Acute care utilization for ambulatory care sensitive conditions among publicly insured children

Abstract

Background:

Although characteristics of preventable hospitalizations for ambulatory care sensitive conditions (ACSCs) have been described, less is known about patterns of emergency and other acute care utilization for ACSCs among children who are not hospitalized. We sought to describe patterns of utilization for ACSCs according to the initial site of care and to determine characteristics associated with seeking initial care in an acute care setting rather than in an office. A better understanding of the sequence of healthcare utilization for ACSCs may inform efforts to shift care for these common conditions to the medical home.

Methods:

We performed a retrospective analysis of pediatric encounters for ACSCs between 2017 and 2019 using data from the IBM Watson MarketScan Medicaid database. The database includes insurance claims for Medicaid-insured children in 10 anonymized states. We assessed the initial site of care for ACSC encounters, which were defined as either acute care settings (emergency or urgent care) or office-based settings. We used generalized estimating equations clustered on patient to identify associations between encounter characteristics and the initial site of care.

Results:

Among 7,128,515 encounters for ACSCs, acute care settings were the initial site of care in 27.9%. Diagnoses with the greatest proportion of episodes presenting to acute care settings were urinary tract infection (52.0% of episodes) and pneumonia (44.6%). Encounters on the weekend (aOR: 6.30, 95% CI: 6.27–6.34 compared with weekday) and among children with capitated insurance (aOR: 1.55, 95% CI: 1.54–1.56 compared with fee-for-service) were associated with increased odds of seeking care first in an acute care setting.

Conclusions:

Acute care settings are the initial sites of care for more than 1 in 4 encounters for ACSCs among publicly insured children. Expanded access to primary care on weekends may shift care for ACSCs to the medical home.

Introduction

Ambulatory care sensitive conditions (ACSCs) are conditions for which timely outpatient care may prevent hospitalizations and adverse health outcomes.^1–3 The American Academy of Pediatrics affirms that the optimal setting for children to receive care for ACSCs and other nonemergent acute health concerns is in the medical home.⁴ However, ACSCs are diagnosed in up to half of visits to pediatric emergency departments (EDs) and urgent care centers, with the vast majority discharged home.^5,6 This finding suggests potentially inefficient utilization of acute care settings (emergency departments and urgent care centers) for conditions that could be managed in primary care.

Although prior studies have described characteristics of preventable hospitalizations for ACSCs,^1,2,7–9 less is known about patterns of acute care utilization for ACSCs among children who are not hospitalized. Studies of children with ACSCs and other low-acuity conditions have generally emphasized overall utilization of acute care services without exploring the sequence of care utilization that may lead to an encounter outside the medical home.^10–13 A better understanding of where children first obtain care for ACSCs—along with patient- and visit-level characteristics associated with initial care-seeking in acute care settings rather than in primary care—will provide greater clarity on the potentially avoidable instances of acute care utilization for ACSCs. This understanding may also inform interventions to shift the initial care for ACSCs to the medical home.

Our objectives were to describe patterns of care-seeking for ACSCs according to the initial site at which children seek care and to determine characteristics associated with seeking care in an acute care setting rather than in an office setting.

Methods

Study Design.

We performed a retrospective cohort analysis of encounters by children <18 years of age using the IBM Watson MarketScan Medicaid Claims and Encounters database (IBM Corp.) between 2017 and 2019. The database contains insurance claims from hospitals, pharmacies, and ambulatory care settings for more than 6 million enrollees in Medicaid from 7–10 anonymized states (depending on the year) distributed across all United States Census regions.¹⁴ Because data were de-identified, this study was deemed to be non-human subjects research by the Institutional Review Board of the Ann & Robert H. Lurie Children’s Hospital of Chicago.

Study Setting and Population.

