CHANGING IRREVERSIBLE NETWORKS

Ole Hanseth1,3 and Eric Monteiro1,2

1Department of Informatics, University of Oslo

2Department of Information and Computer Science, Norwegian University for Technology and Science

3Department of Informatics, University of Gøteborg

E-mail: Ole.Hanseth@ifi.uio.no, Eric.Monteiro@ifi.ntnu.no

ABSTRACT

This paper focuses on the importance of institutional aspects of work practice and the infrastructural aspects of technology through a study of patient record systems in hospitals. Patient records are anything but isolated artefacts. They are essential for the coordination and cooperation of a large collection of actors, both within and outside the hospital. They are accordingly involved in large, complex and highly entrenched actor-networks. Due to the interdependencies of the elements such networks are difficult to change. This fact explains the failure of many change projects. We propose some tools and strategies to ease change. The most important ones are gateways.

1 INTRODUCTION

This paper focuses on the importance of institutional aspects of work practice and the infrastructural aspects of technology through a study of patient record systems in hospitals. Patient records are anything but isolated artefacts. They are essential for the coordination and cooperation of a large collection of actors, both within and outside of the hospital. They are accordingly involved in large, complex and highly entrenched actor-networks.

This paper analyses and discusses strategies for changing large networks with a high degree of irreversibility. Patient records are used as an illustration for such a class of networks. There has over the last decades been numerous efforts at developing and introducing hospital wide electronic patient records (EPRs). Within primary health care such systems are commonplace. In Norway, for instance, the vast majority of general practitioners use such systems. In hospitals, however, practical experience with EPRs lag significantly behind the repeatedly voiced expectations. In a study published in 1993, Ernest & Young expected the growth of EPRs in UK would grow from about 20 systems in use in 1993 to 300 in 1996. What happened was that the number declined close to zero (Quinn 1996). To the best of our knowledge, no systematic, comprehensive and in-depth evaluation exist of these efforts despite the substantial amount of resources spent. It is worth noting, however, that these apparently rule-like failures are related to the introduction of hospital wide record systems, hopefully replacing the paper based record. In contrast to this fact, most hospitals contain a growing number of smaller systems close to a local EPR for single or a few wards.

We argue that the numerous failures of projects developing and introducing EPRs are to a large extent due to a lack of appreciation of the infrastructural aspects of such systems and the institutionalised character of the practices into which they have to fit. Gaining insight into development and use of technology and work settings with a high degree of infrastructural aspects grows increasingly relevant and extends well beyond patient record systems.

In section 2 we illustrate the size and complexity of the networks patient records are part of. This is intended to portray the extent and nature of the entrenched and institutionalised practices surrounding patient records which make them irreversible. Section 3 presents what we take to be relevant theoretical frameworks and concepts which conceptualise stability and reproduced routines. Section 4 discusses how institutional and infrastructural perspectives might be combined. In section 5 strategies for changing existing, irreversible networks are presented.

This paper is basically conceptual in nature but makes use of empirical illustrations. Methodologically, it draws on several sources. We are following the current efforts to develop and introduce by Siemens-Nixdorf at the five largest hospitals in Norway. This effort has undergone many phases, project reorganisations and merges over its nearly ten year span. We have conducted unstructured interviews with staff in the project organisation and at the hospital and studied relevant project documentation in addition to tutoring borrowing observations from master and Ph.D. students' field studies at six wards in two hospitals the last two years. We have also made use of case studies of patient record systems from the literature.

2 THE PATIENT RECORD AS INFRASTRUCTURE

The major source of difficulties in designing patient record systems is to appreciate the infrastructural aspects as "[M]edical records do not have to do a single job, they have to be configured in such a way that a set of organizations are served by them. (...) [T]hey have to serve as infrastructure" (Berg and Bowker 1997). We will here identify and illustrate important infrastructural aspects of the patient record.

2.1 Shared by large communities

The patient record used to be the individual physicians' private notebook. Over time, the patient record plays an increasing number of roles, having got a character of being a shared infrastructure for large groups rather than tools for individuals. The record is still, of course, an important tool for the individual physicians but in addition an array of other actors have an interest in it. Within a ward, the medical record mediates and coordinates the work of nurses, secretaries, managers, various medical specialists and other health professions. Within a hospital, the patient records are exchanged and circulated as the patient receives different kinds of treatments. This kind of cooperation and coordination is increasing as a result of the trend from acute to chronic illnesses and a change from isolated illnesses to combinations of several illnesses at the same time - "illness complexes"- as well as an increase in the number of medical specialities (Strauss et al. 1985). In a study at the Norwegian National Hospital it was found that a patient would be treated by up to 30 different specialities and 105 different people for just one single diagnosis. For patients having several diagnoses at the same time the number of people involved can be significantly higher.

