Cervical Cancer Screening g among g Medicaid Managed Care Members: The Roles of Substance Use Disorders, Mental Illness and y Status Disability Sharada Weir, Weir DPhil Coverage and Care for Vulnerable Populations Under Medicaid and CHIP June 27, 2010 Study Coauthors and Collaborators - Center for Health Policy and Research, UMMS: - - Sharada Weir, DPhil Heather Posner, Posner MSPH Jianying Zhang, MPH Whitney Jones, PhD Candidate Georgianna Willis, Willis PhD Jeffrey Baxter, MD Robin Clark, PhD CHPR Office of Clinical Affairs MassHealth Office of Acute and Ambulatory Care MassHealth Behavioral Health Program MassHealth Information Analysis Importance of Cervical Cancer Screening g - Cervical cancer is the second leading cause of death from cancer (after breast cancer) in women ages 20-39. - Cervical C cancer is treatable, but evidence ffrom the SEER database showed that five year survival drops from 92% if diagnosed early to 17% if diagnosed late. - As early stage cervical cancers are typically asymptomatic, Papanicolaou (Pap) smear tests are key to early detection. - According to the American Cancer Society, between 60 80% of women with newly diagnosed advanced 60-80% cervical cancer did not have a Pap test in the past 5 years. Ratee (% Use of Pap smears among US women 18 years of age and over, by selected years 90.0 88.0 86.0 84.0 82.0 80.0 78.0 76.0 740 74.0 72.0 70.0 ≥18 years 18-24 yyears 25-44 years 45-54 years 55-64 years 2000 2003 Year (NHIS) 2005 Medicaid Population - Medicaid women eligible for the cervical cancer screening measure are insured but vulnerable - 26%: Disability basis for Medicaid eligibility - Different types of disability associated with various barriers to care - 74%: Low income basis for Medicaid eligibility - Food/housing insecurity - Low levels of educational attainment - Racial/ethnic minority status - Behaviors such as drug/alcohol abuse and smoking and comorbidities including mental illness HEDIS Cervical Cancer Screening Measure - National Committee for Quality Assurance’s (NCQAs) Healthcare Effectiveness Data and Information Set (HEDIS) is the standard by which plans are compared p on q quality y health p - Roughly compatible with clinical guidelines - ACS & USPSTF recommend screening every 11 2 years for women ages 21-29 and every 2-3 years thereafter if 3 consecutive normal tests y recorded; screening may be discontinued beyond ages 65-70 HEDIS Cervical Cancer Screening Measure Component Criteria Denominator Being female, aged between 21 and 64 at the end of the 2006, and being continuously enrolled in a MassHealth managed care plan during 2004, 2005 and 2006 with at most one gap in enrollment of no more than 45 days during each calendar year. Numerator The numerator event was having at least one Papanicolaou (Pap) smear test in the past three years, 2004-2006. Data collection Hybrid: numerator status determined from a methodology combination of insurance claims and medical records Massachusetts and National Medicaid Managed Care 2007 Reported Rates HEDIS Cervical Cancer Screening Measure R t Rate Plan† Rate Range MassHealth Mean 79% 74 - 85% National Medicaid Mean 66% National Medicaid 90th percentile 77% † Five health plans serve Massachusetts Medicaid Managed Care Members Conceptual (Andersen) Model Predisposing Characteristics • Age cohort • Race/ethnicity • Primary language • Disability status • Neighborhood educational attainment Enabling Resources Community y • Residential county • Neighborhood income Insurance • Health plan Provider • Provider type • Setting (ED vs. office) Need for Care • CDPS illness burden score†; • Diagnoses†: substance abuse,, mental health disorders • Domestic violence • Tobacco use Cervical Cancer Screening † Andersen considered past medical history to be a ‘predisposing’ factor and current symptoms and diagnoses as ‘need’. Our data on illness burden and specific (generally, chronic) diagnoses were collected within the two-year HEDIS data collection period and may be expected to affect contemporaneous perceived need for screening. Data - W Working ki with ith th the MassHealth M H lth plans, l CHPR was able to access member-level HEDIS data - Dependent variable: numerator hit status - Indicates whether the member met the measure’s requirements - Independent variables: - Individual, family and neighborhood characteristics; diagnoses; disability status; health care utilization; region; provider type; and health plan Analysis - M Multivariate lti i t llogistic i ti regression i using i STATA version 9.2 - Adjusted odds ratios presented j with robust - Standard errors adjusted Huber-White Sandwich variance estimator - Significance evaluated at alpha≤0.05 - Model passed the Hosmer-Lemshow goodness-of-fit test - 83% of cases were correctly classified Sample Description - 12% of women sampled for the