Despite the high number of premature terminations in psychotherapy and their importance both from the clinical and the research point of view (Clarkin & Levy, 2004; Lambert & Ogles, 2004), our knowledge of the possible predictors of this phenomenon is still unsatisfactory, especially for what concerns client variables. Although this topic has been the object of several studies, it is difficult to organize systematic studies with post-tests for those patients that abandon the clinical context suddenly and without giving explanations. The drop-out phenomenon is even more important in the institutions, where the lack of flexibility of the therapeutic setting, the impossibility for the patient to choose his/her therapist and the multiplicity of therapeutic contexts, often uncoordinated among themselves, undermine the dyadic relationship between the therapist and the patient. Identifying the factors responsible for premature terminations in a lack of motivation on the patient’s part, or in the clinician’s failure appears too simplistic: a research by Fava and Masserini (2002) indicates the need to analyze with greater precision the characteristics of both patients and therapists in the cases of drop-out.
The main results of the scientific literature on this topic are reviewed by Garfield in the fourth edition of the Handbook of Psychotherapy and Behavior Change (Bergin & Garfield, 1994) and by Clarkin and Levy in the fifth edition of that manual, edited by Lambert (2004).
According to these reviews, the client variables that have been most frequently investigated in the last ten years are the following:
1. Age, gender and psychiatric diagnosis: Garfield (1994) reports that his results do not show a clear and one-to-one relationship between these characteristics and premature terminations. Clarkin and Levy (2004) agree that age and sex are not specific risks, although the diagnosis of personality disorders generally associated to high risk of premature termination. Other studies (Gunderson, 1996; Clarkin at al., 1999) are more specific and mention Borderline Personality Disorder, or even only some of its traits, as risk indicators.
2. Social class, education, race: according to Garfield patients that are not Caucasian, that belong to the lower classes, or that have received less education are at higher risk (in researches carried out in the US, these three characteristics are often found in the same patients). Related to these are other variables, which may act both separately and together: according to some researches, socially and educationally challenged patients frequently are only interested in getting rid of their symptoms, are not interested in psychotherapy, do not understand the rationale of psychotherapy and have different values as the therapist, who frequently belong to the middle class. Garfield underlines that this can create serious communication difficulties between the patient and the therapist, since the very beginning of their relationship. Clarkin and Levy confirm these results and add on the necessity that supervisors seriously consider these difficulties, especially the ones related to racial and ethnic identities.
3. Patient's expectations: Garfield reports results from several researches which show that patients (unlike therapists) expect very short psychotherapies - i.e. ten sessions - and this has negative effects both on the outcome of the treatment and on the premature terminations. Clarkin and Levy add that if patients don't expect positive results from the treatment, the chances of a negative outcome increase.
It seems that no research carried out in the last years has produced any new perspectives on the issue of the relationship between the characteristics of patients and their premature terminations, as well as about these characteristics and the outcome of treatment. If we compare this datum with the great development of the research on interaction variables (i.e. therapeutic alliance) one might think that the interest for patients' characteristics has faded. Furthermore, Clarkin and Levy, in the article quoted above, state that, although Garfield underlined the role of client variables in isolation, more recently the trend is to underline the interaction between these variables and those pertaining to the therapist and the treatment. To this purpose, Stiles and his colleagues write:
“Client variables have a plausible impact on the therapy, but as soon as therapy begins, the client variables are in a dynamic and ever changing context of therapist variables and behaviours. […] The therapist’s responsiveness to client variables and behaviour will determine the statistical relationship of the client variable to outcome.” (Stiles et al., 1998, p. 441).
One thus could think that, at the moment, the study of client variables suffers from a loss of interest. Fortunately this is not the absolute truth, and in the already-cited article of 1994, Garfield, even if emphasizing the limits of the research on clients variables, made a very interesting comment on another aspect of research:
“Nevertheless, the research on client variables has produced data that are of some real practical value. […] On the basis of the existing research results, it would appear worthwhile for each clinical setting to evaluate its own pattern of continuation and outcome: Since settings differ, such relatively direct and uncomplicated investigation may provide findings that are meaningful and can help improve the quality of professional service at the particular clinical centre.”
On the basis of these statements, in January 1999 we decided to begin a intervention-research aimed at evaluating outcomes of psychotherapy in a Community Mental Health Centre of Rome. Within this research project, which is still ongoing, the present paper deals with the specific aim of examining predictors of premature terminations.
From the methodological point of view, because of the limited predictive value of the differences relative to psychiatric symptoms, we focused on some personality traits. To this issue we used the Italian version of the Adjective Check List (Gough, Heilbrun & Fioravanti, 1980), especially because this classical test can be administered easily and can be used for both self-description and description by an external observer. This last aspect seemed particularly useful because the comparison between the two descriptions – that, in our case, were the patient himself and the therapist – makes it possible to obtain an index of an interactional variable, in a simple way. Nonetheless, we assessed the psychiatric symptoms by means of the SCL-90 test.
The main aim of this study was to gather information that could offer to the clinicians a clue to identify which client variables could be used as predictors of outcome and of premature terminations in their clinical settings, in order to deal with this problem more successfully.
Hypotheses were:
1) psychiatric diagnosis should not predict the premature termination status, as indicated by previous research, with the exception of personality disorders;
2) symptoms severity, which is related to treatment efficacy, should not necessarily predict premature termination;
3) the patient’s motivation to change (measured by some subscales of the ACL) and a low consistence in the patients/therapist evaluations should predict premature termination.
Data collection is still running but we think that it could be useful to communicate preliminary results, which may be replicated in other studies.

