Scandinavian Statistics and the Central Population Registry in Denmark

While introduced in the twentieth century, central population registries have built on existing traditions and routines of recording the population. Before national-level statistics of the Scandinavian countries, the Lutheran church had played a key role in producing population statistics.Footnote ⁷ The first nominative censuses of Denmark-Norway date back to 1769 and 1787 and in 1797 the Danish-Norwegian tabel-kontor was opened in Copenhagen (Danmarks Statistik 2011; Johansen Reference Johansen2002). For the case of Sweden, Lisbet Koerner (Reference Koerner1999) has described the significance of cameralist sciences and their role regarding the economic autarky of the nation state.Footnote ⁸

In Denmark population records were kept at municipal levels and censuses conducted at regular intervals. In the twentieth century two moments have shaped population records in Denmark: the law on centralization of the population registries in 1924 and the introduction of permanent registration numbers for each resident in 1968. During the food shortages after World War I, organizing and overseeing food provisions (rationeringsordning) had been assigned to the Bureau of Statistics in 1917. Upon developing the CPR-system in the 1960s, system developers stressed the significance of the ration cards register. In their view, this infrastructure “stimulated the action to obtain an improved recording system. . . . In 1918, after the end of the war, the capital – Copenhagen – went on updating these ration card registers on their own initiative because they recognized the aid of having an identification register in performing their daily administrative work” (Nielsen Reference Nielsen and Kasher1967, 253).

With the 1924 act and the centralization of municipal registers, each municipality had to appoint a “public register manager.”Footnote ⁹ Supervision by the Bureau of Statistics (since 1966 Danmarks Statistik) granted these local register managers a certain degree of independence from other authorities.Footnote ¹⁰ This relative autonomy of registry managers from their municipalities and the central coordination made possible standardized procedures. The records of the population moved from parish lists and vital statistics tables, to a system of individual paper cards that could flexibly be sorted. In turn, this allowed aggregate accounts at the central level and stratification of the population data for selected age groups, for example to quickly generate lists for schooling, recruitment to the army, pensions, and welfare recipients. After World War II, punch-card technologies widely introduced in the public services changed the registration system. With punch-card technology, this system kept one punch-card per resident, including items on family-relatedness and connecting the individual citizen to the central register. Thus technologies of data processing co-shaped record-keeping of twentieth-century national populations and their governance.Footnote ¹¹

The development of information technologies and electronic data processing also drove the reform of the 1924 law that was prepared during the late 1960s. The Danish Ministry of Interior set up a commission to examine possibilities for electronic data processing and an account number (kontonummer) for each citizen in connection with change in taxation procedures (Willumsen Reference Willumsen1999). The number was conceptualized as means for employers to report salaries and directly deduct taxes at income (kildeskatten). The approach was to first work out technical details of a personal identification number and to postpone the political issues until later. The committee in 1963 recommended an entirely centralized national register with a 10-digit permanent personal number and a central organ to coordinate the reform. The government largely followed this proposalFootnote ¹² but installed, instead of a “central organ,” a preliminary committee with representatives from the Ministry of the Interior (Datacentralen af 1959, and Kommunernes hulkortcentralen). From 1965, this committee named “Sekretariat for personregistrering” was in charge of working out a system of personal identification numbers. For instance, they formulated “minimum technical requirements,” which included that each number was unambiguously assigned to one person, that it was permanent, and that it should be applied in as many business and public services contacts as possible.Footnote ¹³ The Sekretariat also consulted with the population registry in Sweden, the first among the three Scandinavian countries to introduce a personal number in 1947, shifting to permanent national personal numbers in 1967. Also the Norwegian parliament had accepted a proposal to implement such a register without debate in 1964 (Lie and Roll-Hansen Reference Lie and Roll-Hansen2001). Different from Denmark, the demand for personal numbers in Norway didn't first come from the Statistics Bureau, but from professional associations (næringsorganisationer) and banks.

The political issues postponed earlier were addressed by drafting a CPR-law tailored to the technology in 1965. The parliamentary hearings treated the issue as purely a routine matter of technical modernization. Most speakers referred to the plan as an administrative need, recommending that one should “go with the technology” and conceiving of electronic data processing as something “belonging to our modern times.”Footnote ¹⁴ Perhaps surprisingly, there was little concern or debate over this state measure. An exception to the general tone of the debate were the remarks by Member of Parliament Lindegaard Rasmussen who cautioned that “we may actually be facing the introduction of a kind of revolution in the relation between the individual citizen and the public authorities”Footnote ¹⁵ (Folketingstidende 1968). Danish Parliament passed the act on the CPR-law in fall 1967.

