Comparison of Cardiovascular and Safety Outcomes of Chlorthalidone vs Hydrochlorothiazide to Treat Hypertension

This cohort study compares cardiovascular and safety outcomes of chlorthalidone and hydrochlorothiazide as first-line therapies for adults with hypertension.


Study Design
This study is a retrospective, observational, new-user cohort study. By 'retrospective' we mean the study uses data already collected at the start of the study. By 'observational' we mean no intervention takes place in the course of this study. By 'new-user' we mean we only analyze the first exposure of a subject to the treatment of interest. By 'cohort study' we mean two cohorts, a target and comparator cohort, are followed from index date (start of first exposure) to some end date, and assessed for the occurrence of the outcomes of interest.

Data Source(s)
The analyses are performed across a network of observational healthcare databases. All databases have been transformed into the OMOP Common Data Model, version 4 or OMOP Common Data Model, version 5. The complete specification for OMOP Common Data Model is available [2]. The following databases are included in this analysis: • IBM MarketScan Commercial Claims and Encounters (CCAE) • Optum ClinFormatics (Optum) • Optum® de-identified Electronic Health Record Dataset (PanTher)

Truven MarketScan Commercial Claims and Encounters (CCAE)
Truven Health MarketScan® Commercial Claims and Encounters Database (CCAE) represent data from individuals enrolled in United States employer-sponsored insurance health plans. The data includes adjudicated health insurance claims (e.g. inpatient, outpatient, and outpatient pharmacy) as well as enrollment data from large employers and health plans who provide private healthcare coverage to employees, their spouses, and dependents. Additionally, it captures laboratory tests for a subset of the covered lives. This administrative claims database includes a variety of fee-for-service, preferred provider organizations, and capitated health plans.

Optum ClinFormatics (Optum)
Optum Clinformatics Extended DataMart is an adjudicated US administrative health claims database for members of private health insurance, who are fully insured in commercial plans or in administrative services only (ASOs), Legacy Medicare Choice Lives (prior to January 2006), and Medicare Advantage (Medicare Advantage Prescription Drug coverage starting January 2006). The population is primarily representative of commercial claims patients (0-65 years old) with some Medicare (65+ years old) however ages are capped at 90 years. It includes data captured from administrative claims processed from inpatient and outpatient medical services and prescriptions as dispensed, as well as results for outpatient lab tests processed by large national lab vendors who participate in data exchange with Optum. This dataset also provides date of death (month and year only) for members with both medical and pharmacy coverage from the Social Security Death Master File (however after 2011 reporting frequency changed due to changes in reporting requirements) and location information for patients is at the US state level.

Study population
All subjects in the database will be included who meet the following criteria: (note: the index date is the start of the first treatment for hypertension) • Exposure to one of the treatments of interest • At least 365 days of observation time prior to the index date • No exposure of any hypertension treatment before the index date • A diagnosis of hypertensive disorder on or preceding the index date • No diagnosis of the outcome of interest preceding the index date For the purpose of determining previous hypertension treatment, the following medications were used:

