Optimal Donor Selection Across Multiple Outcomes For Hematopoietic Stem Cell Transplantation By Bayesian Nonparametric Machine Learning.

Allogeneic hematopoietic cell transplantation (HCT) is one of the only curative treatment options for patients suffering from life-threatening hematologic malignancies; yet, the possible adverse complications can be serious even fatal. Matching between donor and recipient for 4 of the HLA genes is widely accepted and supported by the literature. However, among 8/8 allele matched unrelated donors, there is less agreement among centers and transplant physicians about how to prioritize donor characteristics like additional HLA loci (DPB1 and DQB1), donor sex/parity, CMV status, and age to optimize transplant outcomes. This leads to varying donor selection practice from patient to patient or via center protocols. Furthermore, different donor characteristics may impact different post transplant outcomes beyond mortality, including disease relapse, graft failure/rejection, and chronic graft-versus-host disease (components of event-free survival, EFS). We develop a general methodology to identify optimal treatment decisions by considering the trade-offs on multiple outcomes modeled using Bayesian nonparametric machine learning. We apply the proposed approach to the problem of donor selection to optimize overall survival and event-free survival, using a large outcomes registry of HCT recipients and their actual and potential donors from the Center for International Blood and Marrow Transplant Research (CIBMTR). Our approach leads to a donor selection strategy that favors the youngest male donor, except when there is a female donor that is substantially younger.

Isolation and survival: The impact of local and MSA isolation on survival among non-Hispanic Black women diagnosed with breast cancer in the United States using a SEER-Medicare cohort.

BACKGROUND: Residential segregation is an important factor that negatively impacts cancer disparities, yet studies yield mixed results and complicate clear recommendations for policy change and public health intervention. In this study, we examined the relationship between local and Metropolitan Statistical Area (MSA) measures of Black isolation (segregation) and survival among older non-Hispanic (NH) Black women with breast cancer (BC) in the United States. We hypothesized that the influence of local isolation on mortality varies based on MSA isolation-specifically, that high local isolation may be protective in the context of highly segregated MSAs, as ethnic density may offer opportunities for social support and buffer racialized groups from the harmful influences of racism. METHODS: Local and MSA measures of isolation were linked by Census Tract (CT) with a SEER-Medicare cohort of 5,231 NH Black women aged 66-90 years with an initial diagnosis of stage I-IV BC in 2007-2013 with follow-up through 2018. Proportional and cause-specific hazards models and estimated marginal means were used to examine the relationship between local and MSA isolation and all-cause and BC-specific mortality, accounting for covariates (age, comorbidities, tumor stage, and hormone receptor status). FINDINGS: Of 2,599 NH Black women who died, 40.0% died from BC. Women experienced increased risk for all-cause mortality when living in either high local (HR = 1.20; CI = 1.08-1.33; p < 0.001) or high MSA isolation (HR = 1.40; CI = 1.17-1.67; p < 0.001). A similar trend existed for BC-specific mortality. Pairwise comparisons for all-cause mortality models showed that high local isolation was hazardous in less isolated MSAs but was not significant in more isolated MSAs. INTERPRETATION: Both local and MSA isolation are independently associated with poorer overall and BC-specific survival for older NH Black women. However, the impact of local isolation on survival appears to depend on the metropolitan area's level of segregation. Specifically, in highly segregated MSAs, living in an area with high local isolation is not significantly associated with poorer survival. While the reasons for this are not ascertained in this study, it is possible that the protective qualities of ethnic density (e.g., social support and buffering from experiences of racism) may have a greater role in more segregated MSAs, serving as a counterpart to the hazardous qualities of local isolation. More research is needed to fully understand these complex relationships. FUNDING: National Cancer Institute.

Nonparametric failure time: Time-to-event machine learning with heteroskedastic Bayesian additive regression trees and low information omnibus Dirichlet process mixtures.

