Coordination of Care Is Associated With Survival and Health Care Utilization in a Population-Based Study of Patients With Cirrhosis

Abstract

BACKGROUND & AIMS:

Improving care coordination for patients with high-intensity specialty care needs, such as cirrhosis, can increase quality of healthcare and reduce utilization. We examined the relationship between care concentration and risk of hospitalization for patients with cirrhosis.

METHODS:

We performed a retrospective cohort study of 26,006 Medicare enrollees with cirrhosis with more than 4 outpatient visits over 180 days. We collected data on 2 validated measures of care concentration: the usual provider of care (UPC) index, a measure of the proportion of a patient’s total visits that is with their most regularly seen provider, and the continuity of care (COC) index, a measure of care density and dispersion. Both use a scale of 0 to 1. Time to death or liver transplantation was evaluated using a multivariable Cox proportional hazards model. Hospital days and 30-day readmissions per person-year were evaluated in negative binomial models.

RESULTS:

The median COC score was 0.40 (interquartile range, 0.26–0.60) and the median UPC was 0.60 (interquartile range, 0.50–0.80). Increasing care concentration (based on COC and UPC index scores) were associated with increased mortality and hospitalization. The highest 25th percentile of COC and UPC scores were associated with adjusted hazard ratios for mortality of 1.20 (95% CI, 1.10–1.31) and 1.14 (95% CI, 1.06–1.24), adjusted incidence rate ratios for hospital days of 1.12 (95% CI, 1.02–1.23) and 1.10 (95% CI, 1.01–1.20), and adjusted incidence rate ratios for readmissions of 1.19 (95% CI, 1.06–1.34) and 1.12 (95% CI, 1.00–1.25), respectively.

CONCLUSIONS:

Based on a study of Medicare enrollees, care concentration is low among patients with cirrhosis. However, increased concentration is associated with increased mortality and increased healthcare utilization. These data indicate that, to optimize outcomes for persons with cirrhosis, team-based care might be necessary.

Cirrhosis is increasingly common and morbid.^1,2 Over the past decade, its prevalence has doubled and mortality has risen by 65%.^3,4 The rate of hospitalization for cirrhosis now exceeds that of congestive heart failure.^2,5 The cost of liver-related health care is now estimated at $9.8 billion annually,^2,6 half of which is incurred during hospitalizations alone.⁷ Hospitalizations and readmissions in particular are a key driver of both cost and poor outcomes.^8,9 Unfortunately, there are few interventions proven to prevent hospitalization for persons with cirrhosis.⁹

Improving the coordination of care for patients with chronic high-intensity specialty care needs, such as cirrhosis, is increasingly recognized as an important tool to improve quality and reduce health care utilization. Provider continuity is central to effective care coordination.¹⁰ Continuity refers to the degree to which a patient’s medical visits are concentrated among providers. Continuity among providers has been associated with a reduction in preventable hospitalizations, emergency room visits, and mortality among patients with diabetes, inflammatory bowel disease, and hepatocellular carcinoma, and children with complex conditions.^11–14 As a function of its medical complexity and the patient’s frequently disordered social context, cirrhosis is prone to fragmentation or scattered care.⁸ Understanding the factors that contribute to this rate of utilization is essential to inform the design of future interventions to reduce overutilization.⁶

Herein, using a nationally representative cohort of Medicare enrollees with cirrhosis, we define the current care team structure experienced by patients with cirrhosis, including continuity and concentration of care. In this study, we determine the relationship between continuity in health care and the risk of hospitalization for patients with cirrhosis.

Materials and Methods

Study Population

We examined data from a 20% random sample of U.S. Medicare enrollees, 18 years of age or older, with cirrhosis (defined by International Classification of Diseases-Ninth Revision [ICD-9] codes 571.2, 571.5, or 571.6) from 2010 to 2014 (Supplementary Figure 1) using a coding algorithm validated for Medicare data.¹⁵ This ICD-9—based algorithm has been validated in Medicare claims with a positive predictive value of 88% and a sensitivity of 67% (Supplementary Table 1).¹⁵ Continuous Part A and B coverage was required for a minimum of 1 year before cirrhosis diagnosis to 90 days after, and Part D coverage for ≥90 days before and after diagnosis. To ensure that we captured incident diagnoses of cirrhosis complications, we included patients with ≥1 year of follow up before their cirrhosis diagnosis. We further limited our dataset to subjects with ≥4 outpatient visits to primary care, gastroenterology or hepatology, cardiology, or hematology or oncology within 90 days before or after entry into the cohort to reduce the bias of limited observations on calculating continuity of care (COC) measures; a cutoff of 4 visits was used for consistency with published studies using similar measures.¹¹ To account for a potential selection bias that results from excluding patients with <4 outpatient visits, we examined differences in outcomes among the excluded cohort as compared with the study cohort (Supplementary Table 2). Patients who received a liver transplant during the year before diagnosis or 90 days after were excluded. Last, subjects needed to survive for ≥90 days after cirrhosis diagnosis in order to achieve adequate stability in the cohort. All subjects were followed beginning 90 days after the initial cirrhosis date until death, transplant, or for an additional 365 days. This study was approved by the institutional review board (#HUM00131544) at the University of Michigan.

