Quality of Care for Youth Hospitalized for Suicidal Ideation and Self-Harm

Abstract

Objectives:

To examine performance on quality measures for pediatric inpatient suicidal ideation/self-harm care, and whether performance is associated with reutilization.

Methods:

Retrospective observational eight hospital study of patients [N=1090] aged 5-17 years hospitalized for suicidal ideation/self-harm between 9/1/14-8/31/16. Two medical records-based quality measures assessing suicidal ideation/self-harm care were evaluated, one on counseling caregivers regarding restricting access to lethal means and the other on communication between inpatient and outpatient providers regarding the follow-up plan. Multivariable logistic regression assessed associations between quality measure scores and 1) hospital site, 2) patient demographics, and 3) 30-day emergency department return visits and inpatient readmissions.

Results:

Medical record documentation revealed that, depending on hospital site, 17% - 98% of caregivers received lethal means restriction counseling (mean 70%); inpatient-to-outpatient provider communication was documented in 0-51% of cases (mean 16%). The odds of documenting receipt of lethal means restriction counseling was higher for caregivers of female patients compared to caregivers of male patients (aOR 1.51, 95% confidence interval [CI]: 1.07-2.14). The odds of documenting inpatient-to-outpatient provider follow-up plan communication was lower for Black patients compared to White patients (aOR 0.45, 95% CI: 0.24-0.84). All-cause 30-day readmission was lower for patients with documented caregiver receipt of lethal means restriction counseling (aOR 0.48, 95% CI: 0.28-0.83).

Conclusions:

This study revealed disparities and deficits in the quality of care received by youth with suicidal ideation/self-harm. Providing caregivers lethal means restriction counseling prior to discharge may help to prevent readmission.

INTRODUCTION

Suicide is the 10^th leading cause of death in the United States (US)¹ and since 2011, the second most common cause of death in the adolescent population.² For 10-14 year olds, the suicide rate almost tripled between 2007-2017; and for 15-19 year olds, the rate increased 76%.³ From 2010 to 2014, there were 874,872 youth visits to the emergency department (ED) due to suicidal ideation (SI)/suicide attempt.⁴ Between 2014 and 2016, these ED visits increased by 38% for youth aged 5 to 11 years and by 82% for youth aged 15-17.⁴ Similarly, hospitalizations have increased for youth with severe mental illness, including SI and self-harm.^5,6 An analysis of data from US children’s hospitals shows that compared to hospitalizations for SI/self-harm in 2008, in 2015 the odds for inpatient care for SI/self-harm had increased by 3.4.⁷ Compared to the historically higher male suicide rates,⁸ a recent study has found that the increasing rates of youth suicide since 2007 have been associated with a narrowing of the gap between males and females, although male suicide rates have continued to increase and remain high.⁹

The growing need for more accessible mental health care services in both the outpatient and inpatient settings is well established.^5,7,10,11 However, even when accessible, very little is known about the quality of the services received, particularly in the inpatient setting.^12,13 Federal health policies have identified pediatric mental health care as a target area for quality improvement, making optimizing care for youth with mental illness a national goal.^14,15 In response to this national prioritization, one of the Pediatric Quality Measures Program Centers of Excellence¹⁶ developed and field tested two medical-records based quality measures related to inpatient care for youth with SI/self-harm; one examining whether caregivers were counseled on lethal means restriction prior to their child or adolescent being discharged to home and the other focused on inpatient-to-outpatient provider communication regarding the follow-up plan.¹⁷ This measure field test demonstrated significant performance gaps, variation in performance across hospitals, and disparities in the care provided.¹⁷ While these findings further support the need for quality improvement (QI) focused on inpatient pediatric mental health care, the study was conducted in only three free-standing tertiary care children’s hospitals in major metropolitan areas and associations between measure performance and outcomes were not assessed. Establishing the predictive validity of these measures before they are used for accountability purposes is important.¹⁸