We included all ambulatory care encounters for ACSCs that occurred in primary and subspecialty care offices, urgent care centers, and EDs among children with continuous enrollment during the study period. Ambulatory care encounters were defined according to Centers for Medicare and Medicaid Services place of service codes and Current Procedural Terminology codes (Supplemental Table 1). ACSCs included both acute conditions (gastroenteritis, bronchiolitis, cellulitis, urinary tract infection, pneumonia, and upper respiratory tract infection) and chronic conditions (asthma, diabetes mellitus, and seizure). ACSCs were defined based on previously published definitions using International Classification of Diseases, Tenth Revision, Clinical Modification, (ICD-10-CM) codes.^1,2,9,15

We defined an illness episode to be a 14-day period starting with a first encounter for an ACSC, during which additional visits may have occurred for some patients.¹⁶ We excluded illness episodes that resulted in hospitalization at any point during the illness episode, as these were more likely reflective of greater disease severity and therefore may not have been amenable to office-based management. Upon examination of the data, it was noted that a small proportion (<1%) of patients had highly frequent (e.g., near-daily) healthcare utilization for ACSCs. Given our limited ability to distinguish frequent healthcare utilization for acute illnesses related to an ACSC from frequent coding of ACSCs (e.g., diabetes or asthma) at health maintenance visits for non-acute illnesses in children with specialized healthcare needs, we excluded children with >15 ambulatory care episodes for ACSCs during the study period. The choice of >15 episodes was made empirically upon examination of the data and corresponded to 5 episodes per patient per year, or approximately 1% of illness episodes.

Outcomes and Measures.

Our primary outcome was the site of initial care during the illness episode, which we defined as either an acute care setting or an office setting. Acute care settings included EDs and urgent care centers. All other included settings were classified as office settings. We elected to combine urgent care centers with EDs, as these settings typically exist outside of the medical home and may have additional testing and treatment capabilities when compared with office settings.⁴ If acute care and office encounters occurred on the same day, we assumed that the office encounter occurred first (e.g., a patient presented to an office setting and subsequently visited the ED).

Data Analysis.

We described the proportion of illness episodes in which patients sought care in acute care settings, first overall and then stratified by diagnosis. To identify associations between visit characteristics and the initial site of care, we used generalized estimating equations clustered on patient. We chose to include variables in the models a priori based on an assessment of their potential to be associated with the outcome and based on prior studies of characteristics associated with ambulatory care utilization.^12,17,18 These characteristics included visit year, age, sex, race and ethnicity, whether the visit occurred on a weekend, insurance plan (capitated vs. fee for service), presence of a complex chronic condition,¹⁹ the specific ACSC diagnosis, and overall ambulatory care utilization during the study period (i.e., the number of encounters for ACSCs and non-ACSC diagnoses during the study period). We included the social constructs of race and ethnicity based on previously described differences in both access to primary care and utilization of acute care services according to these constructs.^12,20 Whereas in fee-for-service plans clinicians are reimbursed for specific procedures, tests, and treatments, in capitated plans the clinician is reimbursed a fixed amount per patient. Complex chronic conditions were defined using ICD codes assigned within the year prior to the index visit according to the scheme developed by Feudtner et al, which defined complex chronic conditions as conditions expected to last at least 12 months and to likely require hospitalization in a tertiary care center.¹⁹

To assess for differences in the associations between visit characteristics and the primary outcome, we performed separate models for each of the nine ACSC diagnoses. We additionally performed 2 sensitivity analyses. First, we performed the regression analysis after excluding chronic ACSCs (asthma, diabetes, seizure), as it is possible that some of the ICD-10 codes assigned for these conditions could have reflected the presence of a chronic diagnosis rather than an acute problem. Second, we excluded upper respiratory tract infections, as these conditions comprised a large proportion of the encounters and therefore had the potential to disproportionately impact the results of the regression analyses.