Exposure Variable Definitions

Our primary focus was to evaluate the impact of care coordination, as defined by concentration of care, on clinical outcomes. We used 2 validated measures based on the number of office visits a patient has with their providers. Claims-based COC measures often incorporate a combination of density of visits, dispersion of providers, and sequence of provider visits.¹⁶ These measures have been correlated with patient-reported measures of COC, validating their use in administrative data.^17,18 While many different COC measures exist, we focused on 2 of the most commonly used measures, the Usual Provider of Care (UPC) index and the Bice-Boxerman COC index. First, we determined the UPC index, a density index that measures the proportion of a patient’s total visits that is with their most regularly seen provider.¹⁹ Second, we calculated the COC index, a dispersion index, that measures care density and dispersion, using the sum of the squared fractions of visits by all providers for that patient (Supplementary Figure 2).^16,20,21 Both measures range from 0 to 1; a score of 1 suggests complete concentration of care, a marker of more coordinated care. For example, if a patient had a total of 8 visits over a 1-year period split evenly between 2 providers, they would have a COC of 0.429 and a UPC of 0.5. On the other hand, if that same patient had 7 visits with 1 provider, and 1 visit with a second provider, they would have a COC of 0.75 and a UPC of 0.875.¹¹ Unique providers were identified using the National Provider Index and cross-walked to taxonomy codes, using the NPPES (National Plan and Provider Enumeration System) dissemination file, to determine provider specialty. For ease of interpretation, the COC and UPC indices were reported in quartiles.

For a complete description of the cohort and risk adjustment, we also included age, sex, race, urban or rural status, region, Charlson comorbidity index, liver disease etiology, and complications of cirrhosis as covariates. All clinical factors were identified using inpatient, outpatient, or professional claims incurred in the year before cohort entry. As performed in multiple prior studies,²² we also classified a group of patients who had cirrhosis (ICD-9 code: 571.5) but lacked any diagnostic codes for viral hepatitis or alcohol-related use disorder as nonviral non-alcohol-related cirrhosis. Liver disease severity was assessed using a combination of diagnostic (eg, ascites, variceal bleeding), and procedure codes (eg, paracentesis and transjugular intrahepatic portosystemic shunt [TIPS] placement).

Outcomes

Our primary outcome was all-cause mortality or liver transplantation. We had 2 secondary outcomes, hospital-days per person-year and 30-day all-cause readmissions per person-year. All-cause 30-day readmissions were calculated based on eligible hospital discharges, defined by absence of death, transfer to hospice, and at least a 1-day gap between discharge and readmission. Transfers were considered as a single hospitalization.

Statistical Analysis

Time to death or liver transplantation was evaluated using a multivariable Cox proportional hazards model. Hospital days and 30-day readmissions were evaluated in negative binomial models and presented as the incidence per person-years and incidence rate ratio (IRR) with 95% confidence interval (CI). We used negative binomial regression to account for overdispersed count data, which is when the conditional variance exceeds the conditional mean. It has the same mean structure as Poisson regression but includes an extra parameter to model for overdispersion. The correlation between UPC and COC was examined using a Spearman correlation test. In all cases, the P values presented were 2-tailed, with a <.05 threshold for significance. All analyses were performed using Stata/MP version 13 (StataCorp, College Station, TX) and SAS version 9.4 (SAS Institute, Cary, NC).

