The Association between Therapeutic Alliance and Parental Health Outcomes following a Child's Death in the Pediatric Intensive Care Unit

Abstract

Therapeutic alliance reflects the strength and quality of the physician–patient/family relationship. We investigated the association between therapeutic alliance and bereaved parents' mental health and perceived overall health following their child's death in a pediatric intensive care unit (PICU). Bereaved parents were surveyed 6 months after their child's death in a PICU affiliated with the Collaborative Pediatric Critical Care Research Network. Parents were evaluated for complicated grief, depression, and post-traumatic stress using the Inventory of Complicated Grief (ICG), the Patient Health Questionnaire (PHQ-8), and the Short Post-Traumatic Stress Disorder Rating Interview (SPRINT), respectively. Overall health was evaluated using a single item. Therapeutic alliance between parents and their deceased child's PICU physicians was assessed using the Human Connection scale (HCS). Two hundred and thirty-five parents of 158 deceased children completed surveys. Mean ICG score was 34.4 ± 14.9 with 142 (60.4%) parents screening positive for complicated grief. Mean PHQ-8 score was 9.1 ± 6.2 with 102 (43.4%) screening positive for at least moderate depression. Mean SPRINT score was 14.6 ± 8.2 with 122 (51.9%) screening positive for post-traumatic stress disorder. Overall health was perceived as fair for 47 (20.0%) parents and poor for 10 (4.3%). Using multivariable modeling, higher HCS score (greater therapeutic alliance) was significantly associated with lower (better) ICG score (−0.23, 95% CI −0.42, −0.04, p  = 0.018). HCS score was not significantly associated with PHQ-8, SPRINT, or overall health scores. We conclude that bereaved parents experience a high level of adverse mental health symptoms including complicated grief, depression, and post-traumatic stress symptoms. Greater therapeutic alliance with PICU physicians may lessen symptoms of complicated grief during bereavement.

Introduction

Bereaved parents often suffer greatly after their child's death. ¹ Although grief is highly individualized, normal grief typically includes a painful sense of longing, sadness, and isolation which gradually diminishes in intensity over time. ² Complicated grief, also referred to as prolonged grief, has been described as a mental health disorder by many bereavement experts. ^{3 4 5} Complicated grief is a maladaptive form of grief with symptoms that include intense yearning, preoccupation with the deceased, a sense of loss of meaning or purpose without the deceased, and inability to accept the reality of the death. ^{3 4 5} Complicated grief symptoms are similar to normal grief but differ in their intensity, duration, and interference with daily living. Prior research conducted among parents whose children died in a pediatric intensive care unit (PICU) found bereaved parents are at risk for complicated grief symptoms. ^{6 7} Insecure attachment style is another risk factor for complicated grief. ⁸ Attachment style refers to an individual's expectations and interactions in close relationships. Securely attached individuals are comfortable being close to others whereas insecurely attached individuals are overly anxious about the availability of others or prefer to avoid relying on others. ⁸

Many parents experience symptoms of major depression or post-traumatic stress disorder (PTSD) after their child's death. ^{9 10 11 12} Although some overlap in symptoms exists, depression and PTSD differ from complicated grief in several ways. Predominant symptoms of depression include a reduced ability to feel pleasure and persistent feelings of worthlessness whereas predominant symptoms of complicated grief are yearning and sadness over the loss while self-esteem is often maintained. ^{3 4 5} PTSD is characterized by anxiety and fear of recurrent danger, rather than yearning and sadness over the lost child. Bereaved parents are also at risk for significant physical morbidity with higher frequency of acute illnesses, habitual smoking or alcohol consumption, insomnia, use of health services, and medication changes. ^{13 14}

