Extent of Health Care Fragmentation in Different Payer Populations: Evidence from the Hudson Valley of New York

Abstract

Health care fragmentation occurs when patients see multiple ambulatory providers, but no single provider accounts for a substantial proportion of visits. Most previous studies have measured fragmentation in Medicare, which may not be generalizable. The study objective was to compare the extent of fragmented ambulatory care across commercially insured, Medicare, and Medicaid populations. The authors conducted a cross-sectional study of adults (N = 256,047) in the Hudson Valley region of New York, who were continuously insured (through 5 commercial payers, Medicare, or Medicaid), were attributed to a primary care physician, and had ≥4 ambulatory visits in the study year. Fragmentation was calculated using a reversed Bice-Boxerman Index, which captures both dispersion of care across providers and the relative share of visits by each provider. Chi-square tests, t tests, and correlation were used to compare patient characteristics and patterns of care across payers. Patients with Medicare had more chronic conditions (45% had ≥5 chronic conditions) than patients with commercial insurance (20%) or Medicaid (23%) (P < 0.01). However, mean fragmentation scores were comparable across all 3 payer populations: 0.73 (commercial insurance), 0.74 (Medicare), 0.72 (Medicaid). The correlation between number of chronic conditions and fragmentation was weak across payers, ranging from r = 0.004 to r = 0.12. If the extent of fragmentation does not vary with payer type or with the number of chronic conditions, it suggests that the causes of fragmentation may be more numerous and more complex than medical need alone.

Introduction

American health care is often described as “fragmented.”¹ One type of fragmentation occurs in the ambulatory setting, when patients see multiple providers but no single provider has a substantial portion of the visits.^2,3 Having fragmented care increases the burden of care coordination across those providers and increases the risk of gaps in communication.^4,5 Having more fragmented care has been associated with having more emergency department visits and more avoidable hospitalizations, compared to having less fragmented care.^6

–9

Most studies describing the extent of fragmented ambulatory care have been based on claims for Medicare beneficiaries.^4
–6,8 A recent national study has shown that care provided to Medicare beneficiaries may not be generalizable to commercially insured patients.¹⁰ Similarly, patterns of care for Medicare beneficiaries, who are mostly elderly adults, may not be generalizable to patterns of care for Medicaid beneficiaries, who are typically younger.

In addition, the relationship between health care fragmentation and chronic conditions is poorly understood, as most previous studies have adjusted for chronic conditions rather than explored them directly.^6,8 If health care fragmentation is correlated with number of chronic conditions (that is, patients with more chronic conditions consistently have more fragmented care than patients with fewer chronic conditions), then much of health care fragmentation may be clinically appropriate. If, on the other hand, health care fragmentation is not correlated with number of chronic conditions, then this may suggest that the causes of fragmentation may be more complex than medical need alone.

This study sought to compare the extent of health care fragmentation among patients with commercial insurance, Medicare, and Medicaid, and to determine the relationship between number of chronic conditions and the extent of fragmentation within each payer population.

Methods

Overview

The study team conducted a cross-sectional study of adults who sought care from physicians in the Hudson Valley region of New York in 2010. Institutional Review Boards at Weill Cornell Medical College and Kingston Hospital approved the protocol.

Setting

The Hudson Valley consists of 7 counties immediately north of New York City (Dutchess, Orange, Putnam, Rockland, Sullivan, Ulster, and Westchester). At the time of the study, most health care in the region was delivered by physicians in small- and medium-sized private practices, using fee-for-service reimbursement from multiple payers.¹¹

Data sources

The study team used administrative claims data for 2010 from 5 commercial payers (Aetna, Capital District Physicians' Health Plan, Hudson Health Plan, MVP Health Care, UnitedHealthcare), which had been aggregated by a third-party company. The team separately used administrative claims for 2010 from Medicare and Medicaid. These 7 payers together cover approximately 75% of insured adults in this region.

Variables

The claim-level variables used were: patient study ID, patient date of birth, patient sex, date of service, rendering provider ID, Current Procedural Terminology (CPT) codes, and International Classification of Diseases, Ninth Revision (ICD-9) codes.

Ambulatory visits were identified using CPT codes, using a modified version of the definition by the National Committee for Quality Assurance.¹² Modifications restricted the definition to evaluation and management visits for adults that would take place in an office setting, excluding management-only visits (eg, dialysis, chemotherapy, physical therapy) and excluding non–office-based visits (eg, home visits, visits in nursing facilities). This definition also excluded emergency department visits.

