BACKGROUND

Advance care planning (ACP), which includes discussing end-of-life healthcare preferences with a clinician, has been linked with goal-concordant care and improved patient well-being at the end of life.1,2 In 2016, Medicare introduced outpatient and inpatient billing codes for ACP, “including the explanation and discussion of advance directives such as standard forms (with completion of such forms, when performed), by the physician or other qualified health care professional.”3 Medicare reimburses clinicians $80–$86 for the first 16–30 min of these face-to-face discussions and $75 for each 30 min thereafter,3,4 with the goal of providing financial incentives and allowing dedicated time for physicians to engage in ACP.5,6 Clinicians can bill for ACP multiple times for the same beneficiary. However, as of 2017, clinicians billed these ACP codes infrequently and for a small fraction of eligible patients, even among seriously ill individuals who may most benefit from these conversations.7,8

Adoption and use of ACP and ACP billing codes are likely linked to individual clinician- and patient-level factors.9,10,11,12,13,14,15 Conceptually, successful ACP billing requires clinicians to be comfortable doing ACP, know about the Medicare ACP benefit and how it applies to their patients, and find time to introduce it during visits. Patients, in turn, must be receptive to having the conversation with their clinicians. In kind, studies have found lower ACP uptake among individuals with fewer economic resources9,12 or from minoritized racial/ethnic groups (Black, Hispanic, Asian10,16,17), and among clinicians exhibiting hesitancy in initiating ACP.13,14,15

Organizational factors, such as those found at the physician practice level, may also play an important role in ACP billing uptake, as they do in other aspects of health care utilization.18 Conceptually, practices can encompass multiple levels of organization, including single clinicians, a small number of physicians at a single location, or large practices with several hundred physicians across multiple sites. Network analyses have demonstrated organizational influences on adoption of new clinical approaches through informal sharing of these approaches, workflows, and culture, and formal sharing of policies and billing systems.19,20,21 The role of practices in ACP billing utilization specifically is not well understood.22 Qualitative studies indicate organizational factors such as administrative, documentation, and leadership support for ACP;22,23,24,25 monitoring and benchmarking;23 and training in conducting and documenting ACP13,23,25 may influence ACP billing. To our knowledge, a large-scale quantitative analysis of organizational-level characteristics has not been conducted.

Building on this prior work, we used Medicare claims data to analyze the understudied relationship between practice-level factors and ACP billing in outpatient settings. We chose factors that may influence billing uptake based on literature review and an adaptation of Reschovsky’s conceptual model of multilevel factors contributing to clinical approaches.18 For example, we considered number of physicians and share of primary care physicians (PCPs) in the practices given evidence linking practice size to care quality26 and that PCPs most commonly offer ACP.27 We also examined the size and sociodemographic composition of their attributed Medicare beneficiary population—for example race/ethnicity10,16,17 and socioeconomic status9,12—since structural and individual biases related to the practice’s patient population may influence practice-wide ACP use.28 Improved understanding of practice-level barriers to ACP billing may help identify types of practices to target to increase uptake of ACP billing.

METHODS

Data and Study Population

This retrospective cohort study used 2016–2018 claims data from a 20% random sample of fee-for-service, continuously enrolled Medicare beneficiaries aged 65 years and older. Information was used from the outpatient, carrier, and Medicare Data on Provider Practice and Specialty (MD-PPAS) files. Consistent with prior studies of practice-level characteristics based on Medicare claims,28,29,30,31,32,33 we identified practices based on taxpayer identification numbers (TINs) and limited our analysis to practices with at least 10 attributed beneficiaries.28,29,30,31 Though TINs are well-established proxies for practices and are reliably linked to national claims, we acknowledge they do not reflect a consistent organizational level (a single TIN may reflect multiple practices within a large health system or an individual practice.).

Clinician and Beneficiary Attribution

We identified clinicians by national provider identifier (NPI) number and attributed each to a practice using the MD-PPAS file based on the TIN that accounted for the most months of billing.34 Consistent with Medicare Shared Savings Program rules for beneficiary assignment,31 we attributed beneficiaries to the practice that accounted for the largest proportion of their qualifying outpatient visits. Qualifying office visits included annual wellness visits, problem-based evaluation and management visits, chronic care management, and transitional care management visits during the study period.28 We also attributed beneficiaries to the individual clinician that accounted for the greatest share of qualifying outpatient visits during the study period, regardless of specialty.28(See online Appendices I and II for additional details.)