Exposures and outcomes
The two study exposures were chlorthalidone and hydrochlorothiazide in any dose. A detailed definition of the exposures is given in Section 3.
The outcomes for effectiveness and safety were selected based on their relevance to hypertension treatment (not limited to chlorthalidone and hydrochlorothiazide) as part of the OHDSI LEGEND study. The effectiveness outcomes were acute myocardial infarction (AMI) (hospitalized), hospitalization for heart failure (HF), ischemic or hemorrhagic stroke (also hospitalized), and a composite cardiovascular disease outcome including the first three outcomes and sudden cardiac death. Fifty-five safety outcomes were included. All outcomes are enumerated and defined in Section 3.
Outcomes were selected based on being initially identified from endpoints from clinical trials reporting in clinical guidelines and systematic reviews, and augmented by adverse events listed on US structured product labels of hypertension drugs [3]. Each outcome was then assessed to determine if it was evaluable in observational data by developing and evaluating an operational phenotype definition. All phenotypes were designed, implemented and evaluated using the same approach: Initial design was guided by literature review in PubMed of prior observational studies of administrative claims or electronic health records where the phenotype was used as a study outcome, and if available, validated through source record verification or other means. Literature review was complemented by review of eMERGE PheKB phenotype entries [4]. Clinical definitions of cardiovascular outcomes were determined from clinical guidelines and systematic review of clinical trials of hypertension treatments [5][6][7]. Where possible, concept sets originated with published code lists (e.g. ICD-9-CM and ICD-10) and were augmented through lexical search and semantic exploration of the OHDSI standardized vocabularies. Cohort definitions and associated concept sets were then reviewed by a clinical adjudicator. Cohort definitions were developed using the OHDSI open-source platform ATLAS [8]. These definitions were initially executed across seven databases (the three from this study, CCAE, Optum, and PanTher, and four others: IBM MarketScan Medicare Supplemental Database, IBM MarketScan Multi-State Medicaid Database, Japan Medical Data Center Claims Database, IQVIA IMS Germany) to identify qualifying patients. Because these databases do not all consistently contain laboratory values, outcomes involving electrolyte imbalance (hypokalemia, hypomagnesemia, hyponatremia) were defined by diagnosis records only. These cohorts were used to produce characterization of outcome incidence, stratified by age decile, gender, and index year, to determine if outcomes were being observed in a frequency that was generally consistent across data sources and sufficiently aligned with epidemiologic expectations of the disease. No source record verification or other methods for validation was performed.

Negative control outcomes
Negative controls are concepts known to not be associated with the target or comparator cohorts, such that we can assume the true relative risk between the two cohorts is 1. Negative controls are selected using a similar process to that outlined by Voss et al. [9]. Person counts of all potential drug-condition pairs are reviewed in observational data; this person count data helps determine which pairs are even probable for use in calibration. Given the list of potential drug-condition pairs, the concepts in the pairs must meet the following requirements to be considered as negative controls: (1) that there is no Medline abstract where the MeSH terms suggest an association between the drug and the condition [10], (2) that there is no mention of the drug-condition pair on a US Product Label in the "Adverse Drug Reactions" or "Postmarketing" section [11], (3) there are no US spontaneous reports suggesting that the pair is in an adverse event relationship [12,13], (4) that the OMOP Vocabulary does not suggest that the drug is indicated for the condition, (5) that the concepts are usable (i.e. not too broad, not suggestive of an adverse event relationship, not pregnancy related), and (6) the exact concept itself is utilized in patient level data (i.e. concepts that are not usually used within the data are usually indicative a broad concept that has a child that is more specific). The remaining concepts are "optimized", meaning parent concepts remove children as defined by the OMOP Vocabulary (e.g. if both "Non-Hodgkin's Lymphoma" and "B-Cell Lymphoma" we selected, child concept "B-Cell Lymphoma would be removed for its parent "Non-Hodgkin's Lymphoma"). Once potential negative control candidates were selected, manual clinical review to exclude any pairs that may still be in a causal relationship or similar to the study outcome was be performed to select the top concepts by patient exposure. The final list can be found here: Abnormal  For each negative control outcome, a patient enters the negative control outcome cohort at the occurrence of a diagnose code identified by the concepts listed above, or any one of its descendant codes.

Positive control outcomes
In addition to negative control outcomes, we will also include synthetic positive control outcomes. These are outcomes based on the real negative controls, but where the true effect size is artificially increased to a desired effect size by injection of additional, simulated outcomes [14]. To preserve confounding, these additional outcomes are sampled from predicted probabilities generated using a fitted predictive model. For each negative control outcome, three positive control outcomes are generated with true relative risk is 1.5, 2, and 4. Using both negative and positive controls, we fit a systematic error model and perform confidence interval calibration [14].

Propensity score covariates
Propensity scores (PS) are used as an analytic strategy to reduce potential confounding due to imbalance between the target and comparator cohorts in baseline covariates. The propensity score is the probability of a patient being classified in the target cohort vs. the comparator cohort, given a set of observed covariates.
The types of baseline covariates used to fit the propensity score model will be: All covariates that occur in less than 0.1% of the persons between the target and comparator cohorts combined are excluded prior to model fitting for computational efficiency. Each condition, drug, class, etc. is counted as a separate covariate, resulting in over 60,000 covariates per database for this study.

