Abstract
Objective
To examine the impact of Oregon's Coordinated Care Organizations (CCOs), an accountable care model for Oregon Medicaid enrollees implemented in 2012, on neonatal and infant mortality.
Data Sources
Oregon birth certificates linked with death certificates, and Medicaid/CCO enrollment files for years 2008‐2016.
Study Design
The sample consisted of the pre‐CCO birth cohort of 135 753 infants (August 2008‐July 2011) and the post‐CCO birth cohort of 148 650 infants (August 2012‐December 2015). We used a difference‐in‐differences probit model to estimate the difference in mortality between infants enrolled in Medicaid and infants who were not enrolled. We examined heterogeneous effects of CCOs for preterm and full‐term infants and the impact of CCOs over the implementation timeline. All models were adjusted for maternal and infant characteristics and secular time trends.
Principal Findings
The CCO model was associated with a 56 percent reduction in infant mortality compared to the pre‐CCO level (−0.20 percentage points [95% CI: −0.35; −0.05]), and also with a greater reduction in infant mortality among preterm infants compared to full‐term infants. The impact on mortality grew in magnitude over the postimplementation timeline.
Conclusions
The CCO model contributed to a reduction in mortality within the first year of birth among infants enrolled in Medicaid.
Keywords: Health Care Organizations and Systems, Integrated Delivery Systems, Medicaid, Pediatrics, State Health Policies
1. INTRODUCTION
In the United States, infants and toddlers under 3 years of age are most likely to live in poverty.1 Research has linked having low‐income parents to poor health outcomes among infants and young children, such as preterm birth, low birthweight, acquired chronic conditions,2, 3, 4, 5, 6 and especially increased risk of infant death.5, 6 Having insurance coverage may contribute to improving access to care and subsequently reducing infant mortality.7, 8, 9 However, more than insurance coverage is required to further improve infant survival. A health care delivery system that ensures timely and quality care for women and their infants is also crucial to improving infant's health outcomes.
Oregon's Medicaid program, known as the Oregon Health Plan (OHP), is the largest health insurer for children in Oregon. OHP currently covers approximately 50 percent of all births and over 400 000 children (0‐18 years old) from low‐income families (income lower than 200 percent federal poverty level).10, 11, 12 OHP provides a comprehensive benefit package with services under the Early and Periodic Screening, Diagnosis, and Treatment (EPSDT) program for children with no or little cost sharing. As of 2011, nearly all Medicaid beneficiaries in Oregon were mandatorily enrolled in fully capitated Medicaid Managed Care Organizations (MCOs). Although MCOs covered physical health care services, behavioral health and dental care services were frequently not covered.13 In addition, different MCOs provided physical, behavioral, and dental health services separately without any coordination. This fragmentation often led to greater costs and poorer health outcomes.14
Since August 2012, Oregon has transformed its health care delivery model for Medicaid enrollees with the implementation of Coordinated Care Organizations (CCOs). Fifteen regional CCOs are currently serving more than 90 percent of Oregon's Medicaid population, both adults and children, providing integrated care for physical, behavioral, and dental health.12 The CCO arrangements are considered a unique Medicaid Accountable Care Organization (ACO) model. CCOs share similarities with ACOs in that they are held accountable for both cost and quality of care, and through the use of payment incentives that reward improved health outcomes and saved cost. However, this is a mandatory enrollment model with required integration of physical, behavioral, and dental health care in which enrollees are served by a geographically defined network of providers. CCOs receive a global budget and focus on primary care through patient‐centered medical homes to improve care for its beneficiaries. Each CCO has its governance from not only health care providers but also members of a community advisory council that ensure community's health needs are met.14, 15, 16, 17, 18
Medicaid ACOs have been shown to improve access to care and care quality in some areas, for example, increasing in preventive care, decreasing in emergency department visits or inpatient days, and containing cost growth.18, 19, 20, 21, 22, 23 Few studies have, however, investigated the impact of pediatric Medicaid ACO models on health care utilization, cost, and care quality23, 24, 25 partly because of modest growth of pediatric ACOs.26 Only one study to our knowledge examined the effect of Medicaid ACOs on childbirth outcomes and found no impact on infant inpatient mortality.27 Specific for Oregon, a recent study investigated the impact of CCO implementation on neonatal and infant outcomes, including infant mortality.28 Findings from this study indicated that CCO implementation was not associated with a reduction in infant mortality after 1 year of implementation. Another study, however, found that following Oregon's CCO implementation, women on Medicaid experienced a significant increase in receiving timely prenatal care.19 The CCO model is designed to enhance access to integrated care for women before and during pregnancy as well as infants after birth, via primary care homes. Further evidence is needed to document whether CCO implementation has an impact on infant mortality over a longer time frame.
