Phdthesis lmoura 20210907 final

Comparative Safety and Effectiveness of Anticonvulsants Among Older Adults Citation Moura, Lidia Maria Veras Rocha de. Comparative Safety and Effectiveness of Anticonvulsants Among Older Adults. Doctoral dissertation, Harvard University Graduate School of Arts and Sciences. Permanent link https://nrs.edu/URN-3:HUL.INSTREPOS:37370172 Terms of Use This article was downloaded from Harvard University’s DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at http:// nrs.edu/urn-3:HUL.terms-of-use#LAA Share Your Story The Harvard community has made this article openly available.

Please share how this access benefits you. Accessibility Title Page Comparative Safety and Effectiveness of Anticonvulsants Among Older Adults A dissertation presented By Lídia Maria Veras Rocha de Moura to The Department of Epidemiology Harvard T. Chan School of Public Health in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the subject of Population Health Sciences Harvard University Cambridge, Massachusetts August 2021 Copyright ã2021 Lidia Maria Veras Rocha de Moura All rights reserved Dissertation Advisor: Author: Sonia Hernández-Díaz Lídia Maria Veras Rocha de Moura Comparative Safety and Effectiveness of Anticonvulsants Among Older Adults Abstract Anticonvulsant use among older adults has increased 200-fold over the past 25 years.1 In our previous work, we estimated that one in eight older adults in the US are currently receiving anticonvulsants.2 The risk of adverse events approaches 20%, and 5% of them are potentially life- threatening.2-4 Older adults with Alzheimer’s disease and Alzheimer’s-related dementias (AD/ADRD) are more susceptible to these adverse effects.5-10 The most common indications for anticonvulsants are seizure prophylaxis, insomnia, pain, and agitation, which are all common conditions among older adults who survive an Acute Ischemic Stroke (AIS).11,12 In the US alone there are 7 million stroke survivors, and 800,000 new or recurrent stroke cases annually.11,13-15 The increasing in-hospital use of neurophysiology studies has led to greater detection of epileptiform abnormalities of uncertain clinical significance, greater perceived seizure risk, and increasing prescriptions of seizure prophylaxis with drugs like levetiracetam or phenytoin.16-19 Additionally, the rate of prescriptions of benzodiazepines for patients with post-AIS insomnia, periprocedural anxiety, and delirium (often manifested with agitation) continues to rise.2,20-23 In the meantime, major medical and psychiatric organizations suggest avoidance of anticonvulsants in those ≥65 years.2,20-28 Rigorous evidence is urgently needed about the effectiveness and safety of these drugs for older adults.5-10 However, randomized trials arguably are impractical because of large numbers needed for safety studies, the high cost of such studies, and difficulty recruiting older adults into clinical trials. Prior iii observational studies have severe limitations: small samples, poor data granularity and validity, selection bias, and confounding by indication.

Many factors may be associated with anticonvulsant initiation, including socio-demographic, clinical, and health-care utilization characteristics.29-32 To overcome these challenges, we linked data from a prospective stroke registry (Get-With-The Guidelines-Stroke) to comprehensive electronic health records (EHRs) and applied novel statistical methods to evaluate the effects of treatment initiation strategies among adults ≥65 years who are survivors of an AIS on adverse health outcomes, initially focusing on mortality in the next months. Moreover, to enable later work toward this overall goal in national Medicare data, we conducted a validation study linking Medicare claims and EHRs to optimize claims-based definitions of AD/ADRD, an important stratification measure. In Chapter 1, we examined the accuracy of claims-based diagnoses of AD/ADRD with respect to expert clinician adjudication using a novel database with individual-level linkages between electronic health record (EHR) and claims. A model that regressed reference standard dementia on a refined-count definition (i., the number of days with a qualifying diagnostic claims), and included age and sex, had excellent calibration to the reference standard with calibration-in-the-large [CV CITL] <0.001 and calibration slope of 0.

The adjudicated prevalence of dementia was 7% (n=2854, 95%CI: 6%-9%) in the target population. Using a predicted probability threshold of 0.5 to classify dementia, that model demonstrated negative predictive value of 90.5%), positive predictive value of 77.8%), and specificity of 93. Next, we linked the prospective stroke registry to EHRs and examined the comparative effectiveness and safety of benzodiazepines and seizure prophylaxis initiation strategies among older adults post-AIS. In Chapter 2, we focused on benzodiazepine prescriptions, which are typically short-term (i., one or few doses) used to treat insomnia and agitation, and may cause serious adverse effects within few hours of initiation.

