Adolescent Intentional Ingestions in a Community Hospital

This study was a retrospective study from 2 emergency departments (EDs) and 1 community hospital in a suburban region in the Pacific Northwest over a 2-year period. Both EDs have general emergency medicine providers, and patients are admitted to the pediatric unit attached to 1 of the ED sites. The study was approved by the hospital’s institutional review board. Data were obtained through a query of the enterprise data warehouse by using International Classification of Diseases, 10th Revision codes T36 to T65 from July 1, 2016, to July 1, 2018, for patients 18 years of age and younger. The study cohort then underwent a full chart review by the study’s principal investigator. Codes in categories T36 to T65 are combination codes that include substances related to adverse effects, poisonings, toxic effects, and underdosing as well as the external cause. We, therefore, used the broadest approach to capture all possible ingestions. Each chart was then manually reviewed to evaluate for intentionality. Inclusion criteria included an intentional ingestion and age >9 years at the time of presentation. Exclusion criteria included the following: age <9 or >18 years, duplicate diagnosis entries for the same ingestion, adverse effect of prescription medication, recreational ingestion, accidental ingestion, and unknown reason for ingestion (Fig 1).

The substances ingested were grouped into the following 9 distinct categories: ibuprofen, acetaminophen alone or acetaminophen with other medications, salicylates, diphenhydramine or other antihistamine, illicit substance, ethanol, psychiatric medications, other prescription medications, and cleaning solution or unknown substance. The illicit substance category included marijuana, cocaine, methamphetamines, opiates, and lysergic acid diethylamide. Psychiatric medications included selective serotonin reuptake inhibitors (SSRIs), atypical antipsychotics, and other antidepressants. Psychiatric medications were not analyzed by specific subclasses but, instead, as a group. Prescription medications included benzodiazepines, anticonvulsants, attention-deficit/hyperactivity disorder (ADHD) medications, and other prescription medications. Psychiatric medications were not included in the prescription medication group.

Descriptive statistics were calculated overall as well as across inpatient medical and psychiatric admission statuses. For continuous data, mean and SD were used to describe normally distributed data (age). Median, 25th percentile, and 75th percentile were used to describe highly skewed data, including both ED and inpatient length of stay (LOS). P values were then generated by using the following tests: χ² tests for categorical data, Fisher’s exact tests for sparse categorical data (expected cell size <5), independent sample t tests for continuous and normally distributed data (age), and 2-sample Wilcoxon rank tests for highly skewed continuous data (ED and inpatient LOS).

Multivariable logistic regression models were used to determine if there were patient or event characteristics that were associated with having an inpatient medical or psychiatric admission. All multivariable models included sex, age, insurance, ED LOS, and whether a therapist was established before the event. We further controlled for clustering by facility using an indicator variable. Because we were limited by a small sample size, we were not able to include all substances ingested as predictors of admission. Instead, we removed substances with P > .20 in the multivariable model until all substances were removed from the model or all remaining substances had P < .20.

All analyses were 2 sided with an α of .05 as the threshold for statistical significance. Analyses were conducted in SAS version 9.4 (SAS Institute, Inc, Cary, NC).

Over the 2-year study period, there were 248 intentional ingestion encounters among 233 distinct individuals, an average of 1 ingestion every 3.2 days. The average age in the cohort was 15.3 years, with a female predominance of 80.3%. The prevalence of the type of substance ingested is described in Table 1, with the top 3 classes including psychiatric medications (30.9%), prescription medications (28.3%), and ibuprofen (24.0%). More than 1 type of substance was ingested in 32.6% of the cohort. The median LOSs for ED and inpatient settings were 5.0 (1.1–194.8 hours) and 68.4 hours (35.1–133.7 hours), respectively. Nearly one-third of patients (33.9%) required admission to a medical hospital, whereas one-quarter (24.9%) required admission to a psychiatric hospital. The majority of patients (56.7%) had public insurance, and more than half (54.9%) were already established with a therapist. Notably, there were no deaths during the study period.

Descriptive analyses revealed that female sex (P = .022), shorter ED LOS (P < .001), ingestion of psychiatric medications (P = .002), and ingestion of prescription medications (P = .042) were significantly associated with an inpatient medical admission (Table 1). Ingestion of ibuprofen was not associated with inpatient medical admission.

Results from a multivariable model revealed that a decreased ED LOS (odds ratio [OR] 0.8; 95% confidence interval [CI]: 0.8–0.9), as well as ingestion of salicylates (OR 5; 95% CI: 1.0–24.2), psychiatric medications (OR 4.5; 95% CI: 1.9–11), and prescription medications (OR 3.1; 95% CI: 1.3–7.0), was associated with an inpatient medical admission (Table 2). Furthermore, there was a nonsignificant trend (.05 < P < .10) revealing a relationship between female sex and lack of an established therapist with inpatient medical admission. There was also a highly significant difference in odds of inpatient admission and inpatient psychiatric admission between the 2 ED locations (OR 18.1; 95% CI: 4.9–66.6).