In this study, we followed CCO implementation for 4 years to examine the extent to which CCO implementation had an effect on mortality of infants enrolled in Medicaid during the first year of birth. Preterm infants are considered a high‐risk population because they have higher risk for chronic conditions, mortality, and greater needs for follow‐up care and care coordination compared to full‐term babies.29, 30, 31 Because CCOs target high‐risk and high‐cost patients, we investigated whether the effects of CCOs on mortality differed between preterm infants and full‐term infants. We also tested whether the impact of CCOs on infant mortality increased over the first 4 years of implementation.
2. METHODS
2.1. Data sources
Data came from multiple sources, including Oregon birth certificates, death certificates, and Medicaid and CCO enrollment data for years 2008 through 2016. Every infant had a unique person identification number that allowed the same infants across the different data sources to be linked deterministically with linkage rate of 100 percent. Also, each infant had a unique mother‐to‐child identification number that enabled linkage between infant's information and maternal characteristics. Data linkage was performed by the Oregon Health Authority.
2.2. Sample
Using Oregon 2008‐15 birth certificates, we created the pre‐ and post‐CCO cohorts of births. The pre‐CCO birth cohort included births from August 2008 to July 2011. CCO implementation started in August 2012, and we, therefore, excluded births from August 2011 to July 2012 because health care for these infants could be partially provided by CCOs during the first year after birth. The cohort of babies born between August 2012 and December 2015 was defined as the post‐CCO births. The birth certificates were then linked to death certificates to identify infants who died within a year after birth. For Medicaid‐financed births, Medicaid provides coverage for infants from birth to 1 year of age. Therefore, we then linked with Medicaid/CCO enrollment files to identify infants enrolled in Medicaid. Infants enrolled in Oregon's Medicaid program include those who are eligible for Medicaid or the Children's Health Insurance Program. An infant enrolled in Medicaid, henceforth Medicaid infant, was defined as being enrolled in Medicaid on the date of birth or within 2 weeks of birth to account for possible delay in Medicaid enrollment for newborns. Infants not enrolled in Medicaid, henceforth non‐Medicaid infants, were defined as those who were not enrolled in Medicaid anytime during their first year after birth.
Beginning in August 2012, almost all Medicaid beneficiaries were enrolled in 16 CCOs, except for a small number of people exempt from CCO enrollment (eg, American Indians, Alaska Natives, members eligible for both Medicaid and Medicare, or members with special health needs).12, 32, 33 We excluded 2592 infants (2.8 percent of post‐CCO Medicaid infants) who were not enrolled in CCOs during the post‐CCO period. Therefore, all Medicaid infants in our post‐CCO cohort were enrolled in CCOs. We also excluded 1639 infants who had congenital anomalies because these infants require specific neonatal/pediatric care and have a higher risk of mortality compared to those who do not have the conditions. The sample included 135 753 pre‐CCO infants (77 850 Medicaid infants and 57 903 non‐Medicaid infants) and 148 650 post‐CCO infants (88 683 Medicaid infants and 59 967 non‐Medicaid infants).
2.3. Outcome variables
We examined two mortality outcomes: neonatal mortality and infant mortality. Neonatal mortality was defined as deaths within 28 days after birth, and infant mortality was deaths within a year of birth.
2.4. Main independent variables
The indicator of the post‐CCO period took a value of 1 if an infant was born during the post‐CCO period. We included an indicator for CCO infant (ie, Medicaid infant) that took a value of 1 if an infant had a Medicaid enrollment record on the date of birth or within 2 weeks of birth. We also used an indicator of preterm infant as defined by babies born before the 37th week of pregnancy using information on gestation weeks from birth certificates.
To estimate how the impact of CCOs grew over the implementation timeline, we used a monthly time indicator for the length of CCO implementation. This variable equaled 0 if an infant was born before the implementation of CCOs (pre‐CCO infants). It took a value of 1 if an infant was born during the month when a CCO started its implementation and increased by 1 each month post‐CCO.
2.5. Covariates
We examined the effects of CCOs on mortality while controlling for a set of maternal characteristics, including age, race/ethnicity, education, marital status, rurality, body mass index (BMI) at delivery, smoking during pregnancy, and number of previous births. Mother's age and BMI at delivery were measured continuously while other maternal characteristics were binary indicators, for example, dichotomous variables for each race/ethnic category, each education level, binary variables indicating whether mother was married at the time of birth, or smoking during pregnancy. We applied the Rural‐Urban Commuting Area (RUCA) criteria to the ZIP code of residence to create a rurality variable with three categories, including urban, large rural, and small rural areas.34
The impacts of CCO implementation were also adjusted for important birth characteristics, that is, presence of birth risk factors, and birth plurality. We also adjusted the effects of CCO implementation for time trends in the outcomes over the study period.
