Relationships between Empathy, Symptoms, and Functioning in First Episode Psychosis

1. Introduction

Social cognition is an important predictor of functional outcomes for those with schizophrenia, accounting for more variance in social functioning than neurocognition (Fett et al., 2011). Social cognition has been investigated among those with first-episode psychosis (FEP) and is similarly linked to social functioning (Addington et al., 2006; Bonfils, Ventura, et al., 2019; Stouten et al., 2017), though one domain of social cognition remains underexplored - empathy. Empathy is a broad construct referring to our abilities to both know how others feel (cognitive empathy) and to experience emotions with others (affective empathy; Decety & Jackson, 2006). Empathy is impaired among those with schizophrenia (Bonfils et al., 2016, 2017; Derntl et al., 2009) and linked to clinical and functional outcomes (Bonfils, Lysaker, et al., 2019; Horan et al., 2015; Michaels et al., 2014; Pijnenborg et al., 2013). Work is needed to understand whether empathy is linked to these important illness characteristics among those with FEP.

Though clinically relevant, the construct of empathy is under-represented in the field of major psychopathology. Indeed, social cognition is a major focus of research among those with schizophrenia, but seminal papers have not included empathy as a construct of interest (e.g., Green et al., 2008; Pinkham et al., 2018), focusing instead on emotion processing, theory of mind, social perception, and social knowledge. Theory of mind shares features with cognitive empathy, as include the ability to understand others’ emotions, but other aspects are distinctive (Dvash & Shamay-Tsoory, 2014). While empathy holds historical significance, noted in early works by Kraepelin (1919) and Bleuler (Bleuler & Zenkin, 1911) as an area of impairment linked to social difficulties in schizophrenia, today’s scholars have struggled to define and measure empathy in a manner comparable to other social cognitive constructs. Indeed, in schizophrenia research, the predominant approach to measuring social cognition is performance-based. This is preferred, given the known discrepancy between how people with schizophrenia report on their abilities and their actual performance (Bonfils et al., 2016; Sabbag et al., 2012). However, though performance-based measures exist (e.g., Derntl et al., 2009; Lee et al., 2011), empathy is most often measured via self-report. Self-reported empathy questionnaires measure something meaningful, given consistent correlations with important outcome variables (Horan et al., 2015; Michaels et al., 2014; Pijnenborg et al., 2013), and may be the most appropriate method for measuring empathy as a construct. Unlike performance-based measures, self-report questionnaires can ask about tendencies to engage in certain acts, which may better reflect how one uses empathy in daily life (Davis, 1983). Performance-based measures also require differentially valued answers (e.g., right or wrong, graded responses), which is not a requirement for self-report instruments. This may be key for measurement of empathy, particularly affective empathy, as accuracy or ability may be less relevant when compared to one’s tendencies to share emotions, a phenomenon hard to replicate in the lab (Zaki & Ochsner, 2012).

While empathy has been examined more extensively in participants with or at-risk for schizophrenia (Bonfils et al., 2017; Kuis et al., 2021; Montag et al., 2020), few published studies have examined FEP-specific samples. Achim and colleagues (Achim et al., 2011; n=31) found that self-reported cognitive and affective empathy did not differ between FEP and healthy control participants (n=31), and that cognitive empathy was linked to social anxiety (Achim et al., 2011). Pepper and colleagues (2019) found that those with early psychosis (n=51) had worse cognitive empathy than neurotypical controls (n=26). Canty and colleagues (2021) found that cognitive empathy predicted significant variance in community functioning for both early (n=26) and chronic schizophrenia (n=32) samples, even when accounting for theory of mind. While preliminary work points to the importance of empathy for those with FEP, no study to date has comprehensively examined the intersection of self-reported empathy and both clinical and functional variables.

Research is particularly lacking in FEP samples using the Interpersonal Reactivity Index (IRI; Davis, 1983), the most commonly used self-report measure of empathy in schizophrenia samples (Bonfils et al., 2017), and the Questionnaire of Cognitive and Affective Empathy (QCAE; (Michaels et al., 2014; Reniers et al., 2011), a more recently developed measure. The IRI was built from a multidimensional model of empathy emphasizing both cognitive and affective aspects of one’s reactions to others in the social environment (Davis, 1983). The IRI includes four subscales, measuring perspective-taking, or the cognitive tendency to take the perspective of another person, fantasy, or the tendency to understand and relate to fictitious characters, empathic concern, or the affective tendency to feel other-oriented emotions during social interactions, and personal distress, or the affective tendency to feel self-oriented emotions during tense or uncomfortable social interactions (Davis, 1980, 1983). The QCAE was developed, in part, to improve upon the IRI and other empathy measures (Reniers et al., 2011). Like the IRI, the QCAE ascribes to a multidimensional model of empathy, including cognitive and affective components, but emphasizes a distinction from behavioral and sympathetic responses, which arise from empathy but do not represent internal empathic experiences (Reniers et al., 2011). The QCAE was developed by combining face-valid items from existing empathy measures (Michaels et al., 2014; Reniers et al., 2011); statistically-derived subscales include perspective-taking (tendency to take others’ perspectives), online simulation (tendency to attempt to mentally represent others’ emotions), emotion contagion (tendency to match self-oriented emotions of close others), proximal responsivity (emotional responsiveness to others’ moods), and peripheral responsivity (emotional responsiveness to distant others, such as fictitious characters) (Michaels et al., 2014; Reniers et al., 2011).

