A CASE-BASED DIALOGUE SYSTEM FOR INVESTIGATING

THERAPY INEFFICACY

Rainer Schmidt and Olga Vorobieva

Institute for Medical Informatics and Biometry, University of Rostock, Rembrandtstr. 16/17, Rostock, Germany

Keywords: Case-Based Reasoning, Therapy Support, Medicine, Endocrine, Psychiatry.

Abstract: ISOR is a Case-Based Reasoning system for long-term therapy support in the endocrine domain and in

ychiatry. ISOR performs typical therapeutic tasks, such as computing initial therapies, initial dose

recommendations, and dose updates. ISOR deals especially with situations where therapies become

ineffective. Causes for inefficacy have to be found and better therapy recommendations should be

computed. In addition to former already solved cases, ISOR uses further knowledge forms, especially

medical histories of query patients themselves and prototypes. Furthermore, the knowledge base consists of

therapies, conflicts, instructions etc. So, different forms and steps of retrieval are performed, while

adaptation occurs as an interactive dialog with the user.

1 INTRODUCTION

In medical practice, therapies prescribed according

to a certain diagnosis sometimes do not give desired

results. Sometimes therapies are effective for some

time but suddenly stop helping any more. There are

many different reasons. A diagnosis might be

erroneous, the state of a patient might have changed

completely or the state might have changed just

slightly but with important implications for an

existing therapy. Furthermore, a patient might have

caught an additional disease, some other

complication might have occurred, or a patient

might have changed his/her lifestyle (e.g. started a

diet) etc.

For long-term therapy support in the endocrine

dom

ain and in psychiatry, we have developed a

Case-Based Reasoning system, named ISOR, that

not only performs typical therapeutic tasks but also

especially deals with situations where therapies

become ineffective. Therefore, it first attempts to

find causes for inefficacy and subsequently

computes new therapy recommendations that

should perform better than those administered

before.

ISOR is a medical Case-Based Reasoning system

at deals with the following tasks:

- choose appropriate (initial) therapies,

- compute doses for chosen therapies,

- update dose recommendations according to

lab

oratory test results,

- establish new doses of prescribed medicine

according to changes in a pat

ient’s medical status

or lifestyle,

- find out probable reasons why administered

erapies are not as efficient as they should,

- test obtained reasons for inefficacy and

make

sure that they are the real cause, and

- suggest recommendations to avoid inefficacy of

presc

ribed therapies.

ISOR deals with long-term diseases, e.g.

psychiatric

diseases, and with diseases even lasting for a

lifetime, e.g. endocrine malfunctions.

For psychiatric diseases some Case-Based

easoning systems have been developed, which deal

with specific diseases or problems, e.g. with

Alzheimer’s disease (Marling and Whitehouse,

2001) or with eating disorders (Bichindaritz, 1994).

Since we do not want to discuss various psychiatric

problems but intend to illustrate ISOR by

understandable examples, in this paper we focus

mainly on some endocrine and psychiatric disorders,

namely on hypothyroidism and depressive

symptoms. Inefficacy of pharmacological therapy

for depression is a widely known problem (e.g.

(Barbee and Jamhour, 2002), (Cuffel et al, 2003),

(Hirschfeld, Montgomery et al, 2002), (Keitner,

Posternak, and Ryan, 2003), (Lam, Wan, Cohen, and

Kennedy, 2002)). There are many approaches to

171

Schmidt R. and Vorobieva O. (2007).

A CASE-BASED DIALOGUE SYSTEM FOR INVESTIGATING THERAPY INEFFICACY.

In Proceedings of the Second International Conference on Software and Data Technologies - PL/DPS/KE/WsMUSE, pages 171-178

 SciTePress

solve this problem. Guidelines and algorithms have

been created (e.g. (Alacorn, Glover, Boyer, and

Balon, 2000), (Expert Consensus Guideline Series,

2000) (Osser and Patterson, 1998)). ISOR gives

reference to a psychopharmacology algorithm (Osser

and Patterson, 1998) that is available on the website

htp://mhc.com/Algorithms/Depression.

The paper is organized as follows. Firstly, we

introduce typical therapeutic tasks, subsequently

we present the architecture of ISOR and finally we

illustrate how it works by examples.