Given that (1) the conditions studied are common and (2) the database reflects healthcare utilization among a very large population of publicly insured individuals in the included states, we did not perform an a priori sample size calculation. Analyses were performed using SAS 9.4 (SAS Institute, Inc.). We followed STROBE reporting guidelines.²¹

Results

There were 7,128,515 ACSC illness episodes among 2,011,339 unique patients. The median age was 8 years (interquartile range, 4–13), and the median number of illness episodes per patient was 3 (interquartile range, 1–5; Table 1). Complex chronic conditions were present in 13.8% of patients, and 64.0% had capitated (rather than fee-for-service) insurance. The most common diagnoses were upper respiratory tract infection (74.1% of episodes), asthma (9.5%), gastroenteritis (5.1%), and cellulitis (3.4%). Initial visits occurred on the weekend in 11.9% of illness episodes.

Table 1.

Characteristics of the cohort

Characteristic	N (%)
Patient-level characteristics
Number of children	2,011,339
Age (years) at start of study period [median, IQR]	8 [4, 13]
Female sex	1,023,829 (50.9)
Race and ethnicity
Non-Hispanic White	1,009,614 (50.2)
Non-Hispanic Black	604,561 (30.1)
Hispanic	164,985 (8.2)
Other	73,325 (3.6)
Missing	158,854 (7.9)
Insurance
Fee for service	724,539 (36.0)
Capitated	1,286,800 (64.0)
Complex chronic condition	277,432 (13.8)
Number of ACSC episodes during the study period
1	582,878 (29.0)
2–4	881,337 (43.8)
5–9	436,191 (21.7)
10–15	110,933 (5.5)
Number of office visits for non-ACSC diagnoses (median, IQR)	11 [6, 19]
Episode-level characteristics
Number of episodes	7,128,515
Year
2017	2,654,590 (37.2)
2018	2,310,428 (32.4)
2019	2,163,497 (30.3)
Weekend visit	851,038 (11.9)
Diagnosis
Asthma	678,285 (9.5)
Bronchiolitis	29,575 (0.4)
Cellulitis	245,369 (3.4)
Diabetes	48,018 (0.7)
Gastro	362,168 (5.1)
Pneumonia	90,389 (1.3)
Seizure	181,630 (2.5)
Upper respiratory infection	5,280,973 (74.1)
Urinary tract infection	212,108 (3.0)

Site of first care.

The site of first care was an acute care setting in 27.9% of episodes and an office setting in 72.1% of episodes. Among episodes that initially presented in acute care settings, EDs comprised 64.2%. The diagnoses with the greatest proportion of episodes presenting to acute care settings were urinary tract infection (52.0%) and pneumonia (44.6%), whereas those with the smallest proportion were diabetes (12.8%) and asthma (17.6%) (Figure 1).

Figure 1.

Proportion of episodes with site of initial care in an acute care setting

URI, Upper respiratory infection; UTI, urinary tract infection

Characteristics associated with the site of first care.

The characteristic most strongly associated with an acute (rather than office-based) site of initial care was occurrence of the encounter on a weekend (adjusted odds ratio [aOR]: 6.30, 95% CI: 6.27–6.34; Table 2). Compared with non-Hispanic white race and ethnicity, non-Hispanic black race and ethnicity was associated with increased odds of an initial acute site of care (aOR: 1.89, 95% CI: 1.88–1.90). Compared with fee-for-service insurance, capitated insurance was associated with increased odds of an initial acute site of care (aOR: 1.55, 95% CI: 1.54–1.56).

Table 2.

Characteristics associated with first seeking care in an acute care setting

Characteristic	Adjusted odds ratio (95% CI)
Year
2017	REFERENCE
2018	0.98 (0.98, 0.99)
2019	0.99 (0.99, 1.00)
Age (per year)	1.02 (1.01, 1.02)
Female sex	1.04 (1.04, 1.05)
Weekend visit	6.30 (6.27, 6.34)
Race and ethnicity
Non-Hispanic White	REFERENCE
Non-Hispanic Black	1.89 (1.88, 1.9)
Hispanic	1.12 (1.1, 1.13)
Other	0.91 (0.90, 0.93)
Missing	1.60 (1.59, 1.62)
Insurance
Fee for service	REFERENCE
Capitated	1.55 (1.54, 1.56)
Complex chronic condition	1.04 (1.03, 1.05)
Each additional ACSC episode during the study period	0.96 (0.96, 0.96)
Each additional office visit for non-ACSC diagnoses during the study period	0.99 (0.99, 0.99)

^*Adjusted for diagnosis

Regression analyses stratified by condition (Figure 2) and excluding chronic ACSCs and URIs (Supplemental Table 2) identified similar associations between visit characteristics and the site of first care.