Comparison of the IRI and QCAE along with their correlates among those with FEP would represent a substantial increase in our knowledge about this important social cognitive domain in a vulnerable group. Accordingly, our aims were to 1) examine characteristics of two self-reported empathy measures, their inter-correlations, and their relationships with other social cognitive measures; 2) correlate empathy with symptomatic and functional variables; and 3) examine the incremental contribution of empathy to the prediction of clinical variables over and above other social cognitive constructs. We expected the empathy scores for the two scales to be inter-correlated and, based on prior research in schizophrenia samples (Bonfils, Lysaker, et al., 2019; Horan et al., 2015; Michaels et al., 2014), associated with negative symptoms and social functioning. Lastly, given some evidence to suggest affective empathy may be relatively intact in those with FEP (Achim et al., 2011; Atoui et al., 2018; Bonfils et al., 2016) and prior work primarily linking cognitive empathy to clinical and functional variables (Canty et al., 2021; Horan et al., 2015), we anticipated that cognitive facets of empathy would account for additional variance in clinical outcomes over and above other social cognitive constructs, whereas affective empathy would not.

2. Method

3. Analyses

Analyses were conducted in SPSS version 29 (IBM Corp., 2022) and JASP version 0.95.1 (JASP Team, 2021). To accomplish aim 1, we examined means and standard deviations of the IRI and QCAE and conducted two types of reliability analyses: Cronbach’s alpha (Cronbach, 1951) and McDonald’s omega (McDonald, 1999). Omega total (Revelle & Zinbarg, 2009) overcomes some limitations of alpha, including by accounting for differences in how individual items relate to the measured construct; i.e., omega does not assume tau equivalence (Dunn et al., 2014).

We next computed Pearson’s R correlations with two other measures of social cognition, the MSCEIT-ME and SAT-MC, and examined intercorrelations between the IRI and QCAE. To accomplish aim 2, we used Pearson’s R to examine associations between the empathy measures (IRI, QCAE) and everyday functioning and symptoms. Given the number of correlations examined, we conducted Benjamini-Hochberg corrections (Benjamini & Hochberg, 1995) as a sensitivity analysis to inform robustness of findings. To accomplish aim 3, we conducted hierarchical multiple linear regressions. To limit alpha inflation and ensure meaningful results, we selected models based on correlation results, such that dependent variables were selected based on whether they were related at a bivariate level to both an empathy construct (IRI, QCAE) and the MSCEIT-ME and/or SAT-MC. Models were then designed to test the contribution of the social cognition measure(s) in step 1, followed by the addition of empathy measure(s) in step 2. We examined total model statistics as well as R² change to determine if empathy measures contributed unique variance to relationships with functioning and symptoms.

4. Results

Participants identified as men (65%) and women (35%). Just under one third each identified as mixed race (31%) or White (29%), with the remainder identifying as Asian (18%), Black (10%), or Pacific Islander or Indian (3%). One third (33%) identified as Hispanic. On average, participants were 23.4 years old (SD = 5.4). The majority of participants were single (93%). Most participants (94%) were diagnosed with schizophrenia, with a minority diagnosed with schizoaffective disorder, depressed type (6%).

5. Discussion

Our results indicate that self-reported measures of empathy perform relatively poorly among FEP participants. Despite this limitation, we found that self-reported empathy in FEP is associated with clinical characteristics, specifically negative symptoms, with lower affective empathy emerging as the most consistent predictor of higher negative symptoms across the domains of alogia, anhedonia, and affective flattening. Moreover, IRI empathy subscales contributed to symptoms and functioning over and above performance-based social cognition measures, in one case accounting for an additional 18% of variance in anhedonia. These findings are notable, given the prevalence of anhedonia among those with schizophrenia (Krzyzanowski et al., 2022) and its links with quality of life, functioning, and other symptoms among those with FEP (Pelizza et al., 2020). Together, findings suggest that empathy is a construct worthy of continued focus in FEP.

