How Much Does the US Public Value Equity in Health? A Systematic Review.

OBJECTIVES: This systematic review aims to summarize and qualitatively assess published evaluations on the US public's preferences for health equity and their willingness to trade-off efficiency for equity. METHODS: Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses literature search extension guidelines, we searched MEDLINE and Embase for relevant peer-reviewed publications on this topic before February 2021. We included English-language articles that solicited US preferences regarding efficiency-equity trade-offs and prioritizing healthcare resources based on socioeconomic status, race, disability, or burden of disease. Quantitative and qualitative data captured were decided a priori and iteratively adapted as themes emerged. RESULTS: Fourteen studies were found over a 25-year span. Only 4 focused on resource allocation across social groups. Three distinct notions of fairness were studied: equal distribution of resources, priority to the worse-off, and equal health achieved. We found modest support for equal distribution of resources and willingness to sacrifice efficiency for equity in the United States. Prioritizing the underserved was relatively less studied and received less support and was more preferred when resources were scarce, when allocating resources between social groups, or when participants were informed about the fundamental origins of health inequities. Equal health was the least studied, but received nontrivial support. CONCLUSIONS: The existing literature evaluating the US public's understanding and preferences toward equity was severely limited by the lack of rigorous quantitative studies and heterogeneous attribute selection and fairness definitions. High-quality studies that clearly define fairness, focus on social groups, and apply rigorous methods to quantify equity preferences are needed to integrate the public's value on equity into healthcare decisions.

A Scoping Review of the Use of Machine Learning in Health Economics and Outcomes Research: Part 1-Data From Wearable Devices.

OBJECTIVES: With the emerging use of machine learning (ML) techniques, there has been particular interest in using wearable data for health economics and outcomes research (HEOR). We aimed to understand the emerging patterns of how ML has been applied to wearable data in HEOR. METHODS: We identified studies published in PubMed between January 2016 and March 2021. Studies that included at least 1 HEOR-related Medical Subject Headings term, applied an ML, and used wearable data were eligible for inclusion. Two reviewers abstracted information including ML application types and data on which ML was applied and analyzed them using descriptive analyses. RESULTS: A total of 148 studies were identified from PubMed, among which 32 studies met the inclusion criteria. There has been an increase over time in the number of ML studies using wearable data. ML has been more frequently used for monitoring events in real time (78%) than to predict future events (22%). There has been a wide range of outcomes examined, ranging from general physical or mental health (24%) to more disease-specific outcomes (eg, disease incidence [19%] and progression [13%]) and treatment-related outcomes (eg, treatment adherence [9%] and outcomes [9%]). Data for ML models were more often derived from wearable devices with specific medical purposes (60%) than those without (40%). CONCLUSION: There has been a wide range of applications of ML to wearable data. Both medical and nonmedical wearable devices have been used as a data source, showing the potential for providing rich data for ML studies in HEOR.

Association Between a Population Health Intervention and Hypertension Control.

INTRODUCTION: As part of the Centers for Medicare and Medicaid Innovation Practice Transformation Network, an integrated healthcare system implemented a multimodal, population health-based hypertension clinical pathway program (HCPP) focused on hypertension management. AIM: To determine whether the HCPP was associated with changes in hypertension control or process-of-care measures and whether associations varied for sites serving higher versus lower proportions of historically underserved patients. SETTING: An integrated academic health system encompassing 5 clinic networks and 85 primary and specialty care sites. PROGRAM DESCRIPTION: The HCPP was implemented at some sites (adopters) but not others (non-adopters) and had four components: (1) stakeholder engagement; (2) clinical staff retraining; (3) electronic health record-based prompts; and (4) performance monitoring and feedback. Program goals were to encourage clinical teams to increase the frequency of follow up visits and adopt standardized approaches to blood pressure (BP) measurements and antihypertensive medication regimen advancement defined as adding or titrating existing medication. PROGRAM EVALUATION: This quasi-experimental study used 2017-2019 data from 63,497 patients with hypertension and multivariable difference-in-differences analyses to evaluate changes in outcomes at 19 adopter versus 39 non-adopter sites before and after HCPP implementation. Adoption was associated with 3.5 times differentially greater odds of a BP reassessment (OR 3.5, 95% CI 3.3-3.8), 11% differentially greater odds of BP control (BP<140/90 mmHg) (OR 1.11, 95% CI 1.07-1.15), and 12% differentially greater odds of having non-severely elevated BP (systolic BP < 155 mmHg) (OR 1.12, 95% CI 1.05-1.19). HCPP adoption was not associated with differential changes in 90-day follow-up BP measurement. Adoption was associated with 23% differentially greater odds of appropriate medication advancement (OR 1.23, 95% CI 1.04-1.46). A similar pattern was observed when limiting comparisons to sites caring for a higher proportion of historically underserved populations. DISCUSSION: A multimodal population health approach to transforming hypertension care was associated with improved BP outcomes.