2 TYPICAL THERAPEUTIC

TASKS

As a consequence of our experiences with ICONS

(Schmidt and Gierl, 2001) a system for antibiotic

therapy advice, and with therapy support programs

for hypothyroidism (Vorobieva, Gierl, and

Schmidt, 2002), we believe that four tasks exist for

medicinal therapies. The first one means

computing an initial therapy, secondly an initial

dose has to be determined, later on dose updates

may be necessary, and finally interactions with

further diseases, complications, and especially with

already administered therapies have to be

considered.

In the following we illustrate the four tasks by

our programs that deal with therapy support for

hypothyroid patients. The antibiotics therapy

adviser ICONS deals only with two of these tasks:

computing initial therapies and initial doses.

2.1 Computing an Initial Therapy

Probably, the most important task for therapies is

the computation of initial therapies. The main task

of ICONS is to compute promising antibiotic

therapies even before the pathogen that caused the

infection is determined in the laboratory. However,

for hypothyroidism ISOR does not compute initial

therapies but only initial doses, because for

hypothyroidism only one therapy is available: it is

thyroid hormone, usually in form of levothyroxine.

2.2 Computing an Initial Dose

In ICONS the determination of initial doses is a

rather simple task. For every antibiotic a specific

calculation function is available and has to be

applied.

For hypothyroidism the determination of initial

doses (figure 1) is more complicated. Firstly, a

couple of prototypes exist. These are

recommendations that have been defined by expert

commissions. Though we are not sure whether they

are officially accepted, we call them guidelines. The

assignment of a patient to a fitting guideline is

obvious because of the way the guidelines have been

defined. With the help of these guidelines a range

for good doses can be calculated.

Figure 1: Determination of an initial levothyroxine dose.

To compute a dose with best expected impact, we

retrieve similar cases whose initial doses are within

the calculated ranges. Since cases are described by

few attributes and since our case base is rather

small, we use Tversky's sequential measure of

dissimilarity (Tversky, 1977). On the basis of those

retrieved cases that had best therapy results an

average initial therapy is calculated. Best therapy

results can be determined by values of a blood test

after two weeks of treatment with the initial dose.

The opposite idea to consider cases with bad

therapy results does not work here, because bad

results can also be caused by various other reasons.

To compute optimal dose recommendations, we

apply two forms of adaptation. First, a calculation of

ranges according to guidelines and patients attribute

values. Secondly, we use compositional adaptation.

That means, we take only similar cases with best

therapy results into account and calculate the

average dose for these cases, which has to be

adapted to the query patient by another calculation.

2.3 Updating the Dose in a Patient's

Lifetime

For monitoring a hypothyroidism patient, three basic

laboratory blood tests (TSH, FT3, FT4) have to be

undertaken. Usually the results of these tests

correspond to each other. Otherwise, it indicates a

more complicated thyroid condition and additional

tests are necessary. If the results of the basic tests

ICSOFT 2007 - International Conference on Software and Data Technologies

172

show that the patients thyroid hormone level is

normal, it means that the current levothyroxine dose

is OK. If the tests indicate that the thyroid hormone

level is too low, the current dose has to be increased

by 25 or 50 μg, if it is high, the dose has to be

decreased by 25 or 50 μg ((DeGroot, 1994),

(Hampel, 2000)). So, for monitoring, adaptation

means calculating according to some rules, which

are based on guidelines. Since an overdose of

levothyroxine may cause serious complications for a

patient, a doctor cannot simply consider test results

and symptoms that indicate a dose increase but

additionally he/she has to investigate reasons why

the current dose is not appropriate any more. In

ISOR this situation is described as a problem of

therapy inefficiency. In most cases the solution is

obvious, e.g. puberty, pregnancy etc. These

situations are covered by adaptation rules.

Sometimes cases are observed in which the

hypothyroidism syndromes are unexplained. For

these cases ISOR uses the problem solving program.

2.4 Additional Diseases or

Complications

It often occurs that patients do not only have

hypothyroidism, but they suffer from further chronic

diseases or complications. Thus, a levothyroxine

therapy has to be checked for contraindications,

adverse effects and interactions with additionally

existing therapies. Since no alternative is available

to replace levothyroxine, if necessary additionally

existing therapies have to be modified, substituted,

or compensated ((DeGroot, 1994), Hampel, 2000)).

