Why Threshold Target Performance-Based Metrics May Not Improve Population Health as Much as Treatment Pathways

Many value-based payment agreements (VBAs) and programs incentivize physicians to provide care based on patients’ achievement of treatment targets, especially threshold-based metrics such as maintaining a satisfactory biomarker value. This may not ultimately be the optimal design for a population health program, as they may incentivize physicians and health systems to expend more resources on patients who require relatively less effort to bring to goal compared with patients who are more difficult to bring to goal. Modeling can be helpful to assess whether changing a threshold-based target scheme might yield improvements in health and resource allocation for chronic diseases. In the continuing drive to move the US health care system from a primarily volume-based system to a value-based system, risk-shifting in the form of VBAs only works when the delivery system is designed to achieve maximal overall health improvements.   

To shift incentives in the health care sector to favor “value” over “volume,” private insurers, public payers, and purchasers apply quality measures to monitor processes and outcomes for health plan accreditation and to adjust fee-for-service payments based on performance (eg, The National Committee for Quality Assurance [NCQA] Health Plan Accreditation Program, Centers for Medicare & Medicaid Services [CMS] Medicare Stars program, and MACRA^1-6). In addition, alternative payment models further target and incentivize the achievement of population health goals (eg, CMS alternative payment models).⁷ While advanced value-based payment models (eg, bundled payments and shared savings) are now becoming more popular, pay-for-performance is still common and often used alongside other payment methods.^8,9  

Using quality measures allows for standardized assessment of care across a range of patients,¹⁰ but uniform measures may not adequately capture the diversity of patient experiences with treatment or the challenges physicians face, especially as treatment becomes more personalized. The many value-based payment programs that incentivize physicians to provide care based on patients’ achievement of treatment targets,¹¹ especially threshold-based metrics like maintaining a satisfactory biomarker value (eg, blood pressure under recommended levels), must be designed carefully, as they may incentivize physicians to expend more resources on patients who require relatively less effort to bring to goal (eg, those with marginally elevated blood pressure) compared with patients who are more difficult to bring to goal (eg, those with refractory hypertension). Whether reallocation of effort incentivized by this type of threshold achievement scheme would result in better outcomes at lower costs depends on how the health improvements for the less healthy patients compare with those of the relatively healthier patients.  

The improvement in health from intensifying treatment for a relatively healthy patient may differ from the improvement in health from intensifying treatment for a relatively unhealthy patient. In some cases, it may be that the less healthy group benefits relatively more from treatment even if they do not reach the threshold. In addition, treatment intensification among those in the relatively healthier group may only improve health marginally. 

Diabetes is chronic, progressive, prevalent, and costly ($327 billion in 2017¹²), and there is a long history of value-based arrangements that include threshold-based goals.¹³ Good blood glucose control has been shown to improve patient outcomes, reducing rates of expensive and debilitating diabetes-related complications.¹⁴ Comprehensive Diabetes Care is part of NCQA’s HEDIS measure set, by which health plans are evaluated for accreditation, and blood glucose testing and goal attainment (ie, HbA_1c poor control >9.0%, HbA_1c control <8.0%, HbA_1c control <7.0% for a selected population) are core components.^15,16 Three features of diabetes and its treatment suggest that intensification might provide more health improvements for patients with poor blood glucose control than those with relatively good glucose control: (1) diabetes complication risks generally increase exponentially (ie, increase at an increasing rate) with rising blood glucose levels¹⁴; (2) the benefits of a 1% point reduction in HbA_1c increase in absolute terms as complication risks increase^14,17,18; and (3) individuals with higher HbA_1c levels experience larger absolute HbA_1c reductions from treatment, regardless of drug class or mode of action.^19,20 Taken together, these features make it probable that individuals with higher HbA_1c values will experience greater absolute benefits from treatment (even if they do not meet target HbA_1c goals).  