Results

Patient and clinical characteristics are reported in Table 1. Overall, 26,006 patients with cirrhosis were included. Half of our cohort (48.8%) was male, a majority (76.9%) of patients were white, with a median age of 69 (interquartile range [IQR], 60–76) years. Comorbidities were common with 68% having a Charlson comorbidity index score of at least 3, and 45.1% received social security disability. Cirrhosis-related complications were common, with 7.8% having hepatic encephalopathy, 4.2% having varices, 0.7% having ascites, and 0.2% having spontaneous bacterial peritonitis. Patients saw a median of 8 (IQR, 5–13) providers in the year before cohort entry, of which a median of 1 (IQR, 0–4) provider was a gastroenterologist or hepatologist, hematologist or oncologists, or cardiologist. There were 66 (0.2%) patients with missing regional data and were dropped from the subsequent model results.

Table 1.

Cohort Characteristics

Patient characteristics	Total (N = 26,006)	Continuity of care	Usual provider of care
	—	0.40 (0.26–0.60)a	0.60 (0.50–0.80)a
Age
<65 y	9049 (34.8)	0.43 (0.26–0.62)	0.63 (0.50–0.80)
65–74 y	9655 (37.1)	0.40 (0.25–0.60)	0.60 (0.50–0.80)
>75 y	7302 (28.1)	0.40 (0.27–0.60)	0.63 (0.50–0.80)
Sex
Male	12,702 (48.8)	0.40 (0.27–0.60)	0.6 (0.50–0.80)
Female	13,304 (51.2)	0.40 (0.26–0.60)	0.6 (0.50–0.80)
Race
White	19,985 (76.9)	0.40 (0.27–0.60)	0.60 (0.50–0.80)
Black	3028 (11.6)	0.40 (0.20–0.60)	0.60 (0.50–0.80)
Hispanic	1386 (5.3)	0.43 (0.27–0.62)	0.63 (0.50–0.80)
Other	1607 (6.2)	0.43 (0.27–0.67)	0.63 (0.50–0.82)
Medicaid status
Full	9390 (36.1)	0.44 (0.27–0.67)	0.67 (0.50–0.83)
Partial	3260 (12.5)	0.40 (0.27–0.60)	0.63 (0.50–0.80)
None	13,356 (51.4)	0.40 (0.25–0.59)	0.60 (0.50–0.78)
Disabled	11,730 (45.1)	0.42 (0.26–0.61)	0.63 (0.50–0.80)
Location of residence
Urban	21,036 (80.9)	0.40 (0.25–0.60)	0.60 (0.50–0.80)
Region
Midwest	5248 (20.2)	0.40 (0.24–0.60)	0.60 (0.50–0.80)
Northeast	4821 (18.6)	0.40 (0.24–0.60)	0.60 (0.50–0.80)
South	10,670 (41.1)	0.40 (0.27–0.60)	0.60 (0.50–0.80)
West	5201 (20.1)	0.46 (0.29–0.67)	0.67 (0.50–0.83)
Unknown	66 (0.2)	0.40 (0.29–0.52)	0.60 (0.50–0.75)
Characteristics of cirrhosis
Alcoholic cirrhosis	2612 (10.0)	0.47 (0.28–0.67)	0.67 (0.50–0.83)
Hepatitis C virus-related cirrhosis	5134 (19.7)	0.40 (0.22–0.60)	0.60 (0.50–0.78)
Nonviral nonalcohol-related cirrhosis	18,345 (70.5)	0.42 (0.26–0.64)	0.60 (0.50–0.80)
Ascites	183 (0.7)	0.40 (0.27–0.61)	0.60 (0.45–0.80)
Varices	1082 (4.2)	0.37 (0.24–0.50)	0.57 (0.50–0.75)
Hepatic encephalopathy	2037 (7.8)	0.40 (0.27–0.60)	0.63 (0.50–0.80)
Hepatocellular carcinoma	650 (2.5)	0.36 (0.20–0.51)	0.57 (0.44–0.75)
Transjugular intrahepatic portosystemic shunt	185 (0.7)	0.33 (0.24–0.50)	0.57 (0.43–0.75)
Paracentesis	951 (3.7)	0.38 (0.25–0.54)	0.60 (0.45–0.75)
Spontaneous bacterial peritonitis	40 (0.2)	0.37 (0.30–0.58)	0.63 (0.55–0.77)
Charlson comorbidity index
0	1507 (5.8)	0.47 (0.30–0.67)	0.67 (0.50–0.83)
1	3182 (12.2)	0.44 (0.29–0.67)	0.67 (0.50–0.82)
2	3642 (14)	0.42 (0.27–0.60)	0.63 (0.50–0.80)
3+	17,675 (68)	0.40 (0.25–0.60)	0.60 (0.50–0.80)
End-stage renal disease	1413 (5.4)	0.40 (0.22–0.60)	0.60 (0.50–0.80)

NOTE. Values are median (interquartile range).