2.6. Statistical analyses
We employed a difference‐in‐differences (DD) approach to estimate the effect of the CCO implementation on mortality. Our main DD model compared the average change in probability of mortality between the pre‐CCO and post‐CCO periods among Medicaid infants, who were affected by the CCO implementation, to the average change in mortality between the two periods among non‐Medicaid infants, who were not influenced by CCOs. To justify the DD approach, we tested for a common trend of mortality between Medicaid and non‐Medicaid infants in the pre‐CCO period. The tests were not statistically significant (P > .05), which indicated that mortality trends in the pre‐CCO period were similar between Medicaid and non‐Medicaid infants.
We estimated probit models and included an interaction term of the post‐CCO period and CCO infant indicators. This interaction effect represented a difference in mortality outcomes between Medicaid infants and non‐Medicaid infants, attributable to the CCO implementation. Standard errors were adjusted for mother clustering to acknowledge there were infants with the same mother during the study period. Because coefficients of interaction terms in probit models are misleading for the interpretation of interaction effects,35 average marginal effects of the interaction term with bootstrapped standard errors from 1000 repetitions were then computed to estimate the average change in predicted probabilities of mortality in Medicaid infants after the implementation of CCO. Logit models yielded almost identical goodness of fit but slightly larger DD estimates.
2.6.1. Heterogeneous effects of CCOs on mortality separately for preterm and full‐term infants
To examine heterogeneous effects of CCOs on mortality between preterm and full‐term infants, we estimated our models in each subgroup, that is, full‐term infants and preterm infants. Computed average marginal effects estimated the change in predicted mortality in each subgroup after the implementation of CCOs.
2.6.2. Impacts of CCOs on mortality over the implementation timeline
We compared pre‐CCO averages to post‐CCO averages by month to investigate the extent to which the impact of CCOs on mortality changed over the implementation timeline. We included a three‐way interaction term between the post‐CCO infant, CCO infant indicators, and the monthly time indicator for the length of CCO implementation. The coefficients of this interaction term represented the impacts of the CCO implementation on the monthly change in mortality among Medicaid infants. We estimated linear probability models in order to gauge the magnitude of the monthly change in mortality. Coefficients from linear probability models of the interaction term were the average monthly change in predicted probabilities of death among Medicaid infants after the implementation of CCOs.
2.6.3. Sensitivity analysis
Because CCOs have been shown to increase prenatal care that resulted in improved infant health outcomes,19, 28 improving prenatal care could be a pathway through which CCOs had an effect on mortality. Receiving adequacy of prenatal care could also be considered as a proxy of access to care or insurance status during pregnancy. Thus, we compared our main models with the models controlling for the indicator of adequacy of prenatal care to investigate whether CCOs had an impact on mortality through other pathways other than improving prenatal care for pregnant women. Adequacy of prenatal care was measured using the Kotelchuck index.36 The binary indicator of adequacy of prenatal care was defined as prenatal care initiation within the first 4 months of pregnancy and completing at least 80 percent of the American Congress of Obstetricians and Gynecologists (ACOG) recommended number of visits for the gestational age of the infant.37
Oregon expanded its Medicaid program in 2014 under the Affordable Care Act (ACA). Although children were not the broad focus of ACA Medicaid expansion, Medicaid expansion could have impacts on access to care for women before pregnancy and ultimately infant health outcomes. As mothers who were in the Medicaid expansion group could have different characteristics, we estimated the impacts of CCOs on mortality with the exclusion of infants whose mothers were in the Medicaid expansion group.
Although impacts of CCO implementation on mortality were adjusted for a comprehensive set of maternal and birth characteristics as well as time trends of the outcomes over the study period, our estimates might be biased if there were omitted variables that affected both the main exposure and mortality outcomes such as the quality of care received by Medicaid and non‐Medicaid mothers and infants. To mitigate this concern, we adjusted our estimates using propensity score weighting to weight the four groups, that is, Medicaid pre‐CCO, Medicaid post‐CCO, non‐Medicaid pre‐CCO, and non‐Medicaid post‐CCO, to be more similar based on observable characteristics. To estimate the propensity scores, we used the method proposed by Stuart et al38 and fitted a multinomial logistic regression for the group indicator as a function of all covariates included in the main models, smoking status during prepregnancy and pregnancy periods, and mother's county.
Analyses used Stata software, version 15.1. The study was approved by institutional review boards at the Oregon Health Authority and at Oregon State University.
3. RESULTS
3.1. Characteristics of the sample
Table 1 presents characteristics of our sample of infants. Mothers of Medicaid infants were younger, more likely to be Hispanic, and less likely be white or to be married and had lower education levels compared to mothers of non‐Medicaid babies. The proportion of infants living in rural residence was also higher in Medicaid infants. Mothers of Medicaid infants were more likely to smoke during pregnancy and less likely to receive adequate prenatal care during pregnancy. The proportion of preterm births was higher in Medicaid infants than non‐Medicaid infants.
