Mortality among psychiatric patients in Taiwan: Results from a universal National Health Insurance program.(accepted)

(1)

Mortality among psychiatric patients in Taiwan

—Results from a universal National

Health Insurance programme

Yi-Hua Chen

a

, Hsin-Chien Lee

b,c

, Herng-Ching Lin

d,

⁎

a

School of Public Health, Taipei Medical University, Taipei, Taiwan

b_{Department of Psychiatry, Taipei Medical University Hospital, Taipei, Taiwan} c

School of Medicine, Department of Psychiatry, Taipei Medical University, Taipei, Taiwan

d

School of Health Care Administration, Taipei Medical University, Taipei, Taiwan

a b s t r a c t

a r t i c l e i n f o

Article history:

Received 21 February 2008 Received in revised form 24 July 2008 Accepted 31 July 2008

Keywords: Mortality

National Health Insurance Psychiatric inpatient

This study investigated 6-year follow-up mortality rates and cause of death for persons younger than 45 years old with a history of hospitalisation for major psychiatric disorders after the introduction of the National Health Insurance (NHI). Linkage data combining death certificates with Taiwan NHI research claims data were used. The study cohort was comprised all patients under the age of 45 years, who had been hospitalised for major psychiatric disorders in 1998. Patients agedb45 years undergoing an appendectomy were selected as a control group. Cox proportional hazard regressions were performed to compute the adjusted 6-year hazard ratios. For patients with schizophrenia, major depression, or bipolar disorder, the adjusted risks of dying during the follow-up period were significantly 4.614, 3.707 and 3.866, respectively, times higher than that for appendectomy patients. The adjusted hazard ratios of non-natural dying during the follow-up period were significantly 16.316, 14.626 and 8.481 times for female patients with schizophrenia, major depression, and bipolar disorder, respectively, as high as for female appendectomy patients. The continuing excess mortality among psychiatric patients, from both natural and unnatural causes, still remains even after implementation of a NHI.

1. Introduction

Over the years, studies have consistently reported higher mortality

rates among psychiatric patients of both genders, in various countries and

ethnic groups, especially among younger patients and speci

ﬁcally during

the

ﬁrst year after admission (

Black et al., 1985; Casadebaig and Quemada,

1991; Chen et al., 1996; Räsänen et al., 2003

). Despite some studies

reporting a distinct decline in excess mortality (

Sims, 1987; Casadebaig

and Quemada, 1991

), others have observed a persistent or even widening

disparity in health outcomes between the general population and

psychiatric patients (

Lawrence et al., 2003; Räsänen et al., 2003

).

Compared with the general population, excess mortality among

psychiatric patients is attributed to both natural and unnatural causes

of death (e.g., suicide and accidents). Higher risk of suicide was

reported for schizophrenic, manic, and depressive patients, especially

among male patients with affective disorders during the

ﬁrst decade

after a psychiatric hospitalisation (

Tsuang, 1978

). Although similar

patterns of death from natural causes (e.g., cardiovascular diseases

and lung diseases) were reported comparing psychiatric groups and

the general population, an elevated risk of mortality was documented

among mentally ill people (

Tsuang and Woolson, 1978; Corten et al.,

1988; Mortensen and Juel, 1990, 1993

).

Consistent with this picture, higher mortality, both natural and

unnatural deaths, among psychiatric inpatients in Taiwan has likewise

been reported (

Chen et al., 1996

). Using a nationwide cohort of

Taiwanese psychiatric inpatients admitted between 1987 and 1988,

inclusive,

Chen et al. (1996)

reported a higher 6-year mortality rate for

psychiatric inpatients (i.e., standardized mortality ratios of 3.1 for men

and 4.8 for women) compared with the general population.

Mortality studies have contributed considerably as indicators of

health-care quality for inpatient and outpatient services (

Tsuang and

Simpson,1985; Hewer et al.,1995

). It is often asked whether equal access

to health services including specialised medical procedures has been

achieved for a population in general (

Corten et al.,1991

). Many countries

with advanced economies provide universal health insurance coverage

to achieve more equal health-care access and to improve the health of

the general population. In Taiwan, before 1995, about 57% of the people

were insured through three separate programmes, the Labor, the

Government Employee, and the Farmers' Insurance Programs. In March

Psychiatry Research 178 (2010) 160–165

⁎ Corresponding author. School of Health Care Administration, College of Medicine, Taipei Medical University, 250 Wu-Hsing St., Taipei 110, Taiwan. Tel.: +886 2 2736 1661x3613; fax: +886 2 2738 4831.

E-mail address:henry11111@tmu.edu.tw(H.-C. Lin).

Contents lists available at

ScienceDirect

Psychiatry Research

(2)

1995, Taiwan initiated its National Health Insurance (NHI) programme

to

ﬁnance health care for all citizens. Taiwan's NHI has a unique

combination of characteristics: universal coverage, a single-payer

payment system with the government as the sole insurer, very low

out-of-pocket copayments, comprehensive bene

ﬁts, unrestricted access

to any medical institution of the patient's choice and a wide variety of

providers including primary care physicians. The NHI also covers the

cost of medication.

