ComFu.ter Methods and Programs m Biomedicine, 32 (1990) 171-175 Elsevier
171
COMMET 01098
What are computers and models good for? Pros and Cons K. Piwernetz 1 a n d U. Fischer 2 1 Diabetes Center, Hospital Bogenhausen, Munich, F.R..G., and 2 Central Institute of Diabews "Ge,'hordt Katsch; Karlsburg~ G.D.R.
This issue contains selected p r e s e n t : ~ n s from the XIIIth International Karlsburg Symposium on Problems of Diabetes, which was held in Dresden, G.D.R., from October 16-18, 1989, on the topic 'Models and Computers in Diabetes Research and Diabetes Care'. Under this headline, papers and computer demonstrations were presented on strategies of model-based diagnosis and treatment, on computer-aided systems for information and decision support, on the verification of metabolic models, and on clinical studies employing computer-aided systems. The concluding panel discussion * of "he meeting highlighted some basic problems of computer application in clinical medicine and its Pros and Cons in diabetes care - - from the following viewpoints. Several problems in health care are caused by an increasing burden due to administration, to an old-fashioned style of commum.~ation, and also to unstructured documentation systems. Thus, it becomes more and more difficult to detect the really important findings in an abundance of routine data. In turn, the overlooking of important facts may result in additional problems for the patient and thereby increase the costs of public health service provision.
Correspondence: Dr. Dr. K. Piwernetz, Diabetes Zentrum Miinchen-Bogenhausen, Englschalkingerstr. 77, D-8000 Miinchen 81, F.R.G. * Panelists: K. Piwernetz, (Munich, F.R.G., chairman), K.P. Ratzmann (Berlin, G.D.R.), J. Schrezenmeir (Mainz, F.R.G.), P. S~nksen (London, U.K.), H. Thoma (Vienna, Austria). Programme Committee: U. Fischer (Karlsburg, G.D.R.).
In fact, computer programs were designed to support complex situations in health care but most of them have failed to become incorporated into routine environments. One reason for this is the fact that most of these programs were developed by informaticians, i.e., by people who by themselves are interested in the functionality of computer-based systems rather than in the usec interface or in the impact on the health care environment. The concrete requirements !n routine work on the one side and the potential support by computers on the other have not yet been balanced to fit in smoothly without introducing new problems. The overall outcome results from benefits provided and new problems which the clinician is faced with. Thus it has to be critically evaluated, whether the net effect of computer application is merely positive or negative. Particular concern has been expressed about the impact of computers on the health care environment. It is mainly the relationship between the patient and the doctor or the nurse which is considered sensitive to influences from outside. Nothing should be allowed to enter this field which might eventually distract the attention of medical staff. It is, therefore, most important that any system or program which is to be introduced should fulfil certain requirements. Unfortunately, little work has been published so far on basic requirements for medical programs or systems and indeed, many of the commercially available programs are not suitable for broad clinical application. The analysis of the impact of information technology (IT) on health care has always to start
0169-260"//90/$03.50 O 1990 Elsevier Science Publishers B.V. (Biomedical Division)
172
from the medical problem. Assuming that a medical problem occurs in the process of providing health care and a computer system and program be available to help solving it, the scene is set. However, to assist in solving the actual problem additional requirements have to be fulfilled: the respective user, i.e., the patient or the physician, must be able to detect the problem. In order to be able to support the solution effectively, the functions of the program must be carefully formulated. The functions, however, can only be used via the user interface which, therefore, must be tailored to the environment where care is provided. If this aspect is disregarded, then the user is not able to make adequate use of the program (Fig. 1). Also, the implementation of 1T tools into the medical environment will modify the health care process. Thus, in most cases routine activities will be supported and time be saved in favour of the patient; but care has to be taken to ensure that the net effect on health care provision is not negative, neither with respect to time nor to costs. In addition, the age and the reputation of the potential user of IT should be thought of. It is a general experience that with increasing age it becomes more and more difficult to learn new topics, to
adopt new ways of thinking, or simply to change habits. Most of the computer programs have been designed by young computer specialists who are sometimes not able to consider these problems. Unfortunately, behavioural science has not yet looked at this problem. Imagine what most programs do: they give more detailed and more precise insights into a problem but they do not solve it, i.e., they prepare the available information in a way to enable the user to apply his knowledge more efficiently. Most of the programs are directed at the health care professional who is in charge of the patient on a day-to-day basis. It may, therefore, be difficult for a medical officer who is older and not involved in the daily routine to make use of the programs or even to judge them appropriately (Piwernetz). In principle, most available computers are now based on serial prog~'amm~r:g and operati,~n. It is, however, obvious that the human brain operates in a different manner: data processing is in most cases governed by routines of paraiiel processing of elements which are probabilistic and to a certain extent stochastic. At the moment, the processes of decision making in the human brain are transplanted into serial expert systems. The:/
MEDICAL PROBLEM
Physician/Patient
oviroooeni' sero r' oe I