The patient record is central also for the communication with actors outside the hospital: to regional and local hospitals, primary care centres and as a basis for various kinds of reporting to governmental health authorities.

The increased focus on patient rights increases the visibility and accountability of patient records. A steady stream of governmental regulations and requirements determine the contents and access to patient records. Quality assurance is increasingly being focused as an important factor for improving patient treatment and care. Quality assurance is tried realised by specifying rules for treatment and care which will ensure intended quality when followed. Such rules include specification of what kind of information that should be included in the patient record. Rules are also defined in order to enable external control of how medical work is performed. Similar rules are specified in order to make judgements about whether patients are treated properly in cases when the patients make claims they are not. A more systematic and comprehensive documentation of the medical work is a key challenge. The health authorities see it as a vehicle for enhancing the quality and equity of health care delivery to patients. Hospitals are expected to keep adequate documentation of their activities in response to possible inquiries, especially regarding alleged malpractice. As a result of increased attention to patient rights, it is reasonable to expect that more patients will exercise their right to gain access to their own patient record. This increased visibility and accountability of the patient records within modern health care generate pressure for elaborate record keeping.

The patient record is also used for management purposes. Treatment carried out and diagnoses made determine the reimbursement from social insurance, and statistics are generated to help making decisions about future allocation of resources. The physicians want to use the patient record as a collection of empirical data for research. All these different ways of using the patient record defines actors being involved in determining how it should be. These includes these "user groups" as well as authorities in charge of controlling the health care sector (The Health Care Directorate) and authorities proposing and deciding on relevant legislation (ministries and Parliament). Insurance companies are lobbying to gain access to information contained in the patient record but have so far been fended off.

2.2 Ecologies of heterogeneous networks

An infrastructure is a layered patchwork of components and associated routines which emerge historically. It is not well-defined in the sense of having a clear-cut boundary towards the "outside". It is negotiable where to draw the line, to separate core characteristics from more marginal functions. It is a boundary object in the sense of Star and Griesemer (1989) as it is fairly, but not completely, grasped object that gets pragmatically negotiated.

There is no widespread agreement about what a patient record is or what should go into it. This is not surprising but typical for the evolving character of infrastructure. It is probably not possible to establish an infrastructure without allowing for a fairly rich set of alternative interpretations.

In the new requirements for patient records in Norway it is stated that "all" clinical work should be documented. This is hardly the case. There is a lot of work that does not get documented, in particular that of the nurses (Bowker, Timmermans and Star 1995).

In some situations the concept denotes the "official" record, stored in the hospitals' (usually) centralised patient record archive. These records contains the information explicitly required to be included in the record, and the information shared among the departments in the hospitals. In addition, most departments keep local records that are important for the individual physicians' decision making and the coordination of patient directed activities within the departments. These records are also used in the communication between the departments and other hospitals. Note that collaboration between departments in the same hospitals usually means collaboration between different medical specialities like cardiologists, radiologists, and surgeons, while collaboration between hospitals more often means collaboration among the same kind of specialists. This is not in line with espoused conception of patient records as presented in the new national recommendations, the requirements for Medakis or the central patient record project at the National Hospital. It illustrates how infrastructural technology evolves historically as an interlocked patchwork of superimposed, complementary or alternative elements.

Several systems share some functions with a patient record system without being classified as one. At one of the surgical wards at the National Hospital they have a special system called Datacor which is close to patient record system, but not quite one. The reason for its popularity is that it allows for detailed classification of surgical operations and associated clinical information which lends itself to aggregated analysis and statistical manipulations. Research is perceived as very important. The ward is the most prestigious surgical ward in Norway and attracts substantial external funding. Datacor has been attempted introduced at other wards without success. The detailed classification related to surgery is perceived as irrelevant and tedious.

At the Children/Heart department at the National Hospital in Norway, for instance, they use two local patient record systems in addition to the official patient record - one paper based and one computerised. All three contain basic patient information like name address, examinations and treatment performed, diagnoses, and controls. The paper based system, called "The Heart Folder," includes in addition various paper forms produced by medical equipment like ECG curves, printouts of ultrasound images of the heart, reports, mails sent and received, and various work documentation. The computerised system, Berte, includes information about medication and a short summary of the patients' history and status.

Every time a child in Norway is born with some heart malfunction for instance, the National Hospital is informed and information about this is registered. This information is immediately registered in Berte. A large number of these children will visit the National Hospital one or more times for examinations or treatment (surgery), and they will be examined by regional and local hospitals and primary care centres more frequently. They will be examined at regular time intervals during their whole lifetime. Berte plays an important role in the follow up of these patients by producing reports about when a child/person should be requested to visit the hospital for a routine control, as well as supporting the communication with other hospitals during telephone requests from a local hospital asking for advice.