ISOR performs three tests. The first one checks

if another existing therapy is contraindicated to

hypothyroidism. This holds only for very few

therapies, namely for specific diets like soybean

infant formula, which is the most popular food for

babies who do not get enough mother’s milk but it

prevents the effect of levothyroxine. Such diets have

to be modified. Since no exact knowledge is

available to explain how to accomplish this, our

program just issues a warning saying that a

modification is necessary.

The second test considers adverse effects. There

are two ways to deal with them. A further existing

therapy has either to be substituted or it has to be

compensated by another drug. Such knowledge is

available, and we have implemented corresponding

rules for substitutional and compensational

adaptation.

The third test checks for interactions between

both therapies. We have implemented some

adaptation rules, which mainly attempt to avoid the

interactions. For example, if a patient has heartburn

problems that are treated with an antacid, a rule for

this situation states that levothyroxine should be

administered at least four hours after or before an

antacid. However, if no adaptation rule can solve

such an interaction problem, the same substitution

rules as for adverse effects are applied.

3 SYSTEM ARCHITECTURE

ISOR is designed to solve typical problems,

especially inefficacy of prescribed therapies that

can arise in different medical domains. Therefore

most algorithms and functions are domain

independent. Another goal is to cope with

situations where important patient data is missing

and/or where theoretical domain knowledge is

controversial.

ISOR does not generate solutions itself. Its task

is to help users by providing all available

information and to support them to find optimal

solutions. Users shall be doctors, maybe together

with a patient.

Technically, ISOR is implemented in Delphi 7,

the format for the case and knowledge bases is

Paradox 7, and retrieval is performed by SQL.

In addition to the typical Case-Based Reasoning

knowledge, namely former already solved cases,

ISOR uses further knowledge components, namely

medical histories of query patients themselves and

prototypical cases (prototypes). Furthermore,

ISOR’s knowledge base consists of therapies,

conflicts, instructions etc. The architecture is

shown in figure 2.

In this section we explain the components and

in the next chapter we present examples to show

how the main knowledge components work

together.

3.1 Medical Case Histories

Ma and Knight (Ma and Knight, 2003) have

introduced a concept of case history in Case-Based

Reasoning. Such an approach is very useful when

we deal with chronic patients, because often the

same complications occur again, former successful

solutions can be helpful again, while former

unsuccessful solutions should be avoided.

The case history is written in the patient’s

individual base as a sequence of records. A

patient’s base contains his/her whole medical

history, all medical information that is available:

diseases, complications, therapies, circumstances

of his/her life etc. Each record describes an episode

in a patient’s medical history. Episodes often

characterise a specific problem. Since the case base

A CASE-BASED DIALOGUE SYSTEM FOR INVESTIGATING THERAPY INEFFICACY

173

is problem oriented, it contains just episodes and

the same patient can be mentioned in the case base

a few times, even concerning different problems.

Information from the patient’s individual base

can be useful for a current situation, because for

patients with chronic diseases very similar problems

often occur again. If a similar situation is found in

the patient’s history, it is up to the user to decide

whether to start retrieval in the general case base or

not.

In endocrinology, case histories are designed

according to a standard scheme, one record per visit.

Every record contains the results of laboratory tests

and of an interrogatory about symptoms, complaints

and physiological conditions of a patient.

Figure 2: ISOR’s architecture.

Therefore the retrieval of former similar situations

from the individual base of an endocrine patient is

easy to organise.

For psychiatric patients, case histories are often

unsystematic and they can be structured in various

forms. A general formalisation of psychiatric cases

and their histories is not achieved yet. The design of

case histories is problem dependent.

In both domains, we first search in the query

patient’s history for similar problems and for similar

diagnoses.

3.2 Knowledge Base, Case Base, and

Prototypes

The knowledge base contains information about

problems and their solutions that are possible

according to the domain theory. It has a tree

structure and it consists of lists of diagnoses,

corresponding therapies, conflicts, instructions, and

medical problems (including solutions) that can arise

from specific therapies. The knowledge base also

contains links to guidelines, algorithms and

references to correspondent publications ((Alacorn,

Glover, Boyer, and Balon, 2000), (Expert Consensus

Guideline Series, 2000), (Osser, and Patterson,

1998)).

