Sachin H. Jain, MD, MBA

 Complex care management programs have emerged as a promising model to better care for high-need, high-cost patients. Despite their widespread use, relatively little is known about the impact of these programs in Medicaid populations. This study evaluated the impact of a complex care management program on spending and utilization for high-need, high-cost Medicaid patients.

 Randomized quality improvement trial conducted at CareMore Health in Memphis, Tennessee. A total of 253 high-need, high-cost Medicaid patients were randomized in a 1:2 ratio to complex care management or usual care.

 Intention-to-treat analysis compared regression-adjusted rates of spending and utilization between patients randomized to the complex care program (n = 71) and those randomized to usual care (n = 127) over the 12 months following randomization. Primary outcomes included total medical expenditures (TME) and inpatient (IP) days. Secondary outcomes included IP admission, emergency department (ED) visits, care center visits, and specialist visits.

 Compared with patients randomized to usual care, patients randomized to complex care management had lower TME (adjusted difference, —$7732 per member per year [PMPY]; 95% CI, –$14,914 to –$550; 

 = .036), fewer IP bed days (adjusted difference, —3.46 PMPY; 95% CI, –4.03 to –2.89; 

 <.001), fewer IP admissions (adjusted difference, —0.32 PMPY; 95% CI, –0.54 to –0.11; 

 = .014), and fewer specialist visits (adjusted difference, —1.35 PMPY; 95% CI, –1.98 to –0.73; 

 <.001). There was no significant impact on care center or ED visits.

 Carefully designed and targeted complex care management programs may be an effective approach to caring for high-need, high-cost Medicaid patients.

A complex care management program for high-need, high-cost Medicaid patients reduced total medical expenditures by 37% and inpatient utilization by 59%. Based on the design of the program, these results suggest that:

Carefully designed and targeted complex care management can be effective among high-need, high-cost Medicaid patients.

Community health workers and other nontraditional healthcare workers can help engage and activate patients, build trust, and better understand and manage the nonmedical drivers of poor health and avoidable spending.

Targeted interventions focused on modifiable risk factors are an effective and efficient approach for reducing unnecessary utilization.

High-need, high-cost patients—those with chronic medical conditions and co-occurring social and behavioral complexity&mdash;experience poor health outcomes and high rates of potentially preventable spending.

 Improving care quality and lowering spending for this population have become a focus for policy makers, clinicians, payers, and health systems.

Complex care management programs—broadly defined as efforts to coordinate medical and social services, assist patients and caregivers in managing medical and behavioral health conditions, and address the psychosocial drivers of poor health&mdash;have emerged as a promising model to improve care for high-need, high-cost patients.

 In recent years, complex care management programs have been widely adopted by health systems and delivery organizations entering into value-based payment arrangements. A 2018 survey found that 96% of accountable care organizations had implemented complex care management programs for high-need, high-cost patients.

The impact of complex care management has been uneven. Systematic reviews and large-scale program evaluations have not demonstrated a consistent impact on health outcomes, spending, or acute care utilization.

 Some individual programs, however, have led to improvements in patient-reported health and decreases in acute care utilization and total spending.

 Program heterogeneity (eg, practice setting, patient characteristics, care models, staffing) may explain these varied results.

Complex care management programs have traditionally been developed and implemented among Medicare populations.

 Increasingly, these models are being adopted in Medicaid.

 Similar to the Medicare population, high-need, high-cost Medicaid patients have high rates of chronic conditions

 both attributes well suited to complex care management.

 High-need, high-cost Medicaid patients also have high rates of behavioral health disorders and unmet social needs,

 attributes less common among Medicare patients and areas in which the impact of complex care management is uncertain.

Despite the proliferation of complex care management for Medicaid patients, there have been very few rigorous program evaluations, and the overall effect of these programs is unclear. This paper presents the results of a randomized quality improvement trial to evaluate the impact of complex care management on spending and utilization for high-need, high-cost Medicaid patients with complex health and social needs.

This was a randomized quality improvement trial to evaluate the impact of a complex care management program on spending and utilization for high-need, high-cost Medicaid patients. The design and implementation of the program, detailed in the following paragraphs, incorporated attributes of successful complex care management programs in other populations.

 The program was implemented at CareMore Health (CareMore) in Memphis, Tennessee, where CareMore provides comprehensive primary care services for Medicaid patients enrolled in TennCare plans administered by Amerigroup. The program was operated out of a single CareMore care center.

Participants were drawn from adult Medicaid patients attributed to CareMore primary care physicians (PCPs). Program eligibility criteria were aimed at identifying patients at risk for poor outcomes and unnecessary spending, as well as those most likely to benefit from complex care management. The criteria drew from analyses suggesting that combining predictive models, historical claims, and clinician judgment is the most effective approach to identifying patients for complex care management.

Eligible patients were first required to meet at least 1 of the following criteria: top 5% of total medical expenditures (TME) in the prior 12 months, top 5% of Chronic Illness Intensity Index (CI3) score, or care team member nomination. The CI3 score is a predictive model developed by Amerigroup that uses demographic, clinical, and pharmacy claims data to estimate a patient’s predicted cost in the subsequent 12 months compared with the average Amerigroup Medicaid patient.

Patients meeting 1 of the above criteria were then required to meet at least 1 of the following criteria: 2 or more inpatient (IP) admissions in the prior 12 months, 3 or more emergency department (ED) visits in the prior 12 months, or 2 or more chronic conditions.

Patients less likely to benefit from complex care management because of specific comorbidities (cognitive impairment, severe mental illness without medical comorbidity, active malignancy, pregnancy) or current residence in a long-term care facility were excluded.

