There is rising concern that people with mental health problems are not receiving adequate care in medical systems. Mental health parity has been a major challenge in numerous countries, even in countries where universal insurance coverage is provided. In many health care systems, cost-sharing mechanisms have been instituted to reduce moral hazards, restrict unnecessary utilization, and contain costs (
1).
Despite these benefits, opponents of cost sharing argue that cost sharing hinders access to care and causes adverse health effects in vulnerable subpopulations, particularly among people with severe mental illnesses (
2). Cost sharing can result in nonadherence and low compliance with treatment among patients with mental illness (
3,
4). Reducing out-of-pocket expenditures among people with severe mental illnesses may improve access to appropriate mental health services and thus yield favorable health outcomes (
5–
8). The use of copayments and coinsurance is particularly of concern for people who are newly diagnosed as having schizophrenia, because this vulnerable subgroup requires frequent and stable contact with health care providers during the initial treatment period (
9–
11). Therefore, governments commonly institute welfare programs to minimize the adverse consequences caused by cost-sharing obligations.
Paradoxically, a major concern for cost-sharing exemption programs and other similar welfare programs is that not every eligible person enrolls. For example, Medicaid and the Medicare Part D prescription drug program are welfare programs that reduce vulnerable subpopulations’ financial barriers to care. Enrollment is voluntary. Unexpectedly, the proportions of eligible enrollees who participate in these welfare programs range from only 44% to 56% (
12–
14). The low enrollment of persons who are eligible for these welfare programs has become a major challenge to the welfare programs and may pose a substantial hindrance to accomplishing their welfare mission.
Therefore, investigating the risk factors associated with nonenrollment and possible obstacles to enrolling in welfare programs is crucial. The enrollment status of public welfare programs is generally investigated by considering individual characteristics. Previous studies have determined that men, people with low incomes, people with low levels of education, and people who have never had health insurance are less likely to enroll in health-related welfare programs (
12–
15). Moreover, findings on the relationship between age and number of comorbid conditions and enrollment in health-related welfare programs have been presented in other studies (
12–
14).
Difficulty in obtaining program information, expenses associated with program enrollment, and complex enrollment requirements or regulations are also cited as primary obstacles to program enrollment (
13). Some people are concerned that limiting public welfare programs to persons with a severe mental illness, such as schizophrenia, may hinder people from enrolling because of the stigma associated with mental illness or because aspects of the illness, such as poor cognitive function, may make it difficult to access program information or undertake the enrollment process (
6,
16). Given that clinicians are likely to act as major information providers or serve as gatekeepers for public welfare programs, it is important to understand how provider characteristics contribute to potentially low enrollment rates (
17). Unfortunately, the current understanding of the association of providers and enrollment is considerably limited (
12).
Experiences with a cost-sharing exemption program for people with severe mental illnesses in Taiwan may shed light on the association of providers and enrollment. Taiwan has been renowned for the comprehensive coverage and easy access to care provided by the National Health Insurance (NHI) program. To ensure the accessibility of care among vulnerable subpopulations, the government exempts people with major illnesses from cost-sharing obligations for disease-specific health services in the NHI program (
17). Although the exemption policy has existed since the inception of the NHI program in 1995, little is known about who enrolls in the program and whether providers’ characteristics influence enrollment. Therefore, this study investigated one-year and three-year enrollment rates for the cost-sharing exemption program among people newly diagnosed as having schizophrenia and the factors associated with enrollment in the NHI exemption program.
Methods
Data Sources
The main data source was the Longitudinal Health Insurance Database 2000 (LHID2000) in Taiwan. The NHI program provides compulsory universal health insurance for all Taiwanese residents. The program benefit coverage includes outpatient, emergency room, and inpatient care as well as prescription drugs for both general medical and mental disorders. The LHID2000 contains the enrollment and utilization data of one million beneficiaries randomly selected from all NHI beneficiaries during 1996 to 2000. The enrollment file provides information on an enrollee’s monthly wage category, living arrangement, and residential location. The outpatient and inpatient files list an enrollee’s birth date, sex, clinical diagnosis, date of visit or discharge, expenditure of the visit, the hospital or clinic where the visit took place, and the physician who provided the service. The registry for medical personnel contains the physician’s sex, birth date, and specialty.