The unadjusted analysis revealed that having an inpatient medical admission (P < .001), an increased stay during that inpatient medical admission (P < .001), and ingestion of a psychiatric medication (P = .008) were all significantly related to having an inpatient psychiatric admission (Table 3). There were nonsignificant trends (.05 < P < .10) for an increased ED LOS, establishment with a therapist, and ethanol ingestions for those with an inpatient psychiatric admission. Again, ingestion of ibuprofen was not associated with inpatient psychiatric admission.

Patients with an inpatient medical admission were more likely to have an inpatient psychiatric admission (P < .001); of the 58 patients with an inpatient psychiatric admission, 37 (63.8%) also had a medical admission.

Finally, a multivariable model revealed that increased ED LOS (OR 1.1; 95% CI: 1.0–1.1), inpatient medical admission regardless of LOS (OR 8.4 [95% CI 2.4–29.6] for LOS less than the median and OR 57.7 [95% CI 16.4–202.8] for LOS greater than the median), and ED location (OR 10.5; 95% CI 1.2–92.4) were all significantly associated with having an inpatient psychiatric admission (Table 4).

With this study, we add needed literature about adolescent intentional ingestions in the community hospital setting. We describe the incidence of adolescent intentional ingestions in this setting and found multiple factors associated with both medical and psychiatric admission that can help guide providers. Furthermore, by performing an individual chart review, granular data were able to be extracted, which is typically difficult to perform with the administrative billing data that has been used in previous studies.^7,14,19 

Similar to many previous studies, our research revealed that girls composed the majority of patient encounters for intentional ingestions.^8,12,21  Research on adolescent suicidality (but not specific to intentional ingestions) has revealed a similar majority of girls.^7,14,15,19  This is in contrast to findings recently published by Burstein et al⁴  using the National Hospital Ambulatory Medical Care Survey to evaluate SA and SI encounters, which revealed a male predominance. It is unclear why this difference arose in the national data set.

The average patient age in our cohort was 15, consistent with previous studies focused on intentional ingestions, SI, and SA.^{7,8,12,16,21–24}  The most commonly ingested substances in our cohort included psychiatric medications, prescription medications, and ibuprofen. This is similar to multiple studies revealing analgesics and antidepressants to be the most common substances ingested in adolescent SA, most likely related to availability.^{20–23,25,26} 

In previous work, researchers have examined ED LOS for pediatric psychiatric-related visits.^27–29  In other studies, average ED LOS ranged from 3.2²⁸  to 11.1 hours.²⁷  Notably, these studies were not focused on intentional ingestions. In our cohort, the average median LOS in the ED specifically for adolescent intentional ingestions was 5 hours. This question can be complicated when comparing multiple sites because admission and boarding decisions can be dictated by numerous factors, including hospital policy and space.

In our study, close to one-third of patients presenting to the ED with an intentional ingestion required admission to a medical hospital. This is consistent with research by Finkelstein et al,¹²  who evaluated self-poisonings and ingestions. In studies focused on admission to the inpatient medical setting with SA and SI not specific to intentional ingestions, authors have estimated a range of rates of admission, with up to half of patients.^7,8,14,15,19 

Our study revealed that decreased ED LOS was associated with inpatient medical admission, as were ingestions of aspirin, psychiatric medications, and prescription medications. Ibuprofen was 1 of the leading substances ingested but was not associated with an admission. This is most likely related to the significant physiologic effects caused by aspirin, psychiatric medications, and prescription medications necessitating prolonged medical clearance. This is an important point when evaluating emergency care because a large burden has been felt by EDs in terms of bed flow because of increased LOS. Noting that almost one-third of patients required an inpatient admission for medical clearance can be important in hospital policies. Specifically, this knowledge improves the multidisciplinary approach required to coordinate the care involving sitters, nurses, social workers, hospitalists, and psychiatrists. Anecdotally, there is a wide variation in the practice of inpatient admission compared with boarding this population in the ED.

The duration of admission to the inpatient setting for those with SI and SAs has also previously been described. Carbone et al¹⁹  found an average LOS of close to 5 days. Plemmons et al⁷  found that more than half of encounters required <24 hours of admission but that ∼15% required prolonged hospitalization >7 days. Specific to intentional ingestions, Finkelstein et al¹²  found a median LOS of 2 days for their self-poisoning cohort. The average LOS in our cohort was close to 3 days (range 35.1–133.7 hours), underscoring the intense resource use and expenses incurred with intentional ingestions. Given the community setting of our cohort, our findings raise the question of whether the longer LOS is generalizable to other community settings and raise the question of factors contributing to this difference.