HEALTH CARE
PROGRAM FUNCTIONS
Fig. 1. Triangle of Information Technology in the medical environment. Medical problems are recorded by patients, physicians or other health care professionals. Attempts to improve health care affected by these problems always have to take the respective environment into consideration. I~ information technology is available to support a solution, the program functions can only be accessed through the user interface. This shows the close interdependency of the users, the environment and the user interface, which seems to be obvious - - but, nevertheless - - has been disregarded in many attempts to implement information technology into health care,
173
cannot be expected to generate reproducible results. This holds true particularly in highly context-sensitive operations such as in the process of elaborating diagnoses. Much information is used in parallel to guide the decision-making process by enforcing certain ways of thinking and by excluding others. This does not happen via strict 'yes' or 'no' decisions but via some kind of weighted probability. Attempts have recently been made to investigate this technique by means of designing neural networks. Up to now, however, neither the hardware nor the programming techniques have been developed in a form capable of being introduced into medical expert systems. Also, little work has been done to characterise or to quanti!iy medi,~.,d criteria using the image or the appearance of the patients. This is very difficult in numerous cases such as assessing the smell of the patient, its shape, stature, colour of the skin, etc. These are obvious limitations of computer applications into medical sciences, which cannot be overcome in the near future. Nevertheless, there are numerous tasks in medical care which can be supported by methods of IT, if it is applied in the right manner and within the right environment. Basic work now has to be done to produce user interfaces which - - without timeconsuming training - - allow the use of the functionality of programs even by 'illiterate users' from the very beginning. Unfortunately, the scientific world seems to leave this responsibility to industry, including all positive aspects of rapid product delivery and all negative ones of not scientific-based knowledge of user-computer interaction in very sensitive environments. In this context, the 'physician-patient relationship' must be analysed from the clinical viewpoint (Ratzmann): - are patients deprived of their personal integrity? do computers lead to loss of contact or to 'inhumanity'? - do computers raise barriers between the physician and his patient? Both the development and the implementation of computers and programs for medical purposes have to consider two aspects: the medical problem -
and the clinical environment. The problems provide guidelines for the functions, the structure and the organisation of the system as a whole. At the very beginning of model implementation, there is the need to specify which medical or management problem shall be solved and how the solution can be supported by the system. For this purpose, test implementations may be helpful even in programs of limited functionality and in a selected environment. Then there is the need to investigate very critically whether even if old problems might be solved new ones might be introduced! Only if these boundary conditions are well recognised, will there be no negative influence on the sensible medical environment or the physician-patient relationship. Thus, it appears important to have involved all the users from the very beginning and to consider fears and barriers which are always present when new technologies are being introduced into a conservative en~ronment such as the medical one. Special precautions and programming techniques are required to exclude any drawback both for the integrity of the patient and for the relationships between different parties involved in the health care process. This has to be verified at critical checkpoints during the implementation of the system. Clinical medicine requires the consideration of the following points: - medical care is based on a complex combination of knowledge and experience; there is a defined responsibility for the results of decisions applied to each patient; it may be difficult to represent by the program the state of the art. Care must be taken in keeping track of new developments. In the immediate future, the entire field of medical care cannot be supported by computers on a routine basis. There are, however, activities where significant help may be provided right now such as in documentation, patient manageraent, scheduling, and automated reports~ Also, there should be selective access to :he data for statistical, epidemiological, and management evaluations. These features may also support research activities or applications in quality assurance. Especially in the U.S.A., many systems have been designed for budgeting and cost-benefit evaluations. Other -
-
174
fields such as consultation or expert systems are under development but they are not yet ready for routine use. It must be noted, however, that in these knowledge-based systems which support medical decisions the responsibility remains al~vays with the user who is in charge of medical care. In general, these systems are not able to consider all available information which might be relevant for the solution of the problem. Therefore, the provider of IT has to elucidate the prerequisites of the respective programs, what can be expected, and - - even more important - - the limitations. This is important both for the welfare of the patient and for the legal safety of the user. If possible, the user ought to know how the output of the system is produced. However, its complexity does in many cases not allow the medical user to obtain insight. In particular for expert systems no procedures are available to test all constellations in all environments. In these cases not even the designer is able to predict the outcome of the system in total. Also one has to consider that medical knowledge