Fifty years' development and future perspectives of psychiatric register research


  • This article is commented on in two editorial comments in this issue: Stewart R. The big case register. Acta Psychiatr Scand 2014, 130:83-86 and Amaddeo F. The small scale psychiatric case registers. Acta Psychiatr Scand 2014, 130:80-82.



This article illustrates the development of psychiatric register research and discusses the strengths, limitations, and possible directions for future activities.


Examples illustrating the development from the post-World War II introduction of psychiatric register research until today are selected.


The strengths of register research are seen especially within health service. Until recently, when starting linking registers to biobanks, register research had limited value in cause-seeking. Register research benefits from the possibilities for following identifiable persons over long time (lifelong) and the possibilities for linking to other registers and databases. Important limitations of register research are the heterogeneity and questionable validity of the clinical data collected.


Future register research can go in the direction of big is beautiful collecting data from all possible sources creating giga-registers. In that case, low data quality will still be an unsolved problem. Or it can take the direction of smaller local clinical databases which has many advantages, for example, integrating clinical knowledge and experience into register research. However, in that case, registers will not be able to deal with rare conditions and diseases.

Clinical recommendations

  • Modern psychiatric register research has its strength in linking registers, databases, and biobanks.
  • Psychiatric registers are in specific useful tools in health service research.
  • When interpreting psychiatric register research, the reader must be aware of limited data quality.

Additional comments

  • Direction of future organization may choose different ways, for example giga-registers and smaller clinically integrated registers.
  • Huge register research centers risk floating away from daily clinical practice and research.
  • Initiatives from international organizations, for example WHO, to set standards for quality and interpretation of register research should be most welcomed when future directions for psychiatric register research will be developed.


‘The mortality among the severely mentally ill compared with the background population has not decreased’, ‘the length of hospitalization periods continues to decrease’, and ‘readmission rates are still on the increase’ are all statements that are easily verified in any country with a well-functioning health register. What we cannot demonstrate by means of registers is why the mortality is still high, or whether the decreasing length of hospitalizations is due to increasing quality of treatment, or whether is it because of mere administrative decisions. The registers can also not help us in answering why readmission rates are still increasing.

Register research, in its present form, is based on the ability to process large, if not enormous, datasets and, most of all, society's acceptance of the registration of identifiable personal data, and acceptance of data being used for research and administrative purposes.


In a 2009 editorial published in the British Journal of Psychiatry, Perera and coauthors [1] revitalized the now more than 25-year-old definition by Ten Horn [2] of ‘a patient-centered longitudinal record of contacts with a defined set of psychiatric services originating from a defined population’. This definition originates from a 1983 World Health Organization (WHO) working group from Mannheim, Germany [3].

This is a very broad, but useful, definition, which highlights the two principal requirements:

  1. Following a patient on his/her way through a variety of treatment institutions
  2. Covering a well-defined population, that is, a well-defined administrative catchment area, for example health service district, municipality, county, or country, which, until now, is the largest area registered.

Under this umbrella definition, each register must operationalize the definition, for example, as carried out for the Psychiatric Case Register Middle Netherlands by Smeets et al. [4] or for the Danish Psychiatric Central Register [5-7], the latter fitting the following criteria:

  1. clearly defined variables
  2. data are person-identifiable through a unique person-identifiable number making it possible to follow the patients through psychiatry, somatic health service, primary care, social registers, taxation register, etc.
  3. data are exhaustive, for example a defined type of institution covered by the register must report all of them when situated in the catchment area covered
  4. the register is data-excluding, that is if a certain variable, for example auxiliary diagnosis or somatic diagnosis is not reported from all institutions reporting to the register, those reported will be excluded
  5. data are collected continuously—there are no time limits for the existence of the register, every time a patient enters one of the reporting institutions, his/her present data are added to his/her record in the register.

Aims of the study

To illustrate the development of psychiatric register research and discuss the strengths, limitations, and possible directions for future activities.

