On the Causal Interpretation of Rate-Change Methods: The Prior Event Rate Ratio and Rate Difference

Foundational model aided automatic high-throughput drug screening using self-controlled cohort study

Background

Developing medicine from scratch to governmental authorization and detecting adverse drug reactions (ADR) have barely been economical, expeditious, and risk-averse investments. The availability of large-scale observational healthcare databases and the popularity of large language models offer an unparalleled opportunity to enable automatic high-throughput drug screening for both repurposing and pharmacovigilance.

Objectives

To demonstrate a general workflow for automatic high-throughput drug screening with the following advantages: (i) the association of various exposure on diseases can be estimated; (ii) both repurposing and pharmacovigilance are integrated; (iii) accurate exposure length for each prescription is parsed from clinical texts; (iv) intrinsic relationship between drugs and diseases are removed jointly by bioinformatic mapping and large language model - ChatGPT; (v) causal-wise interpretations for incidence rate contrasts are provided.

Methods

Using a self-controlled cohort study design where subjects serve as their own control group, we tested the intention-to-treat association between medications on the incidence of diseases. Exposure length for each prescription is determined by parsing common dosages in English free text into a structured format. Exposure period starts from initial prescription to treatment discontinuation. A same exposure length preceding initial treatment is the control period. Clinical outcomes and categories are identified using existing phenotyping algorithms. Incident rate ratios (IRR) are tested using uniformly most powerful (UMP) unbiased tests.

Results

We assessed 3,444 medications on 276 diseases on 6,613,198 patients from the Clinical Practice Research Datalink (CPRD), an UK primary care electronic health records (EHR) spanning from 1987 to 2018. Due to the built-in selection bias of self-controlled cohort studies, ingredients-disease pairs confounded by deterministic medical relationships are removed by existing map from RxNorm and nonexistent maps by calling ChatGPT. A total of 16,901 drug-disease pairs reveals significant risk reduction, which can be considered as candidates for repurposing, while a total of 11,089 pairs showed significant risk increase, where drug safety might be of a concern instead.

Conclusions

This work developed a data-driven, nonparametric, hypothesis generating, and automatic high-throughput workflow, which reveals the potential of natural language processing in pharmacoepidemiology. We demonstrate the paradigm to a large observational health dataset to help discover potential novel therapies and adverse drug effects. The framework of this study can be extended to other observational medical databases.

Learning about treatment effects in a new target population under transportability assumptions for relative effect measures

Investigators often believe that relative effect measures conditional on covariates, such as risk ratios and mean ratios, are "transportable" across populations. Here, we examine the identification of causal effects in a target population using an assumption that conditional relative effect measures are transportable from a trial to the target population. We show that transportability for relative effect measures is largely incompatible with transportability for difference effect measures, unless the treatment has no effect on average or one is willing to make even stronger transportability assumptions that imply the transportability of both relative and difference effect measures. We then describe how marginal (population-averaged) causal estimands in a target population can be identified under the assumption of transportability of relative effect measures, when we are interested in the effectiveness of a new experimental treatment in a target population where the only treatment in use is the control treatment evaluated in the trial. We extend these results to consider cases where the control treatment evaluated in the trial is only one of the treatments in use in the target population, under an additional partial exchangeability assumption in the target population (i.e., an assumption of no unmeasured confounding in the target population with respect to potential outcomes under the control treatment in the trial). We also develop identification results that allow for the covariates needed for transportability of relative effect measures to be only a small subset of the covariates needed to control confounding in the target population. Last, we propose estimators that can be easily implemented in standard statistical software and illustrate their use using data from a comprehensive cohort study of stable ischemic heart disease.

The relative effectiveness of a high-dose quadrivalent influenza vaccine versus standard-dose quadrivalent influenza vaccines in older adults in France: a retrospective cohort study during the 2021-2022 influenza season

OBJECTIVES

High-dose quadrivalent influenza vaccine (HD-QIV) was introduced during the 2021/2022 influenza season in France for adults aged ≥65 years as an alternative to standard-dose quadrivalent influenza vaccine (SD-QIV). The aim of this study is to estimate the relative vaccine effectiveness of HD-QIV vs. SD-QIV against influenza-related hospitalizations in France.

METHODS

Community-dwelling individuals aged ≥65 years with reimbursed influenza vaccine claims during the 2021/2022 influenza season were included in the French national health insurance database. Individuals were followed up from vaccination day to 30 June 2022, nursing home admission or death date. Baseline socio-demographic and health characteristics were identified from medical records over the five previous years. Hospitalizations for influenza and other causes were recorded from 14 days after vaccination until the end of follow-up. HD-QIV and SD-QIV vaccinees were matched using 1:4 propensity score matching with an exact constraint on age group, sex, week of vaccination, and region. Incidence rate ratios were estimated using zero-inflated Poisson or zero-inflated negative binomial regression models.

RESULTS

We matched 405 385 HD-QIV to 1 621 540 SD-QIV vaccinees. HD-QIV was associated with a 23.3% (95% CI, 8.4-35.8) lower rate of influenza hospitalizations compared with SD-QIV (69.5/100 000 person years vs. 90.5/100 000 person years). Post-matching, we observed higher rates in the HD-QIV group for hospitalizations non-specific to influenza and negative control outcomes, suggesting residual confounding by indication.