is rapidly changing but not all new information is relevant or even sufficiently verified to allow implementation into decisionsapport systems which are intended for general use. There was, therefore, agreement on the necessity of some kind of editing committee which decides on the extent and on the time of implementation of newly available knowledge. There is already some long-standing experience with the application of computers in clinical care. For instance, systems for documentation in outpatient clinics which are operating on a limited set of data. They provide basic reports both to the clinic and to the referring physician which may be handed out directly to the patient, thereby facilitating communication with GPs. It is a general experience that numerous physicians are using these systems but some do strongly refuse to learn how to use them. Tools have to be provided which allow application even in the context of such 'non.responders'. Thus, they may have a standardised list looking like the computer screen and they fill in the data by hand and some assistant is entering them into the computer. It was investigated how a system like this influences procedures and habits in an out-clinic. After a
ran-in period, neither the doctor nor the patient felt disturbed by the related vi!eo-system. It has been shown that - - ~l! together - - no additional time is required to run the program and that the doctor is not distracted from the patient (S~nksen). The near-normoglycaemic insulin substitution (NIS) in insulin-dependent diabetic patients is based on the idea that basal insulin requirement can be calculated according to different rules as compared with the dose to be given before meals. The NIS concept comprises a quantitative approach for derivation of insulin doses from actual blood glucose readings and from the intake of carbohydrate. Some adaptation process is performed by the patient himself without the help of computers. The most promising role of this concept is reported to lie in the support of patient education by means of simulation programs for the calculation of appropriate insulin regimens. This concept is, however, not aimed towards the use of computers by patients for the every day adaptation of insulin doses; but computers may be helpful in the graphic illustration of the whole process (Thoma). Extending this principle to intensified insulin therapy, the patient has to decide several timr.s a day on the insulin doses to be applied in relation to the intended carbohydrate intake. Until now, there is no on-site assistance for the patient, no physician is around to give advice in assessing the complex situation under consideration of the 'metabolic history', of actual blood glucose readings, carbohydrate intake, exercise, etc. In this situation, the computer may support the patient in his decisions. Methods of data processing offer assistance in two directions (Schrezenmeir): (1) the previous records may be stored and the program allows the internal evaluation of data to derive those parameters which are needed to give advice in managing the actual situation; (2) additional information on the present situation provided (blood glucose, carbohydrate intake, exercise, illness . . . . ), so that the program may come up with reproducible proposals. These are not biased by mood or motivation as they are more or less in the usual therapeutic setting. Now, it also becomes easy to assess retrospectively the ap-
175
propriateness of available information and the need for adapting the program or the internal variables, either manually or automatically. Most importantly, any standardised procedure like the ones described will only work properly, if the required data are entered. Because of the great variety of possible developments from a given metabolic situation, specially designed computer program~ can only detect major critical errors in the data input. Hence, the remaining ones may be wrong enough to result in potentially harmful advice for the patient. This imminent problem of all computer-based systems may l~e characterised by the idiom 'Garbage in - Garbage out'! Summarising, there was general agreement in that computers are only tools, even if very powerful ones, to meet certain targets. Their usefulness and the advantages of application are largely dependent (1) on how precisely the targets have been defined before, (2) on whether the manner of computer usage does meet the features of the specific environment, i.e., of the health care system in general and of the specific diabetes management problems in particular, (3) on its availability at the right time and at the right place. In stand-alone systems, the use of the computer may be optional. In many cases, however, the functionality of the system depends strongly on a complete database. This is especially true when
systems contain a variety of different interrelated functions. In particular, in systems supporting medical documentation it is essential that adequate data are available whenever needed. Regulations have to be elaborated on the time, the means, and the persons involved in putting items into the system. In diabetes management there are, for i~nstance, potential fields of application which require computers to bring about specific features, i.e., their function as an electronic logbook and their integration into the educational process which can be accomplished by combining the retrospective evaluation of metabolic data with model-based simulation of the therapeutic process. One hope from the embryonic stage of medical information systems has, however, been abandoned, i.e., the paperless office. Using the available technology, there is no way of getting rid of the paper. On the contrary, care must be taken not to produce additional heaps of paper! Since many advantages from computer application are not primarily addressed at the patient himself but at the physician and the health care sy,:tem per se, a routine application should only be taken into consideration when the targets have been defined precisely and the implementation is evaluated carefully (Fischer).
