OBJECTIVES:

To determine if sociodemographic factors or underlying mental health conditions serve as predictors for prolonged length of stay (pLOS) in children hospitalized for suicidal ideation (SI) or suicide attempt (SA) requiring transfer to psychiatric facilities. We hypothesized an association between certain patient and hospital characteristics and pLOS.

METHODS:

For this retrospective cross-sectional study, we used the National Inpatient Sample. We included children <18 years old hospitalized with a primary or secondary International Classification of Diseases, 10th Edition, Clinical Modification diagnosis of SI or SA who were dispositioned to psychiatric facilities from 2016 to 2017. Exposures were patient sociodemographics, underlying mental health diagnoses, and hospital characteristics. Our outcome was pLOS. Adjusted prevalence ratios with 95% confidence intervals (CIs) were generated with log binomial regression.

RESULTS:

Of 12 715 hospitalizations meeting inclusion criteria, 5475 had pLOS. After adjusting for sociodemographics and hospital characteristics, predictive factors for pLOS were public insurance use (prevalence ratio: 1.40; CI: 1.12–1.78), urban nonteaching hospital location (prevalence ratio: 4.61; CI: 2.33–9.12), urban teaching hospital location (prevalence ratio: 3.26; CI: 1.84–5.76), and underlying diagnosis of mood disorder (prevalence ratio: 1.98; CI: 1.63–3.42). Hispanic patients had decreased probability of pLOS (prevalence ratio: 0.69; CI: 0.52–0.93). Otherwise, age, zip income, sex, and hospital region were not predictive of pLOS.

CONCLUSIONS:

Among children hospitalized for SI or SA requiring transfer to psychiatric facilities, public insurance, urban hospital location, and diagnoses of mood disorder, depression, and bipolar disorder were predictive of pLOS. Further research is needed on how to decrease disparities in length of stay among this vulnerable population.

Suicide is the second-leading cause of death in United States adolescents, with rates rising by >50% from 2007 to 2017.1,2  With a national shortage in pediatric mental health services and psychiatric beds, annual encounters for suicide attempt (SA) and suicidal ideation (SI) have more than doubled in the past decade.3,4  More than half of these encounters result in hospitalization to a medical service, either for medical treatment or for “boarding” until transfer to a psychiatric facility can be made.3,511  Psychiatric hospitalizations to general medical facilities are costly and compose ∼10% of child hospitalizations nationally.1214 

Inpatient length of stay (LOS), or the amount of time from initial treatment of injury to discharge or transfer to a psychiatric hospital, is variable among pediatric SI and SA hospitalizations.3  Predictive factors of LOS for children with SI or SA have been studied in the emergency department (ED) setting.1520  However, few data exist on how socioeconomic and underlying mental health conditions can influence LOS for pediatric patients with SI or SA in the inpatient setting. Thus, research focused on predictors of prolonged length of stay (pLOS) could inform targets for quality improvement efforts and policy change.

In our study, we seek to determine if socioeconomic or underlying mental health diagnoses serve as predictors for pLOS in children hospitalized with SI or SA before transfer to a psychiatric facility. Because there is no standardized definition for pLOS among admitted children with SI or SA, we chose to define pLOS as an LOS longer than the median LOS of our target population.21  We hypothesize that Black and Hispanic race and ethnicity, as well as publicly insured payer status, will be associated with pLOS, given that previous researchers have shown disparities in inpatient LOS among racial and ethnic minorities and publicly insured children.2225 

This study was a retrospective cross-sectional analysis of pediatric inpatient hospitalizations in the United States from 2016 to 2017 conducted by using the Healthcare Cost and Utilization Project National Inpatient Sample (NIS), sponsored by Agency for Healthcare Research and Quality.26  The Healthcare Cost and Utilization Project uses a 2-stage cluster sampling design, which first stratifies participating nonfederal community hospitals by 5 characteristics: urban or rural location, number of beds, geographic region, type of ownership, and teaching status. Then, 20% of hospitals from each stratum are selected by using a systematic random sampling technique. In the second stage, all inpatient hospitalizations from hospitals selected during stage 1 are selected for inclusion in the NIS to create the sample annually. NIS data are taken from hospitals that compose 97% of all discharges in the United States. Its data approximate a 20% systematic sample that is representative of the population of all inpatient hospitalizations on critical hospital and patient characteristics. Forty-seven states participated in the NIS, representing >35 million hospitalizations when weighted in 2017, the latest year for which the data were available at the time of the study. Because NIS data are deidentified and publicly available, this study was classified as exempt by our college’s institutional review board.

Our target population consisted of pediatric patients aged 5 to 17 years with a diagnosis of SI or SA admitted to hospitals in the United States between January 1, 2016, and December 31, 2017, who were discharged to psychiatric facilities. We used International Classification of Diseases, 10th Edition, Clinical Modification (ICD-10-CM) codes to identify patients with primary or secondary diagnoses of SI or SA (Supplemental Table 4).27  Patients’ disposition to psychiatric facilities was obtained from the “Transfer Other: Skilled Nursing Facility (SNF), Intermediate Care Facility (ICF), Another Type of Facility” option under the DISPUNIFORM variable in the NIS data. This disposition option includes patients discharged or transferred to a psychiatric hospital or a distinct psychiatric unit of a hospital. Although the selected disposition option includes being transferred to other facilities in addition to psychiatric facilities, we assume that the patients with a diagnosis of SI or SA were most likely transferred to psychiatric hospitals for detailed evaluation and treatment of their conditions. On the basis of highest prevalence of underlying mental health diagnoses among our study population, we selected 5 conditions for further analyses: depressive episode or recurrent depressive disorder, anxiety disorder, attention-deficit/hyperactivity disorder (ADHD), bipolar affective disorder, and mood disorder.