Given limited resources, eligible patients were randomized in a 1:2 ratio to complex care management or usual care. Patients not continuously attributed to CareMore over the 12-month period following randomization were considered lost to follow-up. This could occur for several reasons: loss of Medicaid eligibility, enrollment in another TennCare plan, or transfer of care to a non-CareMore PCP.

The program was staffed by a multidisciplinary care team consisting of a community health worker (CHW), a social worker (SW), and a PCP. CHWs and other nontraditional healthcare workers have become increasingly common components of complex care management programs.

 CHWs, drawn from local communities, build longitudinal relationships with patients, provide context-specific activation and engagement, assist with care coordination, and help patients navigate the complexities of health systems and social programs.

 CHW interventions have been shown to improve health outcomes and reduce spending, especially among urban, underserved populations.

With guidance and support from the SW and the PCP, the CHW was responsible for patient outreach, engagement, activation, and accompaniment. The SW was responsible for counseling and brief interventions for patients with behavioral health needs and for coordinating referrals to social service agencies and other medical providers. The PCP was responsible for providing comprehensive care for acute and chronic conditions and for coordinating with specialists and inpatient providers.

Patients randomized to the complex care management program were contacted by the CHW via telephone and invited to enroll in the program. The CHW attempted to contact patients on 3 separate occasions, after which no additional attempts at contact were made.

Program activities were designed around a conceptual model of the drivers of poor outcomes and avoidable spending for high-need, high-cost Medicaid patients. Drawing on existing frameworks,

 these drivers were grouped into 4 intersectional domains: (1) medical complexity and health trajectory, (2) care coordination and engagement, (3) self-management and behavioral health, and (4) environmental and social supports. Within each domain, specific drivers of poor health and high costs were identified. For example, unstable housing was a driver within the environmental and social supports domain, and medication management was a driver within the medical complexity and health trajectory domain (see 

 for a complete list of domains and drivers).

Patients who enrolled in the program were scheduled for a 60-minute in-person intake visit within 7 days of enrollment. During the intake visit, the entire care team (ie, CHW, SW, and PCP) conducted a comprehensive assessment of the patient’s medical, behavioral, and social risk factors. The results of this assessment were used to identify and prioritize the most impactful and impactable drivers of poor outcomes and avoidable spending for each patient. Finally, the patient and the care team codeveloped a tailored care plan that outlined interventions, roles, and responsibilities for addressing each of the prioritized drivers. For example, the SW may be responsible for providing counseling and brief interventions for depression and for coordinating referrals to psychiatry. The PCP may be responsible for rapidly addressing poorly controlled diabetes or congestive heart failure. The CHW may be responsible for contacting patients to identify barriers to adherence and self-management. The patient may be responsible for applying for temporary housing assistance, logging physical activity, or checking in daily with the CHW/PCP to provide blood pressure or blood sugar readings.

After the intake visit, there was frequent, structured follow-up. The CHW called patients at least weekly to assess progress and identify and troubleshoot barriers. The care team met formally every week to review and update care plans. Patients were scheduled for monthly, in-person follow-up visits to review and revise the care plan. Patients admitted to an IP or skilled nursing facility were visited by a member of the care team, if possible. The CHW accompanied patients to specialist, social service, and other appointments as needed.

The program launched on March 1, 2017, and ran until February 28, 2018, at which point patients returned to usual care.

The primary outcomes were TME and IP bed days. Secondary outcomes included ED visits, IP admissions, care center visits, and specialist visits. The ED visit outcome measure excluded visits that resulted in an IP admission. The care center visit measure included any in-person visit for medical or behavioral health services at one of CareMore’s outpatient facilities. All outcomes were assessed over the 12-month period following randomization (March 1, 2017, to February 28, 2018).

The following covariates were assessed for each patient: age at randomization, gender, CI3 score at randomization, and rates of each outcome measure over the 12 months prior to randomization (March 1, 2016, to February 28, 2017). Spending and utilization rates were calculated from claims data provided by Amerigroup. CI3 scores were provided by Amerigroup. Demographic data were extracted from the CareMore electronic data warehouse.

Baseline characteristics for patients randomized to complex care management and those randomized to usual care were compared using standardized mean differences (SMDs).

 Proposed cutoffs vary, but SMDs greater than 0.1 or 0.25 are generally considered to represent an important difference between groups.

The effect of the program on the primary and secondary outcomes was analyzed through an intention-to-treat approach based on random assignment. Count-based utilization outcomes (IP admissions, IP bed days, ED visits, care center visits, and specialist visits) were modeled with Poisson regression, and the spending outcome (TME) was modeled with ordinary least squares regression. Both adjusted and unadjusted analyses were performed. Adjusted models included age at randomization, gender, CI3 score at randomization, and the corresponding outcome measure over the 12-month period prior to randomization (eg, the model for IP days included IP days from March 1, 2016, to February 28, 2017, as a covariate). Adjusted and unadjusted results were similar. Results from the adjusted models are presented below, with unadjusted results available in the 

 value of .05 was considered to indicate statistical significance. 

 values were adjusted for multiple comparison using the Holm-Bonferroni method.

 All analyses were performed using R version 3.5.1. The evaluation was implemented as a randomized quality improvement trial without institutional review board approval or waiver.

 depicts the flow of patients through the trial. A total of 253 patients met eligibility criteria and were randomized in a 1:2 ratio to complex care management (n = 93) or usual care (n = 160). More than half (56.9%) of the patients randomized to complex care management ultimately enrolled in the program. The most common reason for not enrolling was an inability for the CHW to reach the patient via telephone (37 of 40 patients). A small number of patients were reached but declined to participate in the program (3 of 40 patients). Across both groups, roughly 20% of patients were lost to follow-up in the 12 months after randomization. The intention-to-treat analysis included 71 patients in the complex care management group and 127 patients in the usual care group. Information on appointment-kept rates is provided in the eAppendix.