An official record is created for a patient the first time she visit the hospital. This implies that Berte not only stores more information about a patient than what is included in the official patient record, it also stores information about a lot of patients for which there is no official record.

Most departments (that is, all we know) keep such local records. And a large part of them includes information about persons not being ordinary patients. At the National Hospital the wards are regularly requested to admit new patients. Whenever they decline, the hospital have to document the information they received and their decision. This is necessary, among other reasons, in case they are sued for denying the admission of a patient (for instance someone injured in a car accident) and the patient dies. The National Hospital has a national responsibility for heart, kidney, and liver transplantations. The departments doing this keep records of possible organ donors. In some cases they collect information about a vast number of potential donors - in one case just written about in the newspapers about one million persons were evaluated as a donor of spinal marrow. In such cases, systems at the local wards are connected to larger international networks collaborating in finding suitable donors and organ transport (Joerges and Braun 1994).

Among the highest ranked priorities when introducing EPRs is to make date more widely available to improve coordination, reduce duplication of work like re- doing tests and examinations, and to establish an improved basis for research. Extending the range of shared data is considered a question about agreeing upon shared definitions of concepts and terms used. However, concepts are more than just labels put on coherent sets of objectively given phenomena. They are rather developed to fit the needs of the working practices where they are used (Star and Griesemer, 1989, Hanseth 1991). An example often mentioned is that "if you ask 10 physicians about what is the limit for high blood pressure you get 10 answers." The reason is, first, that blood pressure may be measured in several different ways using different equipment and technologies. The result obtained may be measured using various types (small and large) of pneumatic instruments giving different results. Further, the results will vary depending on whether it is measures on right or left hand, whether the patient was sitting or lying on bed. Pneumatic technologies give also different results for small and big (fat) patients, etc. Second, the limits for high blood pressure will also vary depending on other attributes of the patient, i.e. the existence of other risk factors contributing to the same problems as high blood pressure like smoking, whether any of her parents have had heart attack, etc. Third, the limit may vary depending on the level of the health services that might be delivered to a person, for instance whether a person will stay in Antarctic for a period or is living in a large city having easy access to all kinds of advanced services, etc.

Lots of information is based on highly subjective judgements made by physicians, for which a standardised vocabulary does not exist. An illustration of this phenomena is a surgeon's description of how the area around the spot where she has taken a biopsy/specimen (for instance in the lungs) looks like, and a pathologist's or technician's description of the surface structure of a specimen prepared for being put under the microscope feel like when moving her thumb across it. Such information is crucial in diagnostic work. According to pathologists at the Radium Hospital in Oslo, it is necessary to know the persons making these observations very well, having collaborated for quite some time to know what they mean by their descriptions. Similar examples is described by Strauss et al (1985).

2.3 Socio-technical networks

The patient record is not an isolated, passive representation of medical work. It is "a tool, and a crucial one at that, aiding memory, communication and so forth -- but it is not a mirror of that work. It does not `represent' the work, but it feeds into it, it structures and transforms it in complex ways" (Berg and Bowker 1997). It is "vowen into the fabric of medical work by providing a structure for ward routines and the smooth flow of work" (Rees 1981, p. 55). It is essential to recognise "how intricately and sensitively reporting procedures are tied to other routinised practices of the clinic. Reporting procedures, their results, and the uses of these results are integral features of the same social orders they describe. Attempts to pluck even single strands can set the whole instrument resonating" and that "bringing record keeping performances under greater and more consistent control, overlooks a central feature and perhaps unalterable feature of patient records as an element of institutionalised practices." (Garfinkel 1967, pp. 192, 197).

This is illustrated by examining how the paper record is interwoven into a surrounding transportation system. Changing from a paper based to an electronic patient record is typically argued to enhance the distribution of the record. This neglects how the current paper based record is well aligned with an elaborate and institutionalised system of transportation with delegated roles.

A tube mail system connects all wards of the hospital and is used for distribution of records. In urgent cases, the record will be carried by a personal courier, that is, a porter, nurse or secretary, who sees to that the record is delivered to the right person. Records are not merely stacked, but are "routed" to one of several pigeon holes according to the level of urgency.

The alleged advantage of an electronic record in eliminating retyping or duplication of information is in a similar way (at least party) compensated by entrenched work routines and special purpose artefacts. Prior to the arrival of planned patients, the paper record is prepared in the central achieve. The archiving personnel supply the record with paper notes with glue containing information about the patient (such as name, social security number, address and relative) which otherwise would have to be retyped on several occasions in connection with the writing of a number of documents (including laboratory requests, transfers and entry in clinical record systems).