For each inpatient hospitalization, the NIS database captures various sociodemographic, clinical, and hospital characteristics. Self-reported race or ethnicity, which is reported differently across states, was standardized by grouping into non-Hispanic white, non-Hispanic Black, Hispanic, or other. Patient age in years was categorized as 5 to 11, 12 to 14, and 15 to 17 years, on the basis of definitions of childhood, early adolescence, and late adolescence.28  We chose a lower age limit of 5 years because the vast majority of mental health conditions typically do not present before the age of 5.29  Socioeconomic status was estimated from the median household income in the patient’s zip code of residence, and estimated values were classified into quartiles. Insurance status was based on the primary payer for the hospitalization and was classified into public payer, private payer, self-pay, or other or missing. “Public payer” includes Medicaid, Children’s Health Insurance Program, and Medicare. Patient place of residence was categorized into metropolitan area, micropolitan area, or other. We also considered several characteristics of the treating hospital that are captured within the NIS database, including US census region (Northeast, Midwest, South, or West) and location or teaching status (urban teaching, urban nonteaching, or rural).

Statistical analyses were performed by using R (version 3∙6∙1) and RStudio (Version 1∙2∙5001). LOS was defined as the length of time a patient was hospitalized to a medical facility before transfer to a psychiatric facility. Descriptive statistics were used to outline the mean LOS among various sociodemographic, clinical, and hospital characteristic groups in our target population described above. Mean LOS was chosen given normal distribution of our data. We defined pLOS as a stay >3 days, which was the median LOS of our target population.21  We calculated the prevalence of pLOS among each of the patient, hospital, and mental health diagnoses. We then conducted bivariate analyses using Pearson’s χ2 test to test the association between each of the inpatient stay characteristics and pLOS. We also conducted analysis of variance testing between each hospitalization characteristic and mean LOS and Tukey’s honest significance test for multiple comparisons within the subgroups.

We conducted adjusted survey log binomial regression to generate adjusted prevalence ratios and 95% confidence intervals (CIs) representing the association between each patient and hospital characteristic as exposure and pLOS as outcome among our target population. We excluded records with missing information from our variables of interest. We also created individual multivariable association models for each of the mental health diagnoses and pLOS while adjusting for patient and hospital characteristics. We assumed a 5% type I error rate for all hypothesis tests.

Of 1 804 764 pediatric hospitalizations during the study period, 139 440 (7.7%) had a primary or secondary diagnosis of SI or SA. Of these SI or SA hospitalizations, 12 715 (9.11%) required transfer to a psychiatric facility. The mean and median LOS for hospitalized pediatric patients requiring transfer to a psychiatric facility were 6.27 and 3 days, respectively. Table 1 summarizes patient mean LOS by selected sociodemographic characteristics and mental health diagnoses. Approximately 57% of our target population identified as non-Hispanic white. The majority were between ages 15 and 17 years, female, residing in metropolitan areas, and hospitalized at urban teaching hospitals. A total of 53% reported public insurance. The mean LOS varied significantly by age, type of health insurance coverage, and underlying mental health diagnoses. The mean LOS was significantly greater among children <13 years old. Self-paying patients had a significantly lower mean LOS. Patients with mood disorder or ADHD experienced a significantly greater mean LOS than those without each of these diagnoses. By contrast, patients with anxiety disorder had similar mean LOS compared with their counterparts without anxiety disorder. Although some subtle differences could be discerned in mean LOS by race and ethnicity and sex, those were not statistically significant.

TABLE 1

Patient Characteristics, Underlying Mental Health Diagnoses, and Mean LOS Among Hospitalized Pediatric Patients With a Diagnosis of SI or SA Who Required Transfer to a Psychiatric Facility (Total N = 12 715)

