Telehealth medication management and health care spending in a Medicare Accountable Care Organization

Abstract

BACKGROUND:

Value-based care is an opportunity for medication optimization services to improve medication management and reduce health care spending. The reach of these services may be extended through telehealth. However, as health care systems and payers grapple with the long-term financing of telehealth, real-world assessments are needed to evaluate the potential economic impact of pharmacy-driven telehealth services.

OBJECTIVE:

To evaluate the impact of a scalable pharmacist-driven telehealth intervention to improve medication management on health care spending for clinically complex patients who were enrolled in a Medicare Next Generation Accountable Care Organization.

METHODS:

Data for this pretest-posttest nonequivalent group design study came from Medicare claims from 2015 to 2020 and linked pharmacist care activity data derived from the electronic medical record. Patients in the intervention group were identified as those who received the telehealth medication management service. Patients in the control group were offered the service and refused or could not be contacted. The primary outcome was total medical spending over a 6-month period, and impact was assessed using a covariate-adjusted difference-in-difference model.

RESULTS:

There were 581 patients who received the intervention and 1,765 who served as controls. The telehealth intervention reduced total medical spending by $2,331.85 per patient over the first 6 months of the service ($388.50 per month; P = 0.0261). Across a range of estimates for the cost of service delivery, we find a return on investment of 3.6:1 to 5.2:1.

CONCLUSIONS:

The $388.50 monthly savings found in this study represent a substantial reduction in health care spending and emphasize the opportunity for telehealth delivery of medication management services to improve value as a part of alternative payment models.

Plain language summary

Accountable care organizations can expand pharmacy services through telehealth. The telehealth medication management service evaluated in this study engages pharmacists in population health management for patients with diabetes, hypertension, chronic obstructive pulmonary disease, or heart failure and either high-risk medication use or use of 10 or more chronic medications. This service reduced medical spending by $388.50 per month for the group that received the service vs those that did not: a return on investment of 3.6:1 to 5.2:1.

Implications for managed care pharmacy

Clinically trained pharmacists can increase the breadth and quality of medication-based interventions. The success of this pharmacy-directed intervention emphasizes the need for multimodal care management interventions that use health information technology to manage risk and reduce the total cost of care for high-risk Medicare patients. Telehealth pharmacy-based interventions may be cost-effective services that could be scaled to impact populations served by the health care system.

Health policy trends in 2021 have predictably focused on COVID-19 and strengthening the public health infrastructure. Related to the pandemic, telehealth expansion and reinforcing the move toward value-based payment models have also prevailed, resulting in opportunities to bolster access to primary care through telehealth services. As payers grapple with the long-term financing of telehealth, more evidence and real-world assessments are needed to evaluate the potential impact of telehealth services on patient care and outcomes.^1,2 Relative to the shift to value, Accountable Care Organizations (ACOs) and health care systems continue to invest in the necessary infrastructure to successfully manage risk and value-based payment models.

Even with the growth of telehealth, challenges with continuity of care,³ especially in complex patients with multiple chronic conditions, are pronounced. According to a recent RAND corporation analysis,⁴ over the course of a year the average patient with 3-4 chronic conditions will visit 12 ambulatory providers who will prescribe 24 medications. COVID-19 exacerbated preexisting concerns with medication mismanagement and exposed additional fractures in continuity of care such as reduced in-person visits and limited health care service hours. Medication optimization services present an opportunity to improve medication management and address some of these care gaps. In collaboration with physicians, pharmacists work directly with patients to optimize medication use through services like comprehensive medication management (CMM). In contrast to other services, such as medication therapy management and comprehensive medication reviews, which have become narrowly targeted programs limited mostly to Medicare Part D beneficiaries, CMM is a broadly focused standard of care that assesses an individual’s medications for appropriateness, safety, and efficacy.⁵ Through telepharmacy CMM services, pharmacists educate, manage, and adjust medication regimens to improve care coordination, engage the patient, and solidify the “last mile” of population health, reaching patients in their home through the use of telehealth.