Figure 2. Characteristics associated with initial care-seeking in an acute care setting, stratified by diagnosis.

Plots reflect results from separate models for each diagnosis. The binary outcome variable in the generalized estimating equations was initial care-seeking in an acute care setting rather than in an office. Predictor variables included in the generalized estimating equations are included on the Y-axis. Points and error bars reflect adjusted odds ratios and 95% confidence intervals.

Discussion

In this large multistate study of ACSC episodes among Medicaid-insured children, we found that more than 1 in 4 sought initial care in acute care settings rather than in office settings. We identified substantial variation in this proportion according to the specific ACSC diagnosis. After adjusting for diagnosis and other visit characteristics, weekend visits had more than 6 times the odds of occurring in an acute care setting compared with weekday visits. Capitated insurance as well as non-Hispanic Black race and ethnicity were also associated with initial care in acute care settings.

Our findings should be interpreted in the context of prior studies of acute care utilization for ACSCs. Previous studies have found that 13–48% of ED and urgent care visits in the United States are for ACSCs and other low-acuity conditions^5,6,18,22—findings that also have been demonstrated in other countries.^23,24. Similarly, a national study showed that among adults with ACSCs evaluated in ambulatory care settings in the United States, 47% are cared for in an ED.²⁵ Compared to ED visits among privately insured children, ED visits by publicly-insured patients are approximately 30% more likely to be for an ACSC.⁵ Together, these studies demonstrate both that a large proportion of pediatric acute care is for ACSCs and that a large proportion of patients with ACSCs receive care in acute care settings, particularly among publicly insured children.

Our study adds to these prior findings by identifying that 28% of children seek initial care in an acute care setting without preceding office-based care. Although some children with ACSCs—for example, those requiring extensive testing, those warranting prolonged observation, or those in whom parents appropriately judged that primary care resources would be insufficient to meet the child’s acute needs²⁶—may be more effectively or efficiently managed in acute care settings without initial office-based care, our finding that more than 1 in 4 initial healthcare encounters for ACSC’s occurs at an acute care setting highlights an opportunity to develop interventions to shift the initial care for these conditions into the primary care setting.

While additional data are needed to determine whether earlier contact with primary care improves clinical outcomes for children with ACSCs, shifting care to the office setting may reduce the costs of care for these common conditions.^27,28 A prior national study demonstrated that office visits for ACSCs were both 2.5 times less expensive and associated with payments that were 2.5 times lower than ED visits for ACSCs.²⁵ Providing early, definitive treatment of ACSCs in primary care clinics may also serve to reduce ED crowding, which is associated with decreased quality of care for ACSCs.^27,28

We found substantial variation in the site of initial care for ACSCs according to diagnosis. In particular, episodes of care for urinary tract infections, pneumonia, and cellulitis were most likely to start in an ED or urgent care center. Potential explanations for these differences include greater illness severity, a more urgent concern about clinical worsening for these bacterial infections (compared with self-limited viral illnesses), fewer testing capabilities in office-based settings compared to acute care settings, or less established relationships with office-based clinicians among children with these conditions compared with those with chronic ACSCs (e.g., asthma, diabetes, or seizures). Additional investigation is needed to explain the reasons for the relatively greater rates of acute care seeking for these specific conditions.