Automating the Registries: Storage and Flow

While there are substantial differences in its usage and significance in everyday life, the maintenance of central population registries with personal numbers is a shared feature in the Nordic countries.Footnote ¹⁶ In Denmark, the development of the CPR-system took place with an administrative reform that updated the 1924 law and followed similar trends in Sweden and Norway. Significantly, information technologies and electronic data processing promised progress, efficiency, and optimization of techno-bureaucracies and the state, which provided a new field for emerging professions of systems developers, automation engineers, and public administrators. The idea of mechanization has long shaped the approaches to organization and management. Data processing also belongs to the bureaucratic work in the sense of Max Weber in that it is the central place of distribution of resources (Weber Reference Weber1922).Footnote ¹⁷ It is the continual work done in offices by messengers, clerks, paper work, registration, filing, sorting, etc. that gives bureaucracy its power (Weber Reference Weber1922, 825). If this is the case, in what ways did the transition from file-based to electronic information processing and automation also remake the science-society-citizen configuration?

In this context, a closer look at an international expert meeting on automation of population registries held in Jerusalem in 1967Footnote ¹⁸ is instructive and gives us a window into engineers’ and administrators’ agendas and their debates over automation and population registries: At the conference, system developers discussed technological developments as part of “new management sciences” which they expected to yield the “most efficient solution for the organization of data flow, storage, and processing” (Chevion Reference Chevion and Kasher1967, 534). Experts envisioned that automation would reshape registration practice in all aspects, including manpower, management approaches, technologies, and safeguards for privacy. As conference host Dov Chevion further pointed out, “the population register system may serve as the best laboratory for the actual work of all these parameters” (ibid., 538). System engineers saw the population registries as an avant-garde laboratory for optimization of administration and governance on a larger scale. Significantly, Scandinavian registry developers were represented at almost all panels of the meeting.Footnote ¹⁹

Representatives from the Danish Sekretariat for person-registering also participated and reported on developments of the CPR-system at the conference. Johs Nielsen, head of the Sekretariat, outlined what he called tomorrow's “total EDB-System” (Nielsen Reference Nielsen and Kasher1967, 262), i.e. the technological infrastructures of the Danish CPR system. Danish presenters stressed possibilities to optimize bureaucratic routines through “further development of CPR partly as a catalyst for mechanization of other administration branches.” For the system engineers, this was about accelerating the flows of data and increasing the range of applications “according to the municipality's own wishes” (ibid., 256):

System developers envisioned that the population registry should become a catalyst of automation in much more general terms. In facilitating such development, their reports centered on three main themes: securing the correct identification of data belonging to a person, ways to deal with errors in data-entry (which they preferred to be machine-based, not manual), and coverage of hitherto non-represented groups such as seafarers, non-registered people, people who had disappeared, commuters from neighboring countries, and immigrants.

An initial purpose in bureaucratic procedures of the usages of the CPR-number was to establish the identity of a person in business relations. As developers put it, the personal number would function as an “information filter,” as a device to regulate data traffic by channeling or interrupting data streams via filters that either allow data to pass or are withheld.Footnote ²⁰ In many situations, system engineers treated errors that occurred as something to be controlled via algorithms rather than avoided at all costs. They distinguished between different kinds of errors (punching error, wrong information) and devised ways to manage them by check-digits (e.g. modulus-11, an algorithm that constructs a check digit based on the other digits in the number). They conducted cost-benefit assessments weighing the costs caused by errors against the costs of technically preventing these errors. Thus, they presumed that large numbers would ultimately control and make up for errors, even when it was impossible to avoid those.

The Danish permanent, birthdate-based personal numbers were implemented as public service infrastructure. The 10-digit-CPR contained birth dates (6-digits, which was similar to the system Sweden had introduced in 1947) and a 4-digit running number with the last digit indicating gender (odd numbers for men, even numbers for women). A core bureaucratic goal in setting up the system was the complete registration of all residents in the territories of Denmark. Consequently, the developers strived to include those not registered or difficult to register, like seafarers, or commuters. The CPR-infrastructure was set up to also include all territories under Danish administration. As of 1972 the registries were reorganized to comply with the CPR-system (Nielsen Reference Nielsen1991), covering Denmark, Greenland, and, with the Nordic agreements, the Faroe Islands that had been exempt from the Danish register system before.

