The authors have declared that no competing interests exist.
Conceived and designed the experiments: DG RM c. Performed the experiments: RM TK LH SP. Analyzed the data: DG. Contributed reagents/materials/analysis tools: DG SP. Wrote the paper: DG RM TK LH AF SP.
Having friends is associated with more favourable clinical outcomes and a higher quality of life in mental disorders. Patients with schizophrenia have fewer friends than other mentally ill patients. No large scale studies have evaluated so far what symptom dimensions of schizophrenia are associated with the lack of friendships.
Data from four multi-centre studies on outpatients with schizophrenia and related disorders (ICD F20-29) were included in a pooled analysis (N = 1396). We established whether patients had close friends and contact with friends by using the equivalent items on friendships of the Manchester Short Assessment of Quality of Life or of the Lancashire Quality of Life Profile. Symptoms were measured by the Brief Psychiatric Rating Scale or by the identical items included in the Positive and Negative Syndrome Scale.
Seven hundred and sixty-nine patients (55.1%) had seen a friend in the previous week and 917 (65.7%) had someone they regarded as a close friend. Low levels of negative symptoms and hostility were significantly associated with having a close friend and contact with a friend. Overall, almost twice as many patients with absent or mild negative symptoms had met a friend in the last week, compared with those with moderate negative symptoms.
Higher levels of negative symptoms and hostility are specifically associated with the lack of friendships in patients with psychotic disorders. These findings suggest the importance of developing effective treatments for negative symptoms and hostility in order to improve the probability of patients with schizophrenia to have friends.
Friendship can be defined as a “distinctively personal relationship that is grounded in a concern on the part of each friend for the welfare of the other, for the other’s sake, and that involves some degree of intimacy”
People with psychotic disorders tend to have fewer friends and social relationships compared to the general population and to patients with other mental and physical disorders
While many factors, such as deficits in neurocognition and social cognition, unemployment, financial difficulties and stigma are likely to reduce patients’ social functioning
However, although many studies have assessed the relationship between psychotic symptoms and patients’ global social networks, few data are available on the associations of symptoms specifically with friendship, with its characteristics of an intimate and supportive relationship.
To our knowledge, only one mixed-methods study, carried out on 151 patients with schizophrenia in south England, has specifically focused on relationships with friends of patients with schizophrenia
Evidence from larger samples is necessary to further understand how different symptoms are specifically associated with contacts with friends. Given the protective effects of friendship, interventions to improve patients’ friendships and, as a consequence, social support, clinical outcomes, and quality of life may need to consider specific symptom dimensions.
This study assessed, through a pooled analysis of individual patient data from four Europe-wide multicentre studies, the association of five symptom dimensions of psychotic disorders (negative symptoms, thought disorders, depression/anxiety symptoms, activation and hostility) with having a close friend and contacts with friends in the community. As age and gender have been found to be associated with patients’ social contacts in previous studies
For this study we analysed data from four multi-centre studies, i.e. one cluster randomized controlled trial and three prospective observational studies.
The DIALOG study
The “Nordic multicentre study”
The “EUNOMIA” study
The “InvolvE” study
Rationale, methods and findings of the studies have been published elsewhere
All studies included in this pooled analysis have obtained approval of relevant ethics committees, and all patients provided written informed consent.
Studies used either the Lancashire Quality of Life Profile (LQOLP) or the Manchester Short Assessment of Quality of Life (MANSA) instruments which contain equivalent items for assessing patients’ friendships
In two studies
Stata 12 for Windows was used for all data analyses
Four datasets were pooled to identify main effects of symptom domains and also interaction effects of symptom domains with age, gender or both. A “pooled analysis”
The correlation between the behavioural item, i.e. contact with friends in the last week, and the subjective item, i.e. patients’ subjective appraisal of having a close friend was explored by the phi test. Univariable and multivariable logistic mixed models, adjusted for heterogeneity across centres and studies, were used to identify the factors associated with behavioural and subjective item for friendship. The multivariable model, adjusted for confounding factors (patients’ age and gender), included all BPRS subscales. In this three-level model, patient-level measurements (level-1) were treated as nested within centres (level-2), and centres as nested within studies (level-3). To illustrate the size and clinical relevance of possible associations between symptom domains and friendship, we divided average scores of the given BPRS subscale in six intervals (i.e. 1, from 1 to 2, 2 to 3, 3 to 4, 4 to 5, 5 to 6) and showed the percentages of patients who had seen a friend in the last week and had a close friend for each symptom interval. For each interval, the higher number was included in the lower interval (i.e. 2 was included in the interval from 1 to 2) and there were no values higher than 6. Then, dichotomous variables were created for each subscale that had a statistically significant association with the two friendship items. In these dichotomous variables all the values of BPRS subscales that were lower than 2 (with 2 included) were coded as “1″ and all the values that were higher than 2 as “0″. The univariable associations of these variables with friendship items were tested by mixed logistic regression models, adjusted for heterogeneity of centres and studies.
