What impact did the COVID-19 pandemic have on psychosis and the delivery of early intervention psychosis services?

Background: In Australia, the first nationwide COVID-19 lockdown occurred in March 2020 bringing challenges for early intervention psychosis (EIP) services. Limited evidence exists on the impacts of the pandemic on treatment outcomes among EIP clients. Methods: This prospective cohort study analysed routine data from 15 headspace Early Psychosis centres in Australia. Participants were 12 to 25 years, meeting criteria for First Episode Psychosis (FEP) or Ultra High Risk of psychosis (UHR) comparing those who commenced treatment ‘pre-COVID-19 ’ (between 16th August 2018 and 15th August 2019), and ‘during-COVID-19 ’ (between 1st March 2020 and 15th September 2020). Clinical symptoms at treatment commencement were assessed using the Brief Psychiatric Rating Scale (BPRS) and the Kessler Psychological Distress Scale (K10). with outcomes after 6 months compared between cohorts using linear mixed-effects regression, controlling for confounders. Results: Of 1246 young people analysed (653 FEP, 596 UHR), significant improvements were observed with treatment in both groups (5 to 13-point reduction in BPRS score per 6-months treatment). Treatment effectiveness reduced during-COVID-19 for psychosis symptoms, with the FEP BPRS treatment effect lower by 4.3 points (95%CI: 0.5, 8.1). UHR clients had lower BPRS negative symptoms during-COVID-19 ( p = 0.020). Service contacts increased during-COVID-19, with increased telehealth services ( p < 0.001). Conclusions: Early intervention remained effective for FEP and UHR, despite the pandemic and transition of EIP services to virtual service delivery. Reduced treatment efficacy in FEP psychosis symptoms may indicate potential limitations of telehealth. Further research to examine longer term clinical and functional outcomes due to the pandemic is required.


Introduction
The COVID-19 pandemic had a catastrophic impact on healthcare worldwide, the significance of which continues to be explored.Governments implemented strict public health measures including border closures, stay at home orders and social distancing in an unprecedented manner following the outbreak (Choukér and Stahn, 2020;Chung et al., 2021;Haug et al., 2020;Basseal et al., 2023;Onyeaka et al., 2021).COVID-19 also created significant stressors including disruption to routine, unemployment, financial stress, fear of infection, and Abbreviations: BPRS, Brief Psychiatric Rating Scale; DUP, Duration of untreated psychosis; EIP, Early intervention psychosis; EPPIC, Early Psychosis Prevention and Intervention Centre; FEP, First Episode Psychosis; hEP, headspace Early Psychosis; K10, Kessler psychological distress scale; MDS, Minimum Data Set; NEET, Not in Education, Employment, or Training; SD, Standard Deviation; UHR, Ultra High Risk.
Early Intervention in Psychosis (EIP) services provide treatment to young people with a First Episode of Psychosis (FEP), and those deemed at Ultra-High Risk (UHR) of developing psychosis, the prodromal phase of a psychotic disorder but not yet meeting threshold for FEP (Early Psychosis Guidelines Writing Group and, EPPIC National Support Program, 2016).The efficacy of EIP services has been established, improving clinical and functional outcomes in individuals (Brown et al., 2022;Correll et al., 2018;Tempelaar et al., 2021).When the COVID-19 pandemic began, EIP programmes had to rapidly adapt, with services required to use personal protection equipment (PPE) and transition to online service delivery, posing unique challenges in the assessment of psychosis worldwide (Tan et al., 2022;Nitzan et al., 2011).The pandemic also contributed to widespread anxiety, social withdrawal, and the rise of 'conspiracy' theories, sometimes present in psychosis (Ahmed et al., 2020;O'Donoghue et al., 2022).
For EIP services, the concerns regarding COVID-19 were three-fold.First, increased exposure to risk factors for psychosis including social isolation, stress, and substance use, could result in increased rates of psychosis, particularly in vulnerable groups (Jauhar et al., 2021).Second, COVID-19 infection could increase the risk of psychosis through several potential pathogenic biomechanisms (Watson et al., 2021).Third, social distancing measures would negatively impact the early detection and treatment of psychosis.Psychotic disorders are most likely to develop during adolescence and early adulthood (McGorry et al., 2011;Kessler et al., 2005).Early identification and intervention aims to reduce delays to first treatment (known as Duration of Untreated Psychosis, DUP) and it was hypothesised that pandemic measures would increase DUP, resulting in an increase in undetected cases of psychosis (O'Donoghue et al., 2021).Available research on FEP incidence rate and presentation patterns during-COVID-19 remains mixed (Chakraborty et al., 2020;Esposito et al., 2021;Segev et al., 2021;Casanovas et al., 2022;Seierstad et al., 2023;O'Donoghue et al., 2022) and there are limited published findings about the impact of the pandemic on treatment effects.
Outcome data collected in Australia before COVID-19 showed considerable improvement in symptomatic and functional outcomes in FEP and UHR clients who received EIP treatment through the headspace early psychosis (hEP) programme (Brown et al., 2022).Using this dataset, we can now explore the impact of the COVID-19 pandemic on the young people accessing these services.The aims of this current study are to 1) compare baseline characteristics and clinical presentation of young people engaging in services pre-COVID-19 to during-COVID-19, and 2) investigate the impact of the pandemic on treatment response and clinical outcomes.

