Original qualitative research - Individual, programmatic and systemic indicators of the quality of mental health care using a large health administrative database: an avenue for preventing suicide mortality

Correspondence: Lise Thibodeau, McGill University, Royal Victoria Hospital – V Building (Rm. V2.11), 687 Pine Avenue West, Montréal, QC H3A 1A1; Tel: 514-934-1934, ext. 44724; Fax: 514-934-8343; Email: lise.thibodeau@rimuhc.ca

Abstract

Suicide is a major public health issue in Canada. The quality of health care services, in addition to other individual and population factors, has been shown to affect suicide rates. In publicly managed care systems, such as systems in Canada and the United Kingdom, the quality of health care is manifested at the individual, program and system levels. Suicide audits are used to assess health care services in relation to the deaths by suicide at individual level and when aggregated at the program and system levels.

Large health administrative databases comprise another data source used to inform population-based decisions at the system, program and individual levels regarding mental health services that may affect the risk of suicide. This status report paper describes a project we are conducting at the Institut national de santé publique du Québec (INSPQ) with the Quebec Integrated Chronic Disease Surveillance System (QICDSS) in collaboration with colleagues from Wales (United Kingdom) and the Norwegian Institute of Public Health.

This study describes the development of quality of care indicators at three levels and the corresponding statistical analysis strategies designed. We propose 13 quality of care indicators, including system-level and several population-level determinants, primary care treatment, specialist care, the balance between care sectors, emergency room utilization, and mental health and addiction budgets, that may be drawn from a chronic disease surveillance system.

Keywords: suicide, services, mental health care, large health administrative databases

Introduction

Suicide mortality is a major public health issue. Several studies show that this phenomenon is associated with the quality of health care services,^{Footnote 1} as well as mental disorders, particularly depression.^{Footnote 1Footnote 2} Surveillance plays an essential role in containing the burden of chronic diseases;^{Footnote 3} however, prevention strategies to reduce suicide risk remain a challenge.^{Footnote 4} To contribute to the prevention and potential reduction of suicide mortality, this status report paper establishes and delineates the development of quality of mental health care indicators using large health administrative databases. We describe how the use of these databases may be maximized in the Canadian context using the Quebec Integrated Chronic Disease Surveillance System (QICDSS) as a case study.

Conceptual framework of the untapped potential of linked health administrative databases for identifying suicide prevention avenues

From a public health perspective, there are four primary determinants of health status: i) genetics; ii) the environment; iii) lifestyle; and iv) healthcare services.^{Footnote 5} These determinants also apply to suicide risk.^{Footnote 6} The relevant information required to analyze mental health care in relation to suicide mortality is available from suicide audits^{Footnote 7} and health administrative databases. Tansella and Thornicroft^{Footnote 8} first proposed this type of conceptual framework using the latter, which included service data and patient-based information. These investigators applied the classic Donabedian's medical services quality model^{Footnote 9} to create a two-dimensional matrix: a temporal axis of the input-processes-outcomes and a geographical axis of the system-program-individual levels. In their model, suicide rates are outcomes at the system and program levels, and the suicide risk at the individual level. Budgets represent input at the system and program levels; the balance between primary and specialist services is a process indicator at the system and program levels; and the quality of follow-up after the detection of depression in primary care or after an admission for a suicide attempt is a process indicator at all levels. Other terminology has been used to refer to the program level, such as the Meso-level by Contandriopoulos, et al.,^{Footnote 10} who also drew their approach from the seminal work of Donabedian.^{Footnote 9}

Table 1 presents our adaptation of the Tansella and Thornicroft^{Footnote 8} model using the untapped potential of the large linked health administrative databases of the QICDSS to investigate quality of mental health care as a determinant of suicide.

