Predicting Hospital Readmission in Medicaid Patients With COPD Using Administrative and Claims Data.

BACKGROUND: The goals of this study were to develop a model that predicts the risk of 30-d all-cause readmission in hospitalized Medicaid patients diagnosed with COPD and to create a predictive model in a retrospective study of a population cohort. METHODS: We analyzed 2016-2019 Medicaid claims data from 7 United States states. A COPD admission was one in which either the admission diagnosis or the first or second clinical (discharge) diagnosis bore an International Classification of Diseases, Tenth Revision code for COPD. A readmission was an admission for any condition (not necessarily COPD) that occurred within 30 d of a COPD discharge. We estimated a mixed-effects logistic model to predict 30-d readmission from patient demographic data, comorbidities, past health care utilization, and features of the index hospitalization. We evaluated model fit graphically and measured predictive accuracy by the area under the receiver operating characteristic (ROC) curve. RESULTS: Among 12,283 COPD hospitalizations contributed by 9,437 subjects, 2,534 (20.6%) were 30-d readmissions. The final model included demographics, comorbidities, claims history, admission and discharge variables, length of stay, and seasons of admission and discharge. The observed versus predicted plot showed reasonable fit, and the estimated area under the ROC curve of 0.702 was robust in sensitivity analyses. CONCLUSIONS: Our model identified with acceptable accuracy hospitalized Medicaid patients with a diagnosis of COPD who are at high risk of readmission. One can use the model to develop post-discharge management interventions for reducing readmissions, for adjusting comparisons of readmission rates between sites/providers or over time, and to guide a patient-centered approach to patient care.

Tutorial: Assessing the impact of nonignorable missingness on regression analysis using Index of Local Sensitivity to Nonignorability.

Data sets with missing observations are common in psychology research. One typically analyzes such data by applying statistical methods that rely on the assumption that the missing observations are missing at random (MAR). This assumption greatly simplifies analysis but is unverifiable from the data at hand, and assuming it incorrectly may lead to bias. Thus we often wish to conduct sensitivity analyses to judge whether conclusions are robust to departures from MAR-that is, whether key findings would hold up even if MAR does not in fact hold. This article describes a class of sensitivity analyses derived from a measure of robustness called the Index of Local Sensitivity to Nonignorability (ISNI). ISNI is straightforward to compute and avoids the estimation of complicated non-MAR missing-data models. The accompanying R package isni implements the method for a range of commonly used regression models; the syntax is simple and similar to that for the regular analysis that assumes MAR. We illustrate the application of the method and software to address the credibility of MAR analyses in a series of analyses of real-world data sets from psychology research. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Predicting hospital readmission in Medicaid patients with diabetes using administrative and claims data.

OBJECTIVES: Readmission is common and costly for hospitalized Medicaid patients with diabetes. We aimed to develop a model predicting risk of 30-day readmission in Medicaid patients with diabetes hospitalized for any cause. STUDY DESIGN: Using 2016-2019 Medicaid claims from 7 US states, we identified patients who (1) had a diagnosis of diabetes or were prescribed any diabetes drug, (2) were hospitalized for any cause, and (3) were discharged to home or to a nonhospice facility. For each encounter, we assessed whether the patient was readmitted within 30 days of discharge. METHODS: Applying least absolute shrinkage and selection operator variable selection, we included demographic data and claims history in a logistic regression model to predict 30-day readmission. We evaluated model fit graphically and measured predictive accuracy by the area under the receiver operating characteristic curve (AUROC). RESULTS: Among 69,640 eligible patients, there were 129,170 hospitalizations, of which 29,410 (22.8%) were 30-day readmissions. The final model included age, sex, age-sex interaction, past diagnoses, US state of admission, number of admissions in the preceding year, index admission type, index admission diagnosis, discharge status, length of stay, and length of stay-sex interaction. The observed vs predicted plot showed good fit. The estimated AUROC of 0.761 was robust in analyses that assessed sensitivity to a range of model assumptions. CONCLUSIONS: Our model has moderate power for identifying hospitalized Medicaid patients with diabetes who are at high risk of readmission. It is a template for identifying patients at risk of readmission and for adjusting comparisons of 30-day readmission rates among sites or over time.

Electronic phenotypes to distinguish clinician attention to high body mass index, hypertension, lipid disorders, fatty liver and diabetes in pediatric primary care: Diagnostic accuracy of electronic phenotypes compared to masked comprehensive chart review.

