The MATCH cohort study in the Netherlands: rationale, objectives, methods and baseline characteristics of patients with (long‐term) common mental disorders

Abstract

Research in the last decades shows that common mental disorders may be long‐term and severely disabling, resulting in severe mental illness (SMI). The percentage of Dutch SMI‐patients with common mental disorders receiving mental health services is estimated at 65–70%. However, it is unclear which patients in fact become SMI‐patients. We need to know more about the possible course of common mental disorders, understand the origins of chronicity in more detail, and have more insight in related care processes and care use of patients with common mental disorders.

The MATCH cohort study is a four‐year multicentre naturalistic cohort study, with yearly assessments in primary, secondary, and tertiary services in three large Dutch mental health services. Socio‐demographics, mental disorders, course and severity of psychopathology, physiological health indicators, neurocognitive functioning, past and present life events, health care use and contact with mental health services, social functioning and quality of life, and recovery and well‐being are assessed.

Baseline findings of 283 participating individuals and their key clinicians are described. The sample appears to appropriately represent the distribution of individuals across diagnostic categories in services, and level of care (outpatient, day treatment, inpatient) in the Netherlands and other developed nations. Copyright © 2016 John Wiley & Sons, Ltd.

Introduction

Mental disorders are very common in developed societies, roughly affecting one in five to one in two people at some point during the lifecycle (Kessler et al., 2007). In the Netherlands, life time prevalence is highest in mood (20.1%), anxiety (19.6%), and substance use disorders (19.1%) (de Graaf et al., 2010). In comparison to high‐impact but low‐prevalence mental disorders, such as schizophrenia (0.4%; Saha et al., 2005) and bipolar disorder (0.8–1.1%; Merikangas et al., 2007) these “common” disorders are often considered as less severe and short‐term or transient. Research in the last decades has however shown that common mental disorders may very well be long‐term and severely disabling (e.g. Spijker et al., 2002; Moffitt et al., 2007; Beesdo et al., 2007; Judd et al., 2000; Solomon et al., 2000).

When indeed long‐term and disabling, these mental disorders may meet the criteria for severe mental illness (SMI): (1) presence of a mental disorder according to DSM‐IV criteria; (2) presence of symptoms and complaints of this disorder (often varying in severity and intensity over time); (3) long‐term course (> two years); and (4) resulting in disabilities in social functioning [Global Assessment of Functioning (GAF)‐score < 50] (e.g. Ruggeri et al., 2000; Parabiaghi et al., 2006). When including all mental disorders, the population prevalence of SMI is estimated at 1.34–3.01% in developed countries (Kessler et al., 2001; Lora et al., 2007; Ruggeri et al., 2000), and at 1.9% in the Netherlands (Delespaul et al., 2013). A substantial part of these patients do not receive mental health care, estimated at 35–50% in developed countries with large cross‐national variation (Demyttenaere et al., 2004). The percentage of Dutch SMI‐patients receiving mental health services is estimated at 50–70%, of whom patients with common mental disorders account for 65–70% (Delespaul, 2013; Dieperink, 2006). Thus, a substantial number of people with common mental disorders become long‐term care users, making up almost two‐thirds of the SMI‐population in mental health services.

While diagnostically heterogeneous, patients with SMI often suffer from similar social problems regardless of diagnosis: interpersonal difficulties, financial hardship, lack of employment, poor housing, societal stigma, etc. While some receive no care at all, others may use care for very long periods, and become dependent on individual professional caregivers, or services (Lindamer et al., 2012; Lucas et al., 2001). Also they may use care in high frequencies, including an array of intensive services like crisis intervention or inpatient care (Chaput and Lebel, 2007; Ledoux and Minner, 2006; Pasic et al., 2005). Last, some patients with SMI may become involved in complex care situations in which they feel stuck, and in which professionals perceive them as “difficult” (Bachrach et al., 1987; Fok et al., 2014; Koekkoek et al., 2011b). However, it is unclear which patients with common mental disorders become long‐term SMI‐patients. We need to know more about the possible course of these mental disorders, understand the origins of chronicity in more detail, and have more insight in related care processes and care use of the patients with common mental disorders, and costs.

Therefore, this study focuses on people with common mental disorders in different settings of mental health services, in different illness stages. Currently, a number of excellent cohort studies on the course of mood and anxiety disorders, are carried out (e.g. Penninx et al., 2008). These studies, however, pay less attention to the relationship between the development of long‐term disorders and the care process in mental health services. For a more accurate matching of limited treatment resources to the projected needs of patients, more detailed knowledge is essential.

This study takes places in the Netherlands, a nation with 16.7 million inhabitants, universal health care coverage, and an average spending of 13 to 14% of the gross national product on health services in (United States: 16.2%, Germany: 10.9%, United Kingdom: 8.9%), of which 6–7% is spent on mental health services (Veerbeek et al., 2015). Specialized services are paid for by (obligatory) health insurance, while access is regulated by the primary care physician. Long‐term services for mental illness are increasingly paid from other sources, governed by local authorities.