Index event and calculation of time-at risk
The index event for this study is the first treatment with chlorthalidone or hydrochlorothiazide. It must be taken as a single anti-hypertensive agent, and no other anti-hypertensive agents may precede them. Subjects are thus new users of one of the two study drugs and new users of antihypertensive agents in general.
Each subject must be observed for at least 365 days before the index event. Baseline covariates are collected from this period up to but not including the index event.
Two time-at-risk periods are used: • On-treatment. Starting one day after the index event and stopping at the termination of the drug or on the addition of a second anti-hypertensive agent. A maximum gap of 30 days is allowed between prescriptions, and the treatment is considered to end after the last prescription runs out.
• Intent-to-treat (ITT): Starting one day after the index event and stopping at the end of observation in the database.

Model Specification
In this study, we compare the chlorthalidone (target) cohort with the hydrochlorothiazide (comparator) cohort for the hazards of outcome during the time-at-risk by applying a Cox proportional hazards model.
The time-to-event of outcome among patients in the target and comparator cohorts is determined by calculating the number of days from the start of the time-at-risk window (the cohort start date), until the earliest event among 1) the first occurrence of the outcome, 2) the end of the time-at-risk window as defined above (i.e. 'on-treatment' or 'intent-to-treat'), and 3) the end of the observation period that spans the time-at-risk start.
Patients with the outcome observed prior to target or comparator cohort entry are excluded from consideration.
We conduct our cohort study using the open-source OHDSI CohortMethod R package [15], whose large-scale analytics are achieved through the Cyclops R package [16]. We use propensity scores as an analytic strategy to reduce potential confounding due to imbalance between the target and comparator cohorts in baseline covariates [17]. The propensity score is the probability of a patient being classified in the target cohort vs. the comparator cohort, given a set of observed covariates. In this study, the propensity score is estimated for each patient, using the predicted probability from a regularized logistic regression model, fit with a Laplace prior (LASSO) and the regularization hyperparameter selected by optimizing the likelihood in a 10fold cross, a starting variance of 0.01 and a tolerance of 2e-7. Covariates to be used in the propensity score model are listed above; they include patient demographics, drug, condition, and procedure covariates generated through the FeatureExtraction R package [18]. The target cohort and comparator cohorts are stratified into ten quantiles of the propensity score distribution. The final outcome model applies a conditional Cox proportional hazard model, conditioned on the propensity score strata. We present PS and covariate balance metrics to assess successful confounding control, and provide HR estimates and Kaplan-Meier survival plots for the outcomes.
Residual study bias from unmeasured and systematic sources can exist in observational studies after controlling for measured confounding [19,20]. To estimate such residual bias, we conduct negative control outcome experiments with 76 negative control outcomes identified through a data-rich algorithm [9]. We fit the negative control estimates to an empirical null distribution that characterizes the study residual bias and is an important artifact from which to assess the study design [21]. Using the empirical null distribution and synthetic positive controls [14], we additionally calibrate all HR estimates, their 95% confidence intervals (CIs) and the p-value to reject the null hypothesis of no differential effect (HR = 1). Empirical calibration serves as an important diagnostic tool to evaluate if residual systematic error is sufficient to cast doubt on the accuracy of the unknown effect estimates.
We compute incidence rates for each outcome in each exposure group, in both the on-treatment and intent-to-treat windows.

Comparative analyses
The following comparative analyses are performed. The above three databases were chosen based on having at least 2,500 subjects observed in both target and comparator cohorts: -56 outcomes of interest -2 time-at-risk definitions: on-treatment and intent-to-treat We also include 76 negative control outcomes, and 3 * 76 synthetic positive control outcomes, so 304 control outcomes.

Output
All output is stored in the LEGEND evidence model, available at the OHDSI LEGEND evidence web site [22].