171
COMMET 01098
What are computers and models good for? Pros and Cons K. Piwernetz 1 a n d U. Fischer 2 1 Diabetes Center, Hospital Bogenhausen, Munich, F.R..G., and 2 Central Institute of Diabews "Ge,'hordt Katsch; Karlsburg~ G.D.R.
This issue contains selected p r e s e n t : ~ n s from the XIIIth International Karlsburg Symposium on Problems of Diabetes, which was held in Dresden, G.D.R., from October 16-18, 1989, on the topic 'Models and Computers in Diabetes Research and Diabetes Care'. Under this headline, papers and computer demonstrations were presented on strategies of model-based diagnosis and treatment, on computer-aided systems for information and decision support, on the verification of metabolic models, and on clinical studies employing computer-aided systems. The concluding panel discussion * of "he meeting highlighted some basic problems of computer application in clinical medicine and its Pros and Cons in diabetes care - - from the following viewpoints. Several problems in health care are caused by an increasing burden due to administration, to an old-fashioned style of commum.~ation, and also to unstructured documentation systems. Thus, it becomes more and more difficult to detect the really important findings in an abundance of routine data. In turn, the overlooking of important facts may result in additional problems for the patient and thereby increase the costs of public health service provision.
Correspondence: Dr. Dr. K. Piwernetz, Diabetes Zentrum Miinchen-Bogenhausen, Englschalkingerstr. 77, D-8000 Miinchen 81, F.R.G. * Panelists: K. Piwernetz, (Munich, F.R.G., chairman), K.P. Ratzmann (Berlin, G.D.R.), J. Schrezenmeir (Mainz, F.R.G.), P. S~nksen (London, U.K.), H. Thoma (Vienna, Austria). Programme Committee: U. Fischer (Karlsburg, G.D.R.).
In fact, computer programs were designed to support complex situations in health care but most of them have failed to become incorporated into routine environments. One reason for this is the fact that most of these programs were developed by informaticians, i.e., by people who by themselves are interested in the functionality of computer-based systems rather than in the usec interface or in the impact on the health care environment. The concrete requirements !n routine work on the one side and the potential support by computers on the other have not yet been balanced to fit in smoothly without introducing new problems. The overall outcome results from benefits provided and new problems which the clinician is faced with. Thus it has to be critically evaluated, whether the net effect of computer application is merely positive or negative. Particular concern has been expressed about the impact of computers on the health care environment. It is mainly the relationship between the patient and the doctor or the nurse which is considered sensitive to influences from outside. Nothing should be allowed to enter this field which might eventually distract the attention of medical staff. It is, therefore, most important that any system or program which is to be introduced should fulfil certain requirements. Unfortunately, little work has been published so far on basic requirements for medical programs or systems and indeed, many of the commercially available programs are not suitable for broad clinical application. The analysis of the impact of information technology (IT) on health care has always to start
0169-260"//90/$03.50 O 1990 Elsevier Science Publishers B.V. (Biomedical Division)
172
from the medical problem. Assuming that a medical problem occurs in the process of providing health care and a computer system and program be available to help solving it, the scene is set. However, to assist in solving the actual problem additional requirements have to be fulfilled: the respective user, i.e., the patient or the physician, must be able to detect the problem. In order to be able to support the solution effectively, the functions of the program must be carefully formulated. The functions, however, can only be used via the user interface which, therefore, must be tailored to the environment where care is provided. If this aspect is disregarded, then the user is not able to make adequate use of the program (Fig. 1). Also, the implementation of 1T tools into the medical environment will modify the health care process. Thus, in most cases routine activities will be supported and time be saved in favour of the patient; but care has to be taken to ensure that the net effect on health care provision is not negative, neither with respect to time nor to costs. In addition, the age and the reputation of the potential user of IT should be thought of. It is a general experience that with increasing age it becomes more and more difficult to learn new topics, to
adopt new ways of thinking, or simply to change habits. Most of the computer programs have been designed by young computer specialists who are sometimes not able to consider these problems. Unfortunately, behavioural science has not yet looked at this problem. Imagine what most programs do: they give more detailed and more precise insights into a problem but they do not solve it, i.e., they prepare the available information in a way to enable the user to apply his knowledge more efficiently. Most of the programs are directed at the health care professional who is in charge of the patient on a day-to-day basis. It may, therefore, be difficult for a medical officer who is older and not involved in the daily routine to make use of the programs or even to judge them appropriately (Piwernetz). In principle, most available computers are now based on serial prog~'amm~r:g and operati,~n. It is, however, obvious that the human brain operates in a different manner: data processing is in most cases governed by routines of paraiiel processing of elements which are probabilistic and to a certain extent stochastic. At the moment, the processes of decision making in the human brain are transplanted into serial expert systems. The:/
MEDICAL PROBLEM
Physician/Patient
oviroooeni' sero r' oe I
HEALTH CARE
PROGRAM FUNCTIONS