METHOD

Participants

All males and females that apply to the public Mental Health Centre where our research was carried out are assessed by means of a semi-structured interview yielding socio-demographic data, and a number of interviews – ranging from two to five – with a psychiatrist or a clinical psychologist, who evaluates also the individual’s global functioning according the criteria of the Global Assessment of Functioning Scale (GAF-Axis V) of the DSM-IV.
All patients referred to the Centre for the first time between January 1999 and January 2004 were invited to participate in the research if she/he fulfilled the following criteria:
1) On the basis of the assessment’s results, the subject was considered suitable to begin psychotherapy.
2) He/she accepted a psychotherapy.
3) He/she was not pharmacologically treated.
4) He/she would not have to be wait-listed for more than 30 days before beginning the treatment.
5) He/she was willing to participate in a research on the assessment of the outcomes of psychotherapies, after being informed of the aims of the project, on the voluntary character of their participation, on the possibility of withdrawing from the research in every moment, and on the treatment of personal data according to the ethical  code of the Italian Association of Psychology (Associazione Italiana di Psicologia, www.aipass.org).

On the basis of the above specified criteria, 28 female patients and 4 male patients were enrolled, who – when data had been analyzed – had either concluded psychotherapy or had interrupted it on their own initiative. Because the number of male patients was too exiguous, it was decided to report only the data about the female patients.
The average age of the patients in our sample was 25.9 (sd=3.8; ranging 19-35). All patients were unmarried except for one who was divorced. Concerning education, 21 possessed a high school diploma, 5 held university degrees, 1 had done vocational school, 1 had terminated the middle school. Concerning working and professional activity, 1 was unemployed, 4 were employed on a non-continuous basis (mostly working in call centres) and 23 were university students (since the Mental health centre is situated in the neighbourhood  of the ‘La Sapienza’ University, a great deal of clients are students).
The patients who concluded their therapy in agreement with their therapist are 10, the patients classified as premature terminators are 18.
The diagnoses, carried out according to the criteria of the DSM-IV (American Psychiatric Association, 1994) are summarized in Table1, separately for those who concluded and those who interrupted prematurely the treatment.

Procedure

Data collection

The patients who accepted to participate in the research were administered the following tests:

• Symptom Check List-90-R  (Derogatis, 1983): the SCL-90 was filled in by the patients at the end of the diagnostic interviews. In order to establish a stable baseline level of the presence and intensity of psychiatric symptom clusters patients were asked to fill the SCL-90 in for three times, with a weekly frequency. The patients were also asked to keep, during the same weeks, a ‘clinical log’ in which they should indicate their psychological wellbeing using a 10 point scale ranging from 1 to 10, where 10 corresponded to the maximum wellbeing.
• Adjective Check List (Gough, Heilbourn, Fioravanti, 1980): this test was used both for self-descriptions and for the description of the patients by an external observer. The patient was asked to describe himself immediately before the beginning of psychotherapy. The psychotherapist was asked to described the patient after the second session.

According to the research protocol, this psychometric assessment was repeated every six months, for the whole length of the treatment, and after the end of the psychotherapy.
In this paper we will report only results of the analyses conducted on data pertaining to the initial assessment, considering them possible predictors of a premature termination or of an agreed-upon conclusion. We have defined ‘premature terminator’ a patient who has been accepted for psychotherapy, who has at least one session of therapy and who discontinues treatment on her/his own initiative by failing to come for any future arranged visit with the therapist. (Garfield, 1994).
For all the patients in our sample, the interval between the end of the first assessment and the beginning of the treatment was less than thirty days. The treatments (individual psychotherapy with weekly frequency) have been done by seven experienced therapists: three males (with a psychodynamic approach) and four females (two with a psychodynamic approach, one with a cognitive-behavioural approach, one with a systemic approach).

