Identifying patterns of pediatric mental and behavioral health at end of life: A national study

Abstract

Children, who enroll in hospice, have complex mental and behavioral health (MHBH) problems. There is limited literature on patterns of these problems among children their end-of-life. Using the national database of 6,195 children enrolled in hospice between 1/1/2011 and 12/31/2013, and latent class analysis, this study identified four distinctive classes of children. Neurodevelopment & Anxiety class (26.5% of the sample) had 41.3% of children of age 15 and older; who averaged with 1.4 MHBH comorbidities and took 17 medications; 29.1% had physical health comorbidities, and 23% were dependent on technology. In the Behavior & Neurodevelopment class (20%) 53.1% of the children were between ages 6 to 14 years, averaged with 2 MHBH comorbidities, and took 17 medications. In the Physiology class (43.4%) 65.5% of children were younger than 5 years, had 1 MHBH comorbidity, and took 19 medications. In the Mood, Behaviors, & Anxiety class (10.2%) more than 90% of children were older than 6 years, had 4 MHBH comorbidities, and took 30 medications. In the latter three classes around half of the children either had physical health comorbidities or were technology dependent. These findings highlight the importance of end-of-life care that accounts for the likely presence of complicated patterns of MHBH conditions.

Each year in the United States more than 40,000 children die of health-related conditions.¹ Among this pediatric population, approximately a third use hospice care at end of life.^2,3 Pediatric hospice care provides children with a 6 month-to-live prognosis and their families a range of supportive care services to manage pain and symptoms, along with psychosocial supports (e.g., expressive therapies). The pediatric hospice team is often composed of nurses, physicians, pharmacists, social workers, child life specialists, chaplains, respiratory therapists, and dieticians. Although up to 60% of children enter hospice with complicated physical health problems, there is emerging evidence that their mental and behavioral health (MHBH) may contribute to a more complex picture of serious health problems.^4,5 MHBH is defined as a person’s psychological well-being, which includes emotions, behaviors and biology.^6,7 Furthermore, MHBH conditions can range from a single episode of anxiety to more chronic conditions such as attention deficit disorder.⁸

Much of what is known about MHBH at end of life comes from the adult literature, which has focused on anxiety and depression. However, recent work has provided preliminary evidence about pediatric MHBH in hospice care. First, MHBH conditions are relatively prevalent among children in hospice. Keim-Malpass and colleagues² found that 34% of children admitted to hospice care had a mental and/or behavioral health diagnosis. Among children in concurrent hospice care, the percentage increased to 42%.⁵ Second, the range of MHBH diagnoses is broader than just anxiety and depression and varies by age group.⁹ Disorders included anxiety (31.0%), depression (33.1%), behavioral disorders (33.9%), and other disorders (34.8%), which were highest among children 15 to 20 years. Finally, there is a relationship between MHBH and the type of care delivered at end of life. We have found that children with MHBH problems were more likely to use pediatric concurrent versus standard hospice care,² even among those with cancer.¹⁰

Although this emerging literature suggests that pediatric MHBH is complex, there are gaps in our knowledge. It is still unclear how MHBH contributes to the medical complexity of children in hospice care. While we know that MHBH conditions often co-occur, we lack information about patterns of pediatric MHBH at end of life. For example, we might expect that substance misuse disorders to accompany bipolar disorders.¹¹ Additionally, groupings of MHBH might have unique demographic and health characteristics, which is unknown.

Improving our knowledge of pediatric MHBH patterns at end of life is important and timely for hospice and palliative care nurses. Given the current hospice environment of COVID-19, the isolation experienced by children at end of life may intensify MHBH conditions. Recent COVID-19 data has shown a significant increase in mental health issues in the general population,¹² and the same would also be expected among children in hospice care who have been very isolated since the outbreak of COVID-19. In addition, the recent move towards telehospice may complicate the hospice teams’ ability to assess pediatric mental and behavior health. While MHBH assessments often include formal questions for the child and family, the environment, physical cues, and reactions contribute to the assessment.¹³ Improving our knowledge of MHBH has implications for nurses caring for children at end of life. Therefore, the purpose of this study was to identify patterns of MHBH and describe the demographic and health characteristics of children in these groups.

