Nurse Education and Hospital Readmissions for Children With and Without a Mental Health Condition Free

In this cross-sectional study, we used data for the state of Florida from the 2015–2016 RN4CAST-US survey, 2016 Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases (SID), and the 2016 American Hospital Association (AHA) Annual Survey. This study received approval from the institutional review board of the University of Pennsylvania.

Patient inclusion criteria were:

1. patients were aged 3 to 21 years; and

2. patients were hospitalized for any medical condition or surgical procedure identified using the primary diagnosis field.²⁸ 

Patients with a primary psychiatric or obstetric diagnosis were excluded because this study is primarily concerned with how mental health conditions may affect the manifestation of medical conditions or surgical procedures in a general pediatric population and how nursing may be influential in preventing readmission. Very young children were excluded because studying them does not address the primary study question. The first eligible admission for each patient was included in the study; any subsequent admissions were counted as readmissions (if in the correct time frame) or excluded.

Hospital inclusion criteria were:

1. hospital had >20 discharges meeting inclusion criteria during the window of observation (January 1, 2016–December 31, 2016);

2. hospital characteristics were available from the HCUP SID and AHA Annual Survey;

3. at least 10 RN4CAST-US nurse survey responses were available from the hospital, the minimum number needed to create reliable nurse education estimates^2,6 ; and

4. complete data on patient readmissions were available for the hospital.

The HCUP SID was used to identify patient hospitalizations for inclusion in the sample. HCUP SID allows for identification of the diagnosis-related group category, all patient diagnoses (including mental health conditions and chronic complex conditions), and demographic characteristics. We also used the HCUP SID data to generate the primary outcome for this study (ie, 7-day readmission), which is only available for certain states, including Florida.

The RN4CAST-US nurse survey provides information on hospital working conditions from RNs in California, Pennsylvania, Florida, and New Jersey.²⁹  There was a 26% initial response rate, and an intensive nonresponse survey yielded an 87% response rate.²⁹  There were no statistically significant differences observed between the initial respondents and nonrespondents on the variables measuring hospital nursing resources, thereby addressing concerns about nonresponse bias.²⁹  For the current study, Florida was the only state providing complete data for patient readmissions and nurse working conditions; therefore our sample was derived from Florida.

The AHA Annual Survey provides information on hospital characteristics such as utilization, personnel, revenue, and expenses.

The primary exposure of interest was nurse education measured as a hospital-level proportion of nurses who reported that their highest nursing degree was a baccalaureate degree or higher.

A readmission was defined as any repeat hospitalization occurring within 7 days of discharge from an index hospitalization. Readmissions were identified using a unique encrypted patient identifier and days-to-event variable provided by HCUP.

We selected covariates for inclusion in multivariable models on the basis of factors known to influence risk of hospital readmissions. Hospital characteristics included bed size, teaching status, and technology status. Bed size was categorized as small (≤100 beds), medium (101–250 beds), or large (>250 beds). Teaching status was categorized on the basis of the ratio of medical residents/fellows per bed. Nonteaching hospitals had no residents/fellows, minor teaching hospitals had <1:4, and major teaching hospitals had ≥1:4. High-technology hospitals performed major organ transplants and/or open-heart surgery. Patient characteristics included age, sex, reason for hospitalization (diagnosis-related group), complex chronic conditions (CCCs), multimorbidity (2 CCCs or more), and LOS.³⁰  Two out of 12 CCC categories were not included as covariates in this analysis ([1] premature and neonatal, and [2] transplantation) because these categories represented <1% of the patient sample (N = 195 patients). For this study, patients with a mental health condition were identified using the Child and Adolescent Mental Health Disorders Classification System.³¹  The Child and Adolescent Mental Health Disorders Classification System was selected because it classifies child mental health disorders across coding systems and aligns with the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition. Conditions captured by this system include both psychiatric and developmental/behavioral diagnoses, which can be viewed elsewhere.³²  All conditions were included in a binary variable indicating whether a mental health condition was present or not.

