When inequity impacts clinical care: an analysis of length of stay and reimbursement rates for medical stabilization for anorexia nervosa based on insurance coverage

Abstract

Background

The recent COVID-19 pandemic has had a significant effect on the rates and severity of anorexia nervosa (AN) leading to rises in admissions for medical stabilization. To improve care and effectively allocate limited resources, we need a better understanding of the factors that lead to prolonged hospitalization in patients with AN.

Methods

We performed a retrospective chart review of 139 adolescent and young adult patients admitted a combined 196 times for medical instability due to AN at a large pediatric hospital in the Midwest region of the United States. We used multiple linear and Poisson regression modeling to look for differences in length of hospital stay and reimbursement rates based on payor type (public vs. private), while controlling for demographic and clinical factors which may impact length of stay. We then looked at hospital charges and reimbursement rates by payor type, using two-sided t-tests.

Results

Multiple regression models demonstrated that patients with public insurance tended to have a longer length of stay, despite similar degrees of medical instability. Reimbursement rates were significantly and substantially less from public insurance companies compared to private.

Conclusions

To our knowledge, this is the first study examining length of stay and reimbursement rates for medical stabilization based on insurance status. Understanding the effects of a tiered insurance scheme on both clinical and fiscal outcomes will be key to improving care provided to patients with AN.

Background

Anorexia Nervosa (AN) is a severe psychiatric illness with the second highest mortality rate of any mental health illness [1, 2]. Those with more severe presentation often require hospital admission due to medical instability. Admission criteria for medical stabilization for patients with AN include physiologic instability, electrolyte abnormalities, electrocardiogram abnormalities (prolonged QTc, bradycardia, arrhythmia), significant sequelae of eating disorder (gastrointestinal bleed, seizures, syncope, cardiac failure, pancreatitis, acute refusal to eat), among many others [3]. The past 5 years came with increasing rates and severity of eating disorders. Monthly hospitalization rates increased between 168 and 208% during the COVID-19 pandemic, with similar rates reported when conducting annual hospitalization analytics [4–7]. Readmission rates were also greatly impacted, as Matthews et al. showed that rehospitalization for youth with AN and atypical AN within the first 30 days after discharge increased eightfold [8].

The origin of the increase of hospitalization rates and readmission rates for patients with AN is multifactorial. Duration of time before accessing outpatient therapy, distances to access care, transition to telemedicine platforms, changes in frequency of therapy sessions, and worsening of comorbid conditions have all been described as contributing factors [9, 10]. Otto et al. found that rises in admissions occurred 9–12 months after the commencement of COVID-19, suggesting that patients faced a lack of, or delay in, access to care. This same study also found that 88% of patients admitted during the pandemic had private forms of insurance versus 80% pre-COVID, which suggests that eating disorder treatment during the early stages of the pandemic may have been less accessible for those with public insurance [11]. In the United States, health insurance is categorized as either private (e.g., provided or subsidized by an employer, directly purchased by patients and families, military-sponsored care) or public (Medicaid, government-funded, “safety net”) coverage. In the U.S., access to medical stabilization and therapeutic resources for patients with AN is highly dependent on insurance status. Generally, for patients to be discharged, medical complications and instabilities need to have resolved or significantly improved, the patient needs to be receiving and accepting nutrition, and a safe discharge care plan needs to be established. The latter is critical and may shed light on inequities in access to appropriate levels of treatment. One study found that patients with public insurance were one-third as likely to receive recommended treatment compared to patients with private insurance [12]. Furthermore, when adjusting for relevant demographic and clinical factors, Latinx and Asian youth were half as likely to receive recommended treatment compared to White youth [12]. In our area, many outpatient eating disorder therapists do not accept any type of insurance (patient pays full balance) or are “out-of-network” (i.e., not a preferred provider of the insurance, thus requiring higher fees), even for those with private insurance. Among those certified in Family Based Treatment (FBT), the most evidence-based therapy modality for children and adolescents with AN, only 5% of clinicians accept Medicaid insurance [13]. This system can be difficult to navigate and often forces families to make difficult financial decisions regarding their child’s care.