Outcomes

We identified outpatient ACP services using CPT codes 99497 and 99498. We analyzed two measures of ACP use. First, we measured whether practices had billed an ACP code at least once for any attributed beneficiary (“ACP billing”) during the study period. Second, we measured the rates of use, calculated as the number of practice-attributed beneficiaries with at least 1 ACP service billed divided by the number of practice-attributed beneficiaries, multiplied by 1,000 for ease of interpretation (“ACP use rate”). For ACP use rate, beneficiaries count one time, regardless of their total number of associated CPT codes.

Practice Characteristics

We identified practice characteristic covariates among beneficiaries and clinicians attributed to the practice. Practice characteristics based on beneficiaries included the following: number of beneficiaries attributed to the practice, by quintiles; percentage of beneficiaries from minoritized racial/ethnic groups (non-Hispanic Black, Hispanic, Asian/Pacific Islander, other/unknown race/ethnicity), Medicare-Medicaid dual enrollment, female sex (grouped as <25%, 25 to <50%, 50 to <75%, and 75–100%); and average age. We grouped individuals from minoritized racial/ethnic groups together due to relatively small numbers of beneficiaries in each group and because these subgroups consistently complete ACP at lower rates.10,16,17 Practice characteristics based on clinicians included the following: practice size (average number of clinicians attributed to the practice seeing beneficiaries (grouped as 1, 2–10, and >10) and specialty mix (percentage of clinicians in primary care, grouped as <25%, 25 to <50%, 50 to <75%, and 75–100%). Primary care specialties were internal medicine, family medicine, geriatrics, and general practice (Appendix III). In Medicare claims data, nurse practitioners do not have specified specialties; we kept them as a separate category of “other specialty.” We used categorical measures when relevant to aid interpretation.

Statistical Analysis

The unit of analysis was the practice. We calculated descriptive statistics for the overall sample of practices and for practices that did and did not bill for ACP, using chi-square and t tests to identify significant differences between these practices. To model the association between practice characteristics and ACP use, we used multivariable logistic regression (for billing) and linear regression (for ACP use rate), including as covariates the practice characteristics described above. We conducted several secondary analyses. First, for all 463,943 ACP services billed in the study period, we identified the proportion provided by a beneficiary’s attributed or another clinician, and the specialty of the ACP-billing clinician. Second, we described clinician patterns of ACP billing within practices with 2 or more clinicians. Finally, we calculated the percentage of ACP-billing clinicians within a practice weighted by practice size.

RESULTS

There were 54,788 practices with at least 10 attributed Medicare beneficiaries during 2016–2018. Of these, 8,226 (15%) billed for ACP and 46,562 (85%) did not (Table 1). In bivariate analysis, on average, ACP-billing practices had a higher percentage of PCPs (67% vs 35% with 50% or more PCPs, p<0.0001). ACP-billing practices also had more attributed beneficiaries overall (82% vs 55% had 22 or more beneficiaries, p<0.0001). While these practices also had statistically significantly higher shares of beneficiaries who were female, not dual-enrolled, and from minoritized racial/ethnic groups, the difference in percentages for these groups was much smaller (<5% for most comparisons) (p values for all comparisons were <0.0001, except race/ethnicity (p=0.0024)). ACP-billing practices and non-billing practices had similar patient and clinician-level characteristics, except average number of beneficiaries (Appendix IV).

Table 1 Characteristics for 54,788 Practices, 20% Sample of Fee-For-Service Medicare Beneficiaries, 2016–2018
Full size table

In multivariable logistic regression models, all examined practice characteristics were significantly associated with ACP billing (Table 2). Compared to practices with no PCPs, those with at least one PCP had greater odds of ACP billing (75–100% PCPs AOR: 10.01, 95%CI: 8.81–11.34). Odds of ACP billing also increased with number of attributed beneficiaries (66+ beneficiaries (compared to 10–14) AOR: 4.55, 95%CI: 4.08–5.08). Greater average age of practice beneficiaries; percentage of practice beneficiaries from minoritized racial/ethnic groups, dual-enrolled, and female; and practice size were modestly associated with ACP billing.

Table 2 Adjusted Odds Ratios in 53,926 Practices Billing for ACP
Full size table

When examining ACP use rates, practices with more beneficiaries and a higher proportion of PCPs had higher ACP use rates than practices with 10–14 beneficiaries or 0 PCPs (66+ beneficiaries mean difference (MD): 39 beneficiaries with an ACP billed per 1,000 eligible beneficiaries, 95%CI: 34–43; 75–100% PCPs MD: 50 beneficiaries per 1,000 eligible, 95%CI: 47–54) (Table 3). Practices with a higher percentage of beneficiaries who were female, dual-enrolled, or from minoritized racial/ethnic groups also had higher ACP use rates compared to practices with <25% of beneficiaries in each of these categories (75–100%: female MD: 17 beneficiaries per 1,000 eligible, 95%CI: 8–25;dual-enrolled MD: 23 beneficiaries per 1,000 eligible, 95%CI: 17–30; minoritized racial/ethnic groups MD: 12 beneficiaries per 1,000 eligible, 95%CI: 7–17). In contrast, practices with fewer clinicians had higher ACP use rates than practices with >10 clinicians (>10 clinicians: MD: −27 beneficiaries per 1,000 eligible, 95%CI: ((−32)–(−22)).