Table 1. The three levels of the quality of mental health care services model

Level	Input	Process	Outcomes	Examples of computing and innovating methodologies	Information for disease management / information for policy making
1. Individual	Age Gender Socioeconomic status Diagnoses	Types of treatments Settings of encounters Types of programs Intensity of contacts Coordination of referrals	Individual risk of suicide at last contact with services considering all previous information available in database	Computer-learning artificial intelligence algorithms (Kessler et al.Footnote 11)	Individual clinicians in health, social and addiction service settings
2. Program	Hospitalization Emergency department Specialist outpatient General practitioner (GP)	Average number of visits Proportion of individuals readmitted within 30 days after hospital discharge Type of diagnosis Follow-up after emergency visit or hospital admission for intentional self-harm Follow-up after primary care diagnosis of depression	Increased risk of suicide in type of setting (i.e., emergency room)	Electronic case-control studies based on the Secure Anonymized Information Linkage (SAIL) (John et al.Footnote 13)	Regional/local mental health and addiction programs
3. System	Regional mental health or addiction budgets	Regional balance of specialist and primary care Quality of follow-up after hospitalized suicide attempts	Regional suicide rates	Ecological study (Tondo et al.Footnote 21)	Country/state/provincial

Each level presented in Table 1 refers to a representation of our conceptual framework of a three-level mental health care system, as shown in Figure 1. This system includes the individual, programmatic and systemic levels embedded within environmental determinants of suicide, population health status and other determinants.

Risk factors at the individual level indicated in Table 1 may be used to conduct analyses that would produce algorithms of trajectories associated with a higher risk of suicide. These algorithms, thus, become indicators of the quality of care at the individual level within the available databases. The algorithms are complex to interpret, as demonstrated with the same databases as recent QICDSS trials in cardiovascular diseases.^{Footnote 11Footnote 12} A set of largely used indicators of quality of care, such as those presented in the literature or Table 2, is considered in establishing the validity of the individual level-defined algorithms, such as audits that informed at the individual level or aggregated for the validity at the program and system levels. Established indicators of the quality of mental health care may be obtained from the literature at the programmatic and system levels, with potential candidates from the QICDSS proposed in Table 2. At the programmatic level, a case-control design between individuals who died by suicide and individuals who did not may be explored. These indicator outcomes should be worse among suicide cases in the year preceding their death than in non-cases with similar mental disorder diagnoses after controlling for other covariates.^{Footnote 13} Finally, the system-level analysis suggested in Table 1 may employ identified risk factors from the two previous levels as indicators of the quality of mental health care, considering other risk factors as co-variates, if available, in the various linked health databases, and applicable at this level for the dependent variable of regional or provincial/state suicide rates.

Table 2. List of the candidate indicators at the programmatic and system levels supported by the health services and public health literature or practices