If a patient requires more therapeutic support, a higher level of care (intensive outpatient, partial hospitalization, or residential program) is recommended. However, these programs are more likely to accept patients with private insurance, thereby limiting access for patients with public insurance. In our area, none of the residential treatment programs take Medicaid, and of the partial hospitalization programs, only one takes one specific kind (managed care organization) of Medicaid. As a result, families with public insurance are often unable to directly link to care following medical hospitalization, placing patients at risk for prolonged hospitalization. Given this landscape, the goal of our current study was to compare length of stay and reimbursement rates for hospitalization from restrictive eating for those with public versus private insurance, while controlling for demographic factors, severity of illness and comorbid conditions. To improve care and effectively allocate limited resources, we must understand the potential factors that lead to prolonged hospitalization in patients with AN. To our knowledge, this is the first study in the United States to examine the financial impact of admissions for medical stabilization of AN since admission rates have increased.

Methods

This study is a retrospective chart review of clinical and billing data for patients aged 9 to 21 years hospitalized for medical stabilization of anorexia nervosa, at a large pediatric hospital in the Midwest region of the United States, between January 2020 and September 2023. Primary outcomes were differences in length of hospital stay and reimbursement rates for patients who were medically unstable secondary to anorexia nervosa, based on payor type (public vs. private). Healthcare organizations are required to look at financial implications of providing medical care, so we included a secondary outcome evaluating fiscal data, including amount billed and amount collected, per hospital day.

Inclusion criteria comprised of patients between the ages of 8 and 25 years admitted for the stabilization of medical complications due to anorexia nervosa, rule-out anorexia nervosa (i.e., strong possibility of AN), or unspecified restrictive eating disorder. Patients were included if any part of their hospitalization occurred during the study timeframe. Exclusion criteria included patients who did not have a restrictive eating disorder as defined above (e.g. patients with avoidant-restrictive food intake disorder) and patients who were admitted for reasons other than medical instability due to the eating disorder (e.g. suicidal ideation, influenza-like illness).

Demographic, clinical, and financial data were obtained from the electronic medical record (EMR) by the Data Analytics and Reporting (DAR) team at our institution. Patients were initially identified through diagnosis codes (anorexia nervosa, malnutrition) or placement on the institution’s eating disorder clinical care guideline (CCG) order set. The initial list was cross-referenced with a list kept in the eating disorder clinical program for process improvement and clinical tracking purposes. Individual chart reviews were performed to obtain missing data and the body mass index Z-score (BMIz) at admission. Chart reviews were completed by teams of two or more research team members to ensure data veracity. For interpreter agreement, one team member retrieved and input data while one to two members reviewed the medical chart and data entry for accuracy.

StataNow Version 18 MP was used for statistical analysis. Descriptive statistics and inter-group baseline clinical comparisons were performed. Because admission and discharge times are influenced by hospital census and staffing, both multiple linear and Poisson regression modeling were used to observe differences in length of stay, and logistic regression for rehospitalization rates. Secondary outcome measures of billing and payment were analyzed using two-sided t-tests.

Results

Demographic variables by insurance type

Our study included 139 patients, with 77 having private insurance and 62 having public insurance. There was no significant difference in mean age between privately insured (M = 15.17 years) and publicly insured patients (M = 14.50 years; p = 0.063). However, there were significant differences in primary language, sex, race, and ethnicity between privately- and publicly-insured patients (Table 1). Primary language differed significantly (p < 0.0005), with privately insured patients primarily speaking English and publicly insured patients showing greater linguistic diversity, mainly English and Spanish. Given language differences between groups, interpreter services were used more frequently by publicly-insured patients, with 27 requiring an interpreter compared to only 3 privately insured patients (p < 0.0005). Patient sex assigned at birth also varied significantly between groups, with privately-insured patients comprising 70 females and 7 males, while publicly insured patients included 49 females and 13 males (p = 0.047). Privately insured patients were predominantly White (n = 52), whereas publicly insured patients were more likely to identify as “Other” (n = 45), with only 11 White patients (p < 0.0005). The majority of those whose racial identity was “Other” in our EMR also ethnically identified as “Hispanic/ Latino”. A higher proportion of privately-insured patients identified as “Not Hispanic/Latino” compared to publicly insured patients (p < 0.0005).

