Geographic proximity to primary care providers as a risk-assessment criterion for quality performance measures.

IMPORTANCE: Previous studies have found a mixed association between Patient-Centered Medical Home (PCMH) designation and improvements in primary care quality indicators, including avoidable pediatric emergency department (ED) encounters. Whether these associations persist after accounting for the geographic locations of providers relative to where patients reside is unknown.
OBJECTIVE: To examine the association between geographic proximity to primary care providers versus hospitals and risk of avoidable and potentially avoidable ED visits among children with pre-existing diagnosis of attention-deficit/hyperactivity disorder or asthma.
METHODS: Retrospective cohort study of a panel of pediatric Medicaid claims data from the South Carolina from 2016-2018 for 2,959 beneficiaries having a pre-existing diagnosis of attention-deficit/hyperactivity disorder (ADD, ages 6-12) and 6,390 beneficiaries with asthma (MMA, ages 5-18), as defined using Healthcare Effectiveness Data and Information Set (HEDIS) performance measures. We calculated differences in avoidable and potentially avoidable ED visits by the beneficiary's PCMH attribution type and in relation to differences in proximity to their primary care providers versus hospitals. Outcomes were defined using the New York University Emergency Department Algorithm (NYU-EDA). Differences in ED visit risk were assessed using generalized estimation equations and compared using marginal effects models.
RESULTS: The 2.4 percentage point reduction in risk of avoidable ED visits among children in the ADD cohort who attended a PCMH versus those who did not increased to 3.9 to 7.2 percentage points as relative proximity to primary care providers versus hospitals improved (p < 0.01). Children in the ADD and MMA cohorts that were enrolled in a medical home, but did not attend one for primary care services exhibited a 5.4 and 3.0 percentage point increase in avoidable ED visit compared to children who were unenrolled and did not attend medical homes (p < 0.05), but these differences were only observed when geographic proximity to hospitals was more convenient than primary care providers. Mixed findings were observed for potentially avoidable visits.
CONCLUSIONS: In several health care performance evaluations, patient-centered medical homes have not been found to reduce differences in hospital utilization for conditions that are treatable in primary care settings among children with chronic illnesses. Analytical approaches that also consider geographic proximity to health care services can identify performance benefits of medical homes. Expanding risk-adjustment models to also include geographic data would benefit ongoing quality improvement initiatives.

Introduction

Ample evidence shows that health outcomes across the United States (US) differ by race, ethnicity, and socioeconomic status [1-5]. For instance, emergency department (ED) utilization is higher among the homeless [6], diabetes prevalence is often attributed to greater food insecurity [7], and social isolation contributes to both stroke and heart disease [8], to name but a few of its determinants and consequences. It is also possible to make a reasonably good prediction as to how long it will take people to access care or the type of treatment they will receive based on knowing where they live [9-11]. For example, recent studies have shown that Medicaid beneficiaries who live closer to EDs have a higher probability of using these services for non-emergent reasons than other patient groups [12].

Inappropriate or non-emergent ED utilizations raise costs, strain resources, as well as increase wait times for care [13,14]. They are also fundamentally linked to social determinants of health, such as poor primary care access or low health literacy [15], as well as to care management on the part of patients and primary care providers [16]. Yet, despite national recommendations on the general management of chronic illnesses and the substantial cost and negative implications of non-emergent ED visits, pre-COVID 19 pandemic studies found that the frequency of avoidable visits were increasing, particularly among children [17]. For example, it is estimated that nearly 75% of the 3 million ED visits for asthma-related care and nearly 33% of the 1.6 million ED visits for mental health-related conditions can be treated in primary care settings [18-21]. Medicaid is often the most frequent payer of ED visits nationwide [22].

Over the past decade, patient-centered medical homes (PCMH) have shown potential to limit avoidable ED utilizations and improve preventative and chronic disease management [23,24]. Akin to other patient-based practice structures, such as Canada’s Family Health Teams [25], medical homes are largely based on a physician-led continuity of care model that include nurses, nutritionists, pharmacists, case managers, and social workers with an emphasis on providing patient-and culturally-centered care [26]. In the broadest sense, they represent a transformative whole-person approach to primary care designed to improve physical health, behavioral health, access to community-based social services, as well as better management of chronic illnesses [27-30]. Aspects of medical home design, such as its use of care coordinators to help with medication reconciliation are particularly relevant for children with chronic conditions who have a diverse set of medical needs [31]. For instance, case managers may work directly with children and their guardians to coordinate care, help families navigate the health care system, as well as coordinate appointments with specialists to improve disease control [32].

Although there is growing consensus of the benefit of PCMH-modeled care on health outcomes overall, including lowered ED utilization, its effectiveness in reducing pediatric ED utilizations have been mixed. Among the general pediatric population, evaluations have shown a reduction in ED use as well as improvements in preventative care delivery [33-35], but typically show little effect on usage rates among children with chronic illnesses, even when evaluated using patient-reported quality metrics (e.g., wait times to care, satisfaction with providers) [36-38]. Mixed findings are thought to stem from the amplified effect that care fragmentation and poor communication across resource systems have on children with greater medical needs [39]. It is also possible that a caregiver’s perception that their child is not receiving sufficient care may be a stronger predictor of ED use regardless of the quality of primary services the child receives. However, missing from previous evaluations is whether the interaction between where patients live and where their health care services are located modifies these associations. Although some PCMH evaluation studies have attempted to control for geography (e.g., rural versus urban clinics) [40,41], this approach ignores the spatial connectivity between patients and providers and cannot answer questions as to whether greater travel distances to care determines why the effects of some medical home innovations are often muted.

In this study, we assessed whether geographic proximity to health care providers modified the association between medical home attendance and risk of avoidable and potentially avoidable ED visits among pediatric Medicaid recipients in South Carolina (SC) having a pre-existing diagnosis of asthma or attention-deficit/hyperactivity disorder, each of which are chronic illnesses that often result in ED visits for disease-specific care that is often treatable in primary care settings [42,43]. Each condition was defined using the National Committee for Quality Assurance’s (NCQA) Healthcare Effectiveness Data and Information Set (HEDIS®) process measures, thus reflecting a measurement standard commonly used for risk adjustment and performance tracking by state Medicaid agencies. Although many initial evaluations of the PCMH transformation on outcomes among Medicaid participants have been positive, pointing to the benefit of better compliance measures, case management, and electronic health records with earlier and better access to care [44-47], some studies have failed to show consistent improvements quality [48,49]. Fig 1 provides an example of why spatial interactions in these and other PCMH evaluations might be warranted. It shows that SC counties that have witnessed the greatest penetration of medical homes are simultaneously among the least dense with respect to the % of the population enrolled in Medicaid (areas in bright green). What is also particularly telling is the ‘I-95 corridor’, often referred to as the ‘corridor of shame’ [50], which stretches from Jasper county in the south to Dillon county in the northeast, is among the areas with the poorest access (areas in bright pink). Eight of the 17 counties that form part of the corridor are on the wrong end of the PCMH availability-to-need spectrum, many of which are among the poorest counties in the country.

Fig 1

Bivariate choropleth map of county proportion of medical homes relative to all primary care providers and the proportion of county population enrolled in Medicaid.

Figure created by the authors using the University of South Carolina’s site license for ArcMap. Data to recreate the map is publicly available through the American Community Survey, the NPI Certification files distributed by the Centers for Medicare and Medicaid Services, and the NCQA data feed file (NCQA data available through subscription).

Given what is known about proximity to care as a determinant of service utilization and outcomes, statistical models that include interactions with spatial data representative of where patients reside relative to where they receive primary and emergency care services may help reveal unmeasured benefits or barriers attributed to medical home-modeled care. Assessing the representation of one type of service accessibility measure (i.e., travel distances) is important for ongoing policy decisions for state Medicaid programs, which almost universally provide support, technical assistance, and resources to clinicians, practices, and families to ensure its beneficiaries gain access to designated medial homes [51]. This emphasis is also timely given the ongoing transformations of US risk-adjustment and pay-for-performance models to include social factors on top of the clinical risk factors already in place [52-54]. The NCQA is one of the most recent organizations to propose health plan accountability for social risk factors [55]. The NCQA is also one of the largest accrediting bodies that certifies primary care clinics as designated medical homes.

Materials and methods

Patient population

Approximately one out of every five of the 5.1 million persons residing in SC is a Medicaid beneficiary, of which 62% are 18 years of age or younger [56]. In 2013, the SC Department of Health and Human Services’ (SCDHHS) Medicaid agency began working with its managed care programs to enroll eligible beneficiaries into designated PCMHs. The intention has been to improve access to primary care, particularly care that is considered to be of higher quality. This study used a 3-year retrospective cohort design of pre-COVID 19 pandemic data, beginning with a baseline (reference) year of 2016, which is consistent with other Medicaid program effectiveness evaluations [57]. Medicaid claims were eligible for inclusion if the beneficiary was continuously enrolled for the entire 2016 reporting year, but the patient was not required to be enrolled for the entire study period. Continuous enrollment in South Carolina allows patients to churn on/off Medicaid for no more than 45 days per year.