A Scoping Review of the Use of Machine Learning in Health Economics and Outcomes Research: Part 2-Data From Nonwearables.

OBJECTIVES: Despite the increasing interest in applying machine learning (ML) methods in health economics and outcomes research (HEOR), stakeholders face uncertainties in when and how ML can be used. We reviewed the recent applications of ML in HEOR. METHODS: We searched PubMed for studies published between January 2020 and March 2021 and randomly chose 20% of the identified studies for the sake of manageability. Studies that were in HEOR and applied an ML technique were included. Studies related to wearable devices were excluded. We abstracted information on the ML applications, data types, and ML methods and analyzed it using descriptive statistics. RESULTS: We retrieved 805 articles, of which 161 (20%) were randomly chosen. Ninety-two of the random sample met the eligibility criteria. We found that ML was primarily used for predicting future events (86%) rather than current events (14%). The most common response variables were clinical events or disease incidence (42%) and treatment outcomes (22%). ML was less used to predict economic outcomes such as health resource utilization (16%) or costs (3%). Although electronic medical records (35%) were frequently used for model development, claims data were used less frequently (9%). Tree-based methods (eg, random forests and boosting) were the most commonly used ML methods (31%). CONCLUSIONS: The use of ML techniques in HEOR is growing rapidly, but there remain opportunities to apply them to predict economic outcomes, especially using claims databases, which could inform the development of cost-effectiveness models.

Cause-specific mortality of type B aortic dissection and assessment of competing risks of mortality.

OBJECTIVE: Natural history studies of type B aortic dissection (TBAD) commonly report all-cause mortality. Our aim was to determine cause-specific mortality in TBAD and to evaluate the clinical characteristics associated with aorta-related and nonaorta-related mortality. METHODS: Clinical and administrative records were reviewed for patients with acute TBAD between 1995 and 2017. Demographics, comorbidities, presentation, and initial imaging findings were abstracted. Cause of death was ascertained through a multimodality approach using electronic health records, obituaries, social media, Social Security Death Index, and state mortality records. Causes of death were classified as aorta related, nonaorta related, or unknown. A Fine-Gray multivariate competing risk regression model for subdistribution hazard ratio was employed to analyze the association of clinical characteristics with aorta-related and nonaorta-related mortality. RESULTS: A total of 275 individuals met inclusion criteria (61.1 ± 13.7 years, 70.9% male, 68% white). Mean survival after discharge was 6.3 ± 4.7 years. Completeness of follow-up Clark C index was 0.87. All-cause mortality was 50.2% (n = 138; mean age, 70.1 ± 14.6 years) including an in-hospital mortality of 8.4%. Cause-specific mortality was aorta related, nonaorta related, and unknown in 51%, 43%, and 6%, respectively. Compared with patients with nonaorta-related mortality, patients with aorta-related mortality were younger at acute TBAD (69.5 ± 11.2 years vs 61.6 ± 15.5 years; P = .001), underwent more descending thoracic aortic repairs (19.4% vs 45.8%; P = .002), and had a shorter survival duration (5.7 ± 3.9 vs 3.4 ± 4.5 years; P = .002). There was clear variation in cause of death by each decade of life, with higher aorta-related mortality among those younger than 50 years and older than 70 years and a stepwise increase in nonaorta-related mortality with each increasing decade (P < .001). All-cause mortality at 1 year, 3 years, and 10 years was 15%, 24%, and 57%, respectively. After accounting for competing risks, the cumulative incidence of aorta-related mortality at 1 year, 3 years, and 10 years was 8.9%, 16.5%, and 27.2%, respectively, and that of nonaorta-related mortality was 2.7%, 7.2%, and 29%, respectively. A maximum descending thoracic aortic diameter >4 cm was associated with an increase in hazard of aorta-related mortality by 84% (subdistribution hazard ratio, 1.84; 95% confidence interval, 1.03-3.28) on multivariate competing risk regression analysis. CONCLUSIONS: TBAD is associated with high 10-year mortality. Those at risk for aorta-related mortality have a clinical phenotype different from that of individuals at risk for nonaorta-related mortality. This information is important for building risk prediction models that account for competing mortality risks and to direct optimal and individualized surgical and medical management of TBAD.

