Comparing the high-dimensional propensity score for use with administrative data with propensity scores derived from high-quality clinical data

The use of gabapentinoids and opioids and risk of developing opioid-induced respiratory depression among older breast cancer survivors with neuropathic pain

PURPOSE

To estimate the combined effect of gabapentinoid and opioid therapy compared to opioid monotherapy on the risk of developing opioid-induced respiratory depression among breast cancer survivors with neuropathic pain.

METHOD

A nested case-control study of Medicare female breast cancer survivors with neuropathic pain receiving both opioids and gabapentinoids, opioid monotherapy, gabapentinoid monotherapy, and none of these drugs was conducted using SEER-Medicare between 2007 and 2015. Cases were survivors with respiratory depression and were matched with controls on the event date (± 1 year), age at diagnosis (± 5 years), and stage at diagnosis. Exposure to opioids and gabapentinoids was assessed 120 days before the event date. Conditional logistic regression was used to assess the impact of exposure among cases and controls.

RESULTS

A total of 657 cases and 11,471 controls were identified. After matching, 656 cases and 5612 controls were retained, and cases were more likely to be diagnosed with mental health disorders (24.4% vs 10.5%, p < 0.0001) than controls. In the primary adjusted analysis, combined opioids and gabapentin use were associated with an increased risk of respiratory depression compared to opioid monotherapy (Adj. OR: 1.513; 95% CI: 1.473-2.350). Additionally, under secondary analysis, combined opioids and gabapentin use were associated with an increased risk of respiratory depression compared to receiving neither of these classes. (Adj. OR: 1.595; 95% CI: 1.050-2.421).

CONCLUSION

There is a need for dose titration strategies of gabapentinoids and caution when co-prescribing opioids and gabapentinoids in older cancer survivors.

A methodological review of the high-dimensional propensity score in comparative-effectiveness and safety-of-interventions research finds incomplete reporting relative to algorithm development and robustness

OBJECTIVES

The use of secondary databases has become popular for evaluating the effectiveness and safety of interventions in real-life settings. However, the absence of important confounders in these databases is challenging. To address this issue, the high-dimensional propensity score (hdPS) algorithm was developed in 2009. This algorithm uses proxy variables for mitigating confounding by combining information available across several healthcare dimensions. This study assessed the methodology and reporting of the hdPS in comparative effectiveness and safety research.

STUDY DESIGN AND SETTING

In this methodological review, we searched PubMed and Google Scholar from July 2009 to May 2022 for studies that used the hdPS for evaluating the effectiveness or safety of healthcare interventions. Two reviewers independently extracted study characteristics and assessed how the hdPS was applied and reported. Risk of bias was evaluated with the Risk Of Bias In Non-randomised Studies - of Interventions (ROBINS-I) tool.

RESULTS

In total, 136 studies met the inclusion criteria; the median publication year was 2018 (Q1-Q3 2016-2020). The studies included 192 datasets, mostly North American databases (n = 132, 69%). The hdPS was used in primary analysis in 120 studies (88%). Dimensions were defined in 101 studies (74%), with a median of 5 (Q1-Q3 4-6) dimensions included. A median of 500 (Q1-Q3 200-500) empirically identified covariates were selected. Regarding hdPS reporting, only 11 studies (8%) reported all recommended items. Most studies (n = 81, 60%) had a moderate overall risk of bias.

CONCLUSION

There is room for improvement in the reporting of hdPS studies, especially regarding the transparency of methodological choices that underpin the construction of the hdPS.

Weight telemonitoring of heart failure versus standard of care in a real-world setting: Results on mortality and hospitalizations in a 6-month nationwide matched cohort study

AIMS

Evaluating the benefit of telemonitoring in heart failure (HF) management in real-world settings is crucial for optimizing the healthcare pathway. The aim of this study was to assess the association between a 6-month application of the telemonitoring solution Chronic Care Connect™ (CCC) and mortality, HF hospitalizations, and associated costs compared with standard of care (SOC) in patients with a diagnosis of HF.