Interlinked central and local databases required a temporal management of updates (ajourføring), the transfer of corrected data, and regular synchronization between municipality and the central registry. Data traffic needed to be managed and the two-way-communication between branches and central archive had to be coordinated in order to control when a defined information status was frozen and transmitted and when incoming data would be added. These procedures needed to be integrated into a workable scheme.Footnote ²¹ Thus, more than merely identifying, observing, and sorting, developers worked on managing circuits, flows, and communication loops in order to design manageable applications and optimize the system at those nodes of identity check, error management, and completeness.

These practices made the CPR system function as a testing ground for novel techniques, administrative efficiency, and technology-mediated governance. Regulatory circuits were used to draft causal chains for various domains of government, administration, and business; notions of “systems” and “circuits of communication” were ubiquitous. In a veritable “databank optimism,” as Lie and Roll-Hansen (Reference Lie and Roll-Hansen2001) put it for the case of Norway, systems of automation and rational infrastructures were designed for all domains of society. In that sense, circuits of optimization came to extend throughout administration from these core bureaucratic practices.

The Danish Laboratory of Population Registries

Mechanisms to assess accuracy, to achieve completeness and manage error were at the core of the automated database. As this expanding archive was kept for the entire resident population, the database came to replace the fieldwork for censuses that in Denmark had been carried out at 5–10 year intervals. Conducting a census now shifted from carrying out surveys with questionnaires to an issue of how to retrieve those data from the different registries that stored them. Denmark then was the first country to carry out a census based entirely on registers in 1981, literally unnoticed by the population and without approaching a single citizen.Footnote ²² This practice drew international attention, expressed for example in the Journal of Official Statistics of 1986 which featured an article titled “Which Country Will Follow the Scandinavian Lead in Taking a Register-Based Census of Population?” (Redfern Reference Redfern1986).

Technological developments in population registries impacted on the routines in the healthcare sector, too. For instance, the Central Cytogenetic Register, founded in 1968 became immediately linked to the CPR-system and worked with magnetic tape and disks at the same time. While this registry used magnetic tape for data storage, disk space was chosen for short-term applications (Videbech and Nielsen Reference Videbech and Nielsen1979). The mere technical possibilities of a combinable central infrastructure prompted more visions of future integration and applications by the developers.

Health scientists also mobilized the long history of record keeping in Scandinavia as a resource for knowledge production. The academic discipline to mine these data was “modern epidemiology,” defined as the “study of distributions and determinants of health and disease in populations and its applications to the improvement of health in the population and population subgroups” (Last Reference Last1983), a definition also adopted by the World Health Organization (WHO). This mode of epidemiological research evolved with comprehensive population studies such as the Framingham study, an early inventory of medical and lifestyle data in the US town of Framingham, MA. It was in this context that the concept of “risk factors” arose, which became increasingly used for a whole range of variables identified through statistical comparisons as associated with disease at the population level.Footnote ²³ For example, hypertension, cholesterol level, or alcohol consumption became explanatory variables – “risk factors” – for coronary heart disease. Epidemiological practices translated these variables into a statistical equation, which modeled constellations of risk factors as influences on disease frequencies at the population level. Thus, epidemiology not only had moved from the study of (infectious) agents to chronic disease, but also to a statistical modeling practice that predicted health and disease as “outcomes,” i.e. as dependent on a set of risk factors now in the focus of preventive medicine. Moreover, with its population-level reasoning, epidemiology became a quantitative approach that also was compatible with economics, psychology, and the quantitative social sciences, not least due to parallel developments in their methodologies and the growing capacities of information technology.

Given this internationally emerging mode of health research, populational concepts of health and disease, approaches of statistical optimization and tools of logistic regressionFootnote ²⁴ brought about new research opportunities given the Scandinavian population data infrastructures. Asger Meulengracht and Mette Madsen from the Danish Institute for Clinical EpidemiologyFootnote ²⁵ called for the broad usage of the Danish registry landscape, the then more than 60 health registries and 15 administrative registries:

Health scientists saw a “particular obligation” in the registry data for medical research and healthcare planning. They advocated research as an ethical “duty” to an international community of health scientists and international health organizations. This practice of taking advantage of the bureaucratic infrastructure served to produce the “facts” on “population health” sought after by policymakers, social, and public health planners. Meulengracht and Madsen (Reference Meulengracht and Madsen1982) also listed the first registries such as the Criminological Registry operated by the police and the Central Psychiatric Registry set up in 1938.Footnote ²⁷ Established in 1954, the Twin Registry succeeded in tracking down and covering birth cohorts since 1870 (Kyvik et al. Reference Kyvik, Christensen, Skytthe, Harvald and Holm1996). New registries initiated in the 1970s, for instance, The National Danish Hospital Registry, stored hospital admission records and were extended to include outpatient and emergency data in 1995 (Andersen et al. Reference Andersen, Madsen, Jorgensen, Mellemkjaer and Olsen1999). Danish epidemiologists viewed these population registries as instrumental in generating internationally relevant research output (Mortensen Reference Mortensen2004, 122). The Danish Cancer RegistryFootnote ²⁸ that Johannes Clemmesen, inspired by the local registries in the US he had visited, had established in 1943 was the template for the nationwide approach taken in Denmark. The cancer registry was the first database of that kind at the national level, setting an internationally recognized standard for cancer registration.