Material and methods

Examples illustrating the development from the post-World War II introduction of psychiatric register research until today are selected.


Types of use

More attempts to classify the use of modern health registers for research and clinical purposes have been launched, for example, by Mortensen in 1995 [8], and Tansella and Ruggeri [9] and summarized by Tansella [10] and, more recently, by Wiedsma et al. [11].

We shall not propose yet another classification or grouping of the use of registers, but restrict ourselves to present and discuss some examples of the most frequent types of research. These will be health service research, outcome studies, identification of representative clinical cohorts, follow-up of clinical cohorts and linkage to biobanks.

As the majority of health registers are established, processed, and used for administrative purposes, they are used mainly for descriptive/administrative research. Registers are absolutely at their best when used in health service research.

Health service research

The simplest forms of health service research, or, perhaps, better medical statistics, are prevalence and incidence studies. It is of interest to the professionals and not least to administrators and politicians to see how the psychiatric services are used, for example regular description of diagnostic profiles in each hospital obtained through sending out questionnaires to each department in a huge catchment area (e.g. Denmark), collecting the data, validating them, and analyzing them was an enormous effort 50 years ago [12]. This has now become very easy and can be carried out almost automatically within hours with the electronic registers—a technical possibility that was already available in the 1970s [12].

Through modern statistical methods, we can achieve a high quality overview of, for example, the development of a changing diagnostic profile, with work by Lay et al. [13] from Canton Zürich finding a 50% reduction in the length of in-patient episodes for patients with psychosis from 1977 to 2004, a 100% increase of in-patient admissions and, parallel to this, a mere 50% reduction in the proportion of schizophrenia among in-patients from 1977 to 1993, this on a background of a two- to threefold increase in other patient groups.

In a thorough review of the prevalence of schizophrenia by Saha et al. [14], a great deal of the papers based on register studies are included.

Pharmacoepidemiological studies help us to point out different treatment principles in different diagnostic and treatment cultures; for example, the use of pharmaceutical compounds prescribed for the treatment of attention deficit hyperactivity disorder (ADHD) using prescription databases shows a sevenfold variation among the five Nordic countries [15]. However, is this not caused by variation in the incidence of ADHD? This question illustrates the most important limitation of administrative health service registers: that we cannot add anything about the causes for our findings. Is the variation due to different treatment cultures? Is it due to different diagnostic habits? Is it due to variations in identification rates of the disease? Or, is it due to true differences in the population of incidence/prevalence of the disease?

Incidence studies represent a register research discipline producing data in the grey zone between health service research for administrative and political use on the one side, and causative research on the other.

As recent examples we could mention some very simple studies, for example an increase in new cases of ADHD over a certain period of time (Fig. 1) [16] or in the use of schizophrenia diagnosis in children and adolescents [17].

Figure 1.

Age-standardized incidence of rate of ADHD [16].

Changes in incidence over time or difference in incidence in different environments give us the possibility of asking what the causes for these differences are. The problem is that the incidences we can identify in health service registers are not population incidences—which is what we need for performing causative research—they are treated incidences, that is, they are dependent on how many cases pass the threshold from the population and primary care into secondary care, a pathway depending on the capacity of secondary care and competences, and the willingness of primary care to refer and the willingness of the population to be treated in secondary care. When having passed these thresholds [18], the patient then faces different diagnostic cultures and traditions among services and over geography, to mention but a few of the most common causes for heterogeneity found in the registers.

The question about how many cases in the population are reaching secondary health care and therefore enter the psychiatric registers has been a topic for debate during the entire 50 years of electronic-based register research. An elegant example validating this has been performed by Weiser et al. [19] through Israeli registers. The authors used data from a population survey and compared their findings with the Israeli National Psychiatric Hospitalization Registry 24 years later, identifying 93% of the cases from the survey.