Methods

Study Design and Data Source

A retrospective, nonexperimental study design was used with the database of national Medicaid claims collected by the Centers of Medicare and Medicaid Services (CMS). These data are person-level files prepared by CMS.¹⁴ For this study, we used the Medicaid Analytic eXtract files which included Personal Summary, Other Therapy, Inpatient, and Prescription Drugs. Medicaid data was used for this study because it is the most common insurer of children with life-limiting conditions.^2,3 These data were selected because they were one of the very few data sources that included national-level information about MHBH conditions of children enrolled in hospices.¹⁵ Data were collected between 2011 and 2013 because they were the latest available at the time of this study.

Sample

The unit of analysis was pediatric decedents. Children were defined as under 21 years. Inclusion criteria were a documented mental health and/or developmental diagnosis, which was based on primary and secondary ICD-9-CM codes from the Medicaid files as recommended by Garfield et al. (2015) and admission to hospice between 1/1/2011 and 12/31/2013. Exclusion criteria were no service-level Medicaid claims or missing data (i.e., date of birth, date of death). The total sample included 6,195 children enrolled in hospice care who were diagnosed with MHBH disorders. The study was reviewed and approved by the University of Tennessee, Knoxville Institutional Review Board.

Measures

MHBH conditions were identified using the 9th revision of the International Classifications of Diseases, Clinical Modification (ICD-9-CM): anxiety (ICD-9-CM: 300, 308), neurodevelopment disorders (ICD-9-CM: 299, 314), psychotic disorders (ICD-9-CM: 293-295, 297, 298, 316), substance misuse (ICD-9-CM: 291-292, 303-305), cognitive (ICD-9-CM: 290, 310, 315, 317-319), mood disorders (ICD-9-CM: 300.4, 296.1-296.9, 311), behavioral disorders (ICD-9-CM: 312-313), and other mental health conditions (ICD-9-CM: 301-302, 306-307).^16-18 Separate variables were also constructed to count the number of MHBH conditions per child.

Demographics included age, gender, race, ethnicity, rural/urban, and region. Age group categories were <1 year, 1 to 5 years, 6 to 14 years, and 15 to 20 years. Gender was dichotomized as female and male. The race was operationalized as White, Black, and other races, while ethnicity was defined as Hispanic and non-Hispanic. Rural/urban was operationalized using the definition of the Health Resources & Services Administration.¹⁹ Regions were categorized using the US Census distinction as Midwest, Northeast, South, and West.

Health characteristics included mental/behavioral health comorbidities, physical health comorbidities, polypharmacy, and technology dependence. Mental/behavioral health comorbidities were the number of mental or behavioral health conditions. Physical health comorbidities were two or more complex chronic conditions.^20,21 Polypharmacy was defined as the number of medications a child was simultaneously receiving.^22,23 Technology dependence was operationalized as whether or not a child was dependent on medical technology or a device.²⁴

Analysis

The goal of this study was to identify patterns of MHBH and describe the demographic and health characteristics of children in these groups. First, counts of MHBH conditions were calculated for the sample. Second, latent class analysis (LCA) was used to identify patterns within the sample. To determine the optimal number of classes or clusters, several LCA models were constructed (1 to 6 classes) and tested with multiple goodness-of-fit indices (i.e., Log-Likelihood (LL) ratio), consistent Akaike information criterion (CAIC), Bayesian information criterion (BIC), sample-size adjusted BIC (SABIC), entropy, and Lo-Mendell-Rubin (LMR test).²⁵ Conditional probabilities were calculated for MHBH conditions within the optimal number of clusters. The conditional probabilities within each class informed how the classes were labeled including probabilities >0.10, inpatient, outpatient, versus home-based, and number of conditions. Finally, descriptive statistics were calculated for class membership in each class. All data analysis was conducted using Stata 15.0.²⁶

Results

Data from 6,195 children with an MHBH disorder between 2011 and 2013 comprised the cohort for this study. Figure 1 presents the prevalence of MHBH conditions. Among all the children in the study, the most common condition was cognitive disorders (n=3,297), followed by behavioral disorders (n=1,698) and neurodevelopment disorders (n=1,533). The least common conditions were psychotic disorders (n=461) and substance misuse (n=441).