ICD-9 codes were used to calculate the total number of chronic conditions for each patient, using the 27 chronic conditions defined by the Centers for Medicare & Medicaid Services' Chronic Conditions Warehouse (Supplementary Data S1).¹³ The 2 dementia categories were combined into 1 category to avoid double counting, yielding 26 unique chronic condition categories.

The study team calculated a fragmentation score for each patient, based on the Bice-Boxerman Index (Supplementary Data S2).² This index was selected because it captures both “dispersion” (the spread of a patients' care across multiple providers)³ and “density” (the relative share of visits by each provider).³ The Bice-Boxerman Index has been used previously to predict emergency department visits, hospitalizations, and costs of care.⁶ It also has been found to be highly correlated with other measures of fragmentation, such as the Herfindahl Index, the Usual Provider of Care Index, and the Sequential Continuity Index.¹⁴

Values of the original Bice-Boxerman Index range from 0 (least continuity, or most fragmentation) to 1 (most continuity, or least fragmentation). Patients receive a raw score of 0 if each visit is with a different provider and a raw score of 1 if all ambulatory visits are with the same provider. Other patterns of visits and providers receive intermediate scores; patterns with high dispersion (many providers) and low density (a relatively low proportion of visits by each provider) receive worse scores (indicating more fragmentation) than patterns with low dispersion and high density. To facilitate interpretation, the study team reversed the scores (calculating 1 minus the Bice-Boxerman Index score) so that higher values reflect more fragmentation. For the remainder of the paper, these transformed scores are referred to simply as scores.

Statistical analysis

The study team analyzed commercial claims, Medicare claims, and Medicaid claims side-by-side, rather than combining them into a single data set, because of the requirements of the data use agreements. The study cohort was defined by first identifying in the claims primary care physicians (general internists and family medicine physicians) who had billing zip codes in the Hudson Valley region. The team then determined which adults (≥18 years old for commercial plans and Medicaid, or ≥65 years old for Medicare) could be attributed to those primary care physicians, based on their claims, using previously defined attribution logic.¹⁵ Adults in commercial plans who had insurance products from Medicare or Medicaid lines of business were excluded in order to avoid misclassification.

The study team included only those adults who were continuously enrolled for the full calendar year. The sample was further restricted to those who had ≥4 ambulatory visits that year (with non-missing providers), because calculating fragmentation with fewer than 4 visits can yield statistically unstable estimates.⁸ The team excluded those who had outlier observations (>99.9^th percentile) for the number of ambulatory visits or number of unique ambulatory providers, because some of those observations seemed erroneous. The subset of those with ≥4 ambulatory visits and without outliers represented 67.1% of commercially insured patients, 90.4% of Medicare beneficiaries, and 75.7% of Medicaid beneficiaries (Supplementary Table S1). For the remainder of the paper, both “patients” and “beneficiaries” will be referred to as “patients.”

Patients within each payer were characterized according to age, sex, and number of chronic conditions (0, 1–2, 3–4, ≥5). The study team also characterized patients' patterns of ambulatory care, generating descriptive statistics within payer for the number of ambulatory visits, number of unique providers, proportion of ambulatory visits with the most frequently seen provider, and fragmentation scores. The team compared the distributions of patient characteristics and patterns of ambulatory care across payers using pairwise t tests and chi-square tests.

In order to understand the relationships among fragmentation scores, payer type, and chronic conditions, the study team calculated the average fragmentation score, stratified by number of chronic conditions and by payer. The team calculated the correlation coefficient between number of chronic conditions and fragmentation score within payer, and also determined the top 10 chronic conditions, with the prevalence of each condition, by payer.

P values less than 0.05 were considered to be statistically significant. Software used for the analyses were SAS version 9.4 (SAS Institute Inc., Cary, NC), Stata version 14 (StataCorp LLC, College Station, TX), and R version 3.3.2 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Study sample

The study team identified 105,566 patients with commercial insurance, 125,955 patients with Medicare, and 24,526 patients with Medicaid who met the study's inclusion criteria (adults who could be attributed to a primary care physician in the Hudson Valley, who were continuously enrolled with their payer, and who had ≥4 ambulatory visits that year).

Patient characteristics

As expected, compared to patients with commercial insurance or Medicaid, patients with Medicare were significantly older (Table 1). More than half of patients in each payer type were women. As expected, Medicare patients had the most chronic conditions compared to patients with commercial insurance or Medicaid (Table 1).

Table 1.