Table 1.
Characteristics of infants
| Variable | Medicaida | Non‐Medicaidb | ||
|---|---|---|---|---|
|
Pre‐CCOc
(N = 77 850) (n/%) |
Post‐CCOc
(N = 88 683) (n/%) |
Pre‐CCO (N = 57 903) (n/%) |
Post‐CCO (N = 59 967) (n/%) |
|
| Maternal characteristics | ||||
| Age (mean [SD]) | 25.9 (5.8) | 27.0 (5.8) | 30.8 (5.0) | 31.4 (4.7) |
| Race/ethnicity | ||||
| White | 42 479 (54.6) | 52 497 (59.2) | 45 668 (78.9) | 47 363 (79.0) |
| Black | 2768 (3.6) | 3360 (3.8) | 654 (1.1) | 790 (1.3) |
| AIANd | 2526 (3.2) | 2842 (3.2) | 779 (1.4) | 755 (1.3) |
| Asian | 2144 (2.8) | 3054 (3.4) | 4771 (8.2) | 5402 (9.0) |
| NHPIe | 920 (1.2) | 1063 (1.2) | 285 (0.5) | 295 (0.5) |
| Other | 659 (0.9) | 737 (0.8) | 493 (0.9) | 538 (0.9) |
| Hispanic | 23 372 (30.0) | 23 224 (26.2) | 3987 (6.9) | 4018 (6.7) |
| Missing | 2982 (3.7) | 1906 (2.2) | 1266 (2.1) | 806 (1.3) |
| Education | ||||
| <High school | 23 520 (30.2) | 20 072 (22.6) | 1635 (2.8) | 1008 (1.7) |
| High school or GEDf | 24 099 (31.0) | 26 503 (29.9) | 6801 (11.8) | 5457 (9.1) |
| College or higher | 27 019 (34.7) | 39 960 (45.1) | 48 111 (83.1) | 52 566 (87.7) |
| Missing | 3212 (4.1) | 2148 (2.4) | 1356 (2.3) | 936 (1.5) |
| Married | 34 003 (43.7) | 40 322 (45.5) | 51 157 (88.4) | 53 925 (89.9) |
| Rurality | ||||
| Urban | 64 186 (82.5) | 73 202 (82.5) | 52 302 (90.3) | 53 818 (89.8) |
| Large rural | 7882 (10.1) | 9831 (11.1) | 3125 (5.4) | 3779 (6.3) |
| Small/isolated rural | 2679 (3.4) | 3259 (3.7) | 1230 (2.1) | 1140 (1.9) |
| Missing | 3103 (4.0) | 2391 (2.7) | 1246 (2.2) | 1230 (2.0) |
| BMI at deliveryg (mean [SD]) | 32.2 (6.5) | 32.4 (6.7) | 30.7 (5.7) | 30.8 (5.7) |
| Smoking during pregnancy | 14 420 (18.5) | 16 112 (18.2) | 2405 (4.2) | 1813 (3.0) |
| Number previous births | ||||
| 0 | 28 819 (37.0) | 31 991 (36.1) | 24 716 (42.7) | 26 274 (43.8) |
| 1 | 21 770 (28.0) | 26 013 (29.3) | 20 361 (35.2) | 21 217 (35.4) |
| 2 | 13 307 (17.1) | 15 732 (17.8) | 7665 (13.2) | 7732 (12.9) |
| ≥3 | 10 983 (14.1) | 13 172 (14.9) | 4004 (6.9) | 4097 (6.8) |
| Missing | 2971 (3.8) | 1777 (1.9) | 1157 (2.0) | 647 (1.1) |
| Adequacy of prenatal care | 52 195 (67.1) | 62 588 (70.6) | 46 831 (80.9) | 49 140 (81.9) |
| Birth characteristics | ||||
| Preterm | 6127 (7.9) | 6688 (7.5) | 4058 (7.0) | 4135 (6.9) |
| Girl | 38 008 (48.8) | 43 467 (49.0) | 28 198 (48.7) | 29 115 (48.5) |
| Presence of birth risk factors | 21 122 (27.1) | 27 368 (30.9) | 16 592 (28.7) | 19 119 (31.9) |
| Birth plurality | 1922 (2.5) | 2397 (2.7) | 2425 (4.2) | 2653 (4.4) |
Medicaid births defined as enrolled in Medicaid on date of birth or within 2 weeks of birth.
Non‐Medicaid status defined as not enrolled in Medicaid during the first year of birth.
Pre‐CCO infants were those born between August 2008 and July 2011, and post‐CCO infants were those born between August 2012 and December 2015.
American Indian or Alaska Native.
Native Hawaiian and Pacific Islander.
High school or General Education Development credential.