Evidence Evaluation
We execute the following analysis diagnostics: • Propensity score distribution, shown as a plot of the preference score distribution [23] • Covariate balance before and after propensity score matching, with the standardized difference of the means of the covariate under 0.1 being considered adequate balance • Estimation for negative and positive controls, to assess residual error • Negative and positive control exposures and outcomes are used to evaluate the potential impact of residual systematic error in the study design, and to facilitate empirical calibration of the p-value and confidence interval for the exposures and outcome of interest.
Negative control outcomes in the context of this study are outcomes that are not believed to have been caused by either exposure in any comparison and where therefore the true hazard ratio is equal to 1. We will execute the same analysis used for the primary hypotheses to produce hazard ratio estimates for the negative controls. The distribution of effect estimates across all negative controls are used to fit an empirical null distribution, which models the observed residual systematic error. The empirical null distribution is then applied to the target exposures and outcome of interest to calibrate the p-value [19].
Positive control exposures and outcomes are pairs of exposures and outcomes where the hazard ratio is known to be of some magnitude greater than 1. We synthesize positive controls by starting with the negative controls defined earlier, and adding additional, simulated outcomes during the time-at-risk until the desired true hazard ratio is achieved. The target hazard ratios are 1.5, 2 and 4. The negative and positive controls together are used to estimate an empirical systematic error model, which inform whether systematic error changes as a function of true effect size. The empirical systematic error model is then applied to the target the target exposures and outcome of interest to calibrate the confidence interval [14].
Empirical calibration serves as an important diagnostic tool to evaluate if the residual systematic error is sufficient to cast doubt on the accuracy of the unknown effect estimate. The calibration effect plot and calibration probability plots are generated for review. We report the traditional and empirically calibrated p-value and confidence interval for each negative control, as well as the hypothesis of interest.

Strengths
• Cohort studies allow direct estimation of incidence rates following exposure of interest, and the new-user design can capture early events following treatment exposures while avoiding confounding from previous treatment effects. New use allows for a clear exposure index date. • PS matching allows balancing on a large number of baseline potential confounders.
• Use of negative and positive control outcomes allows for evaluating the study design as a whole in terms of residual bias. Limitations • Even though many potential confounders will be included in this study, there may be residual bias due to unmeasured or misspecified confounders.

Protection of Human Subjects
The study uses only de-identified data. Confidentiality of patient records is maintained at all times. All study reports contain aggregate data only and do not identify individual patients or physicians.

Management and Reporting of Adverse Events and Adverse Reactions
This study uses coded data that already exist in an electronic database. In this type of database, it is not possible to link (i.e., identify a potential causal association between) a particular product and medical event for any individual. Thus, the minimum criteria for reporting an adverse event (i.e., identifiable patient, identifiable reporter, a suspect product, and event) are not available and adverse events are not reportable as individual adverse events reports. The study results will be assessed for medically important results.

Plans for Disseminating and Communicating Study Results
The study results are posted on the OHDSI LEGEND evidence web site [22].

Effectiveness and safety results for the meta-analysis for each analytic approach (stratified vs. matching; on-treatment vs. intent-to-treat)
(Only meta-analysis results are shown here. Effectiveness and safety results by individual database are available at http://data.ohdsi.org/LegendBasicViewer/ -> Specific Research Questions -> "Hypertension," "Drug or procedure," "Chlorthalidone," "Hydrochlorothiazide," <outcome of interest>.) eTable 9. Effectiveness and safety for meta-analysis for stratified on-treatment (these are the primary results reported in the main paper)