^an (%).

Overall, 4482 patients either died or underwent liver transplantation (0.19 events/person-year): 4441 (17.2%) died, and 41 (0.16%) underwent liver transplantation. Over 23,447 person-years of follow-up, hospitalizations accounted for 6.31 d/person-year. In a subset of 10,906 patients with eligible hospital discharges, 3123 (28.6%) patients had at least 1 readmission within 30 days of any admission. A total of 1879 (17.3%) patients had 1 readmission during follow up, while 1124 (11.4%) had 2 or more. An examination of the excluded cohort revealed no significant difference in hospital days or readmissions but a 2% increase in death (Supplementary Table 2). The most common indications for readmission are reported in Supplementary Table 3.

The median COC was 0.40 (IQR, 0.26–0.60). The median UPC was 0.60 (IQR, 0.50–0.80) (Figure 1). The distribution of UPC and COC by patient characteristics are reported in Table 1 without significant patient- or disease-specific differences. UPC and COC were strongly correlated (r = 0.93; P < .001). A higher COC is associated with a higher likelihood of death or liver transplantation, 30-day readmission, and hospital days (Tables 2–4). Similarly, a higher UPC is associated with a higher likelihood of death or liver transplantation and more hospital days but not with 30-day readmissions (Tables 2–4).

Figure 1.

COC and UPC distribution.

Table 2.

Continuity of Care, Usual Provider of Care, and Death or Liver Transplantation

Variable	Continuity of care	Usual provider of care
Percentile
Lowest 25th	ref	ref
25th-50th	1.08 (0.99–1.17)	1.09 (0.99–1.20)
50th-75th	1.14 (1.05–1.24)	1.15 (1.07–1.24)
Highest 25th	1.20 (1.10–1.31)	1.14 (1.06–1.24)
Sex
Male	ref	ref
Female	0.75 (0.71–0.80)	0.75 (0.71–0.80)
Age
<65 y	ref	ref
65–74 y	1.41 (1.29–1.56)	1.41 (1.29–1.56)
≥75 y	2.24 (2.06–2.53)	2.28 (2.06–2.53)
Regiona
Northeast	ref	ref
Midwest	1.20 (1.09–1.32)	1.20 (1.09–1.32)
South	1.11 (1.02–1.21)	1.11 (1.02–1.21)
West	1.04 (0.95–1.15)	1.05 (0.95–1.15)
Race
White	ref	ref
Black	0.94 (0.85–1.04)	0.94 (0.85–1.04)
Hispanic	0.67 (0.57–0.78)	0.67 (0.57–0.78)
Other	0.82 (0.72–0.94)	0.82 (0.72–0.94)
End-stage renal disease	1.69 (1.52–1.89)	1.69 (1.52–1.88)
Medicaid status
None	ref	ref
Full	0.98 (0.91–1.06)	0.98 (0.91–1.06)
Partial	0.98 (0.88–1.08)	0.98 (0.88–1.08)
Charlson comorbidity index	1.10 (1.09–1.11)	1.10 (1.09–1.11)
Hepatitis C virus	0.86 (0.75–0.98)	0.85 (0.74–0.98)
Alcoholic cirrhosis	1.35 (1.25–1.46)	1.35 (1.25–1.46)
Nonalcoholic, nonviral cirrhosis	1.18 (1.05–1.33)	1.18 (1.05–1.33)
Ascites	1.24 (0.97–1.58)	1.25 (0.98–1.59)
Varices	0.74 (0.64–0.86)	0.74 (0.64–0.86)
Hepatic encephalopathy	1.11 (1.01–1.23)	1.11 (1.01–1.23)
Hepatocellular carcinoma	1.88 (1.64–2.15)	1.88 (1.64–2.15)
Transjugular intrahepatic portosystemic shunt	0.97 (0.71–1.33)	0.97 (0.71–1.33)
Spontaneous bacterial peritonitis	1.41 (0.88–2.27)	1.40 (0.88–2.25)
Paracentesis	1.70 (1.51–1.92)	1.71 (1.51–1.93)
Location
Rural	ref	ref
Urban	0.92 (0.85–0.99)	0.92 (0.85–0.99)
Disability	1.01 (0.93–1.10)	1.01 (0.93–1.10)

Values are hazard ratio (95% confidence interval).

