| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
BRIEF REPORT |
Correspondence: Address correspondence to Tomoe Sakashita, Faculty of Health Sciences, Aomori University of Health and Welfare, 58-1, Mase Hamadate, Aomori 030-8505, Japan. E-mail: t_sakashita{at}auhw.ac.jp
Abstract
Purpose: In this study we evaluate outcomes of a community-based program to prevent suicide among elderly individuals aged 65 and older. Design and Methods: We used a quasi-experimental design with intervention and referent municipalities. The program included a 7-year implementation of depression screening with follow-up by general practitioners and a 10-year implementation of public education conducted in Yasuzuka (population 4,940; elderly suicide rate for women, 275/100,000; for men, 323/100,000). We estimated changes in the risk of completing suicide before and after the 10-year implementation by the incidence-rate ratio (IRR).Results: The risk for women in the intervention area was reduced by 64% (age-adjusted IRR = 0.36; 95% confidence interval = 0.14–0.93), whereas there was no significant change in the risk for men in the intervention area and either men or women in the referent municipalities. A ratio of the IRR for women aged 65 to 74 in the intervention area to that in its prefecture was estimated at 0.23 (90% confidence interval = 0.05–0.99), showing that the risk reduction was greater than the secular trend. Implications:The management of depression by use of community resources involving public health and primary care physicians is effective in the prevention of suicide for elderly women but uncertain for men.
Key Words: Community-based intervention • Controlled clinical trial • Gender difference • Incidence rate ratio • Quasi-experimental design
To our knowledge, there are few community-based suicide-intervention trials that have used rigorous study designs to report on the outcome of completed suicide (Mann, Hendin, Rihmer, Kalmar, & Szanto, 2005). Of those trials, seven were cohort studies that resulted in the lowering of suicides or the suicide rate in localized areas with the following prevention strategies: (a) general practitioner education aimed at improving detection and treatment of depression (Mann et al.; Rutz, Von Knorring, & Walinder, 1989); (b) public education campaigns aimed at improving recognition of depression with suicide risk, help seeking, and developing a community network (Oyama, Koida, Sakashita, & Kudo, 2004; Oyama et al., 2005; Takahashi et al., 1998); (c) depression screening for the general population, aimed at identifying at-risk individuals and directing them to treatment (Mann et al.; Oyama et al., 2004; Takahashi et al.); (d) hotline services and phone checks for the residents who need help, aimed at monitoring and providing prompt intervention (De Leo, Dello Buono, & Dwyer, 2002); and (e) restriction of access to lethal means (Loftin, McDowall, Wiersema, & Cottey, 1991).
A comparison of results from the five studies focused on community-based management of depression shows a gender difference in the effect of management practices on suicide risk. Of the five studies, three reduced the suicide risk for both men and women by using the resources of public health and psychiatric care (Mann et al., 2005; Oyama et al., 2004; Takahashi et al., 1998), and two reduced the risk for women but not men by using the resources of public health and primary care (Rutz et al., 1989; Oyama et al., 2005). These findings suggest that it is difficult both to have an impact on men without psychiatric care and to maintain such psychiatric care resources in communities that do not have a psychiatrist.
We implemented a community-based program including both depression screening and public education for elderly residents, using one community's public health and primary care resources and a neighboring community's psychiatric care resources. Our aim in the present study was to evaluate the effectiveness of this program for the prevention of suicide among elders by using a quasi-experimental design involving comparison of the intervention community with neighboring communities.
Methods
Intervention Program
We implemented the community-based intervention program from January 1991 to December 2000 in the municipality of Yasuzuka, a town located in the southwest of Niigata prefecture in central Japan. Japan is divided into 47 prefectures (large-area local governing units), and each is further divided into municipalities (basic local-level governing units: cities, towns, or villages). Several neighboring towns or villages also constitute a county. Yasuzuka is a rural agricultural area having a high suicide rate among the elderly population. The participatory program included (a) public health education from 1991 to 2000 and (b) screening for depression with follow-up from 1991 to 1997, using the public health and primary care resources in the town with the assistance of the psychiatric care resources in the neighboring county.
Initially, we had all households in the intervention community notified by means of the municipal newspaper that the municipality would implement a program providing voluntary, anonymous, annual depression screening and follow-up referrals, if appropriate, for residents aged 65 and older. Next, elderly people were called upon to attend the mental health workshops in a small district of the town. During the workshops, municipal public health nurses conducted the psychoeducational program promoting awareness of depression and suicide risk, and information about the depression screening, with assistance from our colleagues in the next county.
