Medicare Annual Wellness Visit Association With Healthcare Quality and Costs

Adam L. Beckman, BS; Adan Z. Becerra, PhD; Anna Marcus, BS; C. Annette DuBard, MD, MPH; Kimberly Lynch, MPH; Emily Maxson, MD; Farzad Mostashari, MD, ScM; and Jennifer King, PhD

The Annual Wellness Visit (AWV) was introduced in 2011 by Medicare and made available to all eligible beneficiaries without deductibles or co-payments. Unlike a traditional periodic health examination or annual physical, which may be performed without a specified protocol,1-4 the AWV includes a list of required components5 that prioritize preventive care and investing in the relationship with the patient rather than addressing acute complaints or chronic disease (Table 1). It includes assessing risk factors, inquiring about care support, creating a personalized care plan, and educating beneficiaries on how to maintain their health outside of an acute illness episode.6-8 Notably, the only physical examinations required of the visit are blood pressure measurement and height/weight measurement for body mass index (BMI), reflecting the US Preventive Services Task Force’s recommendation against routine physical examinations for asymptomatic adults 65 years or older.9

Previous work has demonstrated growing adoption of AWVs since their introduction, but modest use overall,10-12 with 7% of Medicare beneficiaries receiving an AWV in 2011, increasing to 16% in 2014.13 More than 90% of AWVs nationally in 2014 were performed by a primary care physician (PCP).13 Although utilization of the service is increasing, the benefits of an AWV for improving patient outcomes and controlling healthcare costs continue to be debated.14-16 This lack of evidence regarding the possible impact of AWVs on important outcomes restrains providers and policy makers from optimally using the service.

In order to address these gaps, we examined the association of an AWV with healthcare costs, utilization, and measures of clinical quality among beneficiaries cared for by 2 PCP-led accountable care organizations (ACOs). We focused on beneficiaries cared for by PCPs in the ACOs formed in 2015 by Aledade, a national network of independent practices.17-21 Aledade has prioritized AWVs to improve quality and focus a primary care relationship on preventing adverse health events. It has supported its partners in performing AWVs by identifying high-risk beneficiaries, building user-friendly technology to schedule AWVs with these beneficiaries, providing face-to-face practice transformation support to optimize workflows, implementing templates in the electronic health record (EHR), providing data tools to support performance monitoring, and facilitating best practice sharing across a network of doctors.22 Evaluating the AWV in the context of a highly motivated and supported physician network leads to greater understanding of how AWVs can contribute to improving healthcare quality and reducing costs under optimal conditions.


Study Design and Sample

Primary data source. We used insurance claims from the CMS Claim and Claim Line Feed23 as our primary data source to assess the association of an AWV with cost and utilization. These data include services provided under Medicare parts A and B for patients assigned to 2 PCP-led ACOs. Data from these specific ACOs were used because they were the first ACOs that Aledade partnered with that had complete follow-up data. The observation period was from January 1, 2014, to December 31, 2016. Permission to use data for the study was granted through the Medicare Shared Savings Program data use agreement, and institutional review board (IRB) approval was granted by Hummingbird IRB (IRB #2017-278).

Intervention and control groups in primary data source. Among patients attributed to ACOs 1 and 2 at the start of 2015, we identified intervention beneficiaries who had an AWV in 2015 (Current Procedural Terminology codes G0402, G0438, or G0439). We excluded beneficiaries who had missing data, died during the study period, or had received an AWV in 2014 (we wanted to focus on the effects from a first-time AWV, assuming that patients who did not receive an AWV in 2014 did not receive one in 2011-2013). To identify a control group, we matched the intervention beneficiaries to beneficiaries who did not have an AWV in 2015 and who met the same inclusion and exclusion criteria as the intervention group (for additional information about the matching process, see the eAppendix [available at]). Control patients were assigned the AWV month of the intervention patient to whom they matched. We removed the month that the AWV was done (and thus the cost of the AWV itself, because we wanted to focus on the subsequent AWV impact), which allowed us to define the 11 calendar months before the AWV as the pre-AWV period and the 11 calendar months after the AWV as the post-AWV period. Because Medicare only bills AWVs once every 12 months per patient, the intervention group by definition did not have any AWVs in the post-AWV period. After matching, we excluded all control patients who had an AWV in the post-AWV period, as well as all patients with outlier spend (to reduce skewness). For a visual definition of this cohort, see eAppendix Figure 1.

