Patient and physician characteristics associated with use of erythropoiesis-stimulating agents in myelodysplastic syndrome patients have not yet been described. Myelodysplastic syndrome patients diagnosed from 2001 to 2005 were identified from the Surveillance Epidemiology and End Results-Medicare database. Multivariate regressions examined the association between patient and physician characteristics and the probability of receiving any erythropoiesis-stimulating agents, and of receiving therapeutic-length (≥8 week) treatment episodes.Among the 6,588 myelodysplastic syndrome patients studied, 65% received erythropoiesis-stimulating agents. Use of erythropoiesis-stimulating agents was lower for blacks compared to whites (OR 0.78; 95% CI:0.61–0.99), single persons compared to married (OR 0.77; 95% CI:0.62–0.97), Medicaid recipients (OR 0.66; 95% CI:0.55–0.79), and those living in census tracts with lower educational attainment. Patients who did not consult a hematology-oncology specialist were less likely to receive erythropoiesis-stimulating agents. Specialist access, financial resources and mobility are key determinants of receipt of erythropoiesis-stimulating agents among myelodysplastic syndrome patients.
6006 Background: Performance status (PS) is one of the most important predictors of cancer outcomes, and a key determinant of treatment. Observational studies using administrative data represent an important component of comparative effectiveness research (CER). Data sources such as SEER-Medicare, used to study cancer treatment patterns, outcomes and costs, do not capture patient PS. Claims based indicators of service use associated with poor PS (e.g. home oxygen, wheelchair) were tested in CER studies in colon and lung cancer, and found to predict treatment and survival. The objective of this research is to further develop methods to incorporate PS into CER studies, using an approach similar to the Charlson Comorbidity Index. We report on initial phases in development of a multivariate regression model that will predict PS as a function of a broad set of claims based indicators. Methods: The Medicare Current Beneficiary Survey (MCBS), a nationally representative survey of community and facility based Medicare beneficiaries, includes self or proxy reported limitations to activities of daily living (ADL), instrumental ADL, and physical strength, and is linked to Medicare (Parts A and B) claims. We used MCBS question-response sets to measure key dimensions in the ECOG PS scale (self-care, light/sedentary activity, and strenuous activities), and developed an algorithm to assign PS scores to respondents from the 2005 MCBS (n = 10,565). Results: PS assignments in the MCBS were 77% ECOG PS 0-1, 16% PS 2, and 7.5% PS 3-4. Among the 1/3 of community residents reporting CHF, COPD, AMI, stroke, or rheumatoid arthritis, 29% were PS > = 2 vs. 2% among others. Facility based residents had higher prevalence of PS 3-4 (74%) vs. 3% in the community. Conclusions: Our research uses a rich source of data on health and functional status, independent of chronological age, for an elderly and disabled population. Application of these data through a model that predicts PS will strengthen CER research based on administrative data, in cancer and other chronic diseases. In ongoing research we are identifying potential claims based indicators of PS, estimating multivariate prediction models, and will validate our predicted PS value in SEER-Medicare CER studies. No significant financial relationships to disclose.
6085 Background: Observational studies analyzing cancer treatment and outcomes in datasets such as Surveillance, Epidemiology and End Results (SEER)-Medicare are unable to capture performance status (PS), a key determinant of treatment. In prior work we developed a multivariate regression model to predict poor PS (ECOG 3-4 versus 0-2) as a function of claims-based service use indicators. We report on initial predictive validation of our model by examining chemotherapy treatment in a cohort of metastatic breast cancer (MBC) patients. Methods: 1,519 female estrogen receptor negative MBC patients aged ≥66 were identified in 1999-2005 SEER registry data linked to Medicare Part A and B claims. We generated the predicted probability of poor PS (PredPS) by creating the set of service use indicators from Medicare Part B claims one year before diagnosis and applying coefficients from the prediction model. We examined the relationship of PredPS to age, race, socioeconomic status, and the Charlson Comorbidity Index (CCI). We also included PredPS as either a discrete (Good/ Poor) or continuous variable in multivariate logistic regressions to predict receipt of chemotherapy within 6 months following diagnosis, obtained from Medicare claims. Results: 258 (17%) had poor PredPS; 494 women (33%) received chemotherapy. Poor PredPS patients were less likely to receive chemotherapy (17% versus 36%; p<0.01); poor PredPS was more prevalent among non whites (31% versus 14% for whites; p<0.01), and the oldest (29% among 85+ versus 14% among patients aged 66-74 years). After adjusting for sociodemographics and CCI, MBC women with poor PredPS were 56% less likely to receive chemotherapy (OR 0.44, 95% CI 0.25, 0.77). Models with continuous PredPS showed that a 1% increase in poor predPS resulted in a 3% decrease in odds of receiving chemotherapy (OR 0.97, 95% CI 0.96, 0.99). Results were similar with alternative PredPS developed from models that included interactions with region and with Medicaid enrollment. Conclusions: A predicted PS measure derived from Medicare claims was associated with receipt of chemotherapy and this effect appears to be independent of sociodemographics and CCI. Ongoing analyses will validate predPS in cancer survival models.
