Research Design and Participants
In 2002, the National Health Insurance Service (NHIS) in South Korea established the “National Health Information Database (NHI DB),” which includes information on patients’ medical records, disease history, prescriptions for health insurance, and medical aid beneficiaries. The DB was systematically sampled and stratified by sex, age, region, and the level of insurance cost, to ensure representativeness. Additionally, the NHIS DB is divided into “Standard Cohort,” “Health Examination Cohort,” “Older Adults Cohort,” and “Infant Cohort” DBs to study specific populations.
This study used data from the “Older Adults Cohort DB” from 2002 to 2015. The full DB includes 558,147 randomly extracted people, who constituted about 10% of the total 5.5 million Korean older adults aged over 60 in December 2002. To obtain a study sample of older adults with dementia, we employed the following four steps. Step 1: We selected 121,235 older adults who were outpatients or inpatients at least once with a disease code related to dementia (F00–F03, Korean standard classification of disease and cause of death, KCD) from 2002 to 2015 (in Figure 1, the “survival period (green box)” refers to the duration of the dementia experience). Step 2: All other causes of death, except suicide (which is a dependent variable in this study), were deleted. Step 3: To meet the baseline of health conditions of older adults in each group according to the type of LTCS, older adults with LTCS grading were selected based on their score for long-term care approval; those who had been judged to need LTCS after the introduction of the LTCI system in 2008 were selected. Step 4: The final sample included 62,282 older adults after creating a balanced panel for 14 years, and excluding individuals with missing values for the main variables used in the analysis.
Figure 1 presents the structure of the sample selected for the study. First, among the older adults who survived until the end of follow-up, Case 1 refers to older adults who had been diagnosed with dementia from the beginning of the observation, while Case 2 comprises older adults diagnosed with dementia during the observation. Second, Cases 3 and 4 had the same criteria for period of dementia diagnosis as Cases 1 and 2; however, the difference was that in the latter cases, the older adults had died before the end of follow-up. Lastly, Casen-1 and Casen are older adults who have the same criteria as Cases 1–4 but represent those who have used services under the LTCI since its introduction.
Measures
Dependent Variable: Suicide
Suicide is the dependent variable in this study, to verify whether the cause of death data included the suicide codes (X60–X84). The binary variable was coded 1 if the older adult had committed suicide and 0 otherwise.
Independent Variable: Long-Term Care Services
This study took the experience of using LTCS as a policy variable, to confirm the effect of LTCS on preventing suicide among the older adults with dementia. The LTCS types were classified into non-user, facility benefits, and in-home benefits. Generally, facility services are used when the disease symptoms are severe. The LTCI provides three types of services (LTCS): facilities, in-home, and cash services. In severe cases or grades of 1 to 3, facility services can be used; by contrast, for grades 4 to 5 and the dementia special grade, patients mainly use in-home services. Therefore, the service type was classified as a facility service if the older adult used both facilities and in-home services. We excluded cash services because the frequency of using cash services is very small.
Control variables: Demographic Characteristics and Health Status
The independent variables in this study included the demographic characteristics of older adults with dementia, that is, sex, age, and income level, as well as caregivers and their health status. Sex comprised male and female, and ages were classified into three categories of 60–69 years, 70–80 years, and above 80 years. The NHI premium for the 10th quantile was used as a proxy for income level because NHI coverage in Korea is subject to compulsory payment of health insurance premiums based on individual income levels. This variable was re-categorized into five quantiles: 1st quantile (medical aid beneficiaries), 2nd quantile (the 1st–2nd quantile, which is the lowest income level), 3rd quintile (3rd–5th), 4th quintile (6th–8th), and 5th quintile (9th–10th). The types of caregivers were categorized into family and neighbors, professional caregivers, and no caregivers. Regardless of what LTCS older people used, they may choose professional caregivers or family caregivers, or they may not use caregivers. The health status included activities of daily living (ADL) and cognitive ability. The ADL was classified into three categories (independent, partially dependent, and fully dependent) by summing the scores for 13 questions (activities): bathing, washing the face, washing hair, brushing teeth, eating, dressing, transfer, sitting up, moving to side sit, going out of the room, going to the toilet, controlling urine, and feces, on the basis of four responses (full independence, incomplete independence, partial dependence, and full dependence). For cognitive ability, the binary variable was divided into “good” and “bad” using the total scores of 10 questions: impairment of long-term or short-term memory, communication, counting, awareness of routines, decline in judgment, inability to recognize instructions, date, place, and age.
Analysis Strategy
The following analysis methods were used. First, we conducted a chi-square test to identify the factors affecting suicide among older adults with dementia. Second, the Kaplan-Meier and Cox regression analyses were conducted to create the yearly survival curve from 2002 to 2015, and the suicide rate was identified by demographic characteristics, health status, and policy. Third, to confirm the effect of reinforcing insurance coverage with the expansion of LTCS’s “dementia special grade,” a difference-in-differences (DID) estimation was conducted to analyze the effect of LTCS on suicide rates by comparing statuses before and after 2014. Model 1 analyzed the effect of the interaction term (before and after expansion x service used or not) on the suicide rate, and model 2 confirmed the effect of the interaction term on the suicide rate after adjusting the control variables. This study used a quasi-experimental design to evaluate the impact of the expansion of LTCI’s “dementia special grade.” In addition to DID analysis, we analyzed panel data from 2009 to 2015 after the introduction of the LTCI system. The effect was estimated by calculating the pre (2009 to 2013)-post (2014 to 2015) difference in outcomes for the policy group (LTCS users) and the pre-post difference in outcomes for the reference group (non-LTCS users).