Ambulatory follow‐up among publicly insured children discharged from the emergency department

While children discharged from the emergency department (ED) are frequently advised to follow up with ambulatory care providers, the extent to which this occurs is unknown. We sought to characterize the proportion of publicly insured children who have an ambulatory visit following ED discharge, identify factors associated with ambulatory follow‐up, and evaluate the association of ambulatory follow‐up with subsequent hospital‐based health care utilization.


INTRODUC TI ON
The emergency department (ED) serves a critical role in the provision of acute care to sick and injured children. Guidelines from the Agency for Healthcare Research and Quality emphasize the importance of a high-quality ED discharge process, which includes coordinating care within the health care system. 1 As such, children discharged from the ED are frequently advised to follow-up with an ambulatory (primary or subspeciality) provider to promote continuity of care. 2 Ambulatory visits may provide an opportunity to ensure necessary additional evaluation, 3 provide patient and parent education, reinforce the primary care provider as a child's medical home, and potentially provide definitive management. In some contexts, such as for patients following surgery, ambulatory follow-up may reduce avoidable readmissions and ED encounters. 4 The benefits of routine post-ED care have not been quantified and their cost to the health care system are unknown.
A detailed evaluation of patterns of ambulatory follow-up and subsequent health care utilization can inform future research evaluating care for children discharged from the ED to ambulatory care and provide a baseline for assessing care patterns following an ED encounter. Proportions of follow-up in all-payer studies have generally reported higher proportions of children having ambulatory follow-up, emphasizing the importance of payer status on this outcome. A study of children discharged from EDs at three children's hospitals in Michigan reported that nearly 70% had a subsequent follow-up visit, though 14-day follow-up were lower among children enrolled in Medicaid. 5 Another study, evaluating data from a university hospital, similarly reported a relatively higher proportion with follow-up (60%), though again, with a lower proportion among those publicly insured (38.5%). 2 These findings underscore the need for a more thorough evaluation of outpatient follow-up among children using a robust data source, particularly for publicly insured patients.
In this study, we evaluated follow-up ambulatory visits within 7 days of an ED encounter. We also sought to identify factors associated with follow-up visits resulting in care escalation to the hospital. We hypothesized that patients with outpatient follow-up would have greater subsequent hospital-based health care utilization.

Data source
We performed a cross-sectional study of pediatric (<18 years) encounters to the ED within the IBM Watson Medicaid MarketScan data set (IBM Corp.) from the year 2019. During this year, Medicaid was the primary insurance type for 37.5% of children in the United States. 6 The database contains medical claims across the care continuum (hospital, pharmacy, and ambulatory care) for 6,026,754 Medicaid enrollees ages 0 to <18 years from seven anonymized states that represent all U.S. geographical regions. This study was approved by the Ann and Robert H. Lurie Children's Hospital Institutional Review Board. seven U.S. states. Our primary outcome was an ambulatory follow-up visit within 7 days of ED discharge. Secondary outcomes were 7-day ED return visits and hospitalizations. Logistic regression and Cox proportional hazards were used for multivariable modeling.

Results:
We included 1,408,406 index ED encounters (median age 5 years, IQR 2-10 years), for which a 7-day ambulatory visit occurred in 280,602 (19.9%). Conditions with the highest proportion of 7-day ambulatory follow-up included seizures (36.4%); allergic, immunologic, and rheumatologic diseases (24.6%); other gastrointestinal diseases (24.5%); and fever (24.1%). Ambulatory follow-up was associated with younger age, Hispanic ethnicity, weekend ED discharge, ambulatory encounters prior to the ED visit, and diagnostic testing performed during the ED encounter. Ambulatory follow-up was inversely associated with Black race and ambulatory care-sensitive or complex chronic conditions. In Cox models, ambulatory follow-up was associated with a higher hazard ratio (HR) of subsequent ED return (HR range 1.32-1.65) visit and hospitalization (HR range 3.10-4.03).

Conclusions:
One-fifth of children discharged from the ED have an ambulatory visit within 7 days, which varied by patient characteristics and diagnoses. Children with ambulatory follow-up have a greater subsequent health care utilization, including subsequent ED visit and/or hospitalization. These findings identify the need to further research the role and costs associated with routine post-ED visit follow-up.