Data on previous psychiatric care in patients presenting with SI or SA are limited, but Olfson et al¹⁵  found that close to half of patients had an outpatient mental health visit in the 6 months before their self-harm event. In our cohort, approximately half of patients presenting with an intentional ingestion had been established with a therapist. We found a nonsignificant trend in not having an established therapist with an inpatient medical admission after an intentional ingestion. Conversely, we found a nonsignificant trend in having a therapist with risk of admission to inpatient psychiatry after an intentional ingestion. Both of these findings require further evaluation in larger multisite cohorts.

Previous research has revealed an increased percentage of self-harm encounters in those with public or government insurance.^3,6,7,14,19  Similar to these results, close to 56% of patients in our study had public insurance. The etiology of this finding is likely multifactorial but may be in part due to challenges to accessing mental health care.³ 

One finding that is interesting in our study is the significant difference in admission rates to inpatient medical and psychiatric settings between the 2 EDs. Although both sites are community hospitals in the same geographic location, we hypothesize that the difference in admission rates may stem from inherent differences in patient volume and acuity between the 2 sites. For instance, site 1 has double the number of any type of ED visit for patients of all ages per year compared with site 2. Furthermore, the attached hospital at site 1 has almost 3.5 times the total number of both adult and pediatric medical admissions per year, in addition to being the only site where pediatric patients can be admitted within the local system. Finally, the acuity of patients at ED site 1 is higher than at ED site 2, with emergency medical services personnel preferentially bringing adolescents with higher acuity to this location.

To our knowledge, this is 1 of the first studies in the United States in which the percentage of adolescent patients who required admission to a psychiatric facility after an intentional ingestion is examined. An Australian cohort study revealed a 13% rate of admission to a psychiatric facility.³⁰  In our study, close to 25% of patients required admission to inpatient psychiatry, with risk factors including an inpatient medical admission, increased LOS during that medical admission, and ingestion of a psychiatric medication. Inpatient medical admission is likely related to the severity of the ingestion, reflecting the need for close monitoring and support. The increased LOS may also reflect the severity of the ingestion but, in addition, may be related to the relative scarcity of inpatient psychiatric beds in the state. Many patients with intentional ingestion are medically cleared within 24 hours and then wait additional days for an inpatient psychiatry bed, which highlights the importance in increasing the infrastructure of psychiatric beds. Unfortunately, we were unable to capture the data on time waiting for medical clearance versus time awaiting psychiatry placement. Whereas in the majority of previous studies, authors have examined data from academic hospitals and national cohorts,^{3,7,12,15,16,19–21}  with our study, we add to the literature by providing data in the community hospital setting. Given previous literature revealing that up to 70% of pediatric admissions occur at general hospitals in the United States,³¹  it is paramount that continued research efforts are focused on this demographic.

With increasing rates of adolescents presenting to hospitals with intentional ingestions, better predictive models for both medical and psychiatric hospitalizations will allow hospitals to improve admissions flow and predict staffing needs, including sitters and specialized mental health clinicians (such as social workers and psychiatrists). Furthermore, a better understanding of the characteristics of this population, including the relative severity of attempts and average length of ED and hospital stay, will allow hospitals to develop and implement targeted safety protocols and interventions to decrease risk on discharge. This may include safety planning, counseling on lethal means restriction, and connection with community resources. A 3-day LOS after an SA provides a critical opportunity to decrease harm and prevent future SAs.

There are several limitations to this study. First, the fact that the study was performed within a single system in the Pacific Northwest limits generalizability, as does the inherent limitation of a retrospective chart review design. However, unlike data obtained from administrative billing, we had full access to clinical details for patients, allowing an improved determination of suicidal ingestions from recreational ingestions. In addition, an exhaustive review of all charts during the study period decreased the likelihood of missing a true suicidal ingestion. The second limitation to this study is the potential that these 2 sites were serving different populations given the large discrepancy in the hospital admission rates of adolescent patients with intentional ingestions between the 2 EDs involved, as previously described. This could partially be explained by the difference in acuity between the 2 sites, with the possibility of emergency medical services personnel preferentially bringing patients with higher acuity to the larger, higher-volume ED. However, other possibilities of provider bias or patterns of practice cannot be excluded. Third, we did not evaluate whether the ingested prescribed medication was prescribed to the actual patient versus a family member or versus being illicitly obtained. Fourth, we were not able to distinguish between subclasses of antidepressants but, instead, analyzed them as a group. Future studies should separate subclasses of this group for analysis to identify further high-risk medications (such as tricyclic antidepressants and bupropion), but generally, these medications require medical clearance necessitating medical admissions. Finally, we did not delineate medical clearance time versus time awaiting psychiatric placement for inpatient LOS.

With this study, we provide important data on the epidemiology of adolescent intentional ingestions in the community hospital setting (a demographic not previously well described). In addition, we provide variables that are associated with a higher incidence of inpatient medical and psychiatric admission. Further research is needed to evaluate these variables by using large, national, multisite cohorts.

The authors thank Dr Joshua Lovvorn and Dan Neilson for their contributions.