Although there are limited studies available, positive effects of

universal health insurance on health and health-care use have been

observed (

Decker and Remler, 2004

). Some researchers have

indicated that changes in a health-care system might contribute to a

decline in the tendency to excess mortality among psychiatric patients

(

Tsuang and Simpson, 1985; Sims, 1987

). Following the

implementa-tion of the NHI in 1995, a signi

ﬁcant change in health-seeking

behaviours and health resources use could be anticipated. It is thus of

interest to re-examine the mortality rates among psychiatric patients

in Taiwan following implementation of the NHI.

In this study, we investigated mortality among psychiatric patients

in a universal National Health Insurance programme in Taiwan. The

6-year follow-up mortality rates and causes of death for persons

younger than 45 years with a history of hospitalisation for a major

psychiatric disorder following the introduction of the NHI were

investigated, using data that combines death certi

ﬁcates with Taiwan

NHI research claims data.

2. Methods 2.1. Database

The hospitalisation data used in this study were obtained from the National Health Insurance Research Dataset (NHIRD) for the years 1996–98 inclusive, as published by the National Health Research Institute in Taiwan. The dataset includes all claims data from Taiwan's National Health Insurance (NHI) programme implemented as a means of financing health care for all Taiwanese citizens. As of April 2005, the NHI program had over 21 million enrollees, representing around 96% of the island's population (Chen et al., 2007). The NHI's reimbursement system ties a hospital's reimbursement level to its patient severity profile. To deter diagnosis upcoding, the NHI Bureau implements routine sample cross-checks of each hospital's claims with its medical charts, followed by punitive measures for coding infractions. Fines for fraudulent diagnosis reporting are generally 100 times the value of the false claim. As a result, hospitals' interests are best served by accurate coding of diagnoses and services provided. It is generally believed that the NHI's checks and balances foster accurate coding, although there have not been sensitivity and specificity studies to document its coding accuracy. Since this dataset is for the purpose of claims, no missing data is likely. As one of the largest databanks of medical/health information in the world, the NHIRD offers a unique opportunity to compare mortalities among young patients admitted to psychiatric departments.

In this study, the date of death was obtained from the 1998–2004 ‘cause of death’ datafile published by the Department of Health (DOH) in Taiwan. The ‘cause of death’ file provides details on marital status, the dates of births and deaths, place of legal residence, underlying cause of death (ICD-9-CM code), and employment status. Given that there is mandatory registration of all deaths in Taiwan, this data should be particularly accurate and comprehensive. The NHIRD was linked to the‘cause of death’ datafile with the assistance of the DOH in Taiwan.

2.2. Study sample

The study cohort consisted of all patients under the age of 45 years who had been hospitalised for schizophrenia (any ICD-9-CM 295 code other than 295.7 for schizoaffective disorder), major depression (296.2X and 296.3X), or bipolar disorder (296.0X, 296.4X, 296.5X, 296.6X, 296.7X, 296.80 or 296.89) between January and December 1998, because patients younger than 45 years are at higher risk of excess mortality (Casadebaig and Quemada, 1991; Meloni et al., 2006). To avoid the potential confounding factors of institutionalisation and chronicity (Cuijpers and Smit, 2002), those who had been hospitalized for schizophrenia, bipolar disorder, or major depression or were under any psychiatric treatment for schizophrenia during the previous 2-year period (1996–97) were excluded from the study cohort. In addition, patients agedb45 years undergoing an appendectomy in 1998 with an ICD-OP code of 47.0 as the primary operative procedure were selected as a control group, since data on the general population are not available from the NHIRD. The reason for selecting appendectomy patients as the control group was that patients undergoing appendec-tomies are relatively indistinguishable from the general population. No statistically signiﬁcant differences were found between the control cohort and the general population in Taiwan with regard to either gender (P N 0.05) or age (P N 0.05).

Appendectomy patients were, however, excluded if they had ever been diagnosed as having any major psychiatric disorder at the time of their recruitment.

In total, 26 374 eligible patients were ultimately selected for this study, including 5515 suffering from schizophrenia, 555 from major depression, and 1581 from bipolar disorder, as well as 18 754 who had undergone an appendectomy. Follow-up of each individual patient was undertaken from recruitment until the end of 2004, with all patients identiﬁed from administrative data to determine whether any had died. Both natural and unnatural causes of deaths were examined, with unnatural causes including suicides, accidents, and homicides, while all other deaths were deﬁned as due to natural causes (Honkonen et al., 2008).

The regression modeling also adjusted for sociodemographic variables including age (b18, 18–24, 25–34, and ≥35 years), gender, medical co-morbidities, level of urbanisation and the geographic location of the community in which the patient resided (northern, central, eastern and southern Taiwan). Details on co-morbid medical disorders, including hypertension, diabetes, renal disease, and chronic obstructive pulmonary disease (COPD), were also extracted from the claims data at the time of the index discharge, because these conditions may exacerbate the risk of mortality.

