Search

Introduction: Previous work had shown that machine learning models can predict inflammatory bowel disease (IBD)-related hospitalizations and outpatient corticosteroid use based on patient demographic and laboratory data in a cohort of United States Veterans. This study aimed to replicate this modeling framework in a nationally representative cohort. Methods: A retrospective cohort design using Optum Electronic Health Records (EHR) were used to identify IBD patients, with at least 12 months of follow-up between 2007 and 2018. IBD flare was defined as an inpatient/emergency visit with a diagnosis of IBD or an outpatient corticosteroid prescription for IBD. Predictors included demographic and laboratory data. Logistic regression and random forest (RF) models were used to predict IBD flare within 6 months of each visit. A 70% training and 30% validation approach was used. Results: A total of 95,878 patients across 780,559 visits were identified. Of these, 22,245 (23.2%) patients had at least one IBD flare. Patients were predominantly White (87.7%) and female (57.1%), with a mean age of 48.0 years. The logistic regression model had an area under the receiver operating curve (AuROC) of 0.66 (95% CI: 0.65−0.66), sensitivity of 0.69 (95% CI: 0.68−0.70), and specificity of 0.74 (95% CI: 0.73−0.74) in the validation cohort. The RF model had an AuROC of 0.80 (95% CI: 0.80−0.81), sensitivity of 0.74 (95% CI: 0.73−0.74), and specificity of 0.72 (95% CI: 0.72−0.72) in the validation cohort. Important predictors of IBD flare in the RF model were the number of previous flares, age, potassium, and white blood cell count. Conclusion: The machine learning modeling framework was replicated and results showed a similar predictive accuracy in a nationally representative cohort of IBD patients. This modeling framework could be embedded in routine practice as a tool to distinguish high-risk patients for disease activity. [ABSTRACT FROM AUTHOR]

Objective: To examine which factors are driving improvement in the Dialysis Facility Compare (DFC) star ratings and to test whether nonclinical facility characteristics are associated with observed longitudinal changes in the star ratings.Data Sources: Data were collected from eligible patients in over 6,000 Medicare-certified dialysis facilities from three annual star rating and individual measure updates, publicly released on DFC in October 2015, October 2016, and April 2018.Study Design: Changes in the star rating and individual quality measures were investigated across three public data releases. Year-to-year changes in the star ratings were linked to facility characteristics, adjusting for baseline differences in quality measure performance.Data Collection: Data from publicly reported quality measures, including standardized mortality, hospitalization, and transfusion ratios, dialysis adequacy, type of vascular access for dialysis, and management of mineral and bone disease, were extracted from annual DFC data releases.Principal Findings: The proportion of four- and five-star facilities increased from 30.0% to 53.4% between October 2015 and April 2018. Quality improvement was driven by the domain of care containing the dialysis adequacy and hypercalcemia measures. Additionally, independently owned facilities and facilities belonging to smaller dialysis organizations had significantly lower odds of year-to-year improvement than facilities belonging to either of the two large dialysis organizations (Odds Ratio [OR]: 0.736, 95% Confidence Interval [CI]: 0.631-0.856 and OR: 0.797, 95% CI: 0.723-0.879, respectively).Conclusions: The percentage of four- and five-star facilities has increased markedly over a three-year time period. These changes were driven by improvement in the specific quality measures that may be most directly under the control of the dialysis facility. [ABSTRACT FROM AUTHOR]