Setting
Data came from the 15 treatment centres of the hEP programme with centres managed by 6 clusters across Australia: Southeast Melbourne, Western Sydney, North Perth, Adelaide, Southeast Queensland, and Darwin.The hEP programme delivers EIP services with an objective to identify young people with FEP or deemed UHR as early as possible and provide them with evidence-based treatment.The care provided is based on the EPPIC (Early Psychosis Prevention and Intervention) model (McGorry et al., 1996;Stavely et al., 2013).This includes specialist case management, medication, peer support, family support, and psychosocial interventions including functional recovery programmes (Brown et al., 2022).The fidelity of the hEP programme to the EPPIC model has been demonstrated (Williams et al., 2021) using a novel fidelity tool to carry out annual assessments since 2017.

Participants
Participants met criteria for FEP or UHR, aged between 12 and 25 years and provided informed consent for their data to be used for service evaluation.The dataset was divided into two cohorts, 'pre-COVID-19' and 'during-COVID-19', to evaluate the impact of the pandemic.Australia implemented the first nationwide lockdown on 23rd of March 2020 which included social distancing measures, stay at home orders and closures of school, universities, and workplaces (Lupton and Lewis, 2023).The 'pre-COVID-19' cohort was therefore defined as the period from the 16th of August 2018 to 15th of August 2019 inclusive, allowing all clients the possibility of 6 months treatment time before COVID-19 without overlap with during-COVID-19 cohort.The 'during-COVID-19' cohort was defined as the period commencing 1st March 2020 to 15th of September 2020 to allow for a potential 6 months of treatment time up to the end point of our dataset (March 2021), resulting in a smaller sample size than the 'pre-COVID' group.

Measures
Outcome measures were completed at entry to the hEP service (baseline) and every 90 days during an episode of care.Baseline demographic information encompassed gender, age, education, work status, living arrangements, language spoken at home and First Nations status.Additionally, data on individuals who were not currently involved in education, employment, or training (NEET) was collected.For those experiencing a FEP, DUP was calculated as the time between the onset of the initial frank psychotic symptoms and the start of treatment provided by hEP.Data relating to care during the first six months of treatment were collected.This included the number and hours of direct treatment services including medical, psychological therapy, alcohol or drug specific intervention, vocational, group or family work, and general support and the percentage of these services provided face to face, via telehealth (phone and video) or other (SMS, email, or webchat).

Clinical outcomes
The 24-item Brief Psychiatric Rating Scale (BPRS) is a standardized rating instrument used to assess the severity of psychiatric symptoms in patients with mental ill-health that has been widely used in research studies with good inter-rater reliability and validity (Overall and Gorham, 1962).The items are designed to measure both positive and negative symptoms of mental ill-health, including thought disturbance, anxiety, depression, hallucinations, and delusions.The BPRS can also be used to create subscales, and this study focused on two specific subscales: BPRS-Psychosis (item 9: Suspiciousness, 10: Hallucinations, 11: Unusual thought content and 15: Conceptual disorganisation) (Lachar et al., 2001); and BPRS-Negative (item 16: Blunted affect, 17: Emotional withdrawal and 18: Motor retardation; total score ranges from 3 to 21) (Shafer, 2005).The Kessler Psychological Distress Scale (K10) is a selfreport measure consisting of 10 items used to assess the level of psychological distress experienced in the past four weeks (Kessler et al., 2002).