The urbanisation levels in Taiwan consist of seven strata, with level 1 referring to the ‘most urbanised’ communities and level 7 referring to the ‘least urbanised’ communities, according to standards published by the Taiwanese National Health Research Institute. However, given that there were only very small numbers of schizophrenia cases in levels 5, 6 and 7, these three levels were combined into a single group, which was thereafter referred to as level 5.

2.3. Statistical analysis

The SAS statistical package (SAS System for Windows, version 8.2) was used to perform the statistical analyses in this study. Pearsonχ2

tests were used to examine differences in sociodemographic characteristics, co-morbid medical disorders, and the risk of stroke development among the four cohorts (schizophrenia, major depression, bipolar disorder and appendectomy). The 6-year survival rate was then estimated using the Kaplan–Meier method, with the log-rank test used to examine differences among cohorts. Cox proportional hazard regressions were also carried out as a means of computing the adjusted 6-year hazard ratios, following adjustment for the variables above. Finally, hazard ratios (HRs) are presented along with the 95% conﬁdence intervals (95% CIs), adopting a signiﬁcance level of 0.05 (Pb0.05) for this study.

3. Results

Of the total sample of 26 374 patients under the age of 45 years,

52.1% were male, 30.6% were aged between 25 and 34 years and 0.3%

had conditions complicated by hypertension, 0.1% by cardiovascular

diseases, 0.5% by diabetes, 0.1% by renal diseases and 0.3% by COPD.

Details of the distribution of demographic characteristics and

co-morbid medical disorders for the sampled patients are provided in

Table 1

by disorder type.

Compared to the appendectomy patients, patients with major

depression were more likely to be female, aged between 35 and

44 years, with conditions complicated by diabetes, residing in

more-urbanized communities and in northern Taiwan at the time of the

index discharge (all P b0.001).

Table 1

also shows that there were

signi

ﬁcant differences in the distributions of gender (Pb0.001), age

(P b 0.001), hypertension (P b 0.001), diabetes (P = 0.009),

urbaniza-tion level (P b 0.001), and geographic region (P b0.001).

The log-rank test indicated that patients with schizophrenia, major

depression, and bipolar disorder had signi

ﬁcantly lower 6-year

survival rates compared with appendectomy patients (P b 0.001).

The 6-year survival rate, crude HRs, and adjusted HRs of mortality for

the four cohorts are presented in

Table 2

, which reveals that 6.8%, 5.8%,

5.6% and 1.2% of patients with schizophrenia, major depression,

bipolar disorder, and an appendectomy, respectively, died during the

6-year follow-up period. For patients with schizophrenia, major

depression or bipolar disorder, the risks of dying during the

follow-up period were 6.113 (95% CI = 5.169

–7.230, Pb0.001), 5.191 (95%

CI = 3.580

–7.526, P b0.001), and 4.952 (95% CI = 3.864–6.347,

P b 0.001) times that for appendectomy patients, respectively. After

adjusting for patient's age, gender, co-morbidities, urbanization level,

and geographic region, the relationships still remained; for patients

with schizophrenia, major depression, or bipolar disorder, the risks of

dying during the follow-up period were 4.614 (95% CI = 3.870

–5.500,

P b 0.001), 3.707 (95% CI = 2.544–5.401, P b 0.001), and 3.866 (95%

CI = 3.003

–4.978, Pb0.001) times that for appendectomy patients,

(3)

respectively. For female patients with schizophrenia, major

depres-sion, or bipolar disorder, the adjusted HRs of dying were all higher

than those for male patients.

Table 2

also shows the HRs by natural and non-natural deaths.

Surprisingly, the adjusted HRs of dying of non-natural causes during

the 6-year follow-up period were 16 316 (95% CI = 9.877

–26.952,

P b 0.001), 14 626 (95% CI = 6.909–30.962, P b 0.001) and 8 481 (95%

CI = 4.357

–16.513, Pb0.001) times for female patients with

schizo-phrenia, major depression and bipolar disorder, respectively, as high

as for female appendectomy patients.

Table 3

describes proportionate mortality by disorder type. Not

surprisingly, over 25% (out of total deaths) of patients with

schizophrenia, major depression, or bipolar disorder died of suicide.

In comparison, only 7.9% (out of total deaths) of patients undergoing

an appendectomy committed suicide. Patients with major depression

or bipolar disorder did not have a signi

ﬁcantly higher rate of death due

to circulatory or respiratory diseases than those undergoing an

appendectomy.

4. Discussion

This is the

ﬁrst report on mortality among psychiatric patients,

who were fully insured, in a universal NHI programme in Taiwan. An

important

ﬁnding is that despite the implementation of the NHI in

1995, providing universal health-care resources to all citizens,

mortality remained high and 6-year survival rates were signi

ﬁcantly

lower for psychiatric patients, with females constituting a higher risk

group. Compared with appendectomy patients, the adjusted hazard

ratios were signi

ﬁcantly elevated among schizophrenic, depressive

and manic patients during a 6-year follow-up period, with a higher

ratio of deaths by unnatural causes.