Many popular survival models rely on restrictive parametric, or semiparametric, assumptions that could provide erroneous predictions when the effects of covariates are complex. Modern advances in computational hardware have led to an increasing interest in flexible Bayesian nonparametric methods for time-to-event data such as Bayesian additive regression trees (BART). We propose a novel approach that we call nonparametric failure time (NFT) BART in order to increase the flexibility beyond accelerated failure time (AFT) and proportional hazard models. NFT BART has three key features: (1) a BART prior for the mean function of the event time logarithm; (2) a heteroskedastic BART prior to deduce a covariate-dependent variance function; and (3) a flexible nonparametric error distribution using Dirichlet process mixtures (DPM). Our proposed approach widens the scope of hazard shapes including nonproportional hazards, can be scaled up to large sample sizes, naturally provides estimates of uncertainty via the posterior and can be seamlessly employed for variable selection. We provide convenient, user-friendly, computer software that is freely available as a reference implementation. Simulations demonstrate that NFT BART maintains excellent performance for survival prediction especially when AFT assumptions are violated by heteroskedasticity. We illustrate the proposed approach on a study examining predictors for mortality risk in patients undergoing hematopoietic stem cell transplant (HSCT) for blood-borne cancer, where heteroskedasticity and nonproportional hazards are likely present.

Medication delivery factors and adjuvant endocrine therapy adherence in breast cancer.

PURPOSE: Over 50% of breast cancer patients prescribed a 5-year course of daily oral adjuvant endocrine therapy (ET) are nonadherent. We investigated the role of costs and cancer medication delivery mode and other medication delivery factors on adherence. METHODS: We conducted a retrospective cohort study of commercially insured and Medicare advantage patients with newly diagnosed breast cancer in 2007-2015 who initiated ET. We examined the association between 12-month ET adherence (proportion of days covered by fills ≥ 0.80) and ET copayments, 90-day prescription refill use, mail order pharmacy use, number of pharmacies, and synchronization of medications. We used regression models to estimate nonadherence risk ratios adjusted for demographics (age, income, race, urbanicity), comorbidities, total medications, primary cancer treatments, and generic AI availability. Sensitivity analyses were conducted using alternative specifications for independent variables. RESULTS: Mail order users had higher adherence in both commercial and Medicare-insured cohorts. Commercially insured patients who used mail order were more likely to be adherent if they had low copayments (< $5) and 90-day prescription refills. For commercially insured patients who used local pharmacies, use of one pharmacy and better synchronized refills were also associated with adherence. Among Medicare patients who used mail order pharmacies, only low copayments were associated with adherence, while among Medicare patients using local pharmacies both low copayments and 90-day prescriptions were associated with ET adherence. CONCLUSION: Out-of-pocket costs, medication delivery mode, and other pharmacy-related medication delivery factors are associated with adherence to breast cancer ET. Future work should investigate whether interventions aimed at streamlining medication delivery could improve adherence for breast cancer patients.

Feasibility of a pharmacist-led symptom monitoring and management intervention to improve breast cancer endocrine therapy adherence.

BACKGROUND: Adjuvant endocrine therapy (AET) for breast cancer reduces mortality, but one-third to one-half of patients discontinue it early or are nonadherent. OBJECTIVE: We developed a pilot single-site study of patients with evidence of early nonadherence to AET to assess the feasibility of a novel, clinical pharmacist-led intervention targeting symptom and medication management. METHODS: Patients with prescription fill records showing nonadherence were enrolled in a single-arm feasibility study. Automated reminders were sent by e-mail or text with a link to symptom monitoring assessments weekly for 1 month and monthly until 6 months. Clinical oncology pharmacists used guideline-based symptom management and other medication management tools to support adherence and ameliorate symptoms reported on the assessments. Patient-reported outcome assessments included physical, mental, and social health domains and self-efficacy to manage symptoms and medications. Feasibility outcomes included completion of symptom reports and pharmacist recommendations. RESULTS: Of 19 participants who were nonadherent who enrolled and completed initial assessments, 18 completed all final study procedures, with 14 completing all assessments and no patient missing more than 3 assessments. All 18 participants reported at least one of 3 symptom types, and the majority reported attempting pharmacist recommendations. Patient-reported measures of physical, mental, and social health and self-efficacy improved, and 44% of the patients became adherent. CONCLUSION: An intervention using pharmacists in an oncology practice to systematically monitor and manage symptoms shows promise to reduce symptoms, enhance support and self-efficacy, and improve adherence to AET.