Table 1.

Patient demographics

Demographic	Insurance	Sample Mean or N	Significance
Age (Avg if multiple Admissions)			p = 0.064
	Private (n = 77)	15.17 year (14.65–15.68)
	Public (n = 62)	14.50 year (14.03–14.97)
Language			p < 0.0005
	Private (n = 77)	English (74)
		Other (specified) (3)
	Public (n = 62)	English (31)
		Spanish (27)
		Other (specified) (4)
Interpreter Used			p < 0.0005
	Private (n = 77)	No (74)
		Yes (3)
	Public (n = 62)	No (35)
		Yes (27)
Sex			p = 0.047
	Private (n = 77)	Female (70)
		Male (7)
	Public (n = 62)	Female (49)
		Male (13)
Race			p < 0.0005
	Private (n = 77)	White (52)
		Other (specified) (6)
		Other/Multiple (17)
	Public (n = 62)	White (11)
		Other (specified)_(3)
		Other/Multiple (47)
		Declined (1)
Ethnicity			p < 0.0005
	Private (n = 77)	Not Hispanic/Latino (64)
		Hispanic/Latino (11)
		Unknown (2)
	Public (n = 62)	Not Hispanic/Latino (16)
		Hispanic/Latino (46)

Cells < 5 are grouped as “Other (specified)” in order to ensure privacy

Clinical health factors by insurance type

A comparison of categorical health factors between patients with private versus public insurance revealed no statistically significant differences across several variables (Table 2). Patients requiring nasogastric (NG) tubes did not significantly differ between payor groups, such that 29.9% of privately insured patients required NG tubes compared to 17.7% of publicly insured patients. Pediatric intensive care unit (PICU) admissions were low across both groups, with 2.6% of privately insured patients and 1.6% of publicly insured patients (p = 0.660).

Table 2.

Patient clinical variables

Patient Clinical Variables	Insurance Type	Sample Mean or N	Significance
Nasogastric Tube			p = 0.098
	Private (n = 77)	No (54) Yes (23)
	Public (n = 62)	No (51) Yes (11)
Minimum Phosphate			p = 0.186
	Private (n = 77)	3.13 (2.98–3.28)
	Public (n = 62)	3.28 (3.12 − 3.43)
Minimum Potassium			p = 0.853
	Private (n = 77) Public (n = 62)	3.68 (3.59–3.77) 3.69 (3.61–3.78)
Max Creatinine			p = 0.079
	Private (n = 77) Public (n = 62)	0.728 (0.689–0.767) 0.667 (0.608–0.726)
Hepatitis (ALT > 90 U/L)			P = 0.077
	Private (n = 77)	No (71) Yes (6)
	Public (n = 62)	No (56) Yes (6)
Minimum HR			p = 0.134
	Private (n = 77) Public (n = 62)	41.7 (39.2–44.2) 44.3 ( 41.9–46.7)
BMI Z Score At Admission			p = 0.329
	Private (n = 77) Public (n = 62)	-2.33 (-3.05– -1.60) -1.88 (2.35– -1.4)
Pre-Admission Weight Loss (kg)			p = 0.713
	Private (n = 77) Public (n = 62)	10.62 (8.539–12.71) 11.19 (8.942–13.43)
Pre-Admission Weight Loss (percentage)			p = 0.986
	Private (n = 77) Public (n = 62)	19.26 (16.46–22.07) 19.30 (16.27–22.33)

Mean values and 95% confidence intervals (CIs) for continuous health factors were compared by payor groups, revealing no statistically significant differences between privately- and publicly-insured patients in the following health factors: minimum phosphate levels (p = 0.186), minimum potassium levels (p = 0.853), maximum creatinine levels (p = 0.079), minimum heart rate (p = 0.137), BMI z-scores at admission (p = 0.329), and pre-admission weight loss (both in kilograms and as a percentage of body weight; see Table 2). There were no significant differences in rates of hypoalbuminemia (albumin ≤ 3.5 g/dL) or hepatitis (defined as ALT ≥ 90 U/L) between groups.