When examining if the quality of mental health care services provided to patients admitted for SI/self-harm is associated with outcomes of care, ideally the incidence of repeated episodes of SI/self-harm and suicide deaths would be the key outcomes assessed. However, the low rates of occurrence for these events requires examining a broader set of outcomes, including reutilization for other mental health diagnoses.¹⁹ This type of reutilization is increasingly understood to be a key indicator of poor quality mental health care.²⁰

In this study, we examined these two previously developed quality measures assessing SI/self-harm care in a larger, geographically dispersed sample of eight hospitals serving more diverse patient populations (greater representation of Hispanic and Asian/Pacific Islander patients). Our objectives were to examine whether measure performance varied by site of care and/or youth demographic characteristics and whether better measure performance was associated with decreased same-cause and all-cause 30-day ED return visits and/or inpatient readmissions. We hypothesized that there would be significant variation by patient demographics and by site of care and that better measure performance would be associated with decreased same-cause and all-cause reutilization.

METHODS

Sample

We examined the quality of care for youth (patients aged 5 to 17 years old) admitted for an inpatient stay with a diagnosis of SI/self-harm between 9/1/2014 and 8/31/2016. The Pediatric Research in Inpatient Settings (PRIS) network recruited the hospitals for this study by sending out an email to those participating in their network. Hospitals become members of PRIS on a volunteer basis, although must have the capacity for grants administration and institutional review board approval. The network is comprised of hospitals interested in performing collaborative pediatric research, often focused on quality improvement.²¹ Of 12 interested hospitals, eight were selected in an effort to increase the geographic and racial/ethnic diversity of the sample. Hospitals were in urban areas in the Northeastern (n=1), Southeastern (n=2), Midwestern (n=2), and Western (n=3) US. There were five freestanding children’s hospitals and three pediatric units within general hospitals; all were affiliated with academic centers. Seven out of the eight hospitals had inpatient psychiatric units and/or treatment programs. Eligible patients were identified using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes and International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes noted at any time during their hospital stay (eAppendix). Patients admitted to the intensive care unit were included if they were transferred to a regular unit (medical or psychiatric) before discharge to home. Patients who were discharged to another inpatient facility were excluded. The participating hospitals’ institutional review boards approved all study procedures.

Quality Measures

We used quality measures previously developed by the Center of Excellence on Quality of Care Measures for Children with Complex Needs, one of the seven Centers of Excellence belonging to the national Pediatric Quality Measures Program.^16,17 Both measures were based on consensus from our Delphi panel and required documented proof. The measure development process is described in detail elsewhere.¹⁷ Briefly, a review of the scientific literature and expert clinical practice guidelines related to hospital-based care for youth with SI/self-harm was performed. Four medical-records based measures that focused on SI/self-harm care were developed based on this review. The measures underwent evaluation for face validity, feasibility of implementation, and reliability.¹⁷ The two measures specifically designed to assess the quality of inpatient mental health care for SI/self-harm were used for the current study; these were: 1) documentation in the hospital record that the patient’s caregiver was counseled on how to restrict access to lethal means prior to discharge and 2) documentation in the hospital record of communication (either by telephone or email exchange) between inpatient and outpatient providers regarding the follow-up plan. Patients required admission for the study; we did not include ED visits discharged home. Outpatient provider was defined as any mental health provider or primary care physician (PCP). To pass the lethal means restriction measure, the parent/caregiver had to receive counseling about the need to protect their child by limiting his/her/their access to lethal means, such as the car, guns, sharps, medications, etc. This is often referred to as a “safety plan.” Completion of this counseling then had to be documented in the medical record. If there was no clear documentation that the counseling took place, documentation that a “safety plan” was discussed with the family was considered acceptable as long as there was also clear documentation that this discussion occurred with the parent/caregiver and not with the patient themselves. Passing the inpatient-to-outpatient provider communication measure required documentation that the follow-up plan was shared with any outpatient mental health care provider or PCP (established or newly identified by the inpatient team). A note simply saying that the PCP or outpatient mental health care provider was contacted, however, was insufficient; the note had to document a telephone call had occurred between the youth’s inpatient and outpatient providers and that the plan of care was relayed to the outpatient provider. Alternatively, a copy of an email exchange between these two providers including the relevant content had to be present in the electronic medical record (EMR). For both measures, each documented occurrence resulted in a dichotomous score of 1 (pass) or 0 (fail). Each hospital had trained medical record abstractors who entered the necessary data elements into the previously developed electronic abstraction tool. Additional information about the abstraction tool has been published elsewhere.¹⁷