DISCUSSION

HD-QIV was associated with lower influenza-related hospitalization rates vs. SD-QIV, consistent with existing evidence, in the context of high SARS-CoV-2 circulation in France and likely prioritization of HD-QIV for older/more comorbid individuals.

Association of Implementation of Postoverdose Outreach Programs With Subsequent Opioid Overdose Deaths Among Massachusetts Municipalities

Importance

Nonfatal opioid overdose is the leading risk factor for subsequent fatal overdose and represents a critical opportunity to reduce future overdose and mortality. Postoverdose outreach programs emerged in Massachusetts beginning in 2013 with the main purpose of linking opioid overdose survivors to addiction treatment and harm reduction services.

Objective

To evaluate whether the implementation of postoverdose outreach programs among Massachusetts municipalities was associated with lower opioid fatality rates compared with municipalities without postoverdose outreach programs.

Design, Setting, and Participants

This retrospective interrupted time-series analysis was performed over 26 quarters (from January 1, 2013, through June 30, 2019) across 93 municipalities in Massachusetts. These 93 municipalities were selected based on a threshold of 30 or more opioid-related emergency medical services (EMS) responses in 2015. Data were analyzed from November 2021 to August 2022.

Exposures

The main exposure was municipality postoverdose outreach programs. Municipalities had various program inceptions during the study period.

Main Outcomes and Measures

The primary outcome was quarterly municipal opioid fatality rate per 100 000 population. The secondary outcome was quarterly municipal opioid-related EMS response (ambulance trips) rates per 100 000 population.

Results

The mean (SD) population size across 93 municipalities was 47 622 (70 307), the mean (SD) proportion of female individuals was 51.5% (1.5%) and male individuals was 48.5% (1.5%), and the mean (SD) age proportions were 29.7% (4.0%) younger than 25 years, 26.0% (4.8%) aged 25 to 44 years, 14.8% (2.1%) aged 45 to 54 years, 13.4% (2.1%) aged 55 to 64 years, and 16.1% (4.4%) aged 65 years or older. Postoverdose programs were implemented in 58 municipalities (62%). Following implementation, there were no significant level changes in opioid fatality rate (adjusted rate ratio [aRR], 1.07; 95% CI, 0.96-1.19; P = .20). However, there was a significant slope decrease in opioid fatality rate (annualized aRR, 0.94; 95% CI, 0.90-0.98; P = .003) compared with the municipalities without the outreach programs. Similarly, there was a significant slope decrease in opioid-related EMS response rates (annualized aRR, 0.93; 95% CI, 0.89-0.98; P = .007). Several sensitivity analyses yielded similar findings.

Conclusions and Relevance

In this study, among Massachusetts municipalities with high numbers of opioid-related EMS responses, implementation of postoverdose outreach programs was significantly associated with lower opioid fatality rates over time compared with municipalities that did not implement such programs. Program components, including cross-sectoral partnerships, operational best practices, involvement of law enforcement, and related program costs, warrant further evaluation to enhance effectiveness.

Economic Assessment of High-Dose Versus Adjuvanted Influenza Vaccine: An Evaluation of Hospitalization Costs Based on a Cohort Study

Two influenza vaccines are licensed in the U.S. exclusively for the 65 years and older population: a trivalent inactivated high-dose influenza vaccine (HD-IIV3) and a trivalent inactivated adjuvanted influenza vaccine (aIIV3). In a recent publication, we estimated a relative vaccine effectiveness (rVE) of HD-IIV3 vs. aIIV3 of 12% (95% CI: 3.3–20%) for influenza-related hospitalizations using a retrospective study design, but did not report the number of prevented hospitalizations nor the associated avoided cost. In this paper we report estimations for both. Methods: Leveraging the rVE of a cohort study over two influenza seasons (2016/17 and 2017/18), we collected cost data for healthcare provided to the same study population. Vaccine costs were obtained from the Medicare pricing schedule. Our economic assessment compared cost of vaccination and hospital care for patients experiencing acute respiratory or cardiovascular illness. Results: We analyzed 1.9 million HD-IIV3 and 223,793 aIIV3 recipients. Average vaccine list prices were $46.23 for HD-IIV3 and $48.26 for aIIV3. The hospitalization rates for respiratory disease in HD-IIV3 and aIIV3 recipients were 187 (95% CI: 185–189) and 212 (195–231) per 10,000 persons-years, respectively. Attributing the average cost per hospitalization of $12,652 ($12,214–$13,090) to the difference in hospitalization rates, we estimate net savings of HD-IIV3 to be $34 ($10–$62) per recipient. Conclusion: Pooled over two predominantly A/H3N2 respiratory seasons, vaccination with HD-IIV3 was associated with lower hospitalization rates and associated costs compared to aIIV3 in senior members of a large national managed health care company in the U.S. Reduced hospitalizations affect healthcare utilization overall, and therefore other costly health outcomes.