This work extends our understanding of empathy measurement in FEP samples. While self-reported empathy measures predominate, we found that neither the IRI nor the QCAE met basic requirements for psychometric reliability (Nunnally & Bernstein, 1994), which was relatively low for subscales across both measures. The QCAE, though developed with the aim to improve empathy measurement, had several subscales with unacceptably low internal consistency, and the total score showed minimal associations with clinical or functional variables. While the IRI appears to be comparatively better suited for empathy research in FEP samples, past research has also revealed lower internal consistency estimates in psychosis populations (Bonfils et al., 2022; Bonfils, Lysaker, et al., 2019; Michaels et al., 2014). Thus, an important finding of this study is that novel or refined empathy measurement is needed for studies on FEP, and likely schizophrenia-spectrum disorders broadly.

Given the novelty of our findings and dearth of empathy research in FEP populations, our findings merit interpretation, despite poor psychometric performance of empathy measures. In particular, the dominance of affective empathy in our results is surprising, as past work has primarily supported links between cognitive empathy and clinical variables in both FEP (Achim et al., 2011; Canty et al., 2021) and schizophrenia (Horan et al., 2015). Indeed, while affective empathy deficits are established in schizophrenia (Bonfils et al., 2016), the effect size is small-to-moderate, and some have highlighted affective empathy as relatively preserved in this population (Horan et al., 2012). Our results suggest, however, that affective empathy may be linked to negative symptoms in FEP participants, across both experiential and expressive domains. Prior work has linked reduced empathy more strongly to worse experiential negative symptoms (Bonfils et al., 2021), but our data suggest that affective empathy adds to the prediction of both alogia and anhedonia, even when accounting for other social cognitive variables. While preliminary, this could suggest that one’s tendencies to share emotions with others are linked to the ability to generate and sustain emotional signals, both internally and externally (Andrzejewska et al., 2017). Moreover, our findings align with some work suggesting that a subset of negative symptoms may be the downstream consequence of social cognitive deficits (Pelletier-Baldelli & Holt, 2020). More work is needed to investigate these possibilities.

Of note, IRI-PD and IRI-F are arguably not core to the construct of empathy (Bonfils et al., 2022; Reniers et al., 2011), but demonstrated interesting relational patterns in our data. First, IRI-PD measures self-oriented distress, which may impair empathy rather than promote it (Batson et al., 1987; Eisenberg & Miller, 1987), instead of emotional matching, which promotes positive social bonding (Davis, 1983). Yet, in this sample, greater endorsement of PD was associated with better functioning in the domains of family relationships and friendships (trending). Given that PD was also associated with less affective flattening, these findings may reflect greater emotional range among participants endorsing higher PD. More work is needed to deepen our understanding of PD’s role in social functioning.

IRI-F, like PD, is also less central to the empathy construct. If we consider empathy to primarily refer to one’s ability to relate to other people in real life (Decety & Jackson, 2004), then Fantasy, which assesses connection to fictional characters, is instead an adjacent construct. In our data, higher IRI-F scores were related to worse alogia, anhedonia, and inattention. Indeed, for both anhedonia and alogia, along with IRI-EC, Fantasy remained a significant predictor even when accounting for other social cognitive constructs. It is possible that greater tendencies to relate to fictional characters may represent greater internal absorption (Rosen et al., 2017; Tellegen & Atkinson, 1974) in this sample, which could be driving relationships with negative symptoms. Of note, however, Fantasy was not linked to any positive symptom, suggesting that while higher Fantasy may be impairing, those with FEP who relate to fictional characters are not necessarily at heightened risk for delusional symptoms, such as ideas of reference.

This study has limitations. First and importantly, correlations between empathy measures and clinical characteristics did not survive FDR correction. Given the novelty of our findings and moderate effect sizes (Cohen, 1992), we interpreted uncorrected results. Future research should replicate our findings, with emphasis on adequate power to detect moderate effects when correcting for multiple comparisons. Second, while the MSCEIT-ME and SAT-MC are established social cognition measures, examining empathy alongside measures such as the Bell-Lysaker Emotion Recognition Task or the Hinting Task may have been a better test of its unique contributions, considering that these measures have been recommended after rigorous testing via the SCOPE study (Pinkham et al., 2018). Third, we were not able to closely examine comorbid conditions. Autism spectrum disorder, in particular, may be of interest as a comorbidity in future research, given known empathy impairments associated with this condition (Fatima & Babu, 2024). Lastly, this study did not include a healthy comparison group. While not examined statistically, IRI and QCAE means in this sample were relatively similar to those previously found in schizophrenia samples (Horan et al., 2015). Future work is needed to examine empathy differences in FEP samples compared to healthy participants.

Taken together, our data indicate that improved empathy measurement is needed for work in FEP samples. This is an important future direction, underscored by our results suggesting that affective empathy, in particular, is meaningfully related to negative symptoms, with direction of relationship dependent on subscale. Moreover, our results suggest that empathy predicts notable variance in clinical characteristics in FEP, even after accounting for other social cognitive constructs. Overall, with replication, these findings demonstrate that empathy may be clinically important in FEP samples.

Supplementary Material