Utilization Outcomes

We chose 30-day reutilization as our outcome validation measure as it is a widely accepted and utilized performance measure.²² Starting with the discharge date of the index mental health admission to the hospital, we assessed 30-day all-cause and same-cause ED return visits and inpatient readmissions using hospital administrative data. Inpatient readmissions occurred either through the ED or were admitted directly.

Covariates

We obtained patient characteristics including sex, age, race and ethnicity from hospital administrative data. Each hospital provided an indicator of Hispanic ethnicity and a separate race variable. For analytic purposes, mutually exclusive race/ethnicity categories were created such that Hispanic patients of any race were classified as Hispanic and non-Hispanic patients were classified as Black, White, Asian/Pacific Islander, or Other. One hospital’s ethnicity variable had 88% of cases coded as non-Hispanic and 12% of cases coded as unknown ethnicity. Therefore, we were unable to identify any Hispanic patients from that hospital. A second hospital’s race variable had only three categories: Black, White, and Other. Asian/Pacific Islander patients from that hospital could not be identified and were categorized as Other race/ethnicity in our analyses. We classified race/ethnicity data for the remaining six hospitals as described above. For adjusted models, missing patient demographic variables (n=49) were imputed with the hospital-level mean for the site where the patient was admitted. We did not have information about underlying mental health comorbidities so were unable to adjust for these in our analysis.

Statistical Analysis

We examined descriptive statistics for all variables including the percentage of patients that passed each quality measure.

To examine variation in performance across hospitals, we fit a logistic regression to each quality measure with hospital indicators, age, sex, and race/ethnicity as predictors. A test for the joint significance of the hospital indicators assessed overall variation. We computed adjusted mean scores for each hospital and tested whether each hospital’s mean score differed from the mean for the seven other hospitals. To generate mean quality scores at the hospital level, all individual patient level scores were summed and divided by the number of eligible patients.

We assessed associations between quality measure performance and demographic characteristics by fitting multivariable logistic regressions where the quality measure score was the outcome, and indicators for patient age, sex, and race/ethnicity were the independent variables. Models included random intercepts for site of care to account for clustering within hospitals. Logistic regression was used to estimate the association between quality measure scores and reutilization outcomes (30-day all-cause and same-cause ED return visits and inpatient readmissions). Each reutilization outcome was regressed on each quality measure separately. Models were adjusted for patient age, sex, and race/ethnicity. Random intercepts for hospitals were included to account for clustering at the hospital level.

RESULTS

A total of 1090 inpatient admissions for SI/self-harm among youth aged 5-17 years were analyzed across the eight hospitals. Most patients with SI/self-harm were 13-17 years old (81%), female (66%), and non-Hispanic White (62%) (Table 1). There were no missing data for age. Sex and race/ethnicity were missing for 1% and 4% of patients, respectively. Ten percent of patients were from the Northeastern US, 29% were from the Southeast, 29% were from the Midwest, and 32% were from the West.

Table 1.

Study Sample Demographics (n=1090)

	No. (%a)
Age
Mean (SD)	14.3 (2.3)
5-12	206 (18.9)
13-17	884 (81.1)
Sex
Female	713 (65.8)
Race/ethnicity b
Hispanic	129 (12.4)
White	648 (62.2)
Black	167 (16.0)
Asian/Pacific Islander	26 (2.5)
Other	71 (6.8)

^aPercentages are calculated excluding missing values.