Transforming Population-Based Depression Care: a Quality Improvement Initiative Using Remote, Centralized Care Management.

INTRODUCTION: With the growing prevalence of value-based contracts, health systems are incentivized to consider population approaches to service delivery, particularly for chronic conditions like depression. To this end, UW Medicine implemented the Depression-Population Approach to Health (PATH) program in primary care (PC) as part of a system-wide Center for Medicare and Medicaid Innovation (CMMI) quality improvement (QI) initiative. AIM: To examine the feasibility of a pilot PATH program and its impact on clinical and process-of-care outcomes. SETTING: A large, diverse, geographically disparate academic health system in Western Washington State including 28 PC clinics across five networks. PROGRAM DESCRIPTION: The PATH program was a population-level, centralized, measurement-based care intervention that utilized a clinician to provide remote monitoring of treatment progress via chart review and facilitate patient engagement when appropriate. The primary goals of the program were to improve care engagement and increase follow-up PHQ-9 assessments for patients with depression and elevated initial PHQ-9 scores. PROGRAM EVALUATION: We employed a prospective, observational study design, including commercially insured adult patients with new depression diagnoses and elevated initial PHQ-9 scores. The pilot intervention group, consisting of accountable care network (ACN) self-enrollees (N = 262), was compared with a similar commercially insured cohort (N = 2527) using difference-in-differences analyses adjusted for patient comorbidities, initial PHQ-9 score, and time trends. The PATH program was associated with three times the odds of PHQ-9 follow-up (OR 3.28, 95% CI 1.79-5.99), twice the odds of a follow-up PC clinic visit (OR 1.74, 95% CI 0.99-3.08), and twice the odds of treatment response, defined as reduction in PHQ-9 score by ≥ 50% (OR 2.02, 95% CI 0.97-4.21). DISCUSSION: Our results demonstrate that a centralized, remote care management initiative is both feasible and effective for large academic health systems aiming to improve depression outcome ascertainment, treatment engagement, and clinical care.

Establishing Clinically Significant Patient-reported Outcomes for Diverticular Disease.

BACKGROUND: Diverticular disease can undermine health-related quality of life. The diverticulitis quality of life (DV-QOL) instrument was designed and validated to measure patient-reported burden of diverticular disease. However, values reflecting meaningful improvement (i.e., minimal clinically important difference [MCID]) and the patient acceptable symptom state (PASS) have yet to be established. We sought to establish the MCID and PASS of the DV-QOL and describe the characteristics of those with DV-QOL above the PASS threshold. MATERIALS AND METHODS: We performed a prospective cohort study of adults with diverticular disease from seven centers in Washington and California (2016-2018). Patients were surveyed at baseline, then quarterly up to 30 mo. To determine the MCID and PASS for DV-QOL, we applied various previously established distribution- and anchor-based approaches and compared the resulting values. RESULTS: The study included 177 patients (mean age 57 y, 43% women). A PASS threshold of 3.2/10 distinguished between those with and without health-related quality of life-impacting diverticulitis with acceptable accuracy (area under the curve 0.76). A change of 2.2 points in the DV-QOL was the most appropriate MCID: above the distribution-based MCIDs and corresponding to patient perception of importance of change (AUC 0.70). Patients with DV-QOL ≥ PASS were more often men, younger, had Medicaid, had more serious episodes of diverticulitis, and had an occupational degree or high-school education or less. CONCLUSIONS: Our study is the first to define MCID and PASS for DV-QOL. These thresholds are critical for measuring the impact of diverticular disease and the evaluation of treatment effectiveness.