Two-way interactions for age and gender were tested to establish whether they influenced associations between BPRS subscales and patients’ contact with friends. Statistical significance of interaction terms was assessed using Wald tests.
Since the sample contained patients with different diagnoses within the spectrum of schizophrenia and related disorders, we conducted a sensitivity analysis, repeating the analyses in the pooled sample only with those patients who had a diagnosis of schizophrenia (F20 according to the ICD-10).
Across the studies, a total of 1396 patients met the inclusion criteria (n = 502 from the DIALOG study; n = 341 from the Nordic Multicentre study; n = 352 from the EUNOMIA study; n = 201 from the InvolvE study). Patients were predominantly male (844, 60.5%), with a mean age of 39.9 years (SD = 11.1). The age span in years was 18–64 in EUNOMIA study (median = 39, quartiles = 29–49), 18–65 in DIALOG study (median = 42, quartiles = 33.5–50), 18–64 in the INVolvE study (median = 36, quartiles = 26–45) and 20–55 in the Nordic Multicentre study (median = 40, quartiles = 32–46).
Seven hundred and sixty-nine patients (55.1%) had seen a friend in the previous week and 917 (65.7%) had someone they regarded as a close friend. Overall patients showed low scores on different BPRS subscales and the distribution of these scores were skewed to the left. The mean scores of the BPRS subscales were: 2.1 (SD = 0.9) on the depression/anxiety subscale; 1.9 (SD = 0.9) on the negative symptoms subscale; 1.9 (SD = 1.0) on the thought disorders subscale; 1.5 (SD = 0.6) on the activation subscale; 1.5 (SD = 0.7) on the hostility subscale.
The main socio-demographic and clinical characteristics of the individual studies and samples and of the pooled sample are reported in
Study sample | DIALOG study19 | Nordic multicentre study20 | EUNOMIA study21 | InvolvE study22 | Total sample |
|
UK, Spain, Netherlands,Sweden, Germany,Switzerland | Sweden, Denmark,Finland, Iceland,Norway | Germany, Poland, Slovakia,Czech Republic, Lithuania,Sweden | England | - |
|
502 | 341 | 352 | 201 | 1396 |
|
MANSA | LQOLP | MANSA | MANSA | – |
|
Randomized controlled trial | Prospective-observational | Prospective-observational | Prospective-observational | – |
|
303 (60.4) | 159 (46.5) | 205 (58.2) | 103 (51.2) | 769 (55.1) |
|
321 (63.9) | 223 (65.2) | 239 (67.9) | 135 (67.2) | 917 (65.7) |
|
42.1 (11.4) | 38.9 (8.7) | 39.7 (11.8) | 36.5 (11.4) | 39.9 (11.1) |
|
169 (33.7) | 133 (38.9) | 192 (54.5) | 59 (29.4) | 552 (39.5) |
|
2.3 (0.9) | 2.3 (0.9) | 1.8 (0.8) | 2.1 (1.0) | 2.1 (0.9) |
|
2.1 (0.9) | 2.0 (0.8) | 1.7 (0.8) | 1.7 (0.8) | 1.9 (0.9) |
|
2.1 (1.0) | 2.0 (1.0) | 1.4 (0.6) | 1.8 (1.1) | 1.9 (1.0) |
|
1.4 (0.6) | 1.9 (0.8) | 1.3 (0.4) | 1.4 (0.6) | 1.5 (0.6) |
|
1.4 (0.6) | 1.7 (0.8) | 1.4 (0.6) | 1.7 (0.9) | 1.5 (0.7) |
Sample size refers to included patients with an ICD-10 clinical diagnosis of schizophrenia, schizotypal, or delusional disorders, for which BPRS-18 and MANSA/LQOLP items on friendship scores were available.