The paper version of the patient record has a number of appealing and highly functional features which are difficult to reproduce in an electronic version (Nygren and Henriksen 1992). Reading from screen is substantially slower than reading from paper. Furthermore, the patient record consists of a set of documents in a predefined order. The patient record is a well functioning instrument for physicians' routine work, not the least due to its fine balance between structure and free text. The documents have different shapes and can be marked with different signs and colours in the margin. It is possible to overview several pages at the same time, and to very rapidly browse through a large number of pages. The patient record can in this way be seen as multi-dimensional, coded in terms of shapes, colours and pictures and structured but not strictly formalised. The speed an experienced user can achieve in `zooming-in' the relevant parts is remarkable, and the amount of information covered by a glance is enormous which makes even voluminous records manageable:

* fast overview and understanding of a case (the thickness of the folder, time-scale, distance between two red stamps)

* re-reading, triggering a memory-picture (during telephone contacts, preparation for consultation, "round" on a ward, interpretation of new test results)

* searching for facts (support in questioning a patient, orientation (fixed order) in the record is essential)

* problem solving (tricky cases, test hypotheses, evaluate strategies of action, requires much navigation in the record)

The patient records are small enough to be held in the hand and can easily be moved about (Luff, Heath and Greatbath 1992). As such, their use can be adapted to a range of situations and contingencies. For example, a doctor may drop a patient record card on the corner of the desk so as to be able to glance at it while conducting an examination of a patient seated before him, or the doctor may read a record while facing the patient rather than turning away to read it on the desk. The ecological flexibility of the record contribution to their collaborative use within both the consultation and the practice more generally. Occasionally, doctors may want a patient to look at an item in their paper-based records. In such instances they can place the document between themselves and the patient or hold it up in front of the latter, so as to make it visible to them. They can also carry a document from one room to another in order to discuss it with colleagues.

Bowker and Star (1994) argue convincingly how all kinds of comprehensive and distributed record keeping need to approach the requirement of diligent record keeping pragmatically (ibid., p. 206). It is naive, "unrealistic and counter-productive" to simply accept the absolute need for increased quality of record keeping. Garfinkel (1967) makes a similar point.

This relaxed and pragmatically negotiated quality of patient record keeping needs to be facilitated also with the electronic version. Attempting to use the introduction of an electronic version as a lever for realising the compelling, but unattainable, idea of "perfect" record keeping is ill-adviced. It would work against strong and deep-seated practices. When the National Hospital introduced Medina at one of the wards, the system had to be reconfigured to reproduce the existing practice of more or less formal, but not substantial, signing of records by doctors.

The need to allow for informal note taking, not only full-fledged records, has recently surfaced at the National Hospital. The inclination to do away with such messiness is deep. The vendors of Medina/Medakis reacted to this new requirement with amazement, "Does anybody need such a thing?!" (Project leader 1997).

3 IRREVERSIBILITY

The institutinalised practices around patient records are highly entrenched for reasons elaborated in the previous section. To work out strategies for intervention, which to a lesser or larger extent is inevitable when introducing electronic patient record systems, one needs to grasp the infrastructural aspects of patient records. We briefly outline a few concepts and frameworks which in different ways account for this kind of stability.

3.1 Institutions and institutionalised practice

The study of institutions has a long history within social science. In recent years, however, scholars describe their field as the "new institutionalism" (Powell and DiMaggio, 1992, March and Olsen, 1989). This new institutionalism is based upon social constructivism (Berger and Luckmann 1967) as well as ethnomethodology (see for instance Zucker, and Powell and DiMaggio). We will here briefly present how institutions and institutionalism are conceived within this field and how it can help us understand the use of patient records as sketched above. The key insight is the way stability of institutions are accounted for.

According to Jepperson (1991, p.145), an "institution represents a social order or pattern that has attained a certain state or property; institutionalization denotes the process of such attainment". By order or pattern, Jefferson refer to conventional, standardised interaction sequences. An institution is then a social pattern that reveals a particular reproduction process. When departures from the pattern are counteracted in a regulated fashion, by repetitively activated, socially constructed, controls - that is by some set of rewards and sanctions - a pattern is referred to as institutionalised. Put another way: "institutions are those social patterns that, when chronically reproduced, owe their survival to relatively self-activating social processes. Institutions are socially constructed, routine-reproduced (ceteris paribus), programs or rule systems" (ibid. p 149). Similarly, "institutions reflect the routine way in which people do what they are supposed to do." (March and Olsen 1989, p. 21).

A key aspect is the self-reproducing nature of institutions, that is, the fact that they exhibit stability. This dynamic is found also in accounts within institutional economy looking at how standards and certain products mobilise support and become stable. In analysing the institutionalisation of technology standards Grindley (1995, p. 27) describes the basic mechanism as follows: the large installed base attracts complementary production and makes the standard cumulative more attractive. A larger base with more complementary products also increases the credibility of the standard. Together these make a standard more attractive to new users. This brings in more adoptions which further increases the size of the installed base, etc.