Descriptive statistics were generated to examine characteristics of study hospitals and determine differences in clinical and demographic characteristics between patients with and without a mental health condition. We employed a series of logistic regression models to estimate the unadjusted and adjusted effects of a mental health condition and baccalaureate education separately on 7-day readmission, and the interaction effects of a mental health condition and baccalaureate education on 7-day readmission in models fully adjusted for patient (age, sex, LOS, multimorbidity, CCCs, and major disease category) and hospital characteristics (bed-size, teaching status, and technology status). The first model identified differences in odds of readmission between individuals with and without a mental health condition. The second model examined differences in odds of readmission between hospitals with different levels of nurse education. The final interaction model answered the question of whether the effect of nurse education education is similar for individuals with and without a mental health condition. We depicted these findings graphically by plotting the predicted probabilities of 7-day readmission at varying proportions of baccalaureate nurse education for hospitals. The statistical significance level was set at 0.05. Analyses were conducted with Stata statistical software (version 17.0; Stata Corp, College Station, TX).

The sample included 35 081 patients admitted to 122 hospitals with 4440 nurses. Table 1 presents characteristics of the 122 hospitals in the sample. The proportion of nurses holding a baccalaureate degree or higher varied between hospitals (range: 8.3%–80.0%, mean: 46.4%). Most of the hospitals in the sample were large and minor teaching institutions with high-technological capacity.

Table 2 presents characteristics of the 35 081 patients admitted to 122 hospitals. About 21.0% of these patients had a diagnosed mental health condition, the most common mental health condition was attention-deficit/hyperactivity disorder (5.1%). These patients were most commonly hospitalized for gastrointestinal and digestive conditions (eg, gastroenteritis and constipation) (19.3%), with a median LOS of 2.0 days. The median age of patients in the sample was 16.0 years, and 52.7% of patients were male. Approximately 9.4% of patients had 2 or more CCCs. When compared with individuals without a mental health condition, those with a mental health condition were more commonly hospitalized for neurologic conditions (eg, epilepsy) (17.9% vs 9.8%) than those without a mental health condition. Individuals with a mental health condition were older (18.0 years versus 15.0 years) and more commonly male (57.6% vs 51.4%), with a higher incidence of multimorbidity (12.5% vs 8.6%) than those without a mental health condition. In terms of our main outcome of interest, individuals with a mental health condition had a 7-day readmission rate of 5.7%, compared with a 7-day readmission rate of 3.9% for those without a mental health condition. Patients with and without a mental health condition had an equivalent median LOS (2.0 days).

Table 3 shows odds ratios (ORs) depicting the direct effects of a mental health condition and baccalaureate education on 7-day readmission separately, and how the interaction between patient mental health condition and baccalaureate education affects odds of 7-day readmissions. When adjusting for hospital and patient characteristics, individuals with a mental health condition were more likely to be readmitted within 7 days of discharge (OR = 1.27, 95% confidence interval [CI] = 1.11–1.45) than those without a mental health condition. For all individuals, each 10% increase in the proportion of nurses holding a baccalaureate degree or higher was associated with 10% lower odds of readmission within 7 days of discharge (OR = 0.90, 95% CI = 0.86–0.94). Although receiving care in a hospital with a greater proportion of baccalaureate-educated nurses had a significant and negative association with 7-day readmission for all individuals, the interaction model suggests that individuals with a mental health condition are more sensitive to variability in baccalaureate education. Each 10% increase in the proportion of baccalaureate-educated nurses was associated with 8% lower odds of 7-day readmission for individuals without a mental health condition (OR = 0.92, 95% CI = 0.87–0.97), but 16% lower odds of readmission for individuals with a mental health condition (OR = 0.84, 95% CI: 0.78–0.91). This interaction is depicted graphically in Fig 1.

In this study, we found that children hospitalized for medical or surgical indications were less likely to be readmitted within 7 days if they received care in a hospital with a higher proportion of baccalaureate-educated nurses. This relationship was especially pronounced among children with a mental health condition. These findings are consistent with previous research demonstrating that being cared for by a baccalaureate-educated nurse is associated with a lower risk of hospital readmission in adults.⁶  These findings also support the National Academy of Medicine initiative to progress to an 80% baccalaureate-educated workforce by 2020.³³ 

This study adds new findings to what is known about the relationship between nurse education and readmission by validating this relationship in pediatric care. Current evidence regarding nurse education and pediatric patient outcomes is limited to pediatric critical care.^10,11  Outcomes in critically ill children are sensitive to variation in baccalaureate education; for example, each 10% increase in baccalaureate-educated nurses is associated with 15% to 17% lower odds of postoperative complications.^10,11 