We used the NCQA’s Healthcare Effectiveness Data and Information Set (HEDIS®) process measures to define a similar patient cohort and as a grouping mechanism to differentiate differences in outcomes achieved from designated and non-designated clinics. HEDIS® measures comprise the core evaluation metrics for pediatric and adult Medicaid programs and relate to numerous public health issues, including cancer, heart disease, mental health, asthma, and diabetes, among others. HEDIS® metrics are used to measure specific aspects of primary care quality, including the effectiveness (e.g., colorectal cancer screening) and experiences of care (e.g., CAHPS surveys), access/availability of care (e.g., access to ambulatory health services), risk adjusted utilizations (e.g., well-child visits), as well as relative resource use (e.g., resource use for people with COPD). In 2016, 37 state Medicaid programs collected or required its Managed Care plans to report on over 60 HEDIS® measures as part of annual external quality reviews [58]. The use of HEDIS® metrics as a grouping mechanism for PCMH evaluations is also common practice when attempting to differentiate outcomes by provider settings [59,60].

Our patient sample included all Medicaid beneficiaries that met the inclusion criteria for the HEDIS® measure of follow-up care for attention-deficit/hyperactivity disorder (ADD) or medication management for people with asthma (MMA). The ADD measure is specific to recipients between 6 and 12 years of age who were newly prescribed ADHD medication. The MMA measure captures child recipients between the ages of 5 and 18. For this analysis, we collapsed the MMA age ranges into a single group. These measures were chosen because they represent conditions that result in frequent presentation in hospital emergency departments for disease-specific care that is often treatable in primary care settings [42,43]. Both measures were also chosen because they have a strong association with social vulnerability and residing in economically deprived areas and may be part of expanded risk factor adjustment metrics [55]. The study protocol (#Pro00093322) was reviewed and approved by the University’s Institutional Review Board (IRB) in accordance with 45 CFR 46.104(d)(4).

PCMH attribution

SC Medicaid recipients select their primary care provider from five similarly designed Managed Care Organizations (MCO) or through fee-for-service (FFS) [61]. Over 80% of all Medicaid beneficiaries in SC are enrolled in an MCO, the majority of which include clinics that are NCQA-designated medical homes. This is in line national recommendations from the American Academy of Pediatrics (AAP), the National Association of Pediatric Nurse Practitioners (NAPNAP), and other organizations that all children have access to comprehensive, accessible, coordinated, and culturally appropriate care [62,63]. At the time of this study, there were no health (e.g., the number of chronic conditions) or plan-related (enrolled in MCO or FSS) restrictions on SC Medicaid beneficiaries for selecting a primary care provider whose practice was or was not a designated/accredited medical home.

It was not possible to identify from the claims data how each beneficiary initially elected to enroll with their primary care provider. However, it was possible to identify from the claims data whether the recipient’s health plan assigned them to a medical home as well as determine the designation status of the their primary care provider using their National Provider Identifier (NPI) and the NCAQ provider file. As such, we modeled two different depictions of PCMH attribution on pediatric ED visits: one based solely on attendance (i.e., attending a PCMH or not attending a PCMH) and one based on a 2x2 matrix of enrollment and attendance combinations (e.g., enrolled and attended a PCMH, enrolled and never attended a PCMH, etc.). We assessed both attribution types as the first method provides an account of care quality when based solely on patient attendance patterns that are typically available in health care registries, whereas the latter provides a finer depiction of a medical home effect in relation to geography. Beneficiaries that did not attend a PCMH or who were un-enrolled in a PCMH and never attended one for their primary care were used as the reference group.

Emergency department visits

We used the New York University Emergency Department Algorithm (NYU-EDA) to define the level of acuity for all ED visits [64]. The NYU-EDA is a widely used measure of avoidable ED utilization [65-70]. The NYU-EDA classifies each ED visit probabilistically on a percentage basis into one of four categories: “non-emergent,” “emergent, but primary care treatable,” “emergent, but preventable/avoidable,” and “emergent, not preventable/avoidable”. All visits are classified based on the ICD diagnosis codes from the patient’s chart. The algorithm excludes uncommon diagnoses and treats mental health and substance abuse diagnoses separately from other causes. As it is possible to have probability scores assigned across all 4 categories, we used modifications to the event diagnosis groupings validated by Ballard et al (2010) to assign “non-emergent” (i.e., avoidable) and “emergent, but primary care treatable” (i.e., potentially avoidable) events into two distinct group, so long as the visit probability was greater than 0.50 for either event [71]. We excluded all visits that were the result of injury, or alcohol or drug problems since these conditions usually require ED services regardless of severity.

Covariates

We were able to identify a limited set of covariate information from the claims data, including age, sex, race/ethnicity, co-morbidity classification, number of primary care visits, whether the ED visit occurred on a weekend, as well as whether the claim was attributed as fee-for-service (FFS). At the time of this study, the race/ethnicity field was not a mandatory data collection point by the South Carolina Medicaid agency. Due to small numbers and the large number of unknown/missing responses, we included only beneficiaries who reported their race/ethnicity as non-White Hispanic, Black, or Hispanic. Co-morbidities were defined using 3MTM clinical risk group classification codes (CRG) [72]. CRGs are a population classification system derived from inpatient and ambulatory diagnosis and procedure codes, pharmaceutical data, as well as patient functional health status. CRG codes categorize patients into one of nine severity-adjusted groups, increasing in scale from heathy/non-users to catastrophic condition status. Through data linkages, we were able to obtain additional information pertaining to the patient’s neighborhood dwelling area type using modifications of the Rural Urban Commuting Area (RUCA) codes, the median household income of their residential census tract, as well as the proportion of medical homes relative to all primary care providers within their county of residence. RUCA codes were fixed using 2010 estimates whereas income and provider ratios were calculated yearly using American Community Survey and NCQA reports.

Geographic assessment

To calculate geographic distances between beneficiaries and providers we built a composite geocoding algorithm that allowed for situs, linear, and area referencing of Medicaid claims data using a previously validated approach [73]. Prior to geocoding, we standardized each address file to US Postal Service mailing format to increase the likelihood of matching the provider address information to the street centerline file. Standardization was done using ZP4 address correction software. Next, we linked address information (e.g., street name, street suffix, ZIP code) from the claims database to street centerline data using ESRI’s Street Map Premium Address File. Average travel distance estimates (in miles) for all visits were then constructed for all unique pairings between the beneficiary’s residential address recorded in the claims database to the location where their primary care and ED visit occurred. The distance metric used in the regression model was the difference in distances between a beneficiary and their primary care provider versus the ED where the visit occurred. Negative values represented more convenient access to primary care providers, whereas positive values represented better proximity to hospitals. For ease of interpretation, relative differences in access were grouped using +/- 15 mile increments, with values of 0 representing no difference in distance between either location relative to the beneficiary’s residential address.

Statistical analysis

In the unadjusted analysis, differences between means of continuous variables across PCMH attribution classifications were compared using ANOVA. Differences in proportions of categorical variables were examined using chi square tests. In the adjusted models, we measured the association between PCMH attribution and ED visits in two ways. First, we constructed a-spatial regression models to assess adjusted differences in event probabilities based on a patient’s PCMH attribution group type, with one model comparing ED visits based solely on PCMH attendance (e.g., attended a PCMH versus non-PCMH attendees) and another comparing visits based on the enrollment and attendance matrix (e.g., enrolled and attended a PCMH, enrolled and never attended a PCMH, etc.). Second, we expanded each model to account for the interaction with travel distance to providers. The spatial interaction models included all previously adjusted covariates.

We used a logistic regression analysis and marginal effects classifications to measure and compare the association between PCMH attribution type and the probability of an avoidable or potentially avoidable ED visit. Since beneficiaries may have had multiple ED visits during the study period as well as may not have been enrolled for the entire 3-year study period, we used a logistic model within a GEE to examine the probability of events while explicitly modeling the within-subject correlation. We initially selected a model using an independent correlation structure due to the small clusters and imbalanced design as well as the lack of complete confounder adjustment from the claims database. We ran quasi likelihood criterion tests to confirm this correlation structure was the best for model fit. All covariates that were statistically significant (p < 0.10) in bivariate analysis for each cohort were included in the regression models as a potential influence on ED visit risk. A full list of the covariates and regression coefficients from the models are provided in the online supplementary appendix. All statistical analyses were completed in STATA for Windows.

Results

The final ADD patient sample consisted of 4,222 person-years from 2,959 unique non-Hispanic White, Black, and Hispanic Medicaid beneficiaries. From a total of 6,059 ED visits among the ADD cohort, 1,355 (22.3%) were classified as being non-emergent (i.e., avoidable) and 1,026 (16.9%) were classified as being emergent but primary care treatable (i.e., potentially avoidable), for an avoidable and potentially avoidable ED visit rate of 32.1 and 24.3 visits per 100 person-years, respectively. The final MMA patient sample consisted of 10,998 person-years from 6,390 unique non-Hispanic White, Black, and Hispanic Medicaid beneficiaries. There was a total of 20,462 ED visits, of which 4,205 (20.6%) and 3,515 (17.2%) were classified as avoidable or potentially avoidable, for a person-year rate of 38.2 and 31.9 avoidable and potentially avoidable ED visits per 100 person-years, respectively.

Tables 1 and 2 show that across all patients and PCMH attribution groups, the cohorts were dissimilar with respect to age, race/ethnicity, number of primary care provider visits, RUCA dwelling type, and county provider ratios (p < 0.05). The travel distances to primary care providers and hospitals were dissimilar among the MMA cohort for both PCMH attribution types, with medical home patients residing closer to all services irrespective of how enrollment and attendance groupings were defined (p < 0.01). There were no statistically significant differences in average travel differences to any service provider for the ADD cohort in the unadjusted comparisons.

Table 1

Person-years summary statistics for ADD outcomes.