METHODS AND RESULTS

From February 2018 to March 2020, a retrospective cohort study was conducted using the largest healthcare insurance system claims database in France (Système National des Données de Santé) linked to the CCC telemonitoring database of adult patients with an ICD-10-coded diagnosis of HF. Patients from the telemonitoring group were matched with up to two patients from the SOC group based on their high-dimensional propensity score, without replacement, using the nearest-neighbour method. A total of 1358 telemonitored patients were matched to 2456 SOC patients. The cohorts consisted of high-risk patients with median times from last HF hospitalization to index date of 17.0 (interquartile range: 7.0-66.0) days for the telemonitoring group and 27.0 (15.0-70.0) days for the SOC group. After 6 months, telemonitoring was associated with mortality risk reduction (hazard ratio [HR] 0.71, 95% confidence interval [CI] 0.56-0.89), a higher risk of first HF hospitalization (HR 1.81, 95% CI 1.55-2.13), and higher HF healthcare costs (relative cost 1.38, 95% CI 1.26-1.51). Compared with the SOC group, the telemonitoring group experienced a shorter average length of overnight HF hospitalization and fewer emergency visits preceding HF hospitalizations.

CONCLUSION

The results of this nationwide cohort study highlight a valuable role for telemonitoring solutions such as CCC in the management of high-risk HF patients. However, for telemonitoring solutions based on weight and symptoms, consideration should be given to implement additional methods of assessment to recognize imminent worsening of HF, such as impedance changes, as a way to reduce mortality risk and the need for HF hospitalizations. Further studies are warranted to refine selection of patients who could benefit from a telemonitoring system and to confirm long-term benefits in high-risk and stable HF patients.

Economic and Clinical Burden of Herpes Zoster Among Patients With Inflammatory Bowel Disease in the United States

Background

Patients with ulcerative colitis (UC) or Crohn's disease (CD) are at increased risk of herpes zoster (HZ); however, relevant cost and healthcare resource utilization (HCRU) data are limited.

Methods

We estimated HCRU (hospitalization, emergency department [ED], and outpatient visits) and costs in patients with UC or CD, with and without HZ, using administrative claims data (October 2015-February 2020). HCRU and costs (2020 US dollars) were compared at 1 month, 1 quarter, and 1 year after the index date, using propensity score adjustment and generalized linear models.

Results

In total, 20 948 patients were included: UC+/HZ+ (n = 431), UC+/HZ- (n = 10 285), CD+/HZ+ (n = 435), and CD+/HZ- (n = 9797). Patients with HZ had higher all-cause HCRU rates and all-cause total healthcare costs relative to those without HZ. In the first month, adjusted incidence rate ratios (aIRRs) for hospitalizations and ED visits for patients with UC and HZ compared with UC alone were 2.87 (95% confidence interval [CI], 1.93-4.27) and 2.66 (95% CI,1.74-4.05), respectively; for those with CD and HZ, aIRRs were 3.34 (95% CI, 2.38-4.70) and 3.31 (95% CI, 2.32-4.71), respectively, compared with CD alone (all P < .001). Adjusted cost differences in UC and CD cohorts with HZ over the first month were $2189 and $3774, respectively, chiefly driven by higher inpatient costs. The incremental impact on HCRU and costs in cohorts with HZ predominantly occurred during the first quarter following diagnosis.

Conclusions

HZ is associated with increased HCRU and costs in patients with UC and CD, especially shortly after diagnosis.

Evaluating the Performance of High-Dimensional Propensity Scores Compared with Standard Propensity Scores for Comparing Antihypertensive Therapies in the CPRD GOLD Database

INTRODUCTION

Propensity score (PS) matching is widely used in medical record studies to create balanced treatment groups, but relies on prior knowledge of confounding factors. High-dimensional PS (hdPS) is a semi-automated algorithm that selects variables with the highest potential for confounding from medical databases. The objective of this study was to evaluate performance of hdPS and PS when used to compare antihypertensive therapies in the UK clinical practice research datalink (CPRD) GOLD database.

