Article Text

Mental health of caregivers of children with medical complexity: group-based trajectory modelling
  1. Apsara Ali Nathwani1,
  2. Nora Fayed2,
  3. Sonia M Grandi1,3,
  4. Julia Orkin4,
  5. Eyal Cohen1,5,6
  1. 1Child Health Evaluative Sciences, The Hospital for Sick Children, Toronto, Ontario, Canada
  2. 2School of Rehabilitation Therapy, Queen's University, Kingston, Ontario, Canada
  3. 3Division of Epidemiology, Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada
  4. 4Paediatrics, The Hospital for Sick Children, Toronto, Ontario, Canada
  5. 5Institute of Health Policy, Management and Evaluation, University of Toronto, Toronto, Ontario, Canada
  6. 6Edwin S H Leong Centre for Healthy Children, University of Toronto, Toronto, Ontario, Canada
  1. Correspondence to Dr Eyal Cohen, Child Health Evaluative Sciences, The Hospital for Sick Children, Toronto, Ontario M5G 1X8, Canada; eyal.cohen{at}


Objective To describe the mental health trajectories of caregivers of children with medical complexity (CMC) and explore child characteristics associated with below-average caregiver mental health.

Design A secondary analysis of prospectively collected data from 123 caregivers of children aged <16 years with medical complexity from a multicentre randomised trial conducted from December 2016 to June 2021.

Main outcome measure The Patient-Reported Outcomes Measurement Information System Global Mental Health Scale was used to measure caregivers' self-reported mental health well-being. Group-based trajectory analysis was used to identify clusters of caregivers with similar changes in mental health across 24 months. Logistic regression was used to identify child-related predictors of mental health among caregivers.

Results A final model with three distinct groups was selected, corresponding to caregivers with average (n=39), moderately below-average (n=65) and severely below-average (n=19) mental health scores, all with stable trajectories and high posterior probabilities (>90%). Moderately and severely below-average caregiver mental health groups, merged into one group, were associated with a greater number of child medical technology devices (adjusted OR (aOR) 1.44, 95% CI 1.01 to 2.04), gross motor difficulties (aOR 3.51, 95% CI 1.02 to 12.05) and worse child emotional (aOR 0.93, 95% CI 0.87 to 0.99) and psychological well-being (aOR 0.93, 95% CI 0.88 to 0.99).

Conclusion Most caregivers of CMC reported persistently below-average mental health. The intensity of caregiving, as indicated by medical technology and child functional needs, is a potential risk factor for below-average caregiver mental health. Future design and evaluation of interventions focused on support for caregivers of CMC are warranted.

  • Mental health
  • Child Health

Data availability statement

Data may be obtained from a third party and are not publicly available. The study has used CCKO clinical trial data sets. These data sets cannot be made publicly available due to data-sharing agreements.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Previous studies suggest that caregivers of children with medical complexity (CMC) are at higher risk of experiencing mental health challenges due to the significant demands associated with caregiving.


  • Using longitudinal data, three caregiver mental health trajectories (severely below-average, moderately below-average and average) were observed over 24 months.

  • Child medical technology dependency, gross motor difficulties, and low emotional and psychological well-being predicted below-average caregiver mental health.


  • Most caregivers of CMC face below-average mental health, likely due to the intensity of caregiving.

  • Studies of interventions to improve support are warranted.


Children with medical complexity (CMC) represent a small but highly vulnerable group of children characterised by chronic conditions, high needs, functional limitations and high healthcare utilisation.1 Despite comprising <1% of the paediatric population, CMC account for over a third of all child healthcare expenditures.2 Survival rates of CMC have substantially improved over recent decades.3 This progress, however, has placed extensive demands on parental caregivers,4 including the need to provide intensive healthcare technologies in the home,5 in addition to meeting their child’s daily physical, social and emotional needs. Other potential sources of caregiver stress may include frequent changes in the child’s health status, medical fragility, prognostic uncertainty and the requirement for multiple healthcare visits in multiple places over time.6

Dealing with such unique challenges can impact caregivers’ health.7 Previous studies have reported that CMC caregivers are at increased risk of mental health difficulties, such as depression, anxiety and sleep difficulties,7 8 as well as chronic stress-associated physical conditions such as cardiometabolic disease.2 9

Studies of CMC caregiver mental health have been primarily limited to cross-sectional designs, providing limited insight into long-term trajectories. Given the interrelation of parental mental health, family functioning and the quality of life of children,10 it is important to understand the mental health trajectory of caregivers of CMC and to identify subgroups who may be at higher risk of poor mental health. Our study aimed to: (1) Identify whether there are distinct mental health trajectories among caregivers of CMC, and, (2) Explore which child clinical characteristics are associated with poor caregiver mental health.


