Addressing Extreme Propensity Scores in Estimating Counterfactual Survival Functions via the Overlap Weights

Impact of geographical cohorting, multidisciplinary rounding and incremental case management support on hospital length of stay and readmission rates: a propensity weighted analysis

Hospital length of stay (LOS) in the USA has been increasing since the start of the COVID-19 pandemic, with numerous negative outcomes, including decreased quality of care, worsened patient satisfaction and negative financial impacts on hospitals. While many proposed factors contributing to prolonged LOS are challenging to modify, poor coordination of care and communication among clinical teams can be improved.Geographical cohorting of provider teams, patients and other clinical staff is proposed as a solution to prolonged LOS and readmissions. However, many studies on geographical cohorting alone have shown no significant impact on LOS or readmissions. Other potential benefits of geographical cohorting include improved quality of care, learning experience, communication, teamwork and efficiency.This paper presents a retrospective study at Duke University Hospital (DUH) on the General Medicine service, deploying a bundled intervention of geographical cohorting of patients and their care teams, twice daily multidisciplinary rounds and incremental case management support. The quality improvement study found that patients in the intervention arm had 16%-17% shorter LOS than those in the control arms, and there was a reduction in 30-day hospital readmissions compared with the concurrent control arm. Moreover, there was some evidence of improved accuracy of estimated discharge dates in the intervention arm.Based on these findings, the health system at DUH recognised the value of geographical cohorting and implemented additional geographically based medicine units with multidisciplinary rounds. Future studies will confirm the sustained impact of these care transformations on hospital throughput and patient outcomes, aiming to reduce LOS and enhance the quality of care provided to patients.

Causal interpretation of the hazard ratio in randomized clinical trials

BACKGROUND

Although the hazard ratio has no straightforward causal interpretation, clinical trialists commonly use it as a measure of treatment effect.

METHODS

We review the definition and examples of causal estimands. We discuss the causal interpretation of the hazard ratio from a two-arm randomized clinical trial, and the implications of proportional hazards assumptions in the context of potential outcomes. We illustrate the application of these concepts in a synthetic model and in a model of the time-varying effects of COVID-19 vaccination.

RESULTS

We define causal estimands as having either an individual-level or population-level interpretation. Difference-in-expectation estimands are both individual-level and population-level estimands, whereas without strong untestable assumptions the causal rate ratio and hazard ratio have only population-level interpretations. We caution users against making an incorrect individual-level interpretation, emphasizing that in general a hazard ratio does not on average change each individual's hazard by a factor. We discuss a potentially valid interpretation of the constant hazard ratio as a population-level causal effect under the proportional hazards assumption.

CONCLUSION

We conclude that the population-level hazard ratio remains a useful estimand, but one must interpret it with appropriate attention to the underlying causal model. This is especially important for interpreting hazard ratios over time.

Complications after peripherally inserted central catheter versus central venous catheter implantation in intensive care unit: propensity score analysis using a nationwide database

BACKGROUND

It remains unclear whether peripherally inserted central catheters (PICCs) are superior to central venous catheters (CVCs); therefore, we compared post-implantation complications between CVC and PICC groups.

RESEARCH DESIGN AND METHODS

Patients who received CVCs or PICCs between April 2010 and March 2018 were identified from the Diagnosis Procedure Combination database, a national inpatient database in Japan. The outcomes of interest included catheter infection, pulmonary embolism, deep vein thrombosis, and phlebitis. Propensity score overlap weighting was used to balance patient backgrounds. Outcomes were compared using logistic regression analyses.

RESULTS

We identified 164,185 eligible patients, including 161,605 (98.4%) and 2,580 (1.6%) in the CVC and PICC groups, respectively. The PICC group was more likely to have overall complications (odds ratio [OR], 1.70; 95% confidence interval [CI], 1.32-2.19), pulmonary embolism (OR, 2.32; 95% CI, 1.38-3.89), deep vein thrombosis (OR, 1.86; 95% CI, 1.16-2.99), and phlebitis (OR, 1.72; 95% CI, 1.27-2.32) than the CVC group. There was no significant intergroup difference in catheter infection (OR, 1.09; 95% CI, 0.39-3.04).