Fig. 1. Triangle of Information Technology in the medical environment. Medical problems are recorded by patients, physicians or other health care professionals. Attempts to improve health care affected by these problems always have to take the respective environment into consideration. I~ information technology is available to support a solution, the program functions can only be accessed through the user interface. This shows the close interdependency of the users, the environment and the user interface, which seems to be obvious - - but, nevertheless - - has been disregarded in many attempts to implement information technology into health care,
173
cannot be expected to generate reproducible results. This holds true particularly in highly context-sensitive operations such as in the process of elaborating diagnoses. Much information is used in parallel to guide the decision-making process by enforcing certain ways of thinking and by excluding others. This does not happen via strict 'yes' or 'no' decisions but via some kind of weighted probability. Attempts have recently been made to investigate this technique by means of designing neural networks. Up to now, however, neither the hardware nor the programming techniques have been developed in a form capable of being introduced into medical expert systems. Also, little work has been done to characterise or to quanti!iy medi,~.,d criteria using the image or the appearance of the patients. This is very difficult in numerous cases such as assessing the smell of the patient, its shape, stature, colour of the skin, etc. These are obvious limitations of computer applications into medical sciences, which cannot be overcome in the near future. Nevertheless, there are numerous tasks in medical care which can be supported by methods of IT, if it is applied in the right manner and within the right environment. Basic work now has to be done to produce user interfaces which - - without timeconsuming training - - allow the use of the functionality of programs even by 'illiterate users' from the very beginning. Unfortunately, the scientific world seems to leave this responsibility to industry, including all positive aspects of rapid product delivery and all negative ones of not scientific-based knowledge of user-computer interaction in very sensitive environments. In this context, the 'physician-patient relationship' must be analysed from the clinical viewpoint (Ratzmann): - are patients deprived of their personal integrity? do computers lead to loss of contact or to 'inhumanity'? - do computers raise barriers between the physician and his patient? Both the development and the implementation of computers and programs for medical purposes have to consider two aspects: the medical problem -
and the clinical environment. The problems provide guidelines for the functions, the structure and the organisation of the system as a whole. At the very beginning of model implementation, there is the need to specify which medical or management problem shall be solved and how the solution can be supported by the system. For this purpose, test implementations may be helpful even in programs of limited functionality and in a selected environment. Then there is the need to investigate very critically whether even if old problems might be solved new ones might be introduced! Only if these boundary conditions are well recognised, will there be no negative influence on the sensible medical environment or the physician-patient relationship. Thus, it appears important to have involved all the users from the very beginning and to consider fears and barriers which are always present when new technologies are being introduced into a conservative en~ronment such as the medical one. Special precautions and programming techniques are required to exclude any drawback both for the integrity of the patient and for the relationships between different parties involved in the health care process. This has to be verified at critical checkpoints during the implementation of the system. Clinical medicine requires the consideration of the following points: - medical care is based on a complex combination of knowledge and experience; there is a defined responsibility for the results of decisions applied to each patient; it may be difficult to represent by the program the state of the art. Care must be taken in keeping track of new developments. In the immediate future, the entire field of medical care cannot be supported by computers on a routine basis. There are, however, activities where significant help may be provided right now such as in documentation, patient manageraent, scheduling, and automated reports~ Also, there should be selective access to :he data for statistical, epidemiological, and management evaluations. These features may also support research activities or applications in quality assurance. Especially in the U.S.A., many systems have been designed for budgeting and cost-benefit evaluations. Other -
-
174
fields such as consultation or expert systems are under development but they are not yet ready for routine use. It must be noted, however, that in these knowledge-based systems which support medical decisions the responsibility remains al~vays with the user who is in charge of medical care. In general, these systems are not able to consider all available information which might be relevant for the solution of the problem. Therefore, the provider of IT has to elucidate the prerequisites of the respective programs, what can be expected, and - - even more important - - the limitations. This is important both for the welfare of the patient and for the legal safety of the user. If possible, the user ought to know how the output of the system is produced. However, its complexity does in many cases not allow the medical user to obtain insight. In particular for expert systems no procedures are available to test all constellations in all environments. In these cases not even the designer is able to predict the outcome of the system in total. Also one has to consider that medical knowledge is rapidly changing but