BACKGROUND

Effective care coordination (CC) encompasses patient assessment, communication between patients and care teams, and facilitation of connections between patients, health systems, and communities.1–4 Ideally, CC results in care plans that are integrated across domains of care3 and aligned with patient and family needs, preferences, and goals.5 Given this relative complexity, providers and staff spend significant time organizing and delivering CC services.6–11 Yet, until relatively recently, formal CC went largely undocumented, as it was not compensated through traditional billing mechanisms. CC is now reimbursed by many US health plans and payers.12 For example, in 2013 Medicare began reimbursing providers for transitional care management, since 2015 for chronic care management, and since 2017 for complex chronic care management services.

In the United States (US), the Veterans Health Administration (VHA) is the largest integrated national health care system caring for approximately 9 million Veterans, many of whom have complex health and social needs.13,14 VHA has undertaken efforts to integrate CC into service delivery,15 focusing on these Veterans.16 For example, the Patient Aligned Care Team (PACT) Intensive Management (PIM) model enhances the PACT model—VHA’s patient-centered medical home—with the addition of transitional and other care management services.17 In a randomized trial, PIM was associated with increased use of primary care and social work services, reduced use of inpatient services, similar total costs of care,18 and improvements in patient experiences.19 In 2016, VHA launched the Social Work PACT Staffing Program to increase the provision of social work services to rural Veterans.20 The program was associated with increased use of social work services and decreased hospital admissions and emergency department visits among high-need rural Veterans.20 In addition, in 2016, VHA began a Care Coordination and Integrated Case Management (CC&ICM) initiative to standardize and integrate CC services across all VHA facilities and points of care.21 The VHA also began the Referral Coordination Initiative (RCI) in 2018 to improve the coordination of care for patients requiring care under VHA-purchased care arrangements. However, with interruption due to the pandemic, full implementation of the CC&ICM and RCI is still in progress, and formal evaluations of these programs have not been completed.

Despite the array of VHA initiatives underway to increase the provision of CC services to Veterans, evidence on the scope of CC delivery in VHA is lacking. For example, little information exists on which Veterans do and do not receive CC services or how much and how often these services are received. Yet, the effectiveness of CC likely depends, at least in part, on these factors and their alignment with patients’ CC needs and preferences. To address this gap, we examined the intensity, timing, and visit setting of VHA CC, as well as factors associated with receipt of VHA CC among a national sample of Veterans at high risk for hospitalization or mortality.

METHODS Participants and Data

Our study population included all Veterans enrolled in VHA during fiscal year (FY) 2019 through FY2021 (ie, October 2018 to September 2021), categorized as high risk and with some contact with the VHA health system during this period (eg, outpatient visit or consult/referral for VHA-delivered or VHA-purchased care) (N=1,890,388). Veterans’ high-risk status was ascertained using care assessment of need (CAN) scores of ≥85 at some point during FY2019-FY2021. CAN scores are a weekly VHA measure that indicates a Veteran’s percentile risk for hospitalization or death within the next year,22 and are calculated using sociodemographic characteristics, chronic illness measures, VHA health care use, vital sign parameters, laboratory test values, and specific medication use and drug-drug interactions. Veterans were excluded from the study if death records indicated they died before their first ≥85 CAN score (ie, died before the start of follow-up) (N=2736) if they resided outside of the 50 US states or the District of Columbia (N=16,501), or if they had missing covariate information (VHA priority group: N=2926; social vulnerability: N=23,810; drive distance to VHA primary care: N=1143). Our final study sample comprised 1,843,272 high-risk Veterans (see Fig. 1, Supplemental Digital Content 1 for a study flow diagram, https://links.lww.com/MLR/C852). We followed these Veterans for one year after their first ≥85 CAN score to ascertain the intensity, timing, and visit setting of CC services and to find associations with Veteran characteristics. Data for this study came from the VHA’s Corporate Data Warehouse,23 VHA-purchased care claims records,24 and the VHA linked Centers for Medicare and Medicaid Services (CMS) enrollment and Medicare fee-for-service claims records.25 The VA Portland Health Care System Institutional Review Board approved this project.