Patient Characteristics and Patterns of Ambulatory Care by Payer ^*

	Commercially insured patients (N = 105,566)	Medicare beneficiaries (N = 125,955)	Medicaid beneficiaries (N = 24,526)
Patient characteristics
Age, years, Mean (SD)	53 (17)	77 (7)	52 (19)
Sex, Female, %	60	61	66
Number of chronic conditions,† %
0	17%	1%	18%
1 or 2	34%	17%	33%
3 or 4	29%	37%	26%
≥5	20%	45%	23%
Patterns of ambulatory care
Number of visits per patient
Mean (SD)	9.7 (6.6)	14.5 (9.6)	15.0 (10.9)
Median	8	12	12
Range	4 to 55	4 to 76	4 to 105
Number of unique providers per patient
Mean (SD)	4.5 (2.3)	5.7 (3.1)	5.4 (2.9)
Median	4	5	5
Range	1 to 17	1 to 22	1 to 22
Proportion of visits with the most frequently seen provider
Mean (SD)	0.48 (0.20)	0.45 (0.19)	0.47 (0.18)
Median	0.44	0.40	0.44
Range	0.09 to 1.00	0.08 to 1.00	0.09 to 1.00
Fragmentation score‡
Mean (SD)	0.73 (0.22)	0.74 (0.20)	0.72 (0.20)
Median	0.80	0.80	0.77
Range	0.00 to 1.00	0.00 to 1.00	0.00 to 1.00

This study included only patients with ≥4 visits in the study year. All comparisons across payers (by t tests for continuous variables and chi-square tests for dichotomous and categorical variables) were statistically significant (P < 0.01).

†

The study team calculated the number of chronic conditions per person out of 27 conditions defined by the Centers for Medicare & Medicaid Services' Chronic Conditions Warehouse. Two overlapping conditions were combined, yielding 26 unique chronic conditions.

‡

The fragmentation score is a reversed Bice-Boxerman Index.² Higher scores reflect more fragmented care.

SD, standard deviation.

Patterns of ambulatory care by payer

The mean number of ambulatory visits per patient and mean number of unique providers per patient are shown for each population in Table 1. On average, nearly half of visits were with the most frequently seen provider in each payer population (Table 1). As a result, the average fragmentation scores were similar across payers.

Chronic conditions and fragmentation by payer

The specific chronic conditions affecting patients were similar across payers, with 6 specific conditions appearing on the top 10 list for all 3 payers, and 4 other conditions appearing on the top 10 list for 2 of the 3 payers (Table 2). The prevalence of these conditions was highest in the Medicare population.

Table 2.

Top 10 Chronic Conditions, with the Prevalence of Each Condition, by Payer

	Commercial	Medicare	Medicaid
1	Hyperlipidemia (54%)	Hypertension (79%)	Hypertension (52%)
2	Hypertension (49%)	Hyperlipidemia (68%)	Hyperlipidemia (41%)
3	Diabetes (21%)	Ischemic heart disease (30%)	Diabetes (27%)
4	Arthritis (RA or OA) (19%)	Arthritis (RA or OA) (29%)	Arthritis (RA or OA) (19%)
5	Anemia (18%)	Diabetes (29%)	Ischemic heart disease (17%)
6	Hypothyroidism (17%)	Cataracts (26%)	Asthma (17%)
7	Cataracts (13%)	Anemia (19%)	Anemia (17%)
8	Ischemic heart disease (13%)	Glaucoma (17%)	Depression (12%)
9	Glaucoma (10%)	Hypothyroidism (17%)	Chronic kidney disease (11%)
10	Asthma (10%)	COPD (15%)	Hypothyroidism (10%)

COPD, chronic obstructive pulmonary disease; OA, osteoarthritis; RA, rheumatoid arthritis.

The number of chronic conditions was weakly correlated with the fragmentation score; the correlation coefficients were 0.004 in the commercial population, 0.12 in Medicare, and 0.04 in Medicaid (Table 3).

Table 3.

Fragmentation Score by Payer, Overall and by Number of Chronic Conditions

	Commercially insured patients	Medicare beneficiaries	Medicaid beneficiaries
Mean fragmentation score (SD)
Overall	0.73 (0.22)	0.74 (0.20)	0.72 (0.20)
By number of chronic conditions
0	0.76 (0.21)	0.74 (0.22)	0.74 (0.20)
1–2	0.73 (0.23)	0.71 (0.23)	0.70 (0.22)
3–4	0.71 (0.22)	0.73 (0.21)	0.71 (0.19)
≥5	0.75 (0.18)	0.77 (0.17)	0.74 (0.17)
Correlation between number of chronic conditions and fragmentation score
r	0.004	0.12^*	0.04^*

^*P < 0.05.