To identify baseline demographics and clinical characteristics associated with healthcare costs among patients with atherosclerotic cardiovascular disease (ASCVD). This retrospective cohort study identified newly diagnosed ASCVD patients aged ≥18 years using claims data from the HealthCore Integrated Research Database (HIRDSM) between 1/1/07 and 11/30/12 (index date=first ASCVD diagnosis date). Patients had both ≥ 12 month pre- and post-index insurance enrollment, valid baseline lipid panel values, and no baseline lipid lowering medication use. Costs were adjusted to 2013 U.S. dollar values. Bivariate analyses and generalized linear models with gamma distribution and log link were used to examine baseline factors associated with 12 month follow-up all-cause and ASCVD-related healthcare costs. In the regression model for all-cause healthcare costs (N=26,388), older age, plan region including South and West (vs. Midwest), higher Quan-Charlson Comorbidity Index, index acute coronary syndrome (ACS), ischemic stroke or transient ischemic attack, baseline depression, pain, obesity, and chronic kidney disease, baseline use of antihypertensive agents, antidiabetic medications, and digoxin, and higher baseline all-cause healthcare costs were positively associated with follow-up all-cause healthcare costs (p<0.05). In addition, female, Northeast plan (vs. Midwest), Health Maintenance Organization (vs. Preferred Provider Organization), Medicare Advantage plans, index coronary heart disease (except for ACS) or peripheral artery disease, baseline dyslipidemia, and baseline goal attainment of low-density lipoprotein cholesterol (<100 mg/dL), high-density lipoprotein cholesterol ( >40/50 mg/dL for males and females respectively), triglycerides (<150 mg/dL), and total cholesterol level (<200 mg/dL) were negatively associated with follow-up all-cause healthcare costs (p<0.05). Similar findings were reported for ASCVD-related healthcare costs (N=26,376). As expected, age, gender, baseline comorbid conditions, baseline use of specific medications, baseline lipid profiles, and more severe index ASCVD were significantly associated with all-cause and ASCVD-related healthcare costs. Geographic location and health insurance type also played a significant role in healthcare costs among ASCVD patients.
Dialect Attitude is conceptualized as an individual's cognitive and affective evaluation of a dialect and its speakers. In the contemporary China, dialect is suffering from significant stigmatization, resulting in social inequalities, which hinder sustainable development. This study aims to reveal the Chinese public attitudes towards dialects, and clarify the potential determinants related to heterogeneous attitudes at a macro level.We combine the crawler technology and sentiment analysis to conduct a provincial cross-sectional study. We collect 1,650,480 microblogs about public attitudes towards dialects from Microblog across 31 specific Chinese provinces. Spatial regression models are utilized to clarify the influence of macro-level determinants on differences in public attitudes.The present study reveals that: (1) The Chinese public generally holds positive attitudes towards dialects, with significant variation between provinces. (2) Political Resource (β = 0.076, SD = 0.036, P<0.05), Economic Development (β = 0.047, SD = 0.022, P<0.05), and Cultural Resource (β = 0.054, SD = 0.021, P<0.05) promote public positive attitudes towards dialects. (3) Political Resource and Culture Resource influence more significant in the relatively advantaged regions, and Economic Development poses a higher influence in the relatively disadvantaged regions.Basing on the combination of crawler technology and sentiment analysis, the present study develops the most comprehensive database which takes 1,650,480 dialects-related microblogs from 31 Chinese provinces, and describes the following scenario: (1) Overall, the Chinese public shares a relatively positive attitude towards dialects with significant variations among different provinces, (2) Political Resource, Economic Development and Culture Resource pose positive effects on Chinese public attitudes towards dialects and (3) Political Resource and Culture Resource influence more significant in the relatively disadvantaged regions, and Economic Development poses a higher influence in the relatively advantaged regions.