n (%)χ2PMean LOS (SD)ANOVA P
Race and ethnicity  .055  <.001 
 Non-Hispanic white 7271 (57.2) — 6.35 (12.1) — 
 Non-Hispanic Black 1601 (12.6) — 6.13 (9.4) — 
 Hispanic 1630 (12.8) — 7.37 (13.7) — 
 Other 980 (7.7) — 6.43 (8.6) — 
 Missing 1233 (9.7) — 4.48 (7.7) — 
Age, y  .008  <.001 
 15–17 7105 (55.9) — 5.88 (9.3) — 
 12–14 4290 (33.7) — 6.67 (12.3) — 
 5–11 1320 (10.4) — 7.14 (17.3) — 
Zip income quartile  .484  .002 
 Highest quartile 2833 (22.3) — 6.4 (10.3) — 
 Third quartile 3298 (25.9) — 5.77 (9.8) — 
 Second quartile 3199 (25.2) — 6.11 (12.4) — 
 Lowest quartile 3230 (25.4) — 6.89 (12.9) — 
 Missing 155 (1.2) — 5.65 (7.5) — 
Primary payer  .031  <.001 
 Private insurance 5265 (41.4) — 5.61 (9.1) — 
 Public insurance 6746 (53.1) — 6.25 (8.2) — 
 Self-pay 669 (5.3) — 5.18 (9.4) — 
 Other or missing 35 (0.3) — 10.86 (15.7) — 
Sex  .092  <.001 
 Male 4128 (32.5) — 6.98 (14.3) — 
 Female 8587 (67.6) — 5.94 (9.7) — 
Place of residence  .492  <.001 
 Metro areas 10 836 (85.2) — 6.35 (11.8) — 
 Micropolitan areas 1029 (8.1) — 6.56 (11.4) — 
 Others 830 (6.5) — 4.5 (2.7) — 
 Missing 20 (0.2) — 4.99 (6.1) — 
Hospital region  .859  <.001 
 Northeast 2600 (20.4) — 7.53 (12.2) — 
 Midwest 3057 (24.0) — 4.72 (7.7) — 
 South 4271 (33.6) — 6.09 (13.2) — 
 West 2787 (21.9) — 7.1 (11.0) — 
Hospital location/ or eaching status  .299  <.001 
 Rural 591 (4.6) — 4.62 (8.3) — 
 Urban nonteaching 1253 (9.9) — 6.65 (7.7) — 
 Urban teaching 10 871 (85.5) — 6.33 (11.9) — 
Underlying mental health diagnoses     
 Mood disorder 4749 <.001 65.32% <.001 
  Depressive episode or recurrent depressive disorder 3254 <.001 44.7% <.001 
  Bipolar affective disorder 655 <.001 52.0% <.001 
  Unspecified mood disorder 840 <.001 61.3% <.001 
 ADHD 1115 (20.4) .009 7.48 (16.8) .004 
 Anxiety disorder 1255 (22.9) .491 6.01 (10.1) .010 
n (%)χ2PMean LOS (SD)ANOVA P
Race and ethnicity  .055  <.001 
 Non-Hispanic white 7271 (57.2) — 6.35 (12.1) — 
 Non-Hispanic Black 1601 (12.6) — 6.13 (9.4) — 
 Hispanic 1630 (12.8) — 7.37 (13.7) — 
 Other 980 (7.7) — 6.43 (8.6) — 
 Missing 1233 (9.7) — 4.48 (7.7) — 
Age, y  .008  <.001 
 15–17 7105 (55.9) — 5.88 (9.3) — 
 12–14 4290 (33.7) — 6.67 (12.3) — 
 5–11 1320 (10.4) — 7.14 (17.3) — 
Zip income quartile  .484  .002 
 Highest quartile 2833 (22.3) — 6.4 (10.3) — 
 Third quartile 3298 (25.9) — 5.77 (9.8) — 
 Second quartile 3199 (25.2) — 6.11 (12.4) — 
 Lowest quartile 3230 (25.4) — 6.89 (12.9) — 
 Missing 155 (1.2) — 5.65 (7.5) — 
Primary payer  .031  <.001 
 Private insurance 5265 (41.4) — 5.61 (9.1) — 
 Public insurance 6746 (53.1) — 6.25 (8.2) — 
 Self-pay 669 (5.3) — 5.18 (9.4) — 
 Other or missing 35 (0.3) — 10.86 (15.7) — 
Sex  .092  <.001 
 Male 4128 (32.5) — 6.98 (14.3) — 
 Female 8587 (67.6) — 5.94 (9.7) — 
Place of residence  .492  <.001 
 Metro areas 10 836 (85.2) — 6.35 (11.8) — 
 Micropolitan areas 1029 (8.1) — 6.56 (11.4) — 
 Others 830 (6.5) — 4.5 (2.7) — 
 Missing 20 (0.2) — 4.99 (6.1) — 
Hospital region  .859  <.001 
 Northeast 2600 (20.4) — 7.53 (12.2) — 
 Midwest 3057 (24.0) — 4.72 (7.7) — 
 South 4271 (33.6) — 6.09 (13.2) — 
 West 2787 (21.9) — 7.1 (11.0) — 
Hospital location/ or eaching status  .299  <.001 
 Rural 591 (4.6) — 4.62 (8.3) — 
 Urban nonteaching 1253 (9.9) — 6.65 (7.7) — 
 Urban teaching 10 871 (85.5) — 6.33 (11.9) — 
Underlying mental health diagnoses     
 Mood disorder 4749 <.001 65.32% <.001 
  Depressive episode or recurrent depressive disorder 3254 <.001 44.7% <.001 
  Bipolar affective disorder 655 <.001 52.0% <.001 
  Unspecified mood disorder 840 <.001 61.3% <.001 
 ADHD 1115 (20.4) .009 7.48 (16.8) .004 
 Anxiety disorder 1255 (22.9) .491 6.01 (10.1) .010 

ANOVA, analysis of variance;—, not applicable.

Frequency comparisons for pLOS among patients with SI or SA requiring transfer to a psychiatric facility by sociodemographic and hospital characteristics are shown in Table 2. Of 12 715 patients requiring transfer to a psychiatric facility, 5475 (43.1%) had a pLOS. The prevalence of pLOS differed significantly by race or ethnicity, with the highest prevalence observed among patients denoted as “other” (50.5%), followed by non-Hispanic white patients (44.7%). Patients in the lowest zip code income quartile and those with public insurance experienced the highest level of pLOS, whereas patients on self-pay coverage had the lowest frequency of pLOS. Male patients had a higher prevalence of pLOS than female patients. Of US geographic regions, residents of the southern part of the United States had the highest frequency of pLOS. Among types of hospital facilities, those located in the rural areas recorded the lowest prevalence of pLOS. The mental health diagnosis with the highest prevalence for pLOS was mood disorder at 65.3%.