Using registries, this mode of epidemiological research tracked individuals and their “status changes.” In his account of differences in epidemiological research cultures in Europe, Luc Berlivet recalled a remark by an epidemiology lecturer on a series of Danish research papers at the London School of Hygiene and Tropical Medicine: “In Denmark you take a seat on a chair in a public place and immediately, there are a lot of sensors that measure a lot of things and the data go directly into a central computer” (Berlivet Reference Berlivet, Jorland, Opinel and Weisz2005, 337). This obvious seminar joke is telling about the data hunger and amazement, and at times disapproval, among researchers of this degree of population surveillance.

The CPR-system made it possible to follow citizens through their life and to link up data at the individual level from different sources and abstract them into aggregate formats at the population level. The administrative routines of the healthcare system produce an invisible, ever growing archive of data.Footnote ²⁹ Epidemiologists link individual records to the population registry, aggregate data in epidemiological study designs and compute population-level risk estimates. These practices bring about “population health” as an epistemic object.Footnote ³⁰ Scandinavian administrative registries were increasingly used as unique resources for entire disciplines and research fields – demography, quantitative social science, marketing, policy research, and public health. As a unique and permanent identifier, the CPR-system became the cornerstone of record linkage between otherwise disparate data.

Automation of population registries, data spreading, and recombination for secondary data usage resonated with both the dreams of welfare state engineers and epidemiologists as to the optimization of databases and health surveillance of the population, respectively. While computer scientists and engineers played a major role in the modernization and development of registries, making epidemiological studies work also required expertise in statistics and modeling and therefore statisticians and mathematicians were recruited, as well as technicians to their teams.

In particular with pharmaco-epidemiologyFootnote ³¹ and pharmaco-genomics as an important focus of biomedical research, scientific journals as well as popular science accounts hailed the “Nordic laboratory” as “goldmines” for health research:Footnote ³² The New York Times reported a “Gold Mine of Psychiatric Data Discovered in Danish Registry!” (Coleman Reference Coleman1994). While attention came mainly from US media and institutions, Danish researchers and journalists actively promoted the capacities of those infrastructures. A Science feature story was entitled “When the Entire Country Is a Cohort” (Frank Reference Frank2000) with regard to the National Birth Cohort set up in the late 1990s (ibid. and Frank Reference Frank2003). The epidemiological term “cohort studies” denotes an analytical study design (also referred to as follow-up study or prospective study) for the comparison over time of disease rates in groups with different exposure levels.Footnote ³³ The cohort approach itself was traced back to a study on tuberculosis conducted by Norwegian physician K. F. Andword, who in 1930 used the terms “generational study” and “population frame.” The Danish cohort studies that covered the whole country built directly on the central population registry and were promoted as “the epidemiologist's dream” (Frank Reference Frank2003). In the National Birth Cohort, however, the general practitioners supposed to recruit for the study did question the study's benefits to participants and clinical practice (Olsen et al. Reference Olsen, Mellemkjaer and Friis2004). That general practitioners challenged the research agendas and the top-down approach and did not recruit participants for the National Birth Cohort, considerably delayed the project.

Record Linkage and Secondary Usage: Epidemiology as Quasi-Experimentation

How did epidemiologists use administrative data infrastructures for research into causes of disease? Epidemiologists conduct statistical comparisons between disease measures in population subgroups through a set of research designs, e.g. in cohort studies, and often without formal enrollment, just by using routine data generated by the public services. The researchers themselves described their practices as taking advantage of data sets and designs “that simulate experiments” (Vineis Reference Vineis and Morabia2005, 349), a key approach in observational epidemiology. In this mode of research, any differences between populations become empirical settings that can be evaluated like an experimental setting, comparing the disease response in the different population groups. Textbooks exemplify such “quasi-experimentation” with John Snow's study on cholera and water supplies in nineteenth-century London. In this “grand experiment,” the story goes, the epidemiologist is “taking advantage of situations that constitute a natural experiment” (Roberts Reference Roberts1978, 187). How does this approach of analyzing “natural experiments” (also referred to as “quasi-experiment” by epidemiologists) play out in specific settings and how did epidemiologists implement their research designs, mobilize and recombine data from administrative registries for public health research?