Candidate indicators	Aim	Literature support	Description	Measure	Data sources
1. Quality of anxiety or depressive disorders mental health services follow-up in primary care	Determine adequate care for patient diagnosed with anxiety and depressive disorders in primary care	Based on number of physician visits by Wang, et al.Footnote 44 and other studiesFootnote 47Footnote 48	Denominator: Individuals aged 15+ years with an anxiety or depressive disorder diagnosis by a General Practitioner (GP) in a given year	Prevalence of individuals 15+ years who received an anxiety or depressive disorder diagnosis with ≥ 4 visits for mental health	QICDSS
			Numerator: Received ≥ 4 visits for mental health in that year
2. Quality of depression disorder mental health services follow-up in primary care	Determine adequate care for patient diagnosed with depression in primary care	Based on number of physician visits by Wang, et al.Footnote 44 and other studiesFootnote 47Footnote 48	Denominator: Individuals aged 15+ years with a diagnosis of depression by a General Practitioner (GP) in a given year	Prevalence of individuals 15+ years who received a depression diagnosis with ≥ 4 visits for mental health	QICDSS
			Numerator: Received ≥ 4 visits for mental health in that year
3. Quality of substance use disorder mental health services follow-up in primary care	Determine adequate care for patient diagnosed with substance use disorder in primary care	Based on 4 visits with a family physician for counseling as recommended by NICEFootnote 58 and the guidelines for American primary care cliniciansFootnote 58	Denominator: Individuals aged 15+ years with a diagnosis of substance use disorder by a General Practitioner (GP) in a given year	Prevalence of individuals 15+ years who received a substance use disorder diagnosis with ≥ 4 visits for mental health	QICDSS
			Numerator: Received ≥ 4 visits for mental health in that year
4. Quality of mental health care services follow-up after hospitalization: readmission within 30 days	Determine the quality of mental specialist health care and in-hospital care	Based on the work of the Canadian Institute for Health Information (CIHI)Footnote 47Footnote 48	Denominator: Individuals aged 15+ years admitted in a hospital with a mental health diagnosis in a given year	Prevalence of individuals 15+ years who were readmitted to a hospital for a mental health diagnosis within 30 days of initial discharge	QICDSS
			Numerator: Individual readmitted for mental health within 30 days of initial discharge
5. Quality of mental health services follow-up in primary care after suicide attempt	Determine the quality of mental health care of readmission rates in the region compared to others	Based on the work of the Canadian Institute for Health Information (CIHI)Footnote 47Footnote 48	Denominator: Individuals aged 15+ years admitted to a hospital for suicide attempt in a given year	Prevalence of individuals 15+ years who received ≥ 1 visit from a physician within 30 days of initial discharge for suicide attempt	QICDSS (linked to MedEcho for suicide attempt)40,41,50
			Numerator: Received ≥ 1 visit to a physician for mental health within 30 days of hospital discharge for suicide attempt
6. Quality of community mental health services	Determine the balance of the community-oriented mental health care system	Based on the typologies of primary and specialist (including in-hospital care) mental health careTable 2 Footnote aFootnote 45Footnote 46Footnote 54 used in the study of suicide attemptsFootnote 55	Denominator: Individuals aged 15+ years with a mental health diagnosis in a given year	Prevalence of individuals 15+ years who received a mental health disorder diagnosis with exclusively outpatient services (psychiatric or GP)	QICDSS
			Numerator: Individuals with exclusively outpatient services – psychiatric or general practitioner (GP)
7. Quality of community mental health services of patients with severe mental illness	Determine the balance of psychiatric outpatient and primary outpatient care depending on the profiles usedTable 2 Footnote aFootnote 55	Based on the associations found for the balance between primary and specialist mental health care and suicide ratesFootnote 2Footnote 46Footnote 60Footnote 61	Denominator: Individuals aged 15+ years with exclusively a GP or a psychiatric outpatient visit for psychotic disorder	Prevalence of individuals 15+ years who received a severe mental illness disorder diagnosis and used exclusively outpatient services by a GP	QICDSS
			Numerator: Number of individuals with exclusively a GP or psychiatrist outpatient visits
8. Quality of community mental health services of patients with common mental disorders	Determine the balance of psychiatric outpatient and primary outpatient care depending on the profiles usedTable 2 Footnote aFootnote 55	Based on the associations found for the balance between primary and specialist mental health care and suicide ratesFootnote 2Footnote 46Footnote 60Footnote 61	Denominator: Individuals aged 15+ years with a psychiatric or a GP outpatient visit for depression	Prevalence of individuals 15+ years who received a common mental disorder diagnosis and used exclusively outpatient services by a GP	QICDSS
			Numerator: Number of individuals with exclusively GP outpatient visits
9. Quality of community mental health services of patients with substance use disorders	Determine the balance of psychiatric outpatient and primary outpatient care depending on the profiles usedTable 2 Footnote aFootnote 55	Based on the associations found for the balance between primary and specialist mental health care and suicide ratesFootnote 2Footnote 46Footnote 60Footnote 61	Denominator: Individuals aged 15+ years with a psychiatric or a GP outpatient visit for substance use disorder	Prevalence of individuals 15+ years who received a substance use disorder diagnosis and used exclusively outpatient services by a GP	QICDSS
			Numerator: Number of individuals with exclusively GP outpatient visits
10. Quality of community mental health services of patients with personality disorders	Determine the balance of psychiatric outpatient and primary outpatient care depending on the profiles usedTable 2 Footnote aFootnote 55	Based on the associations found for the balance between primary and specialist mental health care and suicide ratesFootnote 2Footnote 46Footnote 60Footnote 61	Denominator: Individuals aged 15+ years with exclusively a GP or a psychiatric outpatient visit for personality disorder	Prevalence of individuals 15+ years who received a personality disorder diagnosis and used exclusively outpatient services by a GP	QICDSS
			Numerator: Number of individuals with exclusively a GP or psychiatric outpatient visits
11. Adequate use of emergency room for mental health services	Determine the balance of utilization of emergency room (ER) for mental health reasonsTable 2 Footnote aFootnote 55	Based on the associations found for the balance between primary and specialist mental health care and suicide ratesFootnote 46Footnote 60Footnote 61	Denominator: Individuals aged 15+ years with a diagnosis of a mental health disorder	Prevalence of individuals 15+ years who received a diagnosis of mental health disorder with exclusively ER visits without being admitted	QICDSS
			Numerator: Number of individuals with ER visits without being admitted
12. Program expenditures for mental health services	Determine the strength of the relationship between changes in suicide rates and expenditures for mental health (regional and provincial)	Based on associations found between mental health budget and suicide ratesFootnote 21Footnote 23	Refer to the Gouvernement du QuébecFootnote 43	Dollars per capita spent on mental health programs (provincial and regional)	Annual financial reports from the Ministère de la santé et des services sociaux (MSSS)Footnote 43
13. Program expenditures for addiction services	Determine the strength of the relationship between changes in suicide rates and expenditures for addiction services (regional and provincial)	Based on associations found between mental health budget and suicide ratesFootnote 21Footnote 23	Refer to the Gouvernement du QuébecFootnote 43	Dollars per capita spent on health programs for addiction services (provincial and regional)	Annual financial reports from the MSSSFootnote 43
Abbreviations: CIHI, Canadian Institute for Health Information; ER, emergency room; GP, general practitioner; MSSS, Ministère de la santé et des services sociaux; QICDSS, Quebec Integrated Chronic Disease Surveillance System. Table 2 Footnote a Profile 1: psychiatric inpatient care; profile 2: hospital emergency room (ER); profile 3: psychiatric outpatient care; profile 4: general practitioner (GP) clinics; and profile 5: other medical specialist. Table 2 Return to footnote a referrer