LQOLP, Lancashire Quality of Life Profile; MANSA, Manchester Short Assessment of Quality of Life.
The behavioural (having seen a friend) and subjective (having a close friend) items on friendship were significantly correlated (phi = .589, p<.001).
As shown in
Odds ratio | Odds ratio(95% CI |
P | |
BPRS - depression/anxiety subscale | .856 | .759–.966 | .012 |
BPRS - negative symptoms sub scale | .618 | .538–.710 | <.001 |
BPRS - thought disorder sub scale | .806 | .715–.907 | <.001 |
BPRS - activation sub scale | .690 | .570–.835 | <.001 |
BPRS - hostility sub scale | .714 | .607–.840 | <.001 |
Patients’ age | .979 | .970–.989 | <.001 |
Patients’ gender | .810 | .648–1.013 | .065 |
CI = Confidence Interval.
Adjusted for BPRS-18 subscales, patients’ age and gender | |||
Odds ratio | Odds ratio (95% CI |
P | |
BPRS - depression/anxiety subscale | .983 | .860–1.124 | .804 |
BPRS - negative symptoms sub scale | .693 | .602–.797 | <.001 |
BPRS - thought disorder subscale | .971 | .847–1.114 | .679 |
BPRS - activation sub scale | .870 | .708–.1.070 | .186 |
BPRS - hostility sub scale | .823 | .680–.996 | .046 |
Patients’ age | .980 | .970–.990 | <.001 |
Patients’ gender | .835 | .659–1.058 | .135 |
Sigma_u | .226 | .093–.551 | |
Rho | .015 | .003–.084 |
CI = Confidence Interval.
The univariable and multivariable models that tested the associations between symptom domains and the subjective appraisals of patients of having a close friend are reported in
Odds ratio | Odds ratio(95% CI |
P | |
BPRS - depression/anxiety subscale | .902 | .795–1.023 | .108 |
BPRS - negative symptomssub scale | .638 | .555–.734 | <.001 |
BPRS - thought disorder sub scale | .823 | .729–.930 | .002 |
BPRS - activation sub scale | .761 | .629–.921 | .005 |
BPRS - hostility sub scale | .712 | .604–.838 | <.001 |
Patients’ age | .984 | .974–.994 | .002 |
Patients’ gender | .732 | .577–.930 | .011 |
CI = Confidence Interval.
Adjusted for BPRS-18 subscales, patients’ age and gender | |||
Odds ratio | Odds ratio (95% CI |
P | |
BPRS - depression/anxiety subscale | 1.005 | .872–1.158 | .944 |
BPRS - negative symptoms sub scale | .676 | .583–.783 | <.001 |
BPRS - thought disorder subscale | .960 | .832–1.107 | .574 |
BPRS - activation sub scale | .963 | .772–.1.200 | .738 |
BPRS - hostility sub scale | .813 | .670–.988 | .037 |
Patients’ age | .983 | .972–.994 | .002 |
Patients’ gender | .758 | .588–.977 | .032 |
Sigma_u | .290 | .153–.547 | |
Rho | .025 | .007–.083 |
CI = Confidence Interval.
Higher levels of all the symptom domains, with the exception of depression/anxiety symptoms, were univariably associated with the absence of a close friend. When controlling for age and gender and adjusting for heterogeneity of centres and studies, only the associations of the absence of close friendships with higher levels of negative symptoms (OR = .676; 95% CI = .583–.783; p<.001) and hostility (OR = .813; 95% CI = .670–.988; p = .037) held true. Younger patients were more likely to have a close friend (OR age = .983; 95% CI = .972–.994; p = .002). Male patients reported less frequently than female patients that they had a close friendship (OR = .758; 95% CI = .588–.977; p = .032).
