Database

Data were retrieved from the Taiwan National Health Insurance (NHI) Research Database (NHIRD), which is a database of all NHI beneficiaries, over 99% of the Taiwanese population. This database includes the registration files and all medical claims data for approximately 23 million NHI enrollees since 1995 when the NHI program was implemented. The NHIRD provides de-identified data to Taiwanese researchers, enabling longitudinal studies of patients with medical conditions. A large volume of research has been published based on the NHIRD in peer-reviewed international journals. This study was approved by the Institutional Review Board of Taipei Medical University (TMU-JIRB N201912007). This study adhered to the Declaration of Helsinki.

Study Sample

This study used a retrospective case control design. To select cases, we first identified 122,025 patients with a first-time diagnosis of POAG appearing in a claim for a clinic visit or hospital outpatient department visit from January 1, 2010, to December 31, 2015 (ICD-9-CM codes 365.1, 365.10, and 365.11). We used a look-back period of 3 years to exclude patients with POAG appearing in any claim prior to the index date during the study period. We selected patients with at least 2 claims showing a POAG diagnoses and showing a prescription for topical antiglaucoma medications (for diagnostic validity) and assigned the date of first POAG diagnosis as their index date. We excluded 3,114 patients under 18 years of age because of the low prevalence of POAG and HTN in that age group. The analytic study sample included 112,929 cases with POAG.

To select controls, we first excluded all patients with a medical claim showing a POAG diagnosis during the study period or aged less than 18 years. Out of the remaining NHI beneficiaries we selected a comparison group (controls) similar to cases on demographic and comorbidity variables at baseline. We selected four controls per POAG case by propensity score matching (n = 449,840) using age, sex, monthly income, geographic location (northern, central, southern, and eastern) residential urbanization level, and selected comorbidities expected to be associated with either POAG or HTN, hyperlipidemia, diabetes, coronary heart disease, migraine, hypotension, and obstructive sleep apnea syndrome. We assigned the date of their first utilization of ambulatory care during the index year of the matched case as the control patient’s index date. The final study sample included 112,460 POAG patients and 449,840 patients without POAG as controls.

Exposure Determination

The exposure of interest was HTN preceding the index date (for cases, POAG diagnosis date, and for controls, date of the first ambulatory care claim during the year of diagnosis of their matched case). The outcome of interest was POAG. We identified prior HTN by ICD-9-CM codes 401-405 in ambulatory care visit claims prior to the index date. To strengthen diagnostic validity we identified HTN only if a prescription for antihypertensive medication could be found in the claims.

Statistical Analysis

Analyses were performed using the SAS statistical package (SAS System for Windows, V.9.4, SAS Institute, Cary, NC, USA). We used chi-squared tests to compare cases and controls on sex, monthly income (<NT$15,841, 15,841-25,000, ≥25,001), geographic location (northern, central, eastern, and southern Taiwan), urbanization level (5 levels: 1 most urbanized and 5 least urbanized), hyperlipidemia, migraine, obstructive sleep apnea syndrome, diabetes, coronary heart disease, and hypotension between patients with POAG and their matched controls. We performed multiple logistic regression analysis were to estimate the odds (ORs) of prior HTN among cases vs controls. A value of P = .05 was used to determine statistical significance.