Is there a prospective association between psychological distress as measured by the CORE-OM and treatment attendance and treatment duration? A follow-up study at a Norwegian Community Mental Health Centre

Abstract Background Feasible and reliable methods for identifying factors associated with treatment duration and treatment attendance in mental health services are needed. This study examined to what degree the Clinical Outcomes in Routine Evaluation-Outcome Measure (CORE-OM) at the start of treatment is associated with treatment attendance and treatment duration. Methods Outpatients (N = 124) at a community mental health centre in Norway completed the 34-item CORE-OM questionnaire addressing the domains of subjective well-being, problems and symptoms, functioning and risk at the start of treatment. The CORE-OM subscales and the ‘all’ items total scale were used as predictor variables in regression models, with treatment duration, number of consultations attended, treatment attendance (number of therapy sessions attended divided by number of sessions offered) and termination of treatment (planned versus unplanned) as outcome variables. Results Higher CORE-OM subscale scores and the ‘all’ scale were associated with longer treatment duration. No association was found between CORE-OM scales and number of therapy sessions, treatment attendance (sessions attended/offered) or whether the patients unexpectedly ended treatment. Conclusion Higher patient-reported psychological distress as measured by the CORE-OM at the start of treatment was prospectively associated with treatment duration but not with treatment attendance or drop-out of treatment. The findings imply that patients with higher initial psychological distress need longer treatment but that treatment attendance may be related to factors other than the severity of distress.


Introduction
Enduring treatment without achieving the desired improvement of patients' conditions is a major challenge in mental health services.Feasible and reliable methods for the early identification of patient subgroups at risk of staying in treatment for a long time facilitate the planning of treatment and the use of resources, and they may point to the need for early involvement of necessary alternative support services in the community.Other challenges in mental health specialist services are missed appointments (no-show) and the unplanned ending of treatment (drop-out).Low treatment attendance and drop-out are associated with poorer clinical outcomes and service inefficiencies and may disproportionately affect disadvantaged patient populations [1].Various factors may predict or be related to extended treatment or treatment attrition.For example, treatment effectiveness, patients' experiences of treatment, accessibility or personal or contextual factors may be associated with longer duration of treatment or unplanned ending of treatment.Further knowledge about factors associated with the risk of low treatment attendance or the unplanned ending of treatment is needed.
Such knowledge would help therapists to identify patients at risk of non-attendance or dropping out and, in particular, to motivate these patients to stay in treatment.

Clinical Outcomes in Routine Evaluation-Outcome Measure
The Clinical Outcomes in Routine Evaluation-Outcome Measure (CORE-OM) system is intended to monitor psychological change, to aggregate information on multiple levels, to benchmark and to provide feedback to stakeholders, including therapists, service designers and managers [2][3][4].The measure is widely used in routine clinical settings in Norway and elsewhere.Knowing whether it is related to the patients' course of treatment here is, therefore, of interest.The CORE-OM has also been translated into other Nordic languages [5,6].

CORE-OM and treatment duration
Studies investigating whether the self-report CORE-OM questionnaire predicts treatment duration or treatment non-attendance in terms of consultation no-show or drop-out are hard to find.However, in a large study across a range of mental health services using the CORE Practice-based Evidence National database in the uK, Stiles et al. [7] found that the mean pre-to post-treatment reduction in CORE-OM scores was similar, regardless of the number of sessions attended by patients with planned treatment endings [7].Also in the uK, Evans et al. [8] found that the 10-item short version of the CORE-OM was correlated with the number of psychological therapy sessions and treatment duration in a study on patients using secondary care adult psychological therapy teams [8].The median number of therapy sessions received was 15, with a median of 26 weeks between the first and last sessions.

CORE-OM and treatment attendance
A study by Saxon et al. [9] on counselling and clinical psychology services in the uK based on the CORE Practice-based Evidence National database included data from 10,521 patients from 14 sites [9].In this study, the one-to-one therapy drop-out rate was 34%.The mean (Sd) number of sessions attended by patients who dropped out was 2.8 (1.91) compared to 6.1 (2.68) for patients with a planned end of therapy.The CORE-OM non-risk subscale measures psychological distress, but it excludes items relating to self-harm, suicidal ideation and violent behaviour or threats towards other people.In another study, di Bona et al. [10] looked at the Access to Psychological Therapy (IAPT) services in the uK and found lower CORE-OM non-risk scores, more frequent thoughts of 'being better off dead' (derived from the CORE-OM), either a very recent onset of a common mental disorder (i.e.anxiety or depression) (<1 month) or a long-term condition (>2 years) and therapy site as predictors of not attending the first scheduled therapy session [10].