Comparison of outcomes in women and men following carotid interventions in the Washington state's Vascular Interventional Surgical Care and Outcomes Assessment Program.

BACKGROUND: The benefit for carotid endarterectomy (CEA) to prevent a potential stroke has been shown to be less beneficial for women compared with men and the risk of carotid stenting (CAS) is higher in women than men. We hypothesized that a community-based Washington state registry data would also reveal increased morbidity and mortality for women undergoing carotid interventions. METHODS: Deidentified data for CEA and CAS between 2010 and 2015 were obtained from 19 hospitals participating in the Washington State Vascular-Interventional Surgical Care and Outcomes Assessment Program. Data analysis compared in-hospital composite outcome of stroke and mortality from CEA and CAS between women and men. RESULTS: Over the study period, 3704 individuals underwent CEA (n = 2759; 49.5% symptomatic) and CAS (n = 945; 60.9% symptomatic). Women accounted for 39.5% of the cohort. Women were slightly younger than men (70.0 ± 10.2 vs 71.0 ± 9.6 years respectively; P < .01), less likely to be smokers (70.1% vs 75.6%; P < .01), and less likely to have a diagnosis of coronary artery disease (32.9% vs 46.5%; P < .01). Fewer women underwent CEA for symptomatic carotid disease (46.1% vs 51.8%; P < .01). There were no statistically significant differences in the postoperative in-hospital stroke and mortality among women and men undergoing CEA (asymptomatic, 0.8% vs 1.4% [P = .36]; symptomatic, 1.8% vs 2.2% [P = .58]) and CAS (asymptomatic, 1.4% vs 2.2% [P = .56]; symptomatic, 4.6% vs 2.5% [P = .18]). Hospital duration of stay and discharge disposition were similar for women and men. A subanalysis of the octogenarian cohort undergoing CAS demonstrated a substantial increase in-hospital stroke and mortality among women and men (11.6% [CAS] vs 2.2% [CEA]; P = .024). CONCLUSIONS: In the Washington state Vascular-Interventional Surgical Care and Outcomes Assessment Program registry, hospital composite outcome of stroke and mortality following carotid interventions from 2010 to 2015 were noted to be similar for women and men. The notable exception to this finding was observed in subcohort of women undergoing CAS for symptomatic carotid disease at age 80 years or older. These findings should be taken into account when risk stratifying patients for carotid interventions.

Role of bleeding recognition and evaluation in Black-White disparities in endometrial cancer.

BACKGROUND: Advanced stage at diagnosis is an independent, unexplained contributor to racial disparity in endometrial cancer. OBJECTIVE: We sought to investigate whether, prior to diagnosis, provider recognition of the cardinal symptom of endometrial cancer, postmenopausal bleeding, differs by patient race. STUDY DESIGN: Black and White women diagnosed with endometrial cancer (2001 through 2011) from Surveillance, Epidemiology, and End Results-Medicare who had at least 2 years of claims prior to diagnosis were identified. Bleeding diagnoses along with procedures done prior to diagnosis were captured via claims data. Multinomial logistic regression was used to evaluate the association of race with diagnostic workup and multivariate models built to determine the association of appropriate diagnostic procedures with stage at diagnosis. RESULTS: In all, 4354 White and 537 Black women diagnosed with endometrial cancer were included. Compared to White women, Black women were less likely to have guideline-concordant care: postmenopausal bleeding and appropriate diagnostic evaluation (70% vs 79%, P < .001), with adjusted relative risk ratios of 1.12-1.73 for different nonguideline-concordant pathways: bleeding without diagnostic procedures, alternative bleeding descriptions, and neither bleeding nor procedures. These pathways were associated with higher odds of advanced stage at diagnosis (adjusted odds ratio, 1.90-2.88). CONCLUSION: The lack of recognition and evaluation of postmenopausal bleeding is associated with advanced stage at diagnosis in endometrial cancer. Older Black women are at highest risk for the most aggressive histology types, yet they are less likely to have guideline-concordant evaluation of vaginal bleeding. Efforts aimed at improving recognition-among patients and providers-of postmenopausal bleeding in Black women could substantially reduce disparities in endometrial cancer.

