The Health Care Incentives Improvement Institute collaborated with Community Health Choice, a 350,000-member nonprofit Medicaid HMO in southeastern Texas, to design, implement, and measure a maternity-care bundled-payment pilot program. First-year results of this 2-year effort show that bundled payments for maternity should include the costs and outcomes of pregnancy, delivery, and neonatal care — and that this type of model can help to achieve broader population-health and cost goals.
Comprehensive maternity-episode budgets should include all phases of the care: pregnancy, delivery, and neonatal.
A bundled-payment budget methodology can provide an explicit financial incentive to reduce unwarranted C-sections.
Even a few high-need, high-cost infants can greatly affect a provider’s actual costs and its potential for shared savings in a bundled-payment model.
A comprehensive maternity-care bundled-payment program relies on effective linkage between maternal and newborn IDs at all stages of care.
In 2014, the Association for Community Affiliated Plans sponsored a Bundled Payment Learning Collaborative that was supported by Bailit Health Purchasing and the Health Care Incentives Improvement Institute (HCI3). Nine health plans participated, including Community Health Choice (CHC), a 350,000-member nonprofit Medicaid HMO in southeast Texas that has a network of 76 hospitals and more than 10,000 providers. Three main challenges prompted CHC to pilot a maternity-care bundled-payment program:
- CHC provides care for 50% of the infant births in Greater Houston.
- Government regulators of CHC’s line of business endorse and support the use of alternative payment models.
- A key part of CHC’s mission is to improve maternal and child health outcomes, while reducing costs.
CHC’s goal was to focus its maternity-care bundled-payment pilot on the STAR population (low-income children and pregnant women) served by University of Texas Health (UTHealth) and University of Texas Medical Branch (UTMB) — separate provider systems that each account for 36% of CHC’s infant deliveries. The providers were motivated to participate by their strong relationships with CHC, their interest in receiving the quality assessments that the pilot study involved, and the potential for shared savings. CHC contracted with HCI3 to help support methods and analytics for a 2-year pilot; this case study focuses on the first year, from March 2015 through February 2016.
We started by analyzing at least 2 years of CHC’s most recent available data, to establish a baseline bundled-payment case rate (i.e., a budget for an episode of care). Before the pilot, HCI3 used its proprietary software to run 2 years of historical CHC claims. The data enabled CHC and the providers to arrive at negotiable contract parameters. Here is a small subset of those parameters:
- Each maternity care episode has three components: pregnancy, delivery, and newborn. The delivery itself triggers the care episode.
- The pregnancy episode begins up to 270 days pre-delivery and ends at delivery; the delivery episode begins 3 days before birth and ends 60 days after discharge; the newborn episode begins at birth and extends 30 days post-discharge.
- Care episodes include only those where the provider is responsible for the delivery, regardless of whether it also does the prenatal care.
- Year 1 established upside-only shared savings for the provider (split 50/50 with CHC); Year 2 will include upside and downside risk.
According to the CHC-wide historical data for about 25,000 births in 2012 and 2013, the average episode cost — adjusted for provider-specific prices, case mix, and other pertinent factors — was $8,826: $1,973 for prenatal, $3,704 for delivery, and $3,149 for neonatal. The HCI3 website provides detailed definitions for C-section, vaginal delivery, pregnancy, and newborn care (including the services and procedure/diagnosis codes that define triggers, relevant services, complications, and episode subtypes).
Expected costs: The pregnancy budget was determined using average prenatal costs from each provider’s historical data, prorated according to when the mother actually began receiving care (the majority started between Month 2 and Month 5 of the pregnancy). The delivery budget — derived from a blended C-section/vaginal delivery rate using historical C-section data — accounts for a risk-factor profile for each patient, including demographics, comorbidities, and type and severity of the care episode. The newborn budget was derived using the average historical costs for newborns (coded as American Academy of Pediatrics nursery levels 1, 2, and 3). Maternity episodes involving nursery level 4 infants (representing <4% of babies in the historical data, across all providers) were excluded from the pilot because of that level’s outsize influence on average episode costs. Similarly, maternity-bundle budgets excluded level 4 newborn costs.
Reporting: Quarterly reports for each provider compare the risk-adjusted maternity-episode budgets with actual claims costs for each patient (specified by pregnancy, delivery, and newborn care). Therefore, providers and CHC can continuously and cumulatively assess budget status, drill down to individual patient results, and identify potential causes for going over budget. HCI3 also facilitated meetings between CHC, UTHealth, and UTMB to tailor quality scorecards to the pertinent metrics and collection capabilities of each provider. Most measurement categories were consistent between the providers, but some items were collected for only one provider given the other’s challenges in extracting data from electronic health records. Each provider group separately negotiated scorecard point values for the measures.