TABLE 2

Prevalence of pLOS by Patient Characteristics and Underlying Mental Health Diagnoses Among Hospitalized Pediatric Patients With a Diagnosis of SI or SA Who Required Transfer to a Psychiatric Facility

pLOS nPrevalence % by SubgroupP
Race and ethnicity   <.001 
 Non-Hispanic white 3250 44.7 — 
 Non-Hispanic Black 695 43.4 — 
 Hispanic 670 41.1 — 
 Other 495 50.5 — 
 Missing 365 29.6 — 
Age, y   .950 
 15–17 3055 43.0 — 
 12–14 1855 43.2 — 
 5–11 565 42.8 — 
Zip income quartile   <.001 
 Highest quartile 1190 42.0 — 
 Third quartile 1230 37.3 — 
 Second quartile 1385 43.3 — 
 Lowest quartile 1595 49.4 — 
 Missing 75 48.4 — 
Primary payer   <.001 
 Private insurance 2050 38.9 — 
 Public insurance 3210 47.7 — 
 Self-pay 200 29.9 — 
 Other or missing 15 42.9 — 
Sex   .003 
 Male 1850 44.8 — 
 Female 3625 42.2 — 
Place of residence   .4 
 Metro areas 4680 43.2 — 
 Micropolitan areas 420 40.8 — 
 Others 365 44.0 — 
 Missing 10 50.0 — 
Hospital region   <.001 
 Northeast 1100 42.3 — 
 Midwest 1125 36.8 — 
 South 2060 48.2 — 
 West 1190 42.7 — 
Hospital location or teaching status   <.001 
 Rural 150 25.4 — 
 Urban nonteaching 670 53.4 — 
 Urban teaching 4655 42.8 — 
Underlying mental health diagnoses   — 
 Mood disorder 4749 65.32 <.001 
  Depressive episode or recurrent depressive disorder 3254 44.7 <.001 
  Bipolar affective disorder 655 52.0 <.001 
  Unspecified mood disorder 840 61.3 <.001 
 ADHD 1115 44.5 .102 
 Anxiety disorder 1255 40.0 <.001 
pLOS nPrevalence % by SubgroupP
Race and ethnicity   <.001 
 Non-Hispanic white 3250 44.7 — 
 Non-Hispanic Black 695 43.4 — 
 Hispanic 670 41.1 — 
 Other 495 50.5 — 
 Missing 365 29.6 — 
Age, y   .950 
 15–17 3055 43.0 — 
 12–14 1855 43.2 — 
 5–11 565 42.8 — 
Zip income quartile   <.001 
 Highest quartile 1190 42.0 — 
 Third quartile 1230 37.3 — 
 Second quartile 1385 43.3 — 
 Lowest quartile 1595 49.4 — 
 Missing 75 48.4 — 
Primary payer   <.001 
 Private insurance 2050 38.9 — 
 Public insurance 3210 47.7 — 
 Self-pay 200 29.9 — 
 Other or missing 15 42.9 — 
Sex   .003 
 Male 1850 44.8 — 
 Female 3625 42.2 — 
Place of residence   .4 
 Metro areas 4680 43.2 — 
 Micropolitan areas 420 40.8 — 
 Others 365 44.0 — 
 Missing 10 50.0 — 
Hospital region   <.001 
 Northeast 1100 42.3 — 
 Midwest 1125 36.8 — 
 South 2060 48.2 — 
 West 1190 42.7 — 
Hospital location or teaching status   <.001 
 Rural 150 25.4 — 
 Urban nonteaching 670 53.4 — 
 Urban teaching 4655 42.8 — 
Underlying mental health diagnoses   — 
 Mood disorder 4749 65.32 <.001 
  Depressive episode or recurrent depressive disorder 3254 44.7 <.001 
  Bipolar affective disorder 655 52.0 <.001 
  Unspecified mood disorder 840 61.3 <.001 
 ADHD 1115 44.5 .102 
 Anxiety disorder 1255 40.0 <.001 

pLOS was defined as an LOS greater than our population’s median LOS of 3 d; 5475 out of 12 715 hospitalized pediatric patients requiring transfer to a psychiatric facility met criteria for pLOS. —, not applicable.

Selected sociodemographic characteristics that were assessed as potential predictors of pLOS in hospitalized pediatric patients with SI or SA requiring transfer to a psychiatric facility are presented in Table 3 with adjusted estimates. Compared with non-Hispanic white patients, Hispanic patients had ∼30% lower prevalence of pLOS. Adjusted prevalence of pLOS was not significantly different among non-Hispanic Black individuals compared with non-Hispanic white individuals. Patients with public insurance had 40% increased adjusted prevalence of pLOS compared with those covered by a private payer. Prevalence of pLOS in patients hospitalized in urban nonteaching and teaching hospitals was 3 to 4 times the prevalence of those admitted to rural health facilities. Patient age, zip code income, sex, place of residence, and region of hospitalization were not predictive of pLOS.

TABLE 3

Association Between Various Patient Characteristics and pLOS Among Hospitalized Pediatric Patients With a Diagnosis of SI or SA Who Required Transfer to a Psychiatric Facility