^bEach hospital provided an indicator of Hispanic ethnicity and a separate race variable. We created mutually exclusive race/ethnicity categories such that Hispanic patients of any race were classified as Hispanic and non-Hispanic patients were classified as White, Black, Asian/Pacific Islander, or Other.

Variation in Performance by Site of Care

Overall, medical record documentation revealed that 70% of caregivers were counseled on how to restrict access to lethal means prior to the patient’s discharge. Adjusted hospital-level scores ranged from 17% to 98% (p<0.0001 for variation across hospitals; Table 2a). Documented communication between inpatient and outpatient providers regarding the follow-up plan occurred in 16% of cases overall. Adjusted hospital-level scores ranged from 0% to 51% (p<0.0001; Table 2b).

Table 2.

Variation in Performance on the Quality Measures by Site of Care

a. Documentation in Medical Record that Caregiver was Counseled to Restrict Access to Lethal Means
	n	Adjusted meana (95% CI)	Difference from mean of all other sites (95% CI)
Overall	1083	69.71 (66.97, 72.45)	NA
Hospital
A	130	88.55 (83.12, 93.99)	21.89 (15.84, 27.95)e
B	170	77.97 (71.63, 84.30)	10.71 (3.86, 17.55)d
C	184	16.74 (11.31, 22.17)	−63.05 (−69.23, −56.87)e
D	159	63.73 (56.13, 71.34)	−6.74 (−14.87, 1.38)
E	162	69.36 (62.19, 76.52)	0.99 (−6.53, 8.51)
F	113	96.30 (92.67, 99.92)	29.04 (24.05, 34.02)e
G	13	41.29 (14.09, 68.48)	−33.15 (−59.25, −7.04)c
H	152	97.57 (95.20, 99.93)	32.63 (28.75, 36.51)e
b. Documentation in Medical Record that Follow-up Plan was Discussed with Outpatient Provider
	n	Adjusted meana (95% CI)	Difference from mean of all other sites (95% CI)
Overall	1044	16.48 (14.22, 18.73)	NA
Hospital
A	127	18.62 (11.86, 25.38)	3.58 (−3.83, 10.99)
B	172	25.27 (18.74, 31.79)	10.41 (3.47, 17.35)d
C	178	--b	NA
D	139	30.31 (22.74, 37.88)	15.91 (8.00, 23.82)e
E	151	19.28 (12.93, 25.64)	5.54 (−1.71, 12.78)
F	112	17.99 (9.70, 26.27)	−1.71 (−9.50, 6.08)
G	13	50.66 (22.92, 78.40)	31.95 (4.11, 59.78)c
H	152	3.48 (0.48, 6.48)	−14.83 (−18.87, −10.79)e

Abbreviations: CI, confidence interval.

^aHospital means are adjusted for age, sex and race/ethnicity; missing values for patient demographic characteristics (age, sex, and race/ethnicity) are imputed with the hospital-level means.

^bHospital C had 0 observed passes on the measure; thus, logistic regression cannot be used to obtain an adjusted mean or test for the difference from the mean of all others and construct a confidence interval for the difference.

^cp<0.05

^dp<0.01

^ep<0.001

Variation in Quality by Demographic Characteristics

Adjusted results revealed that counseling caregivers on restricting access to lethal means prior to discharge had higher odds of being documented for female patients than for male patients (aOR 1.51, 95% CI: 1.07-2.14, p=0.02; Table 3a).

Table 3.