CONCLUSIONS

Patients with PICCs had a significantly greater incidence of complications than did those with CVCs. Further research is necessary to explore the factors contributing to these complications.

Recovery Dynamics and Prognosis After Dialysis for Acute Kidney Injury

Importance

The interplay among baseline kidney function, severity of acute kidney disease (AKD), and post-AKD kidney function has significant associations with patient outcomes. However, a comprehensive understanding of how these factors are collectively associated with mortality, major adverse cardiac events (MACEs), and end-stage kidney disease (ESKD) in patients with dialysis-requiring acute kidney injury (AKI-D) is yet to be fully explored.

Objective

To investigate the associations of baseline kidney function, AKD severity, and post-AKD kidney function with mortality, MACEs, and ESKD in patients with AKI-D.

Design, Setting, and Participants

This nationwide, population-based cohort study of patients with AKI-D was conducted between January 1, 2015, and December 31, 2018, using data from various health care settings included in the Taiwan nationwide population-based cohort database. Data analysis was conducted from April 28, 2022, to June 30, 2023.

Exposure

Exposure to severe AKI and baseline and post-AKD kidney function.

Main Outcomes and Measures

The primary outcomes were all-cause mortality and incident MACEs, and secondary outcomes were risks of permanent dialysis and readmission.

Results

A total of 6703 of 22 232 patients (mean [SD] age, 68.0 [14.7] years; 3846 [57.4%] male) with AKI-D with post-AKD kidney function follow-up and AKD stage data were enrolled. During a mean (SD) 1.2 (0.9) years of follow-up, the all-cause mortality rate was 28.3% (n = 1899), while the incidence rates of MACEs and ESKD were 11.1% (n = 746) and 16.7% (n = 1119), respectively. After adjusting for known covariates, both post-AKD kidney function and baseline kidney function, but not AKD severity, were independently associated with all-cause mortality, MACEs, ESKD, and readmission. Moreover, worse post-AKD kidney function correlated with progressive and significant increases in the risk of adverse outcomes.

Conclusions and Relevance

In this cohort study of patients with AKI-D, more than one-quarter of patients died after 1.2 years of follow-up. Baseline and post-AKD kidney functions serve as important factors associated with the long-term prognosis of patients with AKI-D. Therefore, concerted efforts to understand the transition from post-AKD to chronic kidney disease are crucial.

Prognostic variations between 'primary' and 'progressive' muscle-invasive bladder cancer following radical cystectomy: a novel propensity score-based multicenter cohort study

OBJECTIVE

To assess prognostic differences between primary and progressive muscle-invasive bladder cancer (MIBC) following radical cystectomy.

MATERIAL AND METHODS

The Surveillance, Epidemiology, and End Results (SEER) database was used to abstract MIBC data following radical cystectomy from 2000 to 2019. Patients were classified as either 'Primary' MIBC (defined as the presentation of muscle-invasive disease at initial diagnosis) or 'Progressive' MIBC (defined as a non-muscle invasive disease that later progressed to MIBC). Baseline characteristics for the two groups were balanced using a propensity score overlap weight (PSOW) technique. Survival differences between the two groups were analyzed using Kaplan-Meier's plots and log-rank tests. Cox's proportional hazard regression was used to assess risk factors associated with overall survival (OS) and cancer-specific survival (CSS).

RESULTS

Six thousand six hundred thirty-two MIBC patients were identified in the SEER database. Among them, 83.3% ( n =5658) were considered primary MIBC patients, and 16.7% ( n =974) were categorized as progressive MIBC patients. Distribution of baseline covariates, including age, sex, race, T stage, N stage, tumour grade, marital status, and chemotherapy, were well-balanced after PSOWs were applied. After stable PSOW adjustments, Kaplan-Meier survival analysis showed that the CSS for progressive MIBC [hazard ratio (HR)=1.25, 95% confidence interval (CI): 1.12-1.38, P <0.001) was poorer than the primary MIBC group. However, the difference in OS (HR=1.08, 95% CI: 0.99-1.18) was not significant ( P =0.073). Multivariate analysis also suggested that patients with progressive MIBC have significantly poorer CSS (HR=1.24, 95% CI: 1.19-1.38, P <0.001) but not OS (HR=1.08, 95% CI: 0.99-1.18, P =0.089).