not all new information is relevant or even sufficiently verified to allow implementation into decisionsapport systems which are intended for general use. There was, therefore, agreement on the necessity of some kind of editing committee which decides on the extent and on the time of implementation of newly available knowledge. There is already some long-standing experience with the application of computers in clinical care. For instance, systems for documentation in outpatient clinics which are operating on a limited set of data. They provide basic reports both to the clinic and to the referring physician which may be handed out directly to the patient, thereby facilitating communication with GPs. It is a general experience that numerous physicians are using these systems but some do strongly refuse to learn how to use them. Tools have to be provided which allow application even in the context of such 'non.responders'. Thus, they may have a standardised list looking like the computer screen and they fill in the data by hand and some assistant is entering them into the computer. It was investigated how a system like this influences procedures and habits in an out-clinic. After a
ran-in period, neither the doctor nor the patient felt disturbed by the related vi!eo-system. It has been shown that - - ~l! together - - no additional time is required to run the program and that the doctor is not distracted from the patient (S~nksen). The near-normoglycaemic insulin substitution (NIS) in insulin-dependent diabetic patients is based on the idea that basal insulin requirement can be calculated according to different rules as compared with the dose to be given before meals. The NIS concept comprises a quantitative approach for derivation of insulin doses from actual blood glucose readings and from the intake of carbohydrate. Some adaptation process is performed by the patient himself without the help of computers. The most promising role of this concept is reported to lie in the support of patient education by means of simulation programs for the calculation of appropriate insulin regimens. This concept is, however, not aimed towards the use of computers by patients for the every day adaptation of insulin doses; but computers may be helpful in the graphic illustration of the whole process (Thoma). Extending this principle to intensified insulin therapy, the patient has to decide several timr.s a day on the insulin doses to be applied in relation to the intended carbohydrate intake. Until now, there is no on-site assistance for the patient, no physician is around to give advice in assessing the complex situation under consideration of the 'metabolic history', of actual blood glucose readings, carbohydrate intake, exercise, etc. In this situation, the computer may support the patient in his decisions. Methods of data processing offer assistance in two directions (Schrezenmeir): (1) the previous records may be stored and the program allows the internal evaluation of data to derive those parameters which are needed to give advice in managing the actual situation; (2) additional information on the present situation provided (blood glucose, carbohydrate intake, exercise, illness . . . . ), so that the program may come up with reproducible proposals. These are not biased by mood or motivation as they are more or less in the usual therapeutic setting. Now, it also becomes easy to assess retrospectively the ap-
175
propriateness of available information and the need for adapting the program or the internal variables, either manually or automatically. Most importantly, any standardised procedure like the ones described will only work properly, if the required data are entered. Because of the great variety of possible developments from a given metabolic situation, specially designed computer program~ can only detect major critical errors in the data input. Hence, the remaining ones may be wrong enough to result in potentially harmful advice for the patient. This imminent problem of all computer-based systems may l~e characterised by the idiom 'Garbage in - Garbage out'! Summarising, there was general agreement in that computers are only tools, even if very powerful ones, to meet certain targets. Their usefulness and the advantages of application are largely dependent (1) on how precisely the targets have been defined before, (2) on whether the manner of computer usage does meet the features of the specific environment, i.e., of the health care system in general and of the specific diabetes management problems in particular, (3) on its availability at the right time and at the right place. In stand-alone systems, the use of the computer may be optional. In many cases, however, the functionality of the system depends strongly on a complete database. This is especially true when
systems contain a variety of different interrelated functions. In particular, in systems supporting medical documentation it is essential that adequate data are available whenever needed. Regulations have to be elaborated on the time, the means, and the persons involved in putting items into the system. In diabetes management there are, for i~nstance, potential fields of application which require computers to bring about specific features, i.e., their function as an electronic logbook and their integration into the educational process which can be accomplished by combining the retrospective evaluation of metabolic data with model-based simulation of the therapeutic process. One hope from the embryonic stage of medical information systems has, however, been abandoned, i.e., the paperless office. Using the available technology, there is no way of getting rid of the paper. On the contrary, care must be taken not to produce additional heaps of paper! Since many advantages from computer application are not primarily addressed at the patient himself but at the physician and the health care sy,:tem per se, a routine application should only be taken into consideration when the targets have been defined precisely and the implementation is evaluated carefully (Fischer).