We are currently conducting the MATCH cohort study, a multicentre study designed to examine the (determinants and consequences of) long‐term course of people with common mental disorders and mental health services offered to them. Both perspectives, of service users and service providers, are included. In this paper the basic rationales, objectives, and methods of the MATCH cohort study are presented.

Objectives

Primary objective: insight into the determinants of long‐term service use of people with common mental disorders in general, and the emergence of complex care situations within services in particular.

While it has been established that common mental disorders may be long‐term, and result in chronic suffering and decreased functioning, little is known about the role that treatment and services play in this limited recovery of people. Not only may transitions across providers be potentially influential in the recovery process, the provision of substandard care or treatment may also be of relevance (Bauer, 2002; Drake et al., 2001; Francke et al., 2008). Bearing in mind that resources are scarce, service providers tend to offer treatment that matches people's needs, yet this may not always succeed, resulting in both undertreatment (e.g. Wang et al., 2007) and overtreatment (Druss et al., 2007). Either of those may result in negligence by professionals, or unwanted chronicity and dependency on services in patients. While there may be many variables at the patient level, other types of variables should not be overlooked and therefore are assessed in this study.

Part of this objective is insight in the different ways people move across services. It is known that people with mental illness may use different types of services over time, across multiple providers, at various levels of care. Since most services follow up only their own care, continuity of data is rare, and insight into when, how, and why transitions between services are made is lacking (Crawford et al., 2004). The influence of such transitions on the course of people's illness and following use of services is largely unknown since continuity of care (or lack thereof) is often measured imprecisely (Puntis et al., 2015). Therefore, we focus on differences between types of services in both care offered, and patients cared for.

A particular harmful type of contact between patients and professionals may arise when patients are difficult to engage in treatment, and professionals perceive difficulties in engaging patients. Complicated care situations may arise over time, and result in patients being labelled as “bad” or “difficult” – often because they do not improve as easily or rapidly as hoped for by professionals, in combination with behaviours that frustrate professionals (Koekkoek et al., 2011a). These situations are neither rare nor harmless: in previous studies 6–37% of patients were labelled as difficult (Hahn et al., 1996; Jackson and Kroenke, 1999; Lin et al., 1991), with a near‐peak of 28% in a study of people with long‐term common mental disorders (Koekkoek et al., 2011b). Therefore, it is important to explore associations between clinical, treatment and social variables and the emergence of chronicity in general, treatment drop‐out, and the emergence of complex care situations in particular.

The main research question for this objective is: which variables predict long‐term service use or complex care situations?

Secondary objective: insight in the long‐term course of common mental disorders, specifically focusing on the association of social variables with outcomes

While a number of studies have shown insight into symptomatology, as well as possible biological or genetic origins of common mental disorders, less is known about course and outcomes beyond symptoms – for instance social needs, interpersonal skills, help‐seeking style and others. It is now well established that in psychotic disorders, long‐term course and outcomes may be more favourable than previously thought or expected (see Slade, 2009, for an overview). The long‐term outcomes of common mental disorders are less well researched.

Although it is likely that course and outcomes differ across distinct disorders (e.g. depression, anxiety disorder, substance use disorder) responses to long‐term (mental) illness have been found to be quite universal. In fact, it may be that people with different disorders but similar broad patterns of social circumstances, feelings, thoughts, and behaviours experience comparable outcomes. Likewise, people with different patterns but equal diagnoses may experience different outcomes. In other words: in long‐term common mental disorders, symptoms or psychopathology in itself may not be the defining variable of outcomes per se. This study assesses a number of variables potentially related to such outcomes to gain insight into people that suffer from common mental disorders in different stages of both their illness and their duration or intensity of contact with mental health services. In terms of the National Institute of Mental Health's (NIMH's) Research Domain Criteria (Cuthbert and Insel, 2013), this study particularly focuses on environmental variables (https://www.nimh.nih.gov/research-priorities/rdoc/research-domain-criteria-matrix.shtml).

The main research question for this objective is: which socio‐demographic, social, or other characteristics patients with common mental disorders have in common across diagnostic categories, and how do these variables affect course and outcome of their illness?

Methods

Study design

This study is a four‐year multicentre naturalistic cohort study on the course and outcome of 283 individuals (and their key clinicians) with common mental disorders, with yearly assessments in primary, secondary, and tertiary services in three large mental health services in the central and eastern part of the Netherlands.

Setting

Of the three participating Dutch mental health services, one exclusively offered primary services, one offered both primary, secondary, and tertiary services. The last offered both secondary, and tertiary services, but no primary services. Since the meaning and content of primary, secondary, and tertiary services may vary across health systems a brief description of each setting is given in Table 1.

Table 1.