Therapeutic alliance is a concept first described by psychotherapists to reflect the strength and quality of the client–therapist relationship. ¹⁵ Core components include mutual understanding, caring, and trust. More recently, the concept of therapeutic alliance has been extended to other situations, including end-of-life care, where it denotes the collaborative bond between terminally ill patients/family caregivers and their physicians. ^{16 17 18 19} Among adults with advanced cancer, strong patient–physician alliance has been associated with greater emotional acceptance of terminal illness, decreased ICU care at the end-of-life, ¹⁶ and better psychological adjustment of family caregivers during bereavement. ¹⁸ Research conducted among parents bereaved in PICUs found that when parents perceived their needs as met by PICU staff near the time of their child's death, complicated grief symptoms were less. ²⁰ Our objective was to explore the association between therapeutic alliance and bereaved parents' mental health and perceived overall health following their child's death in a PICU. We hypothesized that stronger parent–physician alliance would be associated with better parental health outcomes including reduced symptoms of complicated grief, depression and post-traumatic stress, and better overall health during bereavement.

Methods

Design and Setting

A multisite observational study was conducted across eight urban children's hospitals affiliated with the Eunice Kennedy Shriver National Institute of Child Health and Human Development Collaborative Pediatric Critical Care Research Network (CPCCRN) ²¹ between September 2016 and March 2019. The study was approved by the University of Utah Institutional Review Board (IRB) serving as the central IRB for the CPCCRN.

Participants

Parents (i.e., biological and/or legal guardians) were eligible if their child died in a PICU, they were English or Spanish speaking, and were at least 18 years old. One or more parents of each deceased child were eligible. Medical records of the children were reviewed to obtain parents' contact information and primary language for the purpose of recruitment.

Data Collection

Eligible parents were mailed surveys in English or Spanish 6 months after their child's death. If completed surveys were not returned within 1 month, telephone contact was attempted to offer the parents the option of completing the survey by telephone. If the household was successfully contacted by telephone but the survey was not completed, the parent was categorized as a refusal to participate. If the household could not be contacted by telephone after three or more attempts, the parent was categorized as unable to contact. For parents who completed surveys, their deceased child's medical records were further reviewed to collect information about the child's characteristics and clinical course. All parent information was collected by self-report.

All parents received a study information sheet along with the mailed surveys that offered the assistance of the local PICU social worker for bereavement support or mental health referrals. For parents contacted by telephone, the assistance of the social worker was also offered verbally. Social worker assistance was available regardless of the parent's decision to participate in the study.

Outcomes

Outcomes included complicated grief, depression, and posttraumatic stress symptoms, and perceived overall health among parents. Complicated grief symptoms were assessed using the Inventory of Complicated Grief (ICG). ²² The ICG is a 19-item measure that assesses the frequency of the cognitive, emotional, and behavioral symptoms described in each item. Total scores range from 0 to 76; scores of 30 or greater have been suggested as a positive screen for complicated grief. ²³ Cronbach's α has been reported as 0.94 and test–retest reliability as 0.8. ²² Validity has been demonstrated by association with other grief scales and ability to differentiate functional impairments. ²²

Depression symptoms were assessed using the Patient Health Questionnaire-8 (PHQ-8). ²⁴ The PHQ-8 is an 8-item measure that assesses the frequency of depressive symptoms. Total scores range from 0 to 24; scores of 5, 10, 15, and 20 have been suggested as positive screens for mild, moderate, moderately severe, and severe depressive symptoms. ²⁴ Cronbach's α has been reported to range from 0.82 to 0.88. ^{25 26} The PHQ-8 has been shown to have convergent validity with the PHQ-9 and Hamilton Depression Rating Scale. ²⁵

Post-traumatic stress symptoms were assessed using the Short Post-Traumatic Stress Disorder Rating Interview–Self Report (SPRINT). ²⁷ The SPRINT is an 8-item measure that assesses the frequency of post-traumatic stress symptoms. Total scores range from 0 to 32; scores of 14 or greater have been suggested as a positive screen for PTSD. ²⁷ Cronbach's α has been reported to range from 0.77 to 0.87. ²⁷ Validity has been demonstrated by correlation with other traumatic stress scales. ²⁷

Perceived overall health was assessed by one item ²⁸ : “In general, would you say your overall health is (1) excellent, (2) very good, (3) good, (4) fair, (5) poor.” Test–retest reliability has been reported as 0.69 and validity demonstrated by association with a standard health measure—the General Health Survey SF-12v. ²⁸

Independent Variables

Independent variables included parent and child characteristics. Parent characteristics included sociodemographics, attachment style, and the extent of therapeutic alliance with PICU physicians. Sociodemographics included gender, age, race, ethnicity, marital status, education, relationship to the deceased child, and number of surviving children. Relationship to the deceased child was categorized as biological parent or other legal guardian.