Prevalence Ratio (95% CI)P
Race and ethnicity   
 Non-Hispanic white Reference — 
 Non-Hispanic Black 0.83 (0.59–1.16) .27 
 Hispanic 0.69 (0.52–0.93) .02 
 Other 1.17 (0.85–1.6) .34 
Age, y   
 15–17 Reference — 
 12–14 1.03 (0.86–1.23) .77 
 5–11 0.95 (0.67–1.34) .77 
Zip income quartile   
 Highest quartile Reference — 
 Third quartile 0.84 (0.64–1.11) .23 
 Second quartile 0.96 (0.73–1.27) .8 
 Lowest quartile 1.23 (0.92–1.65) .17 
Primary payer   
 Private insurance Reference — 
 Public insurance 1.40 (1.12–1.78) <.001 
 Self-pay 0.65 (0.42–1.01) .06 
Sex   
 Male Reference — 
 Female 0.92 (0.75–1.12) .41 
Place of residence   
 Metro areas Reference — 
 Micropolitan areas 0.82 (0.55–1.21) .31 
 Others 0.86 (0.56–1.34) .51 
Hospital region   
 Northeast Reference — 
 Midwest 0.83 (0.51–1.33) .43 
 South 1.30 (0.73–2.32) .37 
 West 1.38 (0.84–2.26) .21 
Hospital location or teaching status   
 Rural Reference — 
 Urban nonteaching 4.61 (2.33–9.12) <.001 
 Urban teaching 3.26 (1.84–5.76) <.001 
Underlying mental health diagnosesa   
 Mood disorder 1.98 (1.63–3.42) <.001 
  Depressive episode or recurrent depressive disorder 1.48 (1.16–3.15) <.001 
  Bipolar affective disorder 1.41 (1.05–1.90) .02 
  Unspecified mood disorder 2.30 (1.66–3.19) <.001 
 ADHD 1.07 (0.87–1.33) .52 
 Anxiety disorder 0.91 (0.70–1.19) .5 
Prevalence Ratio (95% CI)P
Race and ethnicity   
 Non-Hispanic white Reference — 
 Non-Hispanic Black 0.83 (0.59–1.16) .27 
 Hispanic 0.69 (0.52–0.93) .02 
 Other 1.17 (0.85–1.6) .34 
Age, y   
 15–17 Reference — 
 12–14 1.03 (0.86–1.23) .77 
 5–11 0.95 (0.67–1.34) .77 
Zip income quartile   
 Highest quartile Reference — 
 Third quartile 0.84 (0.64–1.11) .23 
 Second quartile 0.96 (0.73–1.27) .8 
 Lowest quartile 1.23 (0.92–1.65) .17 
Primary payer   
 Private insurance Reference — 
 Public insurance 1.40 (1.12–1.78) <.001 
 Self-pay 0.65 (0.42–1.01) .06 
Sex   
 Male Reference — 
 Female 0.92 (0.75–1.12) .41 
Place of residence   
 Metro areas Reference — 
 Micropolitan areas 0.82 (0.55–1.21) .31 
 Others 0.86 (0.56–1.34) .51 
Hospital region   
 Northeast Reference — 
 Midwest 0.83 (0.51–1.33) .43 
 South 1.30 (0.73–2.32) .37 
 West 1.38 (0.84–2.26) .21 
Hospital location or teaching status   
 Rural Reference — 
 Urban nonteaching 4.61 (2.33–9.12) <.001 
 Urban teaching 3.26 (1.84–5.76) <.001 
Underlying mental health diagnosesa   
 Mood disorder 1.98 (1.63–3.42) <.001 
  Depressive episode or recurrent depressive disorder 1.48 (1.16–3.15) <.001 
  Bipolar affective disorder 1.41 (1.05–1.90) .02 
  Unspecified mood disorder 2.30 (1.66–3.19) <.001 
 ADHD 1.07 (0.87–1.33) .52 
 Anxiety disorder 0.91 (0.70–1.19) .5 

pLOS was defined as an LOS greater than our population’s median LOS of 3 d; 5475 out of 12 715 hospitalized pediatric patients requiring transfer to a psychiatric facility met criteria for pLOS. —, not applicable.

a

Individual multivariable association models for each mental health diagnosis and pLOS were created, while adjusting for patient and hospital characteristics.

In Table 3, we also detail the adjusted prevalence ratios of pLOS among patients with the most-common underlying mental health diagnoses in our target population, using stepwise regression modeling. The strongest predictor was mood disorder. Patients with the diagnosis of SI or SA had double the prevalence of pLOS if they also had mood disorder compared with similar patients without mood disorder. Anxiety disorder and ADHD were not predictive of pLOS in patients with a diagnosis of SI or SA.

In this study, we are the first to identify predictors for pLOS in pediatric patients hospitalized to a medical service with SI or SA using nationally representative data. The majority of existing data analyzing predictive factors for LOS exist primarily in the ED setting with smaller sample sizes. We identified patient demographics, hospital characteristics, and underlying mental health diagnoses that were associated with pLOS, or an inpatient LOS greater than our population median of 3 days.

In our study, we found that public insurance was associated with increased probability of inpatient pLOS. This is consistent with previous data revealing longer inpatient LOS for pediatric mental health visits covered by public insurance, such as Medicaid, compared with private insurance.5  One explanation for this finding is that public insurance approval processes may be more complex than those of private insurance because of the size of the population covered and financial burden. For example, Medicaid is the largest payer for mental health services in the United States and covers approximately half of inpatient mental health hospitalization costs.5,30  It is also possible that pediatric inpatient psychiatric facilities may be more likely to accept private insurance than public insurance. Although researchers of 1 study found that >90% of pediatric inpatient psychiatric facilities accept Medicaid, this may not apply to all types of Medicaid or out-of-state Medicaid.31  Another hypothesis is that hospitalized children with public insurance may have a higher severity of mental illness in comparison with hospitalized children with private insurance. Children who qualify for public insurance have a higher level of poverty and may have less access to resources and social support that may mitigate psychiatric illness severity. Additionally, a large proportion of children with Medicaid presenting to EDs for mental health visits are discharged to the community, so perhaps the proportion that are admitted have relatively more-complex psychiatric illness.15,32  Of note, our data set does not include what portion of the hospital admission was spent receiving medical treatment and what portion was spent awaiting bed placement at a psychiatric facility. Given that psychiatric boarding admissions are costly to both the patient and the hospital with varying quality, future researchers should examine whether publicly insured children are at higher risk for boarding on the medical service compared with those who are privately insured.8,11,3335 

In our study, we also found a lower probability of pLOS in Hispanic compared with white patient, whereas non-Hispanic Black patients had a similar probability of pLOS compared with white patients. Given known disparities in mental health prevention, diagnosis, access to care, and quality of care among Hispanic and Black populations, these results were surprising to us.32,3644  In the ED setting, researchers in 1 study found a longer LOS for mental health encounters in both Hispanic and Black populations in comparison with white patients.19  Researchers in another study did not find a difference across race.15  Our findings could possibly be explained by differences in psychiatric complexity across ethnicity among admitted children. In 1 study, Hispanic ethnicity was associated with a lower prevalence of depression, anxiety, and behavioral problems.45  Researchers in another study found protective factors against suicidal behavior among the Latino population attributed to cultural norms.46  Thus, perhaps reduced probability of pLOS among Hispanic patients could be secondary to relatively lower illness severity due to social protective factors.