A number of theories on social action describe social order. Bourdieu's (1977) account of practice underscores the regular, structured and stable character of human action. Thus emphasising stability does not, of course, eliminate the unstable, the creative or exceptional. It simply highlights how human action is more stable and more difficult to change than some want to believe as it opposes "the reflexive freedom of subjects `without inertia' "(ibid., p. 56).

The aim of Bourdieu's theory of practice is to capture practical, everyday, immediate action rather than strategic or calculated deliberations. The key concept in the theory which accounts for the stability of action is the habitus. It produces a system of "durable, transposable dispositions" (ibid., p. 53) akin to "an acquired system of generative schemes" (ibid., p. 55) which regulate action. The habitus is a product of history and is inscribed in the body (ibid., p. 59).

The reason action is stable is that the habitus -- regulating action -- is constantly reproduced. It is reproduced by "protect[ing] itself from crisis and critical challenges by providing itself with a milieu to which it is as pre-adapted as possible" (ibid., p. 61), "excludes all `extravagances' (`not for the likes of us')" (ibid., 56) and the "internalisation of externality" (ibid., p. 55).

The habitus is basically individual, inscribed in the body, but extends to social groups well. Hence, group behaviour is reproduced similarly to individual action groups tend to "persist in their way, due to inter alia to the fact that they are composed of individuals with durable dispositions that can outlive the economic and social conditions in which they were produced" (ibid., p. 62). The habitus also accounts for the stability of institutions.

3.2 Infrastructure

In Webster's dictionary infrastructure is defined as:

"a substructure or underlying foundation; esp., the basic installations and facilities on which the continuance and growth of a community, state, etc. depends as roads, schools, power plants, transportation and communication systems, etc." (Guralnik 1970).

Some elements in this definition may be emphasised: infrastructures have a supporting or enabling function, and they are large and complex and are shared by a larger community.

Infrastructures like railways, roads, telecommunication networks, electricity supply systems, water supply, etc. have been analysed under the label "large technical systems" (Summerton 1994). Some scholars have focused on what they find to be important characteristics of such technologies, talking about networking (David 1987, Mangematin and Callon 1995) and systemic technologies (Beckman 1994). The notion of infrastructure is also increasingly used for describing the networks of information systems to be developed to support cooperation and communication within and across sectors like health care, transport, education and research, etc.

Characteristics of such infrastructures are of course their size and complexity concerning the technological elements as well as the number of developers and users involved. Another important aspect is the importance and role of standards. For those actively involved in the design of such networks, specifying the technological standards are indeed what the design is all about.

When computer networks are built, the same effects as those described by Grindley (1995) for standardised technologies unfold. As a network grows, it improves its position compared to the competing ones, and it becomes harder to change. The larger the network becomes, the more irreversible it turns. These issues are very visible in case of the rapid growth of Internet as well as the increasing difficulties in changing it to adapt to need requirements, requirements partly generated by the growth of the network in itself.

4 LINKING INFRASTRUCTURES AND INSTITUTIONS

At the core of this paper is a perspective seeing institutionalised practices and infrastructure technologies as closely related. They are in fact connected in several ways. On the surface level, we observe that an important role played by the patient record is to serve as a shared infrastructure for a large collecting of interconnected communities of practice, at the same time as the practices are highly institutionalised. Further, the practices and infrastructure technologies have been co-developed over long time, mutually adapted to each other into large convergent actor networks, the practice is inscribed into the technology making it difficult to change one of them without changing the other.

Infrastructures and institutions are also related as institutions, according to the ethnomethodological maxim saying that infrastructures at the same time enable as well as constrain, often play the role of a shared infrastructure coordinating interdependent actions. At the same time, infrastructures in use are linked to a network of actions in a way making them institutionalised. More precisely can infrastructures may be described as institutionalised standards.

When looking more closely at infrastructures and institutions, they share the structural properties making them hard to change separately as well as when they are considered a single unit. This makes it conceivable that the change strategies share something as well.

4.1 Changing institutions and infrastructures

Institutional theory, Bourdieu's notion of habitus and ANT all conceptualise and explain important aspects of how stability and reproduction take place. But they offer little or no help in developing strategies for changing irreversible networks. It is a different task to establish a network from scratch than to find a way to adapt to a situation with an already existing network.

Institutional theory is primarily used to explain the role of institutions in organizational change, how institutions are established and diffuse, and how change efforts fail due to clashes with exiting institutions. Planned organizational changes often fail as they include the change of institutionalised practices. Institutionalized practices are linked together into a multiplicity of interconnected networks. Changing the elements in the network constituting one practice will then implicitly imply changes in the other networks (practices) the elements are parts of as well. Focusing on only one practice (i.e. only one network) in isolation, will usually give as result changed elements of this practice not being aligned with the other networks it is also a part of. This means that the proposed changes in one practice causes troubles for others. If these troubles are too serious, as they according to the institutionalists often are, change efforts have to be given up.