This is a secondary analysis of data prospectively collected in a multicentre randomised trial of care coordination for CMC conducted between December 2016 to June 2021.11 Included were caregivers of children aged <16 years, who met the standard operational definition for CMC developed for recruitment into a care coordination programme in Ontario, Canada (Complex Care Kids Ontario, CCKO). These criteria included fragility, chronicity, complexity, and technology dependency and/or users of high-intensity care.1 Details on the CCKO intervention and trial have been described elsewhere.1 11 12 Of note, all the patients in the CCKO Trial provided informed consent and were randomly assigned to a care coordination intervention either at enrolment or after a 12-month waitlist period, which was included in the present study as a predictor variable.

Primary caregivers of CMC completed a variety of surveys about themselves and their children over 2 years of data collection beginning at randomisation within the CCKO Trial, including information on the sociodemographic and clinical child characteristics.

Study outcome

The study outcome was the mental health status of primary caregivers of CMC, measured using the PROMIS (Patient Reported Outcomes Measurement Information System)—Global Health 10 Scale at four time points—at the time of recruitment and at 6 months, 12 months and 24 months after the first assessment. This scale rates global mental health by combining self-reported quality of life, mood, emotional well-being and satisfaction with social activities/relationships on a 5-point bipolar Likert Scale. This rating was converted into t-scores, standardised to the general population with a mean score of 50 points and an SD of 10 points,13 14 whereby lower scores are indicative of poorer mental health. This approach has been validated and has norm reference data for comparison. It has a high internal reliability of 0.86 and good to exceptional psychometric characteristics among caregivers.13 15


Predictor variables were selected based on the conceptual framework shown in figure 1, adapted from Pearlin et al’s ‘Stress Process Model’.16

Figure 1

Study conceptual framework; dotted links and variables highlighted in red are not measured in the present study.*Variables in red: not measured in the dataset. Dotted link: associations not measured in the current study.

Sociodemographic factors

Caregivers also completed a brief questionnaire at the baseline visit to provide information on sociodemographic indicators including the child’s age, gender and race/ethnicity, the caregiver’s age, gender, marital status, educational status (defined as the highest level of self-reported completed education), as well as household density (defined as the number of individuals who reside in the same household as the participant).

Child characteristics

Clinical characteristics, including number of diagnoses, reliance on technology, communication skills and gross motor functioning, were extracted from medical records.

Child indicators were measured at baseline using the following caregiver-proxy measures: the (1) Psychological Well-being (6-items) and (2) General Mood/Emotions (7-items) subscales of the KIDSCREEN-52 tool administered to caregivers to measure the child’s psychological and emotional well-being. KIDSCREEN-52 is a validated structured cross-cultural tool and represents the most suitable content overlap with CMC.11 Scores of each item of the subscale were calculated as standardised t-scores, with higher t-scores representing better psychological well-being and general mood/emotions of the child.17 (3) Physical pain was measured using a 10 cm linear Visual Analogue Scale (VAS).18 Due to the consistency of interpretation among parent-child dyads, test-retest reliability and measurement accuracy, linear VAS is regarded as superior to other pain reports for children and is most suited for measuring functional capacity in medically complex populations.19

Statistical analysis

Characteristics of the study population at baseline were presented as means and SD for normally distributed continuous data, or medians and IQRs for skewed data. Categorical data were presented as proportions and frequencies.

We used group-based trajectory analysis to identify clusters of caregivers experiencing similar mental health trajectories. This approach uses a semiparametrical model to account for longitudinal data assuming that the population consists of distinct subgroups, thus enabling the classification of relatively homogeneous groupings within the population (ie, those having a similar trajectory).20 The model uses the maximum likelihood method to obtain unbiased parameter estimates under the assumption that data are missing at random.21

A censored normal distribution model was estimated. To maintain the longitudinal nature of the analysis, caregivers with <2 time points of data were omitted from the model. Model selection was based on the theoretical caregiving model, iterative estimations of trajectories, shape/order of each trajectory and clinically meaningful z-scores of severity based on PROMIS population norms literature.22

The shape of trajectories, that is, the pattern of change over time, was established by fitting several models with the intercept (0), slope (1), quadratic (2) and cubic (3) functions, and removing non-significant cubic and quadratic polynomials.21 The number of distinct clusters was determined based on the following criteria: (1) The Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and log likelihood test indices, with lower values indicating better fit; (2) Entropy, the average posterior probability of trajectory assignment, with >0.70 indicating successful categorisation; (3) Log Bayes factor to assess model complexity, with a score >30 indicating strong evidence; (4) Trajectory size, whereby >5% denotes a more exact assignment; and, (E) The CI width, with narrower CIs indicating more precise and accurate group estimation.23 The mean differences of caregiver’s mental health scores between time points in each trajectory were explored by standardised differences with 95% CIs—the difference in mean caregiver mental health scores at each time point (eg, 6 months − baseline, 12 months − 6 months scores, etc) divided by the pooled SD.24

A logistic regression model was used to identify predictors of membership in the below-average mental health trajectory group in comparison to the average trajectory14 group (reference). The model was adjusted for group assignment in the original trial, as well as baseline data on household density, caregiver’s sex, education and marital status, and presented as adjusted ORs (aORs) with 95% CIs. Analyses were conducted using Stata V.17.