Statistical methods
The UHR and FEP groups were analysed separately using the same statistical approach.Descriptive statistics were used to describe participant characteristics and clinical outcomes at baseline and to compare the 'pre-'and 'during-COVID-19' cohorts.Categorical outcomes were described using counts and percentages, and continuous variables using either a mean and standard deviation (SD) or medians and an interquartile range for clinical outcomes and service data.Cohorts were compared using chi-squared tests for categorical variables and two sample t-tests for all continuous variables except service outcomes, where Wilcoxon's rank sum test was used.
Linear mixed effects regression models were used to model the effects of treatment (per 6 months), cohort period (pre-or during-COVID-19), and an interaction between them, on each outcome.The interaction term is used to test whether the 6-month treatment effect differed precompared to during-COVID-19.All models were adjusted for potential confounders, age at intake, gender (female, male and non-binary), First Nation status, and an indicator of whether the hEP centre state was in Victoria or New South Wales, the states with the strictest lockdown periods reference.The adjusted models were used to estimate the 6month treatment effect for each cohort period.
Multiple imputation was used to account for missing data in all regression models, however both the imputed and complete case analysis are reported.For the multiple imputation, data were imputed in wide format using the missRanger package in R (Mayer, 2023) which implements a chained random forest algorithm (Wright and Ziegler, 2017) using predictive mean matching.For each outcome, 50 datasets were imputed, analysed and combined using Rubin's rules (Rubin, 1987).

Baseline characteristics
Table 1 presents the demographic characteristics of FEP clients accessing services pre-COVID-19 (n = 428) and during-COVID-19 (n = 225).Overall, there was minimal difference in the characteristics of young people, including DUP length at entry to services.During-COVID-19, there was a shift in current employment and working status compared to pre-COVID-19 (p = 0.05) with a higher percentage of individuals studying only (29.9 %) compared to pre-COVID-19 (20.9 %), and a lower percentage studying and working (5.9 %) compared to pre-COVID-19 (9.5 %).Most individuals in both groups were either NEET or studying only.During-COVID-19, a slightly higher percentage of presenting clients were born outside of Australia/New Zealand (p = 0.028) and reported living in stable situations compared to pre-COVID-19 (p = 0.013).

Clinical profiles
Table 2 shows the baseline clinical symptoms of FEP clients presenting to services pre-and during-COVID-19.There were no differences between BPRS total, BPRS subscales and K10 scores.There was a trend towards increased number and hours of direct services delivered between time periods which did not reach significance.During-COVID-19, the median percent of face-to-face services reduced by 33.6 % (p < 0.001), and the median percent of telehealth services delivered increased by 29.5 % (p < 0.001).
Fig. 1 (A) presents the mean outcome scores (with 95 % CIs) for each cohort over a 6-month period of treatment.The patterns were similar across the BPRS scores and subscales, i.e., the psychosis symptoms which are comparable at baseline appeared to improve at a slower rate during compared to rates pre-COVID-19.For K10, the scores appeared similar at baseline and 3 months, but slightly higher at 6 months during-COVID-19.S1 in Supplementary Material.The results of the complete case and imputed analyses are consistent, so we report the imputed results.For every 6 months of treatment pre-COVID-19, there was a 12.8-point reduction in BPRS score (95 % CI 10.6, 15.0).The treatment effect was reduced during-COVID-19 at 8.5 points (95%CI: 5.3, 11.8), 4.3 points less (95%CI: 0.5, 8.1)   S1.).Similar trends were evident for the BPRS subscales; however, the differences were not statistically significant.For K10 score, every 6 months of treatment led to a mean 5.3-point drop (95 % CI: 4.2, 6.3) in K10 score pre-COVID-19 and a slightly smaller mean drop of 4.3-points during-COVID-19 (95 % CI: 2.7, 5.9), which was not significantly different (interaction p = 0.289).

