• Center on Health Equity & Access
  • Clinical
  • Health Care Cost
  • Health Care Delivery
  • Insurance
  • Policy
  • Technology
  • Value-Based Care

Calibrating Medicaid Payment to Need for Long-term Services and Supports

Publication
Article
The American Journal of Managed CareDecember 2020
Volume 26
Issue 12

Incorporating data from functional status assessments, we developed a Medicaid payment model for long-term services and supports in a community-dwelling population of older adults.

ABSTRACT

Objectives: The Massachusetts Medicaid and Children’s Health Insurance Program, MassHealth, offers comprehensive Senior Care Options (SCO) plans to its Medicare-eligible members 65 years and older. Historically, MassHealth has paid a fixed per-person capitation rate for any “nursing home–certifiable” SCO member despite considerable heterogeneity of need. Our objective was to develop a model to predict long-term services and supports (LTSS) costs for community-dwelling SCO members.

Study Design: Concurrent predictive modeling.

Methods: We studied nursing home–certifiable SCO members who were enrolled for at least 183 days during 2016-2017 and used linear models to predict annual cost of community-based LTSS from demographic, medical, social determinants of health, and functional characteristics. We evaluated model performance using predictive performance (R2) and predictive ratios (observed costs divided by predicted costs) for various vulnerable subgroups.

Results: The modeling population included 35,259 enrollees. Mean (SD) annualized LTSS cost was $14,071 ($13,174). Functional status (ie, activities of daily living [ADLs] and instrumental ADLs) accounted for most of the variability in community LTSS cost (R2 = 18.4%) explainable by available variables. The Massachusetts SCO (MA-SCO) model (R2 = 21.6%) predicts accurately for several high-cost, vulnerable subgroups. Compared with fixed per-member capitation payments for all, the MA-SCO model reduces, for example, the payment to one plan by 28% and increases that to another by 35%.

Conclusions: Predictive models using administrative data and functional status information can appropriately allocate payments for subgroups of members with LTSS needs that differ substantially from average. Calibrating payment to need mitigates incentives for population skimming and promotes the sustainability of mission-oriented organizations.

Am J Manag Care. 2020;26(12)e388-e394. https://doi.org/10.37765/ajmc.2020.88545

_____

Takeaway Points

Predictive models using administrative data and functional status information can appropriately allocate payments for subgroups of older adults with long-term services and supports needs that differ substantially from average.

  • By applying risk adjustment, payments to providers and organizations are better aligned with the underlying needs of their members.
  • In the absence of risk adjustment, providers and organizations with more challenging members do not receive adequate payment to meet the needs of their members.
  • To accurately predict the need for long-term services and supports, functional status information that is not available in traditional administrative data sources was required.

______

Among Medicare/Medicaid–eligible (dual-eligible) older adults, long-term services and supports (LTSS) generate most spending by Medicaid programs.1 The use of community-based, or outpatient, LTSS to increase time spent outside of institutional settings is aligned with patient-centered outcomes and preferences2,3 and could save costs for both Medicare and Medicaid.4,5 Whole-person–oriented senior care organizations responsible for the full continuum of an enrollee’s medical and social needs are paid to manage the care of older adults, including using LTSS to maintain individuals in the community.6,7

As of 2019, 24 states had implemented Medicaid managed LTSS programs.1 When based on an architecture of capitation rather than fee-for-service payments, managed LTSS offer senior care organizations the advantage of flexibility in their service delivery model while providing states the advantage of budget predictability. Since 2004, Massachusetts Medicaid (MassHealth) has administered a Senior Care Options (SCO) program of Dual Eligible Special Needs Plans for older adults (≥ 65 years).8 The Massachusetts SCO plans receive capitated payments and provide comprehensive services and community-based supports to enrollees organized under the leadership of an integrated primary care team. Compared with similar fee-for-service enrollees, Massachusetts SCO enrollees lived longer and spent less time in nursing facilities during the early years of the program.7

MassHealth has submitted a proposal to CMS for Duals Demonstration 2.0 seeking to strengthen the SCO program and more closely align it with the state’s dual-eligible program for younger adults.8,9 MassHealth proposed shifting to risk-adjusted capitation rates because without appropriately calibrating capitation payments to enrollee needs, heterogeneity in medical, functional, and social characteristics across plans may promote inefficient service use, population skimming, loss of community tenure, and financial pressure on mission-oriented SCO plans. Therefore, we sought to develop a model that predicts differences in enrollee needs for LTSS for community-dwelling members enrolled in Massachusetts SCO plans.