Little has been done as far as change strategies are concerned. Brunsson and Olsen (1993) mentions a few conditions for successful change, from which some strategical fragments might be extracted. The success of a change reform depends on the degree of consistency between the value basis and beliefs underlying a proposed reform and the value basis and beliefs of an organization. Further, reformers are more likely to succeed if they try to change organizations in ways consistent with long-term trends in society. According to institutional theory, radical change only takes place in case of major crises or external shock.

Within the literature on large technical systems, infrastructural technologies are studied quite extensively. Changes are described but not thematised explicitly to any extent. David and Bunn (1988) analyse the crucial role of gateway technologies in the development of electricity, described as the battle of systems, that is alternating vs. direct currency.

Within the development of computer networks and information infrastructures an extensive stock of experience about developing infrastructures is being built up. The first and most basic characteristics of such networks is, however, the importance of a shared, stable and standardised communication protocol. In the design of network technologies. the importance of one single shared standard is considered so crucial that how to connect networks based on different protocols or how to change from one version to another is not considered relevant. However, when building networks such issues very soon come become present. Accordingly, there exist extensive experience in dealing with evolving networks, although this experience has not been tried generalised into more general tools and strategies. this is one of our objectives. The most important tools used are various forms of gateways.

The object usually considered the most interesting vehicle for learning about changing information infrastructures is Internet (Kahin and Abbate 1995). This is of course primarily due to its enormous success, but also due to the fact that Internet has appeared to be rather flexible and has changed a lot during its life time. At the moment, the ongoing transformation of the Internet form being based on IP version 4 to the new IP version 6 is a very interesting base for learning (Monteiro 1998; Hanseth, Monteiro, and Hatling 1996), the same is the rapid development and deployment of the Web technology. In the changes Internet has gone through, gateways have played a key role.

5 CONCLUSION: DESIGN REVISITED

We have outlined -- conceptually as well as by empirical illustrations -- how patient records should be viewed as large, complex and to an ever increasing degree interconnected and irreversible actor networks. The challenge when attempting to introduce EPRs, then, is to find strategies to somehow find a "fit" between the EPR and the already existing network. These strategies are what corresponds to design in the world of artefacts that are designed from scratch or in isolation. We describe four generic strategies which should be read as a vehicle for heuristically exploring this terrain. We take the latter two to be the most important ones as they allow for more radical changes. The former two are included more for reasons of completion.

5.1 Jumping between disconnected networks

One way of changing a network is to build a new one from scratch that is unrelated (that is, completely un-aligned) to the existing and require jumping from one network to another. Examples of this kind of changes is not found within medical records (as far as we are aware) but is illustrated by e-mail users subscribing to America Online that "jumped" to Internet. In the health sector, such jumping might happen when a hospital (or ward) having registered data (diagnoses, lab results or surgery information) according to one classification scheme abruptly switch to another. Such a change implies (among other things) that the medical personnel at the hospital (ward) exclude themselves from the network doing any kind of analysis of the data they register (like statistics for medical or management purposes) requiring that the data are coded according to the old classification system, and enrol themselves into the networks of practices using the new system.

Changing a network through this kind of abrupt changes, however, are often difficult to implement as being connected to the first network gives access to a large community of communicating partners, while the new one gives initially access to none, making it unattractive to be the first movers. In spite of this fact, this jumping strategy is really the one implicitly assumed in the definition of OSI protocols, and is claimed to be the main explanation of their failure (Stefferud 1994; Hanseth, Monteiro, and Hatling 1996). Such a jumping strategy might, however, be made more realistic if combined with organization and coordination activities. A simple strategy is to decide on a so-called "flag day" where everybody are expected to jump from the old to the new. This strategy requires, however, that the communicating community has a well defined, central authority and that the change is simple enough to be made in one single step. Changing the Norwegian telephone system in 1994 from 6 to 8 digit numbers was done in this way by the Norwegian Telecom which enjoyed a monopoly status.

Radical changes are often advocated, for instance within business process reengineering (BPR). Empirically, however, such radical changes of larger networks are rather rare. Hughes (1987) concluded that large networks only change in the chaos of dramatic crises (like the oil crises in the early 70s) or in case of some external shock. The same is held by the "new institutionalism". The BPR approach to organizational and technological change is a revolutionary approach but revolutions are costly as well as risky-- lots of blood are spilled, and the usually fail.

5.2 Backwards compatible networks

An already existing, stable network may be changed by stepwise adding nodes which strictly increase the number of services, practices or features the network allows. This is what corresponds to "backward compatibility" in the world of products (Grindley 1995). This kind of enrolment of new elements into a stabilised network contribute only to modest changes.