Baseline characteristics
Table 3 presents the demographic characteristics of UHR clients accessing services pre-COVID-19 (n = 351) and during-COVID-19 (n = 245).Overall, there was minimal difference in the characteristics of young people accessing services with only a difference in sexual orientation between the two groups, with a lower proportion identifying as heterosexual/straight (p = 0.033), referral sources pre-and during-COVID-19 (p < 0.001) and a lower proportion born in Australia/New Zealand (p = 0.013) during-COVID-19.The percentage of clients referred by healthcare or community providers increased from 67.4 % pre-COVID-19 to 80.8 % during-COVID-19, while the percentage of those referred by self, family, or friends decreased from 32.6 % pre-COVID-19 to 19.2 % during-COVID-19.

Clinical profiles
Table 4 presents the baseline clinical symptoms of UHR clients presenting to services before and during-COVID-19.There were no differences between BPRS total score, BPRS psychosis subscale score, or K10 score.The mean BPRS negative symptom subscale score at baseline was lower during-COVID-19 (4.9) compared to pre-COVID-19 (5.4) (p = 0.020).There was an increase in the median number of direct services (p < 0.001) and median number of hours of direct services (p < 0.001) during-COVID-19 while the median percent of face-to-face services reduced by 35.7 % (p < 0.001), and the median percent of telehealth services increased by 26.6 % (p < 0.001).
Fig. 2 (A) presents the mean outcome scores for each cohort over a 6month period of treatment.The patterns were similar across the BPRS scores and subscales with the BPRS and K10 scores recovering slower during-COVID-19.The results from the regression models are summaries in Fig. 2 (B) with details provided in Table S2 in Supplementary Material.For every 6 months of treatment received, there was a significant reduction in all clinical outcomes in the UHR group during both time periods.Pre-COVID-19, there was a 7.9 reduction in BPRS score with six months of treatment (95 % CI: 5.9, 9.9).During-COVID-19, the BPRS score reduced by 5.1 points (95 % CI: − 2.8,7.3)(p = 0.063 for the interaction term, see table S2).There was no evidence of a difference in treatment efficacy for the BPRS subscales or K10 score, pre-and during-COVID-19.