METHODS

Data

We used administrative records from the Medicaid Management Information System and MassHealth Data Warehouse, including member eligibility and enrollment files, encounter records, and functional assessments. These last were collected using the Minimum Data Set Home Care version 2.0 (MDS-HC 2.0) at enrollment in a SCO plan and annually thereafter, unless there was an intervening for-cause change in status assessment. The work was conducted using deidentified data as part of an operations project, and review by an institutional review board was not required.

Population Studied

The MassHealth SCO program includes members aged at least 65 years who are also enrolled in Medicare. MassHealth has historically used 6 rating categories and paid a uniform per-person capitation rate (without any further risk adjustment) for enrollees within each category. In this project, we focused on the rating category with the highest variability in observed Medicaid costs: community-dwelling SCO enrollees with a certifiable need for nursing home level of care. The modeling population included members enrolled in the SCO program for at least 183 days during 2016 and/or 2017, with at least 1 concurrent MDS-HC 2.0 assessment. We sequentially excluded enrollees as follows: without complete encounter records due to enrollment in a SCO plan with subcapitation arrangements (n = 15,928), without LTSS records (indicative of missing data; n = 9479), with less than 180 days of enrollment spent in the community (ie, outside of hospitals or nursing homes; n = 9011), who died with less than 6 months of SCO enrollment (n = 2678), with no MDS-HC 2.0 assessments available (n = 751), and receiving hospice care (n = 742).

Outcome

We predicted annualized costs for community LTSS: personal care attendants, adult foster care, adult day habilitation, home health, nursing, and therapy services (eAppendix Table 1 [eAppendix available at ajmc.com]).

Predictors

Except for functional status information, predictors were operationalized as in the risk-adjustment model used by MassHealth for paying managed care organizations.10 This includes indicators for 10 age/sex categories and the DxCG v4.2 concurrent Medicaid relative risk score (RRS) calculated from diagnoses grouped into condition categories with hierarchies.11 Additional individual-level social determinants of health (SDOH) variables identified from Medicaid claims and enrollment files include disability (disability-based Medicaid entitlement and clients qualifying for specialized services for mental health or intellectual disabilities) and housing problems (either unstable housing as indicated by ≥ 3 addresses within a year or homelessness by International Classification of Diseases, Tenth Revision [ICD-10] code). As described previously,10 we also included a contextual “neighborhood stress score” (NSS), calculated from 7 Census indicators of economic stress (eAppendix) and standardized to have a mean of 0 and SD of 1 in the full MassHealth managed care–eligible population (ie, those for whom MassHealth is the primary payer who were enrolled in primary care clinician–managed fee-for-service or managed care organizations). We considered NSS+ (equal to 0 or the NSS, whichever was larger) for inclusion in modeling because we thought that local stressors in economically disadvantaged neighborhoods might contribute to greater LTSS needs. However, we used the full NSS when evaluating how well each model predicted across the range of NSS scores.

Concurrent cognitive and functional status information was drawn from the earliest MDS-HC 2.0 assessment each year and, if none was available, from the most recent assessment in the preceding year. Cognitive impairment was operationalized using the Cognitive Performance Score (CPS; 0-6), where higher scores indicate greater impairment.12 Functional status information was grouped by the Resource Utilization Groups III Home Care (RUG-III-HC).13 We used the grouper logic13 to stratify members into 1 of 7 hierarchical, mutually exclusive categories (special rehabilitation [most resource-intensive], extensive services, special care, clinically complex, impaired cognition, behavior problems, and reduced physical function [least resource-intensive]). Because of inconsistencies between plans in the values for therapy minutes recorded on the MDS-HC 2.0, we calculated the average amount of physical, occupational, and speech therapy minutes received (a criterion for assignment to the special rehabilitation category) from encounters and claims per 7-day period (eAppendix Table 1). Nested within these 7 RUG-III-HC categories are finer stratifications (for a total of 23 groups) based on derived scores for activities of daily living (ADL: bed mobility, transfer, toilet use, and eating; range, 0-16, with each ADL scored 0-4, with higher scores indicating greater limitation), instrumental ADL (IADL: meal preparation, medication management, and phone use; range, 0-3), and cognitive performance (decision-making, understood short-term memory, and eating). Due to small numbers in several of the 23 categories, we used the 7-category grouping (operationalized as 6 categorical indicators because costs did not increase monotonically). However, for consistency with common practice, we retained the RUG-III-HC hierarchy when assigning each member to a unique category. We also examined ADLs and IADLs as independent predictors.