Despite successful integration into care teams across traditional care settings,^6,7 pharmacists have been noted to be missing from the ACO care teams.⁷ In a 2016 report from Pharmacy Benefit Management Institute, more than half (57%) of ACOs reported employing or contracting with a pharmacist.⁸ As medication misuse, underuse, and overuse continue to cost the health care system more than $500 billion annually, the absence of pharmacists from the team creates missed opportunities to improve care and control costs.⁹ Evidence shows pharmacy services improve care and reduce cost in many care settings.^10,11 Across 4 patient-centered medical home sites in Minnesota, team-based medication management services led by a pharmacist demonstrated significant improvement in diabetes goals and 11% reduction in spending growth compared with sites that did not employ team-based care.¹² In another study, pharmacist medication management services over a 10-year period in a large integrated health care system showed improvement in chronic conditions, high patient satisfaction, and a positive return on investment (ROI) of $86 per encounter.¹³ Pharmacists’ ability to impact ACO quality measures has also been documented.⁶

While previous research supports the value of in-person pharmacy services, the value of pharmacy services delivered via telehealth remains underexplored. This research project reports the results of telehealth clinical pharmacy services within an ACO environment and assesses the potential ROI from implementing a similar service in other settings.

In January 2016, University of North Carolina (UNC) Health established population health management services designed to deliver comprehensive, high-quality care to patients of the health care system and clinically integrated network. These services included care management with health coaches, nurses, social workers, and dieticians, as well as a pharmacist-led CMM service through the Carolina Assessment of Medications Program (CAMP). In January 2017, these services, including the CAMP CMM service, were expanded to the UNC Senior Alliance, a Next Generation Accountable Care Organization (NGACO) for Medicare beneficiaries. The NGACO model was characterized by greater risk and greater reward opportunities than Medicare shared savings program ACOs and included benefit enhancements such as waivers of Medicare service rules around telehealth, postdischarge home visits, and skilled nursing facility eligibility rules.

Patients were recruited to the CAMP CMM service in one of the following 2 ways: by prospective population reviews and direct referral by providers. For both recruitment strategies, patients were required to be older than the age 18 years, not reside in a skilled nursing facility, not be seen by a disease state–specific ambulatory care pharmacist at UNC, and have a primary care provider visit in the last 12 months. For the prospective recruitment, NGACO patients with 1 of 4 disease states (diabetes with hemoglobin A1c ≥9%, hypertension, chronic obstructive pulmonary disease, or heart failure) with either high-risk medication use or 10 or more chronic medications were targeted to enroll in the CMM service. High-risk medication use was defined as using drugs with potential for harm in older adult patients, as well as using inhalers, insulins, and narrow therapeutic index medications. Assessment for enrollment eligibility used an automated approach, which combined ACO claims data with data from the electronic health record (EHR). NGACO patients who did not meet all the prospective recruitment criteria could be referred by a provider if they were otherwise in the scope for CMM services. NGACO claims data exist within the health care system’s data warehouse and are linked to the EHR data using a beneficiary ID to medical record number crosswalk.

The objectives of the pharmacist-led CMM service were to customize medications to meet patient-specific disease state goals, identify and resolve medication therapy problems, and reduce polypharmacy. The goal of the service was to engage patients in at least 2 telehealth visits, and the total number of visits varied based on patient need and pharmacists’ clinical assessment. Figure 1 illustrates a detailed description of the service. Pharmacists in the CAMP CMM service engage with patients across the state of North Carolina via telehealth to overcome geographic or transportation-related barriers. The goal of this study was to evaluate the impact of the CAMP CMM service on medical spending in a real-world setting.

FIGURE 1.

Comprehensive Medication Management Service Description

Methods

A pretest-posttest nonequivalent group design¹⁴ was used in this study to evaluate change in medical spending. This design assesses change in an intervention group over time compared with a control group and is a robust pragmatic method for inferring causality in observational studies. Data for this project included enrollment data, demographic information, EHR documentation related to telephone outreach, and all medical claims for NGACO-attributed patients targeted for the CAMP intervention. Although the NGACO did not begin until January 2017, patients who received CAMP services before the start of the NGACO and were later attributed to the NGACO were eligible for inclusion in this study. As this study used only observational data collected as a part of patient care, it was declared exempt by the University of North Carolina-Chapel Hill Institutional Review Board.

STUDY POPULATION AND OUTCOMES

Intervention patients were defined as those who had 2 or more CAMP CMM visits or who had only 1 visit without status change. Control patients were identified as patients who received an outreach call during the same period but did not enroll in the CAMP clinic or those who enrolled but did not receive the full intervention. The primary reason for not receiving the full intervention included loss to follow-up due to not showing up at appointments and not rescheduling or the inability to contact by phone for additional follow-up appointments. Index dates for the intervention patients were the date on which they had their first CAMP visit. Index dates for the control patients was the earliest status change date recorded after the enrollment phone call.