Description of services provided by three types of settings

	Primary mental health services	Secondary mental health services	Tertiary mental health services
Dutch term	Basis GGZ	Specialistische GGZ	Hoogspecialistische GGZ
Institution 1	X
Institution 2	X	X	X
Institution 3		X	X
Content	• outpatient	• outpatient and inpatient	• high intensive or long‐term outpatient (daily contact/24 hour‐care)
	• low‐intensive or short (one contact/week or less)	• medium‐intensive (up to three contacts/week)
	• short‐duration (limited, < 12 months, < 12 contacts)	• medium‐duration (unlimited, in practice mostly up to 36 xmonths, or 100 contacts)	• specialized inpatient services
			• specialized (partial) hospital including specialized biological treatment options
	• psychosocial support
	• manualized, brief psychotherapy	• psychosocial support
		• psychotherapy
		• medication
		• acute services
		• inpatient and partial hospital
Context	• small units (<10 staff)	• medium‐size units (10–100 staff)	• large units (up to 100 staff)
	• urban and suburban locations		• mostly suburban and rural locations
		• mostly urban and some suburban locations

Sample

All patients in one of the three services with a primary common mental disorder, were eligible for participation. Inclusion was stratified across the three different settings (primary, secondary, tertiary).

Inclusion criteria

• common mental disorder according to DSM‐IV as main/primary diagnosis

• age between 18–65

Exclusion criteria

• psychotic disorder, bipolar I disorder or cognitive disorder as main/primary diagnosis

• unable to read and understand Dutch

Sample size, attrition and power

The primary aim of our study is to identify variables (measured at baseline) that predict long‐term service use in common mental disorders, an outcome that – in the absence of generally established guidelines – may be measured categorically (yes/no chronicity) or numerically (number of months in treatment). We used this objective to power the study, conservatively assuming a categorical level of measurement, aiming to detect relative risks (RRs) of at least 2.0. A RR of this size would occur, for example, when we find a predictor of chronicity associated with a risk of 0.60 of becoming chronic in the exposed group versus 0.30 in the non‐exposed group (exposure status may differ per variable analysed). Power analyses with conventional levels of alpha (0.05) and beta (0.20), thus a power of 0.80, showed that the sample size needed to detect such a predictor is n = 55 in each group. We need to take into account loss‐to‐follow‐up and have estimated this at 10% over each one‐year follow‐up period, based on other cohort studies. This means that at four‐year follow‐up we would expect to have retained at least 0.9⁴ = 65.6% of the original sample. Therefore, we needed to start our study with a sample size of at least (3 × 55)/0.656 = 251 at baseline.

Measures used at baseline

A number of variables is included and measured on both the level of the patient, the key clinician, and the treatment setting (for a time‐line of these parameters see Table 2).

Table 2.

Variables assessed and instruments used in MATCH cohort study

Area	Assessmentsa	Measurement	Source	Timelineb
				BL	FU1	FU2	FU3	FU4
1) Socio‐demographics		Interview	Patient	X
2) Diagnosis of mental disorders	• MINI Plus	Interview	Patient	X
	• SAPAS ‐R	Interview	Patient	X
	• SIDP‐IV	Interview	Patient	X
3) Assessment of course and severity of psychopathology	• HoNOS	12‐item quest	Professional	X	X	X	X	X
	• OQ‐45	45‐item quest	Patient	X	X	X	X	X
4) Physiological health indicators	• Somatic illness list	24‐item quest	Patient	X		X		X
5) Neurocognitive functioning	• EST (Emotional Stroop Task),	Researcher‐directed tasks	Patient	X
	• SRET (Self Referent Encoding Task),		Patient	X
	• TMT (Trail making test)		Patient	X
	• NART (National Adult Reading Task)		Patient	X
6) Past and present life events	• NEMESIS‐childhood trauma questionnaire	10‐item quest	Patient	X
	• Brugha‐questionnaire	10‐item quest	Patient	X		X
7) Health care use and contact with mental health services	• CCCL (Care Content Check List):	Interview	Patient	X	X	X	X	X
	• Tic‐P (Trimbos/iMTA questionnaire for costs associated with psychiatric illness	43‐item quest	Patient	X	X	X	X	X
	• STAR (Scale To Assess the Therapeutic Relationship):	12‐item quest	Patient and professional	X	X	X	X	X
	• DDPRQ (Difficult Doctor Patient Relationship Questionnaire):	10‐item quest	Professional	X	X	X	X	X
	• PPD (Perceived Patient Difficulty):	1‐item quest	Professional	X	X	X	X	X
	• CANSAS Camberwell Assessment of Need Short Appraisal Schedule	Interview	Patient and professional	X	X	X	X	X
	• CSQ‐8 (Client Satisfaction Questionnaire	8‐item quest	Patient	X		X		X
8) Social functioning and quality of life	• EQ‐5D (EuroQol 5D):	14‐item quest	Patient	X		X	X	X
	• MANSA (Manchester Short Assessment of Quality of Life	16‐item quest Interview	Patient	X	X		X	X
	• SNM (Social Network Map		Patient	X		X	X	X
9) Recovery and well‐being	• WHODAS 2.0 (World Health Organization Disability Assessment Schedule II;	12‐item quest	Patient	X		X		X
	• MHC‐SF (Mental Health Continuum Short‐Form):	14‐item quest	Patient	X	X		X
	• AAQ‐II (Acceptance and Action Questionnaire‐II):	10‐item quest	Patient	X		X		X
	• GAF‐score (Global Assessment of Functioning)	Check list	Professional	X	X	X	X	X

^aReferences to assessments given in text.