The registry for medical facilities provides information on the physicians’ accreditation levels and ownership of the institution and whether it is a psychiatric institution. The catastrophic illness file was used to identify people with major diseases or injuries. On the basis of the injury severity score, the NHI major disease list contains records on 30 types of major diseases or injuries, such as cancer, schizophrenia, and end-stage renal disease (
17). A person with any of the major diseases or injuries listed can apply for the cost-sharing exemption program. Enrollees with multiple diseases must apply separately for each disease. They can be exempted from the cost-sharing requirement only for health services associated with the diseases for which they have applied.
Study Design and Sample
For this retrospective cohort study, the study population comprised people who were newly diagnosed as having schizophrenia between January 1, 2000, and December 12, 2007. Persons were identified by a primary diagnosis of schizophrenia (code 295 of the
International Classification of Diseases, Ninth Revision, Clinical Modification (
ICD-9-CM) (
18). Persons ages 15 and older who had at least three outpatient visits or one inpatient admission within one year since the first diagnosis of schizophrenia were included (
19). To confirm that the patients represented incident cases, we excluded patients who were diagnosed as having schizophrenia in any claims or registration records of the NHI program during 1996–1999. Patients diagnosed as having dementia (
ICD-9-CM codes 290 and 331), organic psychotic condition (293 and 294), or schizoaffective disorder (295.7) were also excluded from analysis (
11,
18). Each person newly diagnosed as having schizophrenia was followed up for one year and three years after the first diagnosis of schizophrenia.
Measures
The primary outcome was whether people diagnosed as having schizophrenia enrolled in the cost-sharing exemption program during the one-year and three-year follow-ups. Enrollment was determined by whether the patient was listed in the catastrophic illness files. Patients recorded in the catastrophic illness files were categorized as enrollees, and the remaining patients were categorized as nonenrollees.
Both patients and physicians play major roles in enrollment in the cost-sharing exemption program; therefore, we used patient characteristics (demographic characteristics, socioeconomic status, health status, and residential location) and physician characteristics (demographic characteristics, specialty, and practice setting) as independent variables. The patient demographic characteristics were age, gender, and living arrangement. The monthly income served as a proxy variable for socioeconomic status (SES), which was classified into four income levels: ≥$40,000, $20,000–$39,999, and <$20,000 in New Taiwan (NT) dollars (US $1 equals approximately NT$29.50) and nonregular wage earner (including union or association members and local government enrollees). Employment status was used as another proxy variable for SES.
Two health status variables were constructed: exhibiting any major general medical comorbidity and being hospitalized for schizophrenia during the one-year and three-year follow-ups. Furthermore, because the availability of health resources and level of local development may also influence health-seeking behavior, we constructed a summary of the population density of the area in which the psychiatrists were located and categorized the results into tertiles. The level of urbanization was also constructed, according to Liu and colleagues (
20), and the areas of the visit were categorized as urban, suburban, or rural.
In regard to physician characteristics, the physician’s age was classified into four groups: <35, 35–44, 45–54, or 55 or older. Physician specialty was categorized as psychiatrist or nonpsychiatrist. Regarding the practice setting, accreditation level was categorized as hospital or clinic. Ownership of the practice setting was classified as public or private. Whether the practice setting was a psychiatric institution was also included as an independent variable.