Initial assessment of claims availability found that cost estimates for pharmacy claims data were not available in the NGACO records; therefore, total medical spending, defined as the sum of charges for all medical services, was the outcome of interest for this study. NGACO medical claims for 6 and 12 months before and after index date were eligible for inclusion and were obtained from July 27, 2015, to May 18, 2020. Although summing spending across the 6-month measurement period was used for the primary analysis, the 12-month period was added as a sensitivity analysis. To account for the presence of extreme values, we winsorized the distribution by capping spending values across both cohorts at the 99th percentile for the preperiod and postperiod separately.

STATISTICAL APPROACH

To reduce the risk of bias and confounding, we controlled for differences in age, sex, race, and comorbidities between intervention and control cohorts. Age at the time of index was divided into the following 5 broad categories: 18-49, 50-64, 65-74, and 75+ years. Race was modeled as a binary variable representing White, non-Hispanic vs person of color. The count of Elixhauser comorbidities¹⁵ was used to account for clinical risk and was modeled as 0, 1-2, 3-6, 7-9, and 10+ conditions. In comparison with other measures of disease burden, Elixhauser is more expansive than the Charlson Comorbidity Index¹⁶ and is an appropriate choice for modeling disease burden in younger adult as well as older adult populations. Descriptive statistics were used to evaluate differences in demographics, comorbidities, and medical spending between control and intervention at baseline. To estimate the effectiveness of the CAMP clinic and control for differences in demographics and clinical risk between cohorts, we used a generalized linear model with a repeated measure for patient, logit link, and negative binomial distribution. The primary independent variable of interest, the difference in medical spending between the 2 groups over time, was derived from the interaction of cohort and period (difference-in-difference indicator).¹⁴ Analyses were conducted in SAS v9.4 (SAS Institute Inc.) and PROC GENMOD was used for the statistical models.

Results

There were 581 patients who met criteria for inclusion in the intervention group and 1,765 for the control group. The most common reasons controls did not receive the service were loss to follow-up after the initial outreach attempt and declining the service. Demographically, both samples had a majority of patients aged 65+; were majority White, non-Hispanic; and were majority female (Table 1). No significant differences were observed across cohorts for these 3 variables. For comorbidities, patients in both cohorts had substantial comorbidity burden with a mean of 6 conditions for the intervention group and 5.5 for the control group and a mode category of 3-6 conditions for both groups. These differences in comorbidities were statistically significant across groups for both the count of comorbidities (P < 0.001) and frequency across condition count categories (P = 0.02).

TABLE 1.

Demographics and Comorbidities for the 6-Month Eligibility Cohort

Variable	Intervention (n = 581)	Control (n = 1,765)	P value
Age (years), n (%)
18-49	31 (5.3)	123 (7.0)	0.07
50-64	59 (10.0)	219 (12.4)
65-74	249 (43.0)	661 (37.5)
75+	242 (41.7)	761 (42.1)
Race, n (%)
White, non-Hispanic	456 (78.5)	1,426 (80.8)	0.23
Sex, n (%)
Female	348 (60.0)	1,019 (57.8)	0.37
Male	233 (40.0)	746 (40.2)
Elixhauser comorbidity count (mean, SD)	6.0 (3.2)	5.5 (3.0)	<0.001
Elixhauser comorbidity count, n (%)
0	6 (1.0)	41 (2.3)	0.02
1-2	68 (11.7)	219 (12.4)
3-6	281 (51.8)	914 (48.4)
7-9	134 (22.1)	391 (23.1)
10+	92 (15.8)	200 (11.3)

Winsorizing capped the distribution of spending for the 6 months prior to intervention at $85,533.83 for the preperiod and $82,121.40 for the postperiod. For the intervention group who received the CMM service, unadjusted average actual medical spending decreased from $12,497.12 to $10,305.08 and medical spending for the control group decreased from $10,051.14 to $9,399.65 (Figure 2). When accounting for differences in demographics and comorbidities, the average predicted spending for the intervention group decreased from $12,811.19 to $10,684.16 and spending for the control group increased from $9,581.43 to $9,786.25, also shown in Figure 2. The predicted estimates are the estimated marginal mean estimates from the statistical model and analogous to least square means derived from analysis of variance models. The difference in these predicted mean differences, $2,331.85, measures the impact of the CMM service on total medical spending over the 6 months following enrollment and is statistically significant at P = 0.0261 (Supplementary Table 1 ^{(128.8KB, pdf)} , available in online article). This difference translates to a $388.50 difference in per patient per month spending and is similar to the statistically significant (P = 0.0392) $380.75 per patient per month spending difference observed with the 1-year preperiod postperiod difference subanalysis (Supplementary Table 2 ^{(128.8KB, pdf)} ).

FIGURE 2.