In Chapter 3, we switched the focus to examine anticonvulsants given for seizure prophylaxis, which are often prescribed for 30-90 days or more, and may cause serious adverse effects within days or weeks of initiation. iv Specifically, in Chapter 2, we examined the difference in 30-day mortality risk associated with initiating benzodiazepines within seven days after an AIS compared with no benzodiazepines among patients ≥65 years. We concluded that in the overall population the excess in mortality associated with initiating benzodiazepines was largely due to confounding by indication. Excess deaths per 1,000 stratified by age ranged from 2 (95% CI -16 to 12), to 12 (95% CI -9 to 143), to 20 (95% CI -43 to 81) among patients 65-74, 75-84 years, and ≥ 85 years, respectively.

The RD was 1 (CI, -3 to 9) among patients with mild stroke (NIHSS <5) and 18 (CI, -85 to 148) among those with moderate-to-severe stroke (NIHSS 16-20). In Chapter 3, we examined the 90-day mortality risk associated with initiation of epilepsy-specific seizure prophylaxis within seven days after an AIS in patients ≥65 years. After standardization, the estimated mortality was 127 (95% CI, 64 to 256) and 65 (95% CI, 40 to 89) deaths per 1000 in the two strategies, corresponding to a risk difference (RD) of 62 (95% CI, -8 to 198) excess deaths per 1000 patients and a hazard ratio (HR) of 1. Therefore, we found evidence of a potential excess in 90-day mortality associated with initiating seizure prophylaxis within seven days post-AIS compared with not initiating seizure prophylaxis in the overall population.

However, the estimates were imprecise and residual confounding by indication remained a concern. In the prespecified subgroup analyses, the RD was 5 (CI, -8 to 10) among patients with minor and 88 (CI, -17 to 288) among those with moderate or severe AIS. Similarly, the RDs were 27 (95% CI -5 to 83) and 92 (95% CI -5 to 239) among patients aged 65-74 years and ≥ 75 years, respectively. The HR were similar across age and severity groups.

Our work highlights the urgent need to discuss whether older patients would benefit from or be harmed by initiation of anticonvulsant prophylaxis during the acute and post-acute stroke recovery periods. In future steps, we will examine the distribution of anticonvulsant specific outcomes as indicators of benefits (effectiveness) and harms (adverse reactions,), such as the incidence of seizure-like events and fall-related injuries and seizure-like events stratifying by care settings. We also will examine longer- term risks and benefits of other treatment indications and strategies including the comparative effectiveness and safety of specific anticonvulsants in the elderly. v Table of Contents Title Page.

iii Table of Contents. vi List of Tables. viii List of Figures. 1 Chapter 1 - Identifying Medicare Beneficiaries with Dementia: Validated Claims-Based Definition.

26 Chapter 2 - Short-term Mortality and Benzodiazepine Use After Acute Ischemic Stroke in Elderly. 53 Chapter 3 - Seizure Prophylaxis and Short-Term Outcomes after Acute Ischemic Stroke. Supplementary Materials for Chapter 1. Supplementary Materials for Chapter 2.

Supplementary Materials for Chapter 3. 201 vii List of Tables Table 1. Population Characteristics of Reconstructed Sample. Performance Characteristics: Logistic Model of Diagnostic Claim Count, Refined Claims-Based Definition (Model 7), Overall and by Age Group.

Description of A Target Trial and the Corresponding Observational Study. Characteristics of patients, by benzodiazepine initiator versus non-benzodiazepine initiator. Description of A Target Trial and The Corresponding Observational Study. Characteristics of patients by anticonvulsant exposure.

EHR-Adjudicated Dementia Status of Analytic Sample. ICD-10 Diagnostic Codes, removed from and Added to Refined Claims-Based Definition of Dementia (From Base Definition). Performance Characteristics: Ascertainment of EHR-Adjudicated Dementia by Base Claims- Based Definition. Logistic Regression Results: Ascertainment of EHR-Adjudicated Dementia by Refined Claims- Based Definition.

Logistic Model of Diagnostic Claim Count, Refined Claims-Based Definition. CV AUC from Models 2-7, compared with CV AUC from Model 1 (Base Model). Sensitivity Analysis; Logistic Model of Diagnostic Claim Count, Refined Claims-Based Definition, and Adjusted Reference Standard (Dementia and Mild Cognitive Impairment versus Dementia), Overall. Logistic Regression Results and Performance Characteristics, Claims-Based Definition with Removal of Nonspecific Diagnoses, Only.