Attachment style was assessed using the Revised Adult Attachment Scale (RAAS)—Close Relationship Version. ^{29 30} The RAAS is an 18-item measure that evaluates one's general orientation in close relationships. Subscale scores reflect two fundamental dimensions of attachment, anxiety and avoidance, each of which is assessed along a continuum. Adults with high attachment-related anxiety worry about the availability of others whereas those with low anxiety are more secure in their perceived availability of others. Adults with high attachment-related avoidance prefer not to depend on others whereas those with low avoidance are more secure depending on others and having others depend on them. Scores for attachment-related anxiety and avoidance are a mean of the relevant items and range from 1 to 5. Higher scores reflect more insecure relationship styles whereas lower scores reflect more secure relationship styles. Cronbach's α has been reported to range from 0.76 to 0.85. ^{29 30 31} Validity has been demonstrated by association of subscale scores with related measures. ³⁰

Therapeutic alliance was assessed using the Human Connection Scale (HCS). ¹⁶ The HCS is a 16-item measure originally designed to evaluate the strength and quality of the relationship between adult cancer patients and oncologists. Cronbach's α has been reported as 0.9 and validity demonstrated by positive association with emotional acceptance of death ¹⁶ For this study, the HCS was adapted to evaluate the relationship between parents and their child's PICU physicians. ³² Each adapted item specifically referred to PICU physicians, for example, “How much did you trust your child's doctor(s) in the ICU?” The complete adapted scale is shown in Supplementary Tables 1 – 4 (available in the online version only). Scores range from 16 to 64; higher scores indicate greater therapeutic alliance.

Child characteristics included gender, age at time of death, cause of death, trajectory of death, mode of death, and PICU and hospital lengths of stay. Cause of death was categorized as cardiac, respiratory, malignancy, sepsis/multiple organ failure, neurologic, trauma, or other. Mode of death was categorized as limitation/withdrawal of life support, brain death, or unsuccessful cardiopulmonary resuscitation. Trajectory of death was categorized as sudden unexpected death, death from a potentially curable disease, death from a lethal congenital anomaly, or death from a progressive condition with intermittent crises. ³³

Statistical Analysis

For all measures, scores were calculated based on available data if at least 80% of the items in the measure were completed. For scores that represent a sum (i.e., ICG, PHQ-8, SPRINT, HCS), the sum of the responses to the completed items was standardized for any missing data, being multiplied by the total number of items in the measure and divided by the number of completed items. For scores that represent a mean (i.e., RAAS), the mean of the completed items was determined. If less than 80% of items were completed for any measure, the score was considered missing. Cronbach's coefficient α was also calculated for each of the measures as an index of the magnitude of internal consistency in this population, for the set of items comprising each score.

Data are summarized using frequencies and percentages for categorical variables, means and standard deviations for continuous variables, and medians and interquartile ranges for non-normally distributed continuous variables. Because more than one parent of the same deceased child could participate in the study, and the characteristics and outcomes of parents of the same child are likely correlated, linear mixed models were used with child as a random effect to incorporate this between-parent correlation when identifying variables associated with each of the outcomes. Candidate predictors for model selection were determined for each outcome as those variables with a p -value ≤ 0.2 in univariable linear mixed models. Parent-reported measures (i.e., RAAS, HCS, ICG, PHQ-8, SPRINT, overall health) were treated as continuous variables in the models. Multicollinearity was assessed for all candidate predictors, where two predictors had high correlation (>0.7), the most clinically relevant variable was included in the full model. To construct the final models, manual backward selection was performed on each multivariable mixed model containing candidate predictors until all remaining predictors had a p -value ≤0.1.

