Shah, Nilay D.

Background: Individualized diabetes management would benefit from prospectively identifying well-controlled patients at risk of losing glycemic control. Objectives: To identify patterns of hemoglobin A1c (HbA1c) change among patients with stable controlled diabetes. Research Design: Cohort study using OptumLabs Data Warehouse, 2001–2013. We develop and apply a machine learning framework that uses a Bayesian estimation of the mixture of generalized linear mixed effect models to discover glycemic trajectories, and a random forest feature contribution method to identify patient characteristics predictive of their future glycemic trajectories. Subjects: The study cohort consisted of 27,005 US adults with type 2 diabetes, age 18 years and older, and stable index HbA1c <7.0%. Measures: HbA1c values during 24 months of observation. Results: We compared models with k=1, 2, 3, 4, 5 trajectories and baseline variables including patient age, sex, race/ethnicity, comorbidities, medications, and HbA1c. The k=3 model had the best fit, reflecting 3 distinct trajectories of glycemic change: (T1) rapidly deteriorating HbA1c among 302 (1.1%) youngest (mean, 55.2 y) patients with lowest mean baseline HbA1c, 6.05%; (T2) gradually deteriorating HbA1c among 902 (3.3%) patients (mean, 56.5 y) with highest mean baseline HbA1c, 6.53%; and (T3) stable glycemic control among 25,800 (95.5%) oldest (mean, 58.5 y) patients with mean baseline HbA1c 6.21%. After 24 months, HbA1c rose to 8.75% in T1 and 8.40% in T2, but remained stable at 6.56% in T3. Conclusions: Patients with controlled type 2 diabetes follow 3 distinct trajectories of glycemic control. This novel application of advanced analytic methods can facilitate individualized and population diabetes care by proactively identifying high risk patients.

Abstract The epidemiology and outcomes of tumor lysis syndrome (TLS) are understudied. We used the National Inpatient Sample (NIS), a nationally representative weighted sample of all U.S. hospital discharges, to study outcomes and predictors of mortality in hospitalized patients with TLS. The NIS was queried for patients with a discharge diagnosis of TLS (ICD-9 code 277.88) from 2010–2013. Baseline characteristics and outcomes were analyzed. A multivariable logistic regression analysis was performed to identify predictors of mortality. From 2010–2013, 28,370 patients were discharged with a diagnosis of TLS. The most common malignancies were non-Hodgkin lymphoma (30%), solid tumors (20%), acute myeloid leukemia (19%), and acute lymphocytic leukemia (13%). Overall in-hospital mortality was 21%. The median length of stay was 10 days (IQR 5-22). Sixty-nine percent of patients experienced a severe complication, including sepsis (22%, 95% confidence interval [CI] 21–23), dialysis (15%, 95% CI 14–16), acute respiratory failure (23%, 95% CI 22–24), mechanical ventilation (16%, 95% CI 15–17), gastrointestinal hemorrhage (6%, 95% CI 5–7), cerebral hemorrhage (2%, 95% CI 2–3), seizures (1%, 95% CI 0.6–1), and cardiac arrest (2%, 95% CI 2–3). Predictors of mortality were derived from a multivariable logistic regression and included age, Elixhauser comorbidity score, insurance status, teaching versus nonteaching hospital, and cancer type. Predictors of increased length of stay included age, race, teaching versus nonteaching hospital, and cancer type. In the U.S., many patients with TLS develop life-threatening complications and a quarter die during hospitalization. As more cancer treatments become available, strategies to improve the supportive care of patients with TLS should be a priority.