The case base is problem oriented. Thus a case in

the case base is just a part of a patient’s history,

namely an episode that describes a specific problem

that usually has a solution too. So, the case base

represents decisions of doctors (diagnosis, therapies)

for specific problems, and their generalisations and

their theoretical foundations (see the examples). A

solution is called “case solution”, abbreviated “CS”.

Every case solution has (usually two)

generalisations, which are formulated by doctors.

The first one is expressed in terms of the knowledge

base and it is used as a keyword for searching in the

knowledge base. Such a generalisation is called

“knowledge base solution”, abbreviated “KBS”. The

second generalisation of a solution is expressed in

common words and it is mainly used for dialogues.

It is called “prompt solution”, abbreviated “PS”.

Former cases (attribute value pairs) in the case

base are indexed by keywords. Each case contains

keywords that have been explicitly placed by an

expert. For retrieval three main keys are used: a code

of the problem, a diagnosis, and a therapy. Further

keys such as age, sex etc. can be used optionally.

Prototypes (generalized cases) play a particular

role. Prototypes help to select a proper solution from

the list of probable or available solutions. A

prototype may help to point out a reason of

inefficacy of a therapy or it may support the doctor’s

choice of a drug.

3.3 Retrieval, Adaptation, and

Dialogue

For retrieval keywords are used. Since our system

is problem oriented, the first one is a code that

implies a specific problem. The second keyword is

the diagnosis and the other ones are retrieved from

the knowledge base.

Adaptation takes place as a dialogue between

the doctor, the patient, and the system. The system

presents different solutions, versions of them, and

asks questions to manifest them. The doctor

answers and selects suggestions, while the patient

himself or herself suggests possible solutions that

can be considered by the doctor and by the system.

We differentiate between two steps of

adaptation. The first one occurs as a dialogue

between ISOR and a user. Usually, doctors are the

users. However, sometimes even a patient may take

part in this dialogue. The goal of these dialogues is

to select probable solutions from all information

sources mentioned in sections 3.1 and 3.2. Pieces

ICSOFT 2007 - International Conference on Software and Data Technologies

174

of information are retrieved by the use of

keywords. Specific menus support the retrieval

process. The first step of adaptation can be

regarded as partly user based: ISOR presents lists

of probable solutions and menus of keywords, the

user selects the most adequate ones. The second

adaptation means proving obtained solutions. This

proving is rule based and it includes further

dialogues, laboratory test results, and consultations

with medical experts. While the procedures

supporting the first adaptation step are domain

independent, the adaptation rules of the second step

are mainly domain dependent.

4 EXAMPLES

By three examples we illustrate how ISOR works.

The first and the second one are from the endocrine

domain, the third one deals with a psychiatric

problem.

4.1 Hypothyroidism

4.1.1 Inefficacy of Levothyroxine Therapy

Every morning a mother gives her 10 year-old boy

not only the prescribed Levothyroxine dose but

also vitamin pills. These pills have not been

prescribed but they are healthy and have lately

been advertised on TV. Part of this medication is

Sodium Hydrocarbonate (cooking soda) that causes

problems with Levothyroxine.

Individual base. The same problem, inefficacy

of Levothyroxine therapy, is retrieved from the

patient’s history. The solution of the former

problem was that the boy did not take the drug

regularly. This time it must be a different cause,

because the mother controls the intake.

Knowledge base. It has a tree structure that is

organised according to keys. One main key is

therapy and the keyword is Levothyroxine. Another

keyword is instructions. These instructions are

represented in form of rules that concern the intake

of Levothyroxine. For Levothyroxine a rather long

list of instructions exists. Since the idea is that the

boy may break an instruction, this list is sorted

according to the observed frequency of offences

against them in the case base.

Concerning these instructions a couple of

questions are asked, e.g. whether the boy takes

Sodium Hydrocarbonate together with

Levothyroxine. Since the mother is not aware of

the fact that Sodium Hydrocarbonate is contained

in the vitamin pills, she gives a negative answer

and no possible solution can be established by the

knowledge base. However, soda is generated as

one keyword for retrieval in the case base.

So, the following solutions are retrieved from

the knowledge base, the third one does not fit for

the boy.

Knowledge base solution 1: Sodium Hydrocarbon

Knowledge base solution 2: Soy

Knowledge base solution 3: Estrogene

Case base. Using the keyword soda eight cases

with the following seven solutions are retrieved

(case solution 4 occurs twice).