Length of stay

Both linear and Poisson regression were used to assess differences in length of stay by payor type (Table 3). Multiple linear regression modeling predicted a longer average length of stay for publicly-insured patients (β = 2.88, p = 0.021, 95% CI 0.445–5.32), and a longer maximum length of stay (β = 5.01, p = 0.027, 95% CI 0.588–9.44), while controlling for patient age, ethnicity, NG tube use, presence of hypoalbuminemia, minimum phosphorus and potassium, PICU admission requirement, and BMIz at admission. Because admission to the hospital and discharge from the hospital can vary by many hours due to non-clinical issues such as patient census, staffing, and transportation, days of hospitalization was examined as a count variable. Poisson regression models, while controlling for the same demographic and clinical covariates, also predicted longer average number of days of admission (β = 0.186, p = 0.003, 95% CI 0.0650–0.307) and maximum number of days of admission (β = 0.298, p < 0.001, 95% CI 0.185–0.412). Poisson is logarithmic; this translates to a predicted increase in average stay of 1.53 (95% CI 1.16–2.03) days and predicted increase in maximum stay of 1.99 (1.53–2.58) days (Table 3). Differences in the total number of hospitalizations per patient, or whether or not an individual patient was readmitted during the course of the retrospective chart review, were not observed. Although there was no significance in the length of stay for linear regression when controlling for preadmission weight loss rate in addition to BMIz, we included this as this is critical to consider given the potential guideline changes away from BMIz and use of degree of weight loss when considering admission criteria. Significance may have been lost given that not all patients had pre-admission weight data. There was still statistical and clinical significance when running Poisson regression.

Table 3.

Length of Stay

Fiscal data by insurance type

We used two-sided t-tests to compare the fiscal data between payor types (Fig. 1). No differences were seen in amount billed per day (in U.S. dollars), with publicly-insured patients billed $6,115.74 (95% CI 5,907.73–6,323.76), and privately-insured $7,248.35 (95% CI 5,273.36–9,223.35). However, the amount collected by the hospital differed substantially (p < 0.001): publicly-insured $1,113.66 (95% CI 983.67–1,243.65), privately-insured $4,991.68 (95% CI 4,557.51–5,425.85).

Fig. 1.

Charges and Payments in U.S. Dollars, two-sided t-test

Discussion

As hypothesized, patients with anorexia nervosa, with publicly-funded insurance, had significantly longer hospital stays on average than those with private insurance. They also had a significantly higher maximum length of hospital stay than their privately insured counterparts. Intuitively, longer hospital stays typically indicate a more serious illness or complicated course [14]. Data from our sample, however, suggested that this was not the case. Despite differences in length of stay, patients with public insurance did not appear to have more serious illness. Patients across insurance types had similar rates of health factors and conditions which, when present, indicate more severe illness: PICU admissions, NG tube use, hypoalbuminemia and hepatitis, electrolyte abnormalities, heart rate disturbances, and substantial pre-admission weight loss.

On our unit, patients are typically discharged when medical complications have resolved or significantly improved (e.g. severe bradycardia improves to a rate of > 45 beats per minute), the patient is receiving and accepting nutrition (able to tolerate a prescribed number of calories and has no evidence of severe refeeding syndrome), and a discharge care plan is established. Given the similarities in the above illness severity indicators across insurance groups, it is unlikely that difficulties achieving medical stabilization explain the increased length of stay for publicly insured patients. In addition, as the use of NG tubes was similar across groups, there is limited evidence that one group had more difficulties receiving and accepting nutrition compared to the other group. Rather, the most likely issue contributing to longer hospital stays is difficulties establishing a safe discharge plan for patients with public insurance. The financial burden of the extended hospitalization is then passed on to the hospital; if a patient is medically stable and there is no safe discharge plan, it is incumbent on the hospital to either appeal to the insurer for additional days of hospitalization, or accept the full financial burden of continued hospitalization.