Ecological Study of Variability in the Relationship between Liver Cancer Mortality and Racial Residential Segregation.

Racial segregation has been identified as a predictor for the burden of cancer in several different metropolitan areas across the United States. This ecological study tested relationships between racial segregation and liver cancer mortality across several different metropolitan statistical areas in Wisconsin. Tract-level liver cancer mortality rates were calculated using cases from 2003-2012. Hotspot analysis was conducted and segregation scores in high, low, and baseline mortality tracts were compared using ANOVA. Spatial regression analysis was done, controlling for socioeconomic advantage and rurality. Black isolation scores were significantly higher in high-mortality tracts compared to baseline and low-mortality tracts, but stratification by metropolitan areas found this relationship was driven by two of the five metropolitan areas. Hispanic isolation was predictive for higher mortality in regression analysis, but this effect was not found across all metropolitan areas. This study showed associations between liver cancer mortality and racial segregation but also found that this relationship was not generalizable to all metropolitan areas in the study area.

Pharmacy deserts and patients with breast cancer receipt of influenza vaccines.

BACKGROUND: Yearly influenza vaccination is strongly recommended at age 65 and reimbursed by Medicare without copays or deductibles at pharmacies and clinical settings. Uptake is low among patients with a high risk for influenza complications and good access to specialist care, such as recent cancer survivors. We hypothesized that more accessible pharmacies could be associated with higher immunization uptake in such patients. OBJECTIVES: To determine whether pharmacy access is associated with influenza vaccination in subjects recently diagnosed with breast cancer, and whether this association differs by additional risk factors for influenza complications. METHODS: We examined a cohort of patients with stage 0-III breast cancer diagnosed 2011-2015 from the Surveillance, Epidemiology, and End Results-Medicare cancer registry. All retail pharmacies in the United States were identified, and pharmacy access was measured by assessing supply and demand in each census tract using a 2-stage floating catchment area approach that accounted for pharmacy driving distances recommended by the Centers for Medicare and Medicaid Services. We examined the association of pharmacy access with influenza vaccination after breast cancer diagnosis in regression models. RESULTS: More than 11% of 45,722 patients with breast cancer lived in census tracts where no pharmacies were within recommended driving distances from the population-weighted tract center. Vaccination in the year after diagnosis was less likely for patients in these very low-access tracts (adjusted odds ratio 0.92 [95% CI 0.86-0.96]), black (0.55 [0.51-0.60]) and Hispanic (0.76 [0.70-0.83]) women, and Medicaid recipients (0.74 [0.69-0.79]). Vaccination was inversely associated with per capita income in the subject's census tract, but there was no difference in the pharmacy effect by race, ethnicity, or census tract income. CONCLUSION: Very low pharmacy access is associated with modest reductions in vaccination that could be useful for policy and planning regarding vaccinator resources and outreach.

Mortgage Lending Bias and Breast Cancer Survival Among Older Women in the United States.

PURPOSE: The objective was to examine the relationship between contemporary redlining (mortgage lending bias on the basis of property location) and survival among older women with breast cancer in the United States. METHODS: A redlining index using Home Mortgage Disclosure Act data (2007-2013) was linked by census tract with a SEER-Medicare cohort of 27,516 women age 66-90 years with an initial diagnosis of stage I-IV breast cancer in 2007-2009 and follow-up through 2015. Cox proportional hazards models were used to examine the relationship between redlining and both all-cause and breast cancer-specific mortality, accounting for covariates. RESULTS: Overall, 34% of non-Hispanic White, 57% of Hispanic, and 79% of non-Hispanic Black individuals lived in redlined tracts. As the redlining index increased, women experienced poorer survival. This effect was strongest for women with no comorbid conditions, who comprised 54% of the sample. For redlining index values of 1 (low), 2 (moderate), and 3 (high), as compared with 0.5 (least), hazard ratios (HRs) (and 95% CIs) for all-cause mortality were HR = 1.10 (1.06 to 1.14), HR = 1.27 (1.17 to 1.38), and HR = 1.39 (1.25 to 1.55), respectively, among women with no comorbidities. A similar pattern was found for breast cancer-specific mortality. CONCLUSION: Contemporary redlining is associated with poorer breast cancer survival. The impact of this bias is emphasized by the pronounced effect even among women with health insurance (Medicare) and no comorbid conditions. The magnitude of this neighborhood level effect demands an increased focus on upstream determinants of health to support comprehensive patient care. The housing sector actively reveals structural racism and economic disinvestment and is an actionable policy target to mitigate adverse upstream health determinants for the benefit of patients with cancer.