Figure 2 shows a classic example of medical statistics/health service research. The figure illustrates the percentage of patients discharged from in-patient treatment diagnosed with schizophrenia until their first appearance in an out-patient service within 30 days, 3 and 6 months respectively. The main information from the curves is an increase of 10–15% over a 10-year period, reaching close to 70% registered as out-patients within a 6-month period following discharge from in-patient treatment. This type of result must never be used without being discussed or commented upon. At first glance, there is a positive trend—the percentage of patients continuing in an out-patient setting within a maximum of 6 months of discharge is increasing. More discouraging is the result that fewer than 70% reach an out-patient treatment program within 6 months of discharge, which is a result seemingly without improvement since 2001/2002. However, there are many reservations about the results that need to be dealt with before definitive conclusions can be drawn.

  1. We do not know what kind of patients have been investigated. The data behind the statistics are digits in the register, which inform us that someone in clinical psychiatry has diagnosed a person with schizophrenia and reported it to the register, and nothing else.
  2. We do not know whether it is very severely ill patients.
  3. We do not know whether they are mildly ill.
  4. We do not know what has happened to them, in particular those >30% not initiating an out-patient treatment program within 6 months of discharge from in-patient treatment. Have some of them dropped out from treatment? Have some of them died? Are some of them in good care in social psychiatric institutions, in sheltered livings or in hostels (not reporting to health registers in Denmark)?
  5. Are they homeless and living on the streets or are they living with elderly parents?
  6. Are they well treated by their general practitioner (GP) or by a practicing psychiatric specialist (none of them reporting to the psychiatric register in Denmark)?
  7. Or has something else happened?
Figure 2.

Time from latest schizophrenia discharge to out-patient contact (Source: The Danish Psychiatric Central Research Register).

Our sole conclusion is that between 65% and 70% of those discharged from a psychiatric in-patient treatment having been diagnosed with schizophrenia are showing up in the secondary healthcare out-patient setting within 6 months of discharge. What kind of treatment they get, what the quality of this treatment is, and what has happened to those not included we cannot say because of the simple statistics. Making a regular health service research effort, linking the psychiatric register to the causes of death register, to the register on where people are living, to the GP and practicing psychiatric specialist register identifying days of consultations, to the prescription register to see whether they have obtained the medicine prescribed, to social registers to see whether they have been in contact with any social welfare systems, etc., would help us; however, the quality of treatment and quality of living of those not in any identifiable treatment cannot be ascertained. As to the question about the validity of data from registers, we must conclude that no matter how clear and simple the curve seems, clinical information is very sparse.

The more severe a disease, the higher the possibility is of finding the patient again in the register. In contrast, the milder a case, the lower are the possibility is of finding the person in the register and is even lower the more specialized the service reporting to the register. This severe limitation may even be serious to our understanding of a disease such as schizophrenia because we have very limited knowledge about possible mild cases of schizophrenia if, indeed, they exist. We tend to define schizophrenia based on the cases seen in secondary healthcare-identifiable cases in the registers.

Concerning affective disorders, in particular major depressive disorders, the picture is somewhat different because it is so obvious that only 5–10% (maximum) of the cases reach secondary healthcare [18]. That, however, implies that the value of register research on major depressive disorders is of very little use for the treatment of the 90–95% of major depressive disorders in the population, of which only approximately half reach treatment level in primary healthcare, thus introducing bias into the case register system. When considering Denmark, it should be noted that out-patient visits were included in 1995 and that Denmark shifted from International Classification of Diseases (ICD)-8 to ICD-10 in 1994.

Revisiting schizophrenia, incidence (which is, in fact, treated incidence) has shown tremendous oscillations from around 10 and 5 per 100 000 male and female in the early 1970s to approximately 6 and 4 in the early 1980s and is now around 20 in total in the 2000s [[20, 21] The Danish Psychiatric Central Research Register, personal communication].