Background: Self-reported tobacco pack-year history plays a large role in decisions about low-dose CT screening for lung cancer, yet is challenging to measure accurately.Research Question: To what extent does random measurement error in pack-year information impact screening decisions and screening effectiveness?Study Design and Methods: Retrospective cohort study of 10,449 patients with pack-year history documented at least twice between October 2013 and July 2017 across 8 academic Veterans Affairs sites. Outcome measures included (1) observed reliability of pack-year information based on all repeat measures for the study population and (2) each person's statistically "true" pack-year information based on best linear unbiased predictor from a multilevel linear random effects model. To examine how unreliability leads to misclassification of screening eligibility and inaccuracy in estimating lung cancer risk, we simulated pack-year observations for each person, first comparing simulated pack-year and lung cancer risk values with true values, then comparing outcomes when basing screening decisions on unreliable pack-year information vs true information.Results: Reliability of assessing pack-year information in routine practice varied across sites. Thus, we examined the clinical impact of two different levels of reliability, based on the range of intraclass correlation coefficients observed. Using a ≥ 30-pack-year threshold led to a high rate of eligibility misclassifications (48.1% misclassified with higher reliability pack-year information and 60.7% misclassified with lower reliability information). However, using a lung cancer risk threshold leads to fewer misclassifications (47.3%-49.7% misclassified when using lower reliability pack-year information) and maintains screening effectiveness better when using unreliable pack-year information.Interpretation: Random error in real-world pack-year assessments leads to a substantial rate of misclassifying who should be offered CT screening if a ≥ 30-pack-year criterion is used. However, using a lung cancer risk threshold mitigates the impact of unreliable pack-year information. Decision-makers concerned about the impact of unreliable pack-year information should consider using risk-based approaches to CT screening. [ABSTRACT FROM AUTHOR]

Background and Aims: Chronic hepatitis B (CHB) affects >290 million persons globally, and only 10% have been diagnosed, presenting a severe gap that must be addressed. We developed logistic regression (LR) and machine learning (ML; random forest) models to accurately identify patients with HBV, using only easily obtained demographic data from a population-based data set., Approach and Results: We identified participants with data on HBsAg, birth year, sex, race/ethnicity, and birthplace from 10 cycles of the National Health and Nutrition Examination Survey (1999-2018) and divided them into two cohorts: training (cycles 2, 3, 5, 6, 8, and 10; n = 39,119) and validation (cycles 1, 4, 7, and 9; n = 21,569). We then developed and tested our two models. The overall cohort was 49.2% male, 39.7% White, 23.2% Black, 29.6% Hispanic, and 7.5% Asian/other, with a median birth year of 1973. In multivariable logistic regression, the following factors were associated with HBV infection: birth year 1991 or after (adjusted OR [aOR], 0.28; p < 0.001); male sex (aOR, 1.49; p = 0.0080); Black and Asian/other versus White (aOR, 5.23 and 9.13; p < 0.001 for both); and being USA-born (vs. foreign-born; aOR, 0.14; p < 0.001). We found that the ML model consistently outperformed the LR model, with higher area under the receiver operating characteristic values (0.83 vs. 0.75 in validation cohort; p < 0.001) and better differentiation of high- and low-risk persons., Conclusions: Our ML model provides a simple, targeted approach to HBV screening, using only easily obtained demographic data., (© 2021 American Association for the Study of Liver Diseases.)

Background: Studies have shown that Black patients die more frequently following coronary artery bypass grafting than their White counterparts for reasons not fully explained by disease severity or comorbidity. To examine whether provider care team segregation within hospitals contributes to this inequity, we analyzed national Medicare data., Methods: Using national Medicare data, we identified beneficiaries who underwent coronary artery bypass grafting at hospitals where this procedure was performed on at least 10 Black and 10 White patients between 2008 and 2014 (n=12 646). After determining the providers who participated in their perioperative care, we examined the extent to which Black and White patients were cared for by unique networks of provider care teams within the same hospital. We then evaluated whether a lack of overlap in composition of the provider care teams treating Black versus White patients (ie, high segregation) was associated with higher 90-day operative mortality among Black patients., Results: The median level of provider care team segregation was high (0.89) but varied across hospitals (interquartile range, 0.85-0.90). On multivariable analysis, after controlling for patient-, hospital-, and community-level differences, mortality rates for White patients were comparable at hospitals with high and low levels of provider care segregation (5.4% [95% CI, 4.7%-6.1%] versus 5.8% [95% CI, 4.7%-7.0%], respectively; P =0.601), while Black patients treated at high-segregation hospitals had significantly higher mortality than those treated at low-segregation hospitals (8.3% [95% CI, 5.4%-12.4%] versus 3.3% [95% CI, 2.0%-5.4%], respectively; P =0.017). The difference in mortality rates for Black and White patients treated at low-segregation hospitals was nonsignificant (-2.5%; P =0.098)., Conclusions: Black patients who undergo coronary artery bypass grafting at a hospital with a higher level of provider care team segregation die more frequently after surgery than Black patients treated at a hospital with a lower level.