Optimal Donor Selection for Hematopoietic Cell Transplantation Using Bayesian Machine Learning.

PURPOSE: Donor selection practices for matched unrelated donor (MUD) hematopoietic cell transplantation (HCT) vary, and the impact of optimizing donor selection in a patient-specific way using modern machine learning (ML) models has not been studied. METHODS: We trained a Bayesian ML model in 10,318 patients who underwent MUD HCT from 1999 to 2014 to provide patient- and donor-specific predictions of clinically severe (grade 3 or 4) acute graft-versus-host disease or death by day 180. The model was validated in 3,501 patients from 2015 to 2016 with archived records of potential donors at search. Donor selection optimizing predicted outcomes was implemented over either an unlimited donor pool or the donors in the search archives. Posterior mean differences in outcomes from optimal donor selection versus actual practice were summarized per patient and across the population with 95% intervals. RESULTS: Event rates were 33% (training) and 37% (validation). Among donor features, only age affected outcomes, with the effect consistent regardless of patient features. The median (interquartile range) difference in age between the youngest donor at search and the selected donor was 6 (1-10) years, whereas the number of donors per patient younger than the selected donor was 6 (1-36). Fourteen percent of the validation data set had an approximate 5% absolute reduction in event rates from selecting the youngest donor at search versus the actual donor used, leading to an absolute population reduction of 1% (95% interval, 0 to 3). CONCLUSION: We confirmed the singular importance of selecting the youngest available MUD, irrespective of patient features, identified potential for improved HCT outcomes by selecting a younger MUD, and demonstrated use of novel ML models transferable to optimize other complex treatment decisions in a patient-specific way.

Guideline-concordant treatment predicts survival: a National Cancer Database validation study of novel composite locoregional and systemic treatment scores among women with early stage breast cancer.

BACKGROUND: The aim of this large nationwide study was to validate two novel composite treatment scores that address guideline-concordant locoregional and systemic breast cancer care. We examined the relationship between these two scores and their association with survival. METHODS: Women with Stage I-III unilateral breast cancer were identified within the National Cancer Database. For each woman, a locoregional and a systemic treatment score (0, 1, 2) was assigned based on receipt of guideline-concordant care. Multivariable Cox regression models evaluated the association between the scores and survival. RESULTS: 623,756 women were treated at 1,221 different American College of Surgeons Commission on Cancer (CoC) facilities. Overall, 86% had a locoregional treatment score of 2 (most guideline-concordant), 75% had a systemic treatment score of 2, and 72% had both scores of 2. Median follow-up was 4.5 years. Compared to women with a locoregional treatment score of 2, those with a score of 1 or 0 had a 1.7-fold and 2.0-fold adjusted greater risk of death. Compared to women with a systemic treatment score of 2, those with a score of 1 or 0 had a 1.5-fold and 2.1-fold adjusted greater risk of death. Risk-adjusted 5-year overall survival was 91.6% when both scores were 2 compared to 73.4% when both scores were 0. CONCLUSIONS: In this large national study of CoC facilities, two composite scores capturing guideline-concordant breast cancer care had independent and combined robust effects on survival. These clinically constructed novel scores are promising tools for health services research and quality-of-care studies.

The impact of generic aromatase inhibitors on initiation, adherence, and persistence among women with breast cancer: Applying multi-state models to understand the dynamics of adherence.