Statistical Analysis
Descriptive analyses were conducted to describe the distribution of the sample. The differences in patient, physician, and practice setting characteristics between the patients with schizophrenia who had or did not have a cost-sharing exemption were estimated by using chi square tests. Multivariable logistic regression was then applied to identify factors associated with enrollment status in the cost-sharing exemption program. However, correlations between enrollment and patient characteristics might be influenced by the clustering of patients with providers. To adjust for correlation of enrollment and provider clusters, generalized estimating equations were used to cluster patients into the first level and providers into the second level. Furthermore, sensitivity analyses were also conducted by adjusting for various inclusion criteria, such as by using older ages as an inclusion criterion, using a higher number of outpatient visits as a cutoff, including only patients who were cared for by psychiatrists, and extending the follow-up period. All of the results remained robust for the various definitions. SAS, version 9.2, was used for data linkage and analysis. This study was approved by the institutional review board of Taipei Veterans General Hospital in Taiwan (2013–02–019BCY).
Results
The final sample comprised 1,824 people who were newly diagnosed as having schizophrenia. The upward trend in the program enrollment rate for the sample reached a plateau by the fourth year after diagnosis.
Table 1 shows the distribution of the patient and provider characteristics of enrollees in the first- and third-year samples. Of the sample, only 955 (52%) patients enrolled in the cost-sharing exemption program during the first year after their diagnosis, and 1,059 (58%) enrolled during the third year after their diagnosis. Persons who were ages 65 or older (40% and 42% at one-year and three-year follow-ups, respectively), who earned a monthly wage of at least NT$40,000 (40% and 46%, respectively), who were not hospitalized for schizophrenia during the follow-up period (40% and 41%, respectively), and who were cared for by nonpsychiatrists (39% and 40%, respectively) exhibited relatively low enrollment rates. By contrast, people who were admitted to a psychiatric institution exhibited the highest enrollment rates (66% and 71% at one-year and three-year follow-ups, respectively).
Table 2 shows the multivariable results for the one-year and three-year enrollment rates. After the analyses adjusted for other patient and provider characteristics, people ages 35–44 (adjusted odds ratio [AOR]=1.15), 45–54 (AOR=1.24), and 55–64 (AOR=.21) were significantly more likely than the younger population to enroll in the cost-sharing exemption program within one year of diagnosis. Among all socioeconomic groups, nonregular wage earners exhibited the highest likelihood of enrollment (AOR=1.16). Regarding severity, people who were hospitalized for schizophrenia during the follow-up period were 1.51 times more likely to enroll in the cost-sharing exemption program within one year of diagnosis. Furthermore, although people living in rural areas or regions characterized by moderate or low density of psychiatrists were significantly more likely to enroll, the magnitudes were small. When we extended the follow-up period to three years, the density of the physician’s practice area no longer held a statistically significant association with enrollment.
The multivariate results indicated that provider characteristics also played a critical role in program enrollment. People treated by psychiatrists (AOR=1.10), by providers at psychiatric institutions (AOR=1.10), and by public providers (AOR=1.05) were significantly more likely to enroll in the exemption program within one year of diagnosis. After the follow-up period was extended to three years, being treated by a psychiatrist (AOR=1.12) and by psychiatric institutions (AOR=1.07) remained significantly associated with program enrollment. However, the minor difference between the public and private providers regarding enrollment status observed at one year was no longer observed at the three-year follow-up. After adjustment for other characteristics, the enrollment rate among persons who were newly diagnosed as having schizophrenia increased significantly between 2002 and 2005 but decreased slightly in 2006.
Discussion
Consistent with other public welfare programs for individuals with mental illness, the one-year and three-year enrollment rates for the schizophrenia cost-sharing exemption program were 52% and 58%, respectively (
12–
14). The stagnated enrollment rate may suggest that the major underlying barriers to program enrollment may be factors that persist over time. Inadequate financial incentives, fears of stigmatization, and a requirement for voluntary enrollment are three plausible persistent barriers to enrollment.
First, the low cost-sharing obligations of the NHI program may not give people sufficient financial incentives for program enrollment. The estimated average cost-sharing exemption for one year is approximately NT$7,000 (US$237). If only outpatient visits are incurred, the amount exempted is estimated to be NT$2,000 (US$68). A similar observation was found in the Medicare Part D plan (
14).
