Assessing causal treatment effect estimation when using large observational datasets

Discharge Against Medical Advice After Hospitalization for Sepsis: Predictors, 30-Day Readmissions, and Outcomes

BACKGROUND

Sepsis is a leading cause of death worldwide. However, little has been known concerning the status of discharge against medical advice (DAMA) in sepsis patients.

OBJECTIVE

To identify factors associated with DAMA, evaluate the association of DAMA with 30-day unplanned readmission and readmitted outcomes after sepsis hospitalization.

METHODS

Using the National Readmission Database, we identified sepsis patients who discharged routinely or DAMA in 2017. Multivariable models were used to identify factors related to DAMA, evaluate the association between DAMA and readmission, and elucidate the relationship between DAMA and outcomes in patients readmitted within 30 days.

RESULTS

Among 1,012,650 sepsis cases, patients with DAMA accounted for 3.88% (n = 39,308). The unplanned 30-day readmission rates in patients who discharged home and DAMA were 13.08% and 27.21%, respectively. Predictors of DAMA in sepsis included Medicaid, diabetes, smoking, drug abuse, alcohol abuse, and psychoses. DAMA was statistically significantly associated with 30-day (odds ratio [OR] 2.18, 95% confidence interval [CI] 2.09-2.28), 60-day (OR 1.98, 95% CI 1.90-2.06), and 90-day (OR 1.88, 95% CI 1.81-1.96) readmission. DAMA is also associated with higher mortality in patients readmitted within 30 days (OR 1.38, 95% CI 1.17-1.63), whereas there were no statistically significant differences in length of stay and costs between patients who discharged home or DAMA.

CONCLUSIONS

DAMA occurs in nearly 3.88% of sepsis patients and is linked to higher readmission and mortality. Those at high risk of DAMA should be early identified to motivate intervention to avoid premature discharges and associated adverse outcomes.

Estimation of treatment effects in observational stroke care data: comparison of statistical approaches

Introduction

Various statistical approaches can be used to deal with unmeasured confounding when estimating treatment effects in observational studies, each with its own pros and cons. This study aimed to compare treatment effects as estimated by different statistical approaches for two interventions in observational stroke care data.

Patients and methods

We used prospectively collected data from the MR CLEAN registry including all patients (n = 3279) with ischemic stroke who underwent endovascular treatment (EVT) from 2014 to 2017 in 17 Dutch hospitals. Treatment effects of two interventions – i.e., receiving an intravenous thrombolytic (IVT) and undergoing general anesthesia (GA) before EVT – on good functional outcome (modified Rankin Scale ≤2) were estimated. We used three statistical regression-based approaches that vary in assumptions regarding the source of unmeasured confounding: individual-level (two subtypes), ecological, and instrumental variable analyses. In the latter, the preference for using the interventions in each hospital was used as an instrument.

Results

Use of IVT (range 66–87%) and GA (range 0–93%) varied substantially between hospitals. For IVT, the individual-level (OR ~ 1.33) resulted in significant positive effect estimates whereas in instrumental variable analysis no significant treatment effect was found (OR 1.11; 95% CI 0.58–1.56). The ecological analysis indicated no statistically significant different likelihood (β = − 0.002%; P = 0.99) of good functional outcome at hospitals using IVT 1% more frequently. For GA, we found non-significant opposite directions of points estimates the treatment effect in the individual-level (ORs ~ 0.60) versus the instrumental variable approach (OR = 1.04). The ecological analysis also resulted in a non-significant negative association (0.03% lower probability).

Discussion and conclusion

Both magnitude and direction of the estimated treatment effects for both interventions depend strongly on the statistical approach and thus on the source of (unmeasured) confounding. These issues should be understood concerning the specific characteristics of data, before applying an approach and interpreting the results. Instrumental variable analysis might be considered when unobserved confounding and practice variation is expected in observational multicenter studies.

An introduction to inverse probability of treatment weighting in observational research

In this article we introduce the concept of inverse probability of treatment weighting (IPTW) and describe how this method can be applied to adjust for measured confounding in observational research, illustrated by a clinical example from nephrology. IPTW involves two main steps. First, the probability—or propensity—of being exposed to the risk factor or intervention of interest is calculated, given an individual’s characteristics (i.e. propensity score). Second, weights are calculated as the inverse of the propensity score. The application of these weights to the study population creates a pseudopopulation in which confounders are equally distributed across exposed and unexposed groups. We also elaborate on how weighting can be applied in longitudinal studies to deal with informative censoring and time-dependent confounding in the setting of treatment-confounder feedback.

Managing confounding and effect modification in pediatric/congenital interventional cardiology research

Measuring the effect of a treatment on outcomes is an important goal for research in pediatric/congenital interventional cardiology. The breadth of anatomic and physiologic variations, patient ages, and genetic syndromes and noncardiac comorbid conditions all represent sources of potential confounding and effect modification that are major obstacles to this goal. If not accounted for, these factors can obscure the "true" treatment effect and lead to spurious conclusions about the relative efficacy and/or safety of therapies. In this review, we discuss the importance of confounding and effect modification in pediatric/congenital interventional cardiology research. We define these terms and discuss strategies (both in study design and data analysis) to mitigate error introduced by confounding and effect modification. The importance of confounding by indication in pediatric/congenital cardiology is discussed along with specific methods to address it.

A Method to Estimate the Efficacy vs. Effectiveness in Meta-Analysis of Clinical Trials with Different Adherence Scenarios: A Monte Carlo Simulation Study in Nutrition

Randomized clinical trials (RCTs) evaluating the effectiveness of interventions to promote fruit and vegetable (FV) consumption usually report intention-to-treat (ITT) analysis as the main outcome. These analyses compare the randomly assigned groups and accept that some individuals may not follow the recommendations received in their group. The ITT analysis is useful to quantify the global effect of promoting the consumption of FV in a population (effectiveness) but, if non-adherence is significant in the RCT, they cannot estimate the specific effect in the individuals that increased their FV consumption (efficacy). To calculate the efficacy of FV consumption, a per protocol analysis (PP) would have to be carried out, in which groups of individuals are compared according to their actual adherence to FV consumption, regardless of the group to which they were assigned; unfortunately, many RCTs do not report the PP analysis. The objective of this article is to apply a new method to estimate the efficacy of Meta-analysis (MA) PP which include RCTs of effectiveness by ITT, without estimates of adherence. The method is based on generating Monte Carlo simulations of percentages of adherence in each allocation group from prior distributions informed by expert knowledge. We illustrate the method reanalyzing a Cochrane Systematic Review (SR) of RCTs on increased FV consumption reported with ITT, simulating 1000 times the estimation of a PP meta-analyses, and obtaining means and ranges of the potential PP effects. In some cases, the range of estimated PP effects was clearly more favourable than the effect calculated with the original ITT assumption, and therefore this corrected analysis must be considered when estimating the true effect of the consumption of a certain food.