Excess mortality among psychiatric patients has been previously

reported, with overall standardized mortality ratios of psychiatric to

general populations in different regions and countries, ranging from

approximately 1.5

–3.1 for males and 1.9–4.8 for females (

Brook, 1985;

Wood et al., 1985

). Consistent with those reports, increased mortality,

from both natural and non-natural causes, was still apparent among

major psychiatric patients in the 6-year follow-up period. The

prevalence of certain physical illnesses is also higher among

psychiatric patients. For example,

Johannessen et al. (2006)

reported

an elevated level of hypertension in bipolar patients, and our study

also observed a higher tendency to hypertension and diabetes among

patients with schizophrenia and affective disorders. As regional

differences have been observed in incidence of schizophrenia (

Schelin

et al., 2000

), we found that the distribution of psychiatric inpatients

was associated with urbanisation level and geographic region, with

depressive people living mainly in the most urbanised communities.

Comparing our study results with those reported before the NHI era

in Taiwan, no declining mortality trend was identi

_{ﬁed. Before the}

introduction of the NHI,

Chen et al. (1996)

recruited a cohort of

psychiatric inpatients from 1987 to 1988 and followed them up for

6 years to the end of 1993. Despite applying various sampling methods,

our

ﬁnding of excess mortality among psychiatric inpatients after the

implementation of the NHI appears consistent with that from the

preceding period.

Researchers have proposed that universal health insurance might

increase health care use among both young and elderly populations

(

Card et al., 2004; Decker and Remler, 2004

). Findings from the NHI in

Table 1

Demographic characteristics and comorbid medical disorders of schizophrenia, major depression, bipolar disorder, and appendectomy patients in Taiwan in 1998.

Variable Schizophrenia Depression Bipolar disorder Appendectomy P value

(N = 5515) (N = 555) (N = 1581) (N = 18,723) n % n % n % n % Gender Male 3256 59.0 201 36.2 780 49.3 9497 50.7 b0.001 Female 2259 41.0 354 63.8 801 50.8 9226 49.3 Age (year) b18 169 3.1 17 3.1 89 5.8 6029 32.2 b0.001 18–24 1032 18.7 76 13.7 348 22.0 3938 21.0 25–34 2367 42.9 205 36.9 606 38.2 4887 26.1 35–44 1947 35.3 257 46.3 538 34.0 3869 20.7 Hypertension Yes 29 0.5 3 0.5 12 0.8 31 0.2 b0.001 No 5486 99.5 552 99.5 1569 99.3 18,692 99.8 Cardiovascular disease Yes 6 0.1 2 0.4 3 0.2 15 0.1 0.118 No 5509 99.9 553 99.6 1578 99.8 18,708 99.9 Diabetes Yes 41 0.7 4 0.7 10 0.6 74 0.4 0.009 No 5474 99.3 551 99.3 1571 99.4 18,649 99.6 Renal disease Yes 7 0.1 0 0 1 0.1 17 0.1 No 5508 99.9 555 100 1580 99.9 18,706 99.9 COPD Yes 15 0.3 1 0.2 5 0.3 47 0.3 0.943 No 5500 99.7 554 99.8 1576 99.7 18,676 99.7 Urbanization level 1 1896 34.4 280 50.5 613 38.9 6589 35.2 b0.001 2 2039 37.0 238 42.9 609 38.5 7493 40.0 3 515 9.3 7 1.3 142 9.0 1840 9.8 4 808 14.7 30 5.4 203 12.8 2341 12.5 5 257 4.7 0 0 14 0.9 460 2.5 Geographic region Northern 2234 40.2 274 49.4 720 45.4 8450 45.1 b0.001 Central 1245 22.4 66 11.9 377 23.9 4200 22.4 Southern 1746 31.4 199 35.9 386 24.5 5585 29.8 Eastern 339 6.1 16 2.9 98 6.2 488 2.6

(4)

Taiwan also indicated that discrepancies in access to health care were

indeed signi

ﬁcantly reduced after a National Health Insurance

programme with comprehensive coverage was initiated (

Chen et al.,

2007

). Although we anticipated that higher health-care use would be

associated with better health outcomes, our study was unable to

ﬁnd

evidence of reduced mortality among psychiatric patients. Some prior

studies have obtained similar results,

ﬁnding that although health

insurance coverage might improve self-reported health status (

Card

et al., 2004

), mortality remained intact (

Card et al., 2004; Chen et al.,

2007

).

Several plausible reasons might explain this

ﬁnding. First, 6-year

mortality might not be a suf

_{ﬁciently sensitive index for assessing the}

effects of a health-care system on health outcomes. The NHI program

has only recently been implemented for that long, and mortality rates

are established based upon lifetime investment in health.

Second, although the NHI in Taiwan was reported to promote

greater healthcare use of both inpatient and outpatient services

signi

ﬁcantly among low- and middle-income groups, more than

among the higher-income class, whether similar patterns can be

observed for discrepancies between psychiatric and general

popula-tions remains a question. As reported by

Kisely et al. (2007)

, mentally

ill persons with circulatory disease (including ischaemic heart disease

and stroke), especially those who had been psychiatric inpatients,

were signi

ﬁcantly less likely to undergo specialised or

revascularisa-tion procedures. Psychiatric illness was associated with reduced

access to some medically necessary procedures, even in a universal

health-care system aimed at producing equality in service delivery.