The number and percentage of patients who had contact with friends in the previous week and who stated that they had a close friend are reported in
BPRS subscales score | Have you seen a friend in the last week? | |||||
Negative symptoms | Hostility symptoms | |||||
Intervals | Total |
Yes |
Yes |
Total |
Yes |
Yes |
1 | 321 | 214 | 66.7 | 550 | 328 | 59.6 |
>1 and < = 2 | 545 | 311 | 57.1 | 610 | 336 | 55.1 |
>2 and < = 3 | 369 | 177 | 48.0 | 179 | 84 | 46.9 |
>3 and < = 4 | 125 | 48 | 38.4 | 41 | 14 | 34.1 |
>4 and < = 6 | 19 | 7 | 36.8 | 9 | 2 | 22.2 |
Number of patients at each interval of BPRS subscales score.
Number of patients at each interval of BPRS subscales score who reported to have seen a friend in the last week.
Percentage of patients at each interval of BPRS subscales score who reported to have seen a friend in the last week.
BPRS subscales score | Do you have anyone you would call a close friend? | |||||
Negative symptoms | Hostility symptoms | |||||
Intervals | Total |
Yes |
Yes |
Total |
Yes |
Yes |
1 | 321 | 247 | 77.7 | 550 | 392 | 71.3 |
>1 and < = 2 | 545 | 366 | 67.2 | 610 | 388 | 63.6 |
>2 and < = 3 | 369 | 216 | 58.5 | 179 | 105 | 58.7 |
>3 and < = 4 | 125 | 67 | 53.6 | 41 | 21 | 55.3 |
>4 and < = 6 | 19 | 8 | 42.1 | 9 | 5 | 55.6 |
Number of patients at each interval of BPRS subscales score.
Number of patients at each interval of BPRS subscales score who reported to have a close friend.
Percentage of patients at each interval of BPRS subscales score who reported to have a close friend.
Patients with very low levels of negative symptoms (lower than “2” at BPRS negative symptoms subscale) had almost double the odds of having met a friend in the previous week (OR = 1.745; IC 95% = 1.399–2.176, p<.001) and of having a close friend (OR = 1.838; IC 95% = 1.461–2.313, p<.001) compared to those with higher levels of negative symptoms. The odds ratios are adjusted for heterogeneity of centres and studies.
Patients with very low levels of hostility (lower than “2” on BPRS hostility symptoms subscale) had higher odds of having met a friend in the previous week (OR = 1.520; 95% CI = 1.139–2.028, p = .004) and of having a close friend (OR = 1.498; 95% CI = 1.117–2.011, p = .007) compared to those with higher levels of hostility. The odds ratios are adjusted for heterogeneity of centres and studies.
Among the 770 patients who had very low or absent levels both of negative symptoms and hostility (BPRS subscale scores lower than two), 477 (61.9%) had seen a friend in the last week and 549 (71.3%) had someone they regarded as a close friend.
Patients with at least low-moderate levels of both negative symptoms and hostility (BPRS subscales score higher than two) were 138. Among them, 57 (40.6%) had seen a friend in the last week and 73 (52.1%) had someone they regarded as a close friend.
The association of higher levels of negative symptoms with contacts with friends in the last week was neither influenced by age (
Similarly, no influence of age and gender on the association of higher levels of hostility with contacts with friends was found (Wald test values were Z = 0.75, p = .452 for age and Z = 1.73, p = .084 for gender, respectively).
The association of higher levels of negative symptoms with having no close friends was not influenced by age (Z = −0.60; p = .546) or gender (Z = −0.25; p = .806). The interaction of gender on the association of hostility with having no close friends was not statistically significant (Z = 1.28; p = .201). However, in younger patients the association between hostility and absence of close friendships was stronger (Z = 2.68, p = .007).
When we repeated the analyses in the pooled sample only with those patients who met the criteria for schizophrenia (F20) (n = 1019), the association of higher levels of negative symptoms and hostility with contacts with friends in the last week and the subjective appraisals of having a close friend remained statistically significant. No other symptom domain was significantly associated with friendships in the multivariable analysis. No interactions between symptoms and socio-demographic variables were found. In particular, the interaction between hostility levels and age, that was statistically significant in the global sample, failed to reach statistical significance in the subsample of patients with schizophrenia.