Table 3 Mean Differences in 53,926 Practices Billing for ACP
Full size table

The majority (82%) of ACP services during the study period were billed by PCPs, followed by nurse practitioners (13%). More than half (55%) of ACP services were billed by the beneficiary’s attributed clinician (Table 4). Among all ACP services billed by beneficiaries’ attributed clinicians, 95% were billed by a PCP. ACP services occurring with a non-attributed clinician were billed by a PCP 66% of the time.

Table 4 Distribution of 463,943 ACP Billed Services by Clinician Role/Specialty
Full size table

Among multi-clinician practices billing for ACP services, an average of 33.5% of clinicians within these practices billed for ACP at least once. This weighted proportion of clinician ACP use varied by practice size (Appendices V and VI). In smaller practices, a higher percentage of clinicians within practices billed for ACP; for example, in practices with an average of 4–5 clinicians, 50–100% of clinicians billed for ACP, while in larger practices with an average of 188 clinicians, less than 25% of practice clinicians billed for ACP.

DISCUSSION

Only 15% of practices billed Medicare for ACP during 2016–2018, more often those with more PCPs, with more than 10 clinicians, serving a greater number of Medicare beneficiaries, and serving a higher proportion of older, female, minoritized racial/ethnic, and dual-enrolled beneficiaries. PCPs billed for 82% and nurse practitioners 13% of all ACP services. On average, one-third (34%) of clinicians in multi-clinician, ACP-billing practices billed for ACP, with higher percentages of clinicians billing for ACP in smaller practices.

Our findings on low uptake are consistent with prior analyses of 2016–2017 data that show only 1–6% of eligible beneficiaries receive ACP.7,8,27 Beneficiary, clinician, and practice factors may contribute to the limited uptake at the practice level. Beneficiaries may not prioritize ACP or may be hesitant to discuss it with their clinician.35 Clinicians may feel unprepared and therefore reluctant to initiate ACP with patients,13,14 particularly those with limited English proficiency.15 Alternatively, clinicians—and specialist physicians in particular—may forego ACP billing, even if they discuss ACP with their patients, because these codes are reimbursed at much lower rates than other services they may provide such as procedures.23 Practices may lack the resources needed to bill for ACP,14 such as leadership support for ACP billing, dedicated billing champions or personnel, or electronic health records that support documenting and billing for ACP.13,22,23,24

Larger practices and practices with a greater percentage of PCPs had higher odds of billing for ACP, although ACP use rates were lower in the largest practices. Presence of PCPs was the strongest predictor of ACP billing, and 95% of ACP claims were billed by a PCP or a nurse practitioner (most of whom are in primary care36), whether or not that clinician provided the plurality of the beneficiary’s visits. This finding is consistent with analyses of 2016 Medicare data indicating 93% of ACP claims were billed by internal medicine, family medicine, or geriatrics physicians.37 The ACP reimbursement policy provides financial incentives for having appropriately timed conversations about end-of-life care with a clinician, regardless of care setting. Our findings demonstrate that, as measured by outpatient ACP billing, these discussions happen largely with PCPs, perhaps given this specialty’s focus on longitudinal relationships and holistic care. Primary care–focused practices may be more incentivized to bill for ACP than specialty practices, which may focus billing efforts on higher-paying procedures than ACP.23 Our findings that larger practices are more likely to perform ACP billing yet have lower use rates suggest an opportunity to increase ACP billing uptake in larger practices where a small number of PCPs already bill for ACP; it may be easier to maximize information sharing and increase uptake of clinical approaches within groups where some physicians have already adopted it.19,20,21 Additionally, our findings that relatively few physicians in other specialties billed for ACP suggest an opportunity to expand ACP by targeting physicians and practices specializing in cardiology or oncology, for example, that care for seriously ill beneficiaries, and in many cases serve as those patients’ primary source of care.38

We also observed that within smaller practices, higher percentages of clinicians billed for ACP. Information sharing, workflow, structure, and/or billing practices may also lead to more similar approaches among clinicians in smaller practices.39,40 For example, smaller practices may be better suited to implement new approaches practice-wide—such as ACP billing—because it is easier to ensure all practice clinicians use ACP billing codes.41