^bBL, baseline; FU, follow‐up.

Baseline assessment includes the following.

Socio‐demographic variables

A socio‐demographic list was used to assess marital status, household composition, education, work situation and a number of other relevant socio‐demographic variables.

Diagnosis of mental disorders (DSM – Axis I and II)

To ensure maximal reliability and validity of psychiatric diagnosis, structured interviews were administered. First, the MINI Plus (Mini International Neuropsychiatric Interview Plus) was used for diagnosis. This is a structured interview to assess mental disorders and comorbidity on DSM‐IV Axis I (Sheehan et al., 1998) with good reliability and validity. The MINI Plus is the briefest full psychiatric interview available and takes, dependent on the number of disorders, between 15 and 45 minutes. Older, more intensive interviews [e.g. Structured Clinical Interview for DSM‐IV Axis I Disorders (SCID‐I) or Composite International Diagnostic Interview (CIDI)] take between 90 and 120 minutes.

Administration of the full interview for Axis II personality disorders was preceded by a brief 10‐item self‐report screener [Standardized Assessment of Personality Abbreviated Scale – Self Report (SAPAS‐SR); Germans et al., 2012]. Interviewing took place using the SIDP‐IV (Structured Clinical Interview for DSM Axis II Disorders), a structured interview to assess personality disorders and comorbidity on DSM‐IV Axis II (Pfohl et al., 1997) with good psychometric properties and a mean duration of 90 to 120 minutes.

Assessment of course and severity of psychopathology

The course and severity of psychopathology was professional‐assessed with the HoNOS (Health of the Nation Outcome Scale): a 12‐item clinician‐rated instrument to assess general mental health in predominantly SMI‐patients (Bebbington et al., 1999) with good psychometric properties and a mean duration of 10 minutes.

Psychopathology was also patient‐assessed with the OQ‐45 (Outcome Questionnaire): a 45‐item instrument which assesses treatment outcome, mostly in terms of symptom reduction (Lambert et al., 1996) in patients with common mental disorders with very good psychometric properties and a mean duration of 10 minutes.

Physiological health indicators

Physiological health was assessed with the Somatic illness list: a 24–item check list, developed and used by the Netherlands Study of Depression and Anxiety (NESDA)‐consortium in an eight‐year cohort study, with a mean duration of five minutes (Penninx et al., 2008).

Neurocognitive functioning

An integrated set of neurocognitive measures (NeuCog), was used to assess cognitive biases that are hypothesized to cause and maintain common mental disorders. The following tasks were carried out, taking 30 minutes in total:

• EST (Emotional Stroop Task), an instrument that assesses processing bias for negative and positive information using randomized word sequences, measuring reaction times (Williams et al., 1996)

• SRET (Self Referent Encoding Task), assesses affective memory bias, resulting in two variables: the proportion of self‐referent negative recall and the proportion of self‐referent positive recall (Derry and Kuiper, 1981)

• TMT (Trail Making Test) assesses divided attention and psychomotor speed (Reitan and Wolfson, 1985) by connecting letters and numbers on white paper. The experimenter registers the time needed to complete by stopwatch.

• NART (National Adult Reading Task) assesses the verbal ability of the participant, providing an estimation of the verbal IQ (Nelson, 1982)

Past and present life events

The Brugha‐questionnaire assesses exposure to important negative events during life such as death or serious illness of other family members, unemployment and experiences of violence in 12 items (Brugha et al., 1985)

The Netherlands Mental Health Survey and Incidence Study (NEMESIS)‐childhood trauma questionnaire assesses trauma exposure during childhood (emotional neglect, psychological abuse, physical abuse, sexual abuse and important life‐events in early life) with a structured inventory of 10 items (de Graaf et al., 2010) with sufficient psychometric properties and a mean duration of 10 minutes.

Health care use and contact with mental health services

Content of care was assessed with the CCCL (Care Content Check List): a 20‐item purpose‐developed checklist, concerning the content of psychiatric care recently received, which is taken as an interview by a researcher, with a mean duration of eight minutes. Lessons learned from previously developed instruments (Lloyd‐Evans et al., 2011; Lloyd‐Evans et al., 2007) were used to inform the content of the CCCL [content available from 1st author].

The Tic‐P (Trimbos/iMTA questionnaire for Costs associated with Psychiatric Illness): is a widely used 43‐item, instrument that measures direct costs of medical treatments such as the number of contacts with psychiatric services, the general practitioner (GP) and multiple other care providers (Hakkaart‐van Roijen, 2002) with a mean duration of 10 minutes.

