Factors associated with suicide mortality among the elderly in Brazilian municipalities between 2005 and 2007

ARTIGO ARTICLE

Factors associated with suicide mortality among the elderly in Brazilian municipalities between 2005 and 2007

Fatores associados com a mortalidade por suicídio de idosos nos municípios brasileiros no período de 2005-2007

Liana Wernersbach Pinto^I; Cosme Marcelo Furtado Passos da Silva^II; Thiago de Oliveira Pires^I; Simone Gonçalves de Assis^I

^ICentro Latino-Americano de Estudos de Violência e Saúde Jorge Careli, Escola Nacional de Saúde Pública, Fundação Oswaldo Cruz. Avenida Brasil 4036/700, Manguinhos. 21040-361 Rio de Janeiro, RJ. [e-mail]

^IIDepartamento de Epidemiologia e Métodos Quantitativos, Escola Nacional de Saúde Pública, Fundação Oswaldo Cruz

ABSTRACT

This scope of this paper was to conduct an ecological analysis of suicide mortality of people aged 60 years or more in Brazilian municipalities between 2005 and 2007, by investigating factors associated with the event. Data on suicide deaths were extracted from the Mortality Information System, codes X60 to X86 and Y87.0 (ICD-10). Poisson, negative binomial and zero-inflated negative binomial (ZINB) regression models were adjusted. The latter exhibited the best results when comparing models. The proportion of non-whites (negative association), the rate of hospitalization for mood disorders (positive association) and sex ratio (negative association) were identified as factors associated with suicide.

Key words: Suicide, The elderly, Ecological study, Zero-inflated negative binomial regression model.

RESUMO

Este trabalho teve como objetivo realizar análise ecológica sobre suicídio de pessoas com 60 anos ou mais nos municípios brasileiros no triênio 2005-2007, investigando-se fatores associados ao evento. Foram utilizados dados referentes aos óbitos por suicídio extraídos do Sistema de Informação sobre Mortalidade (SIM), códigos X60 a X86 e Y87.0 (CID-10). Foram ajustados modelos de regressão de Poisson, binomial negativa e binomial negativa inflacionada de zeros (ZINB). Este último exibiu os melhores resultados quando da comparação de modelos. Foram identificados como fatores associados ao suicídio: proporção de não brancos (associação negativa), taxa de internação por transtornos de humor (associação positiva) e razão de sexo (associação negativa).

Palavras-chave: Suicídio, Idosos, Estudo ecológico, Regressão binomial negativa inflacionada de zeros.

Introduction

Approximately one million people worldwide die of suicide each year, generating a global death rate of 16 deaths per 100,000 inhabitants. This is higher than deaths caused by war and homicide combined¹.

Additionally, in several countries worldwide elderly suicide has become a relevant public health problem, which has been intensifying with population ageing. World Health Organization data show that for men rates have gone from 19.2/100,000 inhabitants aged 15 to 24 to 55.4/100,000 inhabitants for men over 75 years old in 2000². Recent studies have been showing differences in factors for the young elderly (younger than 75) and for older individuals (older than 75). In the United States in 1998 mortality by suicide among individuals older than 65 reached 18% and it is likely that this figure will reach 35% in 2030³.

In Brazil in 2009 external causes (ICD-10 Chapter XX) ranked number seven in causes of death among people aged 60 or older, totaling 21,437 deaths. Of those, 1,378 were suicides. It should also be taken into account that 3,364 deaths in that year were classified according to codes Y10-Y34, that is, as events when it is not possible to distinguish between accidents, self-inflicted harm or assault because the action's intent is recorded as unknown.

According to Mitty and Flores³, risk factors for elderly suicide may be classified according to the following categories: demographic, mental problems, physical discomfort and social problems. According to several authors, depression, social isolation, lack of a support network and loneliness, ideations and previous attempts, and access to the means are also reasons involved in risk for suicide^1,4,5. Among social factors, several scholars mention the marital statuses of single, widow(er) or separated. Also considered as very important is experiencing stressful events, such as losing beloved family members, since those events interfere with psychological, psychiatric and biological factors in the elderly. Among psychosocial factors, alcoholism and use of other drugs are considered relevant^4,6,7.

Epidemiological studies about this topic have been demonstrating that suicide occurs more often among white individuals than non-whites in several countries^2,6,7. Indigenous populations are an exception, since they are at high risk in different contexts^8,9. The lack of mental health services, lack of psychiatric hospital beds and of professionals are all aspects related to the occurrence of suicide among the elderly in many parts of the world¹⁰.