Outcome and prediction of outcome

Suicidology has for many years now been a star discipline within register research, perhaps because the event—the suicide—is a very serious incident, a catastrophe, and may also be because the validity of the variable—suicide—is very high in operating registers. The first steps taken 50 years ago in register-based research mostly described incidences of suicide and changes in these incidences and methods of suicide (poisoning, shooting, hanging, etc.). The cases of suicide could be identified in national mortal statistics, and subsequently, the patient could be searched in health registers and vice versa. The next step after mere description of changes in incidence was prediction analysis, which was made possible through the linkage of two or more registers. An early study benefitting from linking several registers using modern computer technology able to handle very large datasets and modern epidemiology and biostatistics is seen in the study from Mortensen's group [22]. Later, we have seen very high quality examples from Sweden from Ösby's group [23] who investigated more than 20 000 suicides, replicating Rosseau and Mortensen's [22] finding of the immediate phase after discharge from psychiatric hospitalization as the most high-risk period [23]. Furthermore, the Danish group [24] identified the important finding that a decrease in suicide rates in schizophrenia is parallel to a decrease in the background population.

These studies illustrate the importance of linking registers, maybe the most dramatic progress in epidemiological research, which may be comparable to the most revolutionary breakthroughs in biological medicine, such as brain imaging techniques and nanotechnology.

Regrettably, the predictors identified in this kind of suicide register research are of limited value in directly preventing clinical and public health interventions because suicide is, however tragic, a very rare event resulting in the predictors receiving a low predictive value; for example, it is not of any clinical use to know that the odds ratio for men to commit suicide compared with female is 2—a 100% doubling. A further severe limitation in predictive register research is the predictor variables available. We can only very rarely use those variables which, from a theoretical point of view, could test our hypothesis. Instead, we are restricted to using the variables already in the registers, mostly for administrative purposes, that is, our analysis will be predominantly explorative instead of hypothesis testing.

In contrast to suicide research, register research aimed at predicting physical illness in mental illness and mortality rates in psychiatrically ill patients from physical illnesses has a much higher impact because of the, unfortunately, very high occurrence of physical illnesses in the mentally ill. This research started peaking in the years following Harris and Barreclough's benchmarking study from 1998 [25], showing increased mortality in mental illness because of physical diseases.

Register-linking studies are a unique tool in mapping this issue, for example the study by Munk-Jørgensen et al. [26], which followed an explorative model examining a broader spectrum of physical diseases in schizophrenia. More specifically, association studies have been performed between schizophrenia and autoimmune diseases [27], between autoimmune diseases and severe infections as risk factor for schizophrenia [28], acute myocardial infarction in depression, anxiety and schizophrenia [29], and hypertension in bipolar disorder, anxiety and schizophrenia [30].

However, having access only to those variables already in the registers and not to those which would be most relevant is still a problem.

Identification of representative clinical cohorts

As we will further discuss in this chapter, register research has its major strengths in having a high level of representativity and therefore generalizability of results. The price to pay is the low data validity and reliability. A certain way of benefitting from the strength, the representativity and avoiding the limitations, the low data validity and data reliability is to combine registers with clinical data. This is carried out by identifying a representative cohort years back and following it in a register and contacting the identified people for further clinical research interview and investigation. An example of this method has been used by Munk-Jørgensen et al. [31]. Such a cohort, however, suffers from low data quality at index.

If the purpose is to create a representative cohort which actually has its first contact with the health service system ‘now’ for future personal research-based interviews and examination, we have almost created the highest quality situation benefitting from as high level of representativity, as well as a high level of data quality. In this situation of high representativity and high data quality, we lose one of the advantages of register research: the historical prospective follow-up in the register, that is, we have the disadvantages of long-term personal research follow-up, which are the high costs, and we still have to rely on retrospective data when collecting data previous to the first admission, which is the first reported to the register.

Also, from Ösby's Swedish group, we find a classical example of combining identification of probands in registers followed by collection of detailed clinical data in casu from case records when analyzing the diagnostic profile among suicide cases [32].