Baseline rates of spending and utilization were high. The average patient incurred TME of $22,714, spent 6.81 days in an IP facility, and visited the ED 3.30 times during the 12-month period prior to randomization. 

 compares baseline characteristics between patients randomized to complex care management and those randomized to usual care. Compared with patients in the usual care group, patients in the complex care management group were slightly younger (mean age of 43.8 vs 45.9 years), had lower predicted future costs (mean CI3 score of 8.56 vs 9.99), and visited the ED more frequently (3.82 vs 3.01 mean visits).

 compares adjusted rates of utilization and spending between patients randomized to complex care management and those randomized to usual care over the 12 months following randomization. The complex care management program was associated with lower TME (adjusted difference, —$7732 per member per year [PMPY]; 95% CI, –$14,914 to –$550; 

 = .014), and fewer specialist visits (adjusted difference, —1.35 PMPY; 95% CI, –1.98 to 0.73; 

 <.001). There was no significant impact on ED or care center visits.

The eAppendix includes summary vignettes for selected patients enrolled in complex care management. These are intended to supplement the quantitative findings from the primary analysis and offer additional detail on the implementation of complex care planning.

A complex care management program reduced spending and inpatient utilization among high-need, high-cost Medicaid patients. Patients randomized to complex care management had TME that were 37% lower than those randomized to usual care, an absolute reduction of $7732 per patient per year. This spending reduction appeared to be driven primarily by decreases in inpatient utilization—bed days were reduced by 59% and admissions by 44%.

Several potential mechanisms could explain the observed decreases in utilization and spending. Improved management of medical, social, and behavioral risk factors may have prevented acute exacerbations of chronic disease. Patient engagement and activation may have led to improved self-management and adherence. Finally, social and behavioral stabilization may have facilitated safe discharge planning, reducing the need for, or duration of, inpatient admissions.

Complex care management led to a small but significant decrease in specialist visits. Although prior studies of complex care management have not examined specialist use, data from primary care improvement initiatives suggest that enhanced primary care services can reduce downstream specialist utilization.

 In this study, specialist utilization could have been decreased by increased engagement with primary care, better management of chronic medical conditions in the primary care setting, or better communication and coordination between the PCP and specialists. Because there were no additional financial or nonfinancial incentives to reduce utilization in the complex care management group, it is less likely that the reduction in specialist utilization was due to restricted access.

Complex care management had no impact on ED utilization. This contrasts with findings from some programs targeting Medicare patients

 and may be explained by a differential approach to ED utilization between Medicare and Medicaid patients. Research suggests that even with access to primary care, Medicaid patients use the ED at high rates.

 Additional research is needed to better understand the drivers of ED utilization among high-need, high-cost Medicaid patients.

Articles

Complex care management for high-need, high-cost Medicaid patients significantly reduced total medical expenditures and inpatient utilization in a randomized quality improvement trial.

Impact of Complex Care Management on Spending and Utilization for High-Need, High-Cost Medicaid Patients

Telemedicine offers a promising solution to the growing physician shortage, but state-based medical licensing poses a significant barrier to the widespread adoption of telemedicine services. We thus recommend a mutual recognition scheme whereby states honor each other’s medical licenses. Successfully implementing mutual recognition requires policy, technological, and administrative changes, including a federal mandate for states to participate in mutual recognition, consistent standards for using and regulating telemedicine, a mechanism to enable interstate data sharing, financial support for states, and a “state of principal license” requirement for physicians. Reforming the United States’ outdated system of state-based medical licensure can help meet patient demand for virtual care services and improve access to care in rural and medically underserved areas.

The United States’ antiquated system of state-based medical licensure creates unnecessary hurdles that reduce the potential of telemedicine to address the growing physician shortage and improve access to care in rural and medically underserved areas.

Mutual recognition, a scheme whereby states honor each other’s medical licenses, is a comprehensive potential solution to enhance license portability and eliminate barriers to virtual care.

Implementing mutual recognition necessitates policy, technological, and administrative reforms to protect patient safety, prioritize the financial well-being of states, and streamline physician reimbursement.

By 2030, the United States could see a shortage of nearly 120,000 physicians.

 Telemedicine can alleviate the impacts of this shortage by helping physicians make use of unused time and see additional patients, allowing patients to access a larger pool of physicians and connecting specialists to hospitals in rural and medically underserved areas.

 Although telemedicine is growing in use and acceptance, state licensing laws create false geographic barriers and pose a significant challenge to widespread adoption.

In the United States, state medical boards regulate physician licensing, creating a patchwork of inconsistent state licensure laws. With few exceptions, physicians must acquire and maintain a license for each state in which they practice medicine. This antiquated system of state-based medical licensure, originally enacted in the 19th century to reduce medical malpractice and protect patient safety, has profound implications for the promise of telemedicine to increase access to care for vulnerable populations and mitigate the impacts of the national physician shortage.

 In this commentary, we argue for moving away from state-based medical licensure and describe policy, technological, and administrative changes necessary for moving toward portable medical licensure.

Physicians must be licensed in each state where current and future patients are located, so physicians practicing telemedicine across state borders may be responsible for obtaining and staying compliant with up to 51 different state practices of medicine at any given time. There are some exceptions—for example, 10 states issue special purpose licenses for physicians who wish to come to their state for a limited time, scope, and purpose, such as to demonstrate a new technique or to educate medical students.

 However, these exceptions are few and far between.

The multistate licensure process is long and expensive; some states require physicians to pay annual license renewal fees, complete additional coursework, submit required documentation, and participate in interviews. Even after physicians complete these requirements, state medical boards can take several months to process licensing applications.

 Along with imposing substantial direct costs on physicians, state-specific licensing reduces competition that could lower healthcare prices, limits opportunities for physicians to gain experience by seeing more patients, and exacerbates health disparities.

