Use of General Primary Care, Specialized Primary Care, and Other Veterans Affairs Services Among High-Risk Veterans

This cross-sectional study assesses the use of general and specialized primary care, medical specialty, and mental health services among patients aat high risk of hospitalization in the Veterans Health Administration.


Introduction
Patients at high risk for hospitalization are heterogeneous 1 and complex to treat, 2,3 and they typically have multiple chronic medical conditions, often compounded by mental health conditions. 4-6 A frequent assumption has been that many of these high-risk patients receive mostly acute emergency care. 7 Another common assumption has been that high-risk patients with complex conditions receive care mostly through medical specialists, 8,9 whereas primary care is predominantly directed at a lower-risk population. High-risk patients are often assumed to have low levels of health literacy and to be uninterested in electronic communication with their health care practitioners. 10 We aimed to inform health care system level planning for high-risk patient care by testing assumptions about whether, where, and how the more than 350 000 highest-risk patients cared for nationally in the Veterans Health Administration (VHA) receive outpatient care.
Primary care settings in integrated health care systems aim to provide care that is coordinated, comprehensive, continuous, and assessible. 11 Specialty or subspecialty settings, on the other hand, predominantly aim to provide accessible, short-term consultative care. Even when specialists provide long-term continuity care, as they often do for some patients, 8,[12][13][14] they typically do not aim to provide full primary care, defined by the Institute of Medicine as "the provision of integrated, accessible health care services by clinicians who are accountable for addressing a large majority of personal health care needs." 15(p1),16 Sometimes, however, specialists or specialized teams serve as alternatives to general primary care. These specialized primary care settings aim to provide comprehensive, continuous primary care to a population with special needs, such as those with HIV infection, 17 homeless veterans, 18 or those with serious mental illnesses. 19 In this study, we sought to characterize patterns of care for the top 5% of the highest-risk patients enrolled in the VHA nationally based on a predictive risk score for near-term hospitalization that is generated for all VHA patients. We expected that high-risk patients, compared with low-risk patients, would use more face-to-face and fewer secure message encounters with primary care. We also expected that half of high-risk patients would be cared for in specialized primary care settings and that more than half of specialized primary care patients would be high risk. We hypothesized that high-risk patients assigned to general primary care, compared with those assigned to specialized primary care, would have both more primary care and more medical specialty care visits. We aimed to compare (1) primary care encounters (face-to-face, telephone, and secure messages) among highrisk vs low-risk patients, (2) the proportions of high-risk patients assigned to general vs specialized primary care, and (3) the use of primary care and medical (nonsurgical) specialty care visits among high-risk patients assigned to general vs specialized primary care. Office of Primary Care because this study was designed for internal VA purposes in support of the VA mission. The findings were designed to be used by and within VA for program and planning purposes.

Patients
Data were deidentified. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Setting
Since national implementation of the patient-centered medical home model in more than 900 locations in the VHA in 2010, 21,22 nearly all patients have been assigned at enrollment to a continuity primary care practitioner (physician, nurse practitioner, or physician assistant) who works with a care team that includes a registered nurse care manager, licensed practical nurse, and clerk. 19,21,[23][24][25][26][27] Patients are assigned to a general or specialized primary care team and may switch to another general or specialized primary care team at any time after enrollment. All teams are tasked with providing comprehensive primary care, including screening and preventive care services. 27,28 Most general primary care sites are community based, whereas most specialized primary care sites other than those for women's health are medical center based. 29 As an integrated health care system, the VHA provides a full range of specialty services, including consultative medical and surgical specialty care as well as integrated primary care and mental health care. 30 In addition, the VHA offers supplementary services for intensifying care, such as telehealth (remote monitoring for chronic medical conditions), 31 palliative care, and housing services. The VHA offers multiple modalities to access primary care, including telephone and secure messaging.

Measures Exposures
We assessed assignment to the type of primary care setting on September 30, 2015, based on administrative data. We assessed for assignment to general primary care and 7 diverse but commonly available types of specialized primary care that together accounted for 98% of patients enrolled in any type of VHA primary care. 32 We focused on geriatric, homelessness, and women's health settings as cognitive, nonprocedural services; HIV and dialysis as technical, specialized services; and homebased care and spinal cord injury as disability-focused services.