PURPOSE: Clinical trials have clearly documented the survival benefit of aromatase inhibitors (AIs); however, many women fail to initiate (primary nonadherence) or remain adherent to AIs (secondary nonadherence). Prior studies have found that costs impact secondary nonadherence to medications but have failed to examine primary nonadherence. The purpose of this study is to examine primary and secondary adherence following the reduction in copays due to the introduction of generic AIs. METHODS: Using Surveillance, Epidemiology, and End Results-Medicare data, we identified 50 054 women diagnosed with incident breast cancer between 2008 and 2013. We compare women whose copays would change and those whose would not, due to the receipt of cost-sharing subsidies before and after generics were introduced using a difference-in-difference (DinD) analysis. To examine primary and secondary nonadherence, we rely on a multistate model with four states (Not yet initiated, User, Not Using, and Death). We adjusted for baseline factors using inverse probability treatment weights and then simulated adherence for 36 months following diagnosis. RESULTS: The generic introduction of AIs resulted in patients initiating AIs faster (DinD = -4.7%, 95%CI = -7.0, -2.3; patients not yet initiating treatment at 6-months), being more adherent (DinD ranging in absolute increase of 8.1%-10.4%) and being less likely to not be using the therapy (DinD range in absolute decrease of 1.2% at 6 months to 8.8% at 24 months) for women that do not receive a subsidy after generics were available. CONCLUSIONS: Introduction of generic alternatives to AIs significantly reduced primary and secondary nonadherence.

Prevalence and scope of advanced practice provider oncology care among Medicare beneficiaries with breast cancer.

PURPOSE: Advanced practice providers (APPs) have increasingly become members of the oncology care team. Little is known about the scope of care that APPs are performing nationally. We determined the prevalence and extent of APP practice and examined associations between APP care and scope of practice regulations, phase of cancer care, and patient characteristics. METHODS: We performed an observational study among women identified from Medicare claims as having had incident breast cancer in 2008 with claims through 2012. Outpatient APP care included at least one APP independently billing for cancer visits/services. APP scope of practice was classified as independent, reduced, or restricted. A logistic regression model with patient-level random effects was estimated to determine the probability of receiving APP care at any point during active treatment or surveillance. RESULTS: Among 42,550 women, 6583 (15%) received APP care, of whom 83% had APP care during the surveillance phase and 41% during the treatment phase. Among women who received APP care during a given year of surveillance, the overall proportion of APP-billed clinic visits increased with each additional year of surveillance (36% in Year 1 to 61% in Year 4). Logistic regression model results indicate that women were more likely to receive APP care if they were younger, black, healthier, had higher income status, or lived in a rural county or state with independent APP scope of practice. CONCLUSIONS: This study provides important clinical and policy-relevant findings regarding national practice patterns of APP oncology care. Among Medicare beneficiaries with incident breast cancer, 15% received outpatient oncology care that included APPs who were billing; most of this care was during the surveillance phase. Future studies are needed to define the degree of APP oncology practice and training that maximizes patient access and satisfaction while optimizing the efficiency and quality of cancer care.

Normative ranges of nasal airflow variables in healthy adults.

PURPOSE: Virtual surgery planning based on computational fluid dynamics (CFD) simulations of nasal airflow has the potential to improve surgical outcomes for patients with nasal airway obstruction (NAO). Virtual surgery planning requires normative ranges of airflow variables, but few studies to date have quantified inter-individual variability of nasal airflow among healthy subjects. This study reports CFD simulations of nasal airflow in 47 healthy adults. METHODS: Anatomically accurate three-dimensional nasal models were reconstructed from cone beam computed tomography scans and used for steady-state inspiratory airflow simulations with a bilateral flowrate of 250 ml/s. Normal subjective sensation of nasal patency was confirmed using the nasal obstruction symptom evaluation and visual analog scale. Healthy ranges for several CFD variables known to correlate with subjective nasal patency were computed, including unilateral airflow, nasal resistance, airspace minimal cross-sectional area (mCSA), heat flux (HF), and surface area stimulated by mucosal cooling (defined as the area where HF > 50 W/m2). The normative ranges were targeted to contain 95% of the healthy population and computed using a nonparametric method based on order statistics. RESULTS: A wide range of inter-individual variability in nasal airflow was observed among healthy subjects. Unilateral airflow varied from 60 to 191 ml/s, airflow partitioning ranged from 23.8 to 76.2%, and unilateral mCSA varied from 0.24 to 1.21 cm2. These ranges are in good agreement with rhinomanometry and acoustic rhinometry data from the literature. A key innovation of this study are the normative ranges of flow variables associated with mucosal cooling, which recent research suggests is the primary physiological mechanism of nasal airflow sensation. Unilateral HF ranged from 94 to 281 W/m2, while the surface area stimulated by cooling ranged from 27.4 to 64.3 cm2. CONCLUSIONS: These normative ranges may serve as targets in future virtual surgery planning for patients with NAO.