Purpose: We investigated the prognostic value of distant metastasis sites among patients with metastatic colorectal cancer (CRC) and the significance of metastasectomy and resection of the primary CRC., Methods: Between 2010 and 2014, patients diagnosed with metastatic colorectal adenocarcinoma were selected using the surveillance, epidemiology, and end results (SEER) database. The prognosis of these patients was compared according to the site of metastasis (liver, lung, bone, and brain). A total of 15,133 patients suffered from isolated organ involvement, while 5135 patients experienced multiple organ metastases., Results: In the isolated organ metastasis cohort, median overall survival (OS) for patients with liver, lung, bone, and brain metastases was 16, 20, 7, and 5 months, respectively. Patients with isolated lung metastases had better cancer-specific survival (CSS) and OS as compared to patients with metastases at any other sites (p < 0.0001 for both CSS and OS). Patients with isolated liver metastases had better prognosis as compared to patients with isolated bone or brain metastases (p < 0.0001 for both CSS and OS). Moreover, patients with a single metastatic site had better prognosis than patients with multiple organs involved (p < 0.0001 for both CSS and OS). Multivariate analysis in patients with isolated organ metastases demonstrated that age ≤ 60 years, rectal cancer, being married, non-black race, N0 stage, and surgery of the primary and distant lesions showed more favorable prognosis., Conclusions: The metastatic site was an independent prognostic factor in stage IV colorectal cancer. Also, carefully chosen patients may benefit from surgery.

There have been notable improvements in survival over the past 2 decades for gastrointestinal (GI) cancer. However, the degree of improvement by age, race, and sex remains unclear. We analyzed data from 9 population-based cancer registries included in the SEER program of the National Cancer Institute (SEER 9) in 1990 to 2009 (n = 288,337). The degree of survival improvement over time by age, race, and sex was longitudinally measured. From 1990 to 2009, improvements in survival were greater for younger age groups. For patients aged 20 to 49 years and diagnosed from 2005 to 2009, adjusted HRs (95% CIs) were 0.74 (95% CI, 0.66-0.83), 0.49 (95% CI, 0.37-0.64), 0.69 (95% CI, 0.65-0.76), 0.62 (95% CI, 0.54-0.69), and 0.56 (95% CI, 0.42-0.76), for cancer of the stomach, small intestine, colon, rectum and anus, respectively, compared with the same age groups of patients diagnosed during 1990 to 1994. Compared with African Americans, whites experienced greater improvement in small intestinal and anal cancer survival. Female anal cancer and regional anal cancer patients experienced no improvement. Our data suggest that different improvement in survival in age, sex and race exists.

Pathologic T1-2N0 rectal cancer shows an excellent prognosis without preoperative or postoperative chemoradiation. However, oncologic outcome of ypT1-2N0 remains unclear and undetermined. Thus, the aim of this study was to compare the survival of ypT1-2 and pT1-2 rectal cancer patients after radical resection and identify risk factors of ypT1-2 rectal cancer in Surveillance, Epidemiology, and End Results Program (SEER)-registered rectal cancer patients. The results showed that ypT1-2N0 rectal cancer after neoadjuvant chemoradiation has lower survival compared with pT1-2N0 rectal cancer and mucinous/signet-ring cancer and less than 12 lymph nodes retrieval were two risk factors in ypT1-2 patients. These results suggest that ypT1-2 patients with one or two risk factors may benefit from postoperative adjuvant chemotherapy.