	ADD cohort
	PCMH attribution based on attendance			PCMH attribution based on enrollment and attendance
Characteristic	PCMH(n = 1,140)	Non-PCMH(n = 1,519)	p	Group 1(n = 1,320)	Group 22(n = 924)	Group 3(n = 199)	Group 4(n = 516)	p value
Age (SD)	9.7 (1.7)	9.8 (1.8)	0.038	9.8 (1.8)	9.6 (1.7)	10.1 (1.7)	9.8 (1.7)	0.012
Sex (Male)	953 (66.2)	999 (65.8)	0.812	874 (66.2)	620 (67.1)	125 (62.8)	333 (64.5)	0.590
Race			0.000					0.000
non-Hispanic White	573 (39.8)	721 (47.5)		645 (48.9)	376 (40.7)	76 (38.2)	197 (38.2)
Black	820 (56.9)	747 (49.2)		631 (47.8)	524 (56.7)	116 (58.3)	296 (57.4)
Hispanic	47 (3.3)	51 (3.4)		44 (3.3)	24 (2.6)	7 (3.5)	23 (4.5)
Clinical Risk Group			0.547					0.168
Healthy/non-users	310 (21.5)	362 (22.9)		304 (23.1)	207 (22.4)	58 (29.2)	103 (20.0)
History of significant acute disease	50 (3.5)	56 (3.7)		54 (4.1)	29 (3.1)	2 (1.0)	21 (4.1)
Single minor chronic disease	594 (41.3)	627 (41.4)		546 (41.5)	394 (42.7)	81 (40.7)	200 (38.8)
Minor chronic disease in multiple organ systems	53 (3.7)	46 (3.0)		41 (3.1)	30 (3.3)	5 (2.5)	23 (4.5)
Single dominant or moderate chronic disease	319 (22.2)	326 (21.5)		282 (21.4)	188 (20.4)	44 (22.1)	131 (25.4)
Significant chronic disease in multiple organ systems	109 (7.6)	97 (6.4)		88 (21.4)	72 (7.8)	9 (4.5)	37 (7.2)
Dominant chronic disease in 3+ organ systems	1 (0.1)	0 (0.0)		0 (0.0)	1 (0.1)	0 (0.0)	0 (0.0)
Dominant and metastatic malignancies	1 (0.1)	0 (0.0)		0 (0.0)	0 (0.0)	0 (0.0)	1 (0.2)
Catastrophic condition status	2 (0.1)	1 (0.1)		1 (0.1)	2 (0.2)	0 (0.0)	0 (0.0)
FFS plan	105 (7.3)	142 (9.4)	0.043	139 (10.5)	98 (10.6)	3 (1.5)	7 (1.4)	0.000
Primary care visits (SD)	4.2 (2.3)	3.7 (2.0)	0.000	3.8 (2.1)	4.2 (2.3)	3.5 (1.9)	4.2 (2.1)	0.000
Weekend ED visits	403 (28.0)	410 (27.0)	0.545	351 (26.6)	258 (27.9)	59 (29.7)	145 (28.1)	0.757
Dwelling location			0.000					0.000
Urban	643 (44.9)	504 (33.3)		423 (32.1)	398 (43.4)	81 (40.7)	245 (47.8)
Suburban	321 (22.4)	509 (33.6)		448 (34.0)	200 (21.8)	61 (30.7)	121 (23.6)
Rural	467 (32.6)	503 (33.2)		446 (33.9)	320 (34.9)	57 (28.6)	147 (28.7)
Median household income of census tract (SD)	40,563 (13,960)	40,595 (14,376)	0.928	40,725 (14,653)	40,077 (12,829)	39,772 (12,454)	41,439 (15,765)	0.050
County PCMH proportion	21.5 (16.6)	16.3 (15.3)	0.000	15.5 (15.3)	20.8 (16.3)	21.8 (14.3)	22.7 (16.8)	0.000
Mean distance to PCP (SD)	42.7 (54.6)	43.5 (53.9)	0.535	43.7 (57.1)	41.7 (48.8)	42.6 (23.9)	44.3 (63.4)	0.565
Mean distance to ED (SD)	18.6 (55.1)	18.5 (47.0)	0.955	18.5 (49.5)	17.9 (50.7)	18.2 (26.7)	19.8 (62.3)	0.803
ED visits
Non-emergent (avoidable)	616 (21.0)	739 (23.6)	0.014	621 (23.0)	383 (20.4)	118 (28.0)	233 (22.1)	0.006
Emergent/primary care treatable (potentially avoidable)	514 (17.5)	512 (16.4)	0.235	442 (16.4)	320 (17.0)	70 (16.6)	194 (18.4)	0.512

PCMH attribution groups: (1) un-enrolled and never attended; (2) un-enrolled and attended, (3) enrolled and never attended, (4) enrolled and always attended; PCP = primary care provider; ED = emergency department; SD = standard deviation (from mean); FFS = fee-for-service.

Table 2

Person-years summary statistics for MMA outcomes.

	ADD cohort
	PCMH attribution based on attendance			PCMH attribution based on enrollment and attendance
Characteristic	PCMH(n = 3,274)	Non-PCMH(n = 3,116)	p	Group 1(n = 2,625)	Group 22(n = 1,762)	Group 3(n = 491)	Group 4(n = 1,512)	p value
Age (SD)	9.8 (3.5)	10.2 (3.6)	0.000	10.2 (3.6)	9.7 (3.5)	10.3 (3.6)	9.8 (3.5)	0.000
Sex (Male)	1,902 (58.1)	1,800 (57.8)	0.791	1,518 (57.8)	1,011 (57.4)	282 (57.4)	891 (58.9)	0.826
Race			0.098					0.000
non-Hispanic White	1,095 (33.5)	1,101 (35.3)		953 (36.3)	593 (33.7)	148 (30.1)	502 (33.2)
Black	1,952 (59.6)	1,832 (58.8)		1,529 (58.3)	1,070 (60.7)	303 (61.7)	882 (58.3)
Hispanic	227 (6.9)	183 (5.9)		143 (5.5)	99 (5.6)	40 (8.2)	128 (8.5)
Clinical Risk Group			0.019					0.036
Healthy/non-users	267 (8.2)	291 (9.3)		243 (9.3)	128 (7.3)	48 (9.8)	139 (9.2)
History of significant acute disease	277 (8.5)	327 (10.5)		269 (10.3)	140 (8.0)	58 (11.8)	137 (9.1)
Single minor chronic disease	155 (4.7)	168 (5.4)		138 (5.3)	86 (4.9)	30 (6.1)	69 (4.6)
Minor chronic disease in multiple organ systems	14 (0.4)	15 (0.5)		12 (0.5)	10 (0.6)	3 (0.6)	4 (0.3)
Single dominant or moderate chronic disease	1,970 (60.2)	1,818 (58.3)		1,526 (58.1)	1,072 (60.8)	292 (59.5)	898 (59.4)
Significant chronic disease in multiple organ systems	585 (17.9)	492 (15.8)		433 (16.5)	322 (18.3)	59 (12.0)	263 (17.4)
Dominant chronic disease in 3+ organ systems	2 (0.1)	0 (0.0)		0 (0.0)	2 (0.1)	0 (0.0)	0 (0.0)
Dominant and metastatic malignancies	1 (0.0)	2 (0.1)		2 (0.1)	0 (0.0)	0 (0.0)	1 (0.1)
Catastrophic condition status	3 (0.1)	3 (0.1)		2 (0.1)	2 (0.1)	1 (0.2)	1 (0.1)
FFS plan	172 (5.3)	178 (5.7)	0.420	170 (6.5)	146 (8.3)	8 (1.6)	26 (1.7)	0.000
Primary care visits (SD)	5.7 (3.5)	4.6 (2.5)	0.000	4.6 (2.4)	5.6 (3.5)	4.5 (2.6)	5.8 (3.6)	0.000
Weekend ED visits	1,061 (32.1)	917 (29.4)	0.010	762 (29.0)	586 (33.3)	155 (31.6)	475 (31.4)	0.027
Dwelling location			0.000					0.000
Urban	1,762 (53.9)	1,330 (42.7)		1,065 (40.6)	921 (52.3)	265 (54.0)	841 (55.6)
Suburban	711 (21.7)	897 (28.8)		787 (30.0)	379 (21.5)	110 (22.4)	332 (22.0)
Rural	799 (24.4)	889 (28.5)		773 (30.5)	460 (26.1)	116 (23.6)	339 (22.4)
Median household income of census tract (SD)	42,281 (14,801)	40,508 (14,107)	0.000	40,422 (13,908)	41,988 (14,955)	40,968 (15,125)	42,636 (14,607)	0.000
County PCMH proportion	21.1 (16.0)	16.0 (15.0)	0.000	15.3 (14.8)	21.2 (15.9)	19.4 (16.0)	21.1 (16.0)	0.000
Mean distance to PCP (SD)	36.8 (26.7)	41.6 (31.4)	0.000	41.3 (29.8)	37.6 (28.5)	43.1 (39.2)	35.8 (24.1)	0.000
Mean distance to ED (SD)	15.8 (28.3)	17.3 (40.6)	0.003	16.8 (28.7)	16.2 (22.9)	19.4 (78.0)	15.2 (33.6)	0.000
ED visits
Non-emergent (avoidable)	2,121 (20.8)	2,084 (20.3)	0.359	1,713 (19.8)	1,161 (20.8)	371 (22.9)	960 (20.8)	0.028
Emergent/primary care treatable (potentially avoidable)	1,730 (17.0)	1,785 (17.4)	0.441	1,508 (17.4)	947 (17.0)	277 (17.1)	783 (17.0)	0.877

PCMH attribution groups: (1) un-enrolled and never attended; (2) un-enrolled and attended, (3) enrolled and never attended, (4) enrolled and always attended; PCP = primary care provider; ED = emergency department; SD = standard deviation (from mean); FFS = fee-for-service.