^ahttps://www2.census.gov/geo/pdfs/maps-data/maps/reference/us_regdiv.pdf

Table 4.

Care Coordination and 30-Day Readmissions

Variable	Continuity of care	Usual provider of care
Percentile
Lowest 25th	ref	ref
25th-50th	1.18 (1.05–1.33)	1.04 (0.91–1.19)
50th-75th	1.14 (1.01–1.28)	1.11 (1.00–1.23)
Highest 25th	1.19 (1.06–1.34)	1.12 (1.00–1.25)
Sex
Male	ref	ref
Female	1.03 (0.94–1.12)	1.02 (0.94–1.12)
Age
<65 y	ref	ref
65–74 y	0.89 (0.79–1.01)	0.89 (0.79–1.01)
≥75 y	0.88 (0.76–1.01)	0.88 (0.76–1.02)
Regiona
Northeast	ref	ref
Midwest	0.97 (0.85–1.11)	0.98 (0.86–1.11)
South	0.88 (0.78–0.99)	0.88 (0.79–0.99)
West	0.91 (0.79–1.04)	0.92 (0.80–1.05)
Race
White	ref	ref
Black	1.18 (1.04–1.35)	1.19 (1.05–1.35)
Hispanic	0.84 (0.69–1.03)	0.85 (0.69–1.03)
Other	0.86 (0.71–1.05)	0.87 (0.71–1.06)
End-stage renal disease	1.57 (1.35–1.82)	1.56 (1.35–1.81)
Medicaid status
None	ref	ref
Full	1.08 (0.98–1.20)	1.08 (0.98–1.20)
Partial	1.08 (0.94–1.24)	1.08 (0.94–1.24)
Charlson comorbidity index	1.08 (1.07–1.09)	1.08 (1.07–1.09)
Hepatitis C virus	0.83 (0.69–1.00)	0.84 (0.70–1.01)
Alcoholic cirrhosis	1.28 (1.15–1.44)	1.29 (1.15–1.44)
Nonviral non alcohol-related cirrhosis	1.05 (0.89–1.24)	1.06 (0.90–1.25)
Ascites	0.80 (0.52–1.21)	0.79 (0.52–1.21)
Varices	0.79 (0.65–0.97)	0.80 (0.65–0.97)
Hepatic encephalopathy	1.18 (1.03–1.36)	1.19 (1.03–1.36)
Hepatocellular carcinoma	1.01 (0.78–1.31)	1.02 (0.79–1.33)
Transjugular intrahepatic portosystemic shunt	0.96 (0.65–1.43)	0.97 (0.65–1.44)
Spontaneous bacterial peritonitis	1.19 (0.54–2.61)	1.23 (0.56–2.69)
Paracentesis	1.29 (1.06–1.57)	1.29 (1.07–1.57)
Location
Rural	ref	ref
Urban	1.17 (1.04–1.31)	1.17 (1.04–1.31)
Disability	1.08 (0.96–1.21)	1.08 (0.96–1.21)

Values are incidence rate ratio (95% confidence interval).

^ahttps://www2.census.gov/geo/pdfs/maps-data/maps/reference/us_regdiv.pdf

In a multivariable Cox proportional hazards model for COC, death or liver transplantation was positively associated with older age (for age 65–74 years: hazard ratio [HR], 1.41; 95% CI, 1.29–1.56; for age >75 years: HR, 2.24; 95% CI, 2.06–2.53), and inversely associated with female sex (HR, 0.75; 95% CI, 0.71–0.80) and Hispanic race (HR, 0.67; 95% CI, 0.57–0.78), when controlling for region. Death or liver transplantation was also positively associated with end-stage renal disease (HR, 1.69; 95 CI, 1.52–1.89), a higher Charlson comorbidity index (HR, 1.10; 95% CI, 1.09–1.11), alcohol-related cirrhosis (HR, 1.35; 95% CI, 1.25–1.46), nonviral non-alcohol-related cirrhosis (HR, 1.18; 95% CI, 1.05–1.33), hepatic encephalopathy (HR, 1.11; 95% CI, 1.01–1.23), hepatocellular carcinoma (HR, 1.88; 95% CI, 1.64–2.15), and paracentesis (HR, 1.70, 95% CI, 1.51–1.92), and inversely associated with positive hepatitis C virus status (HR, 0.86, 95% CI, 0.75–0.98) and varices (HR, 0.74; 95% CI, 0.64–0.86) (Table 2). Similar associations exist in the UPC model for death or liver transplantation (Table 2).