For the case of Denmark, the usage of existing data to evaluate “natural experiments” spanned a nearly unlimited range of topics and sources. For example, researchers used company data from all bus and tramway drivers employed by the Copenhagen Traffic Company from 1900 to 1994. For each employee, they estimated the exposure to air pollution taking into account the different time periods of industrialization and traffic development over the twentieth century. Via the Cancer Registry they tracked all cancer cases for this particular group through the CPR-number. Drawing on these exposure and disease data they calculated estimates of cancer risks in relation to air pollution (Soll-Johanning et al. Reference Soll-Johanning, Bach and Jensen2003). Thus, via record linkage from different sources – the company records, the Cancer Registry, and traffic data – distributed information was transformed into a data set that could be analyzed as an experiment. Via record linkage researchers transposed data collected for different reasons into a virtual “quasi-experiment.” While, for reasons of research ethics, actual experiments of these kinds cannot be conducted in human populations, existing data infrastructures are transformed into a platform for simulated experiments. In this sense, differential exposures were used to test hypotheses, for instance associations between air pollution and cancer mortality, based on an empirical data set. The observed study population “quasi-experimented on” can be an epidemiological cohort defined for instance by occupation, as tracked through population registries. One important direction of registry research was occupational epidemiology that focused on disease outcomes using mortality data to compare rates between groups with different professions and thus different exposures. Further research directions included health risk assessments of novel technologies, such as etiologic studies of cancer, a range of clinical and therapeutic practices that were also evaluated using epidemiological designs and techniques. In this way, Denmark's population registries, their completeness and favorable access conditions became a laboratory for population research at comparatively low costs. International agencies and researchers from other countries, for instance from US health authorities, were eager to collaborate with Danish epidemiologists precisely for their data wealth and favorable research conditions. The legal situation was such that, if affiliated with a Danish institution, researchers abroad were able to obtain access to registry data.Footnote ³⁴

In the mid-1990s for example, Danish and US epidemiologists carried out a collaborative investigation into whether cellular phones usage increased cancer risks (Johansen et al. Reference Johansen, Boice, McLaughlin and Olsen2001). Carried out as a nationwide study in Denmark, the study was aimed to address debates over the health effects of electromagnetic exposure with joint funding by Danish and US foundations and the two companies. To utilize distributed existing data for such a “quasi-experiment,” researchers retrieved a list of names that were registered with a cellular phone contract in Denmark between 1982 (when cell phones had become available) and 1995, using company data from the then two cell-phone companies TeleDanmarkMobil and Sonofon. Business contracts were excluded, because the actual users were not known. Then the CPR-numbers of individual users listed in the company's records were linked to the data of the Cancer Registry to find out whether cell phone contract holders had been diagnosed with cancer. To estimate the relative increase in risk in this retrospective cohort study, incidence rates were compared to “background” average incidence rates by age groups for the whole of Denmark. It was the infrastructure of CPR-numbers that made possible the investigation. For most of such register studies, researchers did not need to ask for individual permission. In this case, the possibility to “opt-out” of the study was granted later by the companies. In the article, published in the US Journal of the National Cancer Institute, the authors stated:

Epidemiologists interpreted the low number who opted out – 53 out of 420,095 – as proof of the trust and the general support for population health investigations in Denmark. Researchers not only depended on trust building as part of and within their epidemiological studies, they also took advantage of both the registry data and a certain trust in science and the public services. This made possible high participation rates compared to US large-scale studies that involved interviewing and medical examinations. Epidemiological surveillance spanned the whole range of quiet mode of state governance to practices that enrolled and later actively engaged communities and participant-citizens who, as individuals did not opt out but wanted to contribute to health research (Lind et al. Reference Lind, Mose and Knudsen2007). Epidemiological risk assessments became constitutive of subject formation in terms of more active contributors to and consumers of statistical evidence. These features also became characteristic of a specific science-society relationship associated with Scandinavian public health.