Results

Surveys were mailed to 710 families 6 months after their child's death. One or more parents from 159 (22.4%) families responded to the survey, 181 (25.5%) families refused and 370 (52.1%) could not be contacted. In total, 237 parents responded to the survey; two were later found to be ineligible because their child died in a location other than a PICU. Of the remaining 235 parents, 200 (85.1%) responded in English by mail, 22 (9.4%) in English by telephone, 10 (4.3%) in Spanish by mail, and three (1.3%) in Spanish by telephone. The final dataset included 235 parents of 158 deceased children. The number and percentage of families participating across sites ranged from 8 (5.1%) to 44 (27.8%) ( Supplementary Material 2 , available in the online version only).

Parents were 38.7 ± 10.6 years old; 147 (62.6%) were female, 157 (66.8%) were Caucasian, 177 (75.3%) were married, and 211 (89.8%) had at least a high school education ( Table 1 ). Parents' mean HCS score was 51.4 ± 11.1, and RAAS anxiety and avoidance scores were 2.1 ± 1.0 and 2.6 ± 0.7, respectively. Children had a median age of 5.9 years (IQR 0.64, 13.9 years) at the time of death; 11 (4.7%) were >21 years but had pediatric conditions and were dependent on their parents as caregivers ( Table 1 ). Eighty-seven (55.1%) children were male, 49 (31.0%) had a sudden unexpected death, 40 (25.3%) died from cardiac causes, and 114 (72.2%) underwent limitation or withdrawal of life support. Median PICU and hospital lengths of stay were 6.6 days (IQR 2.1, 21.8) and 8.7 days (IQR 2.4, 35.0 days), respectively.

Table 1. Parent and child characteristics.

Child characteristic	N  = 158
Gender
Female	147 (62.6%)
Male	88 (37.4%)
Age (years)
Mean (SD)	38.7 (10.61)
Race
White/Caucasian	157 (66.8%)
Black/African American	46 (19.6%)
Other	27 (11.5%)
Not Provided	5 (2.1%)
Ethnicity
Non-Hispanic or Latino	193 (82.1%)
Hispanic or Latino	36 (15.3%)
Not Provided	6 (2.6%)
Education
Some high school or less	22 (9.4%)
High school graduate or GED	62 (26.4%)
Vocational school or some college	64 (27.2%)
College degree or higher	85 (36.2%)
Not Provided	2 (0.9%)
Marital status
Married	177 (75.3%)
Single	56 (23.8%)
Not Provided	2 (0.9%)
Relation to child
Biological parent	215 (91.5%)
Other	20 (8.5%)
Additional children
None	42 (17.9%)
One	79 (33.6%)
Two	50 (21.3%)
Three or more	63 (26.8%)
Not Provided	1 (0.4%)
Child characteristic	N  = 158
Gender
Male	87 (55.1%)
Female	71 (44.9%)
Age (years)
Mean (SD)	7.9 (7.7)
Median (IQR)	5.9 (0.64, 13.9)
Trajectory of death
Sudden unexpected death	49 (31.0%)
Death from progressive condition with intermittent crises	44 (27.8%)
Death from potentially curable disease	36 (22.8%)
Death from a lethal congenital anomaly	29 (18.4%)
Mode of death
Limitation or withdrawal of life support	114 (72.2%)
Brain death	25 (15.8%)
Failed CPR	19 (12.0%)
Cause of death
Cardiac	40 (25.3%)
Sepsis or multiple organ failure	30 (19.0%)
Respiratory	25 (15.8%)
Neurologic	25 (15.8%)
Trauma	20 (12.7%)
Malignancy	13 (8.2%)
Other	5 (3.2%)
PICU length of stay (days)
Mean (SD)	22.2 (38.93)
Median (IQR)	6.6 (2.1, 21.8)
Hospital length of stay (days )
Mean (SD)	27.2 (42.4)
Median (IQR)	8.7 (2.4, 35.0)

Abbreviations: CPR, cardiopulmonary resuscitation; GED, General Educational Development test; IQR, interquartile range; PICU, pediatric intensive care unit; SD, standard deviation.