METHODS

Patients initiating antihypertensive treatment with either monotherapy or bitherapy were extracted from the CPRD GOLD database. Simulated datasets were generated using plasmode simulations with a marginal hazard ratio (HRm) of 1.29 for bitherapy versus monotherapy for reaching blood pressure control at 3 months. Either 16 or 36 known covariates were forced into the PS and hdPS models, and 200 additional variables were automatically selected for hdPS. Sensitivity analyses were conducted to assess the impact of removing known confounders from the database on hdPS performance.

RESULTS

With 36 known covariates, the estimated HRm (RMSE) was 1.31 (0.05) for hdPS and 1.30 (0.04) for PS matching; the crude HR was 0.68 (0.61). Using 16 known covariates, the estimated HRm (RMSE) was 1.23 (0.10) and 1.09 (0.20) for hdPS and PS, respectively. Performance of hdPS was not compromised when known confounders were removed from the database.

RESULTS ON REAL DATA

With 49 investigator-selected covariates, the HR was 1.18 (95% CI 1.10; 1.26) for PS and 1.33 (95% CI 1.22; 1.46) for hdPS. Both methods yielded the same conclusion, suggesting superiority of bitherapy over monotherapy for time to blood pressure control.

CONCLUSION

HdPS can identify proxies for missing confounders, thereby having an advantage over PS in case of unobserved covariates. Both PS and hdPS showed superiority of bitherapy over monotherapy for reaching blood pressure control.

Restricted Mean Survival Time Analysis to Estimate SGLT2i-Associated Heterogeneous Treatment Effects on Primary and Secondary Prevention of Cardiorenal Outcomes in Patients With Type 2 Diabetes in Taiwan

IMPORTANCE

Increasing numbers of post hoc analyses have applied restricted mean survival time (RMST) analysis on the aggregated-level data from clinical trials to report treatment effects, but studies that use individual-level claims data are needed to determine the feasibility of RMST analysis for quantifying treatment effects among patients with type 2 diabetes in routine clinical settings.

OBJECTIVES

To apply RMST analysis for assessing sodium-glucose cotransporter-2 inhibitor (SGLT2i)-associated cardiovascular (CV) events and estimating heterogenous treatment effects (HTEs) on CV and kidney outcomes in routine clinical settings.

DESIGN, SETTING, AND PARTICIPANTS

This comparative effectiveness study of Taiwan's National Health Insurance Research Database examined 21 144 propensity score (PS)-matched pairs of patients with type 2 diabetes with SGLT2i and dipeptidyl peptidase-4 inhibitor (DPP4i) treatment for assessing CV outcomes, and 19 951 PS-matched pairs of patients with type 2 diabetes with SGLT2i and DPP4i treatment for assessing kidney outcomes. Patients were followed until December 31, 2018. Statistical analysis was performed from August 2021 to April 2022.

EXPOSURES

Newly stable SGLT2i or DPP4i use in 2017.

MAIN OUTCOMES AND MEASURES

Study outcomes were CV events including hospitalization for heart failure (HHF), 3-point major adverse CV events (3P-MACE: nonfatal myocardial infarction [MI], nonfatal stroke, and CV death), 4-point MACE (4P-MACE: HHF and 3P-MACE), and all-cause death, and chronic kidney disease (CKD). RMST and Cox modeling analyses were applied to estimate treatment effects on study outcomes.