Of 139 CMC caregivers in the clinical trial, 123 (88%) completed at least two out of four surveys with follow-up rates at those time points of 87%, 84% and 78% at 6 months, 12 months and 24 months, respectively (online supplemental eFigure 1). Baseline family, child, and caregiver characteristics of the 123 study participants included in analyses are summarised in table 1. Most caregivers were women (86%), married (83%), and had a postsecondary or higher level of education (85%). The mean (SD) number of diagnoses per child was 6 (3.1), with neurological disorders (48%) and congenital anomalies (30%) being the most prevalent conditions. Baseline characteristics of caregivers included and excluded from the analysis were similar (online supplemental eTable 1).

Table 1

Baseline characteristics of caregivers of children with medical complexity included in the study

Mental health trajectories

The three-class model was found to have the most suitable fit (online supplemental eTable S2) based on low BIC, AIC and log likelihood estimates, high entropy of 0.81, log Bayes Factor estimate of 41, subclass size >5%, and narrow non-overlapping CIs.22 The three trajectories of PROMIS Global Mental Health Scores identified were as follows: (1) Severely below average (n=19 (15%) having mean (SD) scores of 29.5 (6.1) at baseline, 29.8 (6.0) at 6 months, 29.4 (5.0) at 12 months and 30.4 (6.8) at 24 months), (2) Moderately below average (n=65 (53%) with mean (SD) scores of 40.7 (6.0), 41.5 (5.0), 40.5 (4.5) and 41.8 (6.1)), and (3) Average (n=39 (32%) with mean (SD) scores of 53.8 (7.1), 50.6 (7.0), 53.7 (6.4) and 51.6 (6.4)) (table 2). The trajectories had high posterior probabilities ranging from 90% to 95%, indicating high accuracy and reliability of group classification (figure 2). Caregivers in the severely and moderately below-average groups followed flat trajectories with no differences in scores during the study period, while those in the average mental health group showed modest changes; a decrease in scores from baseline to 6 months (standardised difference −0.45, 95% CI −0.90 to 0.00)), followed by an increase from 6 months to 12 months (standardised difference 0.46, 95% CI 0.01 to 0.92) and no change from 12 months to 24 months (standardised difference −0.34, 95% CI −0.80 to 0.12).

Table 2

Caregiver’s Global Mental Health Scores of 123 caregivers included in the study at baseline, 6 months, 12 months and 24 months. All data are presented as means (SDs)

Figure 2

Global mental health trajectories of 123 caregivers of children with medical complexity over 24 months.

Predictors of below-average mental health trajectory

We merged the severely and moderately below-average mental health groups into a single group ‘below-average mental health’ (n=84 (68%)), due to the limited data available for logistic regression analysis.

Caregivers of children with higher technology dependency as well as those caregivers of children with gross motor difficulties had greater odds of having below-average mental health compared with those with lower technology dependency and those without gross motor difficulties (aOR 1.42, 95% CI 1.02 to 1.99 and aOR 3.50, 95% CI 1.01 to 11.87, respectively) (figure 3). Caregivers who reported high levels of emotional and psychological well-being for their children were less likely to have below-average mental health (aOR 0.93, 95% CI 0.87 to 0.99 and aOR 0.93, 95% CI 0.88 to 0.99, respectively) than those who reported low emotional and psychological well-being for their children. No associations were observed for other covariates.

Figure 3

Logistic regression of risk factors of global mental health trajectories among 123 caregivers of children with medical complexity over 24 months. ORs are adjusted for group assignment in the original trial, household density, caregiver’s sex, educational status and marital status.


This study described longitudinal trajectories of mental health of caregivers of CMC, which clustered into three distinct groups. Less than a third of caregivers had average mental health. The pattern of below-average mental health among the majority of the caregivers did not change substantially over 24 months and was associated with higher needs and poorer well-being among the children themselves. The findings demonstrate a concerning pattern of the pervasiveness of suboptimal mental health among CMC caregivers, and the recognition of reliable predictors for those at highest risk.