Real world costs and cost-effectiveness of Rituximab for diffuse large B-cell lymphoma patients: a population-based analysis.

BACKGROUND: Current treatment of diffuse-large-B-cell lymphoma (DLBCL) includes rituximab, an expensive drug, combined with cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP) chemotherapy. Economic models have predicted rituximab plus CHOP (RCHOP) to be a cost-effective alternative to CHOP alone as first-line treatment of DLBCL, but it remains unclear what its real-world costs and cost-effectiveness are in routine clinical practice. METHODS: We performed a population-based retrospective cohort study from 1997 to 2007, using linked administrative databases in Ontario, Canada, to evaluate the costs and cost-effectiveness of RCHOP compared to CHOP alone. A historical control cohort (n = 1,099) with DLBCL who received CHOP before rituximab approval was hard-matched on age and treatment intensity and then propensity-score matched on sex, comorbidity, and histology to 1,099 RCHOP patients. All costs and outcomes were adjusted for censoring using the inverse probability weighting method. The main outcome measure was incremental cost per life-year gained (LYG). RESULTS: Rituximab was associated with a life expectancy increase of 3.2 months over 5 years at an additional cost of $16,298, corresponding to an incremental cost-effectiveness ratio of $61,984 (95% CI $34,087-$135,890) per LYG. The probability of being cost-effective was 90% if the willingness-to-pay threshold was $100,000/LYG. The cost-effectiveness ratio was most favourable for patients less than 60 years old ($31,800/LYG) but increased to $80,600/LYG for patients 60-79 years old and $110,100/LYG for patients ≥ 80 years old. We found that post-market survival benefits of rituximab are similar to or lower than those reported in clinical trials, while the costs, incremental costs and cost-effectiveness ratios are higher than in published economic models and differ by age. CONCLUSIONS: Our results showed that the addition of rituximab to standard CHOP chemotherapy was associated with improvement in survival but at a higher cost, and was potentially cost-effective by standard thresholds for patients <60 years old. However, cost-effectiveness decreased significantly with age, suggesting that rituximab may be not as economically attractive in the very elderly on average. This has important clinical implications regarding age-related use and funding decisions on this drug.

Evaluation of a Multilevel Laboratory Stewardship Intervention Targeted to Cardiac and Thoracic Surgical Services at an Academic Medical Center.

CONTEXT.­: Unnecessary laboratory tests are ordered because of factors such as preselected orders on order sets, clinician habits, and trainee concerns. Excessive use of laboratory testing increases patient discomfort via unnecessary phlebotomy, contributes to iatrogenic anemia, increases risk of bloodstream infections, and increases the cost of care. OBJECTIVE.­: To address these concerns, we implemented a multilevel laboratory stewardship intervention to decrease unnecessary laboratory testing, measured by laboratory tests per day attributed to service, across 2 surgical divisions with high laboratory use. DESIGN.­: The multilevel intervention included 5 components: stakeholder engagement, provider education, computerized provider order entry modification, performance feedback, and culture change supported by leadership. The primary outcome of the study was laboratory tests ordered per patient-day. Secondary outcomes included the number of blood draws per patient-day, total lab-associated costs, length of stay, discharge to a nursing facility, 30-day readmissions, and deaths. A difference-in-differences analytic approach assessed the outcome measures in the intervention period, with other surgical services as controls. RESULTS.­: The primary outcome of laboratory tests per patient-day showed a significant decrease across both thoracic and cardiac surgery services, with between 1.5 and 2 fewer tests ordered per patient-day for both services and an estimated 20 000 fewer tests performed during the intervention period. Blood draws per patient-day were also significantly decreased on the thoracic surgery service but not for cardiac surgery. CONCLUSIONS.­: A multilevel laboratory stewardship intervention targeted to 2 surgical services resulted in a significant decrease in laboratory test use without negatively impacting length of stay, readmissions, or mortality.