6-Month Total Medical Spending Before and After Intervention, Enrolled vs Control

Discussion

The $388.50 in per patient per month savings found in this study represent a substantial reduction in health care spending for the clinically complex population targeted for the CAMP telehealth CMM service. Comparing this estimate with the literature finds that this estimate is on the high end, with estimates of the impact of services on patients with polymedicine ranging from $16 to $414 for experimental studies but only $0.25 to $209 for observational studies.¹⁷

Although the program was associated with substantial savings, operating this service required substantial resource utilization. There, to conceptualize the potential ROI for this program, we extrapolated these savings estimates from the historic cohort to the 369 patients who received CAMP CMM services in the 2020 calendar year (Table 2). At a 6-month savings of $2,331.85 per patient, the program was estimated to have reduced total medical spending by $860,452.65 for patients enrolled in 2020. Based on internal accounting, providing the service in 2020 required 0.7 pharmacist full-time equivalents (FTEs), 0.3 technician FTEs, and 0.1 analyst FTEs. Using FTE estimates, salaries, administrative costs, and overhead costs from the same year ($188,700), we estimate a ROI from the CAMP CMM service of 3.6:1 ([$860,452.65-$188,700])/$188,700).

TABLE 2.

Comprehensive Medication Management Service Cost Savings and Return on Investment Calculation

	Value
(A) Estimated difference between intervention group and comparison groups per patient medical spending over 6 months	$2,331.85
(B) Difference between comparison groups and intervention group medical costs across 369 patients enrolled in intervention (A*369)	$860,452.65
(C) Annual cost of pharmacist services to deliver the intervention	$188,700.00
Pharmacist salary and fringe ($161,642 × 0.7 FTE)	$113,149.00
Technician salary and fringe ($58,734 × 0.3 FTE)	$17,620.00
Analyst salary and fringe ($84,318 × 0.1 FTE)	$8,431.00
Administrative costs ($25/sq. ft × 150 sq. ft per employee × 1.1 FTE)	$49,500.00
(D) Estimated return on investment ([B-C]/C)	3.6:1

FTE = full-time equivalent.

This estimate assumes pharmacists will continue to deliver this telehealth intervention from an office building leased by the health care system, which may or may not be true. Over the past year, the CAMP team worked remotely as a result of the COVID-19 pandemic. However, lease contracts were already secured for this time period and thus were included in the 3.6:1 ROI calculation. As telehealth delivery of services continues permanently, this administrative cost is expected to decrease as the team’s work transitions to permanent remote telehealth delivery. Given the successful delivery of the CMM telehealth service by pharmacists working remotely, organizations developing CMM services can consider creation of a remote service, thereby eliminating the administrative costs of being on-site and resulting in an estimated 5.2:1 ROI. Also of note, revenue was not generated by the CMM service at the time of the study because of complexity of reimbursement for telehealth pharmacist services. Therefore, reimbursement was not included in the ROI calculations, and if revenue were to be included, the ROI would be greater than our calculated estimates, which are based on estimated cost offsets alone.

The significant reductions in medical spending demonstrated in this study provides a compelling argument for continuing this CAMP CMM service as a means to improve care for high-risk Medicare patients within an ACO. Although these ROI estimates demonstrate a robust return, they may be overestimated for a newly formed ACO without a robust data and analytics infrastructure or strong physician engagement strategies. A one-time cost for the program build and implementation process would need to be included in the expense estimate and would reduce the ROI during the initial year of implementation. However, in this example, start-up costs would have had to exceed $670,000 for program costs to exceed program savings.

The success of this pharmacy-directed intervention emphasizes the need for effective use of health information technology aligned to multimodal care management interventions to manage risk and reduce the total cost of care for high-risk Medicare patients.^18,19 Demonstrating the ability to deliver these cost-effective services using telehealth technology is important in order to scale impactful interventions across populations and geographies.¹⁸ Using clinically trained pharmacists to lead pharmacy-directed interventions increases the breadth and complexity of interventions. By overlaying the CAMP program on other ACO population health initiatives, as well as direct physician care, we have demonstrated a positive ROI, contributing to improved patient outcomes and increased shared savings for the ACO.