Like occupational studies and technology assessments, clinical practices became subject to epidemiological scrutiny, using routine reporting and documentation in the health care system. One such transformation of registry data into a retrospective quasi-experiment in society carried out on the Danish and Swedish population was about safety concerns in blood transfusion practices. Here the logistics of organizing transfusions at the same time provided data for epidemiological monitoring through a joint registry. The blood transfusion registry that was central to this investigation was a joint registry, the Scandinavian Donations and Transfusions database (SCANDAT), a combined Danish and Swedish registry that drew together data from computerized blood bank registers and hospital registries from Denmark and Sweden. Safety issues were raised as to blood donors with cancer, in particular with transmissible agents playing some role in cancer etiology (Edgren et al. Reference Edgren, Hjalgrim, Tran, Rostgaard, Shanwell, Titlestad, Jakobsson, Gridley, Wideroff, Jersild, Adami, Melbye, Reilly and Nyren2006). US epidemiologists used the long-term population data in the Scandinavian registry to empirically assess and quantify safety issues. Starting from the records in the binational register, researchers retrieved blood donation and transfusion histories in order to then add data on diagnoses of blood donors and recipients. Retrospectively tracing all recipients of blood transfusion between 1968 and 2002 resulted in a data set of more than a million donors and a million recipients (ibid.). For a restricted subset of these data, record linkage to the Danish and Swedish cancer registries identified tumor diagnoses among the blood donors within five years after donation and also among the recipients. The next step was to calculate cancer rates for both groups, i.e. among recipients of blood from precancerous and non-precancerous donors, respectively, and compare those rates. As these computations detected no significantly increased rates, the conclusion was that no new regulations regarding blood donation were needed (Edgren et al. Reference Edgren, Hjalgrim, Reilly, Tran, Rostgaard, Shanwell, Titlestad, Adami, Wikman, Jersild, Gridley, Wideroff, Nyren and Melbye2007). Only if there had been evidence at a certain level of statistical significance, would this be seen as justifying changes in transfusion practice and regulatory policies.

This rationality of risk management worked in a mode of continual optimization, in particular as to limiting resources spent on regulatory intervention. Epidemiological results soon were used to justify resource allocation in the healthcare systems. Registry research is embedded in a quantitative rationality of quality management of both public health and clinical procedures. Whether it is bus drivers, cell phone users, or transfusion recipients, secondary usage of registry data, at the epistemological level, mimicked controlled experimental configurations. Yet, this mode of observational research in populations was different from actual practice in the controlled spaces of biomedical laboratories. Observational epidemiology deals with “natural experiments” in retrospect; through data retrieval from registers, epidemiological practice formalizes social conditions into exposure differences, which can include socio-economic disparities, life habits, environmental exposures. These techniques mimic other “experimental research designs,” such as clinical trials or controlled intervention trials, linked to institutional practices and policies. Significantly, Trudy Dehue (Reference Dehue2001) has linked the randomized controlled trial less to laboratory practices, but more to social administration. In the context of the late twentieth-century liberal variant of welfare states – in this case the Netherlands – randomized controlled designs also “epitomize liberal values of efficiency and impersonality” (Dehue Reference Dehue2001, 283). In contrast to clinical trials that require an intervention, natural experiments evaluate differential exposures or conditions in the past – as observation of ongoing experiments.

In observational epidemiology, routine administrative databases can be understood as technical infrastructures that quasi-experimenters take beyond the original context in which these databases had been developed. Can quasi-experiments be compared to laboratory experiments in terms of their generative features? To what extent do quasi-experiments take on dynamics similar to the “experimental systems” described by Rheinberger (Reference Rheinberger1997)? Albeit different in terms of the highly inert and less flexible data generation in long-term population statistics, there are some similarities to experimental systems in the lab. Once up and running, population registries provide the empirical base to test hypotheses with empirical data. These are specific contingent constellations that recombine in many different ways, opening up certain (while foreclosing other) “epistemic spaces” (Müller-Wille and Rheinberger Reference Müller-Wille, Rheinberger, Müller-Wille and Rheinberger2007). Drawing on Rheinberger's notion of experimental systems to analyze the routine storage of digital patient data as a future-generating device, Jensen (Reference Jensen2005) has described the generative effects of flexibly recombinable electronic patient records. While there are structural similarities between the principles of statistical testing and in the usage of population databases for hypothesis generation, there are also important differences in the set-up and materiality of large-scale epidemiological data sets, if compared to bench-work in the lab. Where the experimenter is groping with his or her experimental system's resistances, epidemiological experimentation is less mobile and flexible, as it works with inert data formats that are entrenched in long-term institutional practices and their often less flexible classifications.