This is the largest study to date analyzing how specific psychotic symptoms are associated with social contacts of patients with schizophrenia related disorders and the first one focussing specifically on friendships, as a specific and relevant sub-category of social contacts.
Higher levels of negative symptoms and hostility are associated with fewer contacts with friends and absence of close friendships. The association between negative symptoms and contacts with friends is more marked in male patients. Depression/anxiety symptoms, thought disorders and levels of activation were only univariably associated with patients’ friendships. When the associations were adjusted for the influence of other symptoms, no significant association was found between depression/anxiety symptoms, thought disorders and activation and friendships.
Despite the suggestion that all symptoms of schizophrenia are likely to have an impact on patients’ social relationships
We found a high number of patients did not have a close friend and did not see any friend in the previous week, which is in line with the previous literature
The role of hostility in influencing the size of patients’ social networks and their social functioning has already been reported by other studies
While the association of hostility levels with contacts with friends may look rather intuitive, the link of negative symptoms with friendships probably deserves some further exploration. Cognitive-behavioural models
High levels of activation (i.e. excitement, tension, mannerism and posturing) and thought disorders (i.e. thought content disorders, conceptual disorganization, hallucinations and grandiosity) might lead other people to believe that patients are unpredictable and, possibly, dangerous
In our sample, younger patients were more likely to have had recent contacts with friends than older patients. This finding might be interpreted as a consequence of the progressive deterioration of one’s social network related to psychotic disorders
Male patients had less frequently someone they regarded as a close friend. This may be due to a greater fear of intimacy and lower levels of emotional commitment in relationships of males, documented in studies on clinical and non-clinical populations
Although the cross-sectional nature of the study does not allow conclusions to be drawn on causal relationships, we found significant associations of higher levels of negative symptoms and hostility with lack of friendships. Despite suggestions that all symptoms of psychotic disorders can have a role in patients’ difficulties in establishing and maintaining social relationships, the levels of negative symptoms and hostility may be specifically associated with the disruption of more intimate social relationships, i.e. friendships, that can be important sources of social support. It might be hypothesized that a full reduction of moderate to severe symptom levels in these two domains might have a relevant impact on the patient’s chances to have and meet friends.
There is limited evidence on the effectiveness of available treatments specifically for hostility. A range of antipsychotics have been suggested as effective
On the other hand, the treatment of negative symptoms is particularly challenging. Intensive psychosocial treatment has been found to have a beneficial effect on negative symptoms
The limits in effectiveness of available therapies may pose patients with high levels of these symptoms at high risk of poorer psychological and physical health outcomes
Furthermore, even among patients with no or very low levels of both negative symptoms and hostility, about 38% did not see a friend in the last week and 29% reported not having a close friend. It is possible that other factors, such as impairment of neuro-cognitive performance and deficits in social cognition
The large sample size of this study provided a substantial statistical power for multivariable tests, including the testing of interaction effects, and ensured the validity and generalisability of both positive and negative findings. This is also the first multicentre assessment of friendship as an intimate relationship, with its potential to provide support to patients
However, some limitations should be noted: a) Study patients were not representative of all patients in the given service. Selection biases might have influenced the final scores of items on friendship and of BPRS subscales. Nevertheless, the aim of the study was to assess associations between clinical symptoms and patients’ contacts with friends, and associations are usually more robust towards selection bias than absolute levels
Treatment of negative symptoms and hostility may be important for enabling patients with psychotic disorders to engage in friendships. However, the limited effectiveness of currently available treatments for negative symptoms, the reduced adherence to treatment of patients with high levels of hostility, and the significant number of patients that do not have friendships despite low levels of negative symptoms and hostility, suggest that further therapeutic interventions and support need to be developed to address the difficulties of patients with schizophrenia in establishing and maintaining friendships.
Experimental studies are required to longitudinally explore the correlations between symptom domains of psychotic disorders and friendships and to assess to what extent effective treatment of negative symptoms and hostility might indeed be followed by more patient friendships.
(DOC)
The authors gratefully acknowledge the advice of Dr. Stephen Bremner on statistical analysis.