Variation in Performance on the Quality Measures by Demographic Characteristics

a. Documentation in Medical Record that Caregiver was Counseled to Restrict Access to Lethal Means
	n	Unadjusted %	Adjusteda OR (95% CI)
Age
5-12	206	67.0	0.87 (0.57, 1.33)
13-17	877	70.4	Referent
Sex
Male	365	65.5	Referent
Female	711	71.7	1.51 (1.07, 2.14)b
Race/ethnicity c
Hispanic	129	84.5	1.52 (0.82, 2.82)
White	643	65.8	Referent
Black	165	71.5	0.62 (0.38, 1.01)
Asian/Pacific Islander	26	57.7	0.38 (0.14, 1.01)
Other	71	70.4	0.76 (0.40, 1.44)
b. Documentation in Medical Record that Follow-up Plan was Discussed with Outpatient Provider
	n	Unadjusted %	Adjusteda OR (95% CI)
Age
5-12	198	19.2	1.32 (0.85, 2.06)
13-17	846	15.8	Referent
Sex
Male	354	17.5	Referent
Female	683	16.1	0.93 (0.64, 1.35)
Race/ethnicity c
Hispanic	123	20.3	0.80 (0.45, 1.41)
White	626	17.6	Referent
Black	155	9.0	0.45 (0.24, 0.84)b
Asian/Pacific Islander	25	16.0	0.51 (0.16, 1.63)
Other	67	20.9	0.81 (0.42, 1.57)

Abbreviations: OR, odds ratio; CI, confidence interval

^aAdjusted for patient age, sex, and race/ethnicity and site of care (random intercept). Missing values for patient demographic characteristics (age, sex, and race/ethnicity) are imputed with the hospital-level means for the purpose of models (i.e. adjusted ORs). Ns and unadjusted percentages are calculated amongst patients with non-missing values for each respective demographic characteristic (e.g. cases with non-missing values for race/ethnicity).

^bp<0.05

^cEach hospital provided an indicator of Hispanic ethnicity and a separate race variable. We created mutually exclusive race/ethnicity categories such that Hispanic patients of any race were classified as Hispanic and non-Hispanic patients were classified as White, Black, Asian/Pacific Islander, or Other.

Adjusted results also revealed communication between the inpatient and outpatient provider regarding the follow-up plan had lower odds of being documented for Black patients than for White patients (aOR 0.45, 95% CI: 0.24-0.84, p = 0.01; Table 3b).

Association between Measure Performance and Reutilization Outcomes

All-cause ED return visits within 30 days occurred for 8% of patients (87/1083 and 82/1044 for each measure respectively) and all-cause inpatient readmissions within 30 days occurred for 7-8% of patients (85/1083 and 78/1044 for each measure respectively); same-cause ED return visits within 30 days occurred for 2% (22/1083 and 21/1044 for each measure respectively) and same-cause inpatient admission within 30 days occurred for 2% (17/1083 and 16/1044 for each measure respectively). The odds of an all-cause 30-day readmission was significantly lower for patients with documented caregiver receipt of lethal means counseling (aOR 0.48, 95% CI: 0.28-0.83, p=0.008; Table 4). There were no other significant associations between measure performance (for either measure) and the other reutilization outcomes (see Table 4).

Table 4.

Adjusted Association between Quality Measure Performance and 30-day Reutilization

	Number (%) of Events	Adjusteda OR (95% CI)	P-value
Documentation in Medical Record that Caregiver was Counseled to Restrict Access to Lethal Means (n = 1083)
All-cause 30-day ED return	87 (8%)	0.64 (0.37, 1.11)	0.110
Same-cause 30-day ED return	22 (2%)	0.75 (0.23, 2.46)	0.639
All-cause 30-day readmission	85 (8%)	0.48 (0.28, 0.83)b	0.008
Same-cause 30-day readmission	17 (2%)	0.50 (0.14, 1.80)	0.292
Documentation in Medical Record that Follow-up Plan was Discussed with Outpatient Provider (n = 1044)
All-cause 30-day ED return	82 (8%)	1.05 (0.56, 1.97)	0.887
Same-cause 30-day ED return	21 (2%)	0.97 (0.26, 3.69)	0.968
All-cause 30-day readmission	78 (7%)	1.22 (0.66, 2.27)	0.525
Same-cause 30-day readmission	16 (2%)	1.03 (0.21, 5.12)	0.966

Abbreviations: OR, odds ratio; CI, confidence interval; ED, emergency department.

^aAdjusted for patient age, sex, and race/ethnicity and site of care (random intercept). Missing values for patient demographic characteristics (age, sex, and race/ethnicity) are imputed with the hospital-level means.