Registry-based epidemiology takes advantage of a past that is conceived of as “natural experiment” in order to generate projections of the future. Without being aware of their contributions, residents registered in the CPR-system have generated a populational datascape, from which epidemiologists draw up data for population health monitoring and health research. This was by no means a smooth process nor was it determined by technology or registries as such. Yet, registries and information technology afforded specific state-citizen configurations and therefore were an important part in the constitution of registry-based epidemiology as “Scandinavian.” These constellations became infrastructural though fundamental to the rationality within and beyond Scandinavian health care policies.Footnote ³⁵ The continuous data storage and linkage capacity of the CPR-system has become the condition of possibility for public health as social optimization. This mode of knowledge production became a taken-for-granted rationality adopted throughout biomedical research, clinical trials, health services, evidence-based medicine, and public health. The epistemic spaces and regimes of population thinking emerged as a “Scandinavian” mode of efficient organization of society. With the reorientation away from the welfare state during the 1980s and with the emerging genomic economies, the actual usages of these data infrastructures shifted.

Infrastructuring Biomedicine: Populations and Individuals

With the internationalization of epidemiological research in the late twentieth century, the Danish registry infrastructure has gained surplus value. In particular, the emerging genomics data and expectations as to the explanatory power of genetic sequence data prompted researchers to collect DNA samples as part of epidemiological projects. Later, this included the implementation of a host of bio-bank projects and efforts to coordinate bio-banks that store human tissue samples. While the significance of DNA sequence data remained largely unquestioned during the early stages of the human genome project, it soon became clear that neither DNA sequence alone nor consideration of the entire genome would lead to breakthroughs in health research. Researchers advocated the study of gene-lifestyle and gene-environment interaction. This led to new research directions in genetic and genomic epidemiology (Khoury and Little Reference Khoury and Little2000), a situation in which “Scandinavian epidemiology” was viewed as particularly well equipped because of the registries. Whereas in most countries practitioners of epidemiological studies needed to collect variables from questionnaires and obtain participants’ consent to record linkage, Danish legislation allowed epidemiologists working in publicly funded institutions to access a host of individual data through central registries. Epidemiologists have advocated usage of public data records for research for a long time, and collaborated and helped develop quality standards and conditions of access that allow data usage in epidemiological research.Footnote ³⁶ The CPR-number made each individual member of the population traceable across a lifetime – “from the cradle to the grave,” as Danish epidemiologist Jørgen H. Olsen put it (Olsen et al. Reference Olsen, Mellemkjaer and Friis2004, 1459).

Existing in all Scandinavian countries, prescription registries brought about a mode of registry-based research that was less cumbersome than in countries without data retrieval opportunities.Footnote ³⁷ Since this meant greater and cheaper research opportunities, pharmaceutical corporations were attracted, too. Different from the CPR-numbers themselves, the Danish prescriptions registry was an exception given the controversies that came up during the hearings that preceded its introduction in 1993 (Hallas Reference Hallas2001). Major concerns included the then unprecedented size of a database that would record every single prescription as well as state surveillance of prescription practices and possibilities of data misuse. After heated debates, the Ministry of Health decided for an irreversible encryption of the identifiers, so that records here could not be linked as universally as in other Danish registries.

When EU directives on data protection in 1995 put registration practice under restriction, it was the Danish side that intervened and successfully pushed for national legislations on research registries (ibid.). Registry research was exempt from review by ethics boards and overseen by the Danish Data Protection Agency since 1996 (ibid.). Large scale projects at the national level include the Danish National Birth Cohort, according to Science, the “epidemiologist's dream” (Frank Reference Frank2003), in particular when comprehensive bio-banks and databases are envisioned. In a policy response to address the “stagnation of Danish life expectancy” (Danish Ministry of Health 1994) in comparison with the other Scandinavian countries, the Danish National Board of Health began combining some of the social and health registers into the Prevention Registry that combined eight registries under the Danish Ministry of Health and Ministry of Social Affairs. This new data hub was a “super-registry” that was created to ease data linkage between registries of the National Board of Health, such as the National Patient Registry, and the “lifestyle data” held at Danmarks Statistik (Roed et al. Reference Roed, Juhl and Kamper-Jorgensen1999, 269). With analytical techniques established in the 1980s and medical research standards that ranked research designs from randomized clinical trials down to regional or group comparisons, epidemiologists needed data traceable to individuals. The Scandinavian registry systems provided this type of data, a fact that led to many bilateral studies such as those with the US reported above but also studies conducted as EU collaborations, where Denmark and Sweden faced advantages due to their population registries. At the same time, many large-scale studies were designed as multicenter studies organized in EU-consortia, and then were further networked with studies conducted in North America and Asia into global consortia that engage in pooled analyses.Footnote ³⁸ Based on their registries, Scandinavian researchers complete such exercises of global data pooling with less cost and effort than most other countries.

