Association of Participation in a Value-Based Insurance Design Program With Health Care Spending and Utilization

This cohort study evaluates the association of participation in a value-based insurance design program in California with health care spending and utilization.


Introduction
2][3][4][5] VBID encourages members to engage in healthy activities and manage chronic conditions while discouraging unnecessary or unwarranted health care choices to improve member health and reduce total health care costs in the long term.
VBID has been implemented as a cost-sharing model mostly for certain prescription drugs or treatments for specific diseases or chronic conditions or select patient populations.[13] We examined a VBID program that was launched by the California Public Employees' Retirement System (CalPERS) in one of its preferred provider organization (PPO) commercial health plans in 2019.The VBID program was mainly targeted at enhancing primary care for its PPO health plan enrollees.The program's objectives are to provide economic incentives for members to receive high-value coordinated care at the right place and time through a personal primary care physician (PCP); to increase member engagement in health care decisions and reward members for engaging in healthy activities; and to improve outcomes and lower costs and premiums in the long term.
PPO enrollees can choose any in-network physicians when seeking care, which allows freedom of choice of clinician but also increases the risk of fragmented care and higher costs.The CalPERS VBID PPO personal PCP model aims to provide its members a place where their health issues will first be treated by their PCPs, who will guide them through the health care system when referrals to specialists are needed.PCPs also facilitate dialogue between members and clinicians, fostering member participation in decision-making about their own health care, and providing opportunities for disease prevention and health promotion as well as early detection and treatment of conditions, while controlling or reducing costs in the long term. 14e primary VBID interventions include a $10 office visit copayment if a member selects and uses a personal PCP for routine care, including mental health and substance use physician visits.
Otherwise, a $35 office visit copayment is required for a PCP visit, the same as a specialist visit.
Additionally, annual deductibles are reduced from $1000 to $500 for individuals and from $2000 to $1000 for families through completion of the following 5 activities with a $100 credit each for individuals (credit doubled for families): annual biometric screening, annual influenza vaccine, nonsmoking certification, second opinion for elective surgical procedures, and disease management participation.Finally, there is a waiver of inpatient co-insurance (20%) for delivery for expectant mothers enrolled in a healthy mother's program, which provides tools and resources needed for a healthy pregnancy and delivery.
The objective of this study is to evaluate the association of participation in the CalPERS VBID program with health care costs and utilization by its enrollees in 2019 and 2020.We expected a significant VBID association with some of the target services but were unsure about its association with short-term total cost savings.
The VBID interventions were implemented in one of CalPERS PPO health plans, which we label as VBID PPO.We selected another CalPERS PPO plan that did not implement the VBID interventions for comparison, which we call non-VBID PPO.The VBID PPO and non-VBID PPO plan had the same benefit design in terms of copayment, coinsurance, maximum out-of-pocket (OOP) limit, and pharmacy coverage before VBID implementation.The non-VBID PPO had a broader provider (clinicians and hospital systems, etc.) network and higher premium than the VBID PPO, whose premium was set even lower in 2019.After VBID implementation, the VBID PPO annual deductible and PCP visit copayments changed as described previously, but there were no differential changes in copayments for specialist and emergency department visits, coinsurance rate for inpatient hospital admission, annual OOP maximum, and prescription drug coverage and copayments (eTable 1 in Supplement 1).
The VIBD and non-VBID PPO plans are offered during the annual open enrollment period for CalPERS members to compare and choose.The VBID PPO plan provided a mobile health consumer app to increase member engagement through personalized communication and education about the VBID-incentivized activities.CalPERS members also have access to online materials about the VBID PPO incentives, in addition to member outreach and provider communication.The VBID PPO health plan reported that until the fourth quarter of 2019, less than 30% of members had attributed PCPs.
For the 5 incentivized activities, the uptake rates for disease management participation and a second opinion for surgery reached more than 90%.More than half of VBID PPO members received an influenza vaccine, and up to 70% finished a biometric screening and nonsmoking certification.Only 17% of expectant mothers enrolled in a healthy mother's program.
We applied a retrospective cohort study design and a difference-in-differences (DID) approach to compare the relative annual changes in cost and use before and after the 2019 VBID implementation between a cohort of VBID PPO continuous enrollees and a cohort of non-VBID PPO continuous enrollees from 2017 through 2020.The data sources included CalPERS commercial PPO California member enrollment and claims data.
The outcome measures are total allowed payments per member per year for inpatient and outpatient health services.Allowed payments are negotiated prices between health plans and providers, which are often much lower than list prices or charges by providers.The control variables include member age group, sex, region, family relationship and tier, employer type, retirement status, clinical risk score categorical group in 2017 and 2018, and year indicator.