Secondary data source for quality of care. To assess impact of an AWV on clinical quality, we used data on clinical quality measures reported to CMS as part of the ACO program.24 These data are reported to CMS for a different randomly selected sample of beneficiaries for each measure. Beneficiaries met exclusion/inclusion criteria as defined by each measure definition in accordance with Medicare specifications.25 Each sample was then divided into 2 groups: The control group included beneficiaries who did not receive a first-time AWV in 2015 (but could have received one in 2014 or 2016) and the intervention group included beneficiaries who did receive a first-time AWV in 2015 (including the “Welcome to Medicare” visit). We included all beneficiaries who were reported on by CMS regardless of whether they were in our final primary analytic sample.
Outcome Measures

The primary outcome was the differential change in healthcare cost (in the post-AWV period compared with the pre-AWV period) between those who did and did not receive an AWV. Costs in the month that the AWV was performed, including the approximately $175 cost of the AWV itself, were excluded. Cost was evaluated in separate analyses as total Medicare cost (all parts A and B Medicare spending) and category-specific costs within Part A (hospital acute care; hospital outpatient; hospital outpatient non–emergency department [ED]; skilled nursing facility, home health, other outpatient facility spending) and Part B (provider/supplier, durable medical equipment). The secondary outcomes included counts of ED visits and hospitalizations in the pre- and post-AWV periods. The outcomes for the quality measures analysis were 16 clinical quality measures with definitions specified by Medicare in 3 domains: preventive health, clinical care for at-risk populations, and care coordination.25

Statistical Analysis

Similar to other observational studies using real-world evidence, this study’s approach accounted for the likelihood that individuals who received AWVs differed from other patients in substantial ways that might affect the outcomes measured. In particular, issues may include that practices did not reach out to all patients with equal likelihood, practices succeeded in reaching patients at different rates, and patients who were willing and able to come in for a primary care visit differed from patients who were not. We used several analytic methods to explore the impact of such selection bias and account for it. We employed propensity score matching to identify control subjects who were similar to patients who received AWVs in key respects, including level of engagement with primary care in the pre-AWV period. We estimated the impact of AWVs by specifying a series of difference-in-differences (DID) regression models.

Specifically, for the primary data source, we checked the parallel trends assumptions and used a DID study design to assess changes in subsequent 11-month healthcare costs, ED visits, and hospitalizations between beneficiaries who did and did not receive an AWV in the index month. We used a mixed-effects negative binomial model for total cost and mixed-effects zero-inflated negative binomial (ZINB) models for category-specific costs and utilization. For the secondary data source, we used mixed-effects logistic regression models to assess the association between receiving an AWV and quality measures. In order to account for multiple comparisons, the Hochberg sequential procedure was used.26,27 For additional details on the modeling approach, see the eAppendix.

We also conducted a series of sensitivity analyses to evaluate the robustness of our results. First, we evaluated whether the intervention effect differed among those patients who received outreach using the Aledade app in 2015 versus those patients who did not receive outreach in 2015. Second, we used coarsened exact matching instead of propensity score matching to identify the comparison group. Third, we repeated our primary analysis (including the matching) only among beneficiaries who were continuously attributed throughout the entire pre- and post-AWV period. Fourth, we excluded intervention patients who matched to controls who had an AWV in the post-AWV period. Finally, we evaluated whether the AWV associations were different for “early” (January 2015–July 2015) versus “late” AWVs (August 2015–December 2015). (See eAppendix for additional details.)

All statistical analyses were conducted using R version 3.3.3 (R Foundation for Statistical Computing; Vienna, Austria). Propensity score and coarsened exact matching were conducted using the MatchIt package, mixed-effects logistic models were fit using the lme4 package, and the mixed-effects negative binomial and ZINB models were fit using the glmmTMB package.28-30 Graphics and plots were generated using the ggplot2 package.31


Baseline Characteristics by AWV Status

The primary analysis sample of matched intervention and control beneficiaries included 8917 beneficiaries, of whom 4789 (54%) received a first-time AWV in 2015 (Table 223). Differences in clinical and sociodemographic characteristics between the matched intervention and control groups were small. (For characteristics of the population of beneficiaries who entered the matching process, see the eAppendix.) The only covariate that was significantly different between those who did and did not receive an AWV in the final analytic sample was the specific ACO of the patient, and thus it was adjusted for in final regression models.
Association of AWV With Healthcare Costs and Utilization

In the pre-AWV period, average trends in healthcare cost were similar for the intervention and control groups (eAppendix Figure 2). In 2015, a first-time AWV was associated with a 5.7% (95% CI, 0.3%-11.4%) reduction in total healthcare costs in the post-AWV period (excluding the cost of the AWV itself). This association translated to a $38 (95% CI, $9-$67) per-member-per-month (PMPM) reduction over 11 months of follow-up, or approximately $418 per beneficiary (Figure 123). The association between a first-time AWV and reduced costs was stronger among beneficiaries in the top hierarchical condition category (HCC) risk quartile. In this population, the adjusted differential change in total healthcare cost between the intervention and control groups was 6.3%, a PMPM decrease of $81 (95% CI, $12-$150) over 11 months of follow-up (Figure 123).