Within health care, a stabilized network of practices might be extended by adding new procedures for possibly making new diagnoses as new illnesses and deseases are discovered. Another example is using the patient record as it is for new purposes as was the case when it became an instrument for quality inspections or if it in the future will be used to support the analysis of a patient case in telemedicine consultations (as we believe will be the case). This kinds of enrolment of new elements into a stabilized network contribute to only modest changes. As the stabilized network grows, it rather becomes more resistible to major changes.

5.3 Gateways

Gateways may be used to link together two incompatible networks, that is, networks being internally aligned and stabilised but not aligned to each other. The gateway translates back and forth between these two networks thus making it possible for them to co-exist side by side. It also makes it possible, albeit in a restricted form, to make use of one network when situated within the other. When one of the networks are larger than the other, this strategy might be used to buy time to expand and mobilise the smaller network in a fairly sheltered environment.

In the Medakis project, one has followed the strategy of gateways to let paper based records co-exist with the new EPR. In this manner one postpones, or possible sidesteps, a confrontation between these two networks. This is achieved by translating patient records in electronic form to paper through a gateway. To implement this translation is more complicated than one would expect. This is due to the requirement that all information in the paper based patient record follow in exact sequel, one line right after the other one without gaps or overwriting. To implement a gateway which performs this translation according to these requirements, the gateway is in itself a small actor-network. It consists of artefacts and work routines. The secretary fetches the old paper record and uses a ruler to measure the distance from the top of the page down to where the next line of text is to be inserted. The paper record is inserted into the printer and the measured distance is entered into the EPR as an instruction to where to start printing. After printing the new information onto the old paper record, the sheet of paper is inserted into the remaining patient record and stored in the archived the usual way. In this manner, the EPR may co-exist with the paper based record and is allowed to gradually grow more appealing by accumulating more and more electronic records.

At the National Hospital in Oslo, they are just reorganizing the structure of the paper based patient record. About 545 existing forms are replaced by about 100 new ones, and the forms are organised into sections in a new manner. A record according to the new rules will be created as patients are arriving at the hospitals for visits. When new records are created, either the old forms may be put into the new sections, or all information may be transferred from the old to the new forms. In both cases, a computerised system is developed which given a specification of what kind of information is included in an existing, old record will produce a specification of on which new forms the information should be transferred, or into which sections the old forms should be put. In this way the old and new networks will coexist, and a gateway, built of a combination of humans and computer systems.

Star and Griesemer's (1989) concept of boundary objects may also be seen as a gateways enabling communication between different communities of practices.

5.4 Polyvalent networks

There are limits, however, to the degree of incompatibility a gateway may overcome. Such limits do not exist for a different strategy which involves a kind of "multiple attachments." This implies that an element is given polyvalent qualities and may attach to two (or more) different networks at the same time. To enable a complete transition from an old to a new network, first, all nodes must be attached to both networks, second, when all nodes are attached to the new, they may all detach from the old.

The situation at the Children/Heart ward at the National Hospital described earlier illustrates this strategy. Here the staff operated three distinct, but partly overlapping, patient record schemes. The use of each of these patient record schemes corresponds to a different network thereby resulting in three superimposed networks. The staff move fluently and constantly across these networks in the course of their everyday work. They engage polyvalently in any of the three networks in turn as they themselves perform whatever translation that is required to make sense of this criss-crossing.

Another example from the health care sector is the situation, at least in Norway, where it is not allowed to link local computer networks transferring patient information to Internet for security reasons. For this reason they have built a separate network of computers linked to Internet. Several employees access at least two computers, one linked to the internal network transferring patient information, using this network for communication within the hospital and another computer linked to Internet for communication with outside persons. The use of Internet is extending as knowledge about security is improved and security technologies is installed. It is expected that this process will continue until all communication enabled by Internet will move to that network.

So far this discussion has focused on the change of single, aligned networks. We see this as the major challenge. However, change at a higher level, that is, how a larger network consisted of smaller ones, should be considered as well. As we see it, such a "network of networks" is constituted by separate networks linked by gateways and/or polyvalent objects being linked to more then one network. The larger networks change as the separate networks changes as well as the gateways between them. Such decomposition of a network into smaller ones is important as the most basic principle for enabling change is modularisation, i.e. splitting one unit into smaller ones which may be changed independently. This principle, of course, also applies to networks (Hanseth, Monteiro, and Hatling 1996).

References

Berger, P. and Luckmann, T. (1967).The Social Construction of Reality. A Treatise in the Sociology of Knowledge. Penguin Books.

Berg, M. and Bowker, G. (1997). The Multiple Bodies of the Medical Record: Towards a Sociology of an Artifact. Sociological Quarterly.