Figure 1.

MHBH conditions of children in hospice care

The results of the LCA model testing are listed in Table 1. We evaluated models with 1 to 6 classes and, based on the goodness-of-fit indices, the 4-class model was optimal. This model had low values of LL ratio, CAIC, SABIC, and BIC indices. The entropy and LMR tests did not provide evidence for class fit. The composition of the 4-class model is shown in Table 2. Class 1 - Neurodevelopment & Anxiety included 26.5% of the children. Twenty percent of the children were in Class 2 -Behavior & Neurodevelopment. Class 3 - Physiology represented 43.4% of the sample, and Class 4 - Mood, Behaviors, & Anxiety was 10.2%. The distribution of MHBH conditions within each class is graphically represented in Figure 2.

Table 1.

Model fit indices for LCA models (N=6,195)

	LL	CAIC	BIC	SABIC	Entropy	LMR probability
One-class model	−22485.000	45047.852	45039.780	45014.430	NA	NA
Two-class model	−20686.600	41538.635	41512.840	41467.614	.823	<.001
Three-class model	−20474.900	41202.819	41176.830	41094.198	.790	<.001
Four-class model	−20337.100	41014.802	40979.780	40868.582	.871	<.001
Five-class model	−20074.980	40578.146	40534.140	40394.326	.879	<.001
Six-class model	−20108.150	40732.069	40679.080	40510.650	.796	<.001

Note. LL= Log likelihood; CAIC = consistent Akaike information criterion; BIC = Bayesian information criterion; SABIC = sample adjusted BIC; LMR = Lo-Mendell Rubin test; NA = not applicable.

Table 2.

Conditional probabilities of class membership for four-class model (N= 6,195)

	Class 1 Neurodevelopment & Anxiety (n=1,648; 26.49%)	Class 2 Behavior & Neurodevelopment (n=1,284; 19.94%)	Class 3 Physiology (n=2,749; 43.36%)	Class 4 Mood, Behavior, & Anxiety (n=514; 10.21%)
Anxiety	.291	.128	.020	.594
Neurodevelopment	.352	.349	.105	.384
Psychotic	.061	.043	.035	.339
Substance	.148	.036	.006	.218
Cognitive	.077	.276	.999	.227
Mood	.211	.082	.001	.906
Behavior	.012	.999	.001	.701
Other	.211	.090	.057	.193

Figure 2.

Conditional item probability plot for four-class model (N= 6,195)

The class membership demographic and health characteristics of children are presented in Table 3. In Class 1 - Neurodevelopment & Anxiety, almost three-quarters of children (41.3%) were 15 and older; they had on average 1.4 MHBH comorbidities (M=1.4, SD=.67); and took on average 17 medications simultaneously (M=17.1 SD=16.25). Slightly less than one in three children (29.1%) had physical heath comorbidities, and 23% were dependent on technology. The Class 2 - Behavior & Neurodevelopment had a majority of children 6 to 14 years (53.1%) who on average had 2 MHBH comorbidities (M=2.0, SD =.95) and 17 medications (M=16.9, SD=15.6). Approximately half of children in this class either had physical health comorbidities (50.2%) or were dependent on technology (45.2%).

Table 3.