RESULTS

After PS matching, the baseline patient characteristics were comparable between 21 144 patients with stable SGLT2i use (eg, mean [SD] age: 58.3 [10.7] years; 11 990 [56.7%] male) and 21 144 patients with stable DPP4i use (eg, mean [SD] age: 58.1 [11.6] years; 12 163 [57.5%] male) for assessing CV outcomes, and those were also comparable between 19 951 patients with stable SGLT2i use (eg, mean [SD] age: 58.1 [10.7] years; 11 231 [56.2%] male) and 19 951 patients with stable DPP4i use (eg, mean [SD] age: 57.9 [11.5] years; 11 340 [56.8%] male) for assessing kidney outcome. The 2-year difference in RMST between patients with SGLT2i use and patients with DPP4i use was 4.99 (95% CI, 3.56-6.42) days for HHF, 4.12 (95% CI, 2.72-5.52) days for 3P-MACE, 7.72 (95% CI, 5.83-9.61) days for 4P-MACE, 1.26 (95% CI, 0.47-2.04) days for MI, 2.70 (95% CI, 1.57-3.82) days for stroke, 0.69 (95% CI, 0.28-1.11) days for CV death, 6.05 (95% CI, 4.89-7.20) days for all-cause death, and 14.75 (95% CI, 12.99-16.52) days for CKD. Directions of hazard ratios from Cox modeling analyses were consistent with RMST estimates. No association was found between study treatment and the negative control outcome (dental visits for tooth care). Consistent results across sensitivity analyses using high-dimensional PS-matched and PS-weighting approaches supported the validity of primary analysis results. Largest difference in RMST of SGLT2i vs DPP4i use for HHF and CKD was found among patients with established heart failure (30.80 [95% CI, 5.08-56.51] days) and retinopathy (40.43 [95% CI, 31.74-49.13] days), respectively.

CONCLUSIONS AND RELEVANCE

In this comparative effectiveness study, RMST analysis was feasible for translating treatment effects into more clinical intuitive estimates and valuable for quantifying HTEs among diverse patients in routine clinical settings.

Is acid suppression associated with the increased length of stay in diabetic ketoacidosis patients? A nationwide analysis

BACKGROUND

Diabetic ketoacidosis (DKA) patients present with low serum bicarbonate ( HCO3-${\rm{HCO}}_{3}^{-} $ ), and an increase in its level to ≥15 mEq/L is considered one of the criteria for DKA resolution. Both proton pump inhibitors and histamine-2 receptor antagonists inhibit downstream functioning of H⁺ /K⁺ ATPase in the gastric parietal cells, which results in the decreased secretion of HCO3-${\rm{HCO}}_{3}^{-} $ into the bloodstream.

OBJECTIVES

We aimed to introduce the hypothesis that DKA patients on acid-suppressive medications may have a delayed rise in serum HCO3-${\rm{HCO}}_{3}^{-} $ to >15 mEq/L that may cause increased hospital length of stay (LOS) and sought to compare the outcomes of such patients. For the sake of simplicity, conditions requiring acid suppression are grouped under the term peptic ulcer disease (PUD) in this study.

METHODS

This is a retrospective study using Nationwide Inpatient Sample employing International Classification of Diseases (ICD-10) codes for adult patients with a primary diagnosis of DKA. Analyses were performed using STATA, proportions were compared using Fisher exact test, and continuous variables using Student's t-test. Confounding variables were adjusted using propensity matching, multivariate logistic, and linear regression analyses.

RESULTS

DKA patients with PUD had higher adjusted LOS, intensive care unit admission, and total hospital costs. Mortality and morbidity indicators were similar in both groups. The variables found to be independent predictors of increased LOS were malnutrition, Clostridium difficile infection, pneumonia, Glasgow Coma Scale score of 3-8, and higher Charlson comorbidity score.

CONCLUSION

We found that Clostridium difficile and pneumonia predicted longer LOS in DKA patients with concomitant PUD, hinting at the possible role of acid suppression in prolonging the LOS in such patients. However, further studies are needed to examine the effect of lesser HCO3-${\rm{HCO}}_{3}^{-} $ generation on LOS.