Follow-up of clinical cohorts

Clinically established cohorts suffer from two limitations when followed up. First, because of being clinically established, they can, in the vast majority of situations, have only local representativity and therefore limited generalizability; second, they suffer from loss to follow-up because of death, moving out of the catchment area and dropping out of treatment/research program, etc.

When using registers to supplement a personal follow-up, it is possible to identify the addresses and re-establish a personal contact to some of the dropped-out probands and those having left the catchment area. As to those who died, immigrated or in other ways ‘disappeared’, it is possible to identify some, but not great quality, data, for example the number of hospitalizations, the number of prescriptions picked up, and visits to general practitioners or practicing specialists. A fine example of this is 5-year follow-up of a first-episode psychosis cohort by Bertelsen et al. [33].

Some very large sample clinical databases designed for long-term follow-up studies which, however, cannot be defined as registers, are more valuable for cause-searching research than (administrative) health registers, for example the New Zealand Christchurch cohort [34], the UK 1946 birth-cohort [35] and the Finland 1966 birth-cohort [36]. When using these databases, researchers have the opportunity to supplement clinical follow-up data with nationwide register data, improving data completion and, through that, generalizability [36].

More examples of this using a very large sample database followed up in registers can be obtained in the Swedish conscript database, which is linked to national registers by Allbeck's group. This has given us invaluable knowledge as to the discussion about cannabis as risk factor—direct or indirect—for suicide [37]. The same database has used the principle of linking with national registers pointing at the fact that psychiatric diagnosis given in adolescents will increase the risk of suicide not only immediately after, but for many years or decenniums [38].

Linkage to biobanks

As mentioned at the beginning of this chapter, traditional psychiatric register research can describe the medical statistics of any given psychiatric illness, and it can predict course and outcome to a very detailed degree. Over the past 50 years, the methodology for doing research linking several registers has become continuously more sophisticated and, when using modern computer technology, there are no upper limits for how huge datasets can be handled and how many registers can be linked.

However, there is one unbreakable limit: traditional (administrative) register research cannot answer the ‘whys’. Who is at risk of developing a given disease can be predicted, but register research cannot identify the cause of the disease, that is, the pathogenesis. For that purpose, we have to investigate the unique patient and, as soon as we do so, we are in danger of losing the representativity. We can, to some extent, retain a certain degree of representativity using the methods for identifying a cohort for research by means of registers, but we are still down to a limited number of probands. There is a limitation both as to finance and time on how many patients we can identify and examine personally. However, this can, to a vast degree, be overcome by using the very large biobanks that have become established in recent years. Therefore, when linking traditional health service registers with biobanks, it is possible to bypass the time-consuming and expensive part of a research program: identifying the patients, appro-aching them, examining them, and collecting blood samples or other biological samples.

An early example of this is the identification of Toxoplasma gondii as risk factor for early-onset schizophrenia by Mortensen's group [39].


Strengths and limitations

We have touched upon the questions of strengths and limitations connected to register research a couple of times. This will now be further discussed, in particular what concerns register research when use is attempted in answering clinical questions.

Psychiatric register research deals with that part of a total population which has (i) been identified as having a psychiatric illness and (ii) overcome the threshold to an institution covered by registers useable for research.

The registers give no information about the entire population, only about individuals admitted and, furthermore, we know nothing about their diseases; the only information is that a group of individuals have been given a certain diagnosis. This fact is very often forgotten, and the studies are at risk of overconcluding. Examples of this are seen in research about suicide among psychiatric patients/former psychiatric patients compared with suicide in the general population never admitted to psychiatric institutions. The reasons for possible differences/no differences discussed are almost pure guessing; we know nothing about the mental status of those committing suicide in the portion of the population having never been registered in a psychiatric register as having been in contact with a psychiatric institution. These individuals may well have been mentally ill without being treated or having been treated in services not reporting to the register, but are, however, defined as not mentally ill in the analyses.