Background. We test the hypotheses that use of the Chest Pain Choice (CPC) decision aid (DA) would be similarly effective in potentially vulnerable subgroups but increase knowledge more in patients with higher education and trust in physicians more in patients from racial minority groups. Methods. This was a secondary analysis of a multicenter randomized trial in adults with chest pain potentially due to acute coronary syndrome. The trial compared an intervention group engaged in shared decision making (SDM) using CPC to a control group receiving usual care (UC). We assessed for subgroup effects based on age, sex, race, income, insurance, education, literacy, and numeracy. We dichotomized each characteristic and tested for interactions using regression models with indicators for arm assignment and study site. Results. Of 898 patients (451 DA, 447 UC), over 50% were female, over one-third were black, nearly one-third had a high school education or less, and over 60% had “low” health literacy. The DA did not increase knowledge more in patients with higher education ( P for interaction = 0.06) but did increase knowledge more in the “typical” than in the “low” numeracy subgroup (10.6% v. 4.7%, absolute difference [AD] = 5.9%, P for interaction = 0.025). The DA did not significantly increase patient trust in physicians in racial minorities ( P for interaction = 0.06) but did increase trust more in patients with “low” literacy compared with those with “typical” literacy (3.7% v. –1.4%, AD = 5.1, P for interaction = 0.011). Conclusions. CPC benefited all sociodemographic groups to a similar extent, with greater knowledge transfer in patients with higher numeracy and greater physician trust in patients with “low” health literacy. Tailoring SDM interventions to patient characteristics may be necessary for optimal effectiveness.

167 Background: Physicians have a variety of resources available about the treatment of localized prostate cancer. Yet, it is largely unknown what resources and evidence radiation oncologists and urologists commonly use to inform patients and possibly alter clinical practice. In this national survey study, we assessed the types of resources prostate cancer specialists’ use and the level of evidence needed to change treatment recommendations in localized prostate cancer. Methods: From a random sample, 1,422 physicians were sent a mail survey assessing what type of information was used to guide prostate cancer treatment. The survey also queried what level of evidence could alter recommendations. Multivariable logistic regression models were used to identify for differences in physician characteristics for each outcome. Results: Survey response rates were similar for radiation oncologists and urologists (44% vs. 46%; p=0.46). Specialty-specific journals represented the most commonly used resource to inform the clinical practice for radiation oncologists (65%) and urologists (70%). Relative to radiation oncologists, urologists were less likely to utilize medical journals (25% vs. 38.7%, OR: 0.50; p=0.01), cancer journals (21.7% vs. 50.8%, OR: 0.50; p<0.001), but more likely to rely on clinical guidelines (46.3% vs. 37.7%,OR: 1.6; p=0.006). Radiation oncologists and urologists both viewed RCT’s as the most likely source of evidence needed to change treatment recommendations (85% vs. 77%). Conclusions: Both specialties most commonly rely on their specialty-specific journals to inform their clinical practice and view RCT’s as the level of evidence needed to change treatment recommendations for localized prostate cancer. Radiation oncologists embraced greater diversity in literature, while urologists were more likely to use clinical guidelines. Utilizing this information may overcome barriers to the dissemination and adoption of new treatment paradigms for localized prostate cancer.

25 Background: Women with diabetes are at an increased risk of developing breast cancer, present with later stage disease, and experience higher breast cancer-specific mortality compared to nondiabetic women. Thus, it is imperative that diabetic women undergo routine screening mammography. Previous studies have shown conflicting results with smaller studies showing lower screening mammography rates among diabetics while a study using national self-report data showed that diabetic women were equally likely to be screened compared to nondiabetic women. Given the conflicting data and limitations of previous studies, we aimed to study receipt of mammography in diabetic women using a nationally representative administrative claims database. Methods: This was a retrospective study using a nationally representative administrative claims database from January 1, 2010, through December 31, 2012. We compared rates of receipt of mammography within a 2-year time frame in diabetic and nondiabetic women ages 40 to 64. We used Healthcare Effectiveness Data and Information Set (HEDIS) logic to define women with diabetes. We used CPT and ICD-9 codes to identify mammogram services received. We calculated odds ratios with 95% confidence intervals to compare mammogram rates among diabetics compared to nondiabetics. Results: 1,723,674 nondiabetic women and 133,973 diabetic women were identified. Among women ages 40 to 49, diabetic women had significantly decreased mammography rates compared to nondiabetic women (OR 0.84; 95% CI 0.83, 0.85). Rates were also lower in diabetic women ages 50 to 64 (OR 0.88; 95% CI 0.88, 0.91). Despite lower mammography rates among diabetics, diabetic women had 1.78 times more visits than nondiabetic women. Conclusions: Women with diabetes are less likely to undergo routine screening mammography than women without diabetes, despite having more encounters with health care providers. This study serves as a reminder to provide timely preventive health services to all patients regardless of medical comorbidities.