Practice beneficiary factors such as number of beneficiaries and demographics were moderately associated with ACP billing. Practices with more attributed beneficiaries had higher odds of ACP billing, perhaps due to having more experience providing care specific to that population. Greater billing in practices with older, more female beneficiaries is consistent with prior research on ACP and on overall healthcare utilization for these groups. Older adults are more likely to see a physician for serious illness and to receive ACP services.42 Women are also more likely to utilize ACP services,42,43 possibly because they view ACP more positively than men.44 Additionally, practices with a higher percentage of female beneficiaries may have more female clinicians, and increased ACP billing in these practices may be an extension of increased person-centered care more common in female patient-clinician dyads.45

We observed slightly increased ACP billing in practices serving a higher percentage of dual-enrolled beneficiaries and beneficiaries from minoritized racial/ethnic groups. This finding runs contrary to prior research indicating ACP is less common among persons with fewer economic resources and from minoritized racial/ethnic groups,9,10,11,12,37,46 and that physicians in practices with a large proportion of persons from minoritized racial/ethnicgroups—which also tend to serve a large proportion of Medicaid beneficiaries47,48—spend less time with patients, another potential barrier to the time-intensive conversations required for ACP billing.47Dual-enrolled beneficiaries are typically sicker and have more disabilities,49 possibly triggering greater ACP use among practices serving these beneficiaries. Black and Hispanic/Latino persons are less likely to seek preventive care or identify a usual place of care,50,51 and therefore may have received care at ACP-billing practices, but from clinicians with whom they lacked a longitudinal therapeutic relationship conducive to ACP. Although significant, effect sizes for these groups were relatively small, and should be interpreted with care.

Limitations

This study has limitations. First, while TINs are widely used to define practices, they do not allow us to differentiate between different organizational levels that TINs capture.28,29,30,31,32 We addressed this in part by accounting for practice size in our models. Future research should examine how specific organizational structures affect ACP billing. Second, our analysis focused on uptake of Medicare’s ACP billing codes from 2016 to 2018, and we therefore do not observe ACP conversations outside of this period, nor ACP conversations for which clinicians did not bill. Additionally, we cannot evaluate the longitudinal and complex processes involved in ACP; nor can we ascertain the quality or content of ACP discussions or whether discussions resulted in formal documentation of patient preferences, such as an advance directive. However, ACP billing represents a standardized proxy for ACP in a national sample; by measuring ACP billing rates, we are able to assess practices’ realized or missed opportunities to take advantage of Medicare’s ACP initiative.52 Finally, our approach has potential biases. While our models account for a number of practice and beneficiary characteristics, we acknowledge the role of unobserved confounders. Our study is not generalizable to Medicare Advantage patients, those who do not receive any outpatient services, or to physicians and practices who do not provide the outpatient services we analyzed. However, through our attribution approach, 98% of all beneficiaries were assigned to a practice and 95% to a clinician. Of all ACPs billed during the study period, 96% are captured in this analysis (the remainder were in practices with <10 beneficiaries).

Policy Implications

Our analyses suggest Medicare’s ACP billing codes are used by a minority of practices, more commonly among larger, primary care–focused practices, and that only a small proportion of clinicians in larger practices billed for ACP. Policymakers and clinicians might consider multiple approaches to increase ACP billing. Clinician training on ACP conversations may help—not only for PCPs but also for specialists such as cardiologists and oncologists seeing seriously ill patients for whom ACP is particularly appropriate. Efforts to increase ACP billing should also target smaller practices and those with non-PCPs where ACP billing is less common. Increasing the percentage of clinicians billing for ACP in practices where a small number of clinicians already bill for ACP could be an effective and relatively easy way to increase ACP uptake. As such, in larger ACP-billing practices, supporting intra-practice information sharing or education could allow the small number of ACP-billing clinicians to train other clinicians to bill for ACP. Finally, regardless of size, practices need education, technical support, and incentives to introduce ACP billing into existing processes and workflow.

Conclusion

Our analysis of organizational-level uptake of ACP billing finds that several years after policy to reimburse for ACP with their Medicare beneficiaries was introduced, relatively few practices bill for ACP. ACP billing is more likely in large practices but is performed by a relatively small number of largely PCPs. Our findings provide new insights into how Medicare ACP policy is being used and identifies opportunities to improve ACP billing by targeting practices not billing for ACP. Efforts to increase ACP billing uptake by more clinicians in large practices could take advantage of shared workflow, billing, and organizational approaches.