Bowker, G. and Star, S.L. (1994). Knowledge and infrastructure in international information management. Problems of classification and coding. In L. Bud-Frierman, editor, Information acumen. The understanding and use of knowledge in modern business, pages 187-213. Routledge.

Bowker, G., Timmermans, S., and Star, S.L. (1995). Infrastructure and organizational transformation: classifying nurses' work. In (Orlikowski et al., 1995), pp. 344 - 370.

Brunsson, N., and Olsen, J.P. (1989). The Reforming Organization. Sage.

Bourdieu, P. (1977). Outline of a theory of practice. Cambridge: Cambridge Univ. Press, 1990 edition.

David, P.A. (1987). Some new standards for the economics in the information age. In P. Dasgupta and P. Stoneman, editors, Economic policy and technological performance, pages 206-239. Cambridge University Press.

David, P. and Bunn, J.A. (1988). The economics of gateway technologies and network evolutions, Information economics, 3:165-202.

Garfinkel, H. (1967). Studies in Ethnomethodology.

Grindley, P. (1995). Standards, strategy, and policy. Cases and stories. Oxford Univ. Press.

Hanseth, O., Monteiro, E. and Hatling, M. (1996). Developing information infrastructure standards: the tension between standardisation and flexibility. Science, Technology & Human Values, 21(4):407 - 426.

Heath, C. (1982). Preserving the consultation: medical record cards and professional conduct. Sociology of Health and Illness, Vol. 4, No. 1.

Hughes, T.P. (1987). The evolution of large technical systems. In The social construction of technological systems, pages 51-82. MIT Press.

Kahin, B. and Abbate, J. editors. (1995).Standards policy for information infrastructure. MIT Press, Cambridge, MA.

Jepperson, R.L. (1992). Institutions, Institutional Effects, and Institutionalism. In (Powell and DiMaggio, 1995).

Joerges, B., and Braun, I. (1994). How to Recombine Large Technical Systems: The Case of European Organ Transplantation. In (Summerton 1994).

Latour, B. (1986). Visualization and Cognition: Thinking with yes and Hands. Knowledge and Society: Studies in the Sociology of Culture Past and present. Vol 6, pp. 1-40, JAI Press Inc.

Latour, B. (1995). Social theory and the study of computerized work sites. In (Orlikowski et al., 1995) pp. 295 - 307.

Luff, P., Heath, C., and Greatbath, D. (1992). Tasks-in-interaction: paper and screen based communication in collaborative activity. Proceedings of the Conference on Computer Supported Cooperative Work, pp 163 - 170, ACM Press.

Mangemartin, V. and Callon, M. (1995).Technological competition, strategies of the firms and the choice of the first users: the case of road guidance technologies. Research policy, 24:441 - 458.

March, J. and Olsen, J.P. (1989). Rediscovering institutions. The organizational basis of politics. The Free Press, NY.

Monteiro, E. (1998). Scaling information infrastructure: the case of the next generation IP in Internet. The Information Society, To appear, 1998.

Nygren, E., and Henriksson, P. (1992). Reading the medical record. I. Analysis of physicians' ways of reading the medical record. Computer Methods and Programs in Biomedicine, 39 (1992), pp 1 -12. Elsvier Science Publishers.

Orlikowski, W., G. Walsham, M.R. Jones, and J.I. DeGross (eds.). (1995). Information technology and changes in organizational work. Chapman & Hall.

Powell, W.W, and DiMaggio, P.J. (1992). The Iron Cage Revisited: Institutional Isomorphism and Collective Rationality in Organizational Fields. In (Powell and DiMaggio, 1992), 63-82.

Powell, W.W, and DiMaggio, P.J. (eds.) (1992). The New Institutionalism in Organizational Science. The University of Chicago Press.

Quinn, J. (1996). Progress and Potential Pitfalls of Electronic Health Record Systems. Talk given at Conference on the Creation of a European Electronic Health Record, London, 15. Nov., 1996.

Rees, C. (1981). Records and hospital routines. In P. Atkinson and C. Heath, editors, Medical work. Realities and routines, pages 55-70. Gower, Farnborough.

Stefferud, E. and Pliskin, J. (1994). To tunnel or translate, that is the question. In M. Medina and N. S. Borenstein, editors, Upper layer protocols, architectures and applications, pages 277-288. IFIP.

Strauss, A., Fagerhaught, S., Suczek, B. and Wiener, C. (1985).Social organization of medical work. Univ. of Chicago Press.

Summerton, J. (ed.) (1994). Changing large technical systems, Westview.

Star, S. L., and Griesemer, J. G. (1989) Institutional Ecology, 'Translations,' and Boundary Objects: Amateurs and Professionals in Berkeley's Museum of Vertebrate Zoology, 1907-1939. Social Studies of Science, 19: 387- 420.