The quality of the therapeutic alliance in community mental health care, according to the patient and professional, was assessed with the STAR (Scale To Assess the Therapeutic Relationship): a 12‐item self‐assessment instrument (McGuire‐Snieckus et al., 2007), with good psychometric properties and a mean duration of three minutes.

Problems in the relationship between patient and professional were assessed with the DDPRQ (Difficult Doctor Patient Relationship Questionnaire): a 10‐item instrument (Hahn et al., 1996) with very good psychometric properties and a mean duration of 3 minutes.

The patient's perceived “difficulty” was assessed with the PPD (Perceived Patient Difficulty): a professional‐rated single‐item seven‐point Likert score with a mean duration of one minute. This has been used before (Koekkoek et al., 2011b), but no psychometric evaluation has taken place yet.

Met, unmet, and total needs for care as viewed by both client and professional were assessed with the CANSAS (Camberwell Assessment of Need Short Appraisal Schedule), a 23‐item checklist (Phelan et al., 1995) with a mean duration of five minutes. Professionals fill it out themselves, while patients are interviewed by a researcher.

Patient satisfaction with (mental) health care was patient‐assessed with the CSQ‐8 (Client Satisfaction Questionnaire): an eight‐item instrument (Larsen et al., 1979) with good psychometric properties and a mean duration of five minutes.

Social functioning and quality of life

General quality of life was patient‐assessed with two measures. First, with the EQ‐5D (EuroQol 5D): a 14‐item generic instrument that can be used to calculate quality‐adjusted life years (QALYs) (König et al., 2010) with good psychometric properties and a mean duration of five minutes.

Second, with the MANSA (Manchester Short Assessment of Quality of Life): a 16‐item patient‐rated instrument that assesses quality of life in SMI‐patients (Priebe et al., 1999), with good psychometric properties and a mean duration of five minutes.

The extent and quality of the individual's social network was assessed with the SNM (Social Network Map): a structured researcher‐based interview (Tracy and Whittaker, 1990) with sufficient psychometric properties and a mean duration of 15 minutes.

Recovery and well‐being

Disability in broad physical and psychological terms was assessed with the WHODAS 2.0 (World Health Organization Disability Assessment Schedule II; Üstün et al., 2010): a 12–item patient‐assessed instrument (short version) with a mean duration of three minutes.

Positive mental health was assessed with the MHC‐SF (Mental Health Continuum‐Short Form): a 14‐item instrument (Lamers et al., 2011), with good psychometric properties and a mean duration of three minutes.

Psychological flexibility was assessed with the AAQ‐II (Acceptance and Action Questionnaire‐II): a 10‐item instrument with sufficient psychometric properties (Bond et al., 2011) and a mean duration of three minutes.

Psychiatric and social functioning was professional‐assessed with the GAF‐score: a single‐item clinician‐rated composite score between 10 and 100 (Goldman et al., 1992), with a mean duration of one minute.

Procedure

Recruitment

In each participating department of the three institutions, key clinicians sent out information and invitations to each patient in their case‐load. Key clinicians were treating professionals that had most contact with patients. Patients could contact research staff (local investigator or principal investigator) directly, ask questions, and agree to participate. Alternatively, they could turn to their key clinician for information and application. Shortly after having received an initial positive response, local research staff made telephone calls with patients to inquire whether they were indeed willing to participate. If they were, an appointment was made for a face‐to‐face meeting, which took place in either the mental health service, the research facility, or at the patient's home. During this face‐to‐face appointment, verbal and written information was provided, informed consent was obtained and the baseline assessment was made. A gift certificate with a value of 20 euros was handed out.

Staff training and supervision

The MATCH study is centrally coordinated by the Research Group Social Psychiatry and Mental Health Nursing at the Hogeschool Arnhem Nijmegen (HAN) University of Applied Sciences in Nijmegen. The fieldwork coordinator at this site is responsible for the training and supervision of research assistants in all participating mental health services. A group of seven research assistants, all experienced and Master‐trained mental health professionals conducted baseline assessments. Newly appointed research assistants, for follow‐up assessments, received a three‐month training on‐the‐job before engaging in non‐supervised assessments. All baseline‐assessments were audiotaped, and monitored by the fieldwork coordinator in order to address any misunderstandings or errors in performing measurements. Booster sessions took place every six weeks at the research group site.

Data management and control

Data management is conducted by the Research Group Social Psychiatry and Mental Health Nursing at the HAN University of Applied Sciences in Nijmegen. The data manager provides electronic data collection programs, processes and prepares the data for research, monitors and updates administrative data from participants, and provides back‐up procedures.

Ethical issues

The Radboud University Medical Centre Ethical Committee granted approval of the current study in November 2012, and all of the participating centres gave permission to cooperate in this study from their own local Research Committees, in 2012 or 2013. All participants received verbal and written information with regard to the objectives and procedure of the study, with specific attention drawn to their right to refuse or stop participating at any time during the study, as well as specific information on the investment required from the participant.