Nonparametric competing risks analysis using Bayesian Additive Regression Trees.

Many time-to-event studies are complicated by the presence of competing risks. Such data are often analyzed using Cox models for the cause-specific hazard function or Fine and Gray models for the subdistribution hazard. In practice, regression relationships in competing risks data are often complex and may include nonlinear functions of covariates, interactions, high-dimensional parameter spaces and nonproportional cause-specific, or subdistribution, hazards. Model misspecification can lead to poor predictive performance. To address these issues, we propose a novel approach: flexible prediction modeling of competing risks data using Bayesian Additive Regression Trees (BART). We study the simulation performance in two-sample scenarios as well as a complex regression setting, and benchmark its performance against standard regression techniques as well as random survival forests. We illustrate the use of the proposed method on a recently published study of patients undergoing hematopoietic stem cell transplantation.

Factors associated with chronic opioid use after cervical spine surgery for degenerative conditions.

OBJECTIVE: Opioids are commonly prescribed after surgery for painful spinal conditions, yet little is known about postoperative opioid use. The relationship between chronic opioid use and patient-reported outcomes and satisfaction with surgery is also unclear. The purpose of this study was to evaluate factors associated with opioid use 1 year after elective cervical spine surgery for degenerative conditions causing radiculopathy and myelopathy. The authors hypothesized that patients with preoperative opioid use would be more likely to report postoperative opioid use at 1 year, and that postoperative opioid use would be associated with patient-reported outcomes and dissatisfaction with surgery. METHODS: The authors performed a retrospective study of a prospective cohort of adult patients who underwent elective cervical spine surgery for degenerative changes causing radiculopathy or myelopathy. Patients were prospectively and consecutively enrolled from a single academic center after the decision for surgery had been made. Postoperative in-hospital pain management was conducted using a standardized protocol. The primary outcome was any opioid use 1 year after surgery. Secondary outcomes were the Neck Disability Index (NDI); 36-Item Short-Form Health Survey (SF-36) physical function (PF), bodily pain (BP), and mental component summary (MCS) scores; the modified Japanese Orthopaedic Association (mJOA) score among myelopathy patients; and patient expectations surveys. Patients with and without preoperative opioid use were compared using the chi-square and Student t-tests, and multiple logistic regression was used to study the associations between patient and surgical characteristics and postoperative opioid use 1 year after surgery. RESULTS: Two hundred eleven patients were prospectively and consecutively enrolled, of whom 39 were lost to follow-up for the primary outcome; 43.6% reported preoperative opioid use. Preoperative NDI and SF-36 PF and BP scores were significantly worse in the preoperative opioid cohort. More than 94% of both cohorts rated expectations of pain relief as extremely or somewhat important. At 1 year after surgery, 50.7% of the preoperative-opioid-use cohort reported ongoing opioid use, and 17.5% of patients in the no-preoperative-opioid-use cohort reported ongoing opioid use. Despite this, both cohorts reported similar improvements in NDI as well as SF-36 PF, BP, and MCS scores. More than 70% of both cohorts also reported being extremely or somewhat satisfied with pain relief after surgery. Predictors of 1-year opioid use included preoperative opioid use, duration of symptoms for more than 9 months before surgery, tobacco use, and higher comorbidity index. CONCLUSIONS: One year after elective cervical spine surgery, patients with preoperative opioid use were significantly more likely to report ongoing opioid use. However, patients in both groups reported similar improvements in patient-reported outcomes and satisfaction with pain relief. Interventions targeted at decreasing opioid use may need to focus on patient factors such as preoperative opioid use or duration of symptoms before surgery.