Comparability of the populations for the three databases
eFigure 4a. Preference score distribution before propensity score adjustment. The preference score is a transformation of the propensity score that adjusts for differences in the sizes of the two treatment groups. A higher overlap indicates subjects in the two groups were more similar in terms of their predicted probability of receiving one treatment over the other. This plot shows sufficient equipoise (majority of both distributions being between 0.25 and 0.75) in all three databases that propensity score stratification should be able to create balance without discounting a large proportion of the population, but it shows sufficient difference (nonoverlap) that propensity score stratification is necessary. eFigure 4b. Preference score distribution after propensity score adjustment. Same plot as (a) but showing essentially perfect overlap after adjustment (matching shown here). This illustrates the success of the adjustment in achieving balance. eFigure 4c. Covariate balance before and after propensity score stratification. Each dot represents the standardized difference of the means for a single covariate before and after stratification on the propensity score. The figure shows poor balance before but excellent balance after stratification, with all 60,535 to 70,072 covariates under 0.1 and most under 0.05. It demonstrates that all measured variables were successfully balanced by the adjustment and that the two cohorts were in fact similar on all measured aspects. eFigure 4d. Covariate balance before and after propensity score matching. Each dot represents the standardized difference of the means for a single covariate before and after matching on the propensity score. The figure shows poor balance before but excellent balance after stratification, with all 60,405 to 69,356 covariates under 0.1 and most under 0.05. It demonstrates that all measured variables were successfully balanced by the adjustment and that the two cohorts were in fact similar on all measured aspects.

Homogeneity on effectiveness and hypokalemia across databases and analytic approaches
(Only the primary effectiveness and hypokalemia results are shown here. Other outcomes can be seen summarized in the forest plots in the next section, and plots similar to these for all outcomes are available at http://data.ohdsi.org/LegendBasicViewer/ -> Specific Research Questions -> "Hypertension," "Drug or procedure," "Chlorthalidone," "Hydrochlorothiazide," <outcome of interest>.) eFigure 5 (a-t). Homogeneity on effectiveness and hypokalemia by database and analytic approach. Hazard ratios and forest plot of the three databases and the meta-analysis for chlorthalidone versus hydrochlorothiazide for each combination. (Stratified on-treatment represents the paper's primary results and duplicates Figure 2 in the main paper.) e5a. Composite cardiovascular disease for stratified on-treatment e5b. Composite cardiovascular disease for stratified intent-to-treat e5c. Composite cardiovascular disease for matching on-treatment 2.6 Forest plot of safety and effectiveness outcomes across analytic approaches eFigure 6(a-d). Forest plot of safety and effectiveness across analyses. Forest plot of hazard ratio estimates and calibrated 95% confidence intervals for chlorthalidone versus hydrochlorothiazide for 55 safety and effectiveness outcomes for three databases and the metaanalysis. Confidence intervals that include 1 (no effect) are greyed. (The forest plot for stratification on-treatment for the meta-analysis is shown in Figure 3 of the main paper). eFigure 6a. Forest plot of stratification on-treatment

Detailed definitions of exposures and outcomes
We provide here a formal definition for each exposure and outcome. The definitions are based on the OHDSI OMOP data model [24]. They contain logic and concept set expressions, which are lists of codes used to query the database. Each concept set expression is defined in one or more tables below each definition. In each table, the "concept ID" is the unique OHDSI OMOP code for the term, the "domain" tells what kind of data are coded, and the "vocabulary" and "concept name" identify the meaning of the code. If the term is an exclusion for the outcome (include only persons without that code), then "excluded" is set to "yes." If "descendants" is "yes," then all the terms in that vocabulary that are more specific (i.e., descendants) are included in the query. "Mapped" implies that terms from other vocabularies are included. OHDSI maps data from existing vocabularies like ICD9-CM and ICD10-CM to its standard vocabularies like SNOMED CT, so diagnoses in queries are defined by with SNOMED CT codes rather than ICD codes.
For the effectiveness and safety outcomes, we also include a table with a logical description, computer-executable code, and relevant references.

Drug exposures
In this study, we are interested in two exposures, chlorthalidone and hydrochlorothiazide. They were formally defined as follows: Index rule defining the index date: • First exposure to any drug or combination of drugs containing the RxNorm ingredient(s) of interest.
Inclusion rules based on the index date: • At least 365 days of observation time prior to the index date • No exposure to the target or comparator ingredient(s) before the index date • No exposure to any other hypertension treatment on or before the index date • A diagnosis of hypertension on or preceding the index date Note that no prior exposure to other hypertension treatments prior to or on the index date is allowed. For example, when comparing chlorthalidone to hydrochlorothiazide, no prior exposure to lisinopril or any other hypertension treatment is allowed. The end of the exposure cohort is defined as the end of the first exposure, allowing for 30day gaps between consecutive prescriptions.