Analysis of category-specific costs suggested that the primary drivers of this impact were reductions in hospital acute care costs (incidence rate ratio [IRR], 0.88; 95% CI, 0.80-0.97) and hospital outpatient non-ED costs (IRR, 0.93; 95% CI, 0.89-0.97). Table 323 reports the differential change in each category-specific cost between the intervention and control groups. With respect to healthcare utilization, first-time AWVs were not associated with a statistically significant change in the total number of ED visits (IRR, 0.97; 95% CI, 0.83-1.15) or hospitalizations (IRR, 0.95; 95% CI, 0.78-1.11).

Results of the sensitivity analyses were consistent with the main findings. Of note, the AWV association from the main analysis was not systematically different among beneficiaries who were invited by phone or email to schedule an AWV by the PCP practice (defined as receiving outreach) and beneficiaries who were not invited. In addition, using an alternative matching method to identify the comparison group and limiting the sample to those who were continuously attributed were consistent with a robust association between receipt of an AWV and reduced healthcare costs. Additionally, results were consistent with the main analysis when intervention patients who matched to controls who had an AWV in the post-AWV period were excluded. Finally, the effect of the AWV was not different between AWVs conducted in the early versus late part of the calendar year (See eAppendix for additional details.)

Association of AWV With Clinical Quality

Of 16 quality measures evaluated (Figure 224), a first-time AWV in 2015 was significantly associated with greater performance on 7 measures in adjusted analyses (all P <.01): fall risk screening (94% vs 15%), pneumococcal vaccination (86% vs 69%), tobacco screening and cessation (91% vs 77%), depression screening and follow-up planning (87% vs 18%), colorectal cancer screening (69% vs 60%), breast cancer screening (81% vs 66%), and controlled glycated hemoglobin (A1C) (77% vs 65%). AWVs were not statistically significantly associated with diabetes eye exams, use of aspirin, controlled hypertension, BMI screening, medication documentation, blood pressure control, influenza vaccination, diabetes therapy, or heart failure therapy. We speculate that AWVs had a lesser impact on these quality measures because they may already be prioritized in general primary care settings.


In this cohort of Medicare beneficiaries, first-time AWVs were associated with a significant improvement in use of preventive care and a reduction in total healthcare costs compared with matched controls. Rates of screening for fall risk and for clinical depression with follow-up plan, which are not typical components of a traditional evaluation and management visit but are components of an AWV and included in visit templates, were more than 70 percentage points higher among beneficiaries who received an AWV. These beneficiaries were also more likely to experience improved A1C control and to receive other key preventive services, including breast and colorectal cancer screening and tobacco use screening with cessation intervention. Changes in total healthcare costs were greatest among beneficiaries in the highest quartile of HCC and were driven by reductions in hospital acute care and hospital outpatient non-ED spending. Although there was a trend toward reduced hospital utilization, it was not as pronounced as that for hospital costs and did not achieve statistical significance. This would imply that the hospitalizations that did occur tended to have lower severity. These findings suggest that an AWV can be a helpful tool for improving care quality and containing costs within a primary care setting that prioritizes patient engagement, utilization management, and care coordination.

To our knowledge, this study is the first in the peer-reviewed literature to estimate the association of an AWV with measures of healthcare quality, costs, and utilization within the same study. Use of the AWV has been rising slowly since its introduction as a Medicare-reimbursable service in 2011, but there has been a paucity of evidence to guide AWV implementation into routine clinical practice in the primary care setting and to establish the potential value of this type of visit for ACOs and the Medicare program. Our findings add to a growing body of literature suggesting that the AWV can substantially improve rates of preventive services32-36 (which may be directly related to the administration of screening tools) while providing new evidence of substantial near-term effects on total cost of care. By further delineating that the association of the AWV with healthcare costs may be most pronounced among highest-risk patients, our findings lend support to a strategy of population risk segmentation for prioritization of AWV outreach efforts to maximize savings benefits. Furthermore, although there is general consensus that strong primary care is essential to containing healthcare costs,37-39 recent payment and delivery system innovations that intend to enhance primary care services beyond usual care have shown mixed results.40,41 Our study results suggest that a primary care service under a system that provides the right incentives for all may contribute to cost reductions.