Logistic Regression Results and Performance Characteristics, Claims-Based Definition with Removal of Reversible Etiologies of Altered Mental Status, Only. Logistic Model Results: Ascertainment of EHR-Adjudicated Dementia with “High or Moderate” Diagnostic Certainty by Base Claims-Based Definition. Logistic Model Results: Ascertainment of EHR-Adjudicated Dementia with “High or Moderate or Mild” Diagnostic Certainty by Base Claims-Based Definition. Logistic Model Results: Ascertainment of EHR-Adjudicated Dementia with “High or Moderate” Diagnostic Certainty by Refined Claims-Based Definition.

Logistic Model Results: Ascertainment of EHR-Adjudicated Dementia with “High or Moderate or Mild” Diagnostic Certainty by Refined Claims-Based Definition. Logistic Model Results: Ascertainment of EHR-Adjudicated Dementia by Level of Diagnostic Certainty, Count-Based Refined Claims Definition. Performance Characteristics: Logistic Model of Diagnostic Claim Count, Refined Claims-Based Definition (Model 7), Overall and by Age Group, Using only 2018 Claims. Logistic Model of Diagnostic Claim Count, Refined Claims-Based Definition, Using only 2018 Claims.

Terms used in Systematic Search. Diagnosis of Mild Cognitive Impairment (MCI) criteria. Diagnosis of dementia criteria. Dementia Severity Criteria.

Delirium diagnostic Criteria. ICD-9 and ICD-10 for Key Covariates. Characteristics of patients, by benzodiazepine initiation, after standardization. Characteristics of Patients, restricted to Minor Stroke Severity.

Characteristics of Patients, restricted to Moderate Stroke Severity. Characteristics of Patients, restricted to Moderate-to-Severe Stroke Severity. Description of Benzodiazepine Types. Pattern of Anticonvulsant Use Post-Benzodiazepine Initiation.

Mortality Distribution of the Sample Stratified by Exposure Definitions. Model Parameters for Estimating Benzodiazepine Initiation Weights. Main Estimates Aggregated from Datasets 1-5, Based on Rubin’s Rules. Main Results, Stratified by Stroke Severity.

Main Estimates, Stratified by Age Groups. The RECORD Statement. List of Epilepsy-Specific Anticonvulsants. ICD-9 and ICD-10 for Fall-Related Injuries (FRIs).

Mortality Distribution of the Eligible Sample Stratified by Exposure Definitions. Main Standardized Estimates for Stratified Sample - Stroke Severity. Main Standardized Estimates for Stratified Sample - Age Groups. The RECORD Statement.

195 x List of Figures Figure 1. Overview of Sampling Approach. Overlap of Claims-Based Definitions with EHR-Adjudicated Dementia, Overall and by Age. Area Under ROC Curve: Tradeoff Between Sensitivity and Specificity.

Selection of Eligible Patients with New Acute Ischemic Stroke (AIS) ≥65 years, 2014-2020. Crude and Standardized Survival Curves by Benzodiazepine Initiation Strategy. Standardized Survival Curves across Categories of Stroke Severity. Selection of eligible patients with new acute ischemic stroke (AIS) ≥65 years, 2014-2020.

Crude and Standardized Survival Curves by Seizure Prophylaxis Strategy during the first 90 days post stroke admission. Standardized Survival Curves by Epilepsy-Specific Anticonvulsant Initiation Strategy Across Categories of Stroke Severity. Standardized Survival Curves by Epilepsy-Specific Anticonvulsant Initiation Strategy Across Categories of Age. CV Calibration Plot: Logistic Model of Diagnostic Claim Count, Refined Claims-Based Definition.

Logistic Model of Diagnostic Claim Count, Refined Claims-Based Definition. Simplistic Representation of the Problem and Solution. Methods for Multiple Imputation. Crude Survival Curves, Overall and Stratified by Benzodiazepine Initiation.

Standardized Survival Curves, Stratified by Age, by benzodiazepine initiation. Simplistic Representation of the Problem and Solution. Crude Survival Curves, Overall and Stratified. 186 xi Acknowledgments When I was 10 years old, I saw my mom rescue a drowned kid.

At that moment, I decided to become a medical doctor so I could save lives. However, I saw too many people dying because of limited healthcare access and quality of care problems during my initial training. There were times when we would have to decide who would go to an intensive care unit among the patients in need.