Furthermore, psychiatric patients' motivation towards recovery and

compliance with treatment might be involved in the quality of

healthcare they receive.

Third, access to or use of health-care services might not be the

main determinant of health. Other factors, including environment,

lifestyle, or health behaviors are also substantial in in

ﬂuencing a

person's health status (

Marmot and Wilkinson, 1999

). These factors

cannot be directly modi

ﬁed by the introduction of the NHI.

Aside from implementation of the NHI, various factors might help

explain the reasons why psychiatric patients possess higher mortality

risk from both natural and unnatural causes of deaths. Psychiatric

patients may receive poorer quality health care (

Hewer et al., 1995

),

including insuf

ﬁcient diagnosis or treatment of their physical illnesses

(

Corten et al., 1991

). Patients might be limited in their ability to

recognize and communicate physical disturbances to others. Some

demographic, urbanisation, or hazardous health behaviours might

contribute to these differences. For example, higher cardiovascular

morbidity and mortality among depressive patients might possibly be

due their looser control of hypertension (

Davidson et al., 2000

).

Lifestyle, such as alcohol or tobacco use, less physical activity, and

unhealthy eating habits may play a part in the proportionately higher

numbers of deaths (

Brown et al., 2000; Cuijpers and Smit, 2002

).

Factors associated with patients' speci

ﬁc psychiatric symptoms or

side-effects of psychotropic drugs might also be involved in the excess

mortality among persons with mental illness (

Appleby et al., 2000;

Zarate and Patel, 2001

).

Further, higher suicides rates and incidence of accidental deaths

due to hazardous activities were considered to be responsible for

increased numbers of death by unnatural causes. Approximately, 28%

of the excess mortality might be attributed to suicide and 12% to

accidents among schizophrenic patients (

Brown, 1997

). Proportionate

mortality from suicide in our sample was high. Based upon reports

from Department of Health, Executive Yuan, Taiwan, suicide rates rose

from 9.97 (per 100 000) in 1998 to 15.31 (per 100 000) in 2004 among

the general population in Taiwan, with an even higher rise among

psychiatric patients. Further, previous studies have shown that

underreporting of suicide is commonplace (

Dijkhuis et al., 1994;

Wang and Chou, 1997

). Possible reasons for the growing trend might

include increased economic burdens, psychological distress,

unem-ployment rates, psychiatric problems (e.g., alcohol use, major

depression) and other psychosocial situations (e.g., high divorce

rate) (

Yip, 1996; Wang and Chou, 1997

).

Several limitations of this study merit attention. First, typical

concerns associated with mortality analysis include the quality,

extensiveness, and accuracy of death certi

ﬁcates and death registries

Table 2

Six-year survival rates and hazard ratios by disorder type.

Variable Six-year

survival rate

Crude hazard ratio/95% conﬁdence interval (CI)

Adjusted hazard ratioa_{/95% CI} All deaths Total Disorder type Schizophrenia 93.2 6.113 (5.169–7.230)*** 4.614 (3.870–5.500)*** Major depression 94.2 5.191 (3.580–7.526)*** 3.707 (2.544–5.401)*** Bipolar disorder 94.5 4.952 (3.864–6.347)*** 3.866 (3.003–4.978)*** Appendectomy 98.9 1.000 1.000 Male Disorder type Schizophrenia 93.2 4.807 (3.890–5.940)*** 3.477 (2.791–4.329)*** Major depression 92.0 5.664 (3.376–9.501)*** 3.786 (2.245–6.385)*** Bipolar disorder 93.1 4.844 (3.538–6.633)*** 3.733 (2.713–5.135)*** Appendectomy 98.5 1.000 1.000 Female Disorder type Schizophrenia 93.3 8.504 (6.451–11.210)*** 6.925 (5.158–9.299)*** Major depression 95.5 5.687 (3.315–9.757)*** 4.619 (2.659–8.023)*** Bipolar disorder 95.8 5.275 (3.517–7.910)*** 4.406 (2.905–6.681)*** Appendectomy 99.2 1.000 1.000 Natural deaths Total Disorder type Schizophrenia 96.2 4.977 (4.018–6.166)*** 3.457 (2.780–4.342)*** Major depression 97.7 2.975 (1.651–5.361)*** 1.897 (1.048–3.434)* Bipolar disorder 97.1 3.843 (2.749–5.373)*** 2.844 (2.025–3.996)*** Appendectomy 99.2 1.000 1.000 Male Disorder type Schizophrenia 96.6 4.105 (3.112–5.416)*** 2.802 (2.109–3.722)*** Major depression 96.8 3.496 (1.529–7.997)*** 2.145 (0.933–4.932) Bipolar disorder 97.2 3.647 (2.338–5.690)*** 2.710 (1.729–4.247)*** Appendectomy 99.1 1.000 1.000 Female Disorder type Schizophrenia 96.9 6.353 (4.530–8.909)*** 4.584 (3.194–6.577)*** Major depression 98.5 2.901 (1.250–6.732)* 1.980 (0.843–4.652) Bipolar disorder 97.7 4.200 (2.520–6.997)*** 3.220 (1.906–5.439)*** Appendectomy 99.4 1.000 1.000 Non-natural deaths Total Disorder type Schizophrenia 96.3 9.444 (7.244–12.312)*** 7.945 (6.007–10.509)*** Major depression 96.1 9.827 (6.057–15.943)*** 8.276 (5.050–13.563)*** Bipolar disorder 96.8 7.934 (5.535–11.372)*** 6.737 (4.658–9.743)*** Appendectomy 99.6 1.000 1.000 Male Disorder type Schizophrenia 96.3 6.558 (4.750–9.053)*** 5.129 (3.653–7.201)*** Major depression 94.8 9.322 (4.747–18.304)*** 7.050 (3.554–13.988)*** Bipolar disorder 95.7 7.682 (4.982–11.846)*** 6.263 (4.019–9.760)*** Appendectomy 99.4 1.000 1.000 Female Disorder type Schizophrenia 96.2 16.974 (10.505–27.337)*** 16.316 (9.877–26.952)*** Major depression 96.9 14.029 (6.764–29.096)*** 14.626 (6.909–30.962)*** Bipolar disorder 98.0 8.941 (4.666–17.134)*** 8.481 (4.357–16.513)*** Appendectomy 99.8 1.000 1.000 *P b 0.05; ***P b 0.001. a