Racial and Ethnic Bias in Risk Prediction Models for Colorectal Cancer Recurrence When Race and Ethnicity Are Omitted as Predictors.

Importance: Including race and ethnicity as a predictor in clinical risk prediction algorithms has received increased scrutiny, but there continues to be a lack of empirical studies addressing whether simply omitting race and ethnicity from the algorithms will ultimately affect decision-making for patients of minoritized racial and ethnic groups. Objective: To examine whether including race and ethnicity as a predictor in a colorectal cancer recurrence risk algorithm is associated with racial bias, defined as racial and ethnic differences in model accuracy that could potentially lead to unequal treatment. Design, Setting, and Participants: This retrospective prognostic study was conducted using data from a large integrated health care system in Southern California for patients with colorectal cancer who received primary treatment between 2008 and 2013 and follow-up until December 31, 2018. Data were analyzed from January 2021 to June 2022. Main Outcomes and Measures: Four Cox proportional hazards regression prediction models were fitted to predict time from surveillance start to cancer recurrence: (1) a race-neutral model that explicitly excluded race and ethnicity as a predictor, (2) a race-sensitive model that included race and ethnicity, (3) a model with 2-way interactions between clinical predictors and race and ethnicity, and (4) separate models by race and ethnicity. Algorithmic fairness was assessed using model calibration, discriminative ability, false-positive and false-negative rates, positive predictive value (PPV), and negative predictive value (NPV). Results: The study cohort included 4230 patients (mean [SD] age, 65.3 [12.5] years; 2034 [48.1%] female; 490 [11.6%] Asian, Hawaiian, or Pacific Islander; 554 [13.1%] Black or African American; 937 [22.1%] Hispanic; and 2249 [53.1%] non-Hispanic White). The race-neutral model had worse calibration, NPV, and false-negative rates among racial and ethnic minority subgroups than non-Hispanic White individuals (eg, false-negative rate for Hispanic patients: 12.0% [95% CI, 6.0%-18.6%]; for non-Hispanic White patients: 3.1% [95% CI, 0.8%-6.2%]). Adding race and ethnicity as a predictor improved algorithmic fairness in calibration slope, discriminative ability, PPV, and false-negative rates (eg, false-negative rate for Hispanic patients: 9.2% [95% CI, 3.9%-14.9%]; for non-Hispanic White patients: 7.9% [95% CI, 4.3%-11.9%]). Inclusion of race interaction terms or using race-stratified models did not improve model fairness, likely due to small sample sizes in subgroups. Conclusions and Relevance: In this prognostic study of the racial bias in a cancer recurrence risk algorithm, removing race and ethnicity as a predictor worsened algorithmic fairness in multiple measures, which could lead to inappropriate care recommendations for patients who belong to minoritized racial and ethnic groups. Clinical algorithm development should include evaluation of fairness criteria to understand the potential consequences of removing race and ethnicity for health inequities.

Factors associated with lung cancer risk factor documentation.

Objectives: To identify factors associated with the minimum necessary information to determine an individual's eligibility for lung cancer screening (ie, sufficient risk factor documentation) and to characterize clinic-level variability in documentation. Study Design: Cross-sectional observational study using electronic health record data from an academic health system in 2019. Methods: We calculated the relative risk of sufficient lung cancer risk factor documentation by patient-, provider-, and system-level variables using Poisson regression models, clustering by clinic. We compared unadjusted, risk-adjusted, and reliability-adjusted proportions of patients with sufficient smoking documentation across 31 clinics using logistic regression models and 2-level hierarchical logit models to estimate reliability-adjusted proportions across clinics. Results: Among 20,632 individuals, 60% had sufficient risk factor documentation to determine screening eligibility. Patient-level factors inversely associated with risk factor documentation included Black race (relative risk [RR], 0.70; 95% CI, 0.60-0.81), non-English preferred language (RR, 0.60; 95% CI, 0.49-0.74), Medicaid insurance (RR, 0.64; 95% CI, 0.57-0.71), and nonactivated patient portal (RR, 0.85; 95% CI, 0.80-0.90). Documentation varied across clinics. The reliability-adjusted intraclass correlation coefficient decreased from 11.0% (95% CI, 6.9%-17.1%) to 5.3% (95% CI, 3.2%-8.6%), adjusting for covariates. Conclusions: We found a low rate of sufficient lung cancer risk factor documentation and associations of risk factor documentation based on patient-level factors such as race, insurance status, language, and patient portal activation. Risk factor documentation rates varied across clinics, and only approximately half the variation was explained by factors in our analysis.