Furthermore, suicide is believed to result from the interaction of different factors, including genetics^{Footnote 14} and lifestyle/social factors,^{Footnote 15Footnote 16} as illustrated in the other risk factors in Figure 1; several mental illnesses have genetic influences.^{Footnote 17} However, there is no possibility, at this time, to measure genetic risk factors with linked health administrative databases. To accomplish this task, in the future, a reliable test may be discovered and recorded in the electronic medical records available for research.^{Footnote 18Footnote 19}

A brief history of the major initiatives

Previous courses of action have provided important information at each level identified in Figure 1. For example, aggregated individual suicide audits have demonstrated that acting on services may help prevent suicide, as reported by regions following the recommendations outlined in the National Confidential Inquiry into Suicide and Homicide by People with Mental Illness in England and Wales.^{Footnote 20}

Large health administrative databases may provide information relatively promptly for decisions at the system and program levels, whereas the value at the individual level remains experimental. There are reports regarding the examples identified in Table 1. For example, Kessler, et al.^{Footnote 11} addressed individual-level factors and developed an actuarial risk algorithm to predict suicide in the 12 months following inpatient treatment of US Army soldiers for psychiatric disorders using data from the Historical Administrative Data System (HADS) of the Army Study to Assess Risk and Resilience in Service members (Army STARRS). Furthermore, this study relied on machine learning. The findings suggested that the strongest predictors included sociodemographic factors, criminal offenses, prior suicidality, aspects of prior psychiatric inpatient and outpatient treatment, and disorders diagnosed during the focal hospitalization. Similar experimental endeavors have been undertaken with Quebec's linked health administrative databases by Najjar^{Footnote 12} and a project underway in Wales, UK by co-authors John, et al.

At the program level, John, et al.^{Footnote 13} developed a general population-based study of suicide mortality and mental health using routinely collected administrative databases and proposed a case-control design. The services provided were shown to determine the quality of care and define the program determinants of suicide mortality (i.e., general hospital admission; emergency department contact for self-harm and other indications including psychiatric admission and primary care contact in the year prior to probable suicide).

Finally, a system-level ecological study of suicide rates and assessment of mental health services was conducted by Tondo, et al.^{Footnote 21} in the 50 US states. Their study indicated that a higher population density of psychiatrists and physicians and increased levels of federal aid for mental health were associated with lower suicide rates.^{Footnote 22} Leff, et al.^{Footnote 23} demonstrated that efficient mental health services predicted suicide reduction in a sample of publicly managed US facilities. The conceptualization and findings of these studies further substantiated the richness of the information available in administrative health databases used internationally. As suggested by While, et al.,^{Footnote 20} the provision of mental health services may affect suicide rates in clinical populations, and investigations of these services in relation to suicide may help inform future suicide prevention efforts and improve safety for patients receiving mental health care.