Statistical Analysis
A 2-part DID generalized linear model (GLM) was applied: the first logit GLM model predicts whether an enrollee would have any use or positive payments or not, and the second log-gamma GLM model predicts conditional positive payments.All regression models are adjusted for a VBID PPO indicator, its interaction with a year indicator (using 2018 as the reference year), and the control variables, with the average treatment effect on the treated (ATT) weighting and enrollee-cluster robust SEs.The exponentiated coefficients of the interaction terms from the VBID and year indicators are the multiplicative DID estimators of interest: relative odds ratio in the logit GLM model and relative payments ratio in the log-gamma GLM model. 15ATT weighting is based on propensity scores predicted from a logistic regression of the probability of selecting the VBID PPO cohort on the control variables.ATT weighting sets weights of the treatment group as 1s and weights of the control group as the products of propensity scores and inversed (1-propensity scores) to weight the control group equivalent to the treatment group.

Results
The study population consisted of 94 127 CalPERS PPO health plan non-Medicare members; 50% (48 770) were younger than 45 years old and 52% (47 390) were female.The VBID PPO longitudinal cohort enrollees were significantly younger, more likely to be male, dependent, not single, employed by the state or public agencies, not retired, living in Northern California other than the Bay Area or Sacramento, or healthier in 2017 and 2018 when compared with the non-VBID PPO cohort.However, all significant differences became insignificant after the ATT weighting adjustment, which helped balance the control variables and improved the comparability of the VBID and non-VBID PPO cohorts (Table 1).
The unadjusted per-member per-year mean total allowed payments for the VBID PPO were substantially lower than the non-VBID PPO, but with parallel pre-VBID trends.Many spending measures seem to be influenced substantially by the COVID-19 pandemic in 2020, which varied by service setting and type (Figures 1, 2, and 3).
Among those with any utilization or positive allowed payments, the VBID PPO cohort had significantly higher mean total allowed payments per member per year for PCP office visits in both 2019 and 2020 (adjusted relative payments ratio, 1.05; 95% CI, 1.02-1.08;P < .001).The VBID PPO cohorts also had higher mean payments for laboratory test and medication OOP in 2020, but lower mean OOP payments for inpatient admission in 2020 (adjusted relative payments ratio, 0.90; 95% CI, 0.81-1.00;P = .05).The VBID and non-VBID PPO cohorts had no significant differences for the following utilization and payment measures in both 2019 and 2020: inpatient and outpatient combined, outpatient, prescription drug, OOP combined, emergency department, specialist, and psychiatrist visit.
We provide the estimated overall marginal effects in dollars from the fitted 2-part GLM models in eTable 2 in Supplement 1.The differential changes in per-member mean annual payments are small for most measures.Note that although the marginal effects are intuitively appealing in original measurement units, they are not unbiased DID estimates as the interaction item coefficients for GLMs. 15e VBID PPO subcohort, female and aged between 15 and 55 years old, had higher mean total inpatient cesarean-section positive payments than their non-VBID PPO counterpart in 2019 (Table 2).The sensitivity analysis for those with inpatient delivery yielded very similar results or conclusions (eTable 3 and 4 in Supplement 1).
The coefficients of interactions of the VBID PPO indicator and year indicator before 2019 in the 2-part DID GLMs are the model tests for parallel pre-VBID trends assumption for utilization and payments. 15They were insignificant for almost all measures after ATT weighting (eTable 5 in Supplement 1).

Discussion
In this cohort study, we found that in 2019 or 2020, participation in the CalPERS VBID PPO was associated with (1) higher probability of PCP use, immunizations, and outpatient OOP payments, and lower probability of inpatient surgical and total admissions and OOP payments; (2)  The VBID association with some measures, such as PCP visits, immunizations, and inpatient surgical admissions, are as expected, while others are more nuanced.For example, the VBID PPO cohort experienced a higher chance of paying for outpatient visits OOP, but they also had lower positive outpatient OOP allowed payments (Table 2) and lower overall annual outpatient OOP allowed payments, although most are not significant (eTable 2 in Supplement 1).The VBID PPO cohort experienced a lower chance of paying for inpatient admissions OOP, lower positive inpatient OOP payments (Table 2), and lower overall annual inpatient OOP payments (eTable 2 in Supplement 1).These results suggest that even though the VBID PPO may be associated with a higher chance of outpatient OOP payments, it may come with lower annual OOP payments for both outpatient and inpatient services (Table 2; eTable 2 in Supplement 1).
Another unexpected result concerns cesarean sections.The low uptake rate for the healthy mother's program makes drawing meaningful conclusions difficult regarding the VBID PPO higher positive mean total allowed payments for inpatient cesarean deliveries in a single year.Although statistically insignificant, the VBID PPO had a lower probability of inpatient cesarean deliveries than the non-VBID PPO (Table 2; eTable 3 in Supplement 1).Additionally, there were no essential differences in marginal effects on spending (eTable 2 and 4 in Supplement 1).It should be noted that the healthy mother's program has no explicit goal of reducing cesarean delivery.Similarly, the VBID program's significant association with positive laboratory test payments and medication OOP The CalPERS VBID program needs improvement in scope and depth.One area is to enhance its member engagement.The high uptake rates in disease management participation and second opinion for surgery were primarily because VBID PPO members would automatically receive these incentives until they actively declined them.A similar approach may be used to address the low PCP uptake rate.For example, a PCP would be matched to a member if they do not choose a PCP upon open enrollment, but they still have the freedom to change or drop their assigned PCP anytime over the course of the year.
Improving VBID PPO member engagement with PCPs and participation in the incentivized activities should help achieve the program's objectives.7][18][19] Similarly, the VBID PPO may also automatically enroll all expectant mothers who do not explicitly decline program participation.
For example, eliminating drug copayments if members choose the referenced prescription drugs for targeted diseases or conditions to lower member medication OOP payments and improve medication adherence.These measures will increase the operational complexity of administering pharmacy benefits and may face resistance from health plans and pharmacy benefit managers, which could be addressed and resolved during contract solicitation, negotiation, and renewal phases.
Increasing cost sharing for low-value care to reduce health care waste and cost may be another option for CalPERS to expand its VBID scope but with many challenges. 13,38,39[45][46][47] What's more, its association with reducing total cost may not come to fruition in the short term.One empirical study found that substantially increased cost sharing for low-value services was associated with reduced targeted use but not with total cost savings. 48The literature suggests that clinician-