Data Analyses

The differences between the distribution of socio-demographic characteristics or diagnoses between the two groups (premature terminations vs. agreed-upon conclusion) were assessed through the Mann-Withney test or the Chi-square test.
The evaluation of which patients’ characteristics are predictive of premature termination was conducted through discriminant analysis (step-wise method).
Analyses were conducted through Statistical 8 software package.

RESULTS

The average number of sessions for the premature terminators is 18.67 (sd=20.91), that of the patients that concluded therapy in a positive way is 33.36 (sd=22.34). This difference is statistically significant (Mann-Withney 32.05; p=0.004).

Sociodemographic variables, diagnosis, intensity of symptomatic dimensions.
Age, marital status, education and occupation do not differ between the two groups.
For what concerns diagnoses, carried out according to the DSM-IV criteria (American Psychiatric Association, 1994), most patients (13 out of 28) are affected by a personality disorders; only one of them, belonging to the premature terminators group, is also affected by a disorder of  Axis I (panic attacks with agoraphobia).
As shown in table 1, in the subset of personality disorders, the number of premature interruptions is greater than the number of good conclusions, while among patients affected by anxiety disorders, mood disorders and eating disorders, those who concluded therapy and those who prematurely terminated it are distributed quite homogeneously. In order to assess if a diagnosis of personality disorder could be, with respect to other diagnoses, a possible drop-out predictor, the frequencies of premature terminations and conclusions among these patients have been compared using Chi-square test (with Yates’ correction). Results shows the absence of statistically significant differences (Chi-square = 0.82, p= 0. 37).

To evaluate whether the intensity of the different symptom dimensions measured by SCL-90-R was, in our sample, a predictor of premature terminations, we computed as first, the sensitivity and the specificity of a classification based on the mean Global Severity Index (GSI: Premature termination: 1.42; sd=0.57; range = 0.62-2.72. Conclusions: 1.24; sd=0.67; range 0 0.60-2.60). Using 1 as best cut-off point, we obtained a sensitivity of 72.2%, a specificity of 60% and a percentage of subjects correctly classified of 67.9. The classification of patients on the basis of the GSI is reported in table 2.

In order to evaluate whether the motivation to change, self reported by the patients, and the intensity of the symptoms were predictive of premature termination, a discriminant analysis was conducted considering the CHA (Need for Changing) and the CRS (Counseling Readiness Scale) subscales of the ACL and the GSI obtained from the SCL-90 as predictors and the groups (premature terminations vs. agreed-upon conclusion) as criterion. Results of the analysis evidenced no significant discriminant function.
The analysis was repeated considering the same subscales of the ACL assessed by the therapist together with the GSI. Results of this analysis evidenced no significant discriminant function as well.
Finally a third discriminant analysis was conducted considered as predictors the three indices of discrepancy between the patient’s self description and the therapist’s description of the patients, computed as suggested in the ACL Manual (Gough, Heilbrun & Fioravanti, 1980).
Formulas used are the following:

I_1 – Insight 1, obtained as: a/(a+b)
Where:
a = number of items checked in both descriptions
b = number of items checked in the self-description but not in the therapist description

I_2 Insight 2, obtained as: a/ (a+c)
Where:
a = number of items checked in both descriptions
c = number of items checked by the therapist but not in the self-description.

IP (intrapunitivity), obtained as: (d+e)/(d+e+f+g)
Where:
d = number of unfavourable adjectives checked only in the self-description
e = number of favourable adjectives checked only in the therapist description
f = number of favourable adjectives checked only in the self-description
g = number of unfavourable adjectives checked only in the therapist description.

The discriminant function allows the correct classification of 78.6% of the patients. Specifically 88.9 of the premature terminators and 60% of the patients that conclude the psychotherapy can be correctly classified.