Study aim
As the CORE-OM is widely used in clinical settings, we find it relevant to assess whether the CORE-OM scales are associated with important clinical-and system-level outcomes, such as treatment attendance and treatment duration [2,3,11].The aim of the study was to test if patient-reported CORE-OM scales at the start of treatment were prospectively associated with treatment attendance and treatment duration at a Community Mental Health Centre (CMHC).We hypothesised that higher CORE-OM non-risk scores would be associated with longer treatment duration and that higher risk scores would be associated with increased no-show and drop-out rates.

Sample
The present study is a longitudinal study based on CORE measures completed by therapists and patients at the start of treatment and administrative data from the electronic patient record system at a Community Mental Health Centre (CMHC) in Western Norway.The CMHC offered in-and out-patient specialist services for a range of mental health conditions, including depressive and anxiety disorders, psychosis and attention-deficit/hyperactivity disorder.Most treatment was provided during individual consultations at the ambulatory clinic.Additional group courses focused on gaining insight, self-help in mastering symptoms and problems, mindfulness and relaxation techniques.Group sessions were provided within fixed formats of 3, 12 or 15 weeks.Patients' demographic and clinical information at baseline was gathered by therapists using the CORE Therapy Assessment Form [3].Patients completed the CORE-OM self-report questionnaire [2,12].In all, 756 consecutive patients were mailed an invitation to participate in the study, and 179 (24%) accepted the invitation.The valid analysis file included 124 (16%) patients who had completed CORE-OM self-report questionnaires with valid responses and who had valid information about treatment duration in the electronic patient journal system (Table 1).Fifty-six of the 124 patients had individual consultations with their therapists, 10 participated in group treatment only and 58 had both individual and group treatment.All participating patients gave written informed consent to participate.Approval for the study was sought from the Regional Committee for Medical Research Ethics (reference no.2009/1295).This body referred the study to the Norwegian Social Science data Services (NSd), which granted approval (reference no.22920/2).

Clinical Outcomes in Routine Evaluation-Outcome Measure
The CORE system includes three measures: the therapist-completed assessment forms designed to be administered before and after therapy (the CORE Therapy Assessment Form and the CORE End of Therapy Form), and a self-report outcome measure (CORE-OM) questionnaire completed by patients [3].The 34-item CORE-OM questionnaire addresses the domains of subjective well-being, problems and symptoms, functioning and risk [2,3,11,12,15].The measure has a solid theoretical and empirical basis, and its psychometric properties are sound [3,15].
The questionnaire items pertain to the last week.The Norwegian paper and pencil version of the CORE-OM was used to measure psychological distress [15].Patients completed the questionnaire at the start of treatment.Items were scored on a 5-point Likert scale from 0 = Never to 4 = Almost all the time.The eight positively phrased items were reverse scored so that higher scores meant a higher level of psychological distress on all items/scales.
The CORE-OM includes four domains represented by the following subscales: subjective well-being (4 items); functioning (12 items); problems/symptoms (12 items); and risk (6 items).Subjective well-being refers to a patient's sense of life quality and emotional health.Problems/symptoms refer to psychological health issues, such as anxiety and depression symptoms, reactions to trauma and physical complaints.Functioning refers to interpersonal, social and general functioning in daily life.Within the problems/symptom domain, item clusters address anxiety (4 items), depression (4 items), physical problems (2 items) and trauma (2 items).In the the functioning domain item clusters address general functioning (4 items), close relationships (4 items) and social relationships (4 items).
The three subscales subjective well-being, problems/symptoms and functioning can be combined into a general psychological distress scale, non-risk/all minus risk (28 items).The risk/harm portion (6 items) includes items covering harm to self and suicidal ideation (risk-to-self items) and violent behaviour and threats towards other people (risk-to-others items).Finally, all 34 items combined constitute the all items total scale.M: mean; SD: standard deviation; icd-10: international statistical classification of diseases and related health Problems, 10 th revision.a four participants with missing responses.b one participant with missing response.c frequency of respondents scoring above alcohol use disorders identification test (audit) cut-off of 8 for men and 6 for women for medium level of alcohol problems (Who: [13]).d frequency of respondents scoring above drug use disorders identification test (dudit) cut-off of 6 for men and 2 for women for drug related problems [14].
The psychometric properties of the CORE-OM are good in terms of internal consistency, test-retest reliability, discriminant and convergent reliability, sensitivity and specificity when used to classify patients according to clinical versus non-clinical degree of symptomatology [3,12,15].descriptive statistics of the valid CORE-OM responses (N = 124) are shown in Table 2.All scales were calculated as means across items, that is, by dividing the sum of item scores by the number of completed items [3,15].The subjective well-being (4 items) and risk (6 items) subscales each had ≤1 missing item, and the 12-item subscales each had ≤3 missing items.