Racial disparities of liver cancer mortality in Wisconsin.

PURPOSE: To calculate tract-level estimates of liver cancer mortality in Wisconsin and identify relationships with racial and socioeconomic variables. METHODS: County-level standardized mortality ratios (SMRs) of liver cancer in Wisconsin were calculated using traditional indirect adjustment methods for cases from 2003 to 2012. Tract-level SMRs were calculated using adaptive spatial filtering (ASF). The tract-level SMRs were checked for correlations to a socioeconomic advantage index (SEA) and percent racial composition. Non-spatial and spatial regression analyses with tract-level SMR as the outcome were conducted. RESULTS: County-level SMR estimates were shown to mask much of the variance within counties across their tracts. Liver cancer mortality was strongly correlated with the percent of Black residents in a census tract and moderately associated with SEA. In the multivariate spatially-adjusted regression analysis, only Percent Black composition remained significantly associated with an increased liver cancer SMR. CONCLUSIONS: Using ASF, we developed a high-resolution map of liver cancer mortality in Wisconsin. This map provided details on the distribution of liver cancer that were inaccessible in the county-level map. These tract-level estimates were associated with several racial and socioeconomic variables.

The association of pharmacy fill synchronization with breast cancer endocrine therapy adherence.

BACKGROUND: One-third to one-half of patients prescribed adjuvant endocrine therapy are nonadherent during the recommended 5-year endocrine therapy course. This study investigated whether poor pharmacy synchronization of medication fills (requiring refills on different days) acts as a barrier to adherence. METHODS: A cohort of older women with stage 0 to III endocrine receptor-positive breast cancer in 2011 was identified from the Surveillance, Epidemiology, and End Result-Medicare claims-linked cancer registry. Women with endocrine therapy and at least 1 other medication fill were identified, and the 3-month synchronization of their fills was calculated as 1 minus the quotient of the number of pharmacy visits and the number of filled medications. Regression models were used to examine the association between synchronization (in quartiles adjusted for the number of medications) and adherence to endocrine therapy (defined as a medication possession ratio ≥80%) over the subsequent year. RESULTS: During the 3 months after the first endocrine therapy prescription, the study cohort of 3212 women had a mean of 8.6 pharmacy visits (standard deviation, 4.7) with a mean synchronization of 0.3 (standard deviation, 0.2). Those in the third (odds ratio, 1.29; 95% confidence interval, 1.04-1.59) and fourth (most) medication number-adjusted synchronization quartiles (odds ratio, 1.49; 95% confidence interval, 1.19-1.86) were more likely to be adherent than those in the least. Multivariate model predictions showed that the proportion of patients who were adherent over 1 year varied from 68.9% in the least synchronized quartile to 76.6% in the most synchronized one. CONCLUSIONS: Prescription refill synchronization is strongly associated with adherence to endocrine therapy. Efforts to improve adherence should address this.

Housing discrimination and racial cancer disparities among the 100 largest US metropolitan areas.

BACKGROUND: Cancer contributes substantially to the life expectancy gap between US blacks and whites, and racial cancer disparities remain stubborn to eradicate. Disparities vary geographically, suggesting that they are not inevitable. METHODS: The authors examined the relationship between housing discrimination and the size of cancer disparities across large US metropolitan statistical areas (MSAs). MSA-level cancer disparities were measured using data from the US Centers for Disease Control and Prevention. Mortgage discrimination for each MSA was estimated using the Home Mortgage Disclosure Act database, and MSA racial segregation was determined using US Census data. Patterns of housing discrimination and cancer disparities were mapped, and the associations between these place-based factors and cancer disparities across MSAs were measured. RESULTS: Black-to-white cancer mortality disparities (rate ratios) varied geographically, ranging from 1.50 to 0.86; 88% of mortality ratios were >1, indicating higher mortality for blacks. In areas with greater mortgage discrimination, the gap between black and white cancer mortality rates was larger (correlation coefficient [r] = 0.32; P = .001). This relationship persisted in sex-specific analyses (males, r = 0.37; P < .001; females, r = 0.23; P = .02) and in models controlling for confounders. In contrast, segregation was inconsistently associated with disparities. Adjusting for incidence disparities attenuated, but did not eliminate, the correlation between mortgage discrimination and mortality disparities (r = 0.22-0.24), suggesting that cancer incidence and survival each account for part of the mortality disparity. CONCLUSIONS: Mortgage discrimination is associated with larger black-to-white cancer mortality disparities. Some areas are exceptions to this trend. Examination of these exceptions and of policies related to housing discrimination may offer novel strategies for explaining and eliminating cancer disparities.