Another plausible explanation for the low and stagnated enrollment rate is fear of the stigmatization associated with a cost-sharing exemption program for a specific mental illness. When contemplating the trade-off between minimal financial incentives and possible stigmatization, people diagnosed as having schizophrenia may likely opt not to enroll.
More important, substantial variations in patient enrollment by provider group suggest that the requirement for voluntary enrollment may be a barrier to program enrollment. People who were cared for by psychiatrists or in psychiatric institutions were significantly more likely to enroll in the cost-sharing exemption program. To be eligible for enrollment in the program, individuals must obtain an official diagnostic report from a psychiatrist, and clinicians also act as major information providers for the public welfare programs. Thus people who are treated by psychiatrists might have easier access to program information compared with people treated by nonpsychiatrists. Also, mental health specialists may be more experienced and better informed regarding mental health–related welfare policies, mental illness trajectory, and the health care needs of people diagnosed as having schizophrenia. Thus they might have helped and actively encouraged their patients to enroll instead of merely offered information. Providers play a significant role in enrollment (
12).
Another critical issue is whether the application process and eligibility requirement create a barrier for enrollment among more vulnerable individuals. The significantly higher likelihood of enrollment among the most vulnerable populations (people who were hospitalized and more economically disadvantaged) suggests that this is not the case. People with greater financial or health needs were more likely to enroll in the program.
Furthermore, this study indicates that younger people and elderly persons were less likely to enroll in the cost-sharing exemption program. One plausible explanation may be that these two groups may rely more on guardians or family members for health-care seeking and financial supports. Therefore, the decision-making process for enrollment among younger people and elderly persons may be more complex compared with other individuals because it may need to involve family members or caregivers as well as patients and physicians. Future research may help to delineate this complex relationship.
The observation of a significant time pattern for enrollment may also be worth noting. The 2002 NHI policy expanding drug coverage to second-generation antipsychotics (SGAs) may help to explain the increase in enrollment since 2002. The expansive out-of-pocket costs associated with SGAs and the significant health benefits of these drugs may have increased people’s incentive to enroll in the cost-sharing exemption program. However, the minor decrease in enrollment beginning in 2006 may also be attributable to the high cost of SGAs. Increased utilization of expensive SGAs over time may have squeezed the already tightened global budgets of hospitals and clinics, making providers less likely to treat people with schizophrenia and less willing to offer program information or facilitate access to program enrollment. Year-to-year differences in enrollment may serve as clues to important policy considerations. Further investigation of the time pattern and plausible explanations may help to contribute to our understanding of enrollment patterns.
This study had several potential limitations. First, it assessed only enrollment status. Because of data limitations, we could not include attempts to enroll in the analyses. A low enrollment rate can be attributable to either a low number of applications or a high number of failed attempts. Lack of data about denial rates may reduce the significance of the findings in this study. Future research with more detailed data may help in this regard.
Second, we identified only patient or provider characteristics associated with enrollment status. We did not explore possible mechanisms hindering program enrollment. Although we used various observation periods (one-year versus three-year enrollment) to reflect the potential role of stigmatization or access to information as a barrier to enrollment, the exact mechanisms underlying the associations were not evaluated in this study.
Third, administrative data limitations are inherent to the NHI research database, such as the limited availability of clinical and SES information. The insurance category (insurable income and type) was a proxy estimation of SES that was used to reduce potential confounding bias in the outcome. Comorbidity and incidences of hospitalization were used as proxies of the severity of schizophrenia. Finally, this study focused only on schizophrenia, a major mental illness. The generalizability of the findings for other conditions may be limited.
Conclusions
Despite its limitations, this study was the first population-based study to assess enrollment in a disease-specific cost-sharing exemption program among people diagnosed as having schizophrenia. In addition to exploring patient characteristics, we explored the role of providers in welfare program enrollment among a vulnerable subpopulation. The results may provide information to providers and health policy makers regarding the use of welfare subsidies as a major tool for improving mental health parity for vulnerable patients with mental illness. The role or influences of providers must not be overlooked in increasing enrollment in disease-specific welfare programs.