A-spatial adjusted predicted probabilities for ED visit type by clinic attribution grouping for each cohort are shown in Table 3. Among the ADD cohort, the adjusted predicted probability of experiencing an avoidable or potentially avoidable visit by clinic attribution grouping ranged from 0.205 to 0.282 and from 0.162 to 0.174, respectively. The 2.4 percentage point reduction in risk of avoidable ED visits among children in the ADD cohort who attended a PCMH versus those who did not was statistically significant (p < 0.01). The 1.1 percentage point increase in risk of a potentially avoidable ED visit among children in the ADD cohort who attended a PCMH was not statistically significant. When PCMH attribution was based on enrollment and attendance, children in the ADD cohort who were enrolled in a PCMH but never attended one for their primary care exhibited a 5.4 percentage point increase in an avoidable ED visit (p < 0.05) compared to children who were un-enrolled in a medical home and never attended one for primary care (reference group). The 2.4 percentage point decrease in risk for avoidable visits among children that were enrolled and attended a medical home was not statistically significant from the reference group. Among the MMA cohort, the adjusted probability for avoidable or potentially avoidable visits ranged from 0.196 to 0.226 and from 0.167 to 0.173 across all attribution groups, respectively. Children in the MMA cohort who were enrolled in a PCMH and never attended one for primary care exhibited a 3.0 percentage point increase in risk of an avoidable ED visit compared to children in the reference group (p < 0.05). No other comparisons were statistically significant. Regression coefficients for all models are provided in the supplementary online appendix.

Table 3

A-spatial comparisons of predicted probabilities in ED visit category by medical home attribution grouping type.

	ADD Cohort		MMA Cohort
	Avoidable Visits	Potentially Avoidable Visits	Avoidable Visits	Potentially Avoidable Visits
PCMH grouping type 1
Non-PCMH	0.235 (0.008)**	0.163 (0.007)**	0.201 (0.004)**	0.172 (0.004)**
PCMH	0.211 (0.008)**	0.174 (0.007)**	0.205 (0.004)**	0.168 (0.004)**
Δ Differences in margins (type 1)
PCMH versus non-PCMH	-0.024 (0.011)*	0.011 (0.010)	0.004 (0.006)	-0.004 (0.006)
PCMH grouping type 2
Un-enrolled + never attended (1)	0.229 (0.009)**	0.163 (0.008)**	0.196 (0.005)**	0.173 (0.005)**
Un-enrolled + always attended (2)	0.205 (0.010)**	0.169 (0.009)**	0.206 (0.006)**	0.167 (0.005)**
Enrolled + never attended (3)	0.282 (.022)**	0.163 (0.019)**	0.226 (0.016)**	0.169 (0.010)**
Enrolled + always attended (4)	0.223 (0.013)**	0.183 (0.012)**	0.204 (0.007)**	0.169 (0.006)**
Δ Differences in margins (type 2)
group 2 versus group 1	-0.006 (0.016)	0.020 (0.014)	0.007 (0.008)	-0.004 (0.008)
group 3 versus group 1	0.054 (0.024)*	0.004 (0.020)	0.030 (0.012)*	-0.004 (0.011)
group 4 versus group 1	-0.024 (0.013)	0.006 (0.012)	0.009 (0.008)	-0.006 (0.007)

PCMH attribution groups: (1) un-enrolled and never attended; (2) un-enrolled and attended, (3) enrolled and never attended, (4) enrolled and always attended. All group comparisons in bold are statistically significant at

** p < 0.01,

* p < 0.05.

Table 4 shows the results of models that included the spatial interactions. After accounting for relative differences in proximity to providers, the 2.4 percentage point decrease in risk of avoidable ED visits among children in the ADD cohort shown in Table 3 increased to 3.9 to 7.2 percentage points as the relative proximity to primary care providers over hospitals improved (p < 0.05). The differences in risk remained 3.1 percentage points lower so long as the distance to the medical home was not 15 miles further to access than the hospital (p < 0.01). In contrast to the a-spatial model that found no difference in risk of potentially avoidable visits by PCMH attribution type, children in the ADD cohort that attended medical homes exhibited a 3.2 to 7.6 percentage point increase in risk of a potentially avoidable visit so long as proximity to primary care providers was more convenient than EDs (p < 0.05). There were no differences in risk of an avoidable ED visit among the ADD cohort when proximity to hospitals was the most convenient.

Table 4

Comparisons of predicted probabilities in ED visits for ADD outcomes by medical home attribution type after accounting for geographic proximity to providers.

	ADD Cohort
	Avoidable Visits							Potentially Avoidable Visits
	(Greater proximity to PCP)			Equal distance	(Greater proximity to ED)			(Greater proximity to PCP)			Equal distance	(Greater proximity to ED)
	60 miles	30 miles	15 miles	0 miles	15 miles	30 miles	60 miles	60 miles	30 miles	15 miles	0 miles	15 miles	30 miles	60 miles
PCMH attribution (type 1)
Non-PCMH	0.255 (0.025)	0.248 (0.017)	0.245 (0.014)	0.242 (0.011)	0.239 (0.009)	0.236 (0.008)	0.230 (0.012)	0.127 (0.020)	0.139 (0.015)	0.145 (0.012)	0.152 (0.010)	0.158 (0.008)	0.165 (0.007)	0.180 (0.013)
PCMH	0.182 (0.023)	0.192 (0.016)	0.197 (0.013)	0.203 (0.010)	0.208 (0.008)	0.213 (0.008)	0.225 (0.014)	0.204 (0.025)	0.194 (0.012)	0.189 (0.013)	0.184 (0.010)	0.179 (0.008)	0.174 (0.008)	0.165 (0.012)
Δ Differences in margins	-0.072 (0.034)*	-0.056 (0.024)*	-0.048 (0.019)*	-0.039 (0.015)**	-0.031 (0.012)**	-0.023 (0.012)	-0.005 (0.018)	0.076 (0.032)*	0.055 (0.022)*	0.043 (0.018)*	0.032 (0.014)*	0.021 (0.011)	0.009 (0.011)	-0.014 (0.017)
PCMH attribution (type 2)
Unenrolled + never went (1)	0.253 (0.027)	0.245 (0.019)	0.241 (0.015)	0.237 (0.011)	0.232 (0.009)	0.228 (0.009)	0.221 (0.013)	0.134 (0.023)	0.144 (0.017)	0.149 (0.013)	0.154 (0.010)	0.159 (0.008)	0.165 (0.008)	0.176 (0.013)
Unenrolled + always went (2)	0.173 (0.026)	0.184 (0.019)	0.190 (0.016)	0.195 (0.012)	0.201 (0.010)	0.207 (0.010)	0.219 (0.016)	0.193 (0.029)	0.186 (0.020)	0.182 (0.015)	0.178 (0.012)	0.174 (0.010)	0.171 (0.010)	0.165 (0.015)
Enrolled + never went (3)	0.259 (0.067)	0.268 (0.047)	0.272 (0.037)	0.277 (0.029)	0.282 (0.023)	0.286 (0.022)	0.296 (0.037)	0.082 (0.041)	0.105 (0.034)	0.119 (0.030)	0.134 (0.024)	0.151 (0.020)	0.170 (0.020)	0.212 (0.041)
Enrolled + always went (4)	0.199 (0.045)	0.208 (0.032)	0.212 (0.025)	0.216 (0.019)	0.220 (0.014)	0.225 (0.014)	0.233 (0.025)	0.227 (0.047)	0.210 (0.030)	0.202 (0.023)	0.195 (0.016)	0.187 (0.012)	0.180 (0.012)	0.166 (0.019)
Δ Differences in margins
group 2 versus group 1	-0.079 (0.038)*	-0.061 (0.027)*	-0.051 (0.021)*	-0.041 (0.017)*	-0.031 (0.014)*	-0.021 (0.013)	-0.002 (0.021)	0.059 (0.037)	0.042 (0.026)	0.033 (0.020)	0.025 (0.016)	0.016 (0.013)	0.007 (0.013)	-0.011 (0.020)
group 3 versus group 1	0.006 (0.072)	0.023 (0.050)	0.032 (0.040)	0.041 (0.031)	0.049 (0.025)*	0.058 (0.024)*	0.075 (0.039)	-0.052 (0.046)	-0.039 (0.038)	-0.030 (0.033)	-0.020 (0.026)	-0.008 (0.021)	0.005 (0.021)	0.036 (0.044)
group 4 versus group 1	-0.054 (0.053)	-0.037 (0.037)	-0.029 (0.028)	-0.020 (0.022)	-0.012 (0.017)	-0.004 (0.016)	0.013 (0.028)	0.093 (0.052)	0.067 (0.034)*	0.054 (0.026)*	0.041 (0.019)*	0.028 (0.015)	0.015 (0.014)	-0.010 (0.023)

PCMH attribution groups: (1) un-enrolled and never attended; (2) un-enrolled and attended, (3) enrolled and never attended, (4) enrolled and always attended. All group comparisons in bold are statistically significant at

** p < 0.01,

* p < 0.05.