^bp<0.01

DISCUSSION

In this study, we found substantial variation across hospitals in documentation that caregivers were counseled on how to restrict access to lethal means prior to their youth’s discharge. The higher rates of documenting the occurrence of lethal means counseling for caregivers of female youth is noteworthy given the higher suicide completion rate in males.²³ When assessing whether documentation of communication between the inpatient and outpatient providers occurred regarding the follow-up plan, we again observed substantial variation, and poor performance with worse performance on this measure for Black patients compared to White patients. We observed that the odds of an all-cause 30-day readmission was significantly lower for patients with documented caregiver receipt of lethal means counseling. Like the preceding study¹⁷, our study found that male patients were less likely to have documentation of caregiver counseling on lethal means restriction. While the preceding study found that this was also true for Black patients, we found worse performance on the communication measure for this population. Overall, we found more variation on both measures, although similar poor performance on the communication measure. These collective findings, combined with our finding of decreased reutilization associated with one measure, support the need to identify best practices and improve the quality of care for these populations.¹⁷

Disparities relating to lethal means counseling have previously been observed in pediatric patients¹⁷ and are disconcerting given the decrease in suicide rates observed after reducing the availability of access to highly lethal means.^24,25 In other countries, reducing the availability of lethal and commonly used suicide means has been shown to reduce suicide rates by as much as 30-50%.²⁴ Immediately after an initial suicide attempt, youth are at high risk to make a subsequent attempt.²⁶ When access to lethal means is prevented, less deadly means of suicide/self-harm are substituted, fewer attempts are fatal, and through delay, suicide is averted.²⁴

While females statistically have higher numbers of suicidal attempts, the completion rate for male suicide is higher due to their disproportionate use of more lethal means.²⁷ While the literature on low uptake of lethal means counseling is sparse, and does not explain the observed disparity, at certain hospitals, there may be limited awareness of the importance of lethal means counseling or no effective strategies to ensure its receipt. In adult populations, it has been noted that some providers doubt the effectiveness of lethal means counseling or are uncomfortable delivering advice related to safe-storage practice, despite clear evidence of its feasibility and benefits.^28-30 High performing hospitals suggest that meeting this standard of care is feasible. Our study re-affirms that targeted efforts at improving care in this area, especially for male youth, is warranted.

Improving hospital transitional care remains a national priority.^31,32 Previous studies have shown the importance of coordinated preventive care for mental health concerns in decreasing potentially preventable readmissions.³³ Moreover, for those with an index suicide attempt, it has been shown that a scheduled psychiatry follow-up appointment has been associated with decreased subsequent odds of completed suicide.²⁶ The poor overall hospital performance and observed disparities on this quality measure might be explained by decreased access to care, possibly driven by a dearth of mental health specialists, but also by lower acceptance rates for all types of insurance in psychiatry compared to other specialties.³⁴ While lack of access to outpatient mental health services may explain some of the observed poor performance, communication with any outpatient provider, including those in primary care, was given credit for this quality measure. Prior research related to coordination of care indicates that time constraints in the inpatient setting and ambiguity over the receiving provider may also contribute to poor performance.³⁵ As with lethal means counseling, organized hospital efforts or incentives for communicating follow-up plans with the receiving provider may be absent.³⁵

The worse performance on this care transition communication measure for Black patients compared to White patients might be explained by ongoing differential access to treatment³⁶ and/or implicit or explicit bias.^37,38 Prior research on care transition disparities for youth with severe mental illness is sparse, specifically there is little information available on the intersection between race and an established medical home during transitions. However, recent reports call attention to differences in outcomes (pain reduction) by race and ethnicity in pediatric health care.³⁹ Prior studies have also found evidence of unconscious bias in academic faculty.⁴⁰ These studies and others show that bias may impact both clinical decision making and physician attitudes toward the patient.^39,41,42 Our study’s findings indicate a need for further investigation into the etiology of such disparities and the standardization of care to achieve optimal quality.⁴³