 The impact of restricting telemedicine falls hardest on poor patients, the uninsured, and those who rely on state Medicaid programs, many of whom lack access to reliable transportation and cannot travel across state lines to see specialists.

Over the past decade, the Federation of State Medical Boards (FSMB), a national nonprofit organization that represents 70 state medical and osteopathic boards, has advanced several proposals to enhance license portability and reduce regulatory barriers to telemedicine. These include the Uniform Application and the Federation Credentials Verification Service, which are both web-based applications that eliminate the need for physicians to reenter identifying information and credentials when applying for multiple licenses. Most recently, the FSMB instituted the Interstate Medical Licensure Compact, an agreement to expedite the medical licensure process among member states. Physicians in good standing can freely practice in member states as long as they possess a “full and unrestricted” license in their state of principal license (SPL). To date, only 24 states have joined the compact.

These proposals are a step in the right direction but far from the solution. They simply streamline—not eliminate&mdash;the process of applying for multiple medical licenses. Additionally, they do not reduce the cost to doctors of maintaining multiple medical licenses, estimated at $300 million each year.

 As a more comprehensive solution, we recommend a mutual recognition scheme whereby states honor each other’s medical licenses. This model has been successfully adopted in Europe and Australia and by the Veterans Health Administration, US military, and US Public Health Service.

 Furthermore, because standards for medical education apply nationwide and physician training requirements are set by federal agencies such as HHS, mutual recognition is warranted.

Proponents of the status quo argue that mutual recognition compromises patient safety, reduces revenues from state licensing fees, and complicates physician reimbursement. To address these challenges, mutual recognition should be accompanied by (1) a federal mandate, (2) consistent standards for using and regulating telemedicine services, (3) increased data sharing among states, (4) financial support for states, and (5) a requirement for physicians to select an SPL.

To reduce barriers to interstate medical practice, some states have attempted unilateral action. For example, in 2016, the Florida House of Representatives passed a bill with a provision allowing physicians licensed in other states to offer telemedicine services in Florida. However, the Florida Senate eliminated the provision. To avoid similar situations, Congress can require states to participate in mutual recognition. In fact, legal research suggests that federal action to promote interstate telemedicine is justified based on the Commerce Clause of the US Constitution, which states that Congress has the power “to regulate commerce…among the several states.”

 Another benefit of instituting a federal mandate is a consistent set of definitions needed to support mutual recognition (eg, SPL).

Standards for Using and Regulating Telemedicine Services

Each state currently defines the “practice of medicine” differently, making it difficult to discern what constitutes an acceptable telemedicine consultation in any given state. Standards of practice, conduct, and behavior during telemedicine consultations—including requirements related to physician credentialing, patient education, and physician supervision of other healthcare professionals&mdash;vary widely among states.

 Thus, the shift to mutual recognition must be accompanied by efforts to establish consistent standards for using and regulating telemedicine services. These standards should be defined at the federal level.

State medical boards are tasked with the responsibility of monitoring and disciplining physicians licensed in their state. However, due to gaps in information sharing among states, nearly one-third of physicians disciplined in one state are able to practice elsewhere without limitations, repercussions, or public disclosure. This is especially dangerous in a mutual recognition scheme in which out-of-state physicians routinely see and treat patients with little oversight.

For mutual recognition to promote patient safety, information on malpractice, medical errors, and license cancellation or suspension must be shared among states, made publicly available to patients, and used to enforce disciplinary actions across state borders. States should use resources like the National Practitioner Data Bank, established in 1986 as a central data repository for malpractice payments and state disciplinary actions, to conduct rigorous background checks before physicians participate in telemedicine consultations and deliver care across state borders.

Given the administrative and technological costs of implementing mutual recognition and dismantling the existing state-based medical licensure system, HHS could provide states with incentive payments for adopting mutual recognition agreements and eliminating state-specific licensing and renewal fees. Additionally, HHS could offer grant funding to enhance interstate data sharing systems and other functions that promote mutual recognition. While financial support for states is an important first step, mutual recognition can reduce healthcare costs in the long run as telemedicine services increase access to care for rural and underserved populations.

One challenge of mutual recognition is attributing physicians to any given state. This can complicate physician reimbursement for telemedicine services and reduce the effectiveness of state healthcare programs dependent on physician participation. A possible solution is requiring physicians to select a single SPL for questions regarding reimbursement and attribution. This requirement already exists for physicians participating in the Interstate Medical Licensure Compact. Because physicians would fall under the jurisdiction of their SPL’s medical board, the SPL could hold primary responsibility for collating information from interstate data-sharing systems and enforcing disciplinary action against errant physicians.

The environment of medicine is changing. More than three-fourths of patients want access to virtual care services, and patients across all age groups express a desire to use telemedicine to gain easier, more immediate access to physicians.

 To meet growing patient demand and solve our country’s critical physician shortage, we can and must reform our outdated state-based licensing system.

 Harvard Medical School (PC), Boston, MA; CareMore Health System (SHJ), Cerritos, CA; Stanford University School of Medicine (SHJ), Stanford, CA.

 Dr Jain is an employee of CareMore and Aspire Health, which are both multistate provider entities. Ms Chandrashekar reports no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

 Concept and design (PC, SHJ); drafting of the manuscript (PC); critical revision of the manuscript for important intellectual content (PC, SHJ); administrative, technical, or logistic support (SHJ); and supervision (SHJ).

 Sachin H. Jain, MD, MBA, CareMore Health System, 12900 Park Plaza Dr, Cerritos, CA 90703. Email: Sachin.Jain@caremore.com.

1. Dall T, West T, Chakrabarti R, Reynolds R, Iacobucci W. 2018 update: the complexities of physician supply and demand: projections from 2016 to 2030. The Heartland Institute website. heartland.org/_template-assets/documents/publications/aamc_2018_workforce_projections_update_april_11_2018.pdf. Published March 2018. Accessed November 18, 2018.