When assessed by PCMH attribution and enrollment type, children in the ADD cohort who were enrolled in a medical home, but never attended one for their primary care services exhibited a 4.9 to 5.8 percentage point increase in risk of an avoidable ED visit relative to children in the reference group so long as proximity to primary care providers was more convenient (p < 0.05), whereas children who attended medical homes even though they were un-enrolled in one through their MCO showed a 7.9 to 3.1 percentage point decrease in risk of avoidable ED visits relative to children in the reference group (p < 0.05). Similar trends in risk were found among children who were enrolled and attended medical homes, but the reductions in risk were not statistically significant.

As shown in Table 5, there were few statistically significant associations between PCMH attribution type and risk of avoidable and potentially avoidable ED visits among the MMA cohort after including the spatial interactions. As in the a-spatial comparisons, persons who were enrolled in a medical home, but did not attend one for their primary care exhibited an increased risk of avoidable ED visits, ranging in magnitude from 2.9 to 3.7 percentage points, but only when the proximity to the hospital was closer than their primary care provider (p < 0.05). No other spatial interactions were statistically significant.

Table 5

Comparisons of predicted probabilities in ED visits for MMA outcomes by medical home attribution type after accounting for geographic proximity to providers.

	MMA Cohort
	Avoidable Visits							Potentially Avoidable Visits
	(Greater proximity to PCP)			Equal distance	(Greater proximity to ED)			(Greater proximity to PCP)			Equal distance	(Greater proximity to ED)
	60 miles	30 miles	15 miles	0 miles	15 miles	30 miles	60 miles	60 miles	30 miles	15 miles	0 miles	15 miles	30 miles	60 miles
PCMH attribution (type 1)
Non-PCMH	0.191 (0.010)	0.195 (0.007)	0.197 (0.006)	0.199 (0.005)	0.201 (0.005)	0.203 (0.004)	0.207 (0.006)	0.171 (0.011)	0.173 (0.007)	0.174 (0.006)	0.174 (0.005)	0.175 (0.004)	0.176 (0.004)	0.177 (0.006)
PCMH	0.209 (0.012)	0.209 (0.008)	0.208 (0.007)	0.208 (0.005)	0.208 (0.005)	0.208 (0.005)	0.208 (0.007)	0.174 (0.009)	0.173 (0.006)	0.172 (0.005)	0.171 (0.005)	0.171 (0.004)	0.170 (0.004)	0.168 (0.006)
Δ Differences in margins	0.018 (0.015)	0.014 (0.011)	0.012 (0.009)	0.010 (0.007)	0.007 (0.006)	0.005 (0.007)	0.001 (0.009)	0.003 (0.014)	0.000 (0.010)	-0.001 (0.008)	-0.002 (0.007)	-0.005 (0.006)	-0.006 (0.006)	-0.009 (0.008)
PCMH attribution (type 2)
Unenrolled + never went (1)	0.190 (0.010)	0.192 (0.008)	0.194 (0.006)	0.195 (0.005)	0.196 (0.005)	0.198 (0.005)	0.201 (0.006)	0.167 (0.010)	0.170 (0.008)	0.172 (0.006)	0.173 (0.005)	0.175 (0.005)	0.177 (0.005)	0.180 (0.007)
Unenrolled + always went (2)	0.198 (0.017)	0.202 (0.012)	0.204 (0.009)	0.206 (0.007)	0.208 (0.006)	0.210 (0.006)	0.214 (0.010)	0.164 (0.014)	0.166 (0.010)	0.167 (0.008)	0.168 (0.006)	0.169 (0.006)	0.170 (0.006)	0.172 (0.008)
Enrolled + never went (3)	0.207 (0.014)	0.214 (0.018)	0.218 (0.016)	0.222 (0.014)	0.226 (0.012)	0.230 (0.012)	0.238 (0.015)	0.186 (0.032)	0.181 (0.022)	0.179 (0.017)	0.177 (0.014)	0.175 (0.011)	0.173 (0.010)	0.168 (0.015)
Enrolled + always went (4)	0.217 (0.014)	0.213 (0.010)	0.212 (0.009)	0.210 (0.007)	0.209 (0.007)	0.207 (0.007)	0.204 (0.010)	0.181 (0.010)	0.177 (0.008)	0.176 (0.007)	0.174 (0.006)	0.172 (0.006)	0.170 (0.006)	0.166 (0.007)
Δ Differences in margins
group 2 versus group 1	0.008 (0.020)	0.009 (0.014)	0.010 (0.011)	0.011 (0.009)	0.011 (0.008)	0.012 (0.008)	0.013 (0.012)	-0.003 (0.017)	-0.004 (0.012)	-0.005 (0.010)	-0.005 (0.008)	-0.006 (0.007)	-0.007 (0.007)	-0.008 (0.011)
group 3 versus group 1	0.017 (0.026)	0.022 (0.020)	0.025 (0.017)	0.027 (0.015)	0.029 (0.013)*	0.032 (0.013)*	0.037 (0.016)*	0.018 (0.033)	0.011 (0.023)	0.007 (0.018)	0.004 (0.015)	-0.000 (0.012)	-0.004 (0.011)	-0.012 (0.016)
group 4 versus group 1	0.027 (0.017)	0.021 (0.013)	0.018 (0.010)	0.015 (0.009)	0.012 (0.009)	0.009 (0.008)	0.003 (0.001)	0.015 (0.014)	0.008 (0.011)	0.004 (0.009)	0.000 (0.008)	-0.003 (0.008)	-0.007 (0.008)	-0.014 (0.010)

PCMH attribution groups: (1) un-enrolled and never attended; (2) un-enrolled and attended, (3) enrolled and never attended, (4) enrolled and always attended. All group comparisons in bold are statistically significant at

** p < 0.01,

* p < 0.05.

Prior to the analysis, we conducted a sensitivity analysis of ED visits by primary diagnosis codes and PCMH attribution group. For the ADD cohort, the top 10 primary diagnosis codes were identical for 90% of all visits, with the leading causes attributed to J02.9 (Acute pharyngitis, unspecified); J06.9(Acute upper respiratory infection, unspecified); and J02.0 (Streptococcal pharyngitis). Depending on the group, the top visit causes contributed to 20% to 25% of all visits. The diagnosis codes were identical for 88% of the leading visit causes for the MMA cohort, with most claims attributed to J45.901 (Unspecified asthma with (acute) exacerbation); J06.9 (Acute upper respiratory infection, unspecified); and J45.909 (Unspecified asthma, uncomplicated). Depending on group, the top visit causes accounted for 32% - 34% of all visits, respectively.

Discussion

This study analyzed a subset of South Carolina Medicaid pediatric claims data for beneficiaries having a pre-existing diagnosis of asthma or attention-deficit/hyperactivity disorder and assessed different a-spatial and spatial interactions of visit risk based on whether or not the beneficiary attended a designated medical home. We assessed two different depictions of a beneficiary’s PCMH attribution type: one based completely on attendance, and one based on enrollment and attendance records. We found that geographical location between patients, primary care providers, and hospitals does play a role in ED utilization for some pediatric groups with pre-existing chronic illnesses. We also found that these associations varied by PCMH attribution type. Many of the interactions showed that a-spatial findings were driven by instances where children had better access to medical homes versus hospitals. The intersection of these findings suggests that for some pediatric patient populations with pre-existing illnesses, a medical home effect on ED utilization rates may be contingent on where primary providers are located relative to hospitals. These findings provide some evidence that previous pediatric assessments of ED utilizations may be under-counting the benefit of medical home-modeled care owing to a distance decay effect patient’s and providers. That medical home effectiveness gets diluted owing to geographic proximity to providers may be particularly relevant for future evaluations owing to recent findings that primary care practices that elect to transition into designated medical homes tend to locate in more socioeconomically advantaged communities [74].

However, despite these findings, the effects of geographic location on risk reductions attributable to medical home attendance were not consistently found for both cohorts. Contextually, the mixed findings raise questions as to why PCMH attendance would exhibit a benefit to some ED visit types and opposite trends for other visit types once proximity to providers was assessed. That the spatial interactions either increased the a-spatial finding that PCMH attendance lowered risk of ED utilization or revealed trends that were not previously observed suggests that the significance of geographic location is more meaningful than spurious. One explanation might stem from the differences in proximity to providers within the MMA cohort, as its medical home attendees had to travel approximately 5 to 10 miles less to access PCMHs compared to children in the ADD cohort. Patient morbidity, provider recommendations, or time of day may have also accounted for differences in utilization trends for emergent but potentially avoidable visits. Whether these patterns are repeated in other patient cohorts or among adult ED utilization patterns is an important topic for future research.

Of the recent summaries and meta-analyses of medical home evaluations, the reviews conducted by Sinaiko et al (2017), Friedberg et al (2014), and Jackson et al (2013) have shown variation in overall improvements in primary care utilization, ED visits, as well as inpatient hospital admissions [41,75,76]. Factors such as practice size, patient mix, and practice ownership contribute to observed heterogeneity in study findings. Although evidence exists that geography escalates variation in healthcare access and outcomes [77-83], geographic location as a potential cause for variation in PCMH success has not been rigorously examined. This gap is significant as needs-based medical assistance programs such as the Centers for Medicare and Medicaid Services (CMS) rely on information technology platforms such as GIS to evaluate whether Medicaid beneficiaries have timely and adequate access to the providers and services that participate in its network [58,84]. Our findings that higher utilization patterns may have been driven by relative distances to providers or that non-emergent visits were consistently higher among patients that were documented as enrolled in a PCMH but never attended one for their primary care may point to specific instances where network adequacy thresholds that account for relative distances between providers might be warranted. Neither of these findings could have been uncovered without including spatial interactions in the models.