Confidentiality of data is ensured by keeping separate files linking personal information such as name and address to a unique research ID number, which can only be accessed by the principal investigator and the data manager.

Timeline and follow‐up assessments

Recruitment took place from December 2012 to August 2014. Following the baseline measurement (consisting of structured interviews, neurocognitive tasks, and questionnaires), the yearly follow‐up assessments consist of a series of questionnaires containing repeated assessments of core self‐report instruments in order to establish the course of psychosocial functioning, therapeutic alliance, care needs, as well as recent events, and health care utilization (see Table 2 for details). Additionally a telephone interview is conducted, assessing a number of psychosocial variables. Responding is rewarded by a gift certificate that increases in value over time from 10 (first follow‐up) to 20 euros (fourth follow‐up) by Year 4 of the study

Analysis

All cross‐sectional data, longitudinal data and/or hierarchically ordered data will be analysed using SPSS or R, or another software package if appropriate using conventional statistical techniques and linear mixed models. Analyses in this paper were performed with SPSS, using χ ²‐tests for nominal variables, and analysis of variance (ANOVA) for continuous variables.

Missing data

Missing data are a primary concern in a longitudinal naturalistic study. The following precautions have been taken to prevent missing data:

• item non‐response is prevented by using electronic surveys and questionnaires that limit the possibility of missing items to the absolute minimum

• assessment non‐response is electronically tracked, followed up by (1) a reminder by e‐mail, (2) a reminder by telephone call, (3) a second reminder by telephone call, and invitation to assist the participant – either online, by telephone or face‐to‐face – in filling out the assessment, (4) a home visit by a researcher if communicated with, and agreed on by the participant

In the unlikely event of missing items in questionnaires, this is managed by multiple imputation. Intermediate missing assessments (i.e. a participant missing a follow‐up assessment) is handled by estimating coefficients using linear mixed models. If imputation of missing data is deemed necessary, this is done by multiple imputation methods.

Results

Over a period of 19 months, 283 participants were included (see Figure 1) and took part in the baseline measurement. The average participant is most often female (69%), somewhat under 40 years of age (mean 38.4), unmarried (69%), and of Dutch descent (94%) (Table 3).

Figure 1.

Flowchart of the MATCH cohort study.

Table 3.

Baseline socio‐demographic characteristics of MATCH participants per treatment setting

Variable		Primary N = 80	Secondary N = 100	Tertiary N = 103	Total N = 283	Difference (x 2/F)	p‐Value
Gender	Male	21 (26%)	33 (33%)	33 (32%)	87 (31%)	1.08	0.58
	Female	59 (74%)	67 (67%)	70 (68%)	196 (69%)
Age	Mean (SD)	41.2 (11.3)	41.8 (10.6)	32.8 (10.1)	38.4 (11.4)	22.52	0.00
Marital state	Married, registered partnership	27 (34%)	36 (36%)	24 (24%)	87 (31%)	4.12	0.13
	Unmarried	53 (66%)	64 (64%)	78 (77%)	195 (69%)
	Unknown	0	0	1	1
Native country	Netherlands	77 (96%)	93 (93%)	97 (94%)	267 (94%)	5.58	0.85
	Other	3 (4%)	7 (7%)	6 (6%)	16 (6%)
Education	Primary	11 (14%)	11 (11%)	9 (9%)	31 (11%)	25.20	0.03
	Secondary	40 (50%)	58 (58%)	67 (65%)	165 (58%)
	Tertiary	29 (36%)	31 (31%)	27 (27%)	87 (31%)

Most frequent diagnoses (Table 4) are mood disorders (25.1%), anxiety disorders (43.9%), and substance‐related disorders (13.5%). Also, personality disorders are highly frequent among participants (42.0%). Eating disorders (2.6%), somatoform disorders (7.9%), and attention and behaviour disorders (9.2%) are much less frequent. No significant differences in prevalence of the disorders were observed between primary, secondary, and tertiary services. The number of diagnoses per individual (Table 5), however, does differ between treatment settings with comorbidity being most pronounced in tertiary services. Using structured interviews for Axis I and Axis II diagnoses, we found 69 participants (24.4%) without any Axis I or Axis II diagnosis, of whom 31 (11.0%) in primary, 28 (9.9%) in secondary, and 10 (3.5%) in tertiary services.

Table 4.