Analyses

We first described the study population, then we used weighted least squares regression to predict annualized community LTSS costs during each year of the study period. We used concurrent modeling to estimate costs from predictors measured during the same year (or the most recent MDS-HC 2.0 assessment).

We started with the basic framework developed for an SDOH model that MassHealth uses to make capitated payments (exclusive of LTSS costs) to managed care plans. To adapt it for predicting community LTSS costs for SCO enrollees, we added cognitive and functional status information from the MDS-HC 2.0. We fit separate models that included distinct domains of predictors: (1) medical risk and social risk (ie, the MassHealth SDOH model); (2) RUG-III-HC categories; (3) ADLs, ADLs-squared, and IADLs; (4) CPS and CPS-squared; (5) a full model with all terms from models 1 through 4; and (6) model 5, groomed for interpretability and parsimony. All models included indicators for 10 age/sex categories. Because the minimum ADL score in the SCO population eligible for nursing home care was 4 (a criterion for classification in this rating category), we rescaled it (and its squared value) to start at 0. Based on model results from the ADL/IADL only and the “full” model that indicated similar predicted LTSS costs for ADL scores of 4, 5, and 6, we further rescaled the ADL score: Scores of 4 through 6 were recoded to 0, and scores of 6 to 15 were recoded to 1 to 9 (no included member had a score of 16).

Model performance was evaluated in terms of predictive capacity (R2), using 10-fold cross-validation, and with observed to expected predictive ratios (PRs) to examine the extent to which the model accurately predicted community-based LTSS cost for subgroups of interest. A PR greater than 1 indicates that observed costs for the group exceed what the model predicts; a PR is less than 1 when the group costs less than the model predicts.

RESULTS

Patient Characteristics

The modeling population included 35,259 MassHealth SCO enrollees who contributed 32,941 person-years during 2016-2017. The mean (SD) age was 78.1 (7.5) years; 70% were women (Table 1). Most enrollees had moderate limitations in ADLs (mean [SD] = 7.8 [3.2]) and IADLs (mean [SD] = 1.7 [0.9]) and moderate cognitive impairment (mean [SD] CPS = 1.6 [1.3]), but the majority (61%) of enrollees were in the 2 lowest-intensity RUG-III-HC categories. The median (interquartile range) annual expenditure on LTSS was $11,744 ($2320-$21,498) and the mean (SD) was $14,071 ($13,174).

Modeling Results

The SDOH model, which includes only medical and social risk factors, explained little of the variation in LTSS costs (R2 = 3.5%) (eAppendix Table 2). The model including only RUG-III-HC categories performed modestly better (R2 = 7.0%). The CPS alone explained more of the variation in LTSS costs (R2 = 10.4%), and limitations in daily functioning (ADLs and IADLs) explained the most (R2 = 18.4%). A full model including all the potential predictors increased the R2 to 22.9%.

We “groomed” model 5, arriving at a first Massachusetts Senior Care Options (MA-SCO) model, to increase interpretability of model coefficients. The medical RRS and disability status variables were not included in the MA-SCO model because, after accounting for other predictors, they did not improve model performance. Despite not being a statistically significant predictor of LTSS costs after accounting for other variables, the CPS was retained in the MA-SCO model because of the conceptual relationship between increasing levels of cognitive impairment and need for community-based LTSS. Final grooming, leading to the MA-SCO payment model, was only slightly less predictive: Its R2 was 21.6 % (Table 2), and its predictive capacity did not drop in 10-fold cross-validation, suggesting that the model is not overfit.

The MA-SCO model was well calibrated within subgroups defined by enrollee sociodemographic characteristics, medical risk, and functional limitations (Table 3, Table 4, and eAppendix Table 3). Without risk adjustment, observed costs exceeded expected costs by 68% among members with ADL scores in the range of 12 to 15 and by 109% among those with moderate-to-severe cognitive impairment (CPS, 4-6) (Table 3). Members with housing problems (PR, 1.11) and those who lived in the top decile of neighborhood stress (PR, 1.04) also had above-average costs. In less complex members, costs were only 70% of the average among those with the lowest ADL scores (4-6). The MA-SCO model’s predicted costs for most of the subgroups were much closer to observed costs. Underpayment for Black non-Hispanic and Hispanic members was present, both with and without risk adjustment. However, almost half of the study population had missing or unknown race/ethnicity (Table 4).