Finally, we had the annual two-step screening conducted by three municipal public health nurses in order to detect depression at an early stage (Figure 1). In the first-stage screen, nurses distributed a self-report questionnaire based on a Japanese version of Zung's Self-Rating Depression Scale (SDS; Zung, 1965) to all residents aged 65 and older. Family members or community welfare professionals were permitted to help respondents complete the SDS. Those residents who agreed to participate in the program completed the SDS, which the welfare professionals collected.
|
Study Design
We used a quasi-experimental design, with preimplementation and postimplementation registrations covering the population aged 65 and older in the intervention area and the neighboring reference area. We studied the intervention effect by using the prospective registration of all suicide episodes during the study period.
Study Population
We divided the study periods into two 10-year stages: the baseline stage, from January 1981 through December 1990 (before the implementation), and the implementation stage, from January 1991 through December 2000. The study population consisted of two identical dynamic cohorts, that is, an intervention cohort and a reference cohort between 1981 and 2000. The intervention cohort was the population aged 65 and older in Yasuzuka (total population, 4,940; 2,444 women and 2,496 men), where the percentage of elderly individuals was 20.6% and the average suicide rate per 100,000 for men was 322.6 and for women was 274.6 in the period from 1986 to 1990.
The neighboring 10 municipalities were eligible as study references by use of the following criteria: (a) total population 4,000–10,000, (b) elderly percentage 
15%, and (c) elderly suicide rates 150 in both women and men. Of those municipalities, only Kawanishi fulfilled the criteria and was entered as a neighboring reference (total population, 9,091; percentage of elderly individuals, 18.5%; and average suicide rates for people 65 years of age, 212.2 and 151.9 for men and women, respectively, in 1986–1990). Kawanishi is located in a rural agricultural area that is 35 km east of Yasuzuka. Thus, the reference cohort was composed of the population aged 65 and older in Kawanishi.
The 1990 census showed that the unemployment rates (Yasuzuka, 0.44%; Kawanishi, 1.26%) of the study population were less than two thirds of its prefectural mean (Niigata, 2.02%). The census data also showed that the average yearly income of the study population (Yasuzuka, 2,340,000 yen; Kawanishi, 2,177,000 yen) was around 75% of its prefectural mean (2,636,000 yen). These findings suggest that the characteristics of employment rate and income in the intervention area were similar to those in the reference area.
We derived the data on suicide from the prospective registration of all suicide episodes at the Joetsu and Tokamachi Public Health Centers, which surveyed and controlled the mental health activities in the study municipalities, with the exception of the present intervention program. The diagnoses in the register were based on codes from the International Statistical Classification of Diseases, 9th revision, in which confirmed suicides and probable suicides were combined.
Statistical Analysis
We based our statistical analyses on an incidence rate of completed suicide with stratification for age and gender. The suicide incidence rate is the number of incident suicide cases divided by the population. We derived the incidence of suicide cases and populations from the dynamic cohort during each 10-year stage.
Assessment of Risk Change Between Pretests and Post-tests
We expressed changes in the risk of completing suicide associated with exposure to the community program as the incidence-rate ratio (IRR) between the baseline and the implementation. This is because the risk ratio, indicating the risk change, was interchangeable with the IRR under (a) the assumption of constancy of the incidence rate over each stage and (b) the rarity of suicide (incidence rate less than 0.5%; Rothman & Greenland, 1998). We assessed the IRR of the population aged 65 and older between the two stages by using a corrected Mantel–Haenzel stratified chi-square test that we adjusted for 10-year age categories with the maximum likelihood method. We assessed the homogeneity of the IRR in age levels by means of the Breslow–Day test.
Evaluation of Trend Effect on Risk Change
For the suicide risk in the prefecture including the study municipalities, a gradual upward or downward trend over the past two decades occurs independently of the intervention effect. This phenomenon is referred to as a secular trend. Our analysis compared the IRR in Yasuzuka with that in Niigata prefecture. It evaluated the effect of the secular trend, which is indicated by the risk change in the prefecture, on the change in the suicide risk. We calculated the three-factor interaction parameter obtained from a saturated general log linear model for a 2 x 2 x 2 cross-classification by using the maximum likelihood estimation with the chi-square test (df = 1) for comparison. This parameter expresses a ratio of two IRRs (Agresti, 2002).
We calculated descriptive statistics and confidence intervals (CIs) by using the SPSS 10.0 J software package (SPSS Inc., Chicago, IL).
Results
Implementation of Intervention Program
Table 1 shows the number of participants in the annual two-step screening with follow-up during the initial 7 years in the implementation stage. More than 90% of female or male residents aged 65 and older participated in the first-stage screen. Of those participants, 8% to 14% were enrolled as SDS positive. Of the SDS-positive participants, 49% to 62% took part in the second-stage screen. Of those referred in the second-stage screen, very few (1.2% or fewer yearly) were diagnosed with major or minor depression. This could be because depressed people were less likely to participate in the second-stage screen. All of the RDC-positive participants agreed to be referred to the general practitioner in the town, with a psychiatric consultee-centered approach by telephone. During the implementation stage, the general practitioner performed no postmortems on suicide victims because coroners in other counties performed the postmortems.
|
We held the workshops for health education annually over 10 years in one third of the districts of Yasuzuka. During the workshops, we had municipal public health nurses give basic information about the signs and treatment of depression, as well as how to utilize mental health services. We did not emphasize in the present study that suicide is taboo in most world cultures, although such emphasis had been given in the studies by Takahashi and colleagues (1998) and by Oyama and colleagues (2004).