In this study, measure performance was not associated with 30-day same-cause reutilization outcomes, however, better performance on the lethal means restriction counseling measure was associated with fewer 30-day all-cause readmissions. This suggests that this outcome may be useful for assessing the effectiveness of other pediatric inpatient care processes for SI/self-harm. Providing lethal means restriction counseling may be a leading indicator for more thorough discharge planning taking place, resulting in the observed reduced frequency of readmissions. Other solutions to reduce readmissions might relate to policy work focused on lethal means restriction counseling, payment reform to support care across the continuum, and/or other efforts to assist caregiver counseling by hospital providers to enhance the quality of patient care.

Additionally, in this large, multisite study, only 2% of youth had a 30-day same-cause ED return visit and another 2% experienced a 30-day same-cause readmission. The rare occurrence of these outcomes compared to 30-day all-cause reutilization events suggests that these youth may be re-presenting with other relevant diagnoses, which are detected when examining all-cause but not same-cause reutilization events. On further analysis, we found that the top five all-cause readmission diagnoses were F32.9 (Major depressive disorder, single episode, unspecified), F41.1 (Generalized anxiety disorder), F41.9 (Anxiety disorder, unspecified), F33.2 (Major depressive disorder, recurrent severe without psychotic features), and F91.9 (Conduct disorder, unspecified). Among youth with an all-cause readmission, 49.4% had one of these five diagnoses associated with a readmission, and 25.9% had one of these five diagnoses as the primary diagnosis associated with a readmission. These findings highlight the importance of looking at both same-cause and all-cause readmissions when assessing outcomes of care received for SI/self-harm.

Limitations

This study has several limitations. Although conducted in a large, geographically dispersed sample of hospitals, the participating sites were all either freestanding children’s hospitals or pediatric units within academically affiliated adult hospitals located in urban areas. Thus, the results are not generalizable to youth receiving care for SI/self-harm in less urban settings or in non-academic community hospitals. We did not collect information on preferred language. We also did not have information on which patients had an established PCP prior to their hospitalization, which of these PCPs might be more accessible than others, and thus did not factor these variables into our analysis. We did not have information on parental employment, which may have affected caregiver ability to be at bedside for counseling. It is possible that parents were counseled on lethal means restriction and providers communicated with each other about the outpatient follow-up plan, but that these care processes were not documented. However, it seems unlikely that care providers would produce better documentation for White youth than Black youth or for females than males. The multi-stakeholder Delphi panel that endorsed these measures judged that lack of adequate documentation for these aspects of care is itself an indicator of poor quality.¹⁷ Also, we were only able to account for ED return visits or readmissions to the index hospital, thus we may have underestimated reutilization events. Lastly, it is possible that insurance status might prevent reutilization as might demographic characteristics related to socioeconomic status or commuting distance; however, we did not obtain data on insurance status, or location of residence, and thus these were not included in the adjusted models.⁴¹

Conclusions

In this large and diverse sample of youth, we found significant variation in performance across sites of care and by patient demographic characteristics. We observed exceptionally poor performance on communication between inpatient and outpatient providers regarding follow-up plans. Variation, performance gaps, and disparities along with our observation that all-cause 30-day readmissions were significantly lower for patients with documented caregiver receipt of lethal means restriction counseling suggest that proactive quality improvement efforts are needed. These might include policy approaches, payment reform, and/or other efforts to support counseling and care coordination by hospital providers to enhance the effectiveness and equity of patient care for this vulnerable population.

What’s New

Disparities in care for suicidal ideation/self-harm were detected across eight geographically dispersed US hospitals. Both male and Black youth received lower quality care. All-cause 30-day readmission was significantly lower for youth with documented caregiver receipt of lethal means restriction counseling.

Abbreviations:

US

United States

ED

Emergency Department

SI

Suicidal Ideation

QI

Quality Improvement

PRIS

Pediatric Research in Inpatient Settings

PCP

Primary Care Physician

EMR

Electronic Medical Record