Despite widespread use of value-based payment incentives and a myriad of options for controlling blood glucose, the proportion of patients with poor blood glucose control remains high in the United States for all types of insurers. In 2017, the percent of individuals with diabetes with HbA_1c>9.0% ranged across different provider categories from 22.3% for Medicare PPO to 41.2% for Commercial PPO and the percent with HbA_1c>8.0% ranged from 32.8% for Medicare PPO to 51.6% for Medicaid HMO.¹⁵

Whether to expend resources and implement alternative value-based payment schemes in care to address this issue depends on the size of the potential health gains and improvements in resource allocation. Economic modeling (ie, mathematical equations that synthesize available data such as short-run clinical trial outcomes, risk equations and progression rates, and known physiological relationships into a coherent, internally consistent framework) can be used to understand the implications of alternative payment structures on health outcomes. This article presents the results of a modeling exercise in diabetes to illustrate whether changing a threshold-based target scheme might yield improvements in health and resource allocation as implications of these findings may apply to other chronic and life-threatening diseases. In particular, cancer is a disease where there is significant burden and where important economic and outcomes data will be available for modeling in the near future; the Center for Medicare and Medicaid Innovation (CMMI) is now approaching year four of their large scale and bold CMMI Oncology Care Management (CMMI OCM) program.

Economic Modeling in Diabetes 

In diabetes, economic modeling has a long history of use in aiding decision-makers.^21-24  Generally in these models, patients vary by individual characteristics (eg, demographics, biomarker values, presence of diseases), which can be modified over time. Treatments can change biomarker values and result in adverse events. Biomarker values and health status influence the risks of diabetes complications and mortality, which are used to calculate predicted event counts. Model results are usually summarized at a population level and can be costed. Health impacts are often transformed into a standardized disease burden metric known as quality-adjusted life-years (QALYs), which adjusts longevity estimates for patient views on the extent to which health problems decrease their health-related quality of life (HRQoL).^25-28    

To illustrate how modeling can be helpful to assess whether changing a threshold-based target scheme might yield improvements in health and resource allocation, we estimated changes in health outcomes attributable to treatment intensification for a cohort of diabetes patients “Near to Goal” (ie, mean HbA_1c of 7.5%, ranging from 7.0%-8.0%) and compared these outcomes to those of a cohort of patients “Far from Goal” (ie, mean HbA_1c of 9.5%, ranging from 9.0%-10.0%) using an established model (ECHO-T2DM: Economic and Health Outcomes Model of Type 2 Diabetes Mellitus^24,29-33). The exercise was set-up so that outcomes (microvascular and macrovascular event rates over time, HRQoL, and longevity) for each cohort were simulated first without, and second with, the benefit of a hypothetical blood glucose lowering agent. Based on the inverse relationship between glucose lowering and glucose levels discussed above, we assumed that those “Near to Goal” would experience a reduction of 1.0%, while those “Far from Goal” would have a greater reduction of 1.6%.^19,20 Patient characteristics were primarily sourced from real-world data in the United States³⁴ and supplemented with individual-level data from randomized controlled trials where necessary^35,36 In both simulations, when estimated HbA_1c exceeded the treatment target of HbA_1c<7.0%, patients ultimately received insulin on top of their initial treatment to control glucose, with a median lag of 1.5 years to reflect the typical way that insulin is used as an add-on therapy in type 2 diabetes.³⁷ The model also includes treatment-specific risks for hypoglycemic episodes, with higher rates associated with insulin therapy. Given the chronic nature of diabetes, we simulated individuals for up to 25 years or until death to capture the full impact of the hypothetical treatment on outcomes. 

The mean HbA_1c trajectories for the treated and untreated patient groups in the “Near to Goal” and “Far from Goal” cohorts are presented in Figure 1. By design, the initial difference in HbA_1c between the treated and untreated groups was greater for the “Far from Goal” cohort than for the “Near to Goal” cohort. The long-term convergence of the trajectories toward an HbA_1c of 7.0% reflects the increasing use of insulin as an add-on therapy in each group over time.