Parents' mean ICG score was 34.4 ± 14.9 at 6 months after their child's death with 142 (60.4%) having scores ≥30 ( Table 2 ). Mean PHQ-8 score was 9.1 ± 6.2 with 102 (43.4%) having scores ≥10. Mean SPRINT score was 14.6 ± 8.2 with 122 (51.9%) having scores ≥14. Overall health was rated as fair for 47 (20.0%) parents and as poor for 10 (4.3%).

Table 2. Parent health outcomes at 6 months after child's death.

	Overall ( N  = 235)
ICG score a
Mean (SD)	34.4 (14.94)
Min, Max	2.0, 73.0
Complicated grief (ICG ≥30) a	142 (60.4%)
PHQ-8 score b
Mean (SD)	9.1 (6.20)
Min, Max	0.0, 24.0
Depression (PHQ-8) categories b
None to mild (0 to <10)	126 (53.6%)
Moderate to moderately severe (10 to <20)	85 (36.2%)
Severe (≥20)	17 (7.2%)
Unknown	7 (3.0%)
SPRINT score c
Mean (SD)	14.6 (8.19)
Min, Max	0.0, 32.0
PTSD (SPRINT ≥14) c	122 (51.9%)
Overall health
Excellent	17 (7.2%)
Very good	58 (24.7%)
Good	94 (40.0%)
Fair	47 (20.0%)
Poor	10 (4.3%)
Unknown	9 (3.8%)

^aICG is Inventory of Complicated Grief. Total scores range from 0 to 76. Scores ≥30 suggest complicated grief.

^bPHQ-8 is Patient Health Questionnaire. Total scores range from 0 to 24; scores of 5, 10, 15, and 20 represent mild, moderate, moderately severe, and severe symptoms.

^cSPRINT is Short Post-Traumatic Stress disorder Rating Interview. Total scores range from 0 to 32. Scores ≥14 suggest post-traumatic stress disorder (PTSD).

The number of parents having at least 80% of items completed for the HCS, RAAS, ICG, PHQ-8, and SPRINT were 233 (99.1%), 233 (99.1%), 224 (95.3%), 228 (97.0%), and 229 (97.4%), respectively ( Supplementary Material 3 , available in the online version only). Cronbach's α for each measure was as follows: HCS 0.96, RAAS-anxiety 0.73, RAAS-avoidance 0.88, ICG 0.98, PHQ-8 0.99, and SPRINT 0.99.

Univariable analyses exploring associations between parent and child characteristics and parent health outcomes are displayed in Supplementary Material 4 (available in the online version only). Sets of variables with high correlation included hospital and PICU length of stay, and parent and child age. For outcomes having both variables in one of these sets as candidate predictors, PICU length of stay and child age, respectively, were chosen for the full models.

Table 3 shows final multivariable models for parent health outcomes. For each continuous variable included in a model (e.g., age, RAAS, HCS, length of stay), the estimate denotes the effect of a one-unit increase in the value of the variable on the continuous outcome (i.e., ICG, PHQ-8, SPRINT, overall health), adjusted for the effect of other variables also in the model. Higher HCS scores (i.e., greater therapeutic alliance) and longer PICU length of stay were significant predictors of lower (i.e., better) ICG scores. Female parent gender and older child age at the time of death were significant predictors of higher (e.g., worse) PHQ-8 scores. Female parent gender was a significant predictor of higher (e.g., worse) SPRINT scores, and death from lethal congenital anomaly (compared with sudden unexpected death) was a significant predictor of lower (i.e., better) SPRINT scores. Higher RAAS anxiety and avoidance scores (i.e., greater attachment-related anxiety and avoidance) were significant predictors of higher (i.e., worse) scores for all three mental health outcomes. Female parent gender, single parenthood, and higher RAAS anxiety score were significant predictors of worse overall parent health.