The mechanisms explaining the cost reductions of AWVs are not well known, but we speculate that numerous aspects of the AWV might explain its benefits. A successful AWV means that practices are not merely “checking the box.” These screenings can be used to provide updates on medical history and self-reported data as an opportunity to step back from typical acute complaints and meaningfully engage in personalized conversations about risk factors, preventive needs, and a patient’s long-term health goals. This attention on wellness may improve clinical quality, including the delivery of general clinical preventive services and secondary prevention among patients with chronic conditions. An optimal AWV can include medication review and regimen optimization, identification of uncoordinated use of specialty care, and discussion of social or environmental barriers to self-care that may benefit from enhanced care coordination. This up-front investment in preventive care and care coordination may avert subsequent spending. By devoting time to explore the patient’s overall health status, risks, and values, the AWV may enrich the patient–provider relationship, improve patient engagement, and reinforce the core primary care tenets.

Several factors should be considered in the interpretation of this study. First, the results should be interpreted in the context of the specific setting studied. AWVs were a key strategy of the ACOs, and rates of AWVs (54%) were substantially higher than national averages. It is possible that the results of this study would be replicable only in a setting willing to undergo workflow optimization to accommodate a high-value visit. In addition, the cohort studied did not include beneficiaries at the end of life and those who may be less able to receive an AWV due to being homebound, institutionalized, hospitalized, or enrolled in hospice. Thus, the findings cannot be generalized to all Medicare beneficiaries. Second, as with all nonrandomized study designs, a possibility exists for selection bias and residual confounding due to unmeasured differences. However, we used propensity score matching to account for observable differences and a DID design to account for unobservable time-dependent changes in spend, as was done recently in a CMS evaluation of chronic care management.42 A key matching variable for cases and controls was the number of baseline primary care visits, to account for the level of engagement with a PCP and the ability to come in for a visit. Furthermore, the association between receipt of AWV and healthcare cost reductions did not vary by whether or not patients received outreach, providing little evidence to suggest that the intervention effect was driven by differential outreach to patients who were predisposed to favorable cost trajectories. Finally, we were not able to ascertain whether patients had an AWV prior to 2014, although data suggest that fewer than 16% of Medicare patients were receiving them in 2013.13

These findings point to several priorities for payers and providers to consider. Because Medicare reimburses $175 for an AWV per member per year (PMPY), it is important to note that the true cost reduction may be smaller than the $456 PMPY effect estimate that we reported. However, given that the effect estimate is more than 2-fold the cost of the AWV itself, these data suggest that the additional expenditure on primary care can be worth the costs, particularly for a higher-risk population. It is also worth acknowledging that AWVs are one way to improve coding accuracy, and “upcoding” can be balanced by the legislatively afforded renormalization factor to account for risk inflation.43

Future Implications

Future research can help guide policy with respect to whether AWVs should be billable only by the patient’s PCP, who may be in the best position to comprehensively assess patient risk factors and preventive care needs. Our findings show that an AWV may achieve meaningful improvements in cost and quality, lending support that policy makers should further facilitate the adoption of high-quality AWVs by PCPs. Given that underserved populations are less likely to adopt AWVs,44,45 policies should be explored to expand access to AWVs for this important subgroup of patients.

Additional research is still needed to further understand answers to several key questions. Given the recent introduction of the AWV, it will be important to understand whether the impact of an AWV changes over a longer time horizon. Future research should also attempt to differentiate effects of specific AWV components on outcomes. Additional outcomes of interest beyond the scope of this initial study include effects on patient satisfaction, health behaviors, self-management of chronic conditions, and care continuity.


Among beneficiaries cared for by PCPs affiliated with 2 ACOs, the AWV was associated with delivery of greater preventive care and lower total healthcare costs, particularly for those among the highest quartile of HCC risk. The AWV may therefore be an important service for achieving the triple aim of “improving the experience of care, improving the health of populations, and reducing per capita costs of health care.”46 Future research should focus on replicating these results among other populations, given that the current study used data from 2 specific ACOs. It will be important for future studies to test the efficacy of AWVs in different geographical areas and in ACOs with different characteristics in order to provide robust evidence of AWV impact. Furthermore, identifying tactics to further facilitate adoption and optimize the effectiveness of the AWV in primary care practice will be important avenues for future research.


Adam L. Beckman, BS, and Adan Z. Becerra, PhD, contributed equally to this work and are listed as co–first authors.

The authors first acknowledge the 114 independent primary care providers from Delaware ACO and Primary Care ACO who led the ACO efforts through 2016. They also appreciate editorial inputs from Travis Broome, MPH, vice president of policy at Aledade.
Print | AJMC Printing...