SD, standard deviation.

Discussion

This study of more than 250,000 insured adults in a 7-county region found that health care fragmentation is pervasive, affecting patients regardless of payer and regardless of the number of chronic conditions. For example, the median Medicare beneficiary had 12 visits with 5 different providers, and the median Medicaid beneficiary also had 12 visits with 5 providers, even though Medicaid beneficiaries were significantly younger and had fewer chronic conditions. Commercially insured patients had fewer visits and fewer providers (a median of 8 visits to 4 providers), but their most frequently seen provider still accounted for only about half of the visits, giving them fragmentation scores that were just as high as their counterparts in Medicare and Medicaid.

The study team is not aware of other studies that have compared patterns of ambulatory care across payers. Therefore, this study adds to the literature by measuring fragmentation in multiple payer populations and finding that fragmentation is not an issue isolated to the Medicare population (in which it has been measured previously).^6,8,14 This study also adds to the literature with its finding that the extent of fragmentation is only weakly correlated with the number of chronic conditions. This is important, because it suggests that the causes of fragmentation may be more numerous and more complex than medical need alone.

What drives fragmentation for any payer population is not known. The study team cannot determine from the data if the patterns observed are patient driven, physician driven, health system driven, payer driven, or some combination of these. The team also cannot determine to what extent these patterns are intentional or unintentional. That is, the team cannot determine if they are the reflection of patients' preferences and/or physician referral (which would be intentional), or whether at least some of the care reflects more chaotic patterns of visits, in which some patients are having difficulty managing access to their preferred physicians (which would be more unintentional). Difficulty with access could be related to patient factors (eg, whether patients' schedules can accommodate available appointments with their usual providers), provider factors (eg, limited time seeing patients or the size of the provider's panel), health system factors (eg, whether the system enables timely follow-up appointments with the previously seen provider), payer factors (eg, narrow provider networks), or some combination of these.

Future research should explore the drivers of fragmented care in all 3 payer populations, both to determine more precisely why fragmented care occurs and to determine if there are differences in the drivers of fragmented care by payer type. Another important topic for future research will be the consequences of fragmented care across payers. Previous studies have shown that fragmented care can be associated with suboptimal outcomes, such as avoidable hospitalizations in Medicare patients.⁸ Less work has been done on fragmented care in commercially insured patients, but early studies suggest that more fragmented care is associated with more radiology and diagnostic tests.¹⁶ The consequences of fragmented ambulatory care in Medicaid patients are not yet known.

There are several limitations of this study. First, this study included only adults, and its findings may not be generalizable to children. Second, although the study team stratified by the number of chronic conditions, one cannot rule out residual confounding by severity of disease. Third, this study cannot determine whether the patterns observed are medically appropriate. Fourth, this study used data from 2010 and may not reflect the effects of recent initiatives, such as the diffusion of accountable care organizations. However, these data can serve as a baseline against which future data could be compared. Finally, this study takes place in a 7-county region, which may not be generalizable to other regions. However, the Hudson Valley does have many characteristics – such as having predominantly small- and medium-sized private practices, fee-for-service reimbursement, and multiple payers – that are common to many other regions in the United States.

This study has several strengths. It demonstrates a rare situation in which investigators were able to obtain and analyze data across different payer populations in the same region. The claims data used reflect complete records of ambulatory visits across health systems. The study also used a previously validated measure of fragmentation and previously established techniques for ensuring statistical reliability of fragmentation scores (eg, including only those patients with ≥4 visits).⁸

This study has potential implications for patients, providers, and payers alike. Providers are increasingly being asked to take on clinical and financial responsibility for populations of patients (not just individual patients) through value-based purchasing contracts.^17,18 It may not be possible to succeed under these contracts without tracking and improving patients' patterns of ambulatory care (including fragmented care). Payers have tried several strategies to incentivize certain patterns of care (eg, requiring referrals,¹⁹ creating tiered networks,²⁰ creating narrow networks²¹), and the public has had mixed reactions to these programs, sometimes finding them too restrictive.^19
–21 To the study team's knowledge, payers have not tried incentivizing concentrated (rather than fragmented) care; such an approach might serve to minimize excess use of additional providers while preserving patient choice.

In conclusion, this study found that fragmented ambulatory care occurs in multiple payer populations and is only weakly correlated with number of chronic conditions. If the extent of fragmentation does not vary with payer type or with number of chronic conditions, it suggests that the causes of fragmentation may be more numerous and more complex than medical need alone. If future research confirms that highly fragmented care often reflects overuse of care, or care that leads to costly subsequent events and patient harm, then decreasing unnecessary fragmentation may represent a new strategy to improve care.