Table 3. Multivariable Models of Parental Health Outcomes.

Characteristics (N = 207)	Estimate (95% CI)	P-value
ICG Model a
Parent age (years)	−0.19 (−0.38, 0)	0.055
RAAS Anxiety Score b	3.38 (1.27, 5.5)	0.002
RAAS Avoidance Score b	3.7 (0.79, 6.6)	0.013
HCS Score c	−0.23 (−0.42, −0.04)	0.018
PICU length of stay (days) d	−0.05 (−0.11, 0)	0.042
PHQ-8 Model e
Parent Gender (female vs. male)	1.6 (0.07, 3.13)	0.040
Parent Ethnicity (Hispanic vs. non-Hispanic)	−2.23 (−4.5, 0.05)	0.055
RAAS Anxiety Score b	1.83 (0.87, 2.78)	<.001
RAAS Avoidance Score b	1.37 (0.06, 2.68)	0.040
HCS Score c	−0.07 (−0.14, 0)	0.059
Child age at time of death	0.12 (0.01, 0.22)	0.026
SPRINT Model f
Parent Gender (female vs. male)	2.53 (0.75, 4.31)	0.006
RAAS Anxiety Score b	1.96 (0.78, 3.14)	0.001
RAAS Avoidance Score b	3.15 (1.54, 4.77)	<.001
Death trajectory (Lethal congenital anomaly vs. sudden unexpected death	−3.72 (−6.73, −0.71)	0.016
Death trajectory (Potentially curable disease vs. sudden unexpected death)	0.04 (−3, 3.09)	0.977
Death trajectory (Progressive condition with intermittent crises vs. sudden unexpected death)	−1.18 (−4.01, 1.64)	0.405
Overall Health Model g
Parent Gender (female vs. male)	0.26 (0.03, 0.49)	0.026
Marital Status (single vs. married)	0.37 (0.06, 0.68)	0.022
RAAS Anxiety Score b	0.28 (0.13, 0.43)	<.001
RAAS Avoidance Score b	0.2 (−0.01, 0.4)	0.060
HCS Score c	−0.01 (−0.02, 0)	0.071

^aICG is Inventory of Complicated Grief. Higher scores indicate more grief symptoms.

^bRAAS is Revised Adult Attachment Scale. Higher scores indicated greater attachment-related anxiety or avoidance.

^cHCS is Human Connect Scale. Higher scores indicate greater therapeutic alliance.

^dPICU is pediatric intensive care unit.

^ePHQ-8 is Patient Health Questionnaire. Higher scores indicate more depression symptoms.

^fSPRINT is Short Post-Traumatic Stress Disorder Rating Interview. Higher scores indicate more PTSD symptoms.

^gOverall Health Score is parents' self-reported overall health. Parents responded to the following item: “In general, would you say your overall health is (1) excellent, (2) very good, (3) good, (4) fair, (5) poor.”

Discussion

Therapeutic alliance is a term used to reflect the strength and quality of the relationship between patients/families and their physician. This study is the first to explore the association between bereaved parents' perceptions of therapeutic alliance with their child's physician and parents' mental health symptoms following their child's death in a PICU. Overall, levels of therapeutic alliance reported by bereaved parents were similar to studies of adult patients and caregivers that utilized the HCS. ^{16 17 18 19} Similar to prior work, bereaved parents in our study experienced substantial adverse mental health symptoms including symptoms of complicated grief, depression, and post-traumatic stress. ^{6 7 9 10 11 12} Parents with stronger therapeutic alliances with PICU physicians demonstrated less symptoms of complicated grief during bereavement; however, therapeutic alliance was not significantly associated with lesser symptoms of depression, post-traumatic stress, or better overall health.