Also, measuring changes in length of stay over time gives meaning only to those institutions covered by registers. The length of stay may easily decline in monitored psychiatric hospital but remain long, and maybe even increase, in a neighboring private hospital not reporting to the register in question.

A clinical naturalistic follow-up study, for example the Copenhagen High Risk Study [40], has a very high data validity of both explanatory variables and outcome variables because the researchers have followed the probands throughout the entire study collecting data only for research purposes. The representativity may be questionable, including a certain group of probands selected from what is available. It is very time-consuming, and therefore, also the expenses are very high. Moreover, the expenses are a limitation because the naturalistic long-time study requires a lot of manpower (Table 1).

Table 1. Strengths and limitations in register studies compared with some clinical types of studies
 Naturalistic follow-upCase–control studyHistorical register follow-up studyHealth service register research
  1. ↑: strength; ↑↑: important strength; ↓: limitation; ↓↓: severe limitation.

Validity of explanatory variables↑↑↑↑
Validity of outcome variables↑↑↑↑↑↑
Time consumption↓↓↑↑↑↑

The clinical case–control study sampling patients at the time of ‘outcome’ has very limited data quality in regard to explanatory variables because they must be collected from data sources, such as case records and interviews retrospectively, recalling past events. The validity of the outcome data is high because the researchers have the possibility of collecting data directly from the patient about variables designed for the study. The representativity is low because the probands represent only those patients alive and present, for example schizophrenia patients alive and in contact with the service performing the study. We lose all who have died and dropped out, etc., and who should have been included. The time consumption is a relative limitation because it takes time to dig up data from the many various data sources. This makes the expenses rather high, because all patients should be interviewed for collecting outcome data; however, they are not as high as in the long-term follow-up (Table 1).

In contrast to the two methods, the clinical long-term follow-up and the clinical retrospective case–control study, a historical prospective register follow-up study has limited data quality concerning both explanatory variables and outcome variables. These variables carry the impress of being clinical data and are, consequently, characterized by heterogeneity, changing clinical trends and approaches over time, geography and psychiatric schools, etc. The representativity is high, which is one of the strengths because the probands are identified in the registers when they were all present years ago at registration. After their index, they are followed in one or more registers, and it is even possible to supplement the register follow-up with a personal follow-up of those still alive and identified by public civil registers. It is not optimal because they are representatives only for those groups of patients included in the register. Time consumption is minimal, which is another of the strengths, and so are the expenses (Table 1).

To give register research justice, it must be underlined that if the purpose is certain areas of administrative health service research, for example development of consumption of bed days, consumption of out-patient visits, length of stay and readmission rates, etc., the register study is second to none, and all the items listed in the Table 1, validity of both explanatory and outcome variables, representativity, time consumption, and expenses should be given maximum rating for strengths.

We have pointed at the dubious quality of the validity of data in the large (nationwide) administrative health registers. This calls for a continuous validation of register data when used for research. Mors et al. [6] identified five validation studies of diagnoses in the Danish Psychiatric Register compared to case notes. Jørgensen et al. [41] from Sweden documented that register-originated incidence studies of non-affective psychosis would be severely underestimated if not including out-patient services.

Few studies validate register data against research interviews of patients. This method is exemplified by Hansen's validating study from 2000 [42]. These authors examined newly admitted patients for substance use disorders and compared their findings first to clinical case notes and then to register data. They found a high accordance between case notes and register data, but a severe under-reporting of diagnoses in case notes summing up the clinical findings. These findings point at the keeping of case records as the weak link not reporting of case note data to registers.