The hypothetical treatment was associated with reduced risk of diabetes-related microvascular and macrovascular complications in both cohorts. Relative risk reductions associated with treatment vs no treatment were larger in the patients in the “Far from Goal” compared with the “Near to Goal” cohort. (Figure 2) In this exercise, the biggest difference was found for the development of heart failure: treatment vs no treatment in the “Far from Goal” cohort reduced risk by 4.5% and only 2.4% in the “Near to Goal” cohort. Life-year and QALY gains over 25 years for treated vs untreated patients were likewise larger for the “Far from Goal” cohort than for the “Near to Goal” cohort (0.035 vs 0.018 life-years and 0.349 vs 0.145 QALYs, respectively). 

Lessons Learned

This modeling exercise illustrates a case where patients whose HbA_1c levels were far from the recommended glycemic target experienced greater benefits from treatment compared with patients whose HbA_1c levels were closer to the treatment target, based on reasonable assumptions. By allocating resources to bring patients who are far from goal closer to their target, greater long-term returns on the initial investment may be seen due to decreased burden on the health care system for complications that might have developed. Therefore, current incentive structures for treatment that are based on achieving specific treatment targets may be misaligned with a patient-centered treatment approach. In fact, because many factors (eg, comorbidities, life expectancy) can influence treatment, clinicians support the idea that a universal glycemic target may not be suitable for all patients.¹⁰ 

Because all areas of health care seeking VBAs have measurement protocols, the possibility of pathway protocols, or other value-based constructs, authors urge caution; arrangements that define success using discrete threshold metrics may not achieve full population health value due to subtly conflicting incentives. For example, consider the following Oncology Care Model (OCM) metric: Proportion of patients who died who were admitted to hospice for 3 days or more (OCM-3). Success is defined naturally as a higher score. While the intent of this measure is to incentivize appropriate care, it lacks the level of detail needed to accurately distinguish between whom to continue treating and whom not to treat. A practice could score well in curtailing the administration of chemotherapy but never have addressed the issue that a subset of those patients may not have needed as extensive treatment. Additionally, the threshold metric of days in hospice does not address such things as the benefits of early palliative care interventions for certain patients.³⁸  Modeling different scenarios in oncology pathways could help decision-makers and care providers identify algorithms that provide maximal potential health gains. The quality and value-based care environment continues to evolve, and it is essential that it reflects clinical nuance for patient-specific needs.

Conclusion

Modeling exercises can provide estimates of the potential health gains, reduction in risk, and improvements in care provision for different cohorts of patients or different measurement designs to enable clinically sensitive alternative payment models. Moving forward, increasing focus should be placed on the design and implementation of quality measures in incentive programs emphasizing aspects of care provision that will have the greatest benefits for individual patient risk factors and express clinical nuance, rather than relying on absolute thresholds.¹¹

References

1. National Committee on Quality Assurance. Diabetes Recognition Program (DRP). ncqa.org website. https://www.ncqa.org/programs/health-care-providers-practices/diabetes-recognition-program-drp/. Accessed March 27, 2019.

2. NCQA release new standards category – population health management [news release]. Washington, DC: National Committee on Quality Assurance; August 3, 2017.  https://www.ncqa.org/newsroom/details/ncqa-releases-new-standards-category-population-health-management?ArtMID=11280&ArticleID=88&tabid=2659. Accessed March 27, 2019.

3. Fallon Health. 2018 Healthcare Effectiveness Data and Information Set  (HEDIS^®) measures. https://www.fchp.org/providers/resources/hedis-measures.aspx. Updated January 2019. Accessed March 27, 2019.

4. Centers for Medicare & Medicaid Services. Quality measures. cms.gov website. https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/QualityMeasures/index.html?redirect=/QUALITYMEASURES/. Updated March 5, 2019. Accessed March 27, 2019.