2. Bodenheimer TS, Smith MD. Primary care: proposed solutions to the physician shortage without training more physicians. 

. 2013;32(11):1881-1886. doi: 10.1377/hlthaff.2013.0234.

3. Kocher R. Doctors without state borders: practicing across state lines. 

 Blog website. healthaffairs.org/do/10.1377/hblog20140218.036973/full. Published February 18, 2014. Accessed November 18, 2018.

4. Svorny S. Liberating telemedicine: options to eliminate the state-licensing roadblock. Cato Institute website. cato.org/publications/policy-analysis/liberating-telemedicine-options-eliminate-state-licensing-roadblock. Published November 15, 2017. Accessed December 9, 2018.

5. Schwamm LH. Telehealth: seven strategies to successfully implement disruptive technology and transform health care. 

. 2014;33(2):200-206. doi: 10.1377/hlthaff.2013.1021.

6. Vestal C. Why physician licensing is a problem for telemedicine. Governing website. governing.com/news/headlines/why-telemedicine-is-a-problem-for--.html. Published March 7, 2014. Accessed December 9, 2018.

7. Heuser EZ. What do consumers want from virtual visits? Advisory Board website. advisory.com/research/market-innovation-center/research-briefs/2017/virtual-visits-briefing. Published April 27, 2017. Accessed December 9, 2018.&ensp;

In this commentary, the authors argue for moving away from state-based medical licensure and describe policy, technological, and administrative changes necessary for implementing portable medical licensure.

Eliminating Barriers to Virtual Care: Implementing Portable Medical Licensure

ABSTRACT                                     

Costly analogues limit affordability in patients, leading to medication nonadherence and worsening diabetes complications. Evaluating the insulin conversion from analogues to human insulin may determine if this cost savings contributed less toward the initial coverage limit while assessing its impact on glycated hemoglobin (A1C) and medication adherence.

Retrospective and descriptive analysis of patients with type 2 diabetes.

Patients who filled ≥1 insulin prescription between January 1, 2013, and December 31, 2018, were included. No exclusions applied. The primary outcomes assessed percent of human insulin prescribed, cost of both insulins, insulin costs per user, and average last A1C per year. The secondary outcome assessed medication adherence. Outcomes were obtained from pharmacy claims and lab data and compared as pre-insulin conversion (2013-2014), insulin conversion (2015-2016), and postinsulin conversion (2017-2018) periods.

The percentage of human insulin decreased from 11% in 2013 to 10% in 2014 (pre-intervention), increased from 24% in 2015 to 68% in 2016 (intervention), and further increased to 69% in 2017 before dropping to 68% in 2018 (post intervention). Insulin costs per user were $2658.07 in 2013, $3332.65 in 2014, $3119.34 in 2015, $2264.78 in 2016, $2446.78 in 2017, and $2842.10 in 2018. The average last A1C increased by 0.16% from 2013 to 2018. The medication possession ratio ranged from 79% to 86% from 2013 to 2018.

The insulin conversion was associated with a clinically insignificant 0.16% increase in the average last A1C from 2013 to 2018 but improved adherence. Insulin cost did not contribute as much toward the initial coverage limit after 2016.

The high cost of analogue insulin affects healthcare spending and negatively impacts patients by limiting affordability, worsening medication adherence, and leading to microvascular and macrovascular complications. Human insulin is a more cost-effective alternative to expensive analogues, yet analogues are often used for a basal-prandial strategy by many providers. Through this claims analysis that evaluated the insulin conversion from analogues to human insulin since 2015, we found that:

Cost-savings helped prevent or delay reaching the coverage gap, or “donut hole”

There was a clinically insignificant increase in the average last glycated hemoglobin of the year by 0.16%

Medication possession ratio continued to increase from 83% to 86%

Approximately 629 million people are projected to have diabetes by 2045, which constitutes a 48% increase from 2017 to 2045.

 Insulin is considered the most effective medication to achieve glucose control and avoid further complications in patients with diabetes.

 However, 1 in 2 people needing insulin lack access globally. High prices and lack of availability are thought to be key contributors to poor insulin access.

Studies comparing the clinical efficacy of analogue insulin to human insulin are scarce. The American Diabetes Association 

Standards of Medical Care in Diabetes-2019

 currently does not provide guidance for analogues versus human insulin. The guideline states that once-daily basal insulin injection therapy is associated with minimal adverse effects while multiple daily injections may be too complex for older patients.

 The only differentiation made between the 2 insulins by the ADA is the high cost of analogues and the low cost of human insulin. In addition, Fullerton et al’s systematic review comparing short-acting analogue insulin versus regular human insulin in adults with type 2 diabetes (T2D) demonstrated no clear benefit of short-acting analogue insulin over regular human insulin.

CareMore Health, a care delivery system, examined its insulin spending in 2014. The organization found that its Medicare Advantage patients were using analogues that had a high daily injection burden (ie, basal-prandial insulin strategies) and were reaching the Medicare Part D coverage gap.

In February 2015, CareMore started a protocol-driven intervention to switch from higher-cost analogues to lower-cost human insulin in patients with T2D led by clinical pharmacists at various Comprehensive Care Clinics. Patients who were on both basal and prandial insulin receiving at least 50 units/day, without any history of hypoglycemia and adherent to insulin, were eligible for the conversion. Clinical pharmacists determined the total daily dose (TDD), discontinued the basal and prandial insulins and other secretagogues, and converted to premixed insulin by 20% less than the calculated TDD. Patients were given two-thirds of TDD at breakfast and one-third of TDD at dinner then titrated to target predinner

glucose and fasting glucose, respectively. The implementation of the insulin conversion program was further supported by other healthcare providers with experience in chronic disease management.