Genomic technologies such as gene chips speeded up testing procedures by measuring many genetic markers at a time. The data they generated became variables of an emerging “genomic epidemiology” of “gene-lifestyle interactions.” For this reason, linkage capacities gained even more prominence and, in the age of genomics, bioscientists more explicitly stressed “excellent conditions” provided by Danish registries (Loiborg Reference Loiborg2002; Melbye Reference Melbye2001; Olsen et al. Reference Olsen, Mellemkjaer and Friis2004; Mortensen Reference Mortensen2004). The hope for breakthroughs in genetic causation of disease animated practices of record linkage in the promissory culture of late 1990s genomics and, subsequently, genomic epidemiology. Moreover with the 2000 “Act on Processing Personal Data,” the private sector (under certain conditions including research in the public interest, which as a rule includes public health research) has been granted more access to the central registration system CPR-information under the same conditions. A highly centralized database concentrating information, the CPR system is at the same time a multi-purpose database that was supposed to flexibly afford different functions for government, administration, corporate, and transnational biotech knowledge production, and global market interests. In clinical and pharmaceutical research, the Danish registry data would make up “control populations.” Even though data use is regulated by data protection laws, the CPR-resources can under certain conditions be opened to a range of consortia – from public health and biomedical research, and marketing, to consumer-directed information.

While the primary purpose of the CPR-numbers was the new tax system, the CPR system evolved into a yet unprecedented research platform with proliferating usage in the age of genomics. Pharmacogenomics that examines individual susceptibility to pharmaceuticals, the electronic patient record as a data recording device, and large biomedical consortia both expanded and individualized the archive and its usages. In concert with the envisioned personalized precision of epidemiological results, public health policies were redesigned. Strikingly, these techniques of individualizing did function through group categories, enacting chains of subclassifications as individual profiles. Targeting individuals in this personalized way speaks to individuals in a different way, enrolling citizens into more proactive practices, such as managing health by attending to one's risk profile. One example is the design and usage of risk scores that estimates one's risk of lifestyle-related disease; this was one spin-off from epidemiological cohort studies.Footnote ³⁹ In this mode, public health messages to individuals about the management of risk factors produce a kind of interpellation (Althusser Reference Althusser1971) of its citizens. Public health governance operates at both a collective level of optimization, located at the evaluation of “quasi-experiments” in the population on the one hand and the individual-level preventive optimization of risk profiles as “technologies of the self” (Foucault Reference Foucault, Martin, Gutman and Hutton1988) on the other. These two modes are mutually entangled as it is the population-level that became the basis of remodeling individual futures. This exceeds the feedback loops of “reflexive modernity” (Beck et al. Reference Beck, Giddens and Lash1996), in that what takes place here is an active making and remaking of both the population and individual subjectivities through the epidemiological risk computation.Footnote ⁴⁰

As Deleuze (Reference Deleuze1992) noted with regard to the relationship between individual and population in Post-script on Societies of Control: “We no longer find ourselves with the mass/individual pair. Individuals have become ‘dividuals’, and masses, samples, data, markets or ‘banks’” (Deleuze Reference Deleuze1992, 5). Individual and population can no longer be conceived as opposites, as even in the often envisioned individualization and personalization of diagnosis and treatment, this “dividual body” is reassembled and enacted through statistical strata and distributed bodies stored in bio-banks.

Infrastructure itself, as Star has reminded us, never develops de novo; instead, it is “built on an installed base; it wrestles with the inertia of what is already there and inherits strengths and limitations of that base” (Star Reference Star1999, 382). With the growing secondary usages of data infrastructures, cascades of record linkage, hard- and software capacities in data transfer, have transformed epidemiology into big science, often conducted in multinational consortia (Hoover Reference Hoover2007). For their research questions, researchers mobilized older collections of data and samples to take part in a speculative genomic “economy of promises” (Fortun Reference Fortun2005). Combined with epidemiologic study designs the CPR-registry has developed from a mainly bureaucratic machine into a genomics-based prediction-oriented future machine. This generative machine also co-shapes and permeates individual and collective thinking, planning, and acting. Epidemiologists have discussed possibilities to optimize population-based research in Denmark and criticized regulatory conditions that impede epidemiological studies, for instance the data encryption of the Prescriptions Registry by Danmarks Statistics that, unlike other registries, was only made accessible from the registry's institutional servers (Johannesdottir et al. Reference Johannesdottir, Horváth-Puhó, Ehrenstein, Schmidt, Pedersen and Sørensen2012, 304).Footnote ⁴¹