The HEDIS measures chosen for analysis were selected based on the relationship between comorbid conditions and hospital emergency department encounters that are often avoidable or treatable in primary care settings [42,43]. Its use also helped establish baseline similarities across the patient panels that made it possible to exploit differences in outcomes attributed to medical home status and/or geography. Although our study was limited in the number of quality indicators available from the encounter data, our findings are encouraging from the standpoint that additional benefits (or lack thereof) of medical home attribution can be identified for patient populations once details of health services locations are included in the model. Future work could build on this approach to assess whether the geographic location of medical homes modifies other performance measures.

The observations from this study have some noteworthy limitations. Firstly, access to health care services is a multidimensional concept and influenced by factors that have both spatial and a-spatial dimensions. Although our evaluation did allow us to control for key factors required for meaningful risk-adjustment, other clinical and patient-reported measures that we were unable to obtain may further delineate the benefits attributed to medical home access. For example, we could not control for factors such as wait times, historical service use, time or day, or patient satisfaction with ED care versus primary care in our analysis, all of which may have helped to further contextualize differences in use. Similarly, although many state Medicaid agencies are experimenting with some version of PCMH transformation, the varied nature of state Medicaid characteristics make national geographic comparisons difficult to construct and we cannot say whether these trends over or underestimate its effect. Additional state-based tests would be beneficial for context validity as well as for defining appropriate access thresholds for other performance measures. With respect to our spatial approach, the primary limitation of assessing relative differences to providers is that we did not focus on exact distances to care. For example, in our model someone could have been assigned into the 0–15 mile grouping if they lived 20 miles from the ED or 200 miles from the ED, so long as the difference in proximity to their primary care provider remained within 15 miles. Similarly, our analysis may have been biased owing to the lack of information on PCMH enrollment and attendance in the claims data, although we were able to circumvent some of these issues using the enrollment and attendance matrix. Similarly, we could not confirm whether all beneficiaries initiated their care from their place of residence or how they traveled to obtain care (e.g., car, bus, walk), but we could confirm that all claims represented physical encounters at primary care clinics and hospitals. Another limitation is that our analysis may have muted disparities that might have emerged had we included privately insured populations. However, we did find that differences in outcomes among some of the most socially vulnerable populations in the state could be demarcated once the models accounted for relative distances populations must travel to obtain health care services.

Conclusion

In the age of patient-centered medicine, GIS technology is becoming even more important for optimizing case management (e.g., list of available transportation resources) as well as enhancing patient experience with care (e.g., providing web-based service locations of nearby providers). Although geographic variation is an accepted phenomenon in health care services research, the findings in this study illustrate that health care performance measures can be enhanced by knowing how far patients must travel in order to obtain services. Although medical home attendance may not be consistently associated with lower ED visits for all patient groups, our findings suggest that the lack of spatial information may be one factor why the anticipated effects of some PCMH innovations is often muted.

Adjusted GEE coefficients for ED visits where PCMH attribution was based on attendance data.

Covariates were included in the models if bivariate comparisons had p-values at or below 0.10.

(DOCX)

Click here for additional data file.

Adjusted GEE coefficients for ED visits where PCMH attribution was based on attendance and enrollment data.

Covariates were included in the models if bivariate comparisons had p-values at or below 0.10.

(DOCX)

Click here for additional data file.

25 Apr 2022

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Jingjing Qian

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide

3. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For more information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide

4. We note that Figure 1 in your submission contain [map/satellite] images which may be copyrighted. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For these reasons, we cannot publish previously copyrighted maps or satellite images created using proprietary data, such as Google software (Google Maps, Street View, and Earth). For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

We require you to either (1) present written permission from the copyright holder to publish these figures specifically under the CC BY 4.0 license, or (2) remove the figures from your submission:

1. You may seek permission from the original copyright holder of Figure(s) [#] to publish the content specifically under the CC BY 4.0 license.

We recommend that you contact the original copyright holder with the Content Permission Form (http://journals.plos.org/plosone/s/file?id=7c09/content-permission-form.pdf) and the following text:

“I request permission for the open-access journal PLOS ONE to publish XXX under the Creative Commons Attribution License (CCAL) CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). Please be aware that this license allows unrestricted use and distribution, even commercially, by third parties. Please reply and provide explicit written permission to publish XXX under a CC BY license and complete the attached form.”

Please upload the completed Content Permission Form or other proof of granted permissions as an "Other" file with your submission.

In the figure caption of the copyrighted figure, please include the following text: “Reprinted from [ref] under a CC BY license, with permission from [name of publisher], original copyright [original copyright year].”

2. If you are unable to obtain permission from the original copyright holder to publish these figures under the CC BY 4.0 license or if the copyright holder’s requirements are incompatible with the CC BY 4.0 license, please either i) remove the figure or ii) supply a replacement figure that complies with the CC BY 4.0 license. Please check copyright information on all replacement figures and update the figure caption with source information. If applicable, please specify in the figure caption text when a figure is similar but not identical to the original image and is therefore for illustrative purposes only.

The following resources for replacing copyrighted map figures may be helpful:

USGS National Map Viewer (public domain): http://viewer.nationalmap.gov/viewer/

The Gateway to Astronaut Photography of Earth (public domain): http://eol.jsc.nasa.gov/sseop/clickmap/

Maps at the CIA (public domain): https://www.cia.gov/library/publications/the-world-factbook/index.html and https://www.cia.gov/library/publications/cia-maps-publications/index.html

NASA Earth Observatory (public domain): http://earthobservatory.nasa.gov/

Landsat: http://landsat.visibleearth.nasa.gov/

USGS EROS (Earth Resources Observatory and Science (EROS) Center) (public domain): http://eros.usgs.gov/#

Natural Earth (public domain): http://www.naturalearthdata.com/

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: I Don't Know

Reviewer #2: No

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Summary and Overall Impression

Thank you for the opportunity to review this important work.

This work focused attention to an evolving area of health services research that is much needed to non-medical determinants of health. The authors appear to take a very thorough approach in the development of their models to explore the possible linkage between PMCH designation and geographic distance to primary care clinic ED. The use of proxy PCMH attribution status seems relevant and plausible in the absence of an alternative for its proposed application, however, the interpretation of the results given the age groups (children) and uncertainty around these group designations (ie: enrolled vs assigned to PCMH) makes one given some consideration to the generalizability of the findings. The discussion appropriately highlights some key aspects of the work that have relevance to the existing literature, some inherent weaknesses mainly in the short-comings of using administrative data and future work.

As a reader/reviewer, my impression during an intense second read, it became apparent that some of the discussion contained important contextual (background) information that likely could be used to strengthen the introduction - and provide additional points for the reader to better understand the 'how' and 'why' for the analysis was undertaken. I believe this will contextualize the results in a more purposeful fashion. For example: providing additional background the states of S. Carolina, it's population and those on medicaid along with existing rural - urban disparities in access to a family physician. (i.e. what % of the rural population are served by 'X' number of doctors vs those in urban areas). And, for the demographics chosen, are there options also for care to be received by community Pediatricians outside the PCMH umbrella. With some additional focused information on the geographic and demographic variable, perhaps framed in a hypothesis, will greatly support the outcomes found in this work.

Minor

Lines 13-22 This second paragraph contains alot of important information some of which does not

relate directly to the study. Consider a much shorter segue to third paragraph to further focus the message.

Table 3 - contains some valuable information; suggest CRG values assigned be displayed vs descriptive breakdown

for ease of viewing and interpretation.

Miscellaneous

Table 1 - Distance to PCP & Distance to ED - is the average distance?. Further clarification might be helpful.

The sentence in Lines 9-11 ending with 'healthy and prosperous society' seems to be a key statement setting up the study; with some minor editing could be strengthened to highlight to the reader the interplay of (inequity) access to primary care, SDoH and improved patient health outcome and overall population health vs. policies for healthy and prosperous society' any referenced work from Sir Michael Marmot or Barbara Starfield might be valuable to consider.

Lines 18-20 re: Ontario Family Health Team indicates 'specialty care teams', is this accurate, versus community primary care providers working along side allied health (social work, psychologist) and nursing? Specialty care suggests surgery or medical specialist providers.

P-values in results descriptions ex: Pg 10 Lines 12-14 p-values are missing.

Pg10 Line 14 medical ? home.

Reviewer #2: In this three-year retrospective cohort study, authors evaluated the number of pediatric ED visits based on PCMH status and geographic location to their PCP or hospital. Authors concluded an interesting finding that the PCMH model may not lower ED visits but geographic proximity to care may need to be considered. Overall, the manuscript is well-written and addresses a significant issue of health disparities. However, the results do not seem to support the authors’ conclusions. For example, Table 4 provides detailed findings that, overall, do not seem to have significance between PCCMH status nor geographic location; there are minimal differences between PCMH status group comparisons including group 4 that are the most established. While the data are interesting, authors need to emphasis the major significant findings in the abstract, results, and tables so readers will clearly understand the value of this work and the gaps in literature it addresses.

GENERAL

- Change “ED admissions” to “ED visits” throughout the paper. Admission is admitted to the hospital; ED visits are patients coming to the ED but may or may not be admitted to the hospital.