By encouraging patients to use human insulin, this pilot program led to preference for regimens containing fewer daily injections and a delay in reaching the Medicare Part D coverage gap. The insulin switch was found to be associated with a small increase in population-level A1C of 0.14%, and the intervention was not associated with changes in rates of serious hypoglycemia or hyperglycemia.

 Our study further aims to assess the impact of insulin conversion through descriptive analysis in order to determine if this cost savings contributed less toward the initial coverage limit. Overall, this study is an update from the original findings demonstrated by Luo

in which the average last A1C for the year and medication adherence were also evaluated from 2013 to 2018.

Any patients who filled ≥1 insulin prescription between January 1, 2013, and December 31, 2018, were eligible for this study. No exclusions were applied. The primary objectives were to assess the percent of human insulin (calculated from the normalized prescriptions of both insulins), the cost of insulin therapy, insulin costs per user, and the average last A1C per year. The secondary objective was to assess if there was improvement in insulin adherence. Adherence was calculated using medication possession ratio (MPR) based on days supply of basal or mixed insulin. All data were obtained from pharmacy claims and lab data from 2013 to 2018. The overall trends in primary and secondary outcomes were compared in 3 study periods: pre-insulin conversion (2013-2014), insulin conversion (2015-2016), and postinsulin conversion (2017-2018).

The implementation of the insulin switch program at CareMore was part of the QI project and therefore, no internal review board approval was necessary.

Pre-Insulin Conversion Period (2013-2014)

The total normalized prescriptions filled for both insulins increased from 96,968 in 2013 to 101,298 in 2014, which consisted of only 11% and 10% of human insulin prescriptions being filled, respectively (

). The combined cost of both insulins was $25,336,724 in 2013 and $33,936,348 in 2014; the insulin costs per user were $2658.07 and $3332.65, respectively. The average last A1C was 8.06% in 2013 and 8.02% in 2014 while the MPRs were 80% and 81%, respectively (

Insulin Conversion Implementation Period (2015-2016)

The total normalized prescriptions filled for both insulins decreased from 98,779 in 2015 to 85,755 in 2016. Human insulin prescriptions rose from 24% to 68% in this same timeframe (Figure 1). The combined cost of both insulins was $32,880,942 in 2015 and $21,893,628 in 2016, and the insulin costs per user were $3119 and $2265, respectively. The average last A1C was 8.17% in 2015 and 8.31% in 2016, while the MPRs were 79% and 83% (Table 1).

Postinsulin Conversion Period (2017-2018)

The total normalized prescriptions filled for both insulins increased from 90,376 in 2017 to 92,352 in 2018. During this period, 69% and 68% of insulin prescriptions were for human insulin in 2017 and 2018, respectively (Figure 1). The combined cost of both insulins was $24,164,374 in 2017 and $29,313,460 in 2018 while the insulin costs per user were $2446.78 and $2842.10 . The average last A1C was 8.23% in 2017 and 8.22% in 2018, while the MPRs were 85% and 86% (Table 1).

There was an increase in the number of normalized prescriptions filled, but only 11% and 10% were for human insulin in 2013 and 2014, respectively. In 2013, the insulin cost per user was $2658.07 while the initial coverage limit was $2970. However, the costs of some analogue insulins increased in 2014, and the total insulin cost per user increased to $3332.65 while the initial coverage limit was only $2850. This cost suggests that the majority of insulin users would have reached the Medicare Part D coverage gap from insulin use alone. This was the issue that created a need for the pilot program. At this time, a higher percentage of analogues were prescribed and a majority of the clinicians were still practicing the basal-prandial strategy.

There was a sharp drop in analogue use. From 2015 to 2016, human insulin prescriptions jumped from 24% to 68%. This abrupt increase was expected due to the insulin conversion project. In addition, the total cost of insulin decreased substantially, from $32,880,942 in 2015 to $21,893,628 in 2016 due to the more cost-effective human insulin being prescribed. The insulin cost per user also decreased from $3119.34 in 2015 to $2264.78 in 2016. Because the initial coverage limit in 2015 was $2960, patients were already falling into the Medicare Part D coverage gap, from insulin prescriptions alone. However, the initial coverage limit in 2016 was $3310, and insulin users were more likely to avoid reaching the coverage gap due to the increased use of more cost-effective insulin. In addition, in January 2016, select plans moved the analogue products containing glargine, detemir, or aspart from tier 6 ($0 copay) to tier 3 ($37.50 copay for most patients, with additional out-of-pocket payments if patients were in the Medicare Part D coverage gap), while human insulin products remained on a tier with a $0 copay.

 This formulary change may have also influenced the providers to prescribe less of the costly analogues.

During the postconversion period, there was an increase in total normalized prescriptions, while human insulin prescriptions remained relatively stable at 68% to 69%. This suggests that the implementation of the insulin conversion had been sustained since 2015. Similarly, the insulin cost per user remained low at $2446.78 and $2842.10 in 2017 and 2018, respectively. As the initial coverage limit also increased to $3700 in 2017 and $3750 in 2018, more insulin users were able to delay reaching the Medicare Part D coverage gap. The rise in cost during this timeframe may be attributed to 2 formulary changes made for 2018. ReliOn insulin, a lower-cost human insulin, was removed from the formulary. Furthermore, human insulin pens were introduced to the formulary and are more costly than the vials.

The average last A1C levels ranged from 8.02% during the pre-intervention period in 2014 to 8.31% during the intervention period in 2016. The average last A1C levels during the postintervention period also remained relatively stable, at 8.22% to 8.23%. This suggests that there was only a 0.16% increase in the average last A1C over the previous 5 years, which is similar to Luo et al’s finding of population-A1C level increase by 0.14% from the insulin switch.