Our findings offer additional insights into some of the drivers of the observed care-seeking behavior. Specifically, we found that ACSC episodes that occurred on the weekend had more than 6 times the odds of starting in an acute care setting than weekday episodes. This strong association was consistent across all diagnoses and suggests that expanded access to primary care services on the weekend—or, in the case that these services do exist, increased awareness by families of these services—has the potential to reduce acute care utilization for ACSCs.^29–34 Of course, given that (1) expanded outpatient services on weekends or evenings would carry financial and logistical costs and (2) there is a shortage of primary care clinicians to staff these clinics, further evaluation of the cost-effectiveness of expanded office services during off hours is needed. Additionally, innovative approaches to delivering primary care after hours—for example, through telehealth^35,36—may help to shift some after-hours acute care for ACSCs to the medical home.

We found evidence of more frequent acute care utilization without a preceding primary care visit among children enrolled in capitated health plans. This is surprising, as the underlying premise of primary care capitation is closer coordination of health care services with the primary care provider. Based on this premise, it would be expected that capitation would result in closer contact with primary care prior to acute care visits, although the evidence around the benefits of capitation on access and utilization are generally mixed. Additionally, states have variable eligibility for capitated plans. Since states are not identifiable in the database, a detailed assessment of these differences was not feasible. Further exploration of this topic using prospective methods is warranted to evaluate the role of coverage type in healthcare utilization in children.

Our findings that non-Hispanic black race and ethnicity was associated with initial presentation to acute care settings is consistent with prior studies of acute care use for ACSCs and other low-acuity conditions.^18,37,38 Of note, we were not able to adjust for income, health literacy, or composite measures of child opportunity, all of which have been associated with care-seeking patterns.^2,5,10,13,32 Since race is a social construct, it is possible that our findings may reflect confounding by some of these other objective measures or may be indications of decreased access to primary care due to systemic racism.^39–41

Limitations

First, given that there are limited clinical data available in the claims, we were not able to fully adjust for illness severity. Although we accounted for this by excluding children who were hospitalized, we were unable to comment on the appropriateness of presenting to a specific setting in individual cases. Second, although administrative codes are widely used to characterize the reasons for acute care visits, it is possible that in some instances the codes did not accurately capture the reasons for the visit. For example, child with prolonged fever warranting laboratory workup might ultimately be coded as having an upper respiratory infection. This also may be particularly relevant for the chronic ACSCs (i.e., asthma, diabetes, seizure), which may be coded at visits for other acute problems. Our sensitivity analysis excluding these conditions showed similar results to the primary analysis, suggesting that our main results are robust to this potential limitation. Third, the states included in the database were de-identified, so we did not know the specifics of the capitated health plans for included patients. Since there is likely heterogeneity in states’ guidelines and requirements around enrollment in capitated plans, our results regarding the associations between enrollment in these health plans and patterns of utilization should be interpreted with caution. Fourth, the office setting may not have represented the medical home for all patients included in the analysis. Fifth, because we were not able to measure the time of day using the claims data and because many primary care centers do not hold regular after-hours clinics,⁴² we could not assess the extent to which acute care settings were utilized at times when primary care offices were closed. This limitation may be particularly relevant for this Medicaid-insured population, for whom parents’ work hours in many cases may not allow for unanticipated daytime visits in primary care. Sixth, we used data predating the COVID-19 pandemic. While our results should be interpreted in the pre-pandemic context, we believe that the ongoing post-pandemic capacity challenges faced by both emergency and primary care make the findings relevant to current practice. Seventh, because the claims do not capture unbilled phone conversations with office-based clinicians, our results may underestimate the proportion of children who had contact with primary care prior to acute care. Finally, our results in Medicaid-insured children may not be generalizable to children with other types of insurance.

Conclusions

Acute care settings are the initial sites of care for more than 1 in 4 episodes for publicly insured children with ACSCs. Additional research that identifies patient-level preferences for visits in acute care settings rather than in office settings has the potential to reduce costs, prevent overcrowding, and improve clinical outcomes for children with ACSCs.