CONCLUSION

CSS for progressive MIBC patients appears worse than for those with primary MIBC. This highlights the need to direct more resources for this patient population and particularly for high-risk cases of non-MIBC, where timely radical surgery will improve patients prognoses.

Enriching single-arm clinical trials with external controls: possibilities and pitfalls

For the past decade, it has become commonplace to provide rapid answers and early patient access to innovative treatments in the absence of randomized clinical trials (RCT), with benefits estimated from single-arm trials. This trend is important in oncology, notably when assessing new targeted therapies. Some of those uncontrolled trials further include an external/synthetic control group as an innovative way to provide an indirect comparison with a pertinent control group. We aimed to provide some guidelines as a comprehensive tool for (1) the critical appraisal of those comparisons or (2) for performing a single-arm trial. We used the example of ciltacabtagene autoleucel for the treatment of adult patients with relapsed or refractory multiple myeloma after 3 or more treatment lines as an illustrative example. We propose a 3-step guidance. The first step includes the definition of an estimand, which encompasses the treatment effect and the targeted population (whole population or restricted to single-arm trial or external controls), reflecting a clinical question. The second step relies on the adequate selection of external controls from previous RCTs or real-world data from patient cohorts, registries, or electronic patient files. The third step consists of choosing the statistical approach targeting the treatment effect defined above and depends on the available data (individual-level data or aggregated external data). The validity of the treatment effect derived from indirect comparisons heavily depends on careful methodological considerations included in the proposed 3-step procedure. Because the level of evidence of a well-conducted RCT cannot be guaranteed, the evaluation is more important than in standard settings.

Propensity score weighting methods for causal subgroup analysis with time-to-event outcomes

Evaluating causal effects of an intervention in pre-specified subgroups is a standard goal in comparative effectiveness research. Despite recent advancements in causal subgroup analysis, research on time-to-event outcomes has been lacking. This article investigates the propensity score weighting method for causal subgroup survival analysis. We introduce two causal estimands, the subgroup marginal hazard ratio and subgroup restricted average causal effect, and provide corresponding propensity score weighting estimators. We analytically established that the bias of subgroup-restricted average causal effect is determined by subgroup covariate balance. Using extensive simulations, we compare the performance of various combinations of propensity score models (logistic regression, random forests, least absolute shrinkage and selection operator, and generalized boosted models) and weighting schemes (inverse probability weighting, and overlap weighting) for estimating the causal estimands. We find that the logistic model with subgroup-covariate interactions selected by least absolute shrinkage and selection operator consistently outperforms other propensity score models. Also, overlap weighting generally outperforms inverse probability weighting in terms of balance, bias and variance, and the advantage is particularly pronounced in small subgroups and/or in the presence of poor overlap. We applied the methods to the observational Comparing Options for Management: PAtient-centered REsults for Uterine Fibroids study to evaluate the causal effects of myomectomy versus hysterectomy on the time to disease recurrence in a number of pre-specified subgroups of patients with uterine fibroids.