Hazard ratios were adjusted for patient's age, gender, medical comorbidity (including hypertension, diabetes, renal disease, and chronic obstructive pulmonary disease), urbanization, and geographical location.

(5)

(

Martin, 1985; Flanders, 1992

). Because of the mandatory registration

of all deaths in Taiwan, data from the death registry should be

particularly accurate and comprehensive and are thus the most

appropriate data available for mortality analysis. Meanwhile, although

ICD-9-CM codes were used consistently throughout our study period,

the possibility of changes in coding practices on death certi

ﬁcates

should be considered when interpreting trends in natural and

non-natural deaths. In spite of the

ﬂuctuation of non-natural deaths by

various methods, there has been a steady rise in natural deaths which

may re

ﬂect no obvious attitudinal changes in coding practices on

death certi

_{ﬁcates over the study period. However, because of the fact}

that claims and death certi

ﬁcates were used to deﬁne the diagnosis in

this study, the potential bias related to the reliability of diagnoses

existed. Second, the reliability and validity of the NHI claims data and

over- and under-diagnoses should be considered. To deter upcoding,

the NHI Bureau routinely performs sample cross-checks of each

hospital's claims with its medical charts, followed by punitive

measures for coding infractions. It generally accepted that the NHI's

checks and balances promote appropriate coding. However, the

reliability of diagnoses might vary among different psychiatric

disorders, with diagnostic accuracy being higher for schizophrenia

than for major depression (

Chien et al., 2007

). Lower diagnostic

accuracy might attenuate the mortality gap slightly between patients

with major depression and the general population.

Third, social stigmas might lead some patients to deny symptoms.

Because of possible under-representation of psychiatric inpatients in

this study, excess mortality among psychiatric patients might slightly

be under-estimated. Further, the use of psychiatric inpatients might

favour the recruitment of more physically vulnerable persons, because

psychiatric patients with physical illness might be more likely to be

hospitalised than those with mental disorders alone. Finally, this study

was unable to investigate the effects of such risk factors as patients'

socioeconomic status, marital status, education, and family history of

psychiatric conditions related to mortality. Data regarding concurrent

use alcohol, tobacco or other substances were not available.

The continuing excess mortality among psychiatric patients from

both natural and unnatural causes and across different categories of

mental illnesses highlights the ongoing signi

ﬁcance of this problem.

Even after implementation of a universal health-care system in Taiwan

that is free at the point of delivery, this issue remains. Thus, a signi

ﬁcant

next step would be to evaluate the quality of medical care received by

psychiatric patients, ascertaining its weaknesses so that further

improvements can be made. Whether inequitable access to some

medically necessary procedures exists for psychiatric patients under the

NHI and developing strategies to reduce these disparities warrant future

investigation. Consistently high rates of unnatural deaths, especially for

female patients, are warning signals of de

ﬁciencies in psychiatric

services. More deaths from unnatural causes occur particularly during

or shortly after a psychiatric hospitalisation, reinforcing the need for

more intensive care or follow-up during this period. Routine monitoring

of the side effects of different types and doses of psychotropic

prescriptions, compliance with the medication regimen, and

encoura-ging healthier behaviors and lifestyles deserve special and constant

attention.

Our study supports the continuing development of strategies to

ef

_{ﬁciently reduce death rates, such as improvements in health-care}

quality, promoting use of medical services, and keeping track of

disturbing psychotic or physical symptoms among psychiatric

patients. Regular monitoring of mortality trends among psychiatric

patients and the effect of the NHI program on death rates should

continue.

Acknowledgements

This study is based in part on data from the National Health Insurance Research Database provided by the Bureau of National Health Insurance, Department of Health, Taiwan and managed by the National Health Research Institutes. The interpretations and conclusions contained herein do not represent those of the Bureau of National Health Insurance, Department of Health or the National Health Research Institutes.