BACKGROUND/OBJECTIVES: Electronic phenotyping is a method of using electronic-health-record (EHR) data to automate identifying a patient/population with a characteristic of interest. This study determines validity of using EHR data of children with overweight/obesity to electronically phenotype evidence of clinician 'attention' to high body mass index (BMI) and each of four distinct comorbidities. METHODS: We built five electronic phenotypes classifying 2-18-year-old children with overweight/obesity (n = 17,397) by electronic/health-record evidence of distinct attention to high body mass index, hypertension, lipid disorders, fatty liver, and prediabetes/diabetes. We reviewed, selected and cross-checked random charts to define items clinicians select in EHRs to build problem lists, and to order medications, laboratory tests and referrals to electronically classify attention to overweight/obesity and each comorbidity. Operating characteristics of each clinician-attention phenotype were determined by comparing comprehensive chart review by reviewers masked to electronic classification who adjudicated evidence of clinician attention to high BMI and each comorbidity. RESULTS: In a random sample of 817 visit-records reviewed/coded, specificity of each electronic phenotype is 99%-100% (with PPVs ranging from 96.8% for prediabetes/diabetes to 100% for dyslipidemia and hypertension). Sensitivities of the attention classifications range from 69% for hypertension (NPV, 98.9%) to 84.7% for high-BMI attention (NPV, 92.3%). CONCLUSIONS: Electronic phenotypes for clinician attention to overweight/obesity and distinct comorbidities are highly specific, with moderate (BMI) to modest (each comorbidity) sensitivity. The high specificity supports using phenotypes to identify children with prior high-BMI/comorbidity attention.

Sensitivity of estimands in clinical trials with imperfect compliance.

In clinical trials that are subject to noncompliance, the commonly used intention-to-treat estimand is valid as a causal effect of treatment assignment but is sensitive to the level of compliance. An alternative estimand, the complier average causal effect (CACE), measures the average effect of treatment received in the latent subset of subjects who would comply with either assigned treatment. Because the principal stratum of compliers can vary with the circumstances of the trial, CACE too depends on the compliance fraction. We propose a model in which an underlying latent proto-compliance interacts with trial characteristics to determine a subject's compliance behavior. When the latent compliance is independent of the individual treatment effect, the average causal effect is constant across compliance classes, and CACE is robust across trials and equal to the population average causal effect. We demonstrate the potential degree of sensitivity of CACE in a simulation study, an analysis of data from a trial of vitamin A supplementation in children, and a meta-analysis of trials of epidural analgesia in labor.

Survival of cancer survivors with a new pancreatic cancer diagnosis.

BACKGROUND: Persons newly diagnosed with pancreas cancer and who have survived a previous cancer are often excluded from clinical trials, despite limited evidence about their prognosis. We examined the association between previous cancer and overall survival. METHODS: This US population-based cohort study included older adults (aged ≥66 years) diagnosed with pancreas cancer between 2005 and 2015 in the linked Surveillance, Epidemiology, and End Results-Medicare data. We used Cox proportional hazards models to estimate stage-specific effects of previous cancer on overall survival, adjusting for sociodemographic, treatment, and tumor characteristics. RESULTS: Of 32,783 patients, 18.7% were previously diagnosed with another cancer. The most common previous cancers included prostate (29.0%), breast (18.9%), or colorectal (9.7%) cancer. More than half of previous cancers (53.9%) were diagnosed 5 or more years prior to pancreas cancer diagnosis or at an in situ or localized stage (47.8%). The proportions of patients surviving 1, 3, and 5 years after pancreas cancer were nearly identical for those with and without previous cancer. Median survival in months was as follows for those with and without previous cancer respectively: 7 versus 8 (Stage 0/I), 10 versus 10 (Stage II), 7 versus 7 (Stage III), and 3 versus 2 (Stage IV). Cox models indicated that patients with previous cancer had very similar or statistically equivalent survival to those with no previous cancer. CONCLUSIONS: Given nearly equivalent survival compared to those without previous cancer, cancer survivors newly diagnosed with pancreas cancer should be considered for inclusion in pancreas cancer clinical trials.

Analysis of local sensitivity to nonignorability with missing outcomes and predictors.

The ISNI (index of sensitivity to local nonignorability) method quantifies local sensitivity of parametric inferences to nonignorable missingness in an outcome variable. Here we extend ISNI to the situations where both outcomes and predictors can be missing and where the missingness mechanism can be either parametric or semi-parametric. We define the quantity MinNI (minimum nonignorability) to be an approximation to the norm of the smallest value of the transformed nonignorability that gives a nonnegligible displacement of the estimate of the parameter of interest. We illustrate our method in a complete data set from which we synthetically delete observations according to various patterns. We then apply the method to real-data examples involving the normal linear model and conditional logistic regression.