Using Machine Learning to Leverage Biomarker Change and Predict Colorectal Cancer Recurrence.

PURPOSE: The risk of colorectal cancer (CRC) recurrence after primary treatment varies across individuals and over time. Using patients' most up-to-date information, including carcinoembryonic antigen (CEA) biomarker profiles, to predict risk could improve personalized decision making. METHODS: We used electronic health record data from an integrated health system on a cohort of patients diagnosed with American Joint Committee on Cancer stage I-III CRC between 2008 and 2013 (N = 3,970) and monitored until recurrence or end of follow-up. We addressed missingness in recurrence outcomes and longitudinal CEA measures, and engineered CEA features using current and past biomarker values for inclusion in a risk prediction model. We used a discrete time Superlearner model to evaluate various algorithms for predicting recurrence. We evaluated the time-varying discrimination and calibration of the algorithms and assessed the role of individual predictors. RESULTS: Recurrence was documented in 448 (11.3%) patients. XGBoost with depth = 1 (XGB-D1) predicted recurrence substantially better than all other algorithms at all time points, with AUC ranging from 0.87 (95% CI, 0.86 to 0.88) at 6 months to 0.94 (95% CI, 0.92 to 0.96) at 54 months. The only variable used by XGB-D1 was 6-month change in log CEA. Predicted 1-year risk of recurrence was nearly zero for patients whose log CEA did not increase in the last 6 months, between 12.2% and 34.1% for patients whose log CEA increased between 0.10 and 0.40, and 43.6% for those with a log CEA increase >0.40. Compared with XGB, penalized regression approaches (lasso, ridge, and elastic net) performed poorly, with AUCs ranging from 0.58 to 0.69. CONCLUSION: A flexible, machine learning approach that incorporated longitudinal CEA information yielded a simple and high-performing model for predicting recurrence on the basis of 6-month change in log CEA.

Racial Disparities in the Ascertainment of Cancer Recurrence in Electronic Health Records.

PURPOSE: There is growing interest in using computable phenotypes or proxies to identify important clinical outcomes, such as cancer recurrence, in rich electronic health records data. However, the race/ethnicity-specific accuracies of these proxies remain unclear. We examined whether the accuracy of a proxy for colorectal cancer (CRC) recurrence differed by race/ethnicity and the possible mechanisms that drove the differences. METHODS: Using data from a large integrated health care system, we identified a stratified random sample of 282 Black/African American (AA), Hispanic, and non-Hispanic White (NHW) patients with CRC who received primary treatment. Patient 5-year recurrence status was estimated using a utilization-based proxy and evaluated against the true recurrence status obtained using detailed chart review and by race/ethnicity. We used covariate-adjusted probit regression models to estimate the associations between race/ethnicity and misclassification. RESULTS: The recurrence proxy had excellent overall accuracy (positive predictive value [PPV] 89.4%; negative predictive value 96.5%; mean difference in timing 1.96 months); however, accuracy varied by race/ethnicity. Compared with NHW patients, PPV was 14.9% lower (95% CI, 2.53 to 28.6) among Hispanic patients and 4.3% lower (95% CI, -4.8 to 14.8) among Black/AA patients. The proxy disproportionately inflated the 5-year recurrence incidence for Hispanic patients by 10.6% (95% CI, 4.2 to 18.2). Compared with NHW patients, proxy recurrences for Hispanic patients were almost three times as likely to have been misclassified as positive (adjusted risk ratio 2.91 [95% CI, 1.21 to 8.31]). Higher false positives among racial/ethnic minorities may be related to higher prevalence of noncancerous lung-related problems and substantial delays in primary treatment because of insufficient patient-provider communication and abnormal treatment patterns. CONCLUSION: Using a proxy with worse accuracy among racial/ethnic minority patients to estimate population health may misdirect resources and support erroneous conclusions around treatment benefit for these patients.