5. National Committee on Quality Assurance. Health Plan Accreditation (HPA). ncqa.org website. https://www.ncqa.org/Portals/0/Programs/Accreditation/HPA/2018_HPA_SGs.pdf?ver=2018-02-16-150007-887. Accessed March 27, 2019.

6. Department of Health & Human Services, Quality Payment Program. MIPS Overview. qpp.cms.gov website. https://qpp.cms.gov/mips/overview. Accessed March 27, 2019.

7. Department of Health & Human Services, Quality Payment Program. APMs Overview. qpp.cms.gov website. https://qpp.cms.gov/apms/overview. Accessed March 27, 2019.

8. Delbanco SF, Lehan M, Murray, R. The evidence on pay-for-performance: not strong enough on its own? [blog]. Health Affairs. October 24, 2018. https://www.healthaffairs.org/do/10.1377/hblog20181018.40069/full/. Accessed March 27, 2019.

9. Department of Health & Human Services, Quality Payment Program. Advanced Alternative Payment Models (APMS). qpp.cms.gov website. https://qpp.cms.gov/apms/advanced-apms. Accessed March 27, 2019.

10. Cahn A, Raz I, Kleinman Y, et al. Clinical assessment of individualized glycemic goals in patients with type 2 diabetes: formulation of an algorithm based on a survey among leading worldwide diabetologists. Diabetes Care. 2015;38(12):2293-2300. doi:10.2337/dc15-0187

11. Mendelson A, Kondo K, Damberg C, et al. The effects of pay-for-performance programs on health, health care use, and processes of care: a systematic review. Ann Intern Med. 2017;166(5):341-353. doi:10.7326/M16-1881

12. American Diabetes Association. Economic costs of diabetes in the U.S. in 2017. Diabetes Care. 2018;41(5):917-928. doi:10.2337/dci18-0007

13. Damberg CL, Sorbero ME, Lovejoy SL, Martsolf GR, Raaen L, Mandel D. Measuring success in health care value-based purchasing programs: findings from an environmental scan, literature review, and expert panel discussions. Rand Health Q. 2014;4(3):9.

14. Stratton IM, Adler AI, Neil HAW, et al. Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study. BMJ. 2000;321(7258):405-412.

15. National Committee on Quality Assurance. Comprehensive Diabetes Care (CDC). ncqa.org website. https://www.ncqa.org/hedis/measures/comprehensive-diabetes-care/. Accessed March 27, 2019.

16. National Committee on Quality Assurance. 2019 HEDIS Summary Table of Measures, Product Lines and Changes. https://www.ncqa.org/wp-content/uploads/2018/
08/20190000_HEDIS_Measures_SummaryofChanges.pdf. Accessed March 27, 2019.

17. UK Prospective Diabetes Study (UKPDS) Group. Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet. 1998;352(9131):837-853.

18. UK Prospective Diabetes Study Group. Effect of intensive blood-glucose control with metformin on complications in overweight patients with type 2 diabetes (UKPDS 34). Lancet. 1998;352(9131):854-865.

19. DeFronzo RA, Stonehouse AH, Han J, Wintle ME. Relationship of baseline HbA1c and efficacy of current glucose-lowering therapies: a meta-analysis of randomized clinical trials. Diabet Med. 2010;27(3):309-317. doi:10.1111/j.1464-5491.2010.02941.x

20. Bloomgarden ZT, Dodis R, Viscoli CM, Holmboe ES, Inzucchi SE. Lower baseline glycemia reduces apparent oral agent glucose-lowering efficacy: a meta-regression analysis. Diabetes Care. 2006;29(9):2137-2139.

21. Eddy DM, Schlessinger L. Validation of the archimedes diabetes model. Diabetes Care. 2003;26(11):3102-3110.

22. Hoerger TJ, Segel JE, Zhang P, Sorensen SW. Validation of the CDC-RTI Diabetes Cost-Effectiveness Model. Research Triangle Park, NC: RTI International; 2009.