Results: During Year 1, the two provider systems collectively completed 1,246 maternity episodes (UTMB, 774; UTHealth, 472). Overall results differed greatly between the two systems: one was 34% over budget; the other was 4% under budget.
A change in the rate of C-sections was a substantial driver of differences in delivery costs: Compared with the 36% historical rate for each provider, provider A had a higher C-section rate (38%) and provider B had a lower rate (33%). Provider A also was greatly over budget for newborns, in part because babies with birth defects (costs exceeding $150,000 each) were not designated as nursery level 4 and, therefore, not excluded. As a result, in Year 2, the study will use a stop-loss cap rather than exclusion-based nursery-level criteria (see the Lessons Learned below).
We also found that the costs of maternity and delivery have an inverse relationship with the birthweight of the baby, as shown in the figure below. Therefore, costs are associated with an important neonatal outcome.
Year 1 of the pilot yielded lessons in several domains:
Medicaid population: Many mothers in this Medicaid population had transitory membership in CHC. This fact made it more difficult to document their claims history, thereby requiring greater reliance on historical average costs for prenatal care in each provider system. In addition, for some deliveries (2% for one system, 12% for the other), the patient did not receive any Medicaid-paid prenatal care. These delivery-only patients were given a $0 budget for the pregnancy portion but tended to have much higher delivery and newborn costs. Therefore, if we were to exclude delivery-only patients, we would also need to exclude them from the historical data that underlie the budgets.
Blended-delivery budgeting: Creating the expected delivery costs by blending C-section and vaginal-delivery costs (weighted by historical C-section rates) ensures that C-section deliveries always exceed the budget and vaginal deliveries do not. This creates an explicit potential incentive to reduce unwarranted C-sections, an important population health goal.
Indeed, one provider system lowered its C-section rate and achieved a financial benefit; the other did not. Neither provider reported making explicit changes in care to reduce the C-section rate, but the reduction was rewarded for the provider that managed it.
Linkage of mothers and babies: In this pilot, CHC was able to link a mother’s member ID and her newborn’s ID so that the accumulated costs for the newborn were combined with the appropriate prenatal and delivery costs of the mother. Some states, including New York, have statewide reporting and surveillance systems that enable such a link between mother and child. For managed-care organizations to implement a comprehensive maternity bundled-payment program, such a link is essential, if not facilitated by the state then by the health plan itself.
Nursery-level designations: Perhaps the biggest lesson was that the level 4 nursery designation was not an objective indicator of severity and cost. Provider systems vary in how they interpret American Academy of Pediatrics guidelines on this matter. On average, rates of level 4 coding by pilot participants ranged from 2% to 16% (even though, overall, the historical rate was <4% across all providers), with some notable variations among babies with birth defects (as discussed above).
To protect the provider from big asymmetrical losses associated with high-risk newborns who may not need the level 4 designation, and to protect the plan from potential arbitrary newborn nursery-level placement or coding, Year 2 of the pilot will include all nursery levels and set a stop-loss cap of between $30,000 and $60,000. The new stop-loss cap will affect a greater number of patients and substantially diminish the effect of outlier episodes.
Quality scorecards: Both provider systems were passionate about identifying and defining quality measures, making clear the need to tailor scorecards to the institution even though most of the final measures were common to both. Given that quality scorecards had to be realigned halfway through Year 1, Year 2 quality data will be compared with only the second half of Year 1, as well as from quarter to quarter within Year 2.
Shared savings: The pilot revealed a disproportionate distribution of risk between payer and provider given that an episode of care has a minimum base cost (floor) but a high potential ceiling. We determined that a small number of high-cost events should not wipe out a much larger number of smaller-value savings on other newborns or other features of the care for a baby who happened to have a high-cost event. In effect, the provider should have the opportunity to receive a higher share of upside and a lower share of downside risk. Therefore, Year 2’s savings allocations will allow for a greater potential share of savings and a smaller potential share of losses for any given quality score.
Policy: Medicaid pays for roughly 50% of all U.S. births. Harm to mother and child from unwarranted C-sections, premature births, and other causes of low birthweight have long-term consequences for these patients and society at large. In a recent whitepaper, the Health Care Payment Learning and Action Network (LAN), stated: “Often prenatal care, labor and birth, and postpartum care are viewed and delivered as three distinct periods. However, by viewing them as three phases within one episode, there is a potential for incentivizing the types of interactions and care delivery that support positive outcomes.” The CHC pilot study has informed both the LAN whitepaper and the New York State Value-Based Payment Roadmap on maternity bundles, as it illustrates how a Medicaid payer and its affiliated providers can work collaboratively to integrate all phases of maternity care, encourage prenatal care, discourage C-sections, and prevent premature and other low-birthweight deliveries.
Read all NEJM Catalyst articles on bundled payments.
This case study originally appeared in NEJM Catalyst on October 18, 2016.