- Clarification is needed of the phrase “geographic proximity to primary care providers vs hospitals”. Does this mean that the closer individuals were to the type of healthcare entity, the more likely to receive care from that place, e.g., closer to PCP then receive care from them, closer to ED, then receive care from them? If so, this seems intuitive.

ABSTRACT

- Add participant age to Methods

- Results need numerical detail, e.g., population n, % or n of findings, etc. and the corresponding statistical significance as applicable.

INTRODUCTION – overall well-written and outlines the background to the study.

- It is a strength of the study that data were collected prior to COVID since, during the pandemic, it may have been have skewed the data toward ED visits. Authors may want to omit COVID-19 from the intro (pg. 3, line 3) so readers are not confused and highlight this strength in the discussion.

- Pg 2 Line 10. Change “racism” to “race/ethnicity”

- Add the gap authors are filling from this study and the study aim/objective

METHODS – overall detailed and use validated measures for outcomes (NRU ED visits algorithm)

RESULTS

- add numerical values for major findings and their significance

- Typo on Pg 10, line 3: delete among

TABLES/FIGURES

- ensure all abbreviations are defined, e.g., MMA, ADD, HEDIS, etc. and add legends to help the reader understand the meaning/rationale of MMA, ADD, attribution groups, etc.

- Table 4: why is statistical significance considered p<0.10?

- Table 4: There are several bolded items that are not statistically significant

- Table 4: p-values are a probabilities and should not be negative

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Terrence McDonald, MD, MSc, CCFP (SEM), FCFP, Dip. Sport Med

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

6 Jul 2022

I have uploaded (as a word file) an itemized list of our response to the reviewer comments. I am pasting this information here if more convenient. Because HTML will drop formatting, the copy/past should indent our response (plus add a bullet).

June 20, 2022

To the editors of PLOS One,

Thank you for the opportunity to have our manuscript peer-reviewed. We have revised our manuscript based on the thoughtful comments and suggestions from the reviewers and wish to extend our appreciation for their time and effort. All of their comments and suggestions have been addressed in our revised submission.

To summarize: This study evaluated the association between patient-centered medical home (PCMH) enrollment and avoidable ED visits among pediatric Medicaid beneficiaries in South Carolina over a 3-year period prior to the COVID-19 pandemic. It addresses a limitation of ongoing medical home evaluations that suggests they are not effectively reducing avoidable ED visits among children with chronic illness. Our regression models accounted for geographic relationships between patient and primary care provider locations, which is a novel methodological contribution for evaluating these events.

Key findings: Our study provides evidence that medical homes are helping to lower avoidable ED visits among children with pre-existing illnesses once geographical linkages between patients, providers, and hospitals are included in the models. They are important to state Medicaid agencies, payers, and clinical researchers as it points to the value of continuing to invest in medical home-modeled care as it contextualizes the benefits of PCMHs that may be going unmeasured. Our findings are also timely in light of the emergent research on social risk factor adjustment in pay-for-performance models.

We are pleased that both reviewers felt that this was important work. Briefly, the primary concern of reviewer #1 was that some key paragraphs in the discussion section be moved to the introduction to better frame the importance of this work. This resulted in a near complete re-write of the introduction, but one that we believe addresses their comments. Reviewer #1 also identified some minor methodological and grammatical errors, all of which have been addressed. Reviewer #2’s suggestions were also very pragmatic, with emphasis on clarification for why p values < 0.10 were included as well as some recommendations for reworking the abstract. These and their other minor comments have all been addressed in this revised submission.

The following pages (attached separately) provide an itemized list of the reviewer comments (italicized) and how we amended the manuscript (bulleted) and where the changes can be found in the cleaned copy. On behalf of my co-authors, we thank you for continued consideration of this work. All authors listed in this submission fulfill the journal’s requirements for publication and gave final approval of the manuscript submitted for continued review.

Kind regards,

Nathaniel Bell PhD

**************************************

RESPONSE TO REVIEWER COMMENTS FOLLOWS

**************************************

Reviewer #1: Summary and Overall Impression

Thank you for the opportunity to review this important work.

This work focused attention to an evolving area of health services research that is much needed to non-medical determinants of health. The authors appear to take a very thorough approach in the development of their models to explore the possible linkage between PMCH designation and geographic distance to primary care clinic ED. The use of proxy PCMH attribution status seems relevant and plausible in the absence of an alternative for its proposed application, however, the interpretation of the results given the age groups (children) and uncertainty around these group designations (ie: enrolled vs assigned to PCMH) makes one given some consideration to the generalizability of the findings. The discussion appropriately highlights some key aspects of the work that have relevance to the existing literature, some inherent weaknesses mainly in the short-comings of using administrative data and future work.

As a reader/reviewer, my impression during an intense second read, it became apparent that some of the discussion contained important contextual (background) information that likely could be used to strengthen the introduction - and provide additional points for the reader to better understand the 'how' and 'why' for the analysis was undertaken. I believe this will contextualize the results in a more purposeful fashion. For example: providing additional background the states of S. Carolina, it's population and those on medicaid along with existing rural - urban disparities in access to a family physician. (i.e. what % of the rural population are served by 'X' number of doctors vs those in urban areas). And, for the demographics chosen, are there options also for care to be received by community Pediatricians outside the PCMH umbrella. With some additional focused information on the geographic and demographic variable, perhaps framed in a hypothesis, will greatly support the outcomes found in this work.

• Thank you for these thoughtful comments and critiques. We have addressed each of your comments (above) in the revised manuscript.

• The entire introduction section has been heavily edited and re-written to better emphasize the ‘why’ and the ‘how’ and situate the reader within the context of what has been done, what are the major gaps, and why this work is important. These changes can be found on pages 2 through 5, which spans the entire introductory section.

• The methods sub-section for “PCMH attribution” on page 6 was edited to concisely explain why we chose these patient attribution groups. We do agree that streamlining some of these comparisons was necessary. As such, in the revised submission we limit our comparisons to those in which PCMH is based solely on attendance (attended vs. did not attend) and limit our original enrollment/attendance comparisons to be in reference to those who were unenrolled and never attended.

• The introduction section is now significantly longer, moving from broad to general while and emphasizes the importance of this work within the context of risk adjustment and in relation to medical home evaluations. We believe that our reframing of the introduction section based on your thoughtful comments and suggestions gets at the heart of the major comments.

• We have added additional information on SC Medicaid statistics as well as the state’s enrollment policies for medical homes. The major changes can be found on lines 15 – 23 on page 6 as well as lines 1 – 3 on page 7 in the methods sub-section on “PCMH attribution”. Note, we do not have every statistic available that reviewer #1 is mentioning, but we have added a number of contextual points that they are mentioning to better situate the reader within the context of this work.

Minor

Lines 13-22 This second paragraph contains a lot of important information some of which does not relate directly to the study. Consider a much shorter segue to third paragraph to further focus the message.

• Thank you for this suggestion. We have revised the entire introduction section to summarize the key areas that readers need to be aware of that are relevant to this work (1) avoidable ED utilizations, (2) primary care models designed to minimize their impact, (3) gaps in evidence and methodological approaches (e.g., GIS) that may help address the gaps, and (4) our aims/objectives for this research and why it is important and relevant.

Table 3 - contains some valuable information; suggest CRG values assigned be displayed vs descriptive breakdown for ease of viewing and interpretation.

• We agree – thank you for this feedback. We have amended the descriptive statistics tables as well as the logistic regression tables to show how the groups were similar and different.

• Because of the need to have a larger number of tables than average, we have provided some information (e.g., regression coefficients, standard errors) in an online appendix.

Miscellaneous

Table 1 - Distance to PCP & Distance to ED - is the average distance? Further clarification might be helpful.

• Yes, these are average distance values to providers. We have amended lines 7-9 on page 9 to articulate this to the reader in a revised sentence that reads: “Average travel distance estimates (in miles) for all visits were then constructed for all unique pairings between the beneficiary’s residential address recorded in the claims database to the location where their primary care and ED visit occurred.”

• We have also rewritten this paragraph (page 9, lines 1-14) to more clearly explain how relative differences in proximity were assessed.

The sentence in Lines 9-11 ending with 'healthy and prosperous society' seems to be a key statement setting up the study; with some minor editing could be strengthened to highlight to the reader the interplay of (inequity) access to primary care, SDoH and improved patient health outcome and overall population health vs. policies for healthy and prosperous society' any referenced work from Sir Michael Marmot or Barbara Starfield might be valuable to consider.

• Two very important scholars, indeed! Richard Wilkerson’s and Nancy Krieger’s work as well as Robert Evans and Clyde Hertzman’s work (for a Canadian perspective) are also key and I have found memories of meeting with them while attending graduate school at SFU and UBC. We have expanded the number of relevant references in the revised manuscript, but tried to keep the focus on studies that are specific to ED utilization.

Lines 18-20 re: Ontario Family Health Team indicates 'specialty care teams', is this accurate, versus community primary care providers working alongside allied health (social work, psychologist) and nursing? Specialty care suggests surgery or medical specialist providers.

• Yes and no. PCMH are unique to US, but other countries/systems are transitioning (or have transitioned) to team-based care, such as Canada’s Family Health Care teams. These edits can be found on page 2, lines 23-24.

P-values in results descriptions ex: Pg 10 Lines 12-14 p-values are missing.

• Our apology for this oversight. The results section in the revised submission contains p values (when significant) and points to all tables for specific p-values or threshold p values (e.g., < p 0.01, < p < 0.05).

Pg10 Line 14 medical ? home.