In our study, we found that rural hospital designation was associated with lower probability of pLOS compared with urban hospital location. Similar findings have been described in the literature.19  However, the number of children discharged to psychiatric facilities from rural hospitals was relatively small. Thus, our findings could be explained by rural transfer of complex psychiatric cases to urban locations.

Patient sex, age, zip income, place of residence, hospital region, or hospital teaching status were not predictive of pLOS in adjusted modeling. In contrast, studies in the ED setting revealed female sex to be associated with longer LOS in children presenting for psychiatric illness.18,19  Literature on the association between age and LOS for psychiatric encounters in the ED setting is mixed, with some researchers suggesting younger age, some researchers suggesting older age, and others suggesting that there is no association between age and LOS.1719,47  Also in contrast to our findings, researchers in 1 study in the ED setting found metropolitan place of residence and northeastern or southern hospital location to be associated with extended LOS.47  Our findings are consistent with ED literature describing hospital teaching status as not being associated with psychiatric encounter LOS.19  Of note, it is challenging to compare inpatient demographics and hospital characteristics predictive of LOS to those of the ED setting, because EDs have different thresholds for admission of medically cleared psychiatric patients to the hospital.7,47,48 

Finally, we found that patient diagnoses of mood disorders, such as depression and bipolar disorder, to be predictive of pLOS in our target population. This is consistent with literature in the ED setting describing longer LOS among children with these psychiatric conditions.15,19  Given that mood disorder is associated with a higher risk of suicide, our findings could possibly be explained by higher medical illness severity among patients with mood disorder.49 

This study has limitations for consideration. This is a cross-sectional study looking only at admitted children who were later transferred to psychiatric facilities. Thus, this excludes children with SI or SA who were discharged from the ED, transferred to a psychiatric facility from the ED, or never presented to the ED with mental illness. Additionally, the NIS does not provide granular information such as the severity of illness, the proportion of the hospitalization spent on medical clearance, patient readmission rate, or exact geographic location or hospital proximity to psychiatric facilities. Therefore, it is not possible to discern what percentage of the patient’s hospital stay was spent on medical treatment and what percentage was spent “boarding” or awaiting psychiatric placement after medical clearance. We also estimated that the majority of patients admitted with primary or secondary diagnoses of SI or SA who were subsequently transferred to a facility were most likely transferred specifically to a psychiatric facility. However, it is possible that some of our patients were transferred to other types of facilities, such as skilled nursing facilities, or that SI or SA was not the primary reason for their hospitalization. Finally, we may not have captured all injuries secondary to intentional self-harm using ICD-10-CM codes. However, limiting our analysis to ICD-10-CM coding offers many benefits, because InternationalClassification of Diseases, Ninth Revision, Clinical Modification codes do not differentiate between intentional self-inflicted injury with intent to die (SA) and intentional self-inflicted injury without intent to die.27 

In our study, we identified risk factors for pLOS among children hospitalized for SI or SA who require transfer to a psychiatric facility. Public insurance, urban hospital location, and underlying mood disorders were predictive of pLOS, whereas Hispanic ethnicity was associated with decreased prevalence of pLOS. Given the recent rise in hospitalizations for pediatric SI or SA, further research is needed on the reasoning behind these disparities. Improved screening methods for SI in the outpatient setting, especially in publicly insured patients with underlying mood disorders, may lead to early identification and treatment and reduction in illness severity for SI and SA admissions. We also recommend focusing future quality improvement and policy initiatives on streamlining placement of publicly insured patients with SI or SA to psychiatric facilities.

The authors thank Michelle Lopez, MD, MPH for reviewing our work and providing insights into our study design.

Ms Dongarwar conducted the data analysis; and all authors conceptualized and designed the study, drafted the initial manuscript, and approved the final manuscript as submitted.

FUNDING: Supported by grant 1 D34HP31024-01-00 from the Health Resources and Services Administration for the project titled Baylor College of Medicine (BCM) Center of Excellence in Health Equity, Training & Research. The funder or sponsor did not participate in the work