Baseline number of DSM‐IV Axis I and Axis II diagnoses among MATCH participants per treatment setting

	Primary	Secondary	Tertiary	Total	χ 2	p
DSM IV Axis Ia	N = 80	N = 100	N = 103	N = 283	χ2 (12) = 10.41	0.58
Mood disorders	21 (26.6%)	28 (23%)	46 (26%)	95 (25.1%)
Anxiety disorders	33 (41.8%)	62 (50.8%)	71 (40.1%)	166 (43.9%)
Substance‐related disorders	14 (17.7%)	10 (8.2%)	27 (15.3%)	51 (13.5%)
Eating disorders	0 (0%)	3 (2.5%)	7 (4%)	10 (2.6%)
Somatoform disorders	7 (8.9%)	10 (8.2%)	13 (7.3%)	30 (7.9%)
Attention and behaviour disorders	4 (5.0%)	9 (9.0%)	13 (12.6%)	26 (9.2%)
Total number of Axis I diagnoses among participantsb	79 (100%)	122 (100%)	177 (100%)	378 (100%)
DSM IV Axis IIc, d	N = 18	N = 37	N = 64	N = 119	χ2 (16) = 22.32	0.13
Paranoid	1 (7.7%)	6 (12.8%)	9 (11.4%)	16 (11.5%)
Schizoid	1 (7.7%)	0 (0%)	0 (0%)	1 (0.7%)
Schizotypal	0 (0%)	3 (6.4%)	0 (0%)	3 (2.2%)
Antisocial	1 (7.7%)	3 (6.4%)	1 (1.3%)	5 (3.6%)
Borderline	4 (30.8%)	12 (25.5%)	18 (22.8%)	34 (24.5%)
Histrionic	1 (7.7%)	1 (2.1%)	3 (3.8%)	5 (3.6%)
Narcissistic	0 (0%)	1 (2.1%)	3 (3.8%)	4 (2.9%)
Avoidant	3 (23.1%)	16 (34%)	34 (43%)	53 (38.1%)
Dependent	2 (15.4%)	5 (10.6%)	11 (13.9%)	18 (12.9%)
Total number of Axis II	13 (100%)	47 (100%)	79 (100%)	139 (100%)

^aMore diagnoses per person possible.

^bDerived at from MINI‐Plus Interview

^cDerived at from SIDP‐IV interview, following positive screening with SAPAS‐SR.

^dThe N in this part of the table refers to participants with an Axis‐II disorder, which are 119 of 283 participants.

Table 5.

Comorbidity of DSM‐IV diagnoses of MATCH participants per treatment setting

	Primary	Secondary	Tertiary	Total	χ 2	p
DSM‐IV Axis I Currenta	N = 80	N = 100	N = 103	N = 283	χ 2 (10) = 27.39	0.002
Persons without diagnosis	31 (38.8%)	30 (30%)	12 (11.7%)	73 (25.8%)
… with one diagnosis	26 (32.5%)	35 (35%)	32 (31.1%)	93 (32.9%)
… with two diagnoses	18 (22.5%)	24 (24%)	38 (36.9%)	80 (28.3%)
… with three diagnoses	3 (3.8%)	8 (8%)	16 (15.5%)	27 (9.5%)
… with four diagnoses	2 (2.5%)	3 (3%)	4 (3.9%)	9 (3.2%)
… with six diagnoses	0 (0%)	0 (0%)	1 (1%)	1 (0.4%)
DSM‐IV Axis IIb, c	N = 18	N = 37	N = 64	N = 119	χ 2 (10) = 11.95	0.29
Persons without diagnosis	10 (55.6%)	11 (29.7%)	16 (25%)	37 (31.1%)
… with one diagnosis	5 (27.8%)	14 (37.8%)	27 (42.2%)	46 (38.7%)
… with two diagnoses	1 (5.6%)	6 (16.2%)	14 (21.9%)	21 (17.6%)
… with three diagnoses	2 (1.1%)	3 (8.1%)	5 (7.8%)	10 (8.4%)
… with four diagnoses	0 (0%)	3 (8.1%)	1 (1.6%)	4 (3.4%)
… with five diagnoses	0 (0%)	0 (0%)	1 (1.6%)	1 (0.8%)
Persons without any Axis I or Axis II diagnosis	31 (11.0%)	28 (9.9%)	10 (3.5%)	69 (24.4%)

Note: mean number of Axis I diagnoses per person = 1.9 [standard deviation (SD) 1.1], mean number of Axis II diagnoses per person = 0.6 (SD 1.1).

^aDerived from the MINI‐Plus interview.

^bDerived from the SIDP‐IV interview, following positive screening with SAPAS‐SR.

^cThe N in this part of the table refers to participants with an Axis‐II disorder, which are 119 of 283 participants.

Table 6 shows comorbidity patterns, with the combination of an anxiety disorder and a mood disorder being the single most frequent pattern (35.9%), followed by the much less frequent combination of a mood disorder and a substance‐related disorder (12.8%). Among people with personality disorders, the cluster B‐combination of avoidant and borderline personality disorder is most frequent (16.7%), followed by the combination of avoidant personality disorder with dependent and paranoid personality disorder (both 11.1%).

Table 6.