Relation of Statin Use with Esophageal Cancer

The present study evaluated the association of long-term statin use with the diagnosis and mortality of esophageal cancer in a Korean population. The Korean National Health Insurance Service-Health Screening Cohort from 2002 to 2019 was enrolled. Esophageal cancer patients were matched with control participants for demographic variables. The statin prescription histories were collected and grouped into <180 days, 180 to 545 days, and >545 days of duration. Propensity score overlap weighting was applied to minimize the bias between the esophageal cancer and control groups. The odds ratios (ORs) of the duration of statin use for esophageal cancer were analyzed using propensity score overlap weighted multivariable logistic regression analysis. The esophageal cancer group was classified as dead and surviving patients, and the ORs of the duration of statin use for the mortality of esophageal cancer were analyzed in an identical manner. Secondary analyses were conducted according to comorbid factors. Patients with esophageal cancer did not show lower odds for the duration of statin prescription in the overall study population (OR = 1.30, 95% CI = 1.03-1.65, p = 0.027 for 180 to 545 days and OR = 1.29, 95% CI = 1.08-1.55, p = 0.006 for >545 days). Subgroups of nonsmokers, past and current smokers, alcohol consumption ≥ 1 time a week, SBP < 140 mmHg and DBP < 90 mmHg, fasting blood glucose ≥ 100 mg/dL, total cholesterol ≥ 200 mg/dL, CCI score = 0, and nondyslipidemia history demonstrated low odds for the duration of statin prescription. Both types of statins, hydrophilic and lipophilic statins, were not related to a lower rate of esophageal cancer. The mortality of esophageal cancer was not associated with the duration of statin prescription. A subgroup with total cholesterol ≥ 200 mg/dL showed lower odds of statin prescription for mortality of esophageal cancer. The duration of statin prescription was not related to a lower rate or mortality of esophageal cancer in the adult Korean population.

Viability of internal comparisons for epidemiological research in the US astronaut corps

This study aims to determine whether astronauts who have not flown in space can provide an unbiased comparison to astronauts who have flown in space when analyzing long-term health outcomes such as incidence of chronic disease and mortality. Various propensity score methods failed to achieve good balance between groups, demonstrating that even with sophisticated rebalancing methods the group of non-flight astronauts cannot be demonstrated to be an unbiased comparison group for examining the effect of the hazards of spaceflight on incidence and mortality from chronic diseases.

Comparative effectiveness of omalizumab, mepolizumab, and dupilumab in asthma: A target trial emulation

BACKGROUND

Multiple mAbs are currently approved for the treatment of asthma. However, there is limited evidence on their comparative effectiveness.

OBJECTIVE

Our aim was to compare the effectiveness of omalizumab, mepolizumab, and dupilumab in individuals with moderate-to-severe asthma.

METHODS

We emulated a hypothetical randomized trial using electronic health records from a large US-based academic health care system. Participants aged 18 years or older with baseline IgE levels between 30 and 700 IU/mL and peripheral eosinophil counts of at least 150 cells/μL were eligible for study inclusion. The study period extended from March 2016 to August 2021. Outcomes included the incidence of asthma-related exacerbations and change in baseline FEV1 value over 12 months of follow-up.

RESULTS

In all, 68 individuals receiving dupilumab, 68 receiving omalizumab, and 65 receiving mepolizumab met the inclusion criteria. Over 12 months of follow-up, 31 exacerbations occurred over 68 person years (0.46 exacerbations per person year) in the dupilumab group, 63 over 68 person years (0.93 per person year) in the omalizumab group, and 86 over 65 person years (1.32 per person year) in the mepolizumab group (adjusted incidence rate ratios: dupilumab vs mepolizumab, 0.28 [95% CI = 0.09-0.84]; dupilumab vs omalizumab, 0.36 [95% CI = 0.12-1.09]; and omalizumab vs mepolizumab, 0.78 [95% CI = 0.32-1.91]). The differences in the change in FEV1 comparing patients who received the different biologics were as follows: 0.11 L (95% CI = -0.003 to 0.222 L) for dupilumab versus mepolizumab, 0.082 L (95% CI -0.040 to 0.204 L) for dupilumab versus omalizumab, and 0.026 L (95% CI -0.083 to 0.140 L) for omalizumab versus mepolizumab.

CONCLUSIONS

Among patients with asthma and eosinophil counts of at least 150 cells/μL and IgE levels of 30 to 700 kU/L, dupilumab was associated with greater improvements in exacerbation and FEV1 value than omalizumab and mepolizumab.

Effect of intravenous antihypertensives on outcomes of severe hypertension in hospitalized patients without acute target organ damage

BACKGROUND

Treatment of severe inpatient hypertension (HTN) that develops during hospitalization is not informed by guidelines. Intravenous (i.v.) antihypertensives are used to manage severe HTN even in the absence of acute target organ damage; however they may result in unpredictable blood pressure (BP) reduction and cardiovascular events. Our goal was to assess the association between i.v. antihypertensives and clinical outcomes in this population.

