Data source
The Longitudinal Health Insurance Database (LHID), established by the National Health Research Institute (NHRI) in Taiwan, contains patient records of over 99% of the Taiwanese population. The data, derived from across all medical care settings, include information on demographics, diagnoses, and medical procedures and prescriptions. We obtained these records through a formal application to the Health and Welfare Data Science Center (HWDC), Department of Statistics, Ministry of Health and Welfare, Taiwan (http://dep.mohw.gov.tw/DOS/np-2497-113.html). We did not have any special access privileges that others would not have, and interested researchers would be able to access the data in the same manner.
To comply with the Personal Information Protection Act, identifying information of each insurant in the LHID was re-coded. Data files were linked with the identifications of patients that were scrambled with surrogate numbers to protect the confidentiality of the beneficiaries. Therefore, informed consent was subsequently waived. The study protocol was conducted according to the principles described in the Declaration of Helsinki. This study has been approved by the Research Ethics Committee at China Medical University Hospital, Taiwan (CMUH104-REC2-115-AR-4).
Study subjects
To perform this cohort study, we selected patients newly diagnosed with DR [the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) 362.0, 362.01 and 362.02] between January 1, 2000 and December 31, 2012. Patients with at least three outpatient visits for DR were defined as new cases and the first visit date for DR was defined as the index date. Those with a diagnosis of DR prior to 2000 were excluded. The endpoint was a new diagnosis of ptosis (ICD-9-CM 374.3, 374.30, 374.31, and 374.32). The end date of follow-up was December 31, 2013.
Those with a history of ptosis (ICD-9-CM 374.3, 374.30, 374.31, and 374.32) were excluded. Patients with viral hepatitis (ICD-9-CM code 070), cirrhosis (ICD-9-CM code 571, A347), interferon treatment, human immunodeficiency virus (HIV) infection (ICD-9-CM code 042-044, 795.8, V08), tuberculosis (ICD-9-CM code 010-012), syphilis (ICD-9-CM code 091.0, 095.4, 095.8), systemic malignancy (ICD-9-CM code 140-208), autoimmune diseases (ICD-9-CM code 135, 279.49, 283, 443, 571.42, 696, 710, 714, 715), chronic Obstruction Pulmonary Disease (ICD-9-CM code 490-492, 494, 496), and asthma (ICD-9-CM code 493, 494) were also excluded.17 Subjects without outpatient visits for eye diseases were also excluded.
The comorbidities assessed included current smoking (ICD-9-CM code V15.82, 305.1, 794.2), lipid metabolism disorders (ICD-9-CM code 272), and hypertension (HT) (ICD-9-CM code 401-405, A26). Patient identification numbers were encrypted for privacy protection.
Controls were randomly selected from populations without histories of viral hepatitis, interferon treatment, HIV infection, tuberculosis, syphilis, or ptosis. They were frequency-matched by age group (<20, 20-39, 40-64 and 65+ years old), gender, ophthalmologic outpatient department (OPD) before the index date, and index-year at a ratio 4:1. Only patients with at least one ophthalmology clinic visit before enrolling in the study were included. We then matched the ophthalmologic OPD visits between both groups.
Study endpoint
The clinical endpoint was a diagnosis of ptosis. Patients with at least two outpatient visits for ptosis, separated for at least 7 days, were defined as the endpoint to ensure the validity of the diagnosis. All study subjects were followed from the index date until the endpoint. Those without endpoint development were followed until the date of withdrawal from the program or the end of 2012, whichever occurred first. In this study, the demographic characteristics included age group (<20, 20-39, 40-64 and 65+ years old), gender, comorbidities, and ophthalmologic OPD before the index date.
Statistical analyses
We used chi-square testing to determine the difference of demographic characteristics between the DR cohort and comparison cohort from 2000 to 2012. A t-test was employed for the difference of the mean OPD visit for ophthalmology between two cohorts. Continuous variables, such as age and follow-up time, were shown as mean and standard deviation (SD) and analyzed by using the Wilcoxon rank sum test. The cumulative incidences of ptosis for both the DR and comparison cohorts were estimated by using the Kaplan-Meier method. The difference between the two curves was examined by using the log-rank test. A multivariable Cox model was adjusted for continuous age, gender, comorbidities, and OPD visits for ophthalmology before the index date. Univariate and multivariable cox proportional regression analysis were used to measure the hazard ratio (HR) and 95 % confidence interval (CI) to assess the association between DR and the risk of developing ptosis. The incidence density rate of ptosis (per-1,000 years) was calculated for DR cohort and comparison cohort. The risk of ptosis in the DR and comparison cohorts was stratified by age group, gender, and comorbidities, using Cox proportional hazard regression. SAS software (version 9.4 for Windows; SAS Institute, Cary, NC, USA) was used for all statistical analyses and creation of Kaplan-Meier survival curves. A two-sided P < 0.05 was considered statistically significant. The Wilcoxon rank-sum test was used for verification of average age and follow-up time.