Most frequenta comorbidity patterns across DSM‐IV Axis I and Axis II diagnoses

	Combination of diagnoses	Number of respondents
	DSM‐IV Axis I Currentb
1	Anxiety disorder and Mood Disorder	42 (35.9%)
2	Anxiety disorder and Substance‐Related Disorder	15 (12.8%)
3	Anxiety disorder and Mood Disorder and Somatoform Disorder	8 (6.8%)
4	Anxiety Disorder and Somatoform Disorder	7 (6%)
5	Anxiety disorder and Mood Disorder and Substance‐Related Disorder	6 (5.1%)
	DSM‐IV Axis IIc
1	Avoidant and Borderline personality disorder	6 (16.7%)
2	Avoidant and Dependent personality disorder	4 (11.1%)
3	Avoidant and Paranoid personality disorder	4 (11.1%)
4	Borderline and Dependent personality disorder	2 (5.6%)
5	Avoidant and Borderline and Paranoid personality disorder	2 (5.6%)
	Avoidant and Dependent and Paranoid personality disorder	2 (5.6%)

^aOnly comorbidity patterns listed when total frequency > 5%.

^bDerived from the MINI‐Plus interview.

^cDerived from the SIDP‐IV interview, following positive screening with SAPAS.

Table 7 shows services received by participants. As often with service use data, distribution is skewed, due to participants that have received up to 360 months (30 years) of services. Current care differs significantly across treatment settings, with primary services almost exclusively being outpatient (97.5%), while tertiary services are also offered in day treatment and hospital settings (37.9% and 21.4%). Time since first contact with services is relatively short (< two years) for 29.7% of patients, yet the highest proportion (30.8%) is found among people whose first contact with services was between 11 and 25 years earlier.

Table 7.

Services received by MATCH respondents at baseline assessment per treatment setting

Current carea	Primaryb N = 80	Secondaryb N = 100	Tertiaryb N = 103	Duration in months (SD; median; range)c	Total
Outpatient contact	78 (97.5%)	89 (89.0%)	47 (45.6%)	29.0 (49.9; 9.5; 0–360)	214
Supported living or psychiatric home treatment	2 (2.5%)	7 (7.0%)	2 (1.9%)	50.8 (61.1; 16.5; 1–159)	11
Day clinic or partial hospital	0	5 (5.0%)	39 (37.9%)	3.9 (14.4; 1.0; 1–82)	44
Inpatient or residential	2 (2.5%)	6 (6.0%)	22 (21.4%)	4.9 (12.0; 1.0; 2–51)	30

Note: χ ² (6) = 94.25, p < 0.001; data for eight persons is missing.

^aMore forms of current care are possible.

^bRefers to treatment setting in which participant was recruited.

^cTotal months of mental health services received so far (range).

Discussion

In this paper we presented the outline of the MATCH cohort study and some baseline data. We highlight three aspects of the sample, all relevant to our research aims: (1) comorbidity is very common and not limited to having only two concurrent disorders (Tables 5 and 6), which enables us to establish shared characteristics across disorders; (2) in terms of treatment duration, the larger part of our sample receives care shorter than two years (Table 7), which enables us to establish development of chronicity in a substantial number of participants; (3) the sample appears to appropriately represent known populations estimates of people using services, and the distribution of individuals across levels of care (outpatient, day treatment, inpatient). Although national data (de Graaf et al., 2010) is less detailed we found service use in the general population to be lowest for substance abuse disorders relative to anxiety and mood disorders (1: 1.3: 2.2), wheras in our sample the rates were 1: 3.3: 1.9, thus making anxiety disorders relatively overrepresented. Comparisons are difficult in general, since national data reflects only main diagnosis (an administrative term) while our structured diagnostic interviews list all diagnoses. The high percentage of people with Axis II‐disorders may be due to the oversampling, compared to national proportions, of people from tertiary services since personality disorders are known to be more frequent in more intensive services (e.g. Zimmerman et al., 2005). Of all adults receiving mental health care in the Netherlands, 91.5% are in outpatient care (75.6% in our sample) and 8.5% in inpatient care (10.6% in our sample). These differences may be largely explained by our subdivision of outpatient services into day clinics and supported living, which in our study are considered as different forms of care than outpatient services.

When comparing our study to other longitudinal studies among people with common mental disorders, we see that our study has both strengths and weaknesses. The assessment of service use and social variables is more extensive in our study than for instance in the NESDA‐study on anxiety and depression (Penninx et al., 2008), and the Harvard/Brown Anxiey Research Program (HARP) on anxiety disorders (Bruce et al., 2005). Many other studies are community‐based and lack the in‐depth assessment of a clinical cohort like MATCH. Yet, all aforementioned studies have larger samples, and the NESDA‐study incorporates a broad range of instruments, including biological variables.

Conclusion

The MATCH cohort study is a multi‐site, longitudinal, naturalistic cohort study examining the four‐year course and consequences of common mental disorders. This study specifically examines interactions with mental health services and social variables, an element that to date has been researched less intensively.

Declaration of interest statement

The authors have no competing interests.

Koekkoek, B. , Manders, W. , Tendolkar, I. , Hutschemaekers, G. , and Tiemens, B. (2017) The MATCH cohort study in the Netherlands: rationale, objectives, methods and baseline characteristics of patients with (long‐term) common mental disorders. Int J Methods Psychiatr Res, 26: e1512. doi: 10.1002/mpr.1512.