METHODS

This is a multihospital retrospective study of adults admitted for reasons other than HTN who develop severe HTN during hospitalization without acute target end organ damage. We defined severe HTN as BP elevation of systolic >180 or diastolic >110 mmHg. Treatment was defined as receiving i.v. antihypertensives within 3 h of BP elevation. We used overlap propensity score weighted Cox models to study the association between treatment and clinical outcomes during index hospitalization.

RESULTS

Of 224 265 unique, nonintensive care unit hospitalizations, 20 383 (9%) developed severe HTN, of which 5% received i.v. antihypertensives and 79% were untreated within 3 h of severe BP elevation. In the overlap propensity weighted population, patients who received i.v. antihypertensives were more likely to develop myocardial injury (5.9% in treated versus 3.6% in untreated; hazard ratio [HR]: 1.6 [1.13, 2.24]). Treatment was not associated with increased risk of stroke (HR: 0.7 [0.3, 1.62]), acute kidney injury (HR: 0.97 [0.81, 1.17]), or death (HR: 0.86 [0.49, 1.51]).

CONCLUSIONS

Intravenous antihypertensives were associated with increased risk of myocardial injury in patients who develop severe HTN during hospitalization. These results suggest that i.v. antihypertensives should be used with caution in patients without acute target organ damage.

Addressing substantial covariate imbalance with propensity score stratification and balancing weights: connections and recommendations

Objectives

Propensity score (PS) weighting methods are commonly used to adjust for confounding in observational treatment comparisons. However, in the setting of substantial covariate imbalance, PS values may approach 0 and 1, yielding extreme weights and inflated variance of the estimated treatment effect. Adaptations of the standard inverse probability of treatment weights (IPTW) can reduce the influence of extremes, including trimming methods that exclude people with PS values near 0 or 1. Alternatively, overlap weighting (OW) optimizes criteria related to bias and variance, and performs well compared to other PS weighting and matching methods. However, it has not been compared to propensity score stratification (PSS). PSS has some of the same potential advantages; being insensitive extreme values. We sought to compare these methods in the setting of substantial covariate imbalance to generate practical recommendations.

Methods

Analytical derivations were used to establish connections between methods, and simulation studies were conducted to assess bias and variance of alternative methods.

Results

We find that OW is generally superior, particularly as covariate imbalance increases. In addition, a common method for implementing PSS based on Mantel-Haenszel weights (PSS-MH) is equivalent to a coarsened version of OW and can perform nearly as well. Finally, trimming methods increase bias across methods (IPTW, PSS and PSS-MH) unless the PS model is re-fit to the trimmed sample and weights or strata are re-derived. After trimming with re-fitting, all methods perform similarly to OW.

Conclusions

These results may guide the selection, implementation and reporting of PS methods for observational studies with substantial covariate imbalance.

Relationship between the Blood Urea Nitrogen to Creatinine Ratio and In-Hospital Mortality in Non-Traumatic Subarachnoid Hemorrhage Patients: Based on Propensity Score Matching Method

(1) Background: To explore the correlation between the blood urea nitrogen to creatinine ratio (UCR) and in-hospital mortality in non-traumatic subarachnoid hemorrhage patients. (2) Methods: Specific clinical information was collected from the Medical Information Mart for Intensive Ⅳ (MIMIC-Ⅳ) database. The optimal cut-off value of the UCR was calculated with ROC curve analysis conducted using the maximum Youden index for the prediction of survival status. Univariable and multivariable logistic regression analyses were also carried out to assess the prognostic significance of UCR, and the Kaplan−Meier (K−M) analysis was conducted to draw the survival curves. Then, the 1:1 propensity score matching (PSM) method was applied to improve the reliability of the research results while balancing the unintended influence of underlying confounders. (3) Results: This retrospective cohort study included 961 patients. The optimal cut-off value of the UCR for in-hospital mortality was 27.208. The PSM was performed to identify 92 pairs of score-matched patients, with balanced differences exhibited for nearly all variables. According to the K−M analysis, those patients with a UCR of more than 27.208 showed a significantly higher level of in-hospital mortality compared to the patients with a UCR of less than 27.208 (p < 0.05). After the adjustment for possible confounders, those patients whose UCR was more than 27.208 still had a significantly higher level of in-hospital mortality than the patients whose UCR was less than 27.208, as revealed by the multivariable logistic regression analysis (OR = 3.783, 95% CI: 1.959~7.305, p < 0.001). Similarly, the in-hospital mortality remained substantially higher for those patients in the higher UCR group than for the patients in the lower UCR group after PSM. (4) Conclusion: A higher level of the UCR was evidently associated with an increased risk of in-hospital mortality, which made the ratio useful as a prognostic predictor of clinical outcomes for those patients with non-traumatic subarachnoid hemorrhage.