DISCUSSION

The present research has been conducted with the aim of sharing light on the client variables which can potentially predict the drop-out phenomenon. The interest in this kind of prediction is not only theoretical but also clinical, since, as stated by Clarkin and Levy (2004, p. 197), “from the clinician’s point of view, those individuals who drop out of treatment prematurely are not taking advantage of an important resource in their lives”.
In our research the diagnoses according to the criteria of the DSM-IV are not good predictors of premature terminations. Perhaps this is due to the low sample size which do not allowed a direct comparison between each diagnostic category but only the Axis I vs Axis II comparison. However it is noteworthy that the mean number of sessions within the premature termination group is higher than that of previous studies, where it is reported that most patients interrupt prematurely the treatment at the beginning or only after the first session. That is why Clarkin and Levy (2004) suggest that clinicians often offer psychotherapy to those people ready to change by themselves or with whatever help. On the contrary, in the UOSM where the present study was conducted, the restrictive selection criteria for patients to be admitted to the research protocol, could explain the differences found in the results.
As regards the results of the analyses involving the SCL-90-R and the ACL, the low sample size did not allow analyses including all variables assessed. Thus, for statistical reasons, the analyses included only the GSI and two subscales (CHA, CRS) of the ACL. Neither the self evaluation given by patients nor the evaluations given by the therapist were predictive of premature termination. The GSI was non predictive as well, even if classifying patients through a cut-off score of 1, around 68% of them (19 out of 28) are correctly classified also with respect to the treatment status (premature terminations vs. agreed-upon conclusion).
On the contrary, the most interesting results are those produced by the discriminant analysis carried out on the self- and hetero-description discrepancy indices. In this case the analysis has shown a statistically significant discriminant function in which only two predictors (Insight 2 and Intrapunitivity) are included. These functions allow a correct classification of a satisfactory percentage of patients (about 79%) and, despite an increase of false positives, the drop-out prediction is fairly accurate (in fact, 89% of premature termination is correctly classified). Furthermore, from a clinical point of view, since the consequence of a false positive error is to drive the therapist attention to the aspects of the interaction relevant for maintaining the therapeutic relationship, we think that this rate of error could not be considered negative. If we examine the mean scores of the two groups (see table 3), the patients that prematurely terminate therapy – as opposed to those that conclude them – seem to be characterized by less Insight 2 (that is, by a lower capacity to be aware of both positive and negative characteristics than the therapist has detected) and by less Intrapunitivity (that is, by a lower tendency to describe themselves in less favourable terms than they are described by their therapists).
The results of this analysis show that these indices are the main outcome predictors, as they supply precious information on the matching that, as Clarkin and Levy (2004) have emphasised, takes place between the client variables and the therapist variables as soon as the therapy begins. The discrepancies between the values of the two indices can in fact depend on the patient’s intrinsic characteristics, on mistaken evaluations by the therapist, or on both. Furthermore, if they are substantial and not recognised, they can undermine the creation of a constructive therapeutic alliance.
For example, as concerns the I_2 variable, an individual who is not aware of some of his peculiar ways of presenting him/herself – which instead are evident to an outside observer – will however tend experience difficulties in all interpersonal relationships where it is necessary to possess introspective abilities, such as in insight-oriented psychotherapies and especially psychoanalysis. From this point of view, the important contribution of this variable to a correct classification of our subjects certainly depends on the fact that all the seven therapists who participated in this research share a psychodynamic approach. However, it is also true that, from an interactional point of view, this kind of discrepancy between patient and therapist, regardless of the validity of the therapist’s evaluation, increases the probability that his/her intervention might not be understood by the patient or might give him/her the impression of having been ‘unmasked’ or ‘exposed’. In this case, if the therapist is not capable of being empathically attuned to the patient’s difficulties, the chances of premature terminations will increase.
Similarly, as concerns the IP variable, it is true that an individual – like those of our patients that had continued therapy – who has a tendency to consider him/herself worse than he/she appears to an external observer, also manifests depressive traits which, according to the results in outcome research, are more frequently associated with positive outcomes (Roth and Fonagy, 1996). Nevertheless, it is also true, that in an interactional perspective, such a discrepancy between the patient’s and the therapist’s evaluations, regardless of its causes (i.e. patient’s excessive intrapunitivity or therapist’s idealization of the patient) increases the chances that the therapist’s interventions (including his/her interpretations) might decrease the patient’s guilt feelings and enhance his/her self-esteem. In turn, this might decrease the probability of premature terminations, even though an exaggerated idealization of the patient by the therapist can have very negative effects on the therapeutic process.
From a more general point of view, we believe that Michael and Enid Balint’s distinction (Balint & Balint, 1961) between “autogenic disease” (i.e. the patient’s image of his/her new sensations, fears, suspicions and anxieties) and “iatrogenic disease” (i.e. the doctor’s image of the pathological situation, based on the anamnesis and the clinical examination), although proposed more than forty years ago, still keeps its usefulness. According to these authors, in fact, the initial diagnosis must also establish a connection between the iatrogenic and autogenic diseases, which must be understandable by both the patient and the doctor, and emotionally satisfy the former, who otherwise might feel him/herself misunderstood, disappointed, upset, abandoned.
Furthermore, our observations on the agreement between the therapist’s and the patient’s evaluations are largely supported in the recent literature on the ‘therapeutic alliance’, which is considered the most important factor contributing to the therapeutic change (Lingiardi, 2002). It is, in particular, the patient’s ability to trust the therapist – and therefore recognize his/her need for help – that is an essential continuity factor, especially in patients with serious pathologies (Gunderson, 2001). Also the therapist’s monitoring the quality of the clinical relationship and the specific characteristics of that particular patient are considered fundamental components of the therapeutic alliance (Roth & Fonagy, 1996).
In conclusion, we would recommend some caution: the results we are presenting require further empirical evidence on more numerous and different samples. The small number of subjects involved in our study, 28 young women, most of whom university students, highly limits the generalization of our findings. This numerical weakness is not due to a high refusal rate (only one male patient refused to participate). On the contrary, it could depend on a reduced space for psychotherapy in the Italian psychiatric services and on the restricted criteria used for controlling the bias of the waiting list (not more than 30 days of waiting).
For these reasons, in order to increase the number of patients we are trying to increases the number of UOMS involved in the project. In fact, consistently with suggestions of other authors (e.g. Fava, Masserini, 2002; Vigorelli, 2005), we acknowledge the usefulness of studying psychotherapy in public centres. To this issue Clarkin & Levy (2004) state that: “Currently, a major research concern is to extend efficacy research that is conducted on highly selected clients at research centres with carefully selected therapists to research that evaluates the effectiveness of specific therapeutic approaches to a more heterogeneous group of clients in the local community treated by community therapists”.