• Correct. Our apology for this typo. The sentence now includes “medical home”

Reviewer #2: Summary and Overall Impression

In this three-year retrospective cohort study, authors evaluated the number of pediatric ED visits based on PCMH status and geographic location to their PCP or hospital. Authors concluded an interesting finding that the PCMH model may not lower ED visits but geographic proximity to care may need to be considered. Overall, the manuscript is well-written and addresses a significant issue of health disparities. However, the results do not seem to support the authors’ conclusions. For example, Table 4 provides detailed findings that, overall, do not seem to have significance between PCCMH status nor geographic location; there are minimal differences between PCMH status group comparisons including group 4 that are the most established. While the data are interesting, authors need to emphasis the major significant findings in the abstract, results, and tables so readers will clearly understand the value of this work and the gaps in literature it addresses.

• Thank you for this feedback and observations. We have addressed the mismatch between the abstract + results section in the revised submission.

• The abstract results section was revised to better summarize the main findings.

• Please note, in meeting recommendations from Reviewer #1, we amended our PCMH grouping to add two different scenarios: one based on attendance (i.e., PCMH vs non-PCMH) and one based on enrollment + attendance (i.e., identical to our original submission), so some significant changes occurred in the revised manuscript.

GENERAL

Change “ED admissions” to “ED visits” throughout the paper. Admission is admitted to the hospital; ED visits are patients coming to the ED but may or may not be admitted to the hospital.

• We agree. All references to “admissions” have been changed to “visits” throughout. Thank you for this observation.

Clarification is needed of the phrase “geographic proximity to primary care providers vs hospitals”. Does this mean that the closer individuals were to the type of healthcare entity, the more likely to receive care from that place, e.g., closer to PCP then receive care from them, closer to ED, then receive care from them? If so, this seems intuitive.

• Yes, your interpretation is correct. We have amended the “Geographic assessment” sub-section of the Methods section (lines 1 – 14, page 9) to confirm that we are looking at relative distances to one provider over another, which adds a layer of depth to solely looking at distance to EDs or PCPs. This has its limitations, which we have added in the amended limitations section on pages 21 (lines 21- 24, and continued onto page 22) in the last paragraph of discussion section.

ABSTRACT

Add participant age to Methods

• The age groups of the pediatric cohort have been added to the Methods section of the abstract. These age groups are also provided in the methods section (lines 5-9, page 6).

Results need numerical detail, e.g., population n, % or n of findings, etc. and the corresponding statistical significance as applicable.

• Thank you for this observation. The results section of the abstract has been revised so that the key %’s/values are shown numerically.

INTRODUCTION

overall well-written and outlines the background to the study.

• Thank you. This has been an enjoyable and rewarding project to be a part of and we’re excited to share our findings.

It is a strength of the study that data were collected prior to COVID since, during the pandemic, it may have been have skewed the data toward ED visits. Authors may want to omit COVID-19 from the intro (pg. 3, line 3) so readers are not confused and highlight this strength in the discussion.

• Great point – we have added in a sentence in the introduction section (line 16 page 2) and again in the opening line of the Methods section (line 15 page 5) to emphasize that these trends/limitations that we’re comparing and trying to address are all pre-COVID.

Pg 2 Line 10. Change “racism” to “race/ethnicity”

• Thank you – change was made through deletion of the original paragraph during the re-write.

Add the gap authors are filling from this study and the study aim/objective

• Thank you – the last two paragraphs of the introduction (page 4 and 5) section summarize the gap(s) that this study is seeking to address, with the preceding paragraphs providing the rationale for our approach.

METHODS

overall detailed and use validated measures for outcomes (NYU ED visits algorithm)

• Thank you for this feedback.

RESULTS

add numerical values for major findings and their significance

• Thank you – percentage point differences that were significant are listed as well as their corresponding p values throughout the results section.

Typo on Pg 10, line 3: delete among

• Thank you – this grammatical error was removed during the re-write.

TABLES/FIGURES

ensure all abbreviations are defined, e.g., MMA, ADD, HEDIS, etc. and add legends to help the reader understand the meaning/rationale of MMA, ADD, attribution groups, etc.

• Thank you. These have been added throughout when possible at the bottom of the tables.

• We have also included a list of abbreviations at the end of the manuscript.

Table 4: why is statistical significance considered p<0.10?

• This is a longer discussion, but the short end is that many statisticians are trying to get health services researchers/clinicians away from thinking primarily in p values and looking more specifically at the point estimates and the upper/lower bounds of these estimates. For example, take for example a point estimate/risk ratio of 1.45, with an upper bound of 1.94 and a lower bound of 0.98. Obviously, the p value for this association is going to be greater than 0.05 because it is crossing 1.00, but it (in our view) is wrong to say that there is no effect – it’s just that not all data points are compatible with the data. It is more informative to say that “Our results suggest a 45% increase in risk of an avoidable ED visit in patients who do not attend a PCMH. Nonetheless, a risk difference ranging from a 2% decrease, a small negative association, to a 94% increase, a substantial positive association, is also reasonably compatible with our data”. Interpreting the point estimate while acknowledging its uncertainty is a better way to keep us from making false declarations of “no difference” and from making overconfident claims.

• However, to your point, this is still not the predominant view and we have revised our submission to only include p values of 0.05 or lower.

Table 4: There are several bolded items that are not statistically significant

• Thank you - see response above. All values in bold now only correspond to associations with p values < 0.05

Table 4: p-values are a probabilities and should not be negative

• The (-) negative values in table 4 (revised table 4 and table 5) do not refer to p-values. They refer to difference in predicted probabilities between the groups. So negative values suggest a “PCMH effect” of lower ED rate, whereas positive values represent the opposite (e.g., PCMH outcomes are worse than non-PCMH outcomes). The difference in margins were tested and p-values are reported for differences that have p values less than 0.05.

Submitted filename: PONE-D-22-04552_ReviewerResponses.docx

Click here for additional data file.

2 Aug 2022

9 Aug 2022

Reviewer #1: Summary and Overall Impression

Reviewer #1: Overall well written and is of great value adding to the understanding of the impact of the PCMH by assessing geographic proximity and patient outcomes.

The introduction is much more focused and reads well.

Methods again appear appropriately selected and applied.

Results appears in order well-presented.

Discussion reads well and addresses limitations and outlines future work.

Minor

Pg 11 Line 21 ..provider ? ‘over’ another.

• Thank you for this observation. This sentence was actually from the original submission, not from revision 1. No changes were made as the sentence had been removed in our resubmission.

Pg 14 Line 15 ..care that ? ‘is’ considered…

• Thank you for this observation. The sentence (Page 8, Line 14 in methods section) has been amended and now reads: “…particularly care that is considered to be of higher quality”

Reviewer #2: Summary and Overall Impression

Reviewer #2: 1. Rationale for the study improved with the revision but participant rationale is needed. Specifically, it is not clear why only individuals with ADD and asthma were included since the study evaluates health disparities by geographic proximity and avoidable ED visits. Examples/suggestions:

• Thank you for mentioning this. We have amended the Methods section (page 8 of Marked Copy, Paragraph 2 and Paragraph 3) to include an additional information on the HEDIS measures and what they are and why they are important.

• Our original submission specified that these two measures (ADD, MMA) were chosen because they represent two disease conditions that are often treated in hospital EDs for care that could have been obtained by primary care providers. We have added an additional sentence (Page 8, lines 19 – 21) that re-emphasize this point.

a. Abstract: In the background, add the need for studying ADD, Asthma; in the objective, add “potentially avoidable ED visits for children with ADD and asthma”. The conclusion (and manuscript) should state ADD and asthma since not all preexisting illnesses/chronic diseases were studied.

• We have amended the Objective sentence in the abstract. It now reads “To examine the association between geographic proximity to primary care providers versus hospitals and risk of avoidable and potentially avoidable ED visits among children with pre-existing diagnosis of attention-deficit/hyperactivity disorder or asthma. “

b. Introduction: (pg. 2 line 7), provide rationale for ADD and asthma (and perhaps omit homeless and diabetes) as there are many other chronic diseases that utilize EDs

• Thank you – we have left the opening paragraph alone (as it is a framing paragraph), but have modified the 5th paragraph of the introduction section (page 5, lines 6 – 9) so that the contextual information about why our study on risk adjustment included these two metrics is clear.

• Additional information was also provided in the methods section (page 8).

2. Abstract: The conclusion is difficult to understand. Suggest 2-3 sentences stating what authors found and conclude regarding the study aim and results as well as its impact. The summary in authors’ cover letter provides much clearer conclusions.

• Thank you for this observation. We have revised the conclusion section of the abstract. It now reads “In several health care performance evaluations, patient-centered medical homes have not been found to reduce differences in hospital utilization for conditions that are treatable in primary care settings among children with chronic illnesses. Analytical approaches that also consider geographic proximity to health care services can identify performance benefits of medical homes. Expanding risk-adjustment models to also include geographic data would benefit ongoing quality improvement initiatives.”

Submitted filename: PONE-D-22-04552_R2_ReviewerResponses.docx

Click here for additional data file.

16 Aug 2022

Geographic proximity to primary care providers as a risk-assessment criterion for quality performance measures

PONE-D-22-04552R2

Dear Dr. Bell,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Jingjing Qian

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Thanks for the additional revision. All comments have been addressed.

Reviewers' comments:

26 Aug 2022

PONE-D-22-04552R2

Geographic proximity to primary care providers as a risk-assessment criterion for quality performance measures

Dear Dr. Bell:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Jingjing Qian

Academic Editor

PLOS ONE