Among the environmental and social determinants of suicide, in international and Canadian studies, the unemployment rate has been associated with suicide mortality.^{Footnote 24Footnote 25Footnote 26Footnote 27Footnote 28} Seminal work by Durkheim, the founder of modern sociology, at the end of the 19^th century demonstrated that suicide rates are higher in socially and materially deprived areas, where the social capital for support and opportunities is lower.^{Footnote 6}

Large health administrative databases are available in Canada; however, there are few national and provincial studies that utilize them. Research in Alberta^{Footnote 29} using a health services administrative database indicated that approximately 90% of individuals who die by suicide utilized a health service in the year prior to their death, and the majority visited a general practitioner (GP). Moreover, approximately 60% of these individuals had an emergency room (ER) visit, whereas only 39% of their peers had ER visit in the UK and Wales, as reported by Gairin, et al.^{Footnote 30} At the national level, the Canadian Medical Association Journal (CMAJ) recently supported the value of large linked health administrative databases for suicide studies. However, it acknowledged the methodological and analytical challenges noted by Patrick,^{Footnote 31} Quan and Williamson^{Footnote 32} and the international research groups of Benchimol, et al.^{Footnote 33} and Nicholls, et al.^{Footnote 34} The main challenges raised were missing data and, for non-random missing data, variables that created incomplete or inadequate reporting of research based on routinely collected data.^{Footnote 33}

Delimitation of a conceptual framework in the Canadian context: health care services and suicide risk using big data

The creation and growth of national surveillance systems in Canada and their impact on chronic disease and injury prevention have been reviewed,^{Footnote 35} with a primary focus on chronic physical conditions. Our conceptual framework considers health care services, particularly mental illness care to analyze suicide mortality and the possibility of obtaining a better understanding and prevention. Based on this conceptual framework, we will independently test each quality of care indicator (13 indicators). Our general hypothesis is that programs and systems that follow literature guidelines in terms of the quality of mental health care are less likely to present higher suicide rates than programs that do not follow guidelines (Table 2).

In Canada, one example of using mental health care service records from large linked health administrative databases is the QICDSS, from the province of Quebec. Our study was produced in collaboration with colleagues from Wales, UK^{Footnote 13} and the Norwegian Institute of Public Health, and it may accelerate and validate gaps in the publicly managed care system regarding suicide.

QICDSS

The QICDSS was created by the Institut national de santé publique du Québec (INSPQ), which is a public health expertise and reference center that extracts data from five linked health administrative databases. The QICDSS includes all data from all individuals who suffer from one or more chronic diseases, including mental disorders. All mental disorder ICD codes are included, except for dementia. All cases of suicide reported by the coroner in the province of Quebec, whether individuals received a mental disorder diagnosis in the previous year or years, will be investigated.

The QICDSS includes, for all cases, the utilization of all services (hospitalization; emergency room; outpatient specialist; and general practitioners). The QICDSS has been updated annually since 1996.^{Footnote 3} Blais, et al.^{Footnote 3} assessed the essential features and strengths of the QICDSS and determined that it meets all basic requirements of a public health surveillance system.

Underreporting of mental health problems is often called into question; thus, databases and case definitions require validation before being used for epidemiological purposes^{Footnote 36Footnote 37Footnote 38Footnote 39}. Therefore, regarding the data quality of the QICDSS, it has been determined that psychiatrists and pediatricians entered the ICD-9 code in billing files in 95% of cases, and an internal medicine diagnosis code was indicated in 94% of claims.^{Footnote 37Footnote 40} The literature indicates that diagnoses from recorded medical records fully correspond to the entries in administrative databases for other diseases in Quebec, as well as for other provinces.^{Footnote 37Footnote 40} In addition, Dodds, et al.^{Footnote 41} reported that administrative health databases can clearly identify children with autism. In this context, the official data are considered largely reliable. To complement the data in the QICDSS, the determinants listed in Table 3 are provided by a provincial-level statistics organization^{Footnote 42} and a government ministry.^{Footnote 43}