To examine how well the SCO model identified and discriminated between members with the highest and lowest LTSS expenditures, we sorted the population into deciles of model-predicted risk. The average actual cost in the highest decile ($27,133) was more than 5 times the average cost in the lowest decile ($5162). PRs for the highest and lowest deciles of predicted LTSS cost (1.01 and 0.98, respectively) suggest that the model predicts well for both high- and low-cost members.

The implications of applying risk adjustment to a capitation payment for a specific health plan depends on its case mix, which differed among plans. Thus, we applied the MA-SCO model to members enrolled with MassHealth SCO plans in 2016-2017 to examine plan-level implications. We found substantial heterogeneity in case mix across SCO plans. Illustrating this heterogeneity, for 2 plans historically receiving the same per-member capitation payments in the absence of risk adjustment, the MA-SCO model would reduce the payment to one plan by 28% and increase the other’s by 35%.

DISCUSSION

In a community-dwelling, older adult population eligible for nursing home care, detailed functional status information on ADLs and IADLs was necessary to predict Medicaid expenditures for LTSS, whereas a summary medical morbidity score was not useful for predicting LTSS costs after accounting for functional status. Individual-level housing problems (ie, homelessness by ICD-10 code or ≥ 3 addresses) and neighborhood-level economic stress were both associated with higher LTSS expenditures. Without accounting for variation in social and functional characteristics, subgroups of members with LTSS needs that differ from average would be mispriced, and plans that disproportionately serve needier populations would receive capitated LTSS payments well below their costs.

Risk adjustment occupies a central role in the shift from fee-for-service to alternative payment models. Although population-based payment increases the cost consciousness of managed care entities and creates incentives for coordinating and managing care to avoid high-cost episodes, it requires robust risk adjustment to mitigate incentives for stinting on care or avoiding challenging patients.14 High-cost Medicaid managed care enrollees are most likely to switch plans, and high-quality organizations that disproportionately attract complex enrollees often endure the largest negative fiscal impact from such adverse selection.15 During our study period, MassHealth contracted with 6 SCOs of varying size, ranging from a few thousand members for the smallest to nearly 20,000 for the largest. In its Duals Demonstration 2.0, MassHealth has the stated goal of providing statewide access to SCOs.9 Risk adjustment supports this goal by protecting small and mission-oriented organizations against losses from serving disproportionately many high-risk members, thus encouraging broad-based and durable program participation.

Our findings regarding the importance of functional status information for predicting LTSS costs are consistent with prior reports from other state-managed LTSS programs.16 Compared with models developed for frail elderly populations in Wisconsin and New York, which included 38 and 24 predictors, respectively, our model was relatively parsimonious (13 predictors). Predictors included in other state-managed LTSS risk models (but not used in our model) were individual indicators for specific ADLs, IADLs, and diagnoses.16 For Medicaid programs that do not routinely perform MDS-HC 2.0 assessments of community-dwelling older adults, a model similar to ours could be implemented using assessments of ADLs and IADLs only.

Despite considerable attention to social risk factors in recent years,17-19 federal programs do not presently account for these factors in their payment formulas. Massachusetts has used SDOH characteristics for predicting total cost of care (excluding LTSS) among younger (< 65 years) Medicaid managed care enrollees since 2017.10 Our study is the first to examine the role of SDOH characteristics for predicting LTSS costs in an older adult population. Among candidate predictors available in our administrative data, each SD increase above the mean level of neighborhood stress was associated with a $990 increase in annual LTSS costs, whereas the presence of housing problems was associated with a $1531 increase in these costs. Therefore, even after accounting for demographics, functional status, and medical morbidity, it is more expensive to maintain older adults in the community if they are unstably housed (6% of our study population) and/or if they reside in economically stressed neighborhoods.

The additional cost of managing LTSS for socially disadvantaged groups may be due in part to less familial and informal caregiving supports, barriers to transportation, environmental hazards, and other elements linked to socioeconomic positioning.18 In developing the MA-SCO model, we considered social risk factors that were readily available in MassHealth data and from public sources. Better SDOH data would enable fairer payments for those caring for socially vulnerable populations. Thus, we look forward to the success of current initiatives to enrich the ICD-10 codes for SDOH and provide guidance/protocols that encourage their routine use.