Characteristics of children in hospice care within four-class model

Variables	Overall	Class 1 Neurodevelopment & Anxiety (n=1,648)	Class 2 Behavior & Neurodevelopment (n=1,284)	Class 3 Physiology (n=2,749)	Class 4 Mood, Behavior, & Anxiety (n=514)
Demographics
Age Groups (%)
0 to 5 years	39.5	20.6	24.8	65.5	<10.00
6 to 14 years	35.7	38.1	53.1	24.5	42.2
15 to 20 years	24.8	41.3	22.1	10.0	55.8
Gender (%)
Female	45.1	49.4	43.1	42.2	52.3
Male	54.9	50.6	56.9	57.8	47.7
Race (%)
White	57.7	60.0	62.8	52.6	65.0
Black	23.7	26.3	28.7	19.6	24.3
Other Races	18.6	13.7	<10.0	27.8	10.7
Ethnicity (%)
Hispanic	18.1	18.3	17.2	19.0	15.2
Non-Hispanic	81.9	81.7	82.8	81.0	84.8
Rural/Urban (%)
Urban	63.0	61.5	64.6	63.6	60.7
Rural	37.0	38.5	35.4	36.4	39.3
Region (%)
Midwest	29.0	32.5	37.6	22.4	31.5
Northeast	45.6	47.1	46.0	44.7	44.4
South	16.9	13.7	10.8	22.0	14.8
West	<10.0	<10.0	<10.0	10.9	<10.0
Health
Mental/Behavioral Health Comorbidities (M, SD)	1.7 (1.05)	1.4 (0.67)	2.0 (.95)	1.2 (0.47)	4.0 (1.15)
Physical Health Comorbidities (%)	46.7	29.1	50.2	54.7	51.4
Polypharmacy (M, SD)	19.0 (17.47)	17.1 (16.25)	16.9 (15.60)	19.1(17.02)	28.8 (23.44)
Technology Dependence (%)	40.4	23.4	45.2	47.3	45.3

Note. <10.0 is noted for values under 10% per Data Use Agreement; M=mean; SD=standard deviation

In Class 3 – Physiology, more than two-thirds (65.5%) of children were younger than 5, had on average one MHBH comorbidity (M=1.2, SD=0.47), and averaged 19 medications (M=19.1, SD=17.02). A majority of children in this class (54.7%) had physical health comorbidities and 47.3% were dependent on technology. More than 90% of children in Class 4 - Mood, Behavior, & Anxiety were older than 6 years of age, had on average four MHBH comorbidities (M=4.0, SD=1.15), and took almost 30 medications (M=28.8, SD=23.44). Physical health comorbidities were common in 51.4% and technology dependence for 45.3% of children in this class.

Discussion

In a Medicaid-based sample of pediatric decedents, four distinct latent classes adequately accounted for variation in patterns of eight clinically important MHBH conditions. Each of these classes exhibited unique disorder patterns and demographic profiles from one another. Thus, this analysis presented a novel way of understanding clusters of MHBH among children that may inform tailored and targeted end-of-life care for these children.

The findings revealed a significant number and wide variety of MHBH conditions among children in the study. We identified approximately 10,000 diagnoses at end of life among the 6,195 children. In addition, we found that the 4 clusters of diagnoses also varied significantly. As an example, Class 3 – Physiology was predominately composed of only cognitive and neurodevelopment disorders, whereas Class 4- Mood, Behavior, & Anxiety included all diagnoses.

The classes had unique membership composition. Class 4 - Mood, Behavior, & Anxiety was most often older children (15-20 years) and female with significant health issues including the highest incidence of mental/behavioral health comorbidities, physical health comorbidities, and polypharmacy, compared to the other groups. Class 4 - Mood, Behavior, & Anxiety by comparison appeared to be the sickest of the children in the sample. Our prior work examining patterns of physical health of children at end of life identified a cluster of a medically fragile group of children.²⁷ However, the current study builds on that work to suggest that the medically fragile group may also be fragile in their MHBH. The findings highlight a special pediatric population who may have extreme health care needs at end of life. There is a need to continue examining who these children are at end of life and how we might leverage empirical data to refine definitions of clusters.