Body mass index at delivery.
Comparing post‐CCO infants to pre‐CCO infants, mothers of post‐CCO infants were older, were less likely to be Hispanic, had higher education level, and were more likely to be married. During the post‐CCO period, the proportion of mothers receiving adequate prenatal care increased for both Medicaid and non‐Medicaid infants, yet mothers of Medicaid infants experienced a larger increase, that is, from 67.1 to 70.6 percent, than mothers of non‐Medicaid infants, that is, from 80.9 to 81.9 percent. Maternal level of education changed between the pre‐CCO and post‐CCO periods among Medicaid mothers; that is, the proportion of mothers with “college or higher” education level increased from 34.7 percent in the pre‐CCO period to 45.1 percent in the post‐CCO period. In addition, mothers of Medicaid infants in the post‐CCO period were more likely to be white and less likely to be Hispanic compared to the pre‐CCO period. The percentage of preterm births was also lower by 0.4 percent during the post‐CCO period among Medicaid infants. However, post‐CCO infants, both Medicaid and non‐Medicaid, had higher proportion of the presence of birth risk factors.
3.2. Mortality
Table 2 presents the probability of mortality for our sample. Both neonatal mortality and infant mortality were lower among Medicaid infants. For instance, during the pre‐CCO period, infant mortality among Medicaid infants was slightly lower than that among non‐Medicaid infants, that is, 0.34 percent compared to 0.41 percent. After CCO implementation of the CCOs, the difference in infant mortality between Medicaid infants and non‐Medicaid infants was larger, that is, 0.22 percent in Medicaid infants compared to 0.39 percent in non‐Medicaid infants. Mortality decreased during the post‐CCO period but Medicaid infants experienced a larger decline in mortality. Among Medicaid infants, neonatal mortality decreased from 0.16 percent in the pre‐CCO period to 0.09 percent in the post‐CCO period. Meanwhile, the change in neonatal mortality was smaller in non‐Medicaid infants, that is, 0.35 percent in the pre‐CCO period to 0.33 percent in the post‐CCO period.
Table 2.
Mortality among infants
| Population | Mortality | Pre‐CCOa | Post‐CCOa | |||||
|---|---|---|---|---|---|---|---|---|
| All (N = 77 850) | Full‐term (N = 71 723) | Preterm (N = 6127) | All (N = 88 683) | Full‐term (N = 81 995) | Preterm (N = 6688) | |||
| Medicaidb | Neonatal mortality | n | 122 | 35 | 87 | 76 | 17 | 59 |
| % | 0.16 | 0.05 | 1.42 | 0.09 | 0.02 | 0.88 | ||
| Infant mortality | n | 266 | 134 | 132 | 193 | 104 | 89 | |
| % | 0.34 | 0.19 | 2.15 | 0.22 | 0.13 | 1.33 | ||
| All (N = 57 903) | Full‐term (N = 53 845) | Preterm (N = 4058) | All (N = 59 967) | Full‐term (N = 55 832) | Preterm (N = 4135) | |||
|---|---|---|---|---|---|---|---|---|
| Non‐Medicaidc | Neonatal mortality | n | 203 | 38 | 165 | 198 | 40 | 158 |
| % | 0.35 | 0.07 | 4.07 | 0.33 | 0.07 | 3.82 | ||
| Infant mortality | n | 235 | 55 | 180 | 233 | 58 | 175 | |
| % | 0.41 | 0.10 | 4.44 | 0.39 | 0.10 | 4.23 |
Pre‐CCO infants were those born between August 2008 and July 2011, and post‐CCO infants were those born between August 2012 and December 2015.
Medicaid births defined as enrolled in Medicaid on date of birth or within 2 weeks of birth.
Non‐Medicaid status defined as not enrolled in Medicaid during the first year of birth.
3.3. Impacts of CCO implementation on mortality
Coefficients of probit models are presented in Table 3. On average, infants enrolled in Medicaid had lower mortality than those who were not enrolled in Medicaid. Mortality during the post‐CCO period was higher but only statistically significant for infant mortality (P < .05). Our main interest is the marginal effects of the interaction term between CCO infant and post‐CCO period indicators presented in Table 4 that represent the magnitude of CCOs' impacts on mortality. Marginal effects of the interaction term in the main DD model were negative that suggested reduction in mortality among Medicaid infants after the CCO implementation. However, only the reduction in infant mortality was statistically significant. On average, infant mortality among those enrolled in Medicaid decreased by 0.20 percentage points (95% CI: −0.35; −0.05) during the post‐CCO period, equivalent to a 56 percent reduction from the pre‐CCO level.
Table 3.