Measures

CC is documented in VHA as a specific service in line with CMS guidelines and using Current Procedural Terminology (CPT) codes, including those for transitional care management (TCM), chronic care management (CCM), complex chronic care management (CCCM), and behavioral health care management (BHCM) (Table 1, Supplemental Digital Content 1, https://links.lww.com/MLR/C852). Using these procedure codes, dates of service, and VHA stop codes, we constructed five variables to measure receipt of CC services during the 1-year follow-up period after a Veteran’s assignment to high-risk status (ie, first ≥85 CAN score): (1) receipt of any CC service (yes/no), (2) number of CC services received, (3) number of days to first CC service, (4) type of visit setting during which CC services were received, and (5) number of days between CC services among those who received more than one service. The type of visit was ascertained using VHA stop codes, which we classified into visit settings using VHA Health Economics and Resource Center (HERC) Category of Care Assignments categories.26 These included outpatient medicine, dialysis, ancillary services, rehabilitation, diagnostics services, prosthetics, surgery, psychiatry, substance abuse treatment, dental, adult day care, home care, contract extended care, other contracted care, and unassigned. Our CC measures did not include CC services that may occur during a visit or within a care team unless the service was specifically captured via a corresponding CPT code.

Veteran characteristics included for descriptive and analytic purposes included sex (male, female); age group (<45, 45–64, 65–74, 75–84, ≥85); race (American Indian/Alaskan Native, Asian, Black/African American, Multiracial, Native Hawaiian/Pacific Islander, White, another identity/unknown); ethnicity (Hispanic/Latino/a/x, non-Hispanic/Latino/a/x); marital status [married, single/other (including divorced, separated, widowed)]; VHA copay group (none, some, full); Medicare fee-for-service (FFS) (yes/no), Medicare Advantage (yes/no), Medicaid (yes/no), and other insurance coverage (yes/no); rurality of residence (urban, rural, highly rural); assigned VHA facility level of complexity (1a-high, 1b-high, 1c-high, 2-medium, 3-low, excluded, missing; eg, high-complexity facilities are those with the capability to provide high-complexity procedures and treatments); drive time from residence to the nearest VHA primary care facility (0–10, 11–20, 21–30, >30 min); Elixhauser readmission score and individual Elixhauser readmission score comorbidities (yes/no; eg, heart failure, depression, and drug use disorder); and mortality during 1-year follow-up (yes/no). We also examined Veterans’ social vulnerability using the Center for Disease Control and Prevention (CDC) Social Vulnerability Index (SVI), which incorporates factors such as poverty, lack of access to transportation, and crowded housing to create a percentile ranking of the social vulnerability of each census tract along four domains.27 Following an approach used in previous research,28 Veterans were categorized as living in a socially vulnerable area if they resided in a tract that was more vulnerable than ≥90% of tracts in that state for each of the four vulnerability domains: socioeconomic status (yes/no), household composition and disability (yes/no), minority status and language (yes/no), and housing type and transportation (yes/no).

Statistical Analyses

The 5 measures of receipt of CC services, described above, were summarized via frequencies for categorical variables, and means, standard deviations, medians, and histograms for continuous variables. To explore factors associated with receipt of CC services, we first summarized characteristics of the overall sample and stratified by whether Veterans received CC services during 1-year follow-up using descriptive statistics (frequencies for categorical variables; means and standard deviations for continuous variables), and then calculated absolute standardized mean differences29–31 to compare the 2 groups of Veterans. In addition, we estimated a multivariable logistic regression model, computing adjusted odds ratios (ORs) with 95% CIs for receipt of CC services. In our regression model, White race and non-Hispanic/Latino/a/x ethnicity were chosen as the reference groups, since most Veterans identified as White race and non-Hispanic/Latino/a/x ethnicity. All analyses were performed in R version 4.1.2.32 Statistical significance of ORs was determined at alpha=0.05. For SMD, effect sizes were determined at 0.2=small effect, 0.5=medium effect, 0.8=large effect.33

RESULTS Veteran Characteristics

Among 1,843,272 high-risk Veterans, the majority were 65 years of age or above (67%), male sex (93%), White race (75%), non-Hispanic/Latino/a/x ethnicity (95%), and unmarried (54%) (Table 1). Most Veterans had no VHA copays for medical care (55%) and had Medicare FFS insurance coverage in addition to their VHA benefits (79%). Most Veterans had less than a 30-minute drive time to the nearest VHA primary care facility (81%), resided in urban areas (66%), and were assigned to high-complexity VHA facilities (73%). A minority of high-risk Veterans resided in socially vulnerable areas (socioeconomic status: 9%, household composition and disability: 5%, minority status and language: 22%, housing type and transportation: 12%). Regarding their clinical characteristics, Veterans’ mean Elixhauser readmission score was 27.01 (SD 22.05), and their most common Elixhauser readmission comorbidities were hypertension (76%), chronic pulmonary disease (33%), depression (33%), complicated diabetes (32%), and obesity (25%). Approximately one tenth (9%) of Veterans died during 1-year follow-up. Given our cohort construction (ie, cohort entry at first ≥85 CAN score), most Veterans entered the cohort in 2019 (70%).