Complicated grief has been estimated to occur in approximately 7% of bereaved individuals with bereaved parents being among those at highest risk. ³⁴ In the current study, 60.4% of parents demonstrated ICG scores suggestive of complicated grief which is consistent with prior work. ⁶ Parents often rely on health professionals caring for their critically ill child to help meet their informational and emotional needs. Many of these needs are embodied in the concept of therapeutic alliance such as the need for compassion, trust, and honest communication. Prior research suggests that bereaved parents who perceive their needs as met by PICU staff have lower levels of complicated grief symptoms. ²⁰ In a study of cancer-bereaved parents, trust in the health professionals and the provision of practical support was associated with less complicated grief. ³⁵ In the current study, we used an adapted version of the HCS to specifically investigate parents' perceptions of therapeutic alliance with PICU physicians; however, the quality of relationship with other health professionals in the PICU is also relevant to parents' outcomes as demonstrated in prior research. ²⁰ Longer duration of PICU stay was also associated with lower levels of complicated grief symptoms.

Attachment theory posits that the close bonds that develop between infants and parents serve to provide the infant with a safe and secure base from which to explore their environment. ³⁶ Early attachment bonds remain important throughout the lifespan and become internal working models by which later relationships are judged. Attachment is often characterized by two dimensions: attachment-related anxiety and attachment-related avoidance. Individuals with high attachment anxiety tend to have a negative self-image and worry about the availability of others. Those with high avoidance tend to view others negatively and prefer not to rely on others. High attachment anxiety and avoidance represent insecure attachment styles. Individuals with attachment insecurity feel less safe in close relationships whereas those with secure attachment have a healthy approach to interpersonal closeness. In a study of infant loss, parents with insecure attachment styles showed greater psychopathology than those with secure attachment. ³¹ Insecure attachment has also been related to the prevalence of several medical conditions. ³⁷ Given this information, we suspected bereaved parents with attachment insecurity would have more mental health symptoms; this was observed in all three mental health domains, as well as overall health.

An estimated 17.3 million adults in the United States have had at least one major depressive episode, representing 7.1% of all adults. ³⁸ Depression is commonly reported among parents experiencing child loss. ^{9 10 11 12} In our study, mothers had greater symptoms of depression than fathers consistent with prior research on parental bereavement ^{9 10} and higher prevalence of major depressive episodes in women compared with men in the general U.S. population. ³⁸ Greater symptoms of depression were also seen in parents of children that died at an older age. Prior studies of pediatric cancer-related deaths found that parents of deceased adolescents showed an increased risk for depression, possibly due to the breaking of more established affectional bonds or the complex nature of the parent–adolescent relationship. ^{39 40} Depressive symptoms were not significantly associated with the extent of therapeutic alliance.

The estimated lifetime prevalence of PTSD among adult Americans is 6.8% and is higher in women than men. ^{41 42} In our study, bereaved mothers had greater post-traumatic stress symptoms than fathers. In a recent study by Baumann et al, 75% of bereaved parents enrolled in a mental health family support program showed a clinically relevant degree of post-traumatic stress symptoms, and mothers were at higher risk. ⁹ Pohlkamp et al found that mothers have more post-traumatic stress symptoms than fathers in years 1 to 3 after their child's death, but fathers' symptoms increase with time and are similar to those of mothers in years 4 to 5. ¹⁰ In our study, parents of children who had a sudden, unexpected death had more post-traumatic stress symptoms than parents of children who died of congenital anomalies. Landmark bereavement work by Murphy et al found that parents whose children die by accidents, suicides, or homicides have high rates of PTSD that persist for as long as 5 years post-death. ⁴³ Parents whose children die by accidents, suicides of homicides may perceive the events as preventable, prompting symptoms of post-traumatic stress. ⁴³ Post-traumatic stress symptoms were not significantly associated with the extent of therapeutic alliance.