We found that three of the four classes (Class 1 - Neurodevelopment & Anxiety, Class 2 -Behavior & Neurodevelopment, and Class 4 - Mood, Behavior, & Anxiety) were characterized with a diagnosis of anxiety. These three classes represented more than 56% of the sample and anxiety was present in 29.1%, 12.8%, and 59.8%, respectively. This finding was consistent with prior research. Baker and colleagues found in a systematic review and meta-analysis that anxiety is a common pediatric condition at end of life and also that anxiety and depression often occurred together.²⁸ While our findings indicated that anxiety and mood disorders which included depression frequently co-occurred, our evidence suggests that anxiety may be more complicated. It appears to co-occur with more MHBH conditions than just depression. In Class 1 - Neurodevelopment & Anxiety, for example, anxiety presented with neurodevelopment disorders, substance misuse, mood disorders, and other conditions. Anxiety at end of life might be a more complex co-condition that requires a thorough assessment and treatment that accounts for multiple comorbidities.

Among the class membership characteristics, polypharmacy had interesting variation between the classes. Class 2 - Behavior & Neurodevelopment had the lowest average number of medications at 16 medications and Class 4 - Mood, Behavior, and Anxiety had the highest average with 29 medications. Ten or more medications are frequently referred to as excessive polypharmacy,²⁹ while 20 or more medications has been defined as extreme polypharmacy.³⁰ In our study, we found polypharmacy mirrored the increased probability of MHBH conditions in a class. In other words, as the number of conditions rose, so did the number of medications. Although our data did not allow us to separate medications for physical health versus medications for MHBH, the evidence suggests that the management of complex clusters such as Class 4 - Mood, Behavior, & Anxiety may require significantly more medications to treat multiple disorders. This raises an interesting question about medications at end of life for children: How is polypharmacy dealt with in pediatric hospice care? In the adult hospice literature, there is increasing clinical guidance about discontinuing medications that are not related to comfort (e.g., statins).³¹ It is undocumented whether pharmacologically treating MHBH contributes to end-of-life goals of care and the comfort of the child. Future research is needed to explore the role and effectiveness of polypharmacy in pediatric hospice care with a special focus on psychiatric medications.

Our study had several limitations that are worth noting. First, all the data was collected between 2011 and 2013, and all mental diagnoses were classified using the previous 9th version of the International Classification of Diseases (ICD-9). However, several studies have shown that both the 9th and the 10th versions of ICD perform similarly in classifying children with complex chronic conditions enrolled in hospice.³² The second limitation was that all MHBH diagnoses were coded using the ICD classification system instead of a more commonly used Diagnostic and Statistical Manual of Mental Disorders (DSM). This was the limitation of the original database, which did not include DSM codes. Finally, the sample in this study was limited only to children enrolled in Medicaid and did not include children who had only private insurance plans. However, Medicaid is the largest insurer of children at end life in the US which covers children of all socioeconomic statuses and geographic regions.

These findings have important clinical implications for hospice and palliative care nurses. In this study children at end of life had multiple MHBH conditions, which clustered in unique patterns. These findings highlight the importance of end-of-life care that accounts for the likely presence of complicated and multiple MHBH conditions. At least two issues call for the involvement of an experienced psychiatric clinician in the care of this group: 1) the wide variety of disorders observed and 2) the significant level of substance use disorders (over 20% of children in Class 4 - Mood, Behavior, & Anxiety).

Differentiating among multiple diagnoses while avoiding both under and over treatment^9,33 necessitates that an experienced psychiatric clinician conducts a detailed assessment and collaborates with the child and family to develop an individualized plan of care. Furthermore, suicide is now the second leading cause of death among those 10-17 years old and is on the rise in the United States.³⁴ The increased risk for suicide associated with anxiety disorders, depressive disorders, and substance abuse is well known.^35,36

Since many individuals do not respond optimally to first line treatments which often include psychiatric medications,³⁷ children at end of life should be assessed and managed by a psychiatric specialty provider with expertise in selecting from available treatment options. We urge hospice and palliative care nurses to familiarize themselves with local psychiatric treatment resources and colleagues, to facilitate the process of psychiatric referral or ideally, to add a psychiatric clinician as a permanent part of the pediatric hospice and palliative care team.