References

Balint M. & Balint E.(1961). Psychotherapeutic Techniques in Medicine. London:Tavistock Publications.

Clarkin J.F., & Levy K.N. (2004). The influence of client variables on psychotherapy. In Lambert M.J. (Ed.), Bergin and Garfield’s  handbook of psychotherapy and behavior change. New York: Wiley & Sons.

Clarkin J.F., Yeomans F., Kernberg O.F. (1999). Psychotherapy for borderline personality. New York: Wiley & Sons.

Derisley J, Reynolds, S. (2000). The transtheoretical stages of change as a predictor of premature termination, attendance and alliance in psychotherapy. British Journal of Clinical Psychology. 39, 371-382.

Derogatis L.R. (1983): SCL-90. Administration, scoring and procedure manual for the Revised Version. Baltimore (MD): Clinical Psychometric Research.

Fava E., Masserini C. (2002). Efficacia delle psicoterapie nel servizio pubblico. Milano: Franco Angeli.

Garfield S.L. (1994). Research on client variables in psychotherapy. In A.E. Bergin & S.L. Garfield (Eds.), Handbook of psychotherapy and behavior change. New York: Wiley.

Gough G.H., Heilbrun B.A., Fioravanti M.(1980). Manuale della versione italiana della Adjective checklist. Firenze: Organizzazioni Speciali.

Hill C.E. & Lambert M.J. (2004). Methodological issues in studying psychotherapeutic processes and outcome. In Lambert M.J. (Ed.), Bergin and Garfield’s  handbook of psychotherapy and behavior change. New York: Wiley.
 
Lambert M.J. & Ogles B.M. (2004). The Efficacy and Effectiveness of Psychotherapy. In Lambert M.J. (Ed.), Bergin and Garfield’s  handbook of psychotherapy and behavior change. New York: Wiley.

Stiles W.B., Honos-Webb L. & Surko M. (1998). Responsiveness in psychotherapy. Clinical psychology: Science and Practice, 5, 439-58.

Roth A. & Fonagy P. (1996). What works for whom? A Critical Review of Psychotherapy Research. London: Guilford Press.
 
Stiles W.B., Honos-Webb L. & Surko M. (1998). Responsiveness in psychotherapy. Clinical psychology: Science and Practice, 5, 439-58.

Vigorelli M. (a cura di) (2005). Il lavoro della cura nelle istituzioni. Milano: FrancoAngeli.

Notes

* Department of Philosophical Research, University of Rome “Tor Vergata” . Top

** Mental health task force, 3th District, ASL RM A. Top

*** Department of Psychology, University of Rome “La Sapienza”. Top

**** Department of Dynamics and Clinical Psychology, University of Rome “La Sapienza”. Top

1. The English translation is edited by dr. Alessandra Telmon, Psychologist. Top