In a multivariable negative binomial model for COC, hospital days was associated with age >75 years (IRR, 1.44; 95% CI, 1.28–1.62), black race (IRR, 1.24, 95% CI, 1.11–1.37), end-stage renal disease (IRR, 2.09; 95% CI, 1.80–2.42), full Medicaid status (IRR, 1.16; 95% CI, 1.07–1.25), a higher Charlson comorbidity index (IRR, 1.12; 95% CI, 1.11–1.14), alcohol-related cirrhosis (IRR, 1.41; 95% CI, 1.29–1.55), nonviral non–alcohol-related cirrhosis (IRR, 1.25; 95% CI, 1.10–1.43), TIPS (IRR, 1.43; 95% CI, 1.20–1.71), and spontaneous bacterial peritonitis (IRR, 2.32; 95% CI, 1.05, 5.14), and inversely associated with varices (IRR, 0.70; 95% CI, 0.60–0.82) and a positive hepatitis C virus status (IRR, 0.79; 95% CI, 0.69–0.91) (Table 3). Similar associations exist between UPC and hospital days (Table 3).

Table 3.

Care Coordination and Total Hospital Days

Variable	Continuity of care	Usual provider of care
Percentile
Lowest 25th	ref	ref
25th-50th	1.00 (0.91–1.09)	0.99 (0.89–1.09)
50th-75th	1.05 (0.96–1.15)	1.09 (1.00–1.18)
Highest 25th	1.12 (1.02–1.23)	1.10 (1.01–1.20)
Sex
Male	ref	ref
Female	0.98 (0.92,1.05)	0.98 (0.92–1.05)
Age
<65 y	ref	ref
65–74 y	1.11 (1.00–1.22)	1.11 (1.00–1.22)
≥75 y	1.44 (1.28–1.62)	1.44 (1.28–1.62)
Regiona
Northeast	ref	ref
Midwest	0.93 (0.84–1.03)	0.93 (0.84–1.03)
South	0.93 (0.85–1.01)	0.93 (0.85–1.01)
West	0.81 (0.73–0.90)	0.81 (0.73–0.90)
Race
White	ref	ref
Black	1.24 (1.11–1.37)	1.24 (1.11–1.37)
Hispanic	0.72 (0.62–0.84)	0.72 (0.62–0.84)
Other	0.94 (0.82–1.08)	0.94 (0.82–1.08)
End-stage renal disease	2.09 (1.80–2.42)	2.08 (1.80–2.42)
Medicaid status
No	ref	ref
Full	1.16 (1.07–1.25)	1.16 (1.07–1.26)
Partial	1.05 (0.95–1.17)	1.05 (0.95–1.17)
Charlson comorbidity index	1.12 (1.11–1.14)	1.12 (1.11–1.14)
Hepatitis C virus	0.79 (0.69–0.91)	0.79 (0.68–0.91)
Alcoholic cirrhosis	1.41 (1.29–1.55)	1.41 (1.29–1.55)
Nonviral non-alcohol-related cirrhosis	1.25 (1.10–1.43)	1.25 (1.10–1.42)
Ascites	1.01 (0.68–1.49)	1.01 (0.68–1.50)
Varices	0.70 (0.60–0.82)	0.70 (0.60–0.82)
Hepatic encephalopathy	1.13 (1.00–1.27)	1.13 (1.00–1.27)
Hepatocellular carcinoma	1.19 (0.96–1.46)	1.18 (0.96–1.46)
Paracentesis	1.14 (0.79–1.65)	1.43 (1.20–1.71)
Transjugular intrahepatic portosystemic shunt	1.43 (1.20–1.71)	1.15 (0.79–1.66)
Spontaneous bacterial peritonitis	2.32 (1.05,5.14)	2.29 (1.03,5.07)
Location
Rural	ref	ref
Urban	1.10 (1.01–1.20)	1.11 (1.02–1.20)
Disability	1.08 (0.99–1.19)	1.08 (0.99–1.19)

Values are incidence rate ratio (95% confidence interval).