Since change in A1C by 0.5% has been suggested to be clinically meaningful, this study’s results also suggest that the insulin conversion is associated with a clinically insignificant increase in A1C level.

There was an overall increase in MPR from 79% to 86% from 2013 to 2018. The continuously rising MPR—79% in 2015 to 83% in 2016 to 85% in 2017 to 86% in 2018&mdash;suggests that the improvement in insulin adherence may be attributed to the conversion intervention after 2015.

Limitations to this study include no data on the rate of hypoglycemia and hyperglycemia to suggest whether the insulin switch increased the prevalence of serious hypoglycemic and hyperglycemic events since 2016. This study also does not evaluate average daily injections or overall patient satisfaction as a result of the potential reduction in number of daily injections from switching from analogue to human insulin.

This study’s results further demonstrated the impact of cost savings, the average last A1C for each year, and medication adherence after the conversion from analogue insulin to human insulin between 2013 and 2018. Clinical pharmacists’ efforts in researching, developing, and implementing this pilot project in 2015 helped to reduce the cost barrier to insulin and delayed reaching the Medicare Part D coverage gap each year from 2016 to 2018.

The authors would like to thank Jim Rose, PharmD, Ani Davis, PharmD, and Balu Gadhe, MD (along with Sachin H. Jain, MD, MBA) as part of the initial team to launch the insulin conversion program in 2015. The authors would also like to thank the current clinicians (clinical pharmacists, nurse practitioners, physician assistants, and physicians) at CareMore Health for sustaining the insulin implementation since 2015.

Kyunghwa Park, PharmD; Angela Eng Jeong, PharmD; and Eric Guenther-Gleason, PharmD, MS; are all employed by CareMore Health, Cerritos, California.

Ewen M, Joosse JH, Beran D, Laing R. Insulin prices, availability and affordability in 13 low-income and middle-income countries. 

. 2019;4(3):e001410. doi: 10.1136/bmjgh-2019-001410. eCollection2019.

Perez-Nieves M, Jian D, Eby E. Incidence, prevalence, and trend analysis of the use of insulin delivery systems in the United States (2005 to 2011). 

 2015;31(5):891-899. doi: 10.1185/03007995.2015.1020366.

American Diabetes Association. Pharmacologic approaches to glycemic treatment: Standards of Medical Care in Diabetes — 2019. 

, et al. Short-acting insulin analogues versus regular human insulin for adult, non-pregnant persons with type 2 diabetes mellitus. 

. 2018;12:CD013228. doi: 10.1002/14651858.CD013228.

Luo J, Khan NF, Manetti T, et al. Implementation of a health plan program for switching from analogue to human insulin and glycemic control among Medicare beneficiaries with type 2 diabetes. 

 2019;321(4):374-384. doi: 10.1001/jama.2018.21364.

Converting from analogue insulin to human insulin is associated with a clinical insignificant increase in glycated hemoglobin of 0.16% but with improved insulin adherence.

Impact of Switching Analogue Insulin to Human Insulin in Diabetes

Social determinants of health lead to increased healthcare utilization and poor outcomes, but efforts to risk-stratify patients have largely ignored these risk factors.

Advanced analytics applied to electronic health records and primary survey collection can reveal distinct phenotypes of individuals with actionable sociobehavioral contributors to health.

Healthcare organizations can target these sociobehavioral phenotypes with evidence-based interventions.

We propose a research agenda to identify, validate, and apply sociobehavioral phenotypes in healthcare.

Commonly used risk algorithms in healthcare, including indices to predict hospital readmission, define “high-risk” patients by disease-based (eg, comorbidity index) or utilization-based (eg, number of emergency department [ED] visits) factors.

 Clinicians and systems can target interventions to such high-risk groups—so-called “precision delivery.”

 However, current programs that target high-risk individuals, including high-risk case management programs, do not uniformly reduce utilization or adverse outcomes.

Why might such interventions be ineffective? Perhaps the problem is in our conception of risk, rather than the interventions.

An individual’s health phenotype has far-reaching connections with social, behavioral, and environmental factors, such as housing instability and food insecurity. Nevertheless, few risk models include variables associated with social determinants of health, such as income level or educational attainment, and traditional risk models often identify such individuals when it is too late to intervene upon socioeconomic risk factors.

 To complement risk-stratification models based on clinical severity, we introduce the “sociobehavioral phenotype”—actionable risk profiles based on empirically derived social, economic, and behavioral determinants of health.

Defining the Sociobehavioral Phenotype: Lessons From Precision Medicine

To build and implement sociobehavioral phenotypes, clinicians, researchers, and systems leaders can draw lessons from the field of precision medicine. Consider the analogy of metastatic non—small cell lung cancer (NSCLC): Prior to the advent of targeted and immune-based therapy, cytotoxic chemotherapy was the standard of care. Genomic sequencing of tumor cells identified pathologic mutations; researchers later targeted such mutations in drug development. Subsequent development of targeted inhibitors and evaluation in randomized trials led to many more therapeutic options becoming available for NSCLC. As new therapies targeting pathologic “driver” mutations became available, oncologists integrated screening for such mutations into the workup of NSCLC.

Just as many driver mutations exist in lung cancer, driver socioeconomic risk factors (eg, poor social support, housing instability) or behaviors (eg, medication nonadherence) underlie many individuals’ health outcomes, such as mortality or acute care utilization. Clinicians and researchers can analyze large observational data sets to classify 1 or more of these common risk factors into distinct sociobehavioral phenotypes and subsequently test interventions targeted toward those phenotypes.