Limitations

Limitations of our study reflect broader challenges in developing and implementing risk-adjustment models for managed LTSS programs. Enrollees of 2 of the 6 Massachusetts SCO plans were excluded because their LTSS encounter data were incomplete due to subcapitation arrangements. Encounter data for medical services other than LTSS were also incomplete, which likely contributed to the poor predictive capacity of the summary medical morbidity score. Functional status data are collected by the SCOs themselves using the MDS-HC 2.0 assessment and have not previously been used for payment or accountability in the SCO program. Thus, their quality may have varied across plans. We will evaluate the MA-SCO model’s performance using more recent data after the state and SCO plans have had the opportunity to modify and improve data collection and submission processes.

We developed a model to predict LTSS costs to Medicaid because Medicare is the primary payer for most non-LTSS utilization among dual-eligible enrollees. SCO plans are accountable for both Medicare and Medicaid covered services, which should encourage them to invest in community-based LTSS to reduce high-cost acute and postacute care utilization. However, SCO members with particularly complex needs are shifted to the long-term care program when they enter nursing homes. Medicaid programs need mechanisms to identify and reward SCOs that extend their members’ time spent in the community. Holistic evaluations of the implications of risk-adjustment models for both the Medicaid and Medicare programs (ie, modeling the total cost of care) for dual-eligible populations are also needed. During our 2-year study, average enrollment was less than 1 year, underscoring the high risk for loss of community tenure in this population. Future research should examine factors that contribute to successful preservation of members’ independence in the community and identify differences in institutionalization across plans that may signal low-quality care and early shifting of high-cost patients to long-term care settings. Finally, we have described SCO model development. Program administrators often make policy decisions (eg, setting minimum and maximum payment bounds) when implementing a payment formula that may modify model performance.

CONCLUSIONS

Predictive models using administrative data coupled with functional status assessments can identify and enhance payments for subgroups of members with above-average LTSS needs. Functional status variables accounted for most of the variation in LTSS costs explained by our predictors. For the subgroups of enrollees with housing problems and those residing in economically stressed neighborhoods, only models that incorporated social risk factors adequately predicted their LTSS costs. Without risk adjustment, SCOs that disproportionately served members with more functional limitations and social risk factors could be underpaid by as much as 35%.

Acknowledgments

The authors thank the MassHealth staff who collected the data for this study, provided input during its design and conduct, and approved the manuscript for submission. They especially would like to thank David Garbarino, MPA, and Julie Fondurulia, MS, for sharing their extensive knowledge of their program.


Author Affiliations: Department of Population and Quantitative Health Sciences, University of Massachusetts Medical School (MA, EOM, MAC, ASA), Worcester, MA; School of Public Health, Brown University (MAC), Providence, RI.

Source of Funding: This project was funded by MassHealth.

Author Disclosures: The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (MA, EOM, MAC, ASA); acquisition of data (EOM, ASA); analysis and interpretation of data (MA, EOM, ASA); drafting of the manuscript (MA, EOM, MAC, ASA); critical revision of the manuscript for important intellectual content (MA, EOM, MAC, ASA); statistical analysis (MA, EOM, ASA); obtaining funding (ASA); and supervision (MA, ASA).

Address Correspondence to: Matthew Alcusky, PharmD, PhD, University of Massachusetts Medical School, Albert Sherman Bldg, 6th Floor, 368 Plantation St, Worcester, MA 01605. Email: matthew.alcusky@umassmed.edu.

REFERENCES

1. Medicaid and CHIP Payment and Access Commission. Report to Congress on Medicaid and CHIP. Medicaid and CHIP Payment and Access Commission; 2018. Accessed June 20, 2019. https://www.macpac.gov/publication/june-2018-report-to-congress-on-medicaid-and-chip/

2. Barrett L. Home and Community Preferences of the 45+ Population 2014. AARP Research Center; 2014. Accessed July 12, 2019. https://www.aarp.org/content/dam/aarp/research/surveys_statistics/il/2015/home-community-preferences.doi.10.26419%252Fres.00105.001.pdf

3. Lee H, Shi SM, Kim DH. Home time as a patient-centered outcome in administrative claims data. J Am Geriatr Soc. 2019;67(2):347-351. doi:10.1111/jgs.15705