As ACOs evolve, most look to both grow their attributed lives and diversify their value contract and risk portfolio. ACO success across contracts and populations is contingent on investment in resources and initiatives with demonstrated outcomes and value. Effective programs must then be scaled to increase reach and impact. Since 2016, the CAMP team has grown from 5 pharmacists and no support personnel to 10 pharmacists and 6 certified pharmacy technicians. These pharmacists and technicians provide a range of pharmacy services beyond CMM, including disease-specific initiatives for diabetes, hypertension, chronic obstructive pulmonary disease, opioid and benzodiazepine de-escalation, transitions of care, and anticoagulation management. Indeed, the CMM service for NGACO patients accounts for less than 1 FTE out of the nearly 10 current pharmacist FTEs during the time of the study, with the remaining FTE time spent in the non-CMM services mentioned above or with other patient populations supported by the team outside of NGACO. Our current study effectively creates a model and methodology that can be used to understand the impact of all clinical pharmacy services offered, which is an important step in strategically growing pharmacy population health services within a value-based organization.

Determining a clear and reproducible ROI for specific population health initiatives has proven elusive; however, doing so remains critical to account for the full economics of pursuing value-based care. Like many ACOs, UNC Senior Alliance deploys multiple case management and care management interventions, and determining the relative impact of each intervention also remains difficult. While this study finds a meaningful ROI for the CMM service, the risk stratification methodology used by CAMP pharmacists to identify appropriate patients was not refined and has since undergone multiple iterations. It is likely that, with our more advanced risk stratification methodology and evolved care delivery processes, our pharmacy interventions may show even higher ROI.

Although the use of telehealth for the current program increases flexibility and scalability of the service, preparation and wrap-up for a CMM visit is time intensive and the visits are lengthy. Work has been done to reduce the time commitment of a CMM visit by simplifying the previsit planning and charting processes. Despite this, to limit staff burnout, CMM visits are limited to 2 per pharmacist per day because of this complexity. Current efforts to scale the service focus on ensuring the pharmacists focus on medication-related needs by improving workflows to better leverage care management teams and including community health workers for social and lifestyle interventions. At the time of this study, CMM visits were not billed to payors, limiting direct financial support for the service. The health care system’s broader acceptance of chronic care management billing will allow for future state reimbursement of CMM services, which will bring in additional revenue, improving the financial model for services such as this one. Finally, we are exploring alternative tools for risk stratification, patient communication, and outcomes tracking that will streamline the administrative burden of CMM.

Currently, there is no consistent, systematic approach for demonstrating value, reductions in total cost of care, and ROI across care interventions. This research provides an approach and methodology that can be used consistently across population health interventions to inform ACO investments and scalability.

LIMITATIONS

This study has several limitations worth noting. First, although the use of a control group reduces the risk of bias and confounding and is a superior design to the majority of cost-effectiveness studies of clinical pharmacy services,²⁰ the lack of randomization, nevertheless, leaves open the possibility of alternative explanations for the observed decrease in medical spending for the patients who received the CMM service as compared with controls. In particular, baseline spending in the intervention group was higher than the control group, even after accounting for demographic and clinical differences. Randomization would have likely resulted in no baseline spending difference, leaving the possibility that unobserved differences between the 2 groups may contribute to changes in spending independent of the telehealth intervention evaluated in this study. The statistical approached used in this study helps to reduce bias that may arise through lack of randomization, but claims data provide limited ability to observe factors like true disease burden, which may not be accounted for through disease-related summary measures like the one used in this study. Additionally, patients targeted for the CMM service may have been targeted by other care management services around the same time as they were enrolled in the intervention. As such, the observed decrease in spending may be due to a combination of pharmacy services and services from other providers. Furthermore, observed decreases in spending do not account for any increases in pharmaceutical spending, as prescription drug costs were not available in the data received by the NGACO. Although NGACOs are not responsible for managing prescription drug costs, consideration of these costs as a part of total cost of care is relevant to any ACO serving older patients, which does bear financial risk for pharmaceutical spending. Accounting for this prescription drug spending would likely reduce the ROI from this service, as pharmacist-related interventions such as CMM frequently result in increased prescription drug spending. Finally, the population targeted for this service is clinically diverse, and the services provided by the CAMP clinic vary according to clinical need. Although this heterogeneity is reflective of care provided in the real world, it creates practical challenges determining what aspects of the intervention are effective.

Conclusions

As value-based health care expands and matures, interventions are needed that optimize medications to reduce health care expenditures. This study finds a scalable pharmacist-driven telehealth intervention significantly reduced per-patient total medical spending by $2,331.85, or $380.75 per patient per month, over a 6-month period. Although these savings are substantial, additional work is needed to explore the mechanisms by which cost reductions are achieved through medication management interventions and the interaction between pharmacist-delivered programs such as this and those delivered by other health care providers. These results highlight the possibility of medication optimization services as investments to reduce health care spending, and the telehealth service delivery model aligns well with changes to patients’ care delivery preferences as a result of the COVID-19 pandemic.