According to the WHO, the issue of suicide has not been properly addressed by countries because they are unaware of the fact that it is a very important issue - however considered as taboo - in most societies. Only a few countries include suicide prevention among their priorities. The issue of death communication also needs substantial improvement, since suicide is highly under-recorded worldwide². Prevention strategies also should be thought in multi-sector terms involving other sectors besides healthcare, such as education, labor, police, justice, religion, laws, politics, media¹¹.

In this paper we aim to present an ecological analysis of suicide in individuals aged 60 or older in Brazilian municipalities from 2005-2007. We investigate, from the epidemiological point of view, variables associated with the phenomenon.

Methods

Mortality data for 2005-2007 were gathered from the Mortality Information System (Sistema de Informação sobre Mortalidade - SIM); the following 10^th Revised International Classification of Diseases (ICD-10) codes were used: X60 to X84 and Y87.0. Population estimates for the middle year in the period being analyzed (2006) were obtained from Brazil's Institute for Geography and Statistics (Fundação Instituto Brasileiro de Geografia e Estatística - IBGE)¹².

Based on a literature review about the subject^1-3,6,13-15 and on accessibility of municipal information we chose as independent variables: sex ratio in the elderly population (female/male), population percentages by skin color (proportion of non-whites), percentage of elderly individuals who share responsibility for the household and marital status (proportion of unmarried individuals), all extracted from the 2010 Census¹⁶; existence of special police stations for the elderly (yes/no) and presence of violence prevention centers (yes/no), both obtained from IBGE's Munic Survey (2009)¹⁷; and percentage of hospital admissions due to mood disorders throughout 2010, extracted from the Hospital Information System (Sistema de Informações Hospitalares - SIH). Those disorders include problems such as depression, manic disorder and bipolar affective disorders. We also employed as an explanatory variable the Firjan Municipal Development Index (Índice Firjan de Desenvolvimento Municipal - IFDM) for 2007. This tool combines information about employment and income, education and health, generating an index whose result ranges from 0 to 1 (where 1 equals the highest development stage). The dependent variable used consisted of the number of deaths by suicide among people aged 60 or older, and that occurred from 2005 to 2007.

In order to verify variables associated with the outcome we used the following models: Poisson's regression, negative binomial and zero-inflated negative binomial (ZINB). Such models were chosen because they are recommended for count data (deaths by suicide)^18,19.

Poisson's model is the standard approach for analyzing this type of datum; however it assumes mean and variance are equal, which may not occur in series with overdispersion or excess zeros. Using Poisson's Regression in such situations leads to underestimating standard-errors of coefficients of regression models, generating confidence intervals that are too narrow and p-values that are too small.

An alternative to solve the overdispersion or excess zero problem in data is using negative binomial distribution. There is a change in its variance function that makes its dispersion parameter allow for an additional variation¹⁹. Another model that may be used in such situations is the zero-inflated regression model^19-21. According to this model the mean structure is modified to allow excess zeros, which are then considered as coming from distinct processes, an inflated and a non-inflated one (inflated part and noninflated part). Models presume that there are two latent groups: the first one generates zeros only and the second generates a Poisson's distribution or negative binomial, which may take values greater than or equal to zero. Therefore, variables associated with each of those parts (inflated and noninflated) are included in the final model. The interest here lies in analyzing factors associated with the noninflated part.

In the modeling process we initially adjusted Poisson's regression models to each of the explanatory variables. Those which presented a general association with the answer variable with p-value ≤ 0.05 were included next in the negative binomial and ZINB regression models. Thus, model adjustment to verify associations between death by suicide and other variables followed the steps below: (1) bivariate analysis using Poisson's regression model; and (2) multivariate analysis using negative binomial regression and the ZINB model. Model selection was performed comparing their log probability, in addition to assessing the number of zeros estimated by each one; this value was then compared with the amount of zeros actually observed. Municipalities' population data were included in regression models as an offset, using a logarithmic function.

Multicollinearity, that is, the existence of a linear correlation between independent variables²², was verified. Based on this analysis we observed the association between variables: proportion of non-whites and IFDM, and thus we adjusted separate models using each of them.

To perform analyses we used the pscl¹⁹ library of public domain software R 2.12.2²³.

Results

This study aimed at identifying variables associated with death by suicide in individuals aged or older in Brazilian cities, from 2005 to 2007.