The Added Value of the Sociobehavioral Phenotype

The sociobehavioral phenotype aligns well with broader calls for providers to screen for social determinants of health. In resource-constrained environments, it is of paramount importance to target interventions to the individuals most likely to derive benefit from them. However, current interventions may not serve the majority of patients with predominantly sociobehavioral contributors to poor health. Take, for example, a telephonic case management program targeted to individuals with high likelihoods of hospital readmission. Although such a program may benefit some subtypes of patients, it is unlikely that a telephone-based program would succeed for a homeless individual who does not have a reliable address or telephone number. An unknown phone number may also go unanswered due to fear or suspicion. Furthermore, telephonic or remote case management programs may fail to lower health expenditures because certain high-risk patients rely on frequent in-person contact.

 Sociobehavioral phenotyping based on electronic health record (EHR) data may identify driver risk factors to develop and refine effective interventions for such individuals.

Learning from the precision medicine example, a coordinated research agenda with parallel and iterative data mining and experimentation could identify relevant sociobehavioral phenotypes that ameliorate social determinants of health (

). Mining large databases from EHRs and observational studies are necessary first steps to identify sociobehavioral phenotypes. There are examples of this: Using a large sample of clinical notes from the Veterans Health Administration (VHA), researchers have developed and validated natural language processing lexicons to identify individuals who are homeless or at high risk of being homeless.

 Additionally, neural network methods applied to data from the Health and Retirement Study have yielded potential sociobehavioral phenotypes, such as elderly patients with few children and thus little social support, that are predictive of poor health outcomes.

 Clinicians and researchers will need to supplement these EHR-derived phenotypes with primary survey data collection focused on driver social and behavioral risk factors, perhaps as part of regular screens for socioeconomic determinants of health.

 After defining and validating sociobehavioral phenotypes, clinicians and delivery scientists can identify which phenotypes are actionable by developing randomized trials of interventions in specific subgroups. Unlike the situation in metastatic lung cancer a decade ago, healthcare delivery scientists have an armamentarium of community-based interventions to use among selected high-risk individuals. These include home-based and visiting nurse care, telephonic case management, wireless technologies, community health workers, patient-centered medical homes, social service programs, and direct financial assistance. By examining the effect of a social service or patient engagement intervention stratified by a sociobehavioral phenotype, policy makers and clinicians can determine which subpopulations derive benefit from each intervention, which can facilitate more efficient enrollment into these interventions (

Nascent examples of sociobehavioral phenotyping are a harbinger of broader success. Since 2013, the VHA has deployed a screening for homelessness, the Homelessness Screening Clinical Reminder, for all patients who access outpatient VHA care.

 Applying machine learning models to VHA EHR data reliably identifies patient phenotypes at high risk (>90%) of reporting homelessness, a risk factor for acute care utilization.

 For individuals at high risk of homelessness, the VHA deploys interventions, including a homelessness primary care action team and temporary housing, that are associated with reduced ED utilization among those with previously high utilization.

Identifying Heterogeneity Within Phenotypes

As in a genomics-based precision medicine framework, there will still be heterogeneity within a sociobehavioral phenotype. For example, the aforementioned examples of sociobehavioral phenotyping define phenotypes by singular socioeconomic risk factors, such as homelessness. Although homelessness could contribute to healthcare utilization, reasons for homelessness are varied. The single male who faces chronic housing insecurity due to substance use and mental illness would likely have a different set of needs than the mother of 4 children who is fleeing domestic violence. Each of these reasons demands different interventions to address underlying causes of social instability. Unsupervised clustering algorithms applied to very large data sets could help identify heterogeneity within socioeconomic risk factor groups. Such algorithms may differentiate between the constellation of homelessness, substance abuse, and poor primary care follow-up versus a constellation of homelessness, food insecurity, and diabetes. By leveraging data across a large number of patients, delivery scientists and policy makers can identify combinations of social factors that result in a particular pattern of morbidity, utilization, and cost, then prioritize different interventions or sequences of interventions for such phenotypes.

There are several challenges in defining and implementing socio&shy;behavioral phenotypes. Patient-level socioeconomic data are often not readily available in structured data sets. Additionally, accounting for socioeconomic risk factors in individual patient care may increase information overload among overstretched clinicians—who may have trouble managing the data to which they already have access. And certain sociobehavioral phenotypes, such as food insecurity, carry risks of stigma and further alienating patients. Patients who fall into certain sociobehavioral phenotypes will need protections to ensure that they are not excluded from appropriate care or subjected to bias. Partnerships between academia and patient and advocacy groups could aid in refining data-generated phenotypes and associated interventions to ensure maximal uptake and acceptance.

Targeting interventions based on their ability to improve specific sociobehavioral risk factors is more likely to be successful than broad—and often blind&mdash;application of resource-constrained interventions. Just as oncologists in the precision medicine era screen for actionable mutations to guide use of targeted therapies, delivery scientists and clinicians can study and target interventions based on sociobehavioral phenotypes. Perhaps then we may experience a revolution in addressing socioeconomic determinants of health&mdash;similar to the precision medicine revolution of today.

 Perelman School of Medicine and Leonard Davis Institute of Health Economics, University of Pennsylvania (RBP, ASN), Philadelphia, PA; Corporal Michael J. Cresencz VA Medical Center (RBP, ASN), Philadelphia, PA; CareMore Health System/Anthem Inc (SHJ), Cerritos, CA; Stanford University School of Medicine (SHJ), Palo Alto, CA.

 This work was funded in part by the Pennsylvania Commonwealth University Research Enhancement Program, the Leonard Davis Institute Policy Accelerator Program, and the Pennsylvania Department of Health. The department specifically disclaims responsibility for any analyses, interpretations, or conclusions.

Sociobehavioral phenotypes are actionable risk profiles based on empirically derived social, economic, and behavioral factors that, if applied appropriately, can help healthcare organizations address social determinants of health.

Author | Sachin H. Jain, MD, MBA

Articles

New content coming soon, please check back later.