Eligibility criteria
We included ED encounters for children discharged from the ED.
For multiple encounters clustered within a 7-day period, we included the first encounter as the index visit and all subsequent encounters were considered secondary outcomes. We additionally excluded ED encounters resulting in hospitalization (including those defined as observation status 7 ) or transfer to another institution and those for whom no index visit discharge diagnosis code was provided. We excluded transferred patients because of a perceived need for a higher level of care and because they may have received testing prior to transfer that would not be available in the data set. We excluded discharges without a diagnosis given concerns for data integrity and because we sought to examine the association of discharge diagnosis with outpatient follow-up. Patients with mental health, trauma, and dental primary diagnosis as determined by the diagnosis and grouping system (DGS) categorization were excluded because of the unique concerns in these subgroups. 8 Extensive prior work has characterized the unique challenges in obtaining adequate ambulatory care for children with mental health concerns. 9,10 Similarly, children with traumatic or dental diagnoses have follow-up considerations that we considered to be outside the scope of this study. [11][12][13][14][15][16] Outcome Our primary outcome was an ambulatory follow-up visit, with a primary care practitioner or a subspecialist, within 7 days of ED discharge. Secondary outcomes were ED return visits and/or hospitalizations within 7 days of ED discharge. We used 7 days as our return visit window on the basis of literature on ED return visits [17][18][19] and established definitions from government quality programs. 20 We identified ambulatory follow-up visits using billed procedure codes for each encounter (Table S1).

Data acquisition
For each index ED encounter, we identified demographic, utilization, and diagnosis data. Demographics included age, sex, race and ethnicity, complex chronic condition (CCC), and diagnosis. 21 Age was categorized into <1, 1 to <2, 2 to <6, 6 to <12, and 12 to <18 years.
We reported race and ethnicity as provided in the data set in categories of White, Black, Hispanic, other, and missing. We included the social constructs of race and ethnicity as identified in administrative data because of recognized disparities in access to ambulatory care. 22 Index visits with the date of discharge being Saturday or Sunday were classified as weekend visits. We identified whether the index encounter was immediately preceded by an ambulatory visit (i.e., a visit which occurred on the day prior to or the day of the ED encounter) and the number of ambulatory visits during the year 12-month calendar period preceding the ED encounter. For this, we used the MarketScan dataset from 2018 to identify potential prior utilization. These were grouped as zero, one, two to three, four to six, and seven or more visits. The number of diagnostic tests performed at the index visit were grouped as individually billed laboratory tests (zero, one, two, three or four, and five or more tests) and imaging tests (zero, one or two, and three or more studies) for descriptive analyses, but retained as a continuous measure in modeling.
Diagnosis and procedure codes were defined using International Classification of Disease, 10th Revision, Clinical Modification (ICD-10) codes. A participant was classified as having a CCC if one was recorded either at the index visit or during any encounter during the preceding year using ICD-10 diagnosis and procedure codes. CCCs are used to identify diagnoses which are expected to last 12 months or cause hospitalization in a tertiary care hospital. 21 We used diagnosis codes to identify encounters related to an ambulatory care-sensitive condition (ACSC), defined by the Agency of Healthcare Research and Quality as conditions that may avoid hospitalization with timely and effective ambulatory management, such as asthma, bacterial pneumonia, gastroenteritis, dehydration, and urinary tract infection. [23][24][25] Given that ED visits may contain multiple claims and primary diagnoses, we followed a tiered approach to determine the ultimate primary diagnosis in the following order: physician placed diagnosis codes, facility codes, and codes not associated with either a facility or physician. We categorized primary ICD-10 codes using the diagnosis subgroup from the DGS that was developed specifically for pediatric ED encounters. 8 The DGS is a consensus-derived classification structure that uses the ICD diagnosis codes to group ED visits into clinically sensible and comprehensive categories. It is composed of 21 major groups containing 77 subgroups, and includes a more granular list of diagnoses. 26 We combined the DGS subgroups of "not categorized," "not considered" into the "other" group. We additionally reclassified DGS subgroups occurring in fewer than 10,000 encounters into this other group (with this cutoff representing <1% of ED discharge diagnoses.)