Initial exploratory analysis demonstrated that nearly 70% of Brazilian municipalities did not record deaths by suicide in individuals aged 60 or older during the period analyzed. Of all 1,659 municipalities that recoded cases from 2005-2007, 30.1% were in the South, 28.9% in the Southeast and 28.5% in the Northeast. In the North and Middle-West regions 4.5% and 8.1% of municipalities, respectively, recorded at least one death by suicide in individuals aged 60 or older during that period. Table 1 shows the 16 municipalities that recorded over 15 deaths during that period. Among those there are nine capitals. The maximum number of deaths occurred in the city of São Paulo (SP). One also notices that three municipalities have less than 50,000 inhabitants: Venâncio Aires (19 deaths), Caxias do Sul (21 deaths) and Pelotas (23 deaths), and all of them are in the state of Rio Grande do Sul.

Initial analysis of independent variables has shown for sex ratio in the elderly population (female/male) an average of 1.1 women for each man, with municipalities showing a minimum ratio of 0.3 and others a maximum of 1.7. Half of the elderly population studied consists of unmarried individuals (mean = 0.5; SD = 0.9). The proportion of non-whites is on average 0.5 (SD = 0.2). The rate of admission for mood disorders ranged from 0.0 to 13.7 per 100,000 inhabitants, with a 0.2 mean. The percentage of elderly individuals who were co-responsible for households was, on average, 60%. As for the Firjan Index, we observed a 0.6 mean, 0.3 minimum and 0.7 maximum (Table 2). Only 34 municipalities (0.6%) have special police stations for the elderly and 68 (1.2%) have Violence Prevention Centers.

Table 3 shows the final model selected, which includes the following variables: a) noninflated part: proportion of non-whites, rate of admission for mood disorders and sex ratio; b) inflated part: proportion of non-whites and sex ratio. That same table shows that in the noninflated part the variables proportion of non-whites and sex ratio showed negative coefficients, indicating that women and people whose color skin is black, brown, yellow/indigenous commit suicide less often than white individuals and men. The rate of admissions for mood disorders showed a positive coefficient, and this factor is positively associated with the occurrence of suicide. As for the part related to excess zeros, we verified that the variable proportion of non-whites had a positive coefficient and the variable sex ratio had a negative one.

Discussion

With this study we aimed to identify factors associated with the occurrence of suicide in elderly individuals in Brazilian municipalities. The modeling process has shown that the ZINB model better allowed for excess zeros in the series and because of this it was chosen as the phenomenon's best explanatory model.

The negative coefficient found for the variable sex ratio is supported by literature findings which point to a higher occurrence among men^24-26. The adjusted model also resulted in a negative coefficient for the variable proportion of non-whites, thus indicating that there is greater occurrence of suicide among white elderly individuals in Brazilian municipalities. There are examples in other studies that single out white or Caucasian populations as the ones with the highest occurrence of such events^2,13,26,27.

The issue of mood disorders has also been considered a relevant risk factor for elderly suicides, as widely documented in literature, especially depression^{2,3,14,25,28,29}. Lack of openly and community-based mental health services which provide care to elderly individuals at risk of suicide has been noticed in many Brazilian municipalities, despite the National Mental Health Policy³⁰. Such poor service may be related to the occurrence of suicides in this social group¹⁰.

The variable IFDM, which shows a municipality's level of progress, has not proven to be associated with the outcome of suicide cases. Although it is not limited to the issue of economic development, the latter is included in items used to calculate this index. Literature and empirical study findings show that economic issues (such as financial crises and lowering of socioeconomic status) have proven to be related to suicide in elderly individuals^4,6.

The variable percentage of elderly individuals who share responsibility for a household was not associated with the occurrence of suicide. This datum deserves to be analyzed carefully, since municipal data under analysis do not allow us to make inferences about family composition in order to verify the association between living alone and suicide, an aspect mentioned in several studies¹⁴. If a variable had been obtained which included only elderly individuals exclusively responsible for a household, results might have been different.

Studies point to an association between being a widow(er) as a risk factor for suicide^6,25. In this study we did not find any association between the variable marital status (proportion of unmarried individuals) and suicide. This may have occurred because the variable chosen to represent marital status (proportion of unmarried individuals) included the percentages for single and separated individuals and widows(ers), making it difficult to identify an association. A descriptive study conducted by Meneghel et al⁶ using the population of Rio Grande do Sul showed higher coefficients of mortality by suicide among widows(ers).

A few factors have limited the reach of this study, above all the lack of municipal data about several aspects: access to means¹⁵, alcohol and other substance abuse^3,25,31 and other information about self-destructive conduct⁶. The lack of municipal data about Brazil's elderly population also made it difficult to carry out the intended analysis. The modeling performed grouped different realities and incorporated all Brazilian municipalities. There are obviously factors common to all municipalities with large amounts of cases, which has been identified by the model. However, suicide is a complex phenomenon, it has multiple causes and certainly each municipality (or micro-region) has its own features. An analysis by micro-region, including municipalities with more similar realities, could provide important material to design prevention programs. Likewise, qualitative studies about psychosocial autopsies offer an invaluable contribution to local analyses.