1
Curtin
SC
,
Heron
M
.
Death rates due to suicide and homicide among persons aged 10-24: United States, 2000-2017
.
NCHS Data Brief
.
2019
;(
352
):
1
8
2
Heron
M
.
Deaths: leading causes for 2017
.
Natl Vital Stat Rep
.
2019
;
68
(
6
):
1
77
3
Plemmons
G
,
Hall
M
,
Doupnik
S
, et al
.
Hospitalization for suicide ideation or attempt: 2008-2015
.
Pediatrics
.
2018
;
141
(
6
):
e20172426
4
Findling
RL
,
Stepanova
E
.
The workforce shortage of child and adolescent psychiatrists: is it time for a different approach?
J Am Acad Child Adolesc Psychiatry
.
2018
;
57
(
5
):
300
301
5
Torio
CM
,
Encinosa
W
,
Berdahl
T
,
McCormick
MC
,
Simpson
LA
.
Annual report on health care for children and youth in the United States: national estimates of cost, utilization and expenditures for children with mental health conditions
.
Acad Pediatr
.
2015
;
15
(
1
):
19
35
6
Lo
CB
,
Bridge
JA
,
Shi
J
,
Ludwig
L
,
Stanley
RM
.
Children’s mental health emergency department visits: 2007-2016
.
Pediatrics
.
2020
;
145
(
6
):
e20191536
7
Mansbach
JM
,
Wharff
E
,
Austin
SB
,
Ginnis
K
,
Woods
ER
.
Which psychiatric patients board on the medical service?
Pediatrics
.
2003
;
111
(
6 pt 1
).
8
Claudius
I
,
Donofrio
JJ
,
Lam
CN
,
Santillanes
G
.
Impact of boarding pediatric psychiatric patients on a medical ward
.
Hosp Pediatr
.
2014
;
4
(
3
):
125
132
9
Geller
JL
,
Biebel
K
.
The premature demise of public child and adolescent inpatient psychiatric beds: part I: overview and current conditions
.
Psychiatr Q
.
2006
;
77
(
3
):
251
271
10
Geller
JL
,
Biebel
K
.
The premature demise of public child and adolescent inpatient psychiatric beds: part II: challenges and implications
.
Psychiatr Q
.
2006
;
77
(
4
):
273
291
11
Worsley
D
,
Barrios
E
,
Shuter
M
,
Pettit
AR
,
Doupnik
SK
.
Adolescents' experiences during “boarding” hospitalization while awaiting inpatient psychiatric treatment following suicidal ideation or suicide attempt
.
Hosp Pediatr
.
2019
;
9
(
11
):
827
833
12
Herndon
AC
,
Williams
D
,
Hall
M
, et al
.
Costs and reimbursements for mental health hospitalizations at children’s hospitals
.
J Hosp Med
.
2020
;
15
(
12
):
727
730
13
Bardach
NS
,
Coker
TR
,
Zima
BT
, et al
.
Common and costly hospitalizations for pediatric mental health disorders
.
Pediatrics
.
2014
;
133
(
4
):
602
609
14
Egorova
NN
,
Pincus
HA
,
Shemesh
E
,
Kleinman
LC
.
Behavioral health diagnoses among children and adolescents hospitalized in the United States: observations and implications
.
Psychiatr Serv
.
2018
;
69
(
8
):
910
918
15
Hoffmann
JA
,
Stack
AM
,
Monuteaux
MC
,
Levin
R
,
Lee
LK
.
Factors associated with boarding and length of stay for pediatric mental health emergency visits
.
Am J Emerg Med
.
2019
;
37
(
10
):
1829
1835
16
Hughes
JL
,
Anderson
NL
,
Wiblin
JL
,
Asarnow
JR
.
Predictors and outcomes of psychiatric hospitalization in youth presenting to the emergency department with suicidality
.
Suicide Life Threat Behav
.
2017
;
47
(
2
):
193
204
17
Wharff
EA
,
Ginnis
KB
,
Ross
AM
,
Blood
EA
.
Predictors of psychiatric boarding in the pediatric emergency department: implications for emergency care
.
Pediatr Emerg Care
.
2011
;
27
(
6
):
483
489
18
Chakravarthy
B
,
Yang
A
,
Ogbu
U
, et al
.
Determinants of pediatric psychiatry length of stay in 2 urban emergency departments
.
Pediatr Emerg Care
.
2017
;
33
(
9
):
613
619
19
Smith
JL
,
De Nadai
AS
,
Petrila
J
,
Storch
EA
.
Factors associated with length of stay in emergency departments for pediatric patients with psychiatric problems
.
Pediatr Emerg Care
.
2019
;
35
(
10
):
716
721
20
Mahajan
P
,
Alpern
ER
,
Grupp-Phelan
J
, et al
;
Pediatric Emergency Care Applied Research Network (PECARN)
.
Epidemiology of psychiatric-related visits to emergency departments in a multicenter collaborative research pediatric network
.
Pediatr Emerg Care
.
2009
;
25
(
11
):
715
720
21
American Healthcare Association
.
Length of stay calculation
, LTC trend tracker.
2014
. Available at: https://www.ahcancal.org/Data-and-Research/LTC-Trend-Tracker/Documents/Length%20of%20Stay%20Calculation.pdf#search=median. Accessed February 17, 2021
22
McKay
S
,
Parente
V
.
Health disparities in the hospitalized child
.
Hosp Pediatr
.
2019
;
9
(
5
):
317
325
23
Lopez
MA
,
Cruz
AT
,
Kowalkowski
MA
,
Raphael
JL
.
Factors associated with high resource utilization in pediatric skin and soft tissue infection hospitalizations
.
Hosp Pediatr
.
2013
;
3
(
4
)
24
Ravi
P
,
Sood
A
,
Schmid
M
, et al
.
Racial/ethnic disparities in perioperative outcomes of major procedures: results from the national surgical quality improvement program
.
Ann Surg
.
2015
;
262
(
6
):
955
964
25
Pati
S
,
Lorch
SA
,
Lee
GE
,
Sheffler-Collins
S
,
Shah
SS
.
Health insurance and length of stay for children hospitalized with community-acquired pneumonia
.
J Hosp Med