Risk Level
The VHA calculates Care Assessment Needs (CAN) scores 28,33 weekly for all enrolled veterans who are assigned to any (general or specialized) primary care practitioner and are not hospitalized at the time of CAN score generation. We used an updated CAN model 33 (eTable 1 in the Supplement) to assess hospitalization risk using VHA administrative data on demographics, use of VHA health services, comorbidity indicators, prescribed medications, vital signs, and veteran-specific variables.
The top 5% of patients identified by CAN score have an almost 20% risk of VHA hospitalization within the following 90 days. 28 We defined the high-risk patient cohort as all patients with a CAN score in the 95th or greater percentile at any time during April 1, 2015, through September 30, 2015 (n = 351 012). We excluded patients who had died during this period for the purpose of assigning a CAN score. We defined the remainder of patients (n = 3 958 180) as low risk. We used the first date that a patient qualified as high risk and the last date (ie, September 30, 2015) for low-risk patients as the index dates for our analyses.
assessed potentially qualifying characteristics for 4 types of specialized primary care (geriatrics, homelessness, women's health, and HIV) based on age (>70 years of age), housing instability (defined above), sex (female), and HIV status (defined by ICD-9 code).

Use of VHA Services
We counted all patient encounters coded during 2 periods: 1 year before each patient's index date for high-risk identification (in 2015) and 1 year after the VHA fiscal year 2016 (October 1, 2015, to September 30, 2016). We used the VA Support Service Center 35 data definitions for primary care encounter types (in-person or telephone) as recorded in the VHA Corporate Data Warehouse. 36 We used the My HealtheVet patient portal database for data related to secure messages. We counted only 1 encounter per encounter type for any single day. We also counted all nonsurgical specialty encounters (including mental health), inpatient admissions, and emergency department encounters.
We performed a sensitivity analysis on use among patients who were alive throughout the VHA fiscal year 2016 (99.4% of the high-risk cohort) and found minor changes (eTable 4 and eTable 5 in the Supplement).

Outpatient Encounters
We categorized outpatient encounters into the following 5 mutually exclusive and collectively exhaustive groups using established data definitions during VHA fiscal year 2016 37 : any in-person primary care (general or specialized), any in-person mental health (primary care and mental health integration, 30 individual and group psychotherapy, or substance use), any in-person medical specialty (excluding encounters related to procedures and chemotherapy), emergency department, or all other (eg, any telephone, surgical, physical therapy, occupational therapy, dental, nutrition, anticoagulation clinic, procedures, chemotherapy, telehealth, and radiology).

Add-on Services
We defined add-on services as those designed to supplement primary care for added care intensity.
We assessed use of 4 add-on services during VHA fiscal year 2016 with particular relevance to highrisk patients: palliative care, telehealth disease management for chronic conditions, intensive mental health case management (similar to assertive community treatment), 38 and housing services.

Statistical Analysis
We used unadjusted logistic regression to generate odds ratios (ORs) for being a high-risk patient (defined by the CAN score) based on veteran characteristics (eg, sex, age, race/ethnicity, and medical condition). We used 2-sample t tests to assess differences between veterans at high risk for hospitalization vs those at low risk for hospitalization in rates of primary care visits, emergency department visits, and in-patient discharges. We tested mean use differences between patients assigned to general vs specialized primary care using 2-sample t tests and 1-way analysis of variance using the Tukey multiple comparison test. 39 For all statistical tests, we used a 2-sided P < .05 as the a priori significance level and performed 2-tailed hypothesis tests. We performed successive analyses from April 1, 2016, to January 1, 2019. Analyses were conducted with SAS software, version 9.4 (SAS Institute Inc).