Discussion
This study utilised data from a unique naturalistic large clinical cohort to understand the impact of the COVID-19 pandemic on individuals experiencing FEP or at UHR of developing psychosis.Our results suggest that the characteristics and clinical severity at baseline were comparable before and during the COVID-19 pandemic.UHR clients had lower BPRS negative symptoms during-COVID-19.Significant improvements were observed with treatment pre-and during-COVID-19 in both groups.However, there was some evidence of delayed recovery during the pandemic for psychosis symptoms in FEP, even with more intensive care engagement mainly delivered through telehealth.
Clinical outcomes at baseline were comparable between cohorts in the FEP cohort, which is notable as it suggests there was not a significant increase in symptom severity on service entry during-COVID-19.Our findings indicate that UHR clients experienced fewer negative symptoms at baseline during-COVID-19 which is in contrast with existing literature that found UHR individuals experienced increased levels of anhedonia and avolition during-COVID-19 (Strauss et al., 2022).Our finding that there was no difference in positive symptoms at baseline in UHR individuals during-compared to pre-COVID-19, supported current available research (Berglund et al., 2023).We measured the impact of the pandemic on treatment efficacy, finding that treatment delivered was effective in reducing symptom severity in the FEP group during both time periods.However, we observed a reduced treatment effect on BPRS in the FEP cohort during-compared to pre-COVID-19.The reduction in K10 scores over treatment was comparable between both time periods.Overall, these findings suggest that while treatments delivered during-COVID-19 reduced FEP symptom severity, their efficacy may have been negatively impacted by the pandemic and Governments' corresponding policy responses.It also suggests that COVID-19 settings may have impacted psychotic symptoms more than psychological distress symptoms.In the UHR cohort, we found that treatment was effective in reducing symptom severity with a substantial reduction in BPRS and K10 scores during both pre-and during-COVID-19, with a slightly reduced treatment effect during-COVID-19; that did not reach significance.
We found no evidence that treatment services were reduced during-COVID-19.In fact, the median number and hours of direct services were higher during-COVID-19 for both groups.Research examining the impacts of telehealth use on youth mental health service delivery in Australia reported lower cancellation rates during-COVID-19, which supports this finding (Nicholas et al., 2021).Additionally, there was no change in fidelity measures during-COVID-19 (Williams G, Personal Correspondence, Nov 2023) compared to previous years (Williams et al., 2021).This suggests that the quality of care being provided was comparable between periods and did not contribute to the differences in  1 p-values were estimated using Wilcoxon's rank sum test.
treatment effect found.The reduced treatment efficacy during-COVID-19 could possibly be due to widespread disruptions and changes to the mode of service delivery away from face-to-face care.The reduced treatment efficacy could also indicate that pandemic related stressors heightened distress, impeding usual treatment gains.However, we note that baseline presentations were not worse on entry to services.Encouragingly, we found no difference in DUP between periods in our cohorts.We hypothesised that DUP may increase during-COVID-19 due to several pandemic related barriers, which would result in delayed recognition and subsequent treatment delivery (Brown et al., 2020).Our findings suggest this was not the case, which is positive given the benefit of early intervention on functioning and outcomes (Alvarez-Jimenez et al., 2018;Bird et al., 2010) and supports similar results (Casanovas et al., 2022;Nicholls-Mindlin et al., 2023).During-COVID-19, individuals were also spending more time at home, and symptoms may have been detected earlier by household members.Alternatively, individuals experiencing longer DUPs may have experienced greater difficulties accessing EIP care during the pandemic, therefore resulting in an inaccurate representation of overall DUP times in our sample.Studies from North America and Europe reported reduced emergency department (ED) psychiatric presentations during early stages of COVID-19, potentially due to pandemic related measures and fears of contracting the virus (Gonçalves-Pinho et al., 2021;Szmulewicz et al., 2021).This contrasts with findings from Melbourne, Australia, reporting a 6.8 % increase in schizophrenia and acute transient psychosis ED presentations in the first six months of lockdown (Jagadheesan et al., 2021).These findings are notable given that Melbourne experienced extended successive lockdown periods from March 2020, including a 112-day lockdown 8th July -27th Oct 2020 (Griffiths et al., 2022).
The baseline characteristics of young people at service presentation were comparable between time periods.However, we observed a shift in education and employment patterns at baseline for some young people experiencing FEP, with a reduced proportion both working and studying.This is notable, as unemployment and educational disparity in people with psychosis is associated with several psychological and social challenges and can affect functional outcomes (Baksheev et al., 2012;Rinaldi et al., 2010).Research has found that individuals with FEP experienced more pronounced job loss and interruption to education during-COVID-19 compared to the general population (Szmulewicz et al., 2022).As functional recovery is a key goal of EIP services, it is important to consider how the pandemic impacted employment and education rates, and how services can optimise the delivery of functional-recovery orientated treatments during a global crisis like COVID-19.
A slightly higher proportion of individuals accessing services (FEP and UHR) were born outside of Australia/New Zealand during-COVID-19, which could indicate that migrant populations were more vulnerable to pandemic related stressors, as supported by existing studies (Crawley, 2021;Rodríguez-García-De-Cortázar et al., 2021;Kiteki et al., 2022).Positively, more FEP young people reported living in stable housing situations during-COVID-19, which is important for functional recovery (Albert et al., 2011;Roy et al., 2013).The Australian Government's response to homelessness during-COVID-19, which included an unprecedented amount of funding allocated to accommodate the homeless, may have contributed to this finding (Parsell et al., 2020).Additionally, those without stable housing may have encountered increased challenges in accessing care during the pandemic, and therefore may be underrepresented in the during-COVID-19 cohort.Contrary to expectations, a greater number of UHR participants were referred by healthcare or community providers during-COVID-19 compared to self, family, or friends.This is despite changes in healthcare delivery and reduced access to services during-COVID-19.This may be due to lower threshold to refer to EIP services by providers when unable to assess in person and concerns that the COVID-19 may increase psychosis rates.    1 p-values were estimated using Wilcoxon's rank sum test.