.
2012
;
7
(
4
):
304
310
26
Healthcare Cost and Utilization Project (HCUP)
.
HCUP Databases. Agency for Healthcare Research and Quality, Rockville, MD. 2020. Available at: www.hcup-us.ahrq.gov/nisoverview.jsp. Accessed February 17, 2021
27
Hedegaard
H
,
Schoenbaum
M
,
Claassen
C
,
Crosby
A
,
Holland
K
,
Proescholdbell
S
.
Issues in developing a surveillance case definition for nonfatal suicide attempt and intentional self-harm using International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) coded data
.
Natl Health Stat Report
.
2018
;(
108
):
1
19
28
Shaw
JS
,
Duncan
PM
,
Hagan
JF
Jr, eds.
Bright Futures Guidelines for Health Supervision of Infants, Children, and Adolescents
, 4th ed.
Elk Groove Village, IL
:
American Academy of Pediatrics
;
2017
:
889
29
Kessler
RC
,
Amminger
GP
,
Aguilar-Gaxiola
S
,
Alonso
J
,
Lee
S
,
Ustün
TB
.
Age of onset of mental disorders: a review of recent literature
.
Curr Opin Psychiatry
.
2007
;
20
(
4
):
359
364
30
Centers for Medicare and Medicaid Services
.
Behavioral health services
. Available at: https://www.medicaid.gov/medicaid/benefits/behavioral-health-services/index.html. Accessed February 17, 2021
31
Cummings
JR
,
Case
BG
,
Ji
X
,
Marcus
SC
.
Availability of youth services in U.S. mental health treatment facilities
.
Adm Policy Ment Health
.
2016
;
43
(
5
):
717
727
32
Bridge
JA
,
Marcus
SC
,
Olfson
M
.
Outpatient care of young people after emergency treatment of deliberate self-harm
.
J Am Acad Child Adolesc Psychiatry
.
2012
;
51
(
2
):
213
222.e1
33
Fieldston
E
,
Jonas
J
,
Scharko
AM
.
Boarding of pediatric psychiatric patients is a no-fly zone for value
.
Hosp Pediatr
.
2014
;
4
(
3
):
133
134
34
Bardach
NS
,
Burkhart
Q
,
Richardson
LP
, et al
.
Hospital-based quality measures for pediatric mental health care
.
Pediatrics
.
2018
;
141
(
6
):
e20173554
35
Gallagher
KAS
,
Bujoreanu
IS
,
Cheung
P
, et al
.
Psychiatric boarding in the pediatric inpatient medical setting: a retrospective analysis
.
Hosp Pediatr
.
2017
;
7
(
8
):
444
450
36
Bardach
NS
,
Doupnik
SK
,
Rodean
J
, et al
.
ED visits and readmissions after follow-up for mental health hospitalization
.
Pediatrics
.
2020
;
145
(
6
):
e20192872
37
Merikangas
KR
,
He
JP
,
Burstein
M
, et al
.
Service utilization for lifetime mental disorders in U.S. adolescents: results of the National Comorbidity Survey-Adolescent Supplement (NCS-A)
.
J Am Acad Child Adolesc Psychiatry
.
2011
;
50
(
1
):
32
45
38
Costello
EJ
,
He
JP
,
Sampson
NA
,
Kessler
RC
,
Merikangas
KR
.
Services for adolescents with psychiatric disorders: 12-month data from the National Comorbidity Survey-Adolescent
.
Psychiatr Serv
.
2014
;
65
(
3
):
359
366
39
Alegria
M
,
Vallas
M
,
Pumariega
AJ
.
Racial and ethnic disparities in pediatric mental health
.
Child Adolesc Psychiatr Clin N Am
.
2010
;
19
(
4
):
759
774
40
Hacker
KA
,
Penfold
R
,
Arsenault
L
,
Zhang
F
,
Murphy
M
,
Wissow
L
.
Screening for behavioral health issues in children enrolled in Massachusetts Medicaid
.
Pediatrics
.
2014
;
133
(
1
):
46
54
41
Howell
E
,
McFeeters
J
.
Children’s mental health care: differences by race/ethnicity in urban/rural areas
.
J Health Care Poor Underserved
.
2008
;
19
(
1
):
237
247
42
Kataoka
SH
,
Zhang
L
,
Wells
KB
.
Unmet need for mental health care among U.S. children: variation by ethnicity and insurance status
.
Am J Psychiatry
.
2002
;
159
(
9
):
1548
1555
43
Lê Cook
B
,
Barry
CL
,
Busch
SH
.
Racial/ethnic disparity trends in children’s mental health care access and expenditures from 2002 to 2007
.
Health Serv Res
.
2013
;
48
(
1
):
129
149
44
Chow
JC
,
Jaffee
K
,
Snowden
L
.
Racial/ethnic disparities in the use of mental health services in poverty areas
.
Am J Public Health
.
2003
;
93
(
5
):
792
797
45
Ghandour
RM
,
Sherman
LJ
,
Vladutiu
CJ
, et al
.
Prevalence and treatment of depression, anxiety, and conduct problems in US children
.
J Pediatr
.
2019
;
206
:
256
267.e3
46
Oquendo
MA
,
Dragatsi
D
,
Harkavy-Friedman
J
, et al
.
Protective factors against suicidal behavior in Latinos
.
J Nerv Ment Dis
.
2005
;
193
(
7
):
438
443
47
Case
SD
,
Case
BG
,
Olfson
M
,
Linakis
JG
,
Laska
EM
.
Length of stay of pediatric mental health emergency department visits in the United States
.
J Am Acad Child Adolesc Psychiatry
.
2011
;
50
(
11
):
1110
1119
48
Grupp-Phelan
J
,
Mahajan
P
,
Foltin
GL
, et al
;
Pediatric Emergency Care Applied Research Network
.
Referral and resource use patterns for psychiatric-related visits to pediatric emergency departments
.
Pediatr Emerg Care
.
2009
;
25
(
4
):
217
220
49
Shaffer
D
,
Gould
MS
,
Fisher
P
, et al
.
Psychiatric diagnosis in child and adolescent suicide
.
Arch Gen Psychiatry
.
1996
;
53
(
4
):
339
348

Competing Interests

POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.

Supplementary data