Byrne et al. [43] found only 14 studies in a systematic review of validity studies in psychiatric register research. Based on this review, Parker, in an editorial [44], argued strongly for ‘representative constituent dataset examined for validity at appropriate intervals’. Five years earlier Tansella [10], the founder of the South Verona Psychiatric Case Register predicted that there will ‘probably’ be a need for psychiatric case registers and pointed at the lack of ‘direct evidence’ for register cost-effective tools for improving our understanding of the causes, course, and outcome of mental disorders as well for making more rational use of mental health services’. Recently, however, Allebeck [45], in an overview article in which he argues for further utilization of registers in research, underlines the importance of critical consideration of quality with a specific focus on coverage, attrition, representativity and validity of registers. The debate about registers supports the continuous use of registers in psychiatric research; however, pointing at different aspects of validity problems which, in the future, must attract attention.


The first author of this chapter (Povl Munk-Jørgensen) remembers from the late 1970s and the early 1980s how the many instances of control, and administrative and political levels that should give permission for the use of data from registers became overwhelming, at least in Scandinavia; in some research milieus, this was considered as obstructionism. However, he finds that the present situation, 30–40 years later, is at the opposite extreme—it is now very easy to get permission to use register data. In practice, the public is, through its representatives, the politicians, no longer part of the process, for example, in Denmark having used and developed person-identifiable registers for almost half a century. One should anticipate a massive (read: political and/or media created) reaction against the use of public registers and databases within either a shorter or longer period of time. Twenty-five years ago, very large results were obtained through including at that time leading patient and relative organization (in Denmark) as an official collaborator in the use of the Danish Psychiatric Register for research. A massive resistance to the register over a period of 2–3 years changed into positive collaboration and support. Perhaps, it would be fruitful to revitalize a public debate about the extensive use of register research in certain countries.

The future for psychiatric register research

Where should psychiatric case register research go in the future?

We have exemplified some authors' arguments for the continuous use of registers in psychiatric research, but under the condition of an intensified validation of data [10, 43-45] and evaluation of cost-effectiveness of using registers in research [10] (Fig. 3). Wierdsma et al. [11] point out the severe lack of serving healthcare stakeholders with sufficient information for, for example, evaluating healthcare policies and cooperation between mental healthcare and public service; this despite a wide range of research carried out with the registers. One wonders whether the politicians had given that much financial priorities to treatment of first-episode psychosis if they had been fully informed about the more than 12- to 15-fold difference in admission of first-episode psychoses and readmission of schizophrenia [46], the latter a severely burdened group and extremely expensive as concerns public expenses, direct and indirect.

Figure 3.

First ever contact/readmitted with schizophrenia in Denmark, ICD-10 F20 [46].

Perea et al. [1] made a list of recent advantages in modern registers which should facilitate the use of psychiatric case registers.

One way to go may be to continue the classical linking strategy. Instead of ad hoc linking for each research project, this linking could be electronically permanent giving a researcher direct and immediate access to all the datasets needed; or maybe even a super-register could be created—technological advances allow this. However, no matter how many administrative registers we link, it will not solve the problem of the lack of clinical and biological data. However, as case notes are becoming increasingly electronic, it gives us access to full text case notes. This is being used in the South London and Maudsley NHS Foundation Trust Biomedical Research Centre (SLAM BRC) Case Register [47], which covers a 1 000 000+ catchment area. Similar ideas are being discussed presently in Denmark (P. Munk-Jørgensen, personal communication), which would cover 5.5 million inhabitants.

This model, if any, calls for a revitalized public–political debate about the ethical and juridical aspects of using person-identifiable data to such an extreme extent as, for example, called for by Larson in a JAMA View-point discussing the use of Big Data [48].

Another way to go may be an intensifying use of the classical local case register [49], a model sophisticated in, for example, the South Verona Psychiatric Case Register, as will be discussed below. This model has several advantages: It is integrated in clinical daily practice; the clinical working professionals will gain an epidemiological overview of their activities; and the research will be driven by relevant clinical problems and hypothesis. Additional patient-focused research activities will not be removed from the daily clinical treatment of patients, and it will be inspired by clinical problems.