PSA-Net: Deep learning-based physician style-aware segmentation network for postoperative prostate cancer clinical target volumes.

PURPOSE: Automatic segmentation of medical images with deep learning (DL) algorithms has proven highly successful in recent times. With most of these automation networks, inter-observer variation is an acknowledged problem that leads to suboptimal results. This problem is even more significant in segmenting postoperative clinical target volumes (CTV) because they lack a macroscopic visible tumor in the image. This study, using postoperative prostate CTV segmentation as the test case, tries to determine 1) whether physician styles are consistent and learnable, 2) whether physician style affects treatment outcome and toxicity, and 3) how to explicitly deal with different physician styles in DL-assisted CTV segmentation to facilitate its clinical acceptance. METHODS: A dataset of 373 postoperative prostate cancer patients from UT Southwestern Medical Center was used for this study. We used another 83 patients from Mayo Clinic to validate the developed model and its adaptability. To determine whether physician styles are consistent and learnable, we trained a 3D convolutional neural network classifier to identify which physician had contoured a CTV from just the contour and the corresponding CT scan. Next, we evaluated whether adapting automatic segmentation to specific physician styles would be clinically feasible based on a lack of difference between outcomes. Here, biochemical progression-free survival (BCFS) and grade 3+ genitourinary and gastrointestinal toxicity were estimated with the Kaplan-Meier method and compared between physician styles with the log rank test and subsequently with a multivariate Cox regression. When we found no statistically significant differences in outcome or toxicity between contouring styles, we proposed a concept called physician style-aware (PSA) segmentation by developing an encoder-multidecoder network with perceptual loss to model different physician styles of CTV segmentation. RESULTS: The classification network captured the different physician styles with 87% accuracy. Subsequent outcome analysis showed no differences in BCFS and grade 3+ toxicity among physicians. With the proposed physician style-aware network (PSA-Net), Dice similarity coefficient (DSC) accuracy for all physicians was 3.4% higher on average than with a general model that does not differentiate physician styles. We show that these stylistic contouring variations also exist between institutions that follow the same segmentation guidelines, and we show the proposed method's effectiveness in adapting to new institutional styles. We observed an accuracy improvement of 5% in terms of DSC when adapting to the style of a separate institution. CONCLUSION: The performance of the classification network established that physician styles are learnable, and the lack of difference between outcomes among physicians shows that the network can feasibly adapt to different styles in the clinic. Therefore, we developed a novel PSA-Net model that can produce contours specific to the treating physician, thus improving segmentation accuracy and avoiding the need to train multiple models to achieve different style segmentations. We successfully validated this model on data from a separate institution, thus supporting the model's generalizability to diverse datasets.

Estimating the optimal timing of surgery by imputing potential outcomes.

Hypoplastic left heart syndrome is a congenital anomaly that is uniformly fatal in infancy without immediate treatment. The standard treatment consists of an initial Norwood procedure (stage 1) followed some months later by stage 2 palliation (S2P). The ideal timing of the S2P is uncertain. The Single Ventricle Reconstruction Trial (SVRT) randomized the procedure used in the initial Norwood operation, leaving the timing of the S2P to the discretion of the surgical team. To estimate the causal effect of the timing of S2P, we propose to impute the potential post-S2P survival outcomes using statistical models under the Rubin Causal Model framework. With this approach, it is straightforward to estimate the causal effect of S2P timing on post-S2P survival by directly comparing the imputed potential outcomes. Specifically, we consider a lognormal model and a restricted cubic spline model, evaluating their performance in Monte Carlo studies. When applied to the SVRT data, the models give somewhat different imputed values, but both support the conclusion that the optimal time for the S2P is at 6 months after the Norwood procedure.

Survival of patients newly diagnosed with colorectal cancer and with a history of previous cancer.

Patients with previous cancer are often excluded from clinical trials despite limited evidence about their prognosis. We examined the effect of previous cancer on overall and colorectal cancer (CRC)-specific survival of patients newly diagnosed with CRC. This population-based cohort study from the U.S.A. included patients aged ≥66 years and diagnosed with CRC between 2005 and 2015 in linked Surveillance, Epidemiology, and End Results-Medicare data. We estimated the stage-specific effects of a previous cancer on overall survival using Cox regression and on CRC-specific survival using competing risk regression. We also examined the effect of previous cancer type, timing, and stage on overall survival. Of 112,769 patients, 14.1% were previously diagnosed with another cancer--commonly prostate (32.9%) or breast (19.4%) cancer, with many (47.1%) diagnosed <5 years of CRC. For all CRC stages except IV, in which there was no difference, patients with previous cancer (vs. without) had worse overall survival. However, patients with previous cancer had improved CRC-specific survival. Overall survival for those with stage 0-III CRC varied by previous cancer type, timing, and stage; for example, patients with previous melanoma had overall survival equivalent to those with no previous cancer. Our results indicate that, in general, CRC patients with previous cancer have worse overall survival but superior CRC-specific survival. Given their equivalent survival to those without previous cancer, patients with previous melanoma and those with stage IV CRC with any type of previous cancer should be eligible to participate in clinical trials.