Impacts of CCO implementation on neonatal mortality and infant mortality: Coefficients from difference‐in‐differences probit models
| Variables |
Impacts of CCOs on mortality Coefficients (SE) |
|
|---|---|---|
| Neonatal mortality | Infant mortality | |
| CCO (Medicaid) | −0.522*** (0.065) | −0.273*** (0.047) |
| Post‐CCO period | 0.145 (0.096) | 0.156* (0.076) |
| CCO × Post‐CCO period | −0.187** (0.071) | −0.131* (0.053) |
| Maternal characteristics | ||
| Age | −0.008 (*0.004) | −0.014*** (0.003) |
| Race/ethnicity (reference = white) | ||
| Black | 0.258** (0.087) | 0.120 (0.071) |
| AIAN | 0.092 (0.100) | 0.188** (0.065) |
| Asian | −0.105 (0.084) | −0.009 (0.064) |
| NHPI | 0.176 (0.152) | −0.004 (0.134) |
| Other | −0.115 (0.208) | −0.155 (0.174) |
| Hispanic | 0.072 (0.050) | −0.011 (0.040) |
| Education (reference = < high school) | ||
| High school or ED | −0.041 (0.056) | −0.012 (0.041) |
| College or higher | −0.160** (0.061) | −0.099* (0.045) |
| Married | −0.179*** (0.046) | −0.160*** (0.035) |
| Rurality (reference = urban) | ||
| Large rural | 0.007 (0.062) | 0.011 (0.045) |
| Small/isolated rural | −0.259 (0.136) | −0.037 (0.076) |
| BMI at delivery | −0.004 (0.003) | −0.0003 (0.002) |
| Smoking during pregnancy | 0.169** (0.052) | 0.235*** (0.037) |
| Number previous births (reference = 0) | ||
| 1 | −0.097* (0.038) | −0.007 (0.031) |
| 2 | −0.124* (0.055) | 0.010 (0.042) |
| ≥3 | −0.045 (0.060) | 0.086 (0.048) |
| Birth characteristics | ||
| Preterm | 1.270*** (0.040) | 1.035*** (0.029) |
| Girl | −0.028 (0.034) | −0.035 (0.026) |
| Presence of birth risk factors | 0.052 (0.039) | 0.043 (0.030) |
| Birth plurality | 0.144** (0.055) | 0.147** (0.048) |
| Constant | −2.392*** (0.153) | −2.281*** (0.115) |
P < .05.
P < .01.
P < .001.
Table 4.
Impacts of CCO implementation on neonatal mortality and infant mortality among infants enrolled in Medicaid: Marginal effects
| Type of analysis |
Impacts of CCOs on mortality among infants enrolled in Medicaid Marginal effects in percentage points (95% CI) |
|
|---|---|---|
| Neonatal mortality | Infant mortality | |
| Main difference‐in‐differences model | −0.173 (−0.348; 0.002) | −0.196* (−0.346; −0.047) |
| Sub‐group analysis | ||
| Preterm infants | −1.702 (−3.665; 0.261) | −2.109* (−4.007; −0.212) |
| Full‐term infants | −0.046 (−0.120; 0.029) | −0.068 (−0.140; 0.003) |
| Monthly change in mortality | −0.0005 (−0.002; 0.002) | −0.003* (−0.006; −0.001) |
All average marginal effects were computed with bootstrapped 95% CI from 1000 repetitions, except for models on monthly impacts of CCOs where marginal effects were from linear probability models. All models controlled for maternal and birth characteristics and time trend of mortality.
P < .05.
3.4. Heterogeneous effects of CCO implementation between preterm infants and full‐term infants
Table 4 also presents marginal effects from our models investigating whether there were heterogeneous effects of CCO implementation on mortality between preterm infants and full‐term infants. Mortality decreased in both groups after the CCO implementation; however, the decrease in mortality was statistically significant for only infant mortality among preterm infants (ie, −2.11 percentage points, 95% CI: −4.01; −0.21). Reduction in infant mortality was found to be greater among preterm infants compared to that of full‐term infants.
3.5. Impacts of CCOs over the implementation timeline
Marginal effects of models examining whether the impacts of CCOs increased over the implementation timeline are shown in the last row of Table 4. On average, infant mortality of Medicaid infants decreased by 0.003 percentage points (95% CI: −0.006; −0.001) for every month further from the start month of CCO implementation. Neonatal mortality also marginally decreased each month over the implementation timeline, but it was not statistically significant.