The consequences of a lengthy hospitalization are not only financial. Patients and family can suffer unintended consequences in the settings of prolonged hospitalizations, including increased disruption of daily life (leaves of absence from school and work), increased stress and possible long-term psychological sequelae [15]. Lengthy hospitalizations also decrease bed availability and prevent hospitals from serving other children, especially during high patient census times like respiratory virus season. Finally, staff can be impacted by long hospitalizations, resulting in increased compassion fatigue and higher nurse turnover [16].

In the region surrounding this hospital and throughout much of the United States, access to high-quality, evidence-based eating disorder care is challenging. Patients discharged from the hospital may be referred to outpatient clinicians, who often provide Family-Based Treatment (FBT) for eating disorders. While FBT is the gold-standard treatment for eating disorders in adolescent and some young adult patients, it is very demanding for both patients and caregivers to successfully complete, and it has a relatively high failure rate even for those who can complete treatment [17–19]. Higher levels of care, including intensive outpatient, partial hospitalization, or residential programs, are typically recommended for patients who do not respond to FBT or who are not able to do FBT due to its high demands [20]. Unfortunately, these levels of therapy are reserved for those with private insurance, leaving patients who have public insurance with almost no options outside of outpatient follow-up. In addition, not all higher levels of care accept those who are assigned male or identify as male, which is important to note since there was a significantly higher number of males in the publicly insured group further compounding the difficulty in appropriate therapy placement.

Medicaid is a publicly funded healthcare system that is regulated by the federal government but allows states to determine eligibility criteria and allocation of state funding [21]. There are nationwide racial and ethnic disparities among those who are enrolled in Medicaid versus those enrolled in private insurance [22]. Our results mirrored nationwide data that suggests that families who are more likely to have public insurance also tend to be from marginalized groups that already place them at higher risk for experiencing difficulties accessing care [23]. While the expansion of Medicaid in some states, including the state in which this study data was collected, has offered more services for some medical and mental health conditions, treatment centers and individual providers for eating disorders rarely accept Medicaid as a form of coverage, which leaves a vast number of patients in racial and ethnic minority groups with limited access to post-hospital care [13, 18, 24]. This further widens the disparity gap.

In addition to having less access to care due to insurance coverage, racially or ethnically minoritized youth have added barriers finding providers who are bilingual, bicultural, or have access to interpretation services [12, 25]. In this study, when compared with privately-insured patients, patients who had public insurance were more likely to identify as racially diverse and/or “Hispanic/Latino”, more likely to speak a language other than English (typically Spanish), more likely to use an interpreter, and more likely to be male. These factors are critical to consider, given the fact that not all levels of care can support language differences and some programs are reserved solely for those who are assigned female or identify as female.

The findings of our study clearly highlight the individual and systemic impacts that coverage via public insurance may have on patients and families with anorexia nervosa, specifically longer hospitalizations, limited access to post-hospital care, and furthering societal racial/ethnic disparities. And while our study did not directly assess the suspected consequences of longer hospitalizations, one can imagine the sustained stress on patients, families, and providers [15, 16]. This should be evidence enough to encourage providers and hospital leadership to advocate for increased coverage opportunities amongst all levels of eating disorder care. An additional challenge, however, lies within inadequate reimbursement rates for hospital networks and providers. Reimbursement rates for eating disorder treatment are highly variable and are disproportionately lower when patients are publicly insured [26].