Final Considerations

Risk factors for suicide in the elderly pointed out by this ecological study are known internationally, which facilitates prevention strategies already tested in several countries worldwide. Most of them are related to quality of life, to social support, to specific primary and secondary care and the area of mental healthcare, and the prevention of degenerative physical and mental conditions ¹¹. Strategies based on restricting access to means and permanent care being provided to the elderly who have ideations and who have attempted suicide have also shown to be effective reducing suicide rates. Several authors have been making contributions to all aspects mentioned in this paper, showing how warning signs for suicide among the elderly can be recognize and used to draft effective prevention proposals^3,11,15. All such researchers have been showing through data that it is possible to prevent the end from coming too soon.

However the more general factors addressed here need to be further detailed by qualitative studies and treated in different levels. At a local scale, situational diagnosis should point out the most common factors, the presence or lack of social support and the existence and efficiency of healthcare services. At the government level, above all in those where the phenomenon is more relevant, there needs to be coordination, municipal support and personnel training policies. At the national level, it is important that the issue of elderly suicide is treated as a relevant topic when organizing Violence Prevention Centers and in several attempts to improve notifications about violence and accidents.

However, one needs to bear in mind that ultimately all suicides have a component of individual choice, although this choice is influenced by social, psychological, environmental, medical and very painful circumstances. This is why all general factors need to be contextualized and reaching concrete cases. As Cassorla³² shows, "there is no single cause of suicide. It is an event that occurs as the climax of a series of factors that build up in an individual's history; constitutional, environmental, cultural, biological and other problems come into play. What one refers to as "cause" is usually the final link in this chain".

We finish this paper by quoting the same author³²: "I believe life should be lived here and now, with everything we can enjoy from it. And if we are not managing, something is happening. It is time for us to ask for help, even if we are not contemplating suicide. For unhappiness, not using one's potentials, not fighting for dignity may all be considered partial or micro-suicides".

Collaborators

LW Pinto and CMFP Silva participated equally in all stages of drafting this paper. TO Pires participated in data analysis. MCS Minayo, SG Assis participated in the discussion of results and drafting the paper.