Case solution 1: Aspirin Upsa

Case solution 2: Cooking Soda

Case solution 3: Soluble juice

Case solution 4: Alka Seltzer

Case solution 5: “Invite”

Case solution 6: Vitamin “Teddy”

Case solution 7: Lime Pills”

Thus we get a list of drugs and beverages that

contain sodium Hydrocarbonate, all of them belong

to the generalised solution “soluble”.

Solution. The boy admits to take Levothyroxine

together with an instantiation of the generalised

solution “soluble”, namely soluble vitamin.

Recommendation. The boy is told to take

vitamin four hours later than Levothyroxine.

Additionally, further interactions between vitamin

and Levothyroxine must be checked, because it

might be necessary to adjust the Levothyroxine

dose.

4.1.2 Improving the Efficacy by Dose

Updates

Figure 3 shows an example of a case study. We

compared the decisions of an experienced doctor

with the recommendations of ISOR. The decisions

are based on basic laboratory tests and on lists of

observed symptoms. Intervals between two visits

are approximately six months. In this example

there are three deviations between the doctor’s and

ISOR’s decisions, usually there are less. At the

second visit (v2), according to laboratory results

the Levothyroxine should be increased. ISOR

recommended a too high increase. The applied

adaptation rule was not precise enough. So, we

modified it. At visit 10 (v10) the doctor decided to

try to decrease the dose. The doctor’s reasons were

not included in our knowledge base and since his

attempt was not successful, we did not alter any

adaptation rule. At visit 21 (v21) the doctor

increased the dose because of some minor

symptoms of hypothyroidism, which were not

A CASE-BASED DIALOGUE SYSTEM FOR INVESTIGATING THERAPY INEFFICACY

175

included in ISOR’s list of hypothyroidism

symptoms. Since the doctor’s decision was

probably right (visit 22), we added these symptoms

to the list of hypothyroidism symptoms of ISOR.

100

120

V1 V2 V3 V4 V5 V6 V7 V8 V9 V10 V11 V12 V13 V14 V15 V17 V18 V19 V20 V21 V22

Levothyroxine µg/day

Doctor

Program

Figure 3: Dose updates recommended by our program

compared with doctor’s decision. V1 means the first visit,

V2 the second visit and so on.

4.2 Inefficacy of Psychiatric Drugs

Originally, ISOR was developed for the endocrine

domain, especially for hypothyroidism, but later on

it has been generalised. Now it can solve the same

types of problems in different medical domains.

Now we present an example from psychiatry.

A man, 55 years of age, has been treated for

depression for 15 years. Symptoms of depression

appeared after he was cruelly beaten near his house.

Since he did not see any connection between his

depression and the violence, he did not tell it to his

doctor. At first, the antidepressant Tofranil for

intake in the morning and the sedative Tisercin for

better sleep at bedtime were prescribed. Later on,

during another depression stage the antidepressant

Fluoxetine helped. Now, his problem is that neither

Fluoxetine, nor any other proposed antidepressant

helps any more.

Retrieval. Keywords are used to retrieve

information from all data sources. Since optional

keywords about a patient’s feeling (e.g. feeling

worse) are used for retrieval of the patient’s medical

history, even biographical events of a patient can be

retrieved.

Individual base. Since the problem of inefficacy

of an antidepressant never arose in the patient’s past,

no solution can be found. However, indirect

information was retrieved. The keyword feeling

better provided a trip to Switzerland, while the result

of the keyword feeling worse provided a trip to

Israel, where the latest very severe depression

began.

Feeling better: A trip to Switzerland

Feeling worse: A trip to Israel

The knowledge base contains information about

depression, anxiety and other psychiatric diseases,

possible complications and references to their

theoretical grounds ((Davidson, 1991), (Flor-Henry,

P., 1983), (Gelder, Lopez-Ibor, and Andeasen,

2000), (Leonhard, 1979), (Tucker and Liotti,

1989)). References to similar problems are retrieved,

the most remarkable one is a link to the algorithm

for psychopharmacology of depression

(htp://mhc.com/Algorithms/Depression, (Osser and

Patterson, 1998). Though the idea of the algorithm is

to solve the problem of non-response to an

antidepressant, it does not really fit here, because it

does not cover the situation that a therapy helped for

some time and then stopped having an effect.