Data analysis
We described encounters with respect to demographics and diagnoses, stratified by the presence of an ambulatory follow-up within 7 days of the index ED encounter, and assessed differences with chisquare tests. Each visit was categorized based on the subsequent course in the 7-day period following ED-discharge: (1) no further visits, (2) ambulatory visit without further hospital-based care, (3) ambulatory visit followed by an ED encounter without hospitalization, (4) ambulatory visit followed by a hospitalization, (5) ED visit without preceding ambulatory visit, and (6) hospitalization without preceding ambulatory visit.
To identify factors associated with an ambulatory visit, we constructed a multivariable logistic regression model for an outcome of ambulatory encounter within 7 days of the index ED discharge.
Predictors included sociodemographic characteristics (i.e., age, sex, race and ethnicity, CCC), diagnosis data (i.e., DGS subgroup and ASCS), weekend visit, ED encounter investigations (i.e., laboratory and imaging, retained as continuous variable), and prior ambulatory physician utilization (i.e., number of visits in the past year and ambulatory encounter immediately prior to the ED encounter). We expressed these results as adjusted odds ratios (aORs) with 95% confidence intervals (CIs). The model was tested for assumptions of independence of observation, linearity of continuous independent variables, and multicollinearity.
To evaluate the association of ambulatory follow-up with our secondary outcomes, we constructed two multivariable Cox proportional hazards models with ambulatory follow-up as a timedependent variable. In all Cox models, the absence of the outcome within 7 days of the index visit was considered as right-censored data. Results were presented as adjusted hazards ratios (aHRs) with 95% CIs. In these models, an HR was associated with a higher risk of the model outcome. As we hypothesized that follow-up visits on the same day as ED return visits or hospitalizations would have a different effect compared to those occurring at an earlier time, we stratified the time-dependent variable to produce a HR for (1) the day of the return visit and (2) all other days in between the index visit and the ED return visit. As a sensitivity analysis, we repeated our analysis for our primary and secondary outcomes when excluding index ED encounters having a discharge diagnosis of an ACSC. Analysis was performed using SAS 9.4 (SAS Institute, Inc.).  Table 1. Approximately one-third of encounters were for an ACSC, and 10% had a CCC. Most (83.9%) had seen an ambulatory physician at least once in the year prior to their index ED encounter.

Health care utilization following index ED encounter
Overall, 280,602 (19.9%) met our primary outcome of having a follow-up ambulatory encounter within 7 days of the index ED discharge. Greater than half of ambulatory encounters occurred within 3 days of the index ED discharge (Figure 1). Twenty-four of the 77 DGS subgroups contained a primary ED discharge diagnosis accounting for greater than 10,000 ED encounters ( Table 2). Most encounters (n = 1,047,729, 74.4%) had no further health care utilization in the 7-day period following ED discharge; 245,063 (17.4%) had an ambulatory follow-up visit without further hospital-based care, 14,390 (1.0%) had an ambulatory visit followed by an ED visit without hospitalization, 2130 (0.2%) had an ambulatory visit followed by hospitalization, 92,129 (6.5%) had a repeat ED visit without an intervening ambulatory visit, and 4965 (0.4%) had a hospitalization without an intervening ambulatory visit. Encounter characteristics within each subgroup are presented in Table S2.

Factors associated with ambulatory follow-up
In multivariable modeling, the following demographic factors were associated with an ambulatory encounter within 7 days of ED discharge: younger age (with the highest aOR among children <1 year relative to adolescents), male sex, race (lower in Black patients and those with missing race/ethnicity and higher in Hispanic and children of other race and ethnicity), and weekend encounters ( Table 3). ED visits associated with an ACSC had a lower of ambulatory follow-up.
Increasing ambulatory encounters in the prior year and an ambulatory encounter immediately prior to the ED encounter were associated with higher odds of an ambulatory visit following ED discharge.
Additional laboratory and imaging were positively associated with an ambulatory follow-up. ED visits for seizures; asthma; and allergic, immunologic, and rheumatologic disease had greater ambulatory follow-up relative to infectious nose and sinus disorders.

Association of ambulatory follow-up on subsequent ED care
An ambulatory encounter was associated with a higher adjusted aHR of subsequent ED use and subsequent hospitalization (both among all included patients and when limited to the subset of patients with a return ED visit; Table 4). All aHRs were higher among ambulatory visits that occurred the day of the repeat ED encounter. The association of ambulatory visits with these outcomes were similar when limited to specific DGS subgroups, apart from seizures, where ambulatory follow-up was not associated with any specific outcome (Table S3).

Sensitivity analysis
A sensitivity analysis limited to encounters without an ACSC diagnosis on their index visit demonstrated similar findings to our primary outcome (Table S4) and a mild attenuation of HRs for our secondary outcomes (Table S5).