Table 3. List of environmental determinants

Determinants	Aims	Literature support	Measure	Data sources
System determinant of suicide, unemployment rates	Determine the impact of annual regional and provincial unemployment rates	Based on international and Canadian studies that reported a relationship between suicide mortality and unemployment rateFootnote 24Footnote 25Footnote 26Footnote 27Footnote 28	Unemployment rates (provincial and regional)	Institut de la statistique du Québec (ISQ)Footnote 42
Individual socioeconomic determinant of suicide	Determine the impact of social deprivation on suicide	Based on the Pampalon index of material and social deprivation of the census that tracts areas of residence for each individual patient in the databaseFootnote 49Footnote 59	Index constructed in two stages by INSPQFootnote 3Footnote 49	File linking between Canadian Census and INSPQFootnote 3
Abbreviations: INSPQ, Institut national de santé publique du Québec; ISQ, Institut de la statistique du Québec.

Candidate indicators

Table 2 summarizes each of the 13 quality of mental health care candidate indicators produced using the QICDSS and other data sources to specifically cover the determinants presented in our conceptual framework (Figure 1) and at the programmatic and system levels (Table 1).

Our 13 quality of care indicators are based on a literature review of health services research and practices from epidemiological studies. The indicators of receiving four or more visits for mental health in one year^{Footnote 44} following a primary care physician first diagnosis of depression or substance abuse are drawn from US epidemiological studies that defined the quality of depression care or from guidelines for US family physicians for substance-abuse care at the primary care level. The 30-day readmission rate indicator of the quality of specialist care is based on the Canadian Institute of Health Information (CIHI),^{Footnote 45} whereas the balance of specialist and primary care is based on countries with similar socioeconomic profiles and a publicly funded managed care system as Canada.^{Footnote 46}

Interpretation of the indicators proposed in Table 2 all point to a positive relationship with decreased suicide rates, excluding indicator 4, which is the 30-day readmission rate indicator from the CIHI original definition.^{Footnote 45} For example, more expenditures on mental health and addiction services per capita would represent an indicator of the capacity to provide effective and timely services, whereas the two environmental determinants of a lower unemployment rate and a better social and material deprivation index would be expected to be associated with lower programmatic- and regional/provincial/state-level suicide rates and variations. Other variables may be selected,^{Footnote *} since the variables retained are the primary control variables included in analyses in the literature. Overall, by testing the 13 candidate indicators previously described, our aim is to determine the strength of the relationship between the changes in suicide rates and the program or system-level indicators of increased quality of mental health care at the provincial level and in each regional health territory, with consideration of the timeframe allowed for observation by the available linked health administrative databases.

Our work includes patient-centered information to capture the quality of primary care; however, some data are not recorded, such as family history and child abuse, specific treatments, and non-profit organization activities. In particular, the application of specific evidence-based treatments, such as anti-depressant medication and psychotherapy, are not currently available in the QICDSS; however, this information may be available in the future in Quebec or other jurisdictions. Nevertheless, as previously indicated, linking new health databases, such as medication or electronic medical records, may increase the potential determinants considered; however, these databases suffer from missing data or incomplete coverage of the population of interest.^{Footnote 47} Moreover, primary care improvements in the late 1990s may have reduced suicide rates, and thus, it is difficult to estimate their effect. However, this remains important to note.^{Footnote 48}

Next steps: empirical testing of our theoretical framework

The next phase of the project focusses on conducting an empirical analysis to test our theoretical framework. Several statistical models will be tested to explore the relationship between changes in suicide rates and each individual, program and system level indicators of the 13 candidate indicators. The analyses will be performed at the provincial and regional levels. Two models will be considered: 1) a proportional hazard regression model (Cox regression) to investigate the associations between each of the 13 indicators and suicide; and 2) an ecological analysis to assess the association between the two environmental determinants, presented in Table 3, and suicide rates. There will be one model for each indicator – separately. The dependent variables for each Cox regression model will be suicide outcomes, and the exposure variables will be the 13 indicators. Because the indicators may change in value over the course of observation, the Cox regression is an appropriate model to account for time-dependent variables.