Case base. Eleven cases with similar depression

problems are retrieved. Three of them are

characterised by the general idea depression is

secondary to anxiety resulting from a psychical

trauma.

Case solution 1: Severe stress during the World War 2

Case solution 2: Bad experience in a Jail

Case solution 3: Sexual assault in childhood

The other cases have solutions that are

generalised to changes in therapy.

Adaptation. ISOR displays retrieved

information pieces. In this case, two strategies are

offered. The first one suggests trying some other

therapy. This strategy is supported by the majority

of the retrieved cases and partly by theoretical

recommendations. The second strategy means to

check the diagnosis. This strategy is supported by

three retrieved cases and by the patient’s medical

history. The choice between both strategies is up to

the user. In this example the doctor chooses to

attempt the second strategy at first. The doctor is

especially led by the patient’s medical history,

because Switzerland is usually associated with a safe

life (especially in comparison to life in Russia),

while living in Israel is considered as unsafe.

Furthermore, this strategy is supported by the

general situation that some sedative drugs (e.g.

Tisercin at the beginning) had helped for some time.

ISOR offers a list of questions for the favoured

strategy and as a result the doctor concludes that in

this case depression is in fact only second to anxiety.

The man is permanently afraid of possible violence

and anxiety is based on strong fear that occurred

long ago.

Explaining remarks. Diagnosing anxiety needs

good medical skills, because patients try to suppress

traumatic events from their memory (Stein, 2003).

In this example depression even served as a

mechanism of suppression. The accepted case-based

ICSOFT 2007 - International Conference on Software and Data Technologies

176

solution spared the patient unnecessary experiments

with other psychopharmacological drugs.

So, the first problem is solved, a new diagnosis

is ascertained.

The next problem is prescription of a therapy.

According to the domain theory and to our

knowledge base anxiety implies Neuroleptics

(Gelder, Lopez-Ibor, and Andeasen, 2000.),

(Kalinowsky, and Hippius, 1969). Many of them are

available but a good choice is not trivial.

Individual base. From the patient’s history

those sedatives (Neuroleptics) are retrieved that he

took in his lifetime and that had positive effects on

his psychical condition: Tisercin and Paxil, which is

a drug that has both sedative and antidepressive

effects.

Prototype. Among those prototypes that have

been defined by doctors (based on their long

experience with cases) the prototypical solution

Paxil is retrieved.

Adaptation. Before described, every drug must

be checked for conflicts with the patient’s additional

diseases and already existing therapy. Though the

query patient has already taken Paxil in the past, our

system checks all possible conflicts. If necessary,

adaptation has to be performed. In this case no

conflicts are discovered and Paxil is prescribed.

5 CONCLUSION

We have presented a CBR system that helps doctors

to solve medical problems, particularly to

investigate causes of inefficacy of therapies. It

includes different knowledge containers, namely a

case base, a knowledge base, prototypes, and

individual bases of patients that reflect their medical

histories. Information retrieved from these

containers is arranged in form of dialogues.

The case base plays a central role in the dialogue

forming process. It serves as a kind of filter when

the knowledge base suggests too many possible

solutions for the problem (as in the first example). In

this situation the most typical cases are retrieved

from the case base. When a solution from the

knowledge base is not convincing or when it is

hardly adaptable, the case base may provide better

alternatives (as in the third example).

Generalisations, keywords and references to

other knowledge components belong to the case

base. The adaptation program uses them to create

dialogues. In the part that concerns the case base and

the dialogues ISOR can be considered as domain

independent.

The design of the case base and our

implementation allow solving problems from

different medical domains. Specific, domain

dependant features are attributed mostly to the

individual base, because every domain requires a

special design of case histories. The knowledge base

in ISOR is domain-oriented, but all algorithms and

functions are completely domain independent.

ACKNOWLEDGEMENTS

We thank Dr. Monika Mix, Children’s Hospital of

the University Clinic of Rostock, and Prof. Nikolai

Nikolaenko, Sechenov Institute of Evolutionary

Physiology and Biochemistry in St.Petersburg, for

their data and for their help and time during our

consultations.

REFERENCES

Alacorn, R.D., Glover, S., Boyer, W., Balon, R., 2000.

Proposing an algorithm for the pharmacological

treatment of posttraumatic stress disorder. Ann Clin

Psychiatry 12 (4) 239-246

Barbee, J.G., Jamhour, N.J., 2002. Lamotrigine as an

augmentation agent in treatment-resistant depression.