Bereaved parents have reported higher rates of physical morbidity including acute illnesses, substance abuse, insomnia, and utilization of health care services. ^{13 14} Given these known risks we included a single item in our survey about overall health. In our study, mothers reported lower overall health than fathers. In a recent study by Brooten et al, mothers of children who died in a neonatal ICU or PICU 1 to 13 months earlier, had higher physical morbidity than fathers. This was evidenced by 300 acute illnesses and 89 hospitalizations in 176 mothers versus 104 acute illnesses and nine hospitalizations in 73 fathers. ¹⁴ Single parents in our study also reported lower overall health than parents who were married. In the absence of a partner, lower self-reported health could be the consequence of less social support during bereavement. ⁴⁴ Overall health was not significantly associated with the extent of therapeutic alliance in our study.

Strengths of this study include the multicenter design and racial, ethnic, and geographic diversity of the participants. Limitations include 22% response rate, typical of surveys of bereaved parents and likely due to bereavement-related distress and added burden of research participation. Characteristics of nonresponders are unknown because no information beyond parental inclusion criteria and contact information was obtained from the medical records of children whose parents ultimately refused to participate or could not be contacted. Most participating parents were married and had at least a high school education which influences the generalizability of our findings, as does the unequal distribution of parents across sites. HCS, our measure of therapeutic alliance, was validated in an adult oncology population and adapted for use in the PICU; thus, its validity and reliability among bereaved PICU parents is unknown. Although the focus of our study was parent-intensivist alliance, we acknowledge that parent responses to HCS items may have been influenced by a general sense of alliance with PICU staff or other subspecialists rather than alliance with intensivists per se since parents typically interact with multiple health professionals in the PICU environment. Responses to the HCS may also have been affected by parent personality characteristics, attachment style, and normal grief as these may influence how parents perceive their alliance with physicians. Attachment style, in turn, may have been affected by the trauma of child death. HCS scores may have been affected by recall bias since parents completed the measure approximately 6 months after their child's death. The study evaluation time point of 6 months post-death was selected because criteria for diagnosis of complicated grief include persistence of symptoms for at least 6 months. ^{3 4 5} Overall health was assessed by parent self-report rather than by conditions documented by a health care provider. The study design is cross-sectional, and thus associations cannot be taken to represent causality. For example, parents with more complicated grief symptoms 6 months after their child's death may have more negative recollections of their alliance with physicians and vice versa. Ideally, therapeutic alliance between parents and physicians would be evaluated during the child's PICU admission, and parents' mental health symptoms would be evaluated in the months following the child's death; this could have prevented our assessment of alliance from being influenced by parents' mental health during bereavement. Key components of therapeutic alliance (e.g., understanding, caring, trust) are attributes of high-quality physician–patient/family relationships also described in other clinical settings such as palliative care, primary care, and oncology where long-term relationships are the norm. Our findings suggest that even in the short-term, high-tech, high stress settings such as PICUs, relationships matter and may be associated with bereaved parents' grief symptoms. However, the characteristics and behaviors of physicians which facilitate high quality relationships with parents were not investigated and require further study. Based on the extent of mental health symptoms observed in our cohort of bereaved parents, the provision of mental health services by child psychiatrists/psychologists during a child's PICU stay, and the support of mental health professionals for parents after PICU discharge should be considered.

Conclusion

Parents whose children die in PICUs experience high levels of adverse mental health symptoms including complicated grief, depression, and PTSD. Greater therapeutic alliance with PICU physicians is a potential opportunity to lessen complicated grief symptoms; however, therapeutic alliance was not associated with depression or post-traumatic stress. Parent factors such as gender, marital status, and attachment insecurity, as well as child factors such as age, length of stay, and trajectory of death, may influence parental mental health symptoms during bereavement.

Funding Statement

Funding This study was supported in part, by the following cooperative agreements from the Eunice Kennedy Shriver National Institute of Child Health and Human Development: UG1HD083170, UG1HD049981, UG1HD63108, UG1HD083166, UG1HD083171, UG1HD049983, U01HD049934, UG1HD050096.