The individual-level candidate indicators, covariates such as gender, age, comorbidity, and deprivation index,^{Footnote 49} are available and will be integrated into the Cox regression models. Comorbidity will be adjusted at the individual level. Comorbidity refers to the physical conditions described in the Elixhauser comorbidity index.^{Footnote 50} However, mental conditions, such as mood disorders, depression, substance abuse, personality disorders, psychoses and anxiety, will not be included because they are included in the proposed indicators. Furthermore, the Elixhauser index covers more diseases than the QICDSS. The latter covers chronic conditions that have a validated case definition using health administrative data. Some diseases, such as cancer and liver diseases, that are part of the Elixhauser Index, do not have a validated case definition using linked health administrative databases.

For candidate indicators at the regional level, mental health expenditures and the unemployment rate will also be included in the models. The ecological models will be linear models adjusted at the regional level for the unemployment rate and at the dissemination area-level for the socio-economic deprivation index.

Ethics in the Canadian context: respecting access to documents held by public bodies and the protection of personal information

Ethical and legal issues in suicide research and the legal status of suicide have been reviewed in the literature,^{Footnote 51Footnote 52} which highlights concerns regarding the context of common ethical perspectives, the acceptability of suicidal behaviors, and the obligations and limitations in intervening to prevent suicides. Specifically, in this study on suicide and mental health services, legitimate issues of confidentiality arise from the use of administrative databases.

The preliminary work of recording the various chronic diseases in a large database was completed by the INSPQ following authorization from the Commission d'accès à l'information du Québec(CAI). The INSPQ operates with strict access procedures and within secured zones. Only authorized programmers can directly access and extract data from the anonymized QICDSS databases. Other analysts and experts can access the unit for specific projects, and only aggregate results may be reported.^{Footnote 3} Greater access and collaboration with academic institutes may help accelerate the use of and resolve the difficulty in computing the information. The QICDSS sends aggregate data to the Canadian Chronic Disease Surveillance System (CCDSS) of the Public Health Agency of Canada (PHAC), which has produced two interactive reports on overall mental disorders and anxio-depressive disorders by province.^{Footnote 53Footnote 54}

National chronic disease surveillance: actions for suicide and issues for public health

Suicide is a complex phenomenon. The use of a large public health administrative database may provide further opportunities to identify gaps in care and promotion that may be integrated to inform decision-makers as they develop population-based programs. The quality of care indicators defined in this paper support the national, regional and local activities of health advocates throughout Quebec using a web public health portal (InfoCentre). A substantial advantage of our approach is that it may be reproduced in other provinces and may be examined at the national level in Canada because the indicators may be readily obtained from provincial/territorial health administrative databases. All cases of an individual citizen dying by suicide are recorded by a coroner or the medical examiner system. This is the responsibility of each individual Canadian province and territory, and there is no overarching federal authority.^{Footnote 55} Other countries with similar health care systems and access to large health administrative databases, such as the United Kingdom,^{Footnote 13} have begun to examine the associations of mental health care services with suicide and how the innovative use of these data sources may improve prevention strategies. Moreover, a national registry database in Norway has been incorporated by the Norwegian Public Health Institute as part of the effort to monitor mental health and suicidal behaviour.^{Footnote 56}

Therefore, health administrative databases represent a substantive complement to suicide audits and other data sources by identifying opportunities for intervention services that may improve the prevention of suicide. With approximately 4000 deaths by suicide^{Footnote 57} each year in Canada, our study will likely provide useful information for PHAC, Quebec (INSPQ) and other provinces.

Conflicts of interest

The authors have no conflicts of interest to disclose.

Authors' contributions and statement

AL, ER, and LT designed and conceptualized the work. LT, AL, ER, JL, EP, LR, AJ, AR, and KL, contributed to the analytical plan, and LR, LT, ER, and JL assessed the suggested statistical modeling for empirical testing. LT drafted the manuscript, and LT, ER, AL, JL EP, LR, AJ, AR, and KL edited and critically reviewed the manuscript. All authors approved the final manuscript for submission.

The content and views expressed in this article are those of the authors and do not necessarily reflect those of the Government of Canada.