Outcome variables
Administrative personnel at the CMHC extracted the following information about patients from the electronic patient journal system: date of referral, date of first and last consultation, individual or group consultation, number of consultations offered, number of consultations attended, number of consultations not attended and termination of treatment (planned versus unplanned).ICd-10 diagnoses at the end of treatment were extracted from the electronic patient record system.

Treatment duration
A variable representing treatment duration was computed by subtracting the date of the first consultation from the date of the last consultation registered in the electronic patient journal system.The variable was used as outcome variable in the analyses.

Number of consultations and treatment attendance
A variable representing the total number of individual or group consultations patients attended was used as outcome variable in the analyses.Furthermore, a variable representing the fraction of the number of consultations attended, that is, the number of consultations attended divided by the number of consultations offered, was computed and used as outcome variable in the analyses.

Statistical analyses
The internal consistency of the CORE-OM scales was assessed using the Cronbach's alpha.Cronbach's alpha between 0.6 and 0.8 is usually deemed as acceptable and values above 0.8 as good.Associations between predictor variables and treatment duration, number of therapy sessions or the fraction representing therapy attendance (i.e.number of therapy sessions attended divided by number of sessions offered) as outcome variables were tested in linear regression models.In these models, CORE-OM scales were entered as predictor variables, and treatment duration, number of consultations or fraction of attended and offered consultations were entered as outcome variables.Outcome variables were skewed, and therefore reported effect sizes and 95% confidence intervals (CIs) are from analyses performed with non-transformed outcome variables and P-values are from analyses with squared-, log-or square root-transformed outcome variables.The relationship between CORE-OM scales as predictor variables and unplanned ending of treatment as the outcome variable (unplanned = 1, planned = 0) was tested using logistic a regression model.All models were adjusted for patients' age, sex and level of education.Analyses were performed in STATA v. 17.0 (StataCorp LLC, Texas, uSA).Tests were two-tailed, and the P-level was set at .05.

Results
Sociodemographic and clinical characteristics of the total valid sample (N = 124) are shown in Table 1.Table 2 shows descriptive statistics of item scores, domain subscales and the all items scale of the patient-reported CORE-OM.Cronbach's alpha was 0.66 for the subjective well-being subscale, 0.86 for the problems/symptoms subscale, 0.77 for the functioning subscale, 0.78 for the risk subscale, 0.91 for the non-risk subscale and 0.92 for the all items scale of the CORE-OM.descriptive statistics of treatment duration, number of sessions offered and attended, and the way treatment was terminated is presented in Table 3 according to treatment format (individual-, group-, or both).

Treatment duration
Consistent and significant prospective associations were found between all CORE-OM scales at the start of treatment and treatment duration in the linear regression analyses (crude B = 3.3 (95% confidence interval (CI) for B = 0.83; 5.83)), p < .05for the CORE-OM subjective well-being subscale; B = 2.9 (95% CI = 0.43; 5.44), p < .05for the problems/symptoms subscale; and B = 3.4 (95% CI for B = 0.15; 6.62), p < .05for the functioning subscale) (Table 4).Effect sizes were similar for the risk and non-risk subscales.Adjustment for age, sex and level of education did not lead to notable changes in effect sizes for the associations.
Figure 1 shows the duration of treatment (months in intervals) for the total sample (N = 124).The mean duration of treatment was 13.9 months (standard deviation (Sd)=9.51,range 0.5-52.2),and the median duration was 12.2 months.

Number of treatment sessions
The linear regression models with CORE-OM scales as predictor variables and the number of treatment sessions as the outcome variable were all non-significant.However, with the exception of the CORE-OM functioning subscale, there were trends towards statistical significance in the adjusted models (Table 4).

Treatment attendance
No prospective associations were found between CORE-OM subscales at the start of treatment and treatment attendance operationalised as a fraction of the number of therapy sessions attended/offered (all p > .05)(Table 4).
Median treatment attendance for individual consultations expressed as fraction (i.e. the number of sessions attended divided by the number of sessions offered) was 0.77 (mean 0.76, Sd = 0.19, range 0-1) (N = 112).The median fraction of group consultations attended/offered was 0.77 (mean 0.74,

Termination of treatment
Patients with higher CORE-OM scale scores did not have significantly higher odds ratios (ORs) for unplanned ending of treatment (all p > .05)(Table 5).Eleven (9%) of the 124 patients in the valid analysis file dropped out of treatment without a planned ending of their therapy.

Discussion
In the present study, patients' self-reported levels of psychological distress as measured by the CORE-OM at the start of treatment were prospectively associated with treatment duration at a CMHC in Norway.This finding is in line with our hypothesis that higher CORE-OM scales are associated with longer treatment duration.Effect sizes for the associations were similar for both the risk-and non-risk subscales of the CORE-OM.However, the CORE-OM scales were not related to the number of treatment sessions, the degree to which patients met for their planned therapy sessions or the unplanned ending of treatment.