3.6. Sensitivity analysis
In the models which we controlled for the indicator of adequacy of prenatal care, the marginal effects of the interaction term between CCO infant and post‐CCO period indicators were very similar to those in the models without controlling for adequacy of prenatal care (Table 5), which implied that improving prenatal care was not the only pathway through which CCOs influenced mortality among Medicaid infants. After excluding infants whose mothers were in the Medicaid expansion group, the interaction effects were also similar to those in the main models except for one change that the impact of CCOs on infant mortality among full‐term infants became statistically significant (ie, −0.07 percentage points, 95% CI: −0.15; −0.001). Thus, our main estimates of CCOs' impacts were less likely to be biased by the possible effect of Medicaid expansion during the study period. In the models using propensity score weighting, the estimated effects of CCOs were in the same direction but not statistically significant at the 5 percent significance level (Table 5). For instance, infant mortality in those enrolled in Medicaid reduced by 0.25 percentage points but was only statistically significant at the 10 percent significance level (P = .07). We believe that this slight increase in standard errors, which seems a potential limitation of propensity score weighting estimator,39 does not change the implications of our findings substantially.
Table 5.
Sensitivity analysis: Marginal effects
| Type of analysis |
Impacts of CCOs on mortality among Medicaid infants Marginal effects in percentage points (95% CI) |
|
|---|---|---|
| Neonatal mortality | Infant mortality | |
| With controlling for adequacy of prenatal care | ||
| Main difference‐in‐differences model | −0.165 (−0.332; 0.002) | −0.186* (−0.327; −0.044) |
| Subgroup analysis | ||
| Preterm infants | −1.571 (−3.533; 0.391) | −1.932* (−3.806; −0.057) |
| Full‐term infants | −0.055 (−0.116; 0.007) | −0.083* (−0.152; −0.015) |
| Monthly change in mortality | −0.0004 (−0.002; 0.002) | −0.003* (−0.006; −0.001) |
| Excluding infants whose mothers were in Medicaid expansion groups | ||
| Main difference‐in‐differences model | −0.181 (−0.362; 0.0002) | −0.210** (−0.363; −0.058) |
| Subgroup analysis | ||
| Preterm infants | −1.694 (−3.717; 0.330) | −2.068* (−4.024; −0.112) |
| Full‐term infants | −0.052 (−0.133; 0.029) | −0.074* (−0.147; −0.001) |
| Monthly change in mortality | −0.001 (−0.003; 0.001) | −0.004*** (−0.007; −0.001) |
| Models with propensity score weighting | ||
| Main difference‐in‐differences model | −0.168 (−0.466; 0.129) | −0.253 (−0.532; 0.026) |
| Subgroup analysis | ||
| Preterm infants | −1.120 (−5.195; 2.955) | −1.795 (−5.579; 1.989) |
| Full‐term infants | −0.064 (−0.150; 0.021) | −0.095 (−0.225; 0.036) |
| Monthly change in mortality | −0.001 (−0.005; 0.003) | −0.003 (−0.008; 0.002) |
All average marginal effects were computed with bootstrapped 95% CI, except for models on monthly impacts of CCOs where marginal effects were from linear probability models. All models controlled for maternal and birth characteristics and time trend of mortality.
P < .05.
P < .01.
P < .001.
4. DISCUSSION
Our findings indicated that infant mortality declined among infants enrolled in Medicaid after 4 years of the implementation of the Oregon CCO model. Previous studies on the impacts of CCOs or Medicaid ACOs found favorable effects on infant mortality, but these findings were not statistically significant.27, 28 Several reasons may account for the differences between our findings and prior studies. First, it takes time for changes in a health care delivery system to affect health outcomes like mortality. Our study examined the effects on outcomes after 4 years of implementation of the CCO model. Second, in order to change mortality outcomes, Medicaid ACO models may require not only coordinated care but also the delivery of care that addresses the social determinants of health.27 With the ability to examine the effects of CCOs in a longer time frame, our findings suggest that the Oregon CCO model was effective in improving survival within a year after birth of infants enrolled in Medicaid.
We also found that the impacts of CCOs on infant mortality continuously increased over the implementation timeline. Similar to any care delivery system reform, each CCO requires time to develop and adapt its strategies and structures to improve health care and health outcomes for Medicaid enrollees.
Although our findings indicated a favorable impact of CCOs on neonatal mortality, they were not statistically significant. For the next phase of CCO implementation (2020‐2024), the Oregon Health Policy Board has designed multiple specific policies to improve child and family outcomes. For instance, each CCO will be required to implement new value‐based payments in maternity care and children's health care with contracted providers.40 Future research should investigate the extent to which CCOs will have an impact on neonatal mortality when CCOs implement more policies targeting children's health.
These findings coupled with those from previous studies provide evidence that CCOs are moving in the right direction to address infant health outcomes. CCOs have been shown to improve access to prenatal care for pregnant women19 and to reduce some adverse infant's health outcomes, such as low birthweight and abnormal conditions at birth28 that could contribute to a reduction in infant mortality. The CCO model could be considered as a unique Medicaid ACO model that provides integrated care for physical, mental, and dental health to improve care coordination. The CCO model uses quality incentive payments tied to preventive care and care coordination through patient‐centered primary care homes (PCPCHs), which may improve health care for women during the prepregnancy period and follow‐up care for infants after birth. Furthermore, CCOs receive capitated global budgets that allow for flexibility to provide care addressing social determinants of health and health equity that could positively impact children's health. For example, CCOs could use their global budget to pay for nontraditional health care expenses like housing or transportation.41 Also, CCOs are accountable for not only cost and quality of care for Medicaid beneficiaries but also the population health of the region they serve. Each CCO is governed by not only health care providers but also community members, and other stakeholders in local health systems, to ensure community's health needs are met.