COVID-19 associated mortality and cardiovascular disease outcomes among US women veterans.

The burden of COVID-19 has been noted to be disproportionately greater in minority women, a population that is nevertheless still understudied in COVID-19 research. We conducted an observational study to examine COVID-19-associated mortality and cardiovascular disease outcomes after testing (henceforth index) among a racially diverse adult women veteran population. We assembled a retrospective cohort from a Veterans Affairs (VA) national COVID-19 shared data repository, collected between February and August 2020. A case was defined as a woman veteran who tested positive for SARS-COV-2, and a control as a woman veteran who tested negative. We used Kaplan-Meier curves and the Cox proportional hazards model to examine the distribution of time to death and the effects of baseline predictors on mortality risk. We used generalized linear models to examine 60-day cardiovascular disease outcomes. Covariates studied included age, body mass index (BMI), and active smoking status at index, and pre-existing conditions of diabetes, chronic kidney disease (CKD), chronic obstructive pulmonary disease (COPD), and a history of treatment with antiplatelet or anti-thrombotic drug at any time in the 2 years prior to the index date. Women veterans who tested positive for SARS-CoV-2 had 4 times higher mortality risk than women veterans who tested negative (Hazard Ratio 3.8, 95% Confidence Interval CI 2.92 to 4.89) but had lower risk of cardiovascular events (Odds Ratio OR 0.78, 95% CI 0.66 to 0.92) and developing new heart disease conditions within 60 days (OR 0.67, 95% CI 0.58 to 0.77). Older age, obesity (BMI > 30), and prior CVD and COPD conditions were positively associated with increased mortality in 60 days. Despite a higher infection rate among minority women veterans, there was no significant race difference in mortality, cardiovascular events, or onset of heart disease. SARS-CoV-2 infection increased short-term mortality risk among women veterans similarly across race groups. However, there was no evidence of increased cardiovascular disease incidence in 60 days. A longer follow-up of women veterans who tested positive is warranted.

Survival of women diagnosed with breast cancer and who have survived a previous cancer.

PURPOSE: Many women diagnosed with breast cancer have survived previous cancer; yet little is known about the impact of previous cancer on overall and cancer-specific survival. METHODS: This population-based cohort study using SEER-Medicare data included women (age ≥ 66 years) diagnosed with breast cancer between 2005 and 2015. Separately by breast cancer stage, we estimated effect of previous cancer on overall survival using Cox regression and on cause-specific survival using competing risk regression; all survival analyses adjusted for covariates. RESULTS: Of 138,576 women diagnosed with breast cancer, 8% had a previous cancer of another organ site, most commonly colorectal or uterine cancer or melanoma. Many of these women (46.3%) were diagnosed within 5 years of breast cancer. For all breast cancer stages except IV wherein there was no difference, women with vs. without previous cancer had worse overall survival. This survival disadvantage was driven by deaths due to the previous cancer and other causes. In contrast, women with previous cancer generally had favorable breast-cancer-specific survival, although this varied by stage. Overall survival varied by previous cancer type, timing, and stage; previous lung cancer, cancer diagnosed within 1 year of incident breast cancer, and previous cancer at a distant stage were associated with the worst survival. In contrast, women with a previous melanoma had equivalent overall survival to women without previous cancer. CONCLUSION: We observed variable impact of previous cancer on overall and breast-cancer-specific survival depending on breast cancer stage at diagnosis and the type, timing, and stage of previous cancer.

Estimating the optimal timing of surgery from observational data.