A third model may be an intensifying of hypothesis-driven research based on linking data from administrative registers and biobanks. This would benefit the biological cause-seeking research in maintaining a high level of representativity; another advantage would be the possibility for a close connection to population's representatives, the ethical committees, the data surveillance authorities, politicians, and the population in general, making these bodies able to follow and discuss progress in psychiatric (register) research.

The clinically focused register revisited

Some minor registers as, for example, the Italian South Verona Psychiatric Case Register, can be defined as a continuation of the pioneering local registers presented in Psychiatric Case Registers in Public Health [49].

The South Verona Case Psychiatric Case Register has been in operation since 1979. It covers a population of 75 000 and is therefore able to collect several detailed data about psychiatry, somatic and social variables, with high validity and reliability. This register has contributed continuously to the international literature over a span of more than 30 years, documenting the importance of smaller, intensive clinical registers. Examples of this are studies about a highly relevant topic: mortality among psychiatric patients [50, 51], avoidable mortality [52], and health economic aspects of community psychiatry [53].

The South Verona Psychiatric Case Register was established as part of the South Verona Community Mental Health Service. It has therefore contributed to the ongoing and comprehensive evaluation of the service from its very beginning over the consolidating years into routine [54, 55] and further development over more than 30 years and ,gradually, also being used for more cause-seeking research.

The establishing of the register as an integrated part of a new service is an ideal model for how to act responsibly politically, organizationally and academically when a new service is established.

Academic articles published by this service show how well clinical research benefits from using the register (, for example Tansella et al. [56] using data from the register to provide illustrative documentation of the shift from in-patient to out-patient and day-patient treatment over 25 years of the service's treatment profile.

Final remarks

When rereading ten Horn and coeditors' benchmarking book from 1986 [48], it is striking how the authors of the different contributions mastered communicating their messages in a clear language, together with data presentation of high educational quality. This was also recently carried out by, for example, Amaddeo et al. [55].

Having these examples in mind when reading today's register-generated health research, one notices how the language has shifted toward epidemiological/statistical shop lingo and, consequently, is at great risk of losing the average clinical working reader or the public health service worker, both of who should be the key readers of the articles.

We, as the authors of this chapter, should be the first to acknowledge the impressive progress in theory, methodology and information technology capacity, which have characterized the development of register research since the early 1970s. However, a consequence—a side-effect, so to say—has been that register research is, step by step, being hidden away in epidemiological centers, health service research institutes and register research centers staffed with epidemiologists, statisticians and software engineers, some with only a few medically educated people with no, or only homeopathic, clinical experience. This is necessary to handle and develop methods and technique, but we are in danger of losing those colleagues who should benefit from the research.

In this process, the research is in danger of losing the feeling for the clinical topics and losing the prime audience. This could be a challenge to modern register-based research in the coming years.

Over the past 50 years, register research has developed from the local, pioneering clinically based register, established after World War II until now, where we have the possibility of handling enormous datasets in super-registers. The main trend in how these opportunities should be used in the future should not be left to single groups, register researchers and epidemiologists, health trust administrators, politicians, or any other group. We conclude in strong support of Wierdsma et al. [11] when they call for ‘join[ing of] the forces’ and suggest a renewal of the WHO initiative which organized a workgroup on psychiatric case register in Mannheim almost 30 years ago in 1983.


This paper was published in the autumn 2013 as chapter entitled “Psychiatric case registers: their use in the era of global mental health” by Povl Munk-Jørgensen and Niels Okkels in the book “Improving mental health care. The global challenge” [57]. The manuscript for the chapter was assessed and edited by Mirella Ruggeri, Graham Thornicroft and David Goldberg. At the time of the assessment and publication of the book, there were no thoughts and plans about publishing it in the Acta Psychiatrica Scandinavica, that is, the three assessors/editors were not biased in relation to a later publication in the Acta Psychiatrica Scandinavica with editor as first author and themselves as coauthors.

Declaration of interest