References

Appleby, L., Thomas, S., Ferrier, N., Lewis, G., Shaw, J., Amos, T., 2000. Sudden unexplained death in psychiatric in-patients. British Journal of Psychiatry 76, 405–406.

Black, D.W., Warrack, G., Winokur, G., 1985. Excess mortality among psychiatric patients. The Iowa Record-Linkage Study. Journal of the American Medical Association 253, 58–61.

Brook, O.H., 1985. Mortality in the long-stay population of Dutch mental hospitals. Acta Psychiatrica et Neurologica Scandinavica 71, 626–635.

Brown, S., 1997. Excess mortality of schizophrenia. A meta-analysis. British Journal of Psychiatry 171, 502–508.

Brown, S., Inskip, H., Barraclough, B., 2000. Causes of the excess mortality of schizophrenia. British Journal of Psychiatry 177, 212–217.

Card, D., Dobkin, C., Maestas, N., 2004. The impact of nearly universal insurance coverage on health care utilization and health: evidence from Medicare. NBER Working Paper Series, vol. 13065.

Casadebaig, F., Quemada, N., 1991. Changes in mortality among psychiatric inpatients, 1968–1982. Social Psychiatry and Psychiatric Epidemiology 26, 78–82.

Chen, W.J., Huang, Y.J., Yeh, L.L., Rin, H., Hwu, H.G., 1996. Excess mortality of psychiatric inpatients in Taiwan. Psychiatry Research 62, 239–250.

Chen, L., Yip, W., Chang, M.C., Lin, H.S., Lee, S.D., Chiu, Y.L., Lin, Y.H., 2007. The effects of Taiwan's National Health Insurance on access and health status of the elderly. Health Economics 16, 223–242.

Chien, I.C., Kuo, C.C., Bih, S.H., Chou, Y.J., Lin, C.H., Lee, C.H., Chou, P., 2007. The prevalence and incidence of treated major depressive disorder among National Health Insurance enrollees in Taiwan, 1996 to 2003. Canadian Journal of Psychiatry 52, 28–36. Table 3

Proportionate mortality by disorder type (N = 34,157).

Cause of deaths (ICD-9-CM codes 001-999) Schizophrenia Depression Bipolar disorder Appendectomy Total

(n = 377) (n = 32) (n = 88) (n = 216) (n = 713)

n (%) n (%) n (%) n (%) n (%)

Infectious and parasitic diseases (001–139) 8 (2.1) 0 (0.0) 2 (2.3) 1 (0.5) 11 (1.5)

Neoplasms (140–239) 26 (6.9) 0 (0.0) 12 (13.6) 56 (25.9) 94 (13.2)

Endocrine, nutritional and metabolic disease and immunity disorders (240–279) 11 (2.9) 0 (0.0) 3 (3.4) 10 (4.6) 24 (3.4)

Diseases of blood and blood-forming organs (280–289) 0 (0.0) 0 (0.0) 1 (1.1) 0 (0.0) 1 (0.1)

Diseases of the nervous system and sense organs (320–389) 3 (0.8) 0 (0.0) 1 (1.1) 8 (3.7) 12 (1.7)

Disease of the circulatory system (390–459) 30 (8.0) 1 (3.1) 1 (1.1) 18 (8.3) 50 (7.0)

Diseases of the respiratory system (460–519) 15 (4.0) 1 (3.1) 2 (2.3) 5 (2.3) 23 (3.2)

Disease of the digestive system (520–579) 26 (6.9) 5 (15.6) 10 (11.4) 26 (12.0) 67 (9.4)

Diseases of the genitourinary system (580–629) 1 (0.3) 0 (0.0) 1 (1.1) 3 (1.4) 5 (0.7)

Disease of the skin and subcutaneous tissue (680–709) 3 (0.8) 0 (0.0) 1 (1.1) 1 (0.5) 5 (0.7)

Disease of the musculoskeletal system and connective tissue (710–739) 0 (0.0) 0 (0.0) 0 (0.0) 1 (0.5) 1 (0.1)

Congenital anomalies (740–759) 3 (0.8) 0 (0.0) 0 (0.0) 0 (0.0) 3 (0.4)

Symptoms, signs, and ill-deﬁned conditions (780–799) 33 (8.8) 3 (9.4) 4 (4.6) 10 (4.6) 50 (7.0)

Injury and poisoning (800–999) 69 (18.3) 6 (18.8) 17 (19.3) 50 (23.2) 142 (19.9)

Suicide (E950–E959) 101 (26.8) 11 (34.4) 24 (27.3) 17 (7.9) 153 (21.5)

Complications of pregnancy, childbirth, and the puerperium (630–679) 0 (0.0) 0 (0.0) 0 (0.0) 1 (0.5) 1 (0.1)

(6)

Corten, P., Ribourdouille, M., Dramaix, M., 1991. Premature death among outpatients at a community mental health center. Hospital and Community Psychiatry 42, 1248–1251.