Factors Influencing Prescription Drug Synchronization: The Complex Role of Number of Medications.

BACKGROUND: Despite the well-documented association of medication refill synchronization with medication adherence, little is known about how best to measure synchronization at pharmacy visits or about its relationship to number of medications. OBJECTIVE: To examine the relationship of a commonly cited synchronization measure with the number of prescription medications. METHODS: Using a cohort of women aged 66-90 years with stage 0-3 hormone receptor-positive breast cancer from the Surveillance, Epidemiology and End Result (SEER)-Medicare data, we identified women with pharmacy claims for at least 1 endocrine therapy prescription and at least 1 other medication fill. Twelve-month medication refill synchronization was calculated as the quotient of the number of pharmacy visits and the number of filled medications subtracted from 1. Multiple linear regression (including polynomials) was then used to assess the relationship between refill synchronization, number of medications, and other potentially influential factors. RESULTS: Over 47% of cohort subjects took more than 10 unique medications. Subjects made an average (SD) of 29.9 (18.0) pharmacy visits, resulting in a mean (SD) synchronization of 0.28 (0.18, range = 0.0-0.92). The number of medications, including powers through to the fourth, was strongly associated with refill synchronization, with a rapid initial rise followed by a gradual increase after 10 medications. Although patient age and race/ethnicity were not associated with synchronization, there was a significant positive association of receipt of a low-income subsidy and residence in rural areas with synchronization. CONCLUSIONS: There is a complex relationship between refill synchronization and number of prescribed medications, and future research into synchronization should account for this. DISCLOSURES: This study was supported by the National Institute on Minority Health and Health Disparities under grant R01 MD010728. The authors have nothing to disclose. This study was presented as an oral abstract at the Society of General Internal Medicine Meeting; April 13, 2018; Denver, CO.

Decision making and uncertainty quantification for individualized treatments using Bayesian Additive Regression Trees.

Individualized treatment rules can improve health outcomes by recognizing that patients may respond differently to treatment and assigning therapy with the most desirable predicted outcome for each individual. Flexible and efficient prediction models are desired as a basis for such individualized treatment rules to handle potentially complex interactions between patient factors and treatment. Modern Bayesian semiparametric and nonparametric regression models provide an attractive avenue in this regard as these allow natural posterior uncertainty quantification of patient specific treatment decisions as well as the population wide value of the prediction-based individualized treatment rule. In addition, via the use of such models, inference is also available for the value of the optimal individualized treatment rules. We propose such an approach and implement it using Bayesian Additive Regression Trees as this model has been shown to perform well in fitting nonparametric regression functions to continuous and binary responses, even with many covariates. It is also computationally efficient for use in practice. With Bayesian Additive Regression Trees, we investigate a treatment strategy which utilizes individualized predictions of patient outcomes from Bayesian Additive Regression Trees models. Posterior distributions of patient outcomes under each treatment are used to assign the treatment that maximizes the expected posterior utility. We also describe how to approximate such a treatment policy with a clinically interpretable individualized treatment rule, and quantify its expected outcome. The proposed method performs very well in extensive simulation studies in comparison with several existing methods. We illustrate the usage of the proposed method to identify an individualized choice of conditioning regimen for patients undergoing hematopoietic cell transplantation and quantify the value of this method of choice in relation to the optimal individualized treatment rule as well as non-individualized treatment strategies.