TABLE 1 - Veteran Characteristics by Receipt of Care Coordination Status Receipt of care coordination Characteristic Overall, N=1,843,272, n (%) Yes, N=566,097, n (%) No, N=1,277,175, n (%) SMD Cohort entry fiscal year — — — 0.192  2019 1,287,416 (70) 429,158 (76) 858,258 (67) —  2020 292,402 (16) 71,862 (13) 220,540 (17) —  2021 263,454 (14) 65,077 (11) 198,377 (16) — Age category — — — 0.285  <45 115,233 (6.3) 52,995 (9.4) 62,238 (4.9) —  45–64 480,742 (26) 179,468 (32) 301,274 (24) —  65–74 649,698 (35) 182,101 (32) 467,597 (37) —  75–84 335,070 (18) 86,144 (15) 248,926 (19) —  ≥85 262,529 (14) 65,389 (12) 197,140 (15) — Sex 0.039  Male 1,708,709 (93) 520,772 (92) 1,187,937 (93)  Female 134,563 (7.3) 45,325 (8.0) 89,238 (7.0) Race — — — 0.198  American Indian/Alaskan Native 14,699 (0.8) 5067 (0.9) 9632 (0.8) —  Asian 10,411 (0.6) 3161 (0.6) 7250 (0.6) —  Black/African American 376,005 (20) 146,265 (26) 229,740 (18) —  Multiracial 15,763 (0.9) 5405 (1.0) 10,358 (0.8) —  Native Hawaiian/Pacific Islander 13,806 (0.7) 4256 (0.8) 9550 (0.7) —  White 1,387,840 (75) 393,116 (69) 994,724 (78) —  Another identity/unknown 24,748 (1.3) 8827 (1.6) 15,921 (1.2) — Hispanic/Latino/a/x Ethnicity — — — 0.045  Non-Hispanic / Latino/a/x 1,755,010 (95) 535,159 (95) 1,219,851 (96) —  Hispanic / Latino/a/x 88,262 (4.8) 30,938 (5.5) 57,324 (4.5) — Marital status — — — 0.221  Single/other 989,362 (54) 346,669 (61) 642,693 (50) —  Married 853,910 (46) 219,428 (39) 634,482 (50) — VHA copay group — — — 0.202  None 1,010,793 (55) 310,526 (55) 700,267 (55) —  Some 585,200 (32) 203,220 (36) 381,980 (30) —  Full 247,279 (13) 52,351 (9.2) 194,928 (15) — Medicare FFS 1,465,096 (79) 421,103 (74) 1,043,993 (82) 0.178 Medicare advantage 413,219 (22) 121,763 (22) 291,456 (23) 0.032 Medicaid 171,313 (9.3) 78,561 (14) 92,752 (7.3) 0.216 Other health insurance coverage 136,338 (7.4) 34,153 (6.0) 102,185 (8.0) 0.077 Rurality — — — 0.169  Urban 1,225,611 (66) 406,354 (72) 819,257 (64) —  Rural 579,890 (31) 151,872 (27) 428,018 (34) —  Highly rural 37,771 (2.0) 7871 (1.4) 29,900 (2.3) — Drive time to VHA primary care (min) — — — 0.167  0–10 489,005 (27) 173,267 (31) 315,738 (25) —  11–20 668,377 (36) 209,763 (37) 458,614 (36) —  21–30 331,069 (18) 92,145 (16) 238,924 (19) —  >30 354,821 (19) 90,922 (16) 263,899 (21) — Facility complexity — — — 0.089  1a-high complexity 762,455 (41) 238,670 (42) 523,785 (41) —  1b-high complexity 333,791 (18) 103,635 (18) 230,156 (18) —  1c-high complexity 264,244 (14) 86,314 (15) 177,930 (14) —  2-medium complexity 206,878 (11) 55,975 (9.9) 150,903 (12) —  3-low complexity 182,803 (9.9) 51,447 (9.1) 131,356 (10) —  98-excluded 6191 (0.3) 1278 (0.2) 4913 (0.4) —  Missing 86,910 (4.7) 28,778 (5.1) 58,132 (4.6) — Social vulnerability based on socioeconomic status (90th percentile) 158,405 (8.6) 54,322 (9.6) 104,083 (8.1) 0.051 Social vulnerability based on household composition and disability (90th percentile) 84,970 (4.6) 23,822 (4.2) 61,148 (4.8) 0.028 Social vulnerability based on racial/ethnic minority (90th percentile) 404,132 (22) 145,658 (26) 258,474 (20) 0.131 Social vulnerability based on housing and transportation (90th percentile) 223,371 (12) 78,394 (14) 144,977 (11) 0.075 Elixhauser readmission score (mean, SD) 27.01 (22.05) 30.11 (23.40) 25.63 (21.28) 0.200 Individual Elixhauser comorbidities  AIDS/HIV 17,228 (0.9) 6901 (1.2) 10,327 (0.8) 0.041  Alcohol use disorder 283,048 (15) 125,494 (22) 157,554 (12) 0.262  Deficiency anemia 429,702 (23) 139,765 (25) 289,937 (23) 0.047  Rheumatoid arthritis/collagen vascular diseases 103,945 (5.6) 28,034 ((5.0) 75,911 (5.9) 0.044  Chronic blood loss anemia 40,481 (2.2) 13,202 (2.3) 27,279 (2.1) 0.013  Heart failure 314,837 (17) 98,019 (17) 216,818 (17) 0.009  Chronic Pulmonary Disease 603,230 (33) 180,901 (32) 422,329 (33) 0.024  Coagulopathy 105,327 (5.7) 34,609 (6.1) 70,718 (5.5) 0.025  Depression 615,225 (33) 233,455 (41) 381,770 (30) 0.239  Diabetes, uncomplicated 203,298 (11) 54,366 (9.6) 148,932 (12) 0.067  Diabetes, complicated 585,902 (32) 173,080 (31) 412,822 (32) 0.038  Drug use disorder 182,892 (9.9) 97,698 (17) 85,194 (6.7) 0.331  Hypertension 1,406,902 (76) 410,340 (72) 996,562 (78) 0.129  Hypothyroidism 235,579 (13) 67,297 (12) 168,282 (13) 0.039  Liver disease 201,299 (11) 72,684 (13) 128,615 (10) 0.087  Lymphoma 30,188 (1.6) 8530 (1.5) 21,658 (1.7) 0.015  Fluid and electrolyte disorders 361,461 (20) 128,647 (23) 232,814 (18) 0.112  Metastatic cancer 48,436 (2.6) 15,838 (2.8) 32,598 (2.6) 0.015  Other neurological Disorders 365,088 (20) 137,819 (24) 227,269 (18) 0.161  Obesity 466,737 (25) 135,077 (24) 331,660 (26) 0.049  Paralysis 73,313 (4.0) 34,051 (6.0) 39,262 (3.1) 0.142  Peripheral vascular disease 359,248 (19) 105,732 (19) 253,516 (20) 0.030  Psychoses 262,572 (14) 120,097 (21) 142,475 (11) 0.276  Pulmonary circulation disorders 49,286 (2.7) 16,049 (2.8) 33,237 (2.6) 0.014  Renal failure 358,725 (19) 106,231 (19) 252,494 (20) 0.025  Solid tumor without metastasis 298,986 (16) 79,814 (14) 219,172 (17) 0.084  Peptic ulcer disease excluding bleeding 33,977 (1.8) 10,906 (1.9) 23,071 (1.8) 0.009  Cardiac valvular disease 222,687 (12) 60,700 (11) 161,987 (13) 0.061  Weight Loss 141,473 (7.7) 52,795 (9.3) 88,678 (6.9) 0.087  Mortality during 1-year follow-up 165,185 (9.0) 59,558 (11) 105,627 (8.3) 0.077

