 (0.44, 1.11) 
 C0.37 (0.24, 0.57)c0.39 (0.24, 0.64)c 
 D0.69 (0.49, 0.96)c0.76 (0.52, 1.12)c 
 E0.55 (0.39, 0.77)c0.53 (0.35, 0.81)c 
 F0.34 (0.24, 0.50)c0.40 (0.26, 0.62)c 

Univariate analysis of subjects with abnormal Pap tests found CHC, race, insurance status, and language to be significantly associated with timely resolution (Table 5). In the multivariate model, CHC was the only composite variable predicting timely resolution (P<.001). Intergroup differences found 3 sites, CHC C (HR, 0.39; CI, 0.24-0.64), CHC E (HR, 0.53; CI, 0.35-0.81), and CHC F (HR, 0.40; CI, 0.26-0.62), significantly less likely to have timely diagnostic resolution compared with referent group CHC A. In addition, although insurance status was not significant as a composite predictor, those with no health insurance (HR, 0.71; CI, 0.51-0.98) were less likely to have timely follow-up compared with those with private insurance.

DISCUSSION

This study describes delays in diagnostic resolution after an abnormal breast or cervical cancer screening test among a representative population of primarily minority, urban women from a homogeneous socioeconomic strata most at risk for adverse cancer outcomes. The diversity in race/ethnicity across the 6 CHCs is typical of the heterogeneity of populations who receive healthcare at urban safety net institutions.22-24 After a full year of follow-up, diagnostic resolution for all cancer screening abnormalities reached over 90%; however, significant delays existed in those screening abnormalities most likely to lead to a breast cancer diagnosis. Site of care, rather than any sociodemographic characteristic of individuals including race/ethnicity, was the only significant predictor of delay in both cancer screening groups.

We found that any racial/ethnic differences in timely diagnostic resolution were explained by differences in site of care, suggesting that observed differences in timely follow-up may be primarily because of systems issues within each CHC rather than differences in the populations served. These findings are in contrast to much of the published literature, which repeatedly report minority race/ethnicity to predict delays in diagnostic care.8, 13, 15, 16, 18, 25-27 This inconsistency may be explained by the homogeneity in socioeconomic status of this cohort, as suggested by the low rates of private health insurance, even among white subjects—a factor that often confounds racial comparisons. Comparisons to this literature are thus limited by differences in study populations, such that many published studies include diverse socioeconomic strata with various methods of controlling for socioeconomic status.8, 15, 16 One study did identify location of care as an important determinant of timely follow-up16; however, another study including exclusively uninsured and underinsured women did not include such analyses.13

In our study, CHCs with more timely resolution outcomes for a cancer screening test often had delayed resolution for the other cancer screening test. This reinforces the presence of systematic issues within CHCs such that a CHC may have systems to address a particular screening disease, but lack resources for another screening disease. This difference may reflect resource constraints of CHCs and how they prioritize their population's healthcare needs. Although each of the 6 CHC sites had similar resources, such as on-site screening mammography and colposcopy services, programmatic and staffing differences surely existed yet were not measured. Observed differences may reflect systems put in place to reach patients who are at highest risk for delayed follow-up (eg, systems tailored to enhance follow-up of cervical cancer screening abnormalities). The same CHC may not be equipped to handle the systems issues for an older population of breast screening abnormalities.

An alternative explanation for differential outcomes by cancer screening site may be inherent differences in the clinical care for breast and cervical cancer screening. Women screened for breast cancer are willingly participating with forethought; the woman must come to the radiology facility, usually after having made an appointment. This may even be reinforced by the media attention paid to breast cancer screening. In contrast, cervical cancer screening may happen during another healthcare or family planning visit, and may not have been purposeful. In fact, the literature shows improved adherence with cervical cancer screening in vulnerable populations when the screening is done during urgent care visits,28 yet finds longer delays in follow-up care after screening is done in these urgent care settings.29

Finally, subjects across CHCs may differ on their perceptions of cancer risk, which could affect their timely resolution of cancer screening. Although some studies have found differences in the perception of cancer risk in the different ethnic populations,30 survey data from the same health centers failed to identify differences in cancer risk perception (T. A. Battaglia et al, unpublished data).

We found sociodemographic characteristics of study subjects to differ substantially across CHCs, reflecting the sociocultural differences among populations served at health centers within the same zip code. Even within CHCs, we found differences in sociodemographic characteristics for the 2 cancer screening populations, reflecting generational shifts in the community populations given the changing composition of modern cities and immigration patterns. These findings are particularly important given the emerging role of the community health center model in caring for the country's most vulnerable populations.22-24, 31 Our findings highlight the potential diversity both across and within CHCs, underscoring the need to understand the specific sociodemographics of the populations served.

Well over 90% of our subjects achieved diagnostic resolution, supporting the Institute of Medicine's 2002 recognition of the importance of CHCs in increasing access to care and in improving health outcomes for all patients, especially minorities.32 Our findings also support the notion that absence of racial disparities may be related to CHCs' culturally sensitive practices and community involvement—features that other primary care settings may lack—and speaks to the success of CHC models in improving health outcomes for these most vulnerable patients.23, 33, 34

It is important to note that we found the longest delays in follow-up occurred among those most likely to be diagnosed with breast cancer (BIRADS 0, 4, or 5), although this was not true for Pap test abnormalities. This may reflect inherent differences in perception of meaning for different cancer screening abnormalities, including fear of possible cancer. Alternatively, it may reflect system issues in accessing timely breast imaging. Although there is no consensus regarding how long a delay ultimately impacts outcomes, it is clinically feasible that these delays may be a mechanism for the persistent gap in cancer outcomes for vulnerable populations. As such, they speak to the need for community-based interventions targeting such at-risk groups. Patient navigation, an emerging model to address cancer health disparities, is an example of a promising community-based approach to address this gap.11, 12

This study has several limitations, principally that data were collected by retrospective chart review at a single institution for each subject. Specifically, we were limited by CHC record keeping for the years 2004 through 2006; therefore, subjects who achieved diagnostic resolution outside the system will be misclassified as unresolved. In addition, other site-level measures, such as specific funding resources for site-specific cancer programs, were not available from the medical record review. Demographic information was collected at the time of chart abstraction, whereas the abnormality had occurred earlier; thus, for example, our insurance information may not reflect the status at the time the abnormality occurred. We are unaware of any major changes to healthcare coverage in the state during this time period. Provider-level cluster analyses were not performed, as the study sites were not provider-specific systems.

Conclusions

This study found that delays in diagnostic resolution after an abnormal screening test in an urban safety net system are most strongly associated with site of care. Our data support the need for community-based interventions, such as patient navigation, which are culturally targeted, to close the gap in cancer health disparities.

Acknowledgements

We thank Sarah E. Lane for her valuable contributions in the preparation of the manuscript, and the entire Boston Patient Navigation Research Program Research Team and Community Advisory Panel, without whom this study would not have been possible.

CONFLICT OF INTEREST DISCLOSURES

This study was funded by the National Cancer Institute (U01 CA116892-01) and the Susan G. Komen Foundation (POP-05-04,003).

Ancillary