Results
The year than low-risk patients during the same period. Figure 1 shows that most high-risk patients (308 433 of 351 012 [88%]) were assigned to general care than to specialized primary care. Of the remaining high-risk patients, 5% were assigned to women's health, 2% to geriatrics, 2% to home-based primary care, 0.8% to homelessness primary care, 0.8% to HIV care, 0.8% to spinal cord injury care, 0.2% to dialysis, and 0.9% to all other specialized primary care settings.
As shown in Figure 2, among all patients (both high-and low-risk patients) assigned to general primary care, 8% (308 433 of 3 945 631) were high risk. Among all patients assigned to any specialized primary care setting, 12% (42 579 of 363 561) were high risk. The proportions of high-risk   Mean counts of any in-person primary care (general or specialized), any mental health, any medical specialty, or any emergency department (ED) face-to-face visit from October 1, 2015, to September 30, 2016. Other encounters include any outpatient telephone, telehealth, surgical, radiology, rehabilitation, and procedural visits. Quantitative results are also given in eTable 2 in the Supplement. Two-sample t tests were used for each encounter type comparing patients enrolled in general vs specialized primary care. We performed the Tukey multiple comparison procedure to assess differences in means. care had significantly fewer mental health (9. Overall, few high-risk patients used add-on services for intensifying primary care ( Table 2).

Discussion
High-risk patients in the VHA are a small fraction of the general population but account for nearly half of overall health care costs, 40 most of which are accounted for by hospitalizations. 5 Contrary to expectations, we found that most high-risk veterans (88%) were assigned to general primary care rather than specialized primary care. Patients assigned to general primary care had more mental health and primary care visits than medical specialty visits. High-risk patients used all types of encounters, including inperson, telephone, and secure messaging, at higher rates than low-risk patients.
Our findings challenge health care systems and researchers to consider the roles, goals, and implications of general primary care for high-risk patient populations. Health care systems are already providing risk-stratified care management, as in the Medicare Comprehensive Primary Care Plus Initiative, [41][42][43] and turning to specialized programs or intensive primary care models 26 to manage high-risk patient populations. Although specialized programs or intensive primary care models potentially offer important benefits to high-risk patients, these settings are often more resource intensive than general primary care 26,27,[44][45][46][47][48] and more available in urban areas. 49 One reason for investing in specialized primary care teams is to provide expertise and resources to special populations that may be more complex than the general primary care population. Overall, we found that most patients assigned to specialized settings were low risk. We also found that even among patients with a technically demanding condition, such as HIV infection, only half of patients with the condition were followed up in specialized primary care. The single exception was for highrisk women veterans, among whom two-thirds were enrolled in specialized women's health teams.
This finding is likely attributable to the need for sex-specific specialized services and procedures 50,51  Our findings also highlight the burden of mental health conditions and high rates of mental health service use among veterans who are at high risk for hospitalization. Further investigations into the patterns of mental health care among high-risk patients may be helpful for developing a focus on the complex needs of these patients within the primary care and mental health integration program. 30

Limitations
This study has limitations. It is descriptive and does not address causation. Our VHA-based findings may not be generalizable to other health care systems or to fee-for-service health care settings that may incentivize specialty care. 72 Our findings have implications, however, for any large system that aims to provide equitable access to care across an enrolled population. Furthermore, although VHA enrollees have higher rates of psychosocial conditions than non-VHA populations, 72,73 higher psychosocial condition rates are typical for high-risk patients in most health care systems. Additional limitations are that we did not assess non-VA use of care 74 ; we studied only how many encounters of particular types occurred without potentially relevant modeling information, such as patient functional status or distance to specialized care settings. Similarly, our data did not support analyses based on specialized primary care enrollment criteria; we used only very general indicators. In addition, our encounter data do not include slight recent VHA coding changes that resulted in a very small proportion of telephone visits being counted as in-person visits; distributions of in-person visits, however, were similar using alternative VHA data definitions.

Conclusions
Achieving the triple aim of improving patient experience of care, improving population health, and reducing per capita health care costs among high-risk patients 75 has proved to be elusive. 9,26,41,[53][54][55][56] Our data suggest that a better understanding of existing and optimal roles of different types of primary and specialty care for high-risk patient populations may be critical for achieving the triple aim. Planning for high-risk patient care improvement in integrated health care systems such as the VHA should focus on enabling high-quality complex patient care within primary care and mental health services. In this study, many high-risk veterans were avid users of electronic and telephone care, suggesting that opportunities exist to improve care through maximizing the accessibility and effectiveness of non-face-to-face modalities. Overall, our findings provide a foundation for investigation aimed at informing better design of health care programs and resources that can engage patients in their existing locations of care.