Other marital status includes the categories: divorced, separated, and widowed.

AIDS indicates acquired immune deficiency syndrome; FFS, fee-for-service; HIV, human immunodeficiency virus; SMD, standardized mean difference; VHA, Veterans Health Administration.

Intensity, Timing, and Visit Setting of CC

Among all high-risk Veterans in the sample, 31% received CC services within 1 year of their first ≥85 CAN score (Table 1). Also, among all high-risk Veterans, the median number of services received was 0 (IQR [0, 1]) (data not shown). Among high-risk Veterans who received at least 1 CC service, the median number of services received was 2 [IQR (1, 6)] (Fig. 1A). In addition, among those who received at least 1 CC service, the median number of days between their first ≥85 CAN score and their first CC service was 84 [IQR (25, 195)] (Fig. 1B). For high-risk Veterans who received 2 or more CC services, the median number of days between CC services was 26 [IQR (10, 57)] (Fig. 1C). Regarding visit setting, the plurality of CC services was received during outpatient psychiatry visits (46%), followed by outpatient medicine (16%), ancillary services (16%), home care (8%), and substance abuse treatment (7%) visits (Table 2). High-risk Veterans who entered the cohort in 2020 or 2021 had a lower prevalence of receiving CC during 1-year follow-up compared with those who entered the cohort in 2019 (prevalence of receipt of CC in 2019: 33%, 2020: 25%, 2021: 25%) (Table 1).