Outcomes
We first summarize the effectiveness and safety outcomes in a table and then show each detailed definition (including codes used) as described above.  [33] Chest pain or angina The first condition record of chest pain or angina https://github.com/OHDSI/Legen d/blob/master/inst/cohorts/Chest %20pain%20or%20angina.json [69] Chronic kidney disease The first condition record of chronic kidney disease, which is followed by either another chronic kidney disease condition record or a dialysis procedure or observation

Acute myocardial infarction
Acute myocardial infarction events Acute myocardial infarction condition record during an inpatient or ER visit; successive records with > 180 day gap are considered independent episodes Initial Event Cohort People having any of the following: with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: all events per person.
For people matching the Primary Events, include: Having any of the following criteria: • at least 1 occurrences of a visit occurrence of Inpatient or ER visit 1 where event starts between all days Before and 0 days After index start date and event ends between 0 days Before and all days After index start date

Acute pancreatitis
Acute pancreatitis events Acute pancreatitis condition record during an inpatient or ER visit; successive records with >30 day gap are considered independent episodes Initial Event Cohort People having any of the following: • a condition occurrence of Acute pancreatitis 2 with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: all events per person.
For people matching the Primary Events, include: Having any of the following criteria: • at least 1 occurrences of a visit occurrence of Inpatient or ER visit 1 where event starts between all days Before and 0 days After index start date and event ends between 0 days Before and all days After index start date

Acute renal failure
Acute renal failure events A diagnosis of 'acute renal failure' in an inpatient or ER setting; must be at least 30d between inpatient/ER visits to be considered separate episodes Initial Event Cohort People having any of the following: • a condition occurrence of Acute Renal Failure 2 with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: all events per person.
For people matching the Primary Events, include: Having any of the following criteria:

All-cause mortality
All-cause mortality Death record of any type Initial Event Cohort People having any of the following: • a death occurrence from Any Death with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: earliest event per person.
Limit qualifying cohort to: earliest event per person.

End Date Strategy
No end date strategy selected. By default, the cohort end date will be the end of the observation period that contains the index event. Cohort Collapse Strategy: Collapse cohort by era with a gap size of 0 days.

Anaphylactoid reaction
Anaphylactoid reaction events Anaphylactoid reaction condition record during an inpatient or ER visit; successive records with >7 day gap are considered independent episodes Initial Event Cohort People having any of the following: • a condition occurrence of Anaphylactoid reaction 2 with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: all events per person.
For people matching the Primary Events, include: Having any of the following criteria: • at least 1 occurrences of a visit occurrence of Inpatient or ER visit 1 where event starts between all days Before and 0 days After index start date and event ends between 0 days Before and all days After index start date

Anemia
Persons with anemia The first condition record of anemia Initial Event Cohort People having any of the following: • a condition occurrence of Anemia 1 o for the first time in the person's history with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: earliest event per person.
Limit qualifying cohort to: earliest event per person.

Anxiety
Persons with anxiety The first condition record of anxiety, which is followed by another anxiety condition record or a drug used to treat anxiety Initial Event Cohort People having any of the following: • a condition occurrence of Anxiety 1 o for the first time in the person's history with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: earliest event per person.
For people matching the Primary Events, include: Having any of the following criteria:

Cardiac arrhythmia
Person with cardiac arrhythmia The first condition record of cardiac arrhythmia, which is followed by another cardiac arrhythmia condition record, at least two drug records for a drug used to treat arrhythmias, or a procedure to treat arrhythmias Initial Event Cohort People having any of the following: • a condition occurrence of Cardiac arrhythmia 1 o for the first time in the person's history with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: earliest event per person.
For people matching the Primary Events, include: Having any of the following criteria:

Chest pain or angina
Persons with chest pain or angina Initial Event Cohort People having any of the following: • a condition occurrence of Chest pain or angina 1 o for the first time in the person's history with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: earliest event per person.
Limit qualifying cohort to: earliest event per person. End Date Strategy No end date strategy selected. By default, the cohort end date will be the end of the observation period that contains the index event. Cohort Collapse Strategy: Collapse cohort by era with a gap size of 0 days.