Limitations
Our VBID study focused on the cost and utilization of general health care providers and services in a large health benefit public purchaser's commercial PPO health plans using rigorous methods and recent data, but with limitations, nevertheless.We applied a fixed cohort approach by requiring continuous enrollment in the VBID PPO or non-VBID PPO from 2017 through 2020 to reduce favorable selection bias due to lower VBID PPO premium and VBID incentives attracting healthy enrollees in 2019 and 2020. 15However, we may not exclude selection bias on unobserved factors without randomization even though we applied a DID ATT weighted regression adjustment.
Therefore, this study cannot tell whether VBID associations with lower inpatient admissions and surgical procedures were from reduced unnecessary procedures or from limited access, although the former was more likely than the latter for these PPOs.Our study findings may not be generalizable to other states' commercial health plans and the Medicare or Medicaid populations.The study time frame is short, and the COVID-19 pandemic had a shock on the 2020 results, which led to a lack of consistent patterns for many significant findings over the first 2 years.We also did not examine clinical outcomes, which warrant further research.The marginal effect sizes of most outcome measures are financially minimal.It may take a longer time and additional incentives targeting medications for certain conditions or populations for the CalPERS VBID PPO to materialize substantial cost savings.However, the VBID literature to date seem to imply that VBID is more effective in redistributing and enhancing health care value than reducing total cost. 6,28,49

Conclusions
In this study, the CalPERS VBID program was associated with desired changes for some interventions with no added total costs in its initial 2 years of operation: higher PCP and immunization utilization or spending, and lower inpatient total and surgical admissions and OOP payments.However, due to the small marginal effect sizes, the COVID-19 pandemic in 2020 and inconsistent results across years, additional years of observation and study are warranted to assure persistent VBID association.
CalPERS may need to adjust current VBID interventions and keep exploring, implementing, and evaluating other VBID incentives and benefit designs to promote value while controlling cost for its health plan members. 50 Inpatient cesarean-section measure was analyzed for female participants aged between 15 and 55 years old.A sensitivity analysis was performed for female participants with any inpatient deliveries.All statistical hypothesis tests were 2-sided tests with a priori significance level of .05using STATA version 14.2 (StataCorp) and SAS Enterprise Guide version 8.3 (SAS Institute).Data were analyzed from September 2021 to August 2022.JAMA Network Open | Health Policy Participation in a Value-Based Insurance Design Program and Health Care Spending and Utilization

Figure 1 .
Figure 1.Unadjusted Yearly Trends of Selected per Member Annual Category Spending Measures by Cohorts

Figure 2 .
Figure 2. Unadjusted Yearly Trends of Selected per Member Annual Out-of-Pocket (OOP) Spending Measures by Cohorts based interventions are more effective than consumer-based interventions, and multicomponent interventions in addressing both clinician and patient roles have the greatest potential for reducing low-value care.13,39

Figure 3 .
Figure 3. Unadjusted Yearly Trends of Selected per Member Annual Service Spending Measures by Cohorts

JAMA Network Open | Health Policy Participation
higher mean positive spending for PCPs, laboratory tests, inpatient cesarean deliveries and medication OOP in a Value-Based Insurance Design Program and Health Care Spending and Utilization

Table 1 .
California Public Employees' Retirement System VBID Study Longitudinal Cohorts Summary Statistics in 2019 ; and lower inpatient OOP payments; and (3) similar total health care spending changes in both years, suggesting the VBID program has the potential to promote valued health care services while controlling costs for CalPERS PPO plan members.
b The risk score is the Diagnostic Cost Group risk score based on member's age, sex, and diagnosis information, which ranged from close to 0 to up to 147 in baseline years.Higher risk score predicts higher health care utilization or costs.payments