TABLE 2 - Number and Proportion of Care Coordination Services by Visit Setting Visit Setting N (%) Outpatient psychiatry 1,573,393 (46) Outpatient medicine 545,371 (16) Outpatient ancillary services 530,767 (16) Home care 289,061 (8) Outpatient substance abuse treatment 225,528 (7) Outpatient rehabilitation 126,388 (4) Unassigned 60,455 (2) Outpatient surgery 22,167 (0.7) Outpatient dialysis 14,650 (0.4) Outpatient adult day care 12,197 (0.4) Other contract care 2,756 (0.08) Outpatient diagnostics services 1,373 (0.04) Outpatient prosthetics 410 (0.01) Outpatient dental 28 (<0.01) Contract extended care 27 (<0.01)
FIGURE 1:

Intensity and timing of receipt of care coordination. A, Number of care coordination services documented during 1-year follow-up, Veterans who had documentation of having received ≥1 service. B, Number of days between Veterans’ first ≥85 CAN score and first care coordination service, Veterans who had documentation of having received ≥1 service. C, Mean number of days between care coordination services, Veterans who had documentation of having received >1 service. CAN indicates care assessment of need; CC, care coordination.

Associations With Receipt of CC

In unadjusted analyses, high-risk Veterans who did and did not receive CC services were similar in terms of their sex, ethnicity, complexity of assigned VHA facility, social vulnerability, and certain Elixhauser comorbidities (eg, anemias, heart failure, chronic pulmonary disease, diabetes, liver disease) (Table 1). Compared with high-risk Veterans who did not receive CC services, those who were more likely to receive CC services were younger (SMD 0.285), unmarried (SMD 0.221), had Medicare FFS (SMD 0.178) and/or Medicaid (SMD 0.216) coverage in addition to their VHA benefits, lived in urban areas (SMD 0.169), had shorter drive times to VHA primary care (SMD 0.167), had higher Elixhauser readmission scores (SMD 0.200), and had certain Elixhauser readmission comorbidities [e.g., alcohol use disorder (SMD 0.262), depression (SMD 0.239), drug use disorder (SMD 0.331), psychoses (SMD 0.276)]. High-risk Veterans who did and did not receive CC services were also different in terms of their race (SMD 0.198).

Results from the adjusted regression model demonstrated that high-risk Veterans who were older [45–64 y aOR 0.63 (0.60, 0.65); 65–74 y aOR 0.44 (0.42, 0.47); 75–84 years aOR 0.41 (0.39, 0.44); 85+ years aOR 0.39 (0.36, 0.42); reference <45 y]; female sex [aOR 0.84 (0.82, 0.87), reference male]; who resided in rural areas [rural aOR 0.87 (0.82, 0.93), highly rural aOR 0.80 (0.72, 0.88); reference urban], and who had longer driving times to VHA facilities [11–20 min aOR 0.90 (0.87, 0.92), 21–30 min aOR 0.84 (0.80, 0.89), >30 min aOR 0.83 (0.79, 0.88); reference 0–10 min] had lower adjusted odds of receiving CC services during 1-year follow-up (Table 3). High-risk Veterans living in socially vulnerable areas based on household composition and disability also had lower adjusted odds of receiving CC services [aOR 0.89 (0.80, 0.99), reference not vulnerable based on household composition and disability]. Black/African American [aOR 1.25 (1.19, 1.32); reference White] and American Indian/Alaskan Native race [aOR 1.17 (1.10, 1.26); reference White]; Hispanic/Latino/a/x ethnicity [aOR 1.09 (1.04, 1.14); reference non-Hispanic/Latino/a/x]; and Medicaid-enrolled [aOR 1.45 (1.39, 1.50); reference not Medicaid enrolled] high-risk Veterans had greater adjusted odds of receiving CC services during 1-year follow-up. High-risk Veterans with some copay [aOR 1.12 (1.10, 1.15); reference no copay], and with higher Elixhauser readmission scores also had greater adjusted odds of receiving CC services [aOR 1.01 (1.01, 1.01)].