Impact of Prior Underinsurance on Cervical Cancer Screening Among Davidson County, Tennessee Women Diagnosed with Invasive Cervical Cancer, 2008-2018 | BMC Women’s Health
Data sources and population
The Human Papillomavirus Vaccine Impact Surveillance Project (HPV-IMPACT) is an active, multi-site, U.S. population-based cervical cancer surveillance project funded by the Centers for Disease Control and Prevention (CDC) . From 2019, the project expanded its activities to include retrospective surveillance of invasive cervical cancer from 2008. We used data collected by the TN HPV-IMPACT site (catchment area: county of Davidson) which uses diagnostic pathology lab reports to identify high-grade pre-cervicals. – cancerous lesions, i.e. cervical intraepithelial neoplasia grade 2 and above and adenocarcinoma in situ, collectively referred to as CIN2+ and invasive cervical cancer cases to assess the population-level effectiveness of HPV vaccination .
We identified Davidson County, TN, residents diagnosed with invasive cervical cancer between January 1, 2008 and December 31, 2018 who were ≥ 18 years old at diagnosis. The total population of women aged ≥ 18 in this catchment area was 305,982 in 2010 . All cases of invasive cervical cancer were identified through: (1) diagnoses reported to the TN Cancer Registry and (2) a comprehensive review of laboratory-reported pathology records. We excluded all women who had a diagnosis of cervical carcinoma in situ (i.e. non-invasive carcinoma), who lived outside the catchment area at the time of diagnosis, or who had a primary non-cervical tumour.
This project was considered public health surveillance (i.e., not human research) by the Institutional Review Boards of Vanderbilt University Medical Center, TN Department of Health, and CDC and received a non-research decision under 45 CFR 164.512.
Following the identification of all potential cases during the study period from the cancer registry and partner laboratories, trained study personnel systematically extracted sociodemographic and cancer diagnostic information from a detailed review of medical records using standardized case report forms. Although our identifying data sources provide sociodemographic and other cancer-related information, we extracted and validated all data elements for this study through medical record review.
Date of cancer diagnosis was determined as the earliest date that specimen obtained following surgery (including cervical biopsies/excisions, endocervical curettage, or hysterectomy) indicated invasive cervical cancer of the uterus. Cancers diagnosed only by cervical cytology were excluded. We collected the following at the time of cancer diagnosis: race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic and other/unknown), address, insurance status (private, public, none, unknown), stage of cancer (International Federation of Obstetrics and Gynecology [FIGO] staging criteria), any symptoms reported at diagnostic work-up, smoking status, and any immunocompromised conditions (full list of conditions in Supplementary File 1: Table S1). We extracted all available data on previous cervical cancer screening (including cytology and HPV molecular testing), follow-up examinations, and cervical procedures up to 5.5 years before diagnosis cervical cancer. To identify potential contributory barriers to screening, we collected documentation of any of the following over the five years prior to cancer diagnosis (indicated as ‘yes’, ‘no’ or ‘unknown’) from records medical. Underinsurance, our main exposure variable, was defined as any documented history of no insurance or insufficient coverage, including interruptions of coverage or concerns about poor insurance coverage leading to lack or a delay in receiving health services. This information was researched and collected in several ways, including: (1) notes from providers indicating a lack or gaps in insurance coverage affecting screening or health care-seeking behavior; (2) reviewing the insurance information provided in the “login” or “face sheet” information for all prior consultations available at the clinic for any documentation indicating that you are uninsured or underinsured (for example, charity; indigent; help from the Tennessee Breast and Cervical Screening Program). Other potential barriers to screening include documentation of poor English proficiency, substance abuse or use disorder, morbid obesity, or documented body mass index > 40 kg /m2history of homelessness, history of incarceration, and diagnosis of a serious mental health disorder (as defined by the Substance Abuse and Mental Health Services Administration) .
The medical record review involved a comprehensive approach to obtaining all available records for data abstraction. Our data identifying sources (i.e., Cancer Registry and Pathology Laboratories) provided information about the provider who ordered the qualifying diagnostic procedure and the associated provider facility. With this information, medical records were obtained either through electronic health record systems or through paper records at the request of facilities. Missing information following this initial examination was sought by obtaining records from gynecological oncology providers or primary care physicians/obstetrician-gynecologists to whom the women had been initially referred or regularly visited as part of their gynecological care, respectively. This multi-step approach meant that patient records were reviewed by at least two to three different providers per woman and as part of a thorough continuum of care investigation for each woman whenever possible.
Define cervical cancer screening history
To determine cervical cancer screening history, we excluded all cervical cytology and HPV tests performed within six months prior to cancer diagnosis, as we assumed that they were part of the diagnostic process and not part of routine cervical cancer screening. We used a ‘threshold’ of six months as has been used in previous studies [8, 22]. We then categorized the women into one of three mutually exclusive screening history categories based on documentation of a screening test performed according to the United States Preventive Services Task Force (USPSTF) recommended screening intervals. in 2012: six months to 3.5 years for a cytology test only and six months—5.5 years for cytology with HPV molecular test or HPV molecular test only . We assumed that meeting the 2012 guidelines would prevent progression to cervical cancer for most women. The ‘no screening’ category included women who did not have a history of screening (cytology, HPV or cytology/HPV co-testing) within the recommended screening interval immediately prior to diagnosis of cancer. Women in the “no follow-up” group had at least a ≥ 6 month or ≥ 1 year interval between abnormal cytology and/or a positive HPV test and receipt of recommended follow-up management based on grade and severity abnormal screening test, according to the guidelines of the American Society for Colposcopy and Cervical Pathology (ASCCP) . Finally, women whose most recent screening test within the recommended time window before cancer diagnosis was a normal cytology test and/or a negative HPV molecular test or women with adequate follow-up after an abnormal screening test per ASCCP guidelines were categorized as “test/failed screening”. Women with no or inconsistent information about screening tests performed within the recommended screening interval were considered to have an “undetermined” screening history and were excluded from analyzes examining patient characteristics by based on screening history. An algorithm of the screening history determination process is shown in Supplementary File 1: Figure S1.
For women diagnosed with invasive cervical cancer, we summarized demographic, social, and clinical characteristics according to the three categories of screening history and examined the association between screening history and patient characteristics. women using parametric and non-parametric tests (Pearson’s chi-square [two-sided]; Fisher’s exact; sum of Wilcoxon ranks; Kruskal-Wallis) as appropriate. Using similar tests, we examined the association between underassurance (underassurance versus no such history) and women’s sociodemographic and clinical characteristics. Univariate and multivariate logistic regression models were used to test our hypothesis that women with a history of underinsurance in the 5-year period prior to cancer diagnosis had an increased likelihood of having a history of composite “no-screen/no-follow-up” screening. versus “test/screening failure” versus women with no history of underinsurance. The covariates included in the multivariate logistic regression were determined a priori based on pre-existing knowledge of their effects on the exposure (insurance history) and outcome (screening history) variables. We adjusted for race/ethnicity, stage of cancer, smoking status, presence of one or more barriers to screening other than underinsurance, and year of cancer diagnosis. All analyzes were performed using STATA (version 16) with statistical significance determined at α = 0.05. All statistical tests were two-sided.