4. Irvin CV, Bohl A, Stewart K, et al. Money Follows the Person 2015 annual evaluation report. Mathematica. May 11, 2017. Accessed July 12, 2019. https://www.mathematica.org/our-publications-and-findings/publications/money-follows-the-person-2015-annual-evaluation-report

5. Meyer H. A new care paradigm slashes hospital use and nursing home stays for the elderly and the physically and mentally disabled. Health Aff (Millwood). 2011;30(3):412-415. doi:10.1377/hlthaff.2011.0113

6. Ghosh A, Scmitz R, Brown R. Effect of PACE on costs, nursing home admissions, and mortality: 2006-2011. Office of the Assistant Secretary for Planning and Evaluation. March 1, 2015. Accessed March 8, 2019.
https://aspe.hhs.gov/pdf-report/effect-pace-costs-nursing-home-admissions-and-mortality-2006-2011

7. JEN Associates. Massachusetts SCO evaluation—nursing facility residency and mortality summary report. Mass.gov. November 23, 2015. Accessed March 9, 2019. https://www.mass.gov/doc/massachusetts-
sco-evaluation-nursing-facility-residency-and-mortality-summary-report-112315-0/download

8. Massachusetts Medicare-Medicaid integration demonstration: Duals Demonstration 2.0. Mass.gov. August 20, 2018. Accessed January 10, 2019. https://www.mass.gov/doc/final-concept-paper-duals-demonstration-20/download

9. One Care, Senior Care Options (SCO) and Duals Demonstration 2.0: request for information document # FY19EHSCBONECARESCODUALSRFI. Mass.gov. September 5, 2018. Accessed March 9, 2019.
https://www.mass.gov/files/documents/2018/09/07/one-care-sco-and-duals-demonstration-rfi.pdf

10. Ash AS, Mick EO, Ellis RP, Kiefe CI, Allison JJ, Clark MA. Social determinants of health in managed care payment formulas. JAMA Intern Med. 2017;177(10):1424-1430. doi:10.1001/jamainternmed.2017.3317

11. Chen J, Ellis RP, Toro KH, Ash AS. Mispricing in the Medicare Advantage risk adjustment model. Inquiry. 2015;52:0046958015583089. doi:10.1177/0046958015583089

12. Hartmaier SL, Sloane PD, Guess HA, Koch GG, Mitchell CM, Phillips CD. Validation of the Minimum Data Set Cognitive Performance Scale: agreement with the Mini-Mental State Examination. J Gerontol A Biol Sci Med Sci. 1995;50(2):M128-M133. doi:10.1093/gerona/50a.2.m128

13. Björkgren MA, Fries BE, Shugarman LR. A RUG-III case-mix system for home care. Can J Aging. 2000;19(S2):106-125. doi:10.1017/S0714980800013921

14. Miller HD. From volume to value: better ways to pay for health care. Health Aff (Millwood).
2009;28(5):1418-1428. doi:10.1377/hlthaff.28.5.1418

15. Marton J, Yelowitz A, Talbert JC. Medicaid program choice, inertia and adverse selection. J Health Econ. 2017;56:292-316. doi:10.1016/j.jhealeco.2017.04.006

16. Dominiak M, Bohl A. Building managed long-term services and supports risk-adjustment models: state experiences using functional data. Center for Health Care Strategies, Inc. August 2016. Accessed November 18, 2018. https://www.chcs.org/media/CHCS-Building-MLTSS-Risk-Adjustment-Models-08-22-16.pdf

17. Joynt KE, De Lew N, Sheingold SH, Conway PH, Goodrich K, Epstein AM. Should Medicare value-based purchasing take social risk into account? N Engl J Med. 2017;376(6):510-513. doi:10.1056/NEJMp1616278

18. National Academies of Sciences, Engineering, and Medicine. Accounting for Social Risk Factors in Medicare Payment: Identifying Social Risk Factors. The National Academies Press; 2016.

19. National Academies of Sciences, Engineering, and Medicine. Accounting for Social Risk Factors in Medicare Payment. The National Academies Press; 2017.

Related Videos
James Chambers, PhD
Screenshot of Adam Colborn, JD during an interview
Screenshot of an interview with Stuart Staggs
Screenshot of an interview with James Chambers, PhD
Screenshot of an interview with Susan Wescott, RPh, MBA
Stuart Staggs
Lalan Wilfong, MD, during a video interview
dr katherine baker
Julie Patterson, PharmD, PhD
Related Content
© 2024 MJH Life Sciences
AJMC®
All rights reserved.