Impact of rituximab on treatment outcomes of patients with diffuse large b-cell lymphoma: a population-based analysis.

We conducted a multi-institutional population-based analysis of the survival and toxicity associated with the addition of rituximab to chemotherapy for patients with diffuse large B-cell lymphoma (DLBCL), including patients aged ≥ 80 years, who were excluded from published randomized trials. Using population-based registries in Ontario, we identified 4021 patients who received chemotherapy with or without rituximab (R-CHOP [rituximab with cyclophosphamide, doxorubicin, vincristine and prednisone] or CHOP) for DLBCL between 1996 and 2007, including 397 patients aged ≥ 80 years. After propensity score matching, the overall survival (OS) and significant toxicities for R-CHOP and CHOP treatment groups were compared. R-CHOP was associated with a significant increase in 5-year OS compared to CHOP alone (62% vs. 54%; hazard of death = 0·78, P = 0·0004). Survival benefit was seen in all age groups, including those aged ≥ 80 years. Patients treated with rituximab did not have a significant increase in 1-year hospitalization rates for cardiac, pulmonary, gastrointestinal or neurological diagnoses compared to those treated with CHOP alone. The addition of rituximab to CHOP improves survival in the general population of patients with DLBCL and produces early survival benefit for very elderly patients, without any significant increase in the risk of serious toxicity.

Risk of Adverse Financial Events in Patients With Cancer: Evidence From a Novel Linkage Between Cancer Registry and Credit Records.

PURPOSE: Although financial toxicity is a growing cancer survivorship issue, no studies have used credit data to estimate the relative risk of financial hardship in patients with cancer versus individuals without cancer. We conducted a population-based retrospective matched cohort study using credit reports to investigate the impact of a cancer diagnosis on the risk of adverse financial events (AFEs). METHODS: Western Washington SEER cancer registry (cases) and voter registry (controls) records from 2013 to 2018 were linked to quarterly credit records from TransUnion. Controls were age-, sex-, and zip code-matched to cancer cases and assigned an index date corresponding to the case's diagnosis date. Cases and controls experiencing past-due credit card payments and any of the following AFEs at 24 months from diagnosis or index were compared, using two-sample z tests: third-party collections, charge-offs, tax liens, delinquent mortgage payments, foreclosures, and repossessions. Multivariate logistic regression models were used to evaluate the association of cancer diagnosis with AFEs and past-due credit payments. RESULTS: A total of 190,722 individuals (63,574 cases and 127,148 controls, mean age 66 years) were included. AFEs (4.3% v 2.4%, P < .0001) and past-due credit payments (2.6% v 1.9%, P < .0001) were more common in cases than in controls. After adjusting for age, sex, average baseline credit line, area deprivation index, and index/diagnosis year, patients with cancer had a higher risk of AFEs (odds ratio 1.71; 95% CI, 1.61 to 1.81; P < .0001) and past-due credit payments (odds ratio 1.28; 95% CI, 1.19 to 1.37; P < .0001) than controls. CONCLUSION: Patients with cancer were at significantly increased risk of experiencing AFEs and past-due credit card payments relative to controls. Studies are needed to investigate the impact of these events on treatment decisions, quality of life, and clinical outcomes.

What's in a Number? Assessing the Burden of Diverticular Disease.

MINI-ABSTRACT: In this prospective observational cohort of patients with a history of diverticulitis, we assessed the correlation between the diverticulitis quality of life survey (DVQOL) and other patient-reported expressions of disease measures including work and activity impairment, and contentment with gastrointestinal-related health. Then, we assessed whether the DVQOL is better correlated with these measures than diverticulitis episode count. Our study results showed that the DVQOL has a stronger correlation with other disease measures than diverticulitis episode count, and our findings support the broader use of the DVQOL in assessing the burden of diverticulitis and monitoring response to management.