Our study demonstrates how dramatic that difference can be and leads to concern that providers and medical organizations may be hesitant to increase access to patients with public insurance given the potential for significant financial strain. With that said, we envision several opportunities for advocacy and growth, including negotiating higher reimbursement rates, providing parity in coverage, and advocating for alternative payment options such as sliding scale fees and scholarships in an effort to broaden access to coverage. Higher reimbursement rates and full coverage parity may make it more appealing for both inpatient and outpatient providers, hospital networks and treatment centers to accept Medicaid as a form of coverage, with fewer concerns about financial consequences to the institution. Sliding scale fees and scholarship opportunities are unlikely to be sustainable for large numbers of patients, however, would at least serve as a safety-net in scenarios when public insurance is not accepted. We propose that when working in unison, these avenues would increase the number of patients with anorexia nervosa who can be seen and successfully treated post-hospitalization, and thereby prevent prolonged hospitalizations that may result from lack of a safe discharge plan. And with that, we encourage medical providers and hospital leadership to read this discussion as a call to action. We must step outside the walls of our hospitals and clinics in attempt to educate state and federal legislators about the disparities within eating disorder care and advocate for the implementation of strategies that bridge the access and reimbursement gaps. Our patients deserve it and depend on it.

Strengths

This study was the first of our knowledge to examine financial implications of prolonged hospitalizations and readmission rates for eating disorder patients. While the sample size was somewhat small, there was diversity in the sample demographics and an equal representation between patients with private insurance and patients with public insurance. This study was also able to examine important research questions in a novel way and has practical implications that apply to real-world scenarios. There is inequity in eating disorder care that needs to be addressed, and this research was able to demonstrate some of those inequities and barriers that patients and families face, calling for future change and growth opportunities.

Limitations

As this study was limited to a freestanding children’s hospital in one major city in the United States, the results may not be generalizable to other major cities or suburban or rural areas. Furthermore, this was a retrospective chart review. While the data were collected within a recent time frame, limitations of retrospective chart reviews can include inconsistencies in the data and reliance on information in charts that were not originally designed for the purpose of research. Patients could have been re-admitted at an outside hospital during the study period, and this was not captured in our EMR. This study also had a limited sample size, so future research should focus on collecting data from a larger sample.

Future directions

Future research should include examining the consequences of longer hospitalizations and higher readmissions, both on patients and families and on providers. Providers may struggle from compassion fatigue when working with patients for extended periods of time or on a reoccurring basis, families often must sacrifice a great deal of time and energy into the care of their loved one in the hospital, and patients may lose motivation or desire to recover with increased or prolonged hospitalization. Furthermore, future research can assess the long-term impact of limited treatment options through longitudinal studies. Assessment of patients post-hospitalization over a longer period may provide insight into how to provide better care both while in the hospital and post-discharge. In addition, larger sample sizes from different geographic locations would be useful in understanding trends nationwide. Overall, due to the severity of restrictive eating disorders and difficulty with accessing treatment, future research should look to understand earlier and more effective interventions and to promote advocacy for decreasing barriers for all patients with eating disorders.

Conclusions

In a two-tiered healthcare system, anorexia nervosa patients with public insurance tend to have longer medical stabilization hospitalizations when compared to patients with private insurance. At the same time, hospitals are reimbursed markedly less for publicly-insured patient hospitalizations. This combination could have the tragic effect of fewer medical facilities accepting public insurance, to remain financially solvent.

Abbreviations

AN

Anorexia nervosa

FBT

Family–based treatment

EMR

Electronic medical record

DAR

Data analytics and reporting

CCG

Clinical care guideline

BMIz

Body mass index z–score

NG

Nasogastric (tube)

PICU

Pediatric intensive care unit

CI

Confidence interval

Author contributions

A.B., F.M, S.C., K.H, G.M., L.N., and R.A. conceived and designed the study; A.B.; S.C.; K.H.; G.M.; L.N., and K.O. acquired the data; A.B.; C.F., K.H., A.K., G.M., L.N., and L.S. analyzed and interpreted the data; C.F., K.H., G.M., and L.N. wrote the main manuscript; G.M. prepared the tables and figures. All authors reviewed the manuscript.

Funding

This study received no funding.

Data availability

Deidentified data is available upon request, by contacting corresponding author.

Declarations

Ethics and consent to participate

The study was approved by the Institutional Review Board at Ann & Robert H. Lurie Children’s Hospital of Chicago (IRB 2021–4824). Parent/patient consent or assent was not required for this retrospective chart review.

Competing interests

The authors declare no competing interests.