23. Palmer AJ, Roze S, Valentine WJ, et al. Validation of the CORE Diabetes Model against epidemiological and clinical studies. Curr Med Res Opin. 2004;20(suppl 1):S27-S40.

24. Willis M, Asseburg C, He J. Validation of economic and health outcomes simulation model of type 2 diabetes mellitus (ECHO-T2DM). J Med Econ. 2013;16(8):1007-1021. doi:10.3111/13696998.2013.809352

25. Drummond M, Sculpher M, Torrance G, O´Brien B, Stoddart G. Methods for the Economic Evaluation of Health Care Programmes. 3rd ed. United Kingdom: Oxford University Press; 2005.

26. Bagust A, Beale S. Modelling EuroQol health-related utility values for diabetic complications from CODE-2 data. Health Economics. 2005;14(3):217-230.

27. Clarke P, Gray A, Holman R. Estimating utility values for health states of type 2 diabetic patients using the EQ-5D (UKPDS 62). Med Decis Making. 2002;22(4):340-349.

28. Sullivan PW, Ghushchyan VH. EQ-5D scores for diabetes-related comorbidities. Value Health. 2016;19(8):1002-1008.

29. Gupta V, Willis M, Johansen P, et al. Long-term clinical benefits of canagliflozin 100 mg versus sulfonylurea in patients with type 2 diabetes mellitus inadequately controlled with metformin in India. Value Health Reg Issues. 2019;18:65-73.

30. Neslusan C, Teschemaker A, Johansen P, Willis M, Valencia-Mendoza A, Puig A. Cost-Effectiveness of Canagliflozin versus Sitagliptin as add-on to Metformin in patients with type 2 diabetes mellitus in Mexico. Value Health Reg Issues. 2015;8:8-19.

31. Neslusan C, Teschemaker A, Willis M, Johansen P, Vo L. Cost-effectiveness analysis of canagliflozin 300 mg versus dapagliflozin 10 mg added to metformin in patients with type 2 diabetes in the United States. Diabetes Ther. 2018;9(2):565-581.

32. Sabapathy S, Neslusan C, Yoong K, Teschemaker A, Johansen P, Willis M. Cost-effectiveness of canagliflozin versus sitagliptin when added to metformin and sulfonylurea in type 2 diabetes in Canada. J Popu Ther Clin Pharmacol. 2016;23(2):e151-e168.

33. Willis M, Johansen P, Nilsson A, Asseburg C. Validation of the Economic and Health Outcomes Model of Type 2 Diabetes Mellitus (ECHO-T2DM). Pharmacoeconomics. 2017;35(3):375-396.

34. Optum Insight Claims D [proprietary data]. Unpublished analysis; 2014.

35. Lavalle-Gonzalez FJ, Januszewicz A, Davidson J, et al. Efficacy and safety of canagliflozin compared with placebo and sitagliptin in patients with type 2 diabetes on background metformin monotherapy: a randomised trial. Diabetologia. 2013;56(12):2582-2592.

36. Cefalu WT, Leiter LA, Yoon KH, et al. Efficacy and safety of canagliflozin versus glimepiride in patients with type 2 diabetes inadequately controlled with metformin (CANTATA-SU): 52 week results from a randomised, double-blind, phase 3 non-inferiority trial. Lancet. 2013;382(9896):941-950.

37. Khunti K, Gomes MB, Pocock S, et al. Therapeutic inertia in the treatment of hyperglycaemia in patients with type 2 diabetes: A systematic review. Diabetes Obes Metab. 2018;20(2):427-437.

38. Temel JS, Greer JA, Muzikansky A, et al. Early palliative care for patients with metastatic non–small-cell lung cancer. N Engl J Med. 2010;363(8):733-742. doi:10.1056/NEJMoa1000678