Chronic kidney disease
Persons with chronic kidney disease The first condition record of chronic kidney disease, which is followed by either another chronic kidney disease condition record or a dialysis procedure or observation Initial Event Cohort People having any of the following: • a condition occurrence of Chronic kidney disease 1 o for the first time in the person's history with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: earliest event per person.
For people matching the Primary Events, include: Having any of the following criteria: • at least 1 occurrences of a condition occurrence of Chronic kidney disease 1 where event starts between 1 days After and all days After index start date

Depression
Persons with depression The first condition record of depression, which is followed by another depression condition record, at least two drugs used to treat depression without another indication, or two psychotherapy procedures Initial Event Cohort People having any of the following: • a condition occurrence of Depression 1 o for the first time in the person's history with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: earliest event per person.
For people matching the Primary Events, include: Having any of the following criteria: where event starts between 0 days Before and all days After index start date Limit cohort of initial events to: earliest event per person.
Limit qualifying cohort to: earliest event per person. End Date Strategy No end date strategy selected. By default, the cohort end date will be the end of the observation period that contains the index event. Cohort Collapse Strategy: Collapse cohort by era with a gap size of 0 days.

End stage renal disease
Persons with end stage renal disease End stage renal disease (ESRD) is defined by at least one diagnosis in any setting, followed by at least one additional diagnosis of a dialysis-related procedure within 90 days Initial Event Cohort People having any of the following: • a condition occurrence of End stage renal disease 2 with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: all events per person.
For people matching the Primary Events, include: Having any of the following criteria: • at least 1 occurrences of a condition occurrence of End stage renal disease 2

Gout
Persons with gout The first condition record of gout Initial Event Cohort People having any of the following: • a condition occurrence of Gout 1 o for the first time in the person's history with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: earliest event per person.
Limit qualifying cohort to: earliest event per person.

Heart failure
Persons with heart failure The first condition record of heart failure, which is followed by at least 1 heart failure condition record in the following year Initial Event Cohort People having any of the following: • a condition occurrence of Heart Failure 1 with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: all events per person.
For people matching the Primary Events, include: Having all of the following criteria:

Stroke
Stroke (ischemic or hemorrhagic) events Initial Event Cohort People having any of the following: • a condition occurrence of Stroke (ischemic or hemorrhagic) 2 with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: all events per person.
For people matching the Primary Events, include: Having any of the following criteria: • at least 1 occurrences of a visit occurrence of Inpatient or ER visit 1 where event starts between all days Before and 1 days After index start date and event ends between 0 days Before and all days After index start date Sudden cardiac death condition record during an inpatient or ER visit; successive records with > 180 day gap are considered independent episodes Initial Event Cohort People having any of the following: • a condition occurrence of Sudden cardiac death 2 with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: all events per person.
For people matching the Primary Events, include: Having all of the following criteria: • at least 1 occurrences of a visit occurrence of Inpatient or ER visit 1 where event starts between all days Before and 0 days After index start date and event ends between 0 days Before and all days After index start date

Thrombocytopenia
Persons with thrombocytopenia The first condition record of thrombocytopenia Initial Event Cohort People having any of the following: • a condition occurrence of Thrombocytopenia 1 o for the first time in the person's history with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: earliest event per person.
Limit qualifying cohort to: earliest event per person. End Date Strategy No end date strategy selected. By default, the cohort end date will be the end of the observation period that contains the index event. Cohort Collapse Strategy: Collapse cohort by era with a gap size of 0 days. with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: all events per person.

Concept
For people matching the Primary Events, include: Having any of the following criteria: • at least 1 occurrences of a visit occurrence of Inpatient or ER visit 1 where event starts between all days Before and 0 days After index start date and event ends between 0 days Before and all days After index start date with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: earliest event per person.
Limit qualifying cohort to: earliest event per person. End Date Strategy No end date strategy selected. By default, the cohort end date will be the end of the observation period that contains the index event. Cohort Collapse Strategy: Collapse cohort by era with a gap size of 0 days.