5 Background: The 11/2009 USPSTF breast cancer screening update recommended against routine screening mammography for women age 40-49, created confusion and prompted organizations to release opposing statements. We aimed to determine if the USPSTF update for breast cancer screening impacted screening mammography utilization in women age 40-49. Methods: We conducted a time-series analysis utilizing administrative claims data from over 100 health plans. Women ages 40-64 with at least one month of enrollment from 01/2006-12/2010 were included. Medicare claims are not included and thus women ≥65 were excluded. The start date of 01/2006 was chosen to account for possible effects of the recent recession. We developed claims-based algorithms to identify the number of monthly screening mammograms. Time series models were fit using SAS PROC AUTOREG; strong seasonal fluctuations were adjusted by including an autoregressive error of order 12. Results: Over 7.9 million women were included. Prior to the intervention, the baseline monthly mammography rate was 40.9/1,000 women for the 40-49 group and 47.4/1,000 women for the 50-64 group. Based on projections from pre-intervention trends, the update was associated with a 5.72% (1.27,10.18) and 4.97% (1.11,8.84) decrease in mammography rate at 3 and 10 months post-intervention, respectively in the 40-49 group. The intervention had no effect on mammography rates in the 50-64 group. This translates to 53,969 fewer mammograms performed in the year following the update for women ages 40-49 in this dataset. Conclusions: We present the first estimates of the impact of the USPSTF breast cancer screening update using a large nationally representative database. The update was associated with a small but significant decrease in mammography rates for women ages 40-49, while no change was seen for women ages 50-64, which is consistent with the context of the guideline change. A modest effect is also in line with public backlash and the release of numerous conflicting guidelines. These findings underscore the need for further research on benefits and risks of screening mammography as it is difficult to act on numerous sources of contradictory information.

284 Background: MM is an uncommon cancer with annual incidence of only 27,000 cases in the US. Our study determined the extent to which the number of MM patients treated annually in a treatment facility affected all-cause mortality. Methods: We used the National Cancer Data Base (NCDB) to identify adult patients with MM diagnosed from 2003-2011. NCDB is sourced from over 1,500 Commission on Cancer-accredited cancer registries representing 70% of cancer cases in the US. We classified treatment facilities by quartiles based on facility volume (mean patients/year): Quartile 1 (Q1: < 3.6), Quartile 2 (Q2: 3.6-6.1), Quartile 3 (Q3: 6.1-10.3) and Quartile 4 (Q4: > 10.3). We used hot deck imputation to account for missing data, Cox regression to analyze the association between facility volume and time-to-death, and random intercepts to adjust for multiple patients per facility. Results: There were 94,722 MM patients cared for at 1,333 facilities. Most patients (73.5%) were diagnosed and treated in the same facility. The median age at diagnosis was 67 years and 54.7% were males. The median annual facility volume was 6.1 patients/year (IQR: 3.6-10.3; range: 0.2-109.9). The distribution of patients according to facility volume was Q1 (5.2%), Q2 (12.6%), Q3 (21.9%) and Q4 (60.3%). The unadjusted median overall survival by facility volume was: Q1: 26.9 months, Q2: 29.1 months, Q3: 31.9 months and Q4: 49.1 months. After multivariable analysis adjusting for demographic (sex, age, race, ethnicity), socioeconomic (income, education, insurance type), geographic (area of residence, treatment facility location), co-morbidity (Charlson-Deyo score), and disease-specific (year of diagnosis) factors, we show that facility volume remains an independent predictor of all-cause mortality. Compared to patients treated at Q4 facilities, patients treated at lower quartile facilities had a higher risk of death (Q3 HR: 1.16 [95% CI, 1.12-1.20]; Q2: 1.21 [1.17-1.26]; Q1: 1.27 [1.22-1.34]). We observed an inverse volume-outcome relationship up to an annual facility volume of approximately 60 patients. Conclusions: MM patients treated at higher volume facilities had lower risk of mortality compared to those treated at lower volume facilities.