Implications
This study adds to the limited data available on the impact of COVID-19 on EIP services and users (FEP and UHR).Current literature suggests future pandemics are likely (Iserson, 2020;Sachs et al., 2022), which underscores the importance of further research to develop mental health policies and treatment strategies for global pandemics.These findings add to our understanding of how pandemics may adversely impact the effectiveness of treatment delivered in this population.Potential strategies may include additional resources and increased social supports.
Our findings demonstrate that the shift to online service delivery due to the pandemic increased engagement with young people, supporting existing research (Nicholas et al., 2021).However, the increased use of telehealth may have negatively impacted treatment efficacy and clinical outcomes when compared to pre-COVID-19.Further research to examine the efficacy of telehealth to deliver therapeutic interventions in this cohort is required, with new research indicating that face-to-face interactions are superior to those online (Zhao et al., 2023).By examining and attempting to identify the subset of clients for whom telehealth is less beneficial, EIP services could then individually tailor therapeutic interventions, ensuring the most appropriate and effective treatment is delivered.Given the significant impact that early intervention can have on long-term social, vocational, and functional outcomes, it is crucial we study the effects of COVID-19 so we can rapidly optimise EIP service delivery in future global crises (Szmulewicz et al., 2022).Strengths of the study include relatively long lockdown periods studied and large geographical area included.We note that the hEP sites are based in metropolitan areas in Australia, reducing generalisability of results in rural populations.The demographics of both cohorts were similar, strengthening the validity when comparing findings.The hEP programmes are well established, with standardized models of care.Fidelity assessments continued throughout the pandemic, with no significant change from previous years pre-pandemic.No major service changes occurred between time points, except for the transition to virtual service delivery, which was noted in the findings.

Limitations
We also acknowledge some limitations of the study.The data were collected and entered by clinicians, which increases the potential for error (Goldberg et al., 2008).The proportion of missing outcome data was mostly similar or improved between the two time points (See table S3).Multiple imputation was used to account for missing data and minimise bias, but we recognise this may still result in bias.The imputed results, however, were consistent with the unimputed (complete case) analysis suggesting this was not significant.While potential confounding factors were controlled for, it is possible that there may be other changes in service configuration (e.g., changes in staffing profile) during the study period which resulted in differences in the two temporal cohorts, unrelated to the pandemic.The during-COVID-19 period may not have been universal due to the differences in lockdown periods and measures between states and territories in Australia (Meyer et al., 2023).Furthermore, potential biases relating to accessing EIP services during the pandemic limit study conclusions.

Conclusion
COVID-19 brought many challenges for EIP services.Our study found a reduction in overall symptom severity and psychological distress with six months treatments both pre-and during-COVID-19, with slightly reduced treatment efficacy observed in our FEP group during the pandemic.In our UHR cohort, an improvement in clinical outcomes were observed with treatment and there was no significant impact on treatment effect when compared to the same period pre-COVID-19.Slightly surprisingly, DUP did not increase.These findings strengthen current evidence on the efficacy of early intervention treatment for FEP and UHR and highlight the successful transition of EIP services to virtual mental health care delivery, with treatments effects still evident despite the pandemic.While the shift to virtual service delivery increased client engagement, it may have been less effective than in-person service delivery.These results highlight the need for further research to examine the use of telepsychiatry in this cohort and the need for longitudinal studies to inform healthcare planning for future pandemics.

Fig
Fig. 1 (B)  summarises the results of the regression models, with detailed regression coefficients provided in TableS1in Supplementary Material.The results of the complete case and imputed analyses are consistent, so we report the imputed results.For every 6 months of treatment pre-COVID-19, there was a 12.8-point reduction in BPRS score (95 % CI 10.6, 15.0).The treatment effect was reduced during-COVID-19 at 8.5 points (95%CI: 5.3, 11.8), 4.3 points less (95%CI: 0.5, 8.1)

Fig. 1 .
Fig. 1.Treatment trajectories in FEP group (A) Trend of outcome mean scores (and 95 % confidence intervals) for pre-COVID and during COVID cohorts (B) Comparison of mean reduction in outcome scores with 6 months of treatment, pre-COVID and during COVID in the FEP cohorts estimated from imputed and unimputed linear mixed effects regression models.

Fig. 2 .
Fig. 2. Treatment trajectories in UHR group (A) Trend of outcome mean scores (and 95 % confidence intervals) for pre-COVID and during COVID cohorts (B) Comparison of mean reduction in outcome scores with 6 months of treatment, pre-COVID and during COVID in the UHR cohorts estimated from imputed and unimputed linear mixed effects regression models.

Table 1
Demographic characteristics for FEP clients by cohort at program entry.

Table 2
Baseline clinical characteristics and service contacts (within first 6 months) among FEP clients.

Table 3
Demographic characteristics for UHR clients by cohort at program entry.

Table 4
Baseline clinical characteristics and service contacts (within first 6 months) among UHR clients.