Treatment duration
Previous studies on the duration of mental health treatment have shown that higher age [16], problems with living conditions [17], a diagnosis of psychosis [17], higher distress levels [16] and a previous higher number of acute psychiatric inpatient bed days [17] are associated with longer treatment duration.Prior visits to the emergency department and self-harm are predictors of shorter treatment duration [17].Crits-Christoph showed that receiving a non-preferred treatment at any point was a significant predictor of longer treatment duration in a community mental health setting for patients with depression [18].Our expected finding that higher levels of psychological distress is associated with longer treatment is in line with Evans et al. [8] , who found a correlation between the scores of the 10-item short version of the CORE-OM and the treatment duration of patients being treated by secondary care adult psychological therapy teams [8].The study by Evans et al. and our findings suggest that at the group level, patients who report higher levels of psychological distress at the start of treatment are in need of longer treatment contact.Consequently, the self-reported CORE-OM may help therapists and administrators identify patients at risk of enduring therapy at an early point during the patients' contact with specialist mental health services.Being aware of which patients it is that are at risk of long treatment duration may help therapists in planning therapy.Furthermore, higher CORE-OM scores at the start of treatment may indicate the need for involving additional services in the community at an early point in treatment.
Our study did not confirm the association between CORE-OM and the number of therapy sessions found by Evans et al. [8].The lack of such an association in our data implies that recovery time may be independent of therapy intensity.The lack of an association between self-reported psychological distress and number of therapy sessions in our study is in line with earlier findings showing that, on average, clients who have widely different numbers of sessions have similar rates of recovery and improvement [7,19].These findings suggest that patients improve at different rates and leave therapy when they have reached an acceptable level of psychological well-being.For system administrators, the findings imply that attempts to prescribe fixed numbers of therapy sessions may overlook the fact that patients' needs for treatment vary and that recovery time may be irrespective of the number of sessions or treatment intensity.

Strengths and limitations
One strength of the study is that the CORE-OM is a theoretically and empirically sound measure of self-reported psychological distress [3,15].
The naturalistic setting of our study may be a strength, as it contributes to the external or ecological validity of the study and may increase the generalisability of the findings to other public specialised mental health centres.However, the study sample was very heterogeneous with regard to age, diagnosis, sociodemographic and clinical variables.unfortunately, this heterogeneity may reduce the transferability of the findings to specific diagnostic or clinical subgroups.
One limitation of the study may be inclusion bias.Only about one out of six of the patients who were invited to participate in the study were included in the valid analysis file, making the analyses vulnerable to systematic differences between participants and non-participants.For example, patients with more severe mental health conditions or substance abuse issues may be underrepresented in the valid sample.Given the presence of these kinds of biases in our data, the associations we found may be weaker than in the total patient population due to lower variance in CORE-OM variables in the present study.Inclusion biases may have reduced the generalisability of findings to patient sub-populations that probably were under-represented in our study.
Other weaknesses include probable under-reporting of the unplanned ending of treatment and a lack of statistical power in the logistic regression analyses with unplanned ending of treatment as outcome measure.The results of these analyses should therefore be interpreted with caution due to the risk of type II errors.Further, confounding factors such as quality of treatment, patients' living and work situations and use of community mental health services in addition to service at the CMHC may have affected the associations between CORE-OM subscales and outcome variables.For instance, higher complexity of the patients' problems may be associated with need for extended treatment, irrespective of symptom levels as measured by CORE-OM.Finally, we would like to make clear that although significant associations between CORE-OM scales and relevant outcome variables were found in our study, these associations do not imply that there are causal relationships between these variables.

Conclusion
In the present study, higher patient-reported psychological distress as measured by the CORE-OM at the start of treatment was prospectively associated with longer treatment duration but not with treatment attendance or treatment  drop-out.The findings imply that patients with higher psychological distress as assessed by the CORE-OM at start of their contact with mental health services need longer treatment, but that treatment attendance and dropping out are explained by factors other than the severity of distress.As CORE-OM often is used routinely in the specialist mental health services, this new knowledge may help therapists and administrators estimate patients' need for treatment at the individual level and to plan treatment at the system-level.

Table 1 .
sociodemographic and clinical characteristics of the valid sample (N = 124).

Table 2 .
Patient-reported psychological distress as measured by corE-oM item scores a , domain subscales and total scale (N = 124).

Table 3 .
treatment duration, number of sessions offered and attended and termination of treatment (N = 124).descriptive statistics include only 54 patients with valid data on number of sessions. a

Table 4 .
linear regression models with corE subscales as predictor variables and treatment duration, number of therapy sessions and treatment attendance (number of therapy sessions attended/offered) as outcome variables (N = 124).