Data sources
The Taiwan National Health Insurance (NHI) program was launched in 1995 to provide a centralized health insurance for its citizens, and as of 2014 approximately 93% of the nation’s medical care institutions were contracted, with an enrollment rate exceeding 99% of Taiwan’s population (Ho Chan, 2010). The NHIRD is derived from the Taiwan NHI program, and all claims data are released by the Bureau of National Health Insurance for research purposes. The NHIRD uses the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) codes to record diagnoses (Chinese Hospital Association, 2000). The quality and validity of the NHIRD is adequate, and its data have been used in many published studies (Lin et al., 2014; Shen et al., 2013; Shih et al., 2014). In the present study, we used the data sets from the Registry for the one-million Longitudinal Health Insurance Database (LHID) which included comprehensive outpatient and inpatient information, such as demographic data, dates of clinical visits, diagnostic codes, and details of prescriptions, with regard to nearly 1 million beneficiaries over a 13-year period from the LHID in Taiwan (2000–2013).
Ethical approval
To protect patient privacy, patient identity data were scrambled cryptographically in the NHIRD. This study was approved by the Institutional Review Board of Tri-Service General Hospital (TSGHIRB-2-106-05-029), and the written informed consents were waived.
Study design and sampled participants
This study is of a retrospective, matched-cohort design. Patients with PTSD were selected from January 1, 2000 to December 31, 2013, according to the ICD-9-CM codes: 309.81. In addition, each enrolled patient was required to have made at least three outpatient visits within the one-year study period for adult males with PTSD according to these ICD-9-CM codes. The patients diagnosed with ED before 2000 or before the first visit for PTSD were excluded. In addition, all patients aged <20 years were also excluded. A total of 4,310 enrolled patients with the 1,079 subjects with PTSD and 3,237 in the age and index-year matched control group without PTSD in this study, in the 13 years of follow-up to December 31, 2013 (Figure S1).
Covariates
The covariates included age group (20-39 years, ≥40 years), geographic area of residence (north, center, south, west, and east of Taiwan), urbanization level of residence (levels 1–4), levels of hospitals as medical centers, regional hospitals,
and local hospitals, and monthly income (in New Taiwan Dollars [NT$]: <18,000, 18,000–34,999, ≥35,000). Charlson Comorbidity Index (CCI) defined the comorbidity (Charlson et al., 2008; Charlson et al., 1987). The population and various indicators defined the urbanization levels. Level 1 was defined as a population of >1,250,000; level 2 was defined as a population between 500,000 and 1,249,999; and urbanization levels 3 and 4 were defined as a population between 149,999 and 499,999, and <149,999, respectively (Chang et al., 2014).
Comorbidity
Baseline comorbidities (in ICD-9-CM codes) included dementia, schizophrenia, anxiety disorder, bipolar disorder, depressive disorders, stroke, coronary artery diseases, hypertension, diabetes mellitus, asthma, and alcohol-related illness, with the reference from one previous study (Yang et al., 2018). Data on the usage of psychotropic medications, including antidepressants, antipsychotics, and hypnosedatives, were collected. The data of defined daily dose (DDD) were obtained from the WHO Collaborating Centre for Drug Statistics Methodology (https://www. whocc.no/), and the duration of the use of drugs was calculated by
dividing the cumulative doses by the DDD of drugs.
Main outcomes
All of the study participants were followed from the index date until the onset of erectile dysfunction, withdrawal from the NHI program, or the end of 2013. ED was divided into two subgroups: psychogenic ED and organic ED.
Statistical analysis
All statistical analyses were performed using the SPSS software V.22 (SPSS Inc., Chicago, Illinois, USA). χ2 test and t-test were used to evaluate the distributions of the categorical and continuous variables, respectively. The Fisher’s exact test for the categorical variables was used to statistically examine the differences between the two cohorts. The Cox regression model was used to determine the risk of psychiatric disorders, and the results were present as HR with a 95% CI. The difference in the cumulative incidence of psychiatric disorders between the study and control groups was estimated using the Kaplan-Meier method with the log-rank test. A two-tailed p value <0.05 was considered to indicate the statistical significance.