J Clin Psychiatry 63 (8) 737-41

Bichindaritz,I., 1994. A case-based assistant for clinical

psychiatry expertise, Journal of the American Medical

Informatics Association, Symposium Supplement 673-

677

Cuffel, B.J., Azocar, F., Tomlin, M., Greenfield, S.F.,

Busch, A.B., Croghan, T.W., 2003. Remission,

residual symptoms, and nonresponse in the usual

treatment of major depression in managed clinical

practice. J Clin Psychiatry 64 (4) 397-402

Davidson, R.J., 1991. Cerebral asymmetry and affective

disorders: A developmental perspective. In: Cicchetti,

D., Toth, S.L. (eds.): Internalizing and externalizing

expressions of dysfunction. Rochester Symp. on

Developmental Psychopathology 2, Erlbaum,

Hillsdale, 123-133

DeGroot, L.J., 1994. Thyroid Physiology and

Hypothyroidsm. In: Besser, G.M., Turner, M. (eds.):

Clinical endocrinilogy, Wolfe, London, Chapter 15

Expert Consensus Guideline Series Treatment of

Posttraumatic Stress Dosorder, 2000. J Clin

Psychiatry 60 (suppl 16) 1-76

Flor-Henry, P., 1983. Cerebral Basis of Psychopathology.

John Wright, Boston

Gelder, M.G., Lopez-Ibor, U., Andeasen, N.C. (eds.),

2000. New Oxford Textbook of Psychiatry. Oxford

University Press, Oxford

Hampel, R., 2000. Diagnostik und Therapie von

Schilddrüsenfunktionsstörungen. UNI-MED,

Bremen

A CASE-BASED DIALOGUE SYSTEM FOR INVESTIGATING THERAPY INEFFICACY

177

Hirschfeld, R.M., Montgomery, S.A., et al, 2002. Partial

response and nonresponse to antidepressant therapy:

current approaches and treatment options. J Clin

Psychiatry 63 (9) 826-37

Kalinowsky, L., Hippius, H., 1969. Pharmacolological,

convulsive and other somatic treatments in psychiatry.

Grunee&Stratton, New York London

Keitner, G.I, Posternak, M.A., Ryan, C.E., 2003. How

many subjects with major depressive disorder meet

eligibility requirements of an antidepressant efficacy

trial? J Clin Psychiatry 64 (9) 1091-3

Lam, R.W., Wan, D.D., Cohen, N.L., Kennedy, S.H.,

2002. Combining antidepressants for treatment-

resistant depression: a review, J Clin Psychiatry 63 (8)

685-93

Leonhard, K., 1979. The Classification of the Endogenous

Psychoses. John Wiley, New York

Ma, J., Knight, B.A., 2003. Framework for Historical

Case-Based Reasoning. In: Proc ICCBR-2003,

Springer, Berlin, 246-260

Marling, C., Whitehouse, P., 2001. Case-Based Reasoning

in the care of Alzheimer’s disease patients. In: Proc

ICCBR-2001, Springer, Berlin, 702-715

Osser, D.N., Patterson, R.D., 1998. Algorithms for the

pharmacotherapy of depression, parts one and two.

Directions in Psychiatry 18, 303-334

Schmidt, R., Gierl, L., 2001. Case-based Reasoning for

Antibiotics Therapy Advice: An Investigation of

Retrieval Algorithms and Prototypes. Artificial

Intelligence in Medicine 23 (2) 171-186

Stein, M.B., 2003. Attending to anxiety disorders in

primary care. J Clin Psychiatry 64 (suppl 15) 35-39

Tucker, D.M., Liotti, M., 1989. Neuropsychological

mechanisms of anxiety and depression. In: Boller, F.,

Grafman, J. (eds.): Handbook of Neuropsychology,

vol. 3. Elsevier, Amsterdam, 443-456

Tversky, A., 1977. Features of similarity. Psychological

review 84, 327-352

Vorobieva, O., Gierl, L., Schmidt, R., 2002. Case-based

Adaptation in Medicine - Focusing on

Hypothyroidism. In: Proc UK-Workshop on CBR,

University of Paisley Press, Paisley, 61-68

ICSOFT 2007 - International Conference on Software and Data Technologies

178