Relationship between the Hemoglobin-to-Red Cell Distribution Width Ratio and All-Cause Mortality in Septic Patients with Atrial Fibrillation: Based on Propensity Score Matching Method

(1) Objective: To reveal the correlation between the hemoglobin-to-red cell distribution width ratio (HRR) and all-cause mortality (ACM) among the septic patients with atrial fibrillation. (2) Methods: Specific clinical information was collected from the Medical Information Mart for Intensive IV (MIMIC-IV) database. The optimal cut-off value of HRR was calculated through ROC curve analysis conducted by using the maximum Youden index for the prediction of survival status. In addition, univariable and multivariable Cox regressive analyses were carried out to assess the prognostic significance of HRR and the Kaplan-Meier (K-M) analysis was conducted to draw the survival curves. Then, the 1:1 propensity score matching (PSM) method was adopted to improve the reliability of research result while balancing the unintended influence of underlying confounders. (3) Results: There were 9228 patients participating in this retrospective cohort study. The optimal cut-off value of the HRR was determined as 5.877 for in-hospital mortality. The PSM was performed to identify 2931 pairs of score-matched patients, with balanced differences exhibited by nearly all variables. According to the K-M analysis, those patients with a lower HRR than 5.877 showed a significantly higher level of in-hospital mortality, 28-day mortality, and 90-day mortality, compared to the patients with HRR ≥ 5.877 (p < 0.001). After the adjustment of possible confounders, those patients whose HRR was below 5.877 had a significantly higher level of in-hospital mortality than the patients with HRR ≥ 5.877, as revealed by the multivariable Cox regression analysis (HR = 1.142, 95%CI: 1.210−1.648, p < 0.001). Similarly, the ACM remained substantially higher in those patients with a lower HRR than in the patients with higher HRR after PSM. (4) Conclusion: A lower HRR (<5.877) was evidently associated with an increased risk of ACM, which made it applicable as a prognostic predictor of clinical outcomes for those septic patients with atrial fibrillation.

Bootstrap vs asymptotic variance estimation when using propensity score weighting with continuous and binary outcomes

We used Monte Carlo simulations to compare the performance of asymptotic variance estimators to that of the bootstrap when estimating standard errors of differences in means, risk differences, and relative risks using propensity score weighting. We considered four different sets of weights: conventional inverse probability of treatment weights with the average treatment effect (ATE) as the target estimand, weights for estimating the average treatment effect in the treated (ATT), matching weights, and overlap weights. We considered sample sizes ranging from 250 to 10 000 and allowed the prevalence of treatment to range from 0.1 to 0.9. We found that, when using ATE weights and sample sizes were ≤ 1000, then the use of the bootstrap resulted in estimates of SE that were more accurate than the asymptotic estimates. A similar finding was observed when using ATT weights and sample sizes were ≤ 1000 and the prevalence of treatment was moderate to high. When using matching weights and overlap weights, both the asymptotic estimator and the bootstrap resulted in accurate estimates of SE across all sample sizes and prevalences of treatment. Even when using the bootstrap with ATE weights, empirical coverage rates of confidence intervals were suboptimal when sample sizes were low to moderate and the prevalence of treatment was either very low or very high. A similar finding was observed when using the bootstrap with ATT weights when sample sizes were low to moderate and the prevalence of treatment was very high.