Corten, P., Ribourdouille, M., Hermann, P., Rorsman, B., Sims, A., 1988. Epidemiological survey of the“natural” mortality in psychiatry. Acta Psychiatrica Belgica 88, 349–371.

Cuijpers, P., Smit, F., 2002. Excess mortality in depression: a meta-analysis of community studies. Journal of Affective Disorder 72, 227–236.

Davidson, K., Jonas, B.S., Dixon, K.E., Markovitz, J.H., 2000. Do depression symptoms predict early hypertension incidence in young adults in the CARDIA study? Coronary artery risk development in young adults. Archives of Internal Medicine 160, 1495–1500.

Decker, S., Remler, D., 2004. How much might universal health insurance reduce socioeconomic disparities in health? A comparison of the US and Canada. NBER Working Paper Series, vol. 10715.

Dijkhuis, H., Zwerling, C., Parrish, G., Bennett, T., Kemper, H.C., 1994. Medical examiner data in injury surveillance: a comparison with death certiﬁcates. American Journal of Epidemiology 139, 637.

Flanders, W.D., 1992. Inaccuracies of death certiﬁcate information. Epidemiology 3, 3–5. Hewer, W., Rössler, W., Fätkenheuer, B., Löfﬂer, W., 1995. Mortality among patients in psychiatric hospitals in Germany. Acta Psychiatrica et Neurologica Scandinavica 91, 174–179.

Honkonen, H., Mattila, A.K., Lehtinen, K., Elo, T., Haataja, R., Joukamaa, M., 2008. Mortality of Finnish acute psychiatric hospital patients. Social Psychiatry and Psychiatric Epidemiology 43 (8), 660–666.

Johannessen, L., Strudsholm, U., Foldager, L., Munk-Jørgensen, P., 2006. Increased risk of hypertension in patients with bipolar disorder and patients with anxiety compared to background population and patients with schizophrenia. Journal of Affective Disorder 95, 13–17.

Kisely, S., Smith, M., Lawrence, D., Cox, M., Campbell, L.A., Maaten, S., 2007. Inequitable access for mentally ill patients to some medically necessary procedures. Canadian Medical Association Journal 176, 779–784.

Lawrence, D.M., Holman, C.D., Jablensky, A.V., Hobbs, M.S., 2003. Death rate from ischaemic heart disease in Western Australian psychiatric patients 1980–1998. British Journal of Psychiatry 182, 31–36.

Marmot, M., Wilkinson, R.G., 1999. Social Determinants of Health. Oxford University Press, Oxford, UK.

Martin, R.L., 1985. Methodological and conceptual problems in the study of mortality in psychiatry. Psychiatric Developments 3, 317–333.

Meloni, D., Miccinesi, G., Bencini, A., Conte, M., Crocetti, E., Zappa, M., Ferrara, M., 2006. Mortality among discharged psychiatric patients in Florence, Italy. Psychiatry Services 57, 1474–1481.

Mortensen, P.B., Juel, K., 1990. Mortality and causes of death in schizophrenic patients in Denmark. Acta Psychiatrica et Neurologica Scandinavica 81, 372–377.

Mortensen, P.B., Juel, K., 1993. Mortality and causes of death in ﬁrst admitted schizophrenic patients. British Journal of Psychiatry 163, 183–189.

Räsänen, S., Hakko, H., Viilo, K., Meyer-Rochow, V.B., Moring, J., 2003. Excess mortality among long-stay psychiatric patients in northern Finland. Social Psychiatry and Psychiatric Epidemiology 38, 297–304.

Schelin, E.M., Munk-Jørgensen, P., Olesen, A.V., Gerlach, J., 2000. Regional differences in schizophrenia incidence in Denmark. Acta Psychiatrica et Neurologica Scandinavica 101, 293–299.

Sims, A., 1987. Why the excess mortality from psychiatric illness? British Medical Journal (Clinical Research Ed.) 294, 986–987.

Tsuang, M.T., 1978. Suicide in schizophrenics, manics, depressives, and surgical controls. A comparison with general population suicide mortality. Archives of General Psychiatry 35, 153–155.

Tsuang, M.T., Woolson, R.F., 1978. Excess mortality in schizophrenia and affective disorders. Do suicides and accidental deaths solely account for this excess? Archives of General Psychiatry 35, 1181–1185.

Tsuang, M.T., Simpson, J.C., 1985. Mortality studies in psychiatry. Should they stop or proceed? Archives of General Psychiatry 42, 98–103.

Wang, C.S., Chou, P., 1997. An analysis of unnatural deaths between 1990 and 1994 in A-Lein, Taiwan. Injury 28 (3), 203–208.

Wood, J.B., Evenson, R.C., Cho, D.W., Hagan, B.J., 1985. Mortality variations among public mental health patients. Acta Psychiatrica et Neurologica Scandinavica 72, 218–229. Yip, P.S., 1996. Suicides in Hong Kong, Taiwan and Beijing. British Journal of Psychiatry

169 (4), 495–500.

Zarate, C.A., Patel, J., 2001. Sudden cardiac death and antipsychotic drugs: do we know enough? Archives of General Psychiatry 58, 1168–1171.