Machine learning for improving high-dimensional proxy confounder adjustment in healthcare database studies: an overview of the current literature

Controlling for large numbers of variables that collectively serve as 'proxies' for unmeasured factors can often improve confounding control in pharmacoepidemiologic studies utilizing administrative healthcare databases. There is a growing body of evidence showing that data-driven machine learning algorithms for high-dimensional proxy confounder adjustment can supplement investigator-specified variables to improve confounding control compared to adjustment based on investigator-specified variables alone. Consequently, there has been a recent focus on the development of data-driven methods for high-dimensional proxy confounder adjustment. In this paper, we discuss the considerations underpinning three areas for data-driven high-dimensional proxy confounder adjustment: 1) feature generation-transforming raw data into covariates (or features) to be used for proxy adjustment; 2) covariate prioritization, selection and adjustment; and 3) diagnostic assessment. We survey current approaches and recent advancements within each area, including the most widely used approach to proxy confounder adjustment in healthcare database studies (the high-dimensional propensity score or hdPS). We also discuss limitations of the hdPS and outline recent advancements that incorporate the principles of proxy adjustment with machine learning extensions to improve performance. We further discuss challenges and avenues of future development within each area. This manuscript is endorsed by the International Society for Pharmacoepidemiology (ISPE). This article is protected by copyright. All rights reserved.

Transparency of high-dimensional propensity score analyses: guidance for diagnostics and reporting

Purpose

The high‐dimensional propensity score (HDPS) is a semi‐automated procedure for confounder identification, prioritisation and adjustment in large healthcare databases that requires investigators to specify data dimensions, prioritisation strategy and tuning parameters. In practice, reporting of these decisions is inconsistent and this can undermine the transparency, and reproducibility of results obtained. We illustrate reporting tools, graphical displays and sensitivity analyses to increase transparency and facilitate evaluation of the robustness of analyses involving HDPS.

Methods

Using a study from the UK Clinical Practice Research Datalink that implemented HDPS we demonstrate the application of the proposed recommendations.

Results

We identify seven considerations surrounding the implementation of HDPS, such as the identification of data dimensions, method for code prioritisation and number of variables selected. Graphical diagnostic tools include assessing the balance of key confounders before and after adjusting for empirically selected HDPS covariates and the identification of potentially influential covariates. Sensitivity analyses include varying the number of covariates selected and assessing the impact of covariates behaving empirically as instrumental variables. In our example, results were robust to both the number of covariates selected and the inclusion of potentially influential covariates. Furthermore, our HDPS models achieved good balance in key confounders.

Conclusions

The data‐adaptive approach of HDPS and the resulting benefits have led to its popularity as a method for confounder adjustment in pharmacoepidemiological studies. Reporting of HDPS analyses in practice may be improved by the considerations and tools proposed here to increase the transparency and reproducibility of study results.

Statin Discontinuation and Cardiovascular Events Among Older People in Denmark

Importance

Statin use is common in older persons. Given uncertainties in ongoing benefit, changes in health status, and shifting goals of care and preferences, statin discontinuation may be considered in some older persons, although there is currently little evidence to guide this decision.

Objective

To evaluate the association between statin discontinuation and the rate of major adverse cardiovascular events (MACE) among people aged 75 years or older who receive long-term statin treatment.

Design, Setting, and Participants

This cohort study included all persons in Denmark aged 75 years or older who were treated with statins for at least 5 consecutive years as of January 1, 2011. Participants were followed up until December 31, 2016. Data were analyzed from July to November, 2020.

Exposure

Statin discontinuation.

Main Outcomes and Measures

Rate of occurrence of MACE and its components (myocardial infarction, ischemic stroke or transient ischemic attack, coronary revascularization, and death due to myocardial infarction or ischemic stroke) in persons continuing statins compared with those discontinuing statins. Confounding adjustment was done using inverse probability of treatment weighting. Analyses were conducted separately for primary prevention (no history of cardiovascular disease) and secondary prevention (history of cardiovascular disease).