Footnotes

Acknowledgment

The authors thank HealthlinkNY, the participating payers, and the New York State Department of Health for facilitating access to the data. This work does not necessarily represent the views of the New York State Department of Health or the Commonwealth Fund.

Author Disclosure Statement

The authors declare that there are no conflicts of interest. The authors received the following financial support: This study was funded by The Commonwealth Fund (grant #20140960). The funder had no role in the study design, data collection, data analysis, interpretation, preparation or approval of the manuscript.

Supplementary Material

Supplementary Data S1

Supplementary Data S2

Supplementary Table S1

References

Elhauge

, ed. The fragmentation of U.S. Health Care. Oxford, United Kingdom: Oxford University Press, 2010.

Bice

, Boxerman

. A quantitative measure of continuity of care. Med Care, 1977; 15:347–349.

Jee

, Cabana

. Indices for continuity of care: a systematic review of the literature. Med Care Res Rev, 2006; 63:158–188.

Pham

, O'Malley

, Bach

, Saiontz-Martinez

, Schrag

. Primary care physicians' links to other physicians through Medicare patients: the scope of care coordination. Ann Intern Med, 2009; 150:236–242.

Pham

, Schrag

, O'Malley

, Wu

, Bach

. Care patterns in Medicare and their implications for pay for performance. N Engl J Med, 2007; 356:1130–1139.

Hussey

, Schneider

, Rudin

, Fox

, Lai

, Pollack

. Continuity and the costs of care for chronic disease. JAMA Intern Med, 2014; 174:742–748.

Liu

, Einstadter

, Cebul

. Care fragmentation and emergency department use among complex patients with diabetes. Am J Manag Care, 2010; 16:413–440.

Nyweide

, Anthony

, Bynum

, et al. Continuity of care and the risk of preventable hospitalization in older adults. JAMA Intern Med, 2013; 173:1879–1885.

van Walraven

, Oake

, Jennings

, Forster

. The association between continuity of care and outcomes: a systematic and critical review. J Eval Clin Pract, 2010; 16:947–956.

10.

Cooper

, Craig

, Gaynor

, Van Reenen

. for the Health Care Pricing Project. The price ain't right? Hospital prices and health spending on the privately insured. 2015. www.healthcarepricingproject.org/papers/paper-1 Accessed June 12, 2018 .

11.

Stuard

, Blair

. Interval examination: regional transformation of care delivery in the hudson valley. J Gen Intern Med, 2011; 26:1371–1373.

12.

National Committee for Quality Assurance. HEDIS Volume 2: Technical Specifications, 2015. www.ncqa.org/HEDISQualityMeasurement/HEDISMeasures/HEDIS2015.aspx Accessed June 12, 2018 .

13.

Centers for Medicare and Medicaid Services. Chronic Conditions Data Warehouse, 2015. https://www.ccwdata.org Accessed June 12, 2018 .

14.

Pollack

, Hussey

, Rudin

, Fox

, Lai

, Schneider

. Measuring care continuity: a comparison of claims-based methods. Med Care, 2016; 54:e30–e34.

15.

Kern

, Edwards

, Kaushal

. The patient-centered medical home, electronic health records, and quality of care. Ann Intern Med, 2014; 160:741–749.

16.

Kern

, Seirup

, Casalino

, Safford

. Healthcare fragmentation and the frequency of radiology testing. J Gen Intern Med, 2017; 32; 175–181.

17.

Burwell

. Setting value-based payment goals–HHS efforts to improve U.S. health care. N Engl J Med, 2015; 372:897–899.

18.

Medicaid and CHIP Payment and Access Commission. Value-based purchasing, 2016. https://www.macpac.gov/subtopic/value-based-purchasing/ Accessed June 12, 2018 .

19.

Ferris

, Chang

, Blumenthal

, Pearson

. Leaving gatekeeping behind–effects of opening access to specialists for adults in a health maintenance organization. N Engl J Med, 2001; 345:1312–1317.

20.

Sinaiko

. Variations in patient response to tiered physician networks. Am J Manag Care, 2016; 22:420–425.

21.

Giovannelli

, Lucia

, Corlette

. Health Policy Brief: Network adequacy. July 28, 2016. https://www.healthaffairs.org/do/10.1377/hpb20160728.898461/full/ Accessed June 12, 2018 .

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.03 MB