Infants with hypoplastic left heart syndrome require an initial Norwood operation, followed some months later by a stage 2 palliation (S2P). The timing of S2P is critical for the operation's success and the infant's survival, but the optimal timing, if one exists, is unknown. We attempt to identify the optimal timing of S2P by analyzing data from the Single Ventricle Reconstruction Trial (SVRT), which randomized patients between two different types of Norwood procedure. In the SVRT, the timing of the S2P was chosen by the medical team; thus with respect to this exposure, the trial constitutes an observational study, and the analysis must adjust for potential confounding. To accomplish this, we propose an extended propensity score analysis that describes the time to surgery as a function of confounders in a discrete competing-risk model. We then apply inverse probability weighting to estimate a spline hazard model for predicting survival from the time of S2P. Our analysis suggests that S2P conducted at 6 months after the Norwood gives the patient the best post-S2P survival. Thus, we place the optimal time slightly later than a previous analysis in the medical literature that did not account for competing risks of death and heart transplantation.

Approach to Managing a Postmenopausal Patient.

CASE AND PRINCIPLES OF MANAGEMENT: The case of a symptomatic, postmenopausal woman is presented and a full discussion of the approach to her management is discussed. Pertinent guidelines and scientific evidence are emphasized as support for the recommendations.

Flexible, rule-based dose escalation: The cohort-sequence design.

Phase I oncology trials seek to acquire preliminary information on the safety of novel treatments. In current practice, most such trials employ rule-based designs that determine whether to escalate the dose using data from the current dose only. The most popular of these, the 3 + 3, is simple and familiar but inflexible and inefficient. We propose a rule-based design that addresses these deficiencies. Our method, which we denote the cohort-sequence design, is defined by a sequence of J increasing cohort sizes n = ( n 1 , … , n J ) and corresponding critical values b = ( b 1 , … , b J ) . The idea is to begin with a small cohort size n 1 and escalate through the planned doses, increasing the cohort size when we encounter toxicities. By selection of J and a safety threshold tuning parameter θ, one can create a design that will efficiently identify a target toxicity rate, potentially including a built-in dose-expansion cohort. We compared our designs to the 3 + 3 under a range of toxicity scenarios, observing that our approach generally rapidly identifies an MTD without enrolling patients unnecessarily at low doses where both toxicity and response rates are likely to be low. We have implemented the design in the R package cohortsequence.

Underlying Breast Cancer Risk and Menopausal Hormone Therapy.

The recent Collaborative Group on Hormonal Factors in Breast Cancer (CGHFBC) publication calculated the attributable risk of breast cancer from use of estrogen alone and estrogen plus a synthetic progestogen for less than 5 to 15 or more years of use. This CGHFB report calculated attributable risk based on their findings of relative risk from pooled data from 58 studies. Notably, neither the CGHFBC nor other previous studies have examined the effect of underlying risk of breast cancer on attributable risk. This omission prompted us to determine the magnitude of the effect of underlying risk on attributable risk in this perspective. Meaningful communication of the potential risk of menopausal hormonal therapy requires providing women with the estimated risk above their existing underlying risk (ie, attributable risk). Therefore, we have estimated attributable risks from the data published by the CGHFBC, taking into account varying degrees of underlying risk. Based on the Endocrine Society Guideline on Menopausal Hormone Therapy (MHT), we divided groups into 3 categories of risk: low (1.5%), intermediate (3.0%), and high (6.0%) underlying risk of breast cancer over 5 years. In women taking estrogen plus a synthetic progestogen for 5 to 9 years, the attributable risks of MHT increased from 12, to 42, to 85 additional women per 1000 in the low-, intermediate-, and high-risk groups, respectively. The attributable risks for estrogen alone were lower but also increased based on underlying risk. Notably, the attributable risks were amplified with duration of MHT use, which increased both relative risk and breast cancer incidence.

Estimating cost-effectiveness from claims and registry data with measured and unmeasured confounders.

The analysis of observational data to determine the cost-effectiveness of medical treatments is complicated by the need to account for skewness, censoring, and the effects of measured and unmeasured confounders. We quantify cost-effectiveness as the Net Monetary Benefit (NMB), a linear combination of the treatment effects on cost and effectiveness that denominates utility in monetary terms. We propose a parametric estimation approach that describes cost with a Gamma generalized linear model and survival time (the canonical effectiveness variable) with a Weibull accelerated failure time model. To account for correlation between cost and survival, we propose a bootstrap procedure to compute confidence intervals for NMB. To examine sensitivity to unmeasured confounders, we derive simple approximate relationships between naïve parameters, assuming only measured confounders, and the values those parameters would take if there was further adjustment for a single unmeasured confounder with a specified distribution. A simulation study shows that the method returns accurate estimates for treatment effects on cost, survival, and NMB under the assumed model. We apply our method to compare two treatments for Stage II/III bladder cancer, concluding that the NMB is sensitive to hypothesized unmeasured confounders that represent smoking status and personal income.