^ahttps://www2.census.gov/geo/pdfs/maps-data/maps/reference/us_regdiv.pdf

In the COC model, increasing count of 30-day readmissions was associated with black race (IRR, 1.18; 95% CI, 1.04–1.35), end-stage renal disease (IRR, 1.57; 95% CI, 1.35–1.82), a higher Charlson comorbidity index (IRR, 1.08; 95% CI, 1.07–1.09), alcohol-related cirrhosis (IRR, 1.28; 95% CI, 1.15–1.44), hepatic encephalopathy (IRR, 1.18; 95% CI, 1.03–1.36) and paracentesis (IRR, 1.29; 95% CI, 1.06–1.57), and inversely associated with varices (IRR, 0.79; 95% CI, 0.65–0.97) (Table 4). Similar associations exist in the UPC model for 30-day readmissions (Table 4).

Patients with advanced cirrhosis in many ways are different than those with compensated cirrhosis (eg, frequency and complexity of care required). To address this, we performed a subgroup analysis of patients with decompensated cirrhosis. The exposure covariates, COC and UPC, did not change considerably, though unadjusted rates showed higher utilization in this subgroup. For example, death or transplant occurred at a rate of 0.29 events/person-year in the subgroup compared with 0.19 in the full cohort. Hospitalizations accounted for 8.79 d/person-year in the subgroup compared with 6.31 in the full cohort. After adjusting, the third COC quartile in the all-cause mortality or liver transplant model had a similar effect size observed as in the full cohort but it was not significant, whereas the top COC quartile moved further from the null (1.33 in the subgroup compared with 1.20 in the full cohort). UPC exhibited the same pattern, with a nonsignificant third quartile and a larger effect in the top quartile. For hospital days, the top quartile for COC and UPC moved further from the null (remained significant), whereas the bottom 3 quartiles exhibited little fluctuation in effect size and remained insignificant (Supplementary Table 4).

Discussion

Cirrhosis is increasingly prevalent and complex. Optimal management of cirrhosis is a multidisciplinary effort that demands vigilance. In this nationally representative cohort of Medicare beneficiaries with cirrhosis, we show that increasing concentration of care among providers is associated with an increase in mortality and hospitalization. These novel data extend our knowledge of care coordination in cirrhosis in multiple ways.

Risks of Highly Concentrated Care in Cirrhosis

Higher concentrations of care are associated with increased mortality and hospitalization for patients with cirrhosis. We also see a 19% increase in 30-day readmissions for patients with a COC in the highest quartile. This contrasts sharply with findings in other disease states. In the overall population of Medicare enrollees, for every 10% increase in COC, there was a reduction in preventable hospitalizations by 2%.¹⁹ In a cohort of patients with newly diagnosed diabetes, higher COC and UPC both associated with a reduction in hospital days but were not associated with mortality.²³ These data suggest that cirrhosis is a uniquely complex condition. Compared with receiving from a single provider with whom they have a strong COC, patients with cirrhosis may benefit from the active collaboration of multiple clinicians.

Cirrhosis is characterized by unpredictable health care needs affecting multiple organ systems such as hepatic encephalopathy and symptomatic ascites. Its complexity is compounded by a least 2 factors. First, cirrhosis is dependent on medications with narrow therapeutic windows, including lactulose and diuretics, and procedures that require timely care coordination, such as paracentesis. Second, cognitive dysfunction and disability are common even among patients with compensated cirrhosis.²⁴ The factors diminish the patient’s capacity for self-care and increase their dependence on caregivers and providers. For these reasons, cirrhosis demands team-based care, including gastroenterologists, primary care providers, and other specialists, to achieve optimal quality measures and satisfactory clinical outcomes, and reduce inpatient utilization.²⁵ Prior studies have shown that quality of care and outcomes are improved when persons with cirrhosis receive care from gastroenterologists in addition to primary care, through both in-person and telehealth visits,^26,27 as well as when care is provided by nurse practitioners or physician assistants after subspecialty consultation.²⁸ Our study’s findings extend these data to show that a key aspect linking prior work is collaboration and therefore lower concentration of care. In contrast to other disease states, cirrhosis may benefit more from co-management than concentrated care. One clinician alone may not be able to effectively care for patients with cirrhosis.