measure had a substance use disorder (SUD) - 25% were diagnosed with severe mental illness (SMI; mostly major depression) - 26% qualified for Medicaid because of a work-limiting disability - Among women with disabilities, 50% had an SMI and 21% had an SUD - A majority j it off women with ith an SUD h had d att lleastt one co-occurring SMI diagnosis Adjusted Odds Ratios (OR) for Cervical Cancer Screening Adjusted OR Predisposing Individual Characteristics 21 <= Age <= 30 1.62 *** 50 <= Age <= 65 0.61 *** Black race 1.57 Hispanic ethnicity 1.62 Other race/ethnicity 1 36 1.36 Unknown race/ethnicity 1.26 Spanish primary language 1.65 Other primary language 1.21 Has a work-limiting disability 0.65 * Neighborhood Characteristics High school graduate, percent 1.43 College graduates graduates, percent 2 69 2.69 Continued… Continued from previous slide… Adjusted OR Enabling Health Care Utilization ≥ 1 ED visit 0.96 1-29 Ambulatory y Office Visits 7.77 *** 30+ Ambulatory Office Visits 17.30 *** Provider: Community Health Center 1.14 Provider: Single-practice 0.98 Provider: Other type 0.98 N i hb h d Ch Neighborhood Characteristics t i ti Household income, mean ($000) 1.01 Percent minority y race/ethnicityy 1.46 6 Continued… Continued from previous slide… Adjusted OR Need Overall Illness Burden CDPS Illness Burden Score Diagnoses Alcohol/drug abuse Tobacco use Severe mental illnessa Other mental illnessb Domestic violence Hosmer-Lemeshow goodness of fit (p-value) Percent of cases correctly classified Number u be o of obse observations a o s 1 19 1.19 0.54 *** 1.27 1.21 1.24 0.43 * 0.07 82.7% 1,868 ,868 Note. Reference groups for odds ratios in the table are: age cohort,31-49 years old; race/ethnicity, white; primary language language, English;provider type, type group practice. practice All equations are also adjusted for county of residence and health plan. a Major depression, schizophrenia, and bipolar disorder b Other depression, anxiety disorder and other mental illness *<0.05; **<0.01; ***<0.001 Key Findings - Some vulnerable groups less likely to be screened work limiting disabilities (OR=0.65) (OR 0.65) - Women with work-limiting - Women with an SUD, particularly drug abuse or dependence (OR=0 (OR 0.54) 54) - Women with evidence of domestic violence in claims data (OR=0 (OR 0.43) 43) - However, we found no evidence of an association between mental illness and screening and no evidence of racial/ethnic disparities - Provider contacts are important important, but the type of PCP is not Comparison with Prior Studies - - - Some prior S i research hh has shown h that h women with i h di disabilities bili i less likely to be screened, while other studies find no significant association after controlling for socio-demographic variables - We found that substantially lower odds of screening (adjusted for individual demographics and area socioeconomic variables) for women with disabilities Studies show that women who abuse drugs or alcohol more likely to be diagnosed with cervical cancer, yet there are few studies t di on th the iimpactt off substance b t use disorders di d on screening i - We found substantial negative effects of drug abuse/dependence p on regular g Pap p testing g Literature on mental illness mixed - We found no significant effects of mental health conditions as a group or individually i di id ll Study Limitations - Findings may not be generalizable to other states with different Medicaid benefit structures or population characteristics - Use of administrative data - Some member characteristics unobserved or poorly observed (e (e.g., g substance use use, race, domestic violence) - No data on individual circumstances and type of disability to better understand our findings for women with a disability Data Strengths - Administrative claims data offer an advantage over survey data by eliminating recall and social desirability biases, allowing for relatively long, accurate look back periods - Self-reported Pap screening data have been found to overstate screening rates - Claims data contain rich information on comorbid diagnoses diagnoses, usual source of care care, and other variables that may drive results Discussion - Healthy People 2010 set a goal of 90% of women ages 18 and older receiving a Papanicolaou (Pap) smear test within the prior three years - Massachusetts Medicaid Managed Care Organizations had not reached this goal by 2007, though MA Medicaid on average exceeded National Medicaid 90th percentile - Vulnerable groups less likely to be screened Implications - Women with SUDs or disabilities face barriers to cervical cancer screening - Physician contacts may be key to i increasing i routine ti screening i and d reducing preventable cervical cancer morbidity and mortality - Better integration of drug abuse treatment with preventive health care is needed