Previous cross-sectional studies have also reported a high prevalence of poor mental health among caregivers of children with chronic disabilities, ranging from 18% to 60%.25 26 Our findings reinforce previous reports of high caregiving demands associated with complex medical conditions27 with a cost to caregivers’ well-being, even while receive promising complex care coordination services.11 12 Compromised mental well-being of caregivers may hinder their ability to provide sufficient care for their children,28 leading to adverse health consequences not only for the caregivers themselves but also for their CMC and/or other children. The stability of the trajectories described here despite participation in a complex care coordination intervention trial further underscores the chronic nature of caregivers’ struggles and suggests that, although the health status of their children may fluctuate (eg, with intercurrent illnesses), caregiver mental health concerns are persistent.

Our finding that gross motor difficulties is associated with below-average mental health among caregivers, even within a group of children who all had substantial complexity, supports previous reports.29 Caregivers of children with gross motor difficulties often face additional challenges, such as helping with daily activities, transportation and supervision.30 They often face numerous logistical and financial burdens, which could put strain on their mental health.31 32 Likewise, the high OR of below-average mental health among caregivers of children with greater technology dependency can be attributed to the emotional and physical demands, sleep disruptions and time commitments, leaving caregivers with little opportunity for self-care or family care.33 Furthermore, children who rely heavily on medical technology are at a heightened risk of complications, including infections or device malfunction, necessitating vigilant and proactive monitoring from caregivers with prompt actions,34 factors that may be stressful and anxiety-provoking. Caregivers of children with complex needs also encounter societal attitudes and perceptions that portray their child as less capable or different from other children. This stigmatisation may undermine caregiver confidence, and can generate a sense of isolation and exclusion from the broader community, leading to heightened distress.35 Our observation of an association between caregiver reports of good emotional and psychological well-being of the child and caregiver mental health may be due to the impact of their own mental health on the perceptions of their child’s. The knowledge that their child is content and emotionally thriving can generate positive emotions and enhance caregiver satisfaction, resulting in improved mental health outcomes. Whatever the explanation, the observed association reinforces the interdependency of caregiver and child well-being,34 36 consistent with previous literature reporting poor parental mental health among caregivers of children with adverse psychological and emotional problems.37 38

Future studies are important to explore protective factors that promote caregiver resilience and positive mental health outcomes to appropriately tailor interventions and support programmes. Additional considerations for caregiver support include those focused on prevention, screening and/or treatment of mental illnesses. Prevention can focus on reducing caregiver stress, and strengthening social support networks. Screening could involve questions about parental mental health during follow-up visits of CMC.36 Treatment strategies could include improving access to dedicated mental health clinicians within structured complex care teams.39

Strengths and limitations

The study used data from a randomised clinical trial, including high-quality, longitudinal data capturing various types of information like demographic, clinical and patient-reported outcomes with relatively minimal loss to follow-up. While the amount of missing data was relatively small (<15%), it may have nevertheless introduced some potential selection bias. Participants were mostly mothers from high socioeconomic groups, and findings may differ in more diverse populations. Participant responses may have been influenced by their experience with the care coordination intervention being evaluated, although the participants receiving care coordination had similar mental health outcomes as the control group. Our assessment of CMC well-being was assessed entirely by the caregiver proxy as the participants were unable to self-report due to age and/or intellectual disability. Findings may differ with CMC self-report or report by a different proxy reporter. This study is mainly descriptive and it is essential to recognise that caregiver’s mental health challenges may not be solely linked to the complexity of medical issues. Future studies should also explore other demanding aspects of parenting that could contribute to the caregiver’s well-being.


The majority of caregivers of CMC report persistently below-average mental health. The intensity of caregiving for the child is a risk factor for below-average caregiver mental health, suggesting that future design and evaluation of interventions focused on support for caregivers of CMC are warranted.

Data availability statement

Data may be obtained from a third party and are not publicly available. The study has used CCKO clinical trial data sets. These data sets cannot be made publicly available due to data-sharing agreements.

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and was approved by The Hospital for Sick Children’s Research Ethics Board (REB 1000053509). Participants gave informed consent to participate in the study before taking part.



  • Contributors EC, AAN, NF and SMG were involved in the study conception and design. AAN performed statistical analysis, interpreted the results and drafted the initial version of the manuscript. EC, NF, SMG and JO provided expert clinical guidance and made critical revisions to the manuscript. All authors read and approved the final version of the manuscript. EC is responsible for the overall content of the work as guarantor.

  • Funding This study was supported by the Canadian Institute of Health Research (Funding reference numbers: FDN-143315 and PJT-180612) and Ontario Strategy for Patient-Oriented Research Support Unit IMPACT (Innovative, Measurable, Patient-oriented, Appropriate, Collaborative and Transformative) Award.

  • Disclaimer The authors are solely responsible for the analyses, conclusions, opinions and statements expressed in this document. These views do not represent those of the funding or data sources, and no endorsement should be inferred. The funders did not participate in the study's design, conduct, data collection, management, analysis, or interpretation, manuscript preparation, review, or approval, nor did they make the decision to submit the document for publication.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; internally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.