Concentration of Care Metrics

Our study demonstrates a median COC of 0.40 for Medicare beneficiaries with cirrhosis, much lower than that observed in persons with congestive heart failure (COC 0.55) or chronic obstructive pulmonary disease (COC 0.66).²⁰ While the optimal organizational structure of care teams may be different for the management of cirrhosis, as compared with other chronic conditions, patients with congestive heart failure and chronic obstructive pulmonary disease also have complex needs and often require care by both primary care providers and specialists. However, unlike cirrhosis, congestive heart failure and chronic obstructive pulmonary disease have been the target of successful policy efforts to reduce overuse, which have led to more coordinated high-quality care for these conditions.²⁹

Multiple patient and system-related factors influence care concentration. Similar features are noted when considering care concentration as defined by either the COC or UPC indices, as other studies have demonstrated.¹¹ Remarkably, we found limited variation in concentration of care across demographics and clinical features. Still, there are important clues that inform the potential drivers of care concentration. For example, persons with few comorbidities have higher concentration of care than those with many and the same is true for those with TIPS and hepatocellular carcinoma. Together, these data suggest that severity and complexity of illness are linked with decreased concentration of care. However, given that these specific factors are accounted for in multivariable models, it is clear that factors beyond disease severity may mediate the association between concentration of care and adverse outcomes.

Contextual Factors

These data must be interpreted in the context of the study design. First, in order to study meaningful estimates of care concentration, we limited our cohort to those who had ≥4 outpatient visits. While this may select for patients with infrequent in-person care, this exclusion reduces the bias of limited observations on concentration measures; this excluded group has similar outcomes to the study cohort, with the exception of a higher death rate. Second, although the COC and UPC have not been previously applied to persons with cirrhosis, we used 2 of the most widely validated care coordination instruments for the Medicare population. Third, while our cohort was identified using algorithms validated for cirrhosis in Medicare,¹⁵ cirrhosis complication codes are limiting to the documentation decisions of individual clinicians, and identification of details such as clinical deterioration are limited. Fourth, our coordination metrics do not capture qualitative data such as the methods or quality of communication between clinicians. Last, our data may not generalize to many younger patients with cirrhosis given the nature of Medicare data.

Next Steps

Care coordination interventions, such as referral tracking and referral contracts, have targeted communication in a team structure, and tele-visits and e-referrals, have addressed barriers to specialty access.²⁵ However, cirrhosis care has demonstrated to be different than other chronic conditions (eg, congestive heart failure), which have been the emphasis of policies to bridge the quality chasm in medicine. A better understanding of the current state of concentration of care for patients with cirrhosis, and its relationship with outcomes, is warranted to better care for these patients. While both specialists and primary care provider influence cirrhosis care, little is known about care coordination between providers in a cirrhosis care team.²⁵ Therefore, interventions targeting coordination in this setting have been limited in their impact. Future studies should incorporate an in-depth qualitative analysis to better understand the correlation between these coordination measures and day-to-day clinical practice. This is a step toward the long-term goal of developing care coordination interventions and optimizing the organization of care teams to provide high-quality cirrhosis management.

In conclusion, for patients with cirrhosis, coordination in care is relatively low and an optimal threshold of continuity and concentration of care that relates to optimal outcomes likely exists, in the context of team-based cirrhosis care. This examination of coordination of care as it relates to Medicare beneficiaries with cirrhosis is timely as the emphasis on value-based cirrhosis care grows and cirrhosis-related morbidity continues to rise. A gap persists between current practice and optimal achievement of cirrhosis quality measures; this must be addressed by focusing future studies on better understanding the role of COC, both as a target of improvement and quality of care monitoring, as we continue to shift our focus on improving health care value for patients with chronic liver disease.

What You Need to Know.

Background

Improving care coordination for patients with high-intensity specialty care needs, such as cirrhosis, is an important tool to improve quality and reduce health care utilization.

Findings

In an analysis of data from 26,006 Medicare enrollees with cirrhosis, we found that increasing concentration of care among providers is associated with an increase in mortality and hospitalization. There is also an association between increasing concentration of care and a higher likelihood of 30-day readmission.

Implications for patient care

For patients with cirrhosis, coordination in care is relatively low, and there appears to be an optimal threshold of continuity and concentration of care that relates to optimal outcomes, in the context of team-based cirrhosis care.

Abbreviations used in this paper:

CI

confidence interval

COC

Continuity of Care

HR

hazard ratio

ICD-9

International Classification of Diseases, Ninth Revision

IRR

incidence rate ratio

TIPS

transjugular intrahepatic portosystemic shunt

UPC

Usual Provider of Care