DISCUSS ION
In this large cross-sectional data set of publicly insured children, approximately 20% of children discharged from an ED had an ambulatory visit within 7 days, with associations noted by age, race, and ethnicity, preceding ambulatory care utilization, clinical testing, and diagnosis. Outpatient follow-up varied by patient characteristics and ED diagnosis and was associated with an increase in subsequent ED return visits and hospitalizations. Given the unknown cost and benefit of ambulatory follow-up after ED visit, our findings provide an important foundation for beginning to understand barriers to and potential benefits of ambulatory follow-up.
While ED providers often recommend that children receive follow-up care following ED discharge 2 we found that only one-fifth of Medicaid-insured children followed up with an ambulatory physician. Prior work has demonstrated higher proportions having ambulatory overall follow-up among children following an ED encounter (60%-70%) compared to our study. However, these studies have found that follow-up was lower among publicly insured patients (48%-68%) compared to those commercially insured (77%-78%). 2,5 TA B L E 1 Children discharged from the ED stratified by ambulatory follow-up within 7 days of ED visit. Other studies have demonstrated differences in the ability to secure follow-up appointments in ambulatory clinics on the basis of payer status. [27][28][29] Taken together, these findings support previously reported disparities in access to outpatient care adversely affecting publicly insured patients. This may be due to lower reimbursement for publicly insured patients relative to commercial insurers. 30 Furthermore, children who are uninsured and/or publicly insured may preferentially choose to seek care in the ED due to difficulties in securing appropriate outpatient care for lower-acuity conditions. 31 One survey-based study noted that families of children with lowacuity ED conditions elected to use the ED because of concerns in securing a timely primary care appointment. 32 In some of these situations, the subsequent need for outpatient care may be low. The lower odds of follow-up for some of these patients may represent a missed opportunity to reinforce the primary care physician office as a medical home, provide education to families, and ensure adequate management of chronic conditions.
We note important associations between race and ethnicity and ambulatory follow-up, which are potentially due to differences in health care access and quality. 22 Our findings with respect to race from this Medicaid data set are comparable to a Medicare study among adults, which identified that Black race, treatment in a rural hospital, and female sex were associated with less ambulatory follow-up after ED discharge. 33 37 These results are similar to prior work studying TA B L E 2 Ambulatory encounters following index ED visit among the most common diagnoses by DGS subgroups from the index ED visit (limited to diagnosis groups occurring in at least 10,000 encounters). children with asthma discharged from the ED, which report that approximately one-third of children had a follow-up within a month of their encounter. 38,39 An ambulatory visit following discharge from the ED was asso-

LI M ITATI O N S
Our findings are subject to limitations. This was a retrospective crosssectional study of insurance claims data from seven anonymized states within the United States, which provides a large sample size.
However, the specific states studied may have different care patterns including in specific eligibility requirements for Medicaid. The underlying demographics of these states may also differ from others, potentially limiting generalizability. Our findings may not be generalizable to those with private insurance or self-pay. We were unable to evaluate our findings by state as these data are not provided within the MarketScan database. Additionally, the MarketScan provides a means to identify care utilization patterns across publicly insured patients, and it may be subject to inaccuracies in coding and data accuracy; furthermore, our findings may be subject to confounding by indication based on available data. This data source does not contain sufficient data to account for disease severity. Our classification of race and ethnicity was limited by availability in the data set, which does not code these into mutually exclusive groups. We were unable to definitively identify the specialty of the ambulatory provider using this data set and we could not ascertain the reasons why patients do not follow-up with an ambulatory provider (e.g., not advised to seek follow-up, inability to schedule an appointment, or unable or unwilling to miss work). We were unable to capture unbilled telemedicine or telephone follow-ups, which may particularly be the sole point of post-ED contact for children with lower acuity F I G U R E 1 Percentage of ambulatory encounters occurring following ED discharge. Dashed line indicates the cumulative percentage of encounters with ambulatory follow-up over the study period.
illness; this in turn may account for patients with a billed outpatient encounter having a higher risk of subsequent ED utilization or hospitalization. However, we suspect that this occurred less frequently prior to the onset of the COVID-19 pandemic. Our variable relating to prior utilization of ambulatory physicians over the prior year are limited for infants, who could not have 12 months of preceding data.
A 7-day window was used to identify potential follow-up encounters related to the ED visit, recognizing follow-up encounters related to an ED visit may occur both before or after this time. 51 Finally, this study was performed using data prior to the onset of the COVID-19 pandemic, which has resulted in differences in care-seeking pattern (including an increasing utilization of telemedicine).

CON CLUS IONS
Approximately one-fifth of publicly insured children who are discharged from the ED have a billed encounter with an outpatient physician within 1 week. Younger age, race and ethnicity, and specific diagnoses, such as seizures and asthma, were associated ambulatory follow-up. Contrary to our preconceived impression, we found that ambulatory visits were associated with ED revisits and hospitalizations. These findings highlight the importance of further evaluating the role of routine outpatient follow-up after an ED visit in children.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare no conflicts of interest.

Sriram Ramgopal
https://orcid.org/0000-0002-1389-5726 Pradip P. Chaudhari https://orcid.org/0000-0002-7732-632X Stephen B. Freedman https://orcid.org/0000-0003-2319-6192 TA B L E 4 Cox proportional-hazard models to evaluate the role of ambulatory follow-up on subsequent ED and inpatient health care utilization. Note: Data are reported as HR (95% CI) The time-dependent variable for ambulatory visit is stratified produce HR for the day of the return visit and days following the return visit. Models adjusted for age, sex, race/ethnicity, presence of a CCC or ACSC, weekend discharge, ambulatory provider follow-up immediately prior to index visit, weekend discharge of ED encounter, number of ambulatory provider encounters within a year prior to the ED encounter, number of imaging and laboratory tests performed, and diagnosis subgroup.