Results

The study included 67 418 long-term statin users, including 27 463 in the primary prevention analysis (median age, 79 years [IQR, 77-83 years]; 18 134 [66%] female) and 39 955 in the secondary prevention analysis (median age, 80 years [IQR, 77-84 years]; 18 717 [47%] female). In both primary and secondary prevention analyses, the rate of MACE was higher among persons who discontinued statins compared with those who continued statins. In the primary prevention cohort, the weighted rate difference was 9 per 1000 person-years (95% CI, 5-12 per 1000 person-years) and the adjusted sub-hazard ratio was 1.32 (95% CI, 1.18-1.48), corresponding to 1 excess MACE per 112 persons who discontinued statins per year. In the secondary prevention cohort, the weighted rate difference was 13 per 1000 person-years (95% CI, 8-17 per 1000 person-years) and the adjusted sub-hazard ratio was 1.28 (95% CI, 1.18-1.39), corresponding to 1 excess MACE per 77 persons who discontinued statins per year.

Conclusions and Relevance

In this cohort study, among older adults receiving long-term statin treatment, discontinuation of statins was associated with a higher rate of MACE compared with statin continuation in both the primary and the secondary prevention cohorts. These findings suggest a need for robust evidence from randomized clinical trials.

Change in healthcare during Covid-19 pandemic was assessed through observational designs

Objective

Methodological challenges for investigating the changes in healthcare utilization during COVID-19 pandemic must be considered for obtaining unbiased estimates.

Study design and setting

A population-based study in the Lombardy region (Italy) measured the association between the level of epidemic restrictions (increasing exposure during pre-epidemic, post-lockdown, and lockdown periods) and the recommended healthcare (outcome) for patients with schizophrenia, heart failure, chronic obstructive pulmonary disease, breast cancer, and pregnancy women. Two designs are applied: the self-controlled case series (SCCS) and the usual cohort design. Adjustments for between-patients unmeasured confounders and seasonality of medical services delivering were performed.

Results

Compared with pre-epidemic, reductions in delivering recommended healthcare during lockdown up to 73% (95% confidence interval: 63%–80%) for timeliness of breast cancer surgery, and up to 20% (16%–23%) for appropriated gynecologic visit during pregnancy were obtained from SCCS and cohort design, respectively. Healthcare provision came back to pre-epidemic levels during the post-lockdown, with the exception of schizophrenic patients for whom the SCCS showed a reduction in continuity of care of 11% (11%–12%).

Conclusion

Strategies for investigating the changes in healthcare utilization during pandemic must be implemented. Recommendations for taking into account sources of systematic uncertainty are discussed and illustrated by using motivating examples.

Comparative effectiveness of buprenorphine-naloxone versus methadone for treatment of opioid use disorder: a population-based observational study protocol in British Columbia, Canada

INTRODUCTION

Despite a recent meta-analysis including 31 randomised controlled trials comparing methadone and buprenorphine for the treatment of opioid use disorder, important knowledge gaps remain regarding the long-term effectiveness of different treatment modalities across individuals, including rigorously collected data on retention rates and other treatment outcomes. Evidence from real-world data represents a valuable opportunity to improve personalised treatment and patient-centred guidelines for vulnerable populations and inform strategies to reduce opioid-related mortality. Our objective is to determine the comparative effectiveness of methadone versus buprenorphine/naloxone, both overall and within key populations, in a setting where both medications are simultaneously available in office-based practices and specialised clinics.

METHODS AND ANALYSIS

We propose a retrospective cohort study of all adults living in British Columbia receiving opioid agonist treatment (OAT) with methadone or buprenorphine/naloxone between 1 January 2008 and 30 September 2018. The study will draw on seven linked population-level administrative databases. The primary outcomes include retention in OAT and all-cause mortality. We will determine the effectiveness of buprenorphine/naloxone vs methadone using intention-to-treat and per-protocol analyses-the former emulating flexible-dose trials and the latter focusing on the comparison of the two medication regimens offered at the optimal dose. Sensitivity analyses will be used to assess the robustness of results to heterogeneity in the patient population and threats to internal validity.

ETHICS AND DISSEMINATION

The protocol, cohort creation and analysis plan have been approved and classified as a quality improvement initiative exempt from ethical review (Providence Health Care Research Institute and the Simon Fraser University Office of Research Ethics). Dissemination is planned via conferences and publications, and through direct engagement and collaboration with entities that issue clinical guidelines, such as professional medical societies and public health organisations.