References

• ¹
  American Foundation for Suicide Prevention. Facts and Figures: international statistics. [página na Internet]. [acessado 2012 mar 8]. Disponível em: http://www.afsp.org/index.cfm?fuseaction=home.viewPage&page_id=0512CA68-B182-FBB3-2E4CB905983C0AB8
• ²
  Organização Mundial de Saúde (OMS). Relatório mundial sobre violência e saúde Genebra: OMS; 2002
• 3. Mitty E, Flores S. Suicide in late life. Geriatric Nursing 2008; 29(3):160-165.
• 4. Di Mauro S, Leotta C, Giuffrida F, Distefano A, Grasso MG. Suicides and the third age. Arch Gerontol Geriatr 2003; 36(1):1-6.
• 5. O'Connell H, Chin AV, Cunningham C, Lawlor BA. Recent developments: Suicide in older people. BMJ 2004; 329(7471):895-899.
• 6. Meneghel SN, Victora CA, Faria NMX, Carvalho LA, Falk JW. Características epidemiológicas do suicídio no Rio Grande do Sul. Rev Saude Publica 2004; 38(6):804-810.
• 7. Nock MK, Borges G, Bromet EJ, Cha CB, Kessler RC, Lee S. Suicide and suicidal behavior. Epidemiol Rev 2008; 30:133-154.
• 8. Conwell Y, Thompson C. Suicidal behavior in elders. Psychiatr Clin North Am 2008; 31(2):333-356.
• 9. Oliveira CS, Lotufo FN. Suicídio entre povos indígenas: um panorama estatístico brasileiro. Rev Psiquiatr Clin. 2003; 30(1):4-10.
• 10. Shah A, Bhat R. The relationship between elderly suicide rates and mental health funding, service provision and national policy: a cross-national study. Int psychogeriatr 2008; 20(3):605-615.
• ¹¹
  World Health Organization (WHO). Multisite intervention study on suicidal behaviours - SUPRE-MISS: Protocol of SUPRE-MISS Genève: WHO; 2002.
• ¹²
  Departamento de Informática do SUS (DATASUS). Informação de saúde População residente por ano segundo município período: 2005-2007. [documento na Internet]. 2012 jul. [acessado 2012 jul 09]. Disponível em: http://tabnet.datasus.gov.br/cgi/tabcgi.exe?ibge/cnv/popbr.def
• 13. Gallagher-Thompson D, Osgood NJ. Suicide in later life. Behavior Therapy 1997; 28(1):23-41.
• ¹⁴
  Instituto Brasileiro de Geografia e Estatística (IBGE). Pesquisa de Informações Básicas Municipais - MUNIC, 2009, Perfil dos Municípios Brasileiros: Assistência Social 2009. Diretoria de Pesquisas, Coordenação de População e Indicadores Sociais, 2009.
• 15. Conwell Y, Duberstein PR, Cain ED. Risk factors for suicide in later life. Biol Psychiatry 2002; 52(3): 193-204.
• 16. Instituto Brasileiro de Geografia e Estatística (IBGE). Censo Demográfico 2010 [documento na Internet]. 2012 jul. [acessado 2012 jul 09]. Disponível em: http://www.sidra.ibge.gov.br/cd/defaultcd2010.asp?o=2&i=P
• ¹⁷
  Instituto Brasileiro de Geografia e Estatística (IBGE). Pesquisa de Informações Básicas Municipais - MUNIC, 2009, Perfil dos Municípios Brasileiros: Assistência Social 2009. Diretoria de Pesquisas, Coordenação de População e Indicadores Sociais, 2009.
• 18. Khan A, Ullah S, Nitz J. Statistical modelling of falls count data with excess zeros. Injury Prevention 2011; 17(4):266-270.
• 19. Zeileis A, Kleiber C, Jackman S. Regression Models for Count Data. In: Zeileis A, Kleiber C, Jackman S. Research Report Series/Department of Statistics and Mathematics, 53. Vienna: Department of Statistics and Mathematics, WU Vienna University of Economics and Business; 2007.
• 20. Fumes G, Corrente JE. Modelos inflacionados de zeros: aplicações na análise de um questionário de frequência alimentar. Rev Bras Biom 2010; 28(1):24-38.
• 21. Lambert D. Zero-inflated Poisson regression with an application to defects in manufacturing. Technometrics 1992; 34(1):1-14.
• 22. Vasconcello MTL, Portela MC. Índice de Massa Corporal e sua relação com variáveis nutricionais e sócio-econômicas: um exemplo de uso de regressão linear para um grupo de adultos brasileiros. Cad Saude Publica 2001; 17(6):1425-1436.
• ²³
  R 2.12.2. [computer program]. Viena; The R Foundation for Statistical Computing: 2012. [acessado 2012 jul 2]. Disponível em: http://www.r-project.org
  » link
• 24. Pinto LW, Assis SG, Pires TO, Minayo MCS. Mortalidade por suicídio em pessoas com 60 anos ou mais nos municípios brasileiros no período de 1996 a 2007. Cien Saude Colet 2012; 18(8): 1963-1972.
• 25. Johnston M, Walker M. Suicide in the Elderly Recognizing the Signs. Gen Hosp Psychiatry 1996, 18(4): 257-260.
• 26. Szanto K, Prigerson HG, Reynolds CF. Suicide in the elderly. Clinical Neurosci Res 2001; 1(5):366-376.
• 27. Kung HC, Liu X, Juon HS. Risk factors for suicide in Caucasians and in African-Americans: a matched case-control study. Soc Psychiatry Psychiatr Epidemiol 1998; 33(4):155-161.
• 28. Cole MG, Bellavance F, Mansour A. Prognosis of depression in elderly community and primary care populations: a systematic review and meta-analysis. Am J Psychiatry 1999; 156(8):1182-1189.
• 29. Ritchie CW, King MB, Nolan F, O´Connor S, Evans M, Toms N, Kitchen G, Evans S, Bielawski C, Lee D, Blanchard M. The association between personality disorder and an act of deliberate self harm in the older people. International Psychogeriatrics 2011; 23(2):299-307.
• 30. Brasil. Lei nº 10.216 de 6 de abril de 2001. Dispõe sobre a proteção e os direitos das pessoas portadoras de transtornos mentais e redireciona o modelo assistencial em saúde mental. Diário Oficial da União 2001; 9 abr.
• 31. Hawton K, Van Heeringer K. Suicide. The Lancet 2009; 373(9672):1372-1381.
• 32. Cassorla R. Do suicídio: estudos brasileiros. Campinas: Editora Papirus; 1991.

Fatores associados com a mortalidade por suicídio de idosos nos municípios brasileiros no período de 2005-2007

Factors associated with suicide mortality among the elderly in Brazilian municipalities between 2005 and 2007

Publication Dates

• Publication in this collection
  30 Jan 2013
• Date of issue
  Aug 2012

History

• Received
  02 Mar 2012
• Accepted
  2012
• Reviewed
  14 Apr 2012