In addition to the present intervention program, in the intervention and reference communities during the study periods, we implemented secondary prevention programs for individuals with dementia and tertiary prevention programs for individuals with schizophrenia.
Change in Incidence Rates of Suicide in the Study Population
In the intervention area, we observed a 64% reduced risk of suicide among women aged 65 and older (age-adjusted IRR = 0.36, 95% CI = 0.14–0.93) during the implementation stage, compared with the baseline stage (Table 2). In contrast, a 49% reduction of the risk in men aged 65 and older did not reach significance (age-adjusted IRR = 0.51, 95% CI = 0.22–1.19). There was no significant change in the risks for either women or men in the reference community. The study results would have approximately 70% or more statistical power calculated for a two-tailed t test at a p =.05 one-tail level of significance in each cohort by gender.
|
Discussion
Although it is widely known that most elderly people who commit suicide suffer from depression (Conwell, 1997; Shah & De, 1998), there are few interventions proven to reduce suicidal behavior in senior adults (Beautrais, 2002). In the present study, we evaluated a community-based program, which included both depression screening with follow-up by the community's general practitioners and public health education. The suicide risk for elderly women, but not men, in the intervention population was reduced by 64% during the 10-year implementation.
The present study has some limitations resulting from its quasi-experimental design that could significantly affect suicide risk. Because the present study used a preimplementation–postimplementation design with a nonrandomized reference, the following biases could result. First, nonrandomized assignment cannot control for unmeasured confounding variables. For the variables related to suicide, however, the socioeconomic characteristics (i.e., unemployment rates, average yearly income, and welfare service) and the mental health activities (except the present intervention) appear to be similar between the intervention and reference communities, because of similarities in the neighborhoods and because both study populations resided in the same prefecture. Second, the preimplementation–postimplementation design could cause time-dependent confounding, including a secular trend, and a regression effect toward the mean. Nevertheless, the reduction of suicide risk among women in the intervention area was greater than the regional secular trend. A regression effect on the decrease in the suicide rate could be also suspected because of the high baseline rate in the intervention area. However, the suicide rate among elders in the neighboring reference community was continuously high, which counters an explanation of the regression effect.
In addition, the different sources of information might cause source bias because some participants in the first-stage screen completed the SDS with assistance from family or professional staff. The source bias appears not to affect the sensitivity of SDS responses, because there seems to be little possibility that the assistance results in an increase of the number of false-negative participants. There is also no evidence that the ongoing intervention influenced the physicians in that community to alter their death-record-certification habits with respect to suicide, because the coroners in other communities examined all suicide victims.
Lastly, poor participation after the first-stage screen suggests selection bias. The selection bias appears not to reduce the suicide risk because selective dropout of depressed people in the screening process will result in more suicides. Thus, to counter a passive participatory approach, follow-up and intensive outreach to those who drop out would be more effective for suicide prevention among the elderly population.
The findings lend support to the conclusion that the risk reduction by depression screening and public health education, using the community's resources of public health and primary care, was successful for elderly women. The screening with follow-up can intervene on behalf of the overlooked depressed elderly women in the community. Educational activities with information and advice should also improve help-seeking behavior among elderly women.
Of the present results, the reduced risk for women is consistent with the results of three previous studies using community-based screening with the resources of primary or psychiatric care. Our finding of unchanged risk for men with the use of primary care resources contrasts with three prior studies showing reduced risk for men with the use of psychiatric care resources (Mann et al., 2005; Oyama et al., 2004; Takahashi et al., 1998), indicating the benefits of risk reduction among men with the use of psychiatric care resources. This gender difference of the risk changes following the implementation can be explained by (a) the acceptance of more women than men, (b) the lack of psychiatric care ameliorating the suicidal impulse to which men would be more vulnerable than women (Rutz, 2001), and (c) an insufficient sample size to detect the male risk reduction of considerably less than 75% expected in the antecedent study (Oyama et al.).
Footnotes
1 Faculty of Health Sciences, Aomori University of Health and Welfare, Aomori, Japan. 
2 Niigata Psychiatric Center, Nagaoka, Japan.
3 School of Health Sciences, Niigata University, Japan.
4 National Nishigunma Hospital, Shibukawa, Japan.
Decision Editor: Linda S. Noelker, PhD
Received for publication October 6, 2005. Accepted for publication June 30, 2006.
References
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