The RN is central to the care of hospitalized pediatric patients and the detection and prevention of potential causes of readmission. This detection process is known as surveillance, which involves gathering, analyzing, and synthesizing patient data for clinical decision-making.³⁴  The side effects of some psychotropic medications (eg, hypertension and hyperglycemia), combined with the physical and psychological manifestations of mental illness, may result in delays in the detection of medical complications and preventable deaths.³⁵  Symptoms related to the individual’s mental health condition may also impair an individual’s navigation of and access to health care systems; for example, children with autism may have difficulty with the sensory stimuli in health care environments.³⁶  Because of this complexity, individuals with a mental health condition require care from nurses who are able to think critically, coordinate care with their interdisciplinary colleagues, and provide ample discharge teaching. Many nurses feel unprepared to care for patients with mental illness and report that caring for these patients is stressful, difficult, and unrewarding.³⁷  Baccalaureate degree programs emphasize these competencies and prepare nurses to provide holistic care to this vulnerable patient population.

With various paths to baccalaureate education, these findings are actionable. On the basis of the current evidence, the United States is lagging behind on the National Academy of Medicine’s goal of an 80% baccalaureate-educated workforce, with ∼66% of nurses expected to have baccalaureate education by 2025 and an average of 46.4% baccalaureate-educated nurses per hospital in the most recent data from 2015 to 2016.³⁸  There are 2 main pathways to baccalaureate education:

1. nurses may advance from a diploma or associates degree to a bachelor’s degree via a baccalaureate completion program; or

2. nurses may complete a traditional 4-year college program.

About 48% of newly conferred baccalaureate degrees are earned through the first option, RN to Bachelor of Science in Nursing programs.³⁸  Importantly, the educational pathway is not associated with patient outcomes, but instead achievement of a baccalaureate degree in nursing.³⁹  Beyond baccalaureate education, programs to educate nurses and other health care providers in considerations unique to the care of children with a mental health condition are likely to benefit patients and families.

Future research should examine the mechanisms underlying the association between baccalaureate education and readmissions in general pediatrics. A greater volume of research demonstrating the link between baccalaureate nursing education and outcomes may incentivize investments into the nursing workforce. To put these findings into context, an increase in the proportion of baccalaureate-educated nurses within a hospital from 20% to 80% could be associated with 66 fewer readmissions per 1000 patients with a mental health condition. However, this same change would be associated with 19 fewer readmissions per 1000 patients without a mental health condition.

Some limitations in the study design and data should be noted. The cross-sectional design prevents us from drawing conclusions about causality between Bachelor of Science in Nursing education and readmission in this population. The study did not link individual nurses to patients, but instead only hospitals, which limits our ability to conclude the observed associations were because of nursing education. There may be omitted variable bias related to the kinds of hospitals that have more baccalaureate-prepared nurses in their workforce. Additionally, study data are from 2016 and from one state, and we note age and location of the study data as possible limitations to generalizability. The median of the entire cohort was notably older than the average population of hospitalized children, which was consistent with previous literature utilizing these data but may not be representative of a typical pediatric hospital sample.⁴⁰  In the future, researchers should replicate this study with data reflective of a traditional pediatric cohort, and include information on where the patients were bedded. The data also limited our ability to adjust for variables related to physician care quality, experience, or specialty, which may have influenced the outcomes. We were also limited in our ability to discern whether older children and adolescents were hospitalized on a pediatric or adult unit. We have no reason to suspect that the relationships observed have changed over time, and relationships between these variables have remained consistent at different time points and in different settings in other studies.^{18,21–23,26}  Finally, future research should include analyses of subgroups of mental health conditions to help understand the mechanisms behind these findings.

Being cared for in a hospital with more baccalaureate-educated nurses is associated with lower odds of 7-day readmission for all children and adolescents, and the effect of baccalaureate education is strongest in individuals with a mental health condition. Pediatric patients with a mental health condition are more likely to be readmitted within 7 days of discharge than those without a mental health condition, and being cared for in a hospital with a higher proportion of baccalaureate-educated nurses lowers readmission rates. Our findings suggest that transitioning to a largely baccalaureate-educated workforce may improve outcomes for hospitalized children, particularly for the growing population of children who have a mental health condition.

We thank Dr Douglas Sloane, Dr Morgan Peele, and Mr Timothy Cheney for their contributions to data analysis on this study.