Our study is the first, to our knowledge, to examine the impact of the Medicaid ACO models on mortality among preterm infants. We found that the CCO model had a greater impact on infant mortality in preterm infants enrolled in Medicaid than that in full‐term infants. One possible explanation for this result is that most CCOs initially focused on care coordination for high‐risk and high‐cost population.41 Thus, preterm infants could have benefited from improved follow‐up and coordinated care after birth.
Our study has some important limitations. First, although our models controlled for a comprehensive set of maternal and birth characteristics that are associated with mortality in infants, we were not able to control for other time‐variant variables, for example, health care utilization before and during pregnancy of mothers or health care utilization after birth of infants. Nonetheless, our doubly robust estimates from the propensity score–weighted DD models replicate the main findings, mitigating this potential concern. Second, without the availability of information on insurance status of the non‐Medicaid group, we were not able to estimate the effect of CCOs when comparing the Medicaid group separately with privately insured infants or uninsured infants. Also, reporting errors may be present for maternal and birth characteristics in birth certificates. However, such errors are considered as random and should not influence our findings. Finally, our study examined the overall effects of the CCO implementation on mortality. Because the CCO model has different features that could contribute to the reduction in mortality among infants as previously discussed, future study should examine how each specific characteristic of this care delivery model has an effect on mortality among infants enrolled in Medicaid.
5. CONCLUSION
The implementation of CCOs in Oregon was associated with a reduction in infant mortality among infants enrolled in Medicaid. This health care delivery reform also had a greater impact on improving survival among preterm infants compared to full‐term infants in the first year after birth. Future research should investigate whether the CCO implementation has impacts on other infants' health outcomes, especially health outcomes for other high‐risk infant groups.
CONFLICT OF INTEREST
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. “No Other Disclosures.”
Supporting information
ACKNOWLEDGMENTS
Joint Acknowledgment/Disclosure Statements: This work was supported by the National Center for Chronic Disease Prevention and Health Promotion (NCCDPP) of the Centers for Disease Control and Prevention under Award Number 1U01DP004783‐01 to S. Marie Harvey (PI) and Jeff Luck (PI). The content is solely the responsibility of the authors and does not necessarily represent the official views of the Centers for Disease Control and Prevention.
Bui LN, Yoon J, Harvey SM, Luck J. Coordinated Care Organizations and mortality among low‐income infants in Oregon. Health Serv Res. 2019;54:1193–1202. 10.1111/1475-6773.13228
REFERENCES
- 1. Jiang Y, Granja MR, Koball H. Basic facts about low‐income children: children under 3 years, 2015. 2017.
- 2. Currie J, Lin W. Chipping away at health: more on the relationship between income and child health. Health Aff. 2007;26(2):331‐344. [DOI] [PubMed] [Google Scholar]
- 3. Margolis P, Greenberg R, Keyes L, et al. Lower respiratory illness in infants and low socioeconomic status. Am J Public Health. 1992;82(8):1119‐1126. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Kehrer B, Wolin C. Impact of income maintenance on low birth weight: evidence from the Gary experiment. J Hum Resour. 1979;14(4):434‐462. [PubMed] [Google Scholar]
- 5. Singh GK, Kogan MD. Persistent socioeconomic disparities in infant, neonatal, and postneonatal mortality rates in the United States, 1969–2001. Pediatrics. 2007;119(4):E928‐E939. [DOI] [PubMed] [Google Scholar]
- 6. Olson ME, Diekema D, Elliott BA, Renier CM. Impact of income and income inequality on infant health outcomes in the United States. Pediatrics. 2010;126(6):1165‐1173. [DOI] [PubMed] [Google Scholar]
- 7. Currie J, Gruber J. Health insurance eligibility, utilization of medical care, and child health. Q J Econ. 1996;111(2):431. [Google Scholar]
- 8. Bhatt C, Beck‐Sagué C. Medicaid expansion and infant mortality in the United States. Am J Public Health. 2018;108(4):565‐567. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Moss NE, Carver K. The effect of WIC and Medicaid on infant mortality in the United States. Am J Public Health. 1998;88(9):1354. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Child Welfare League of America . Oregon's children 2017. 2017; https://www.cwla.org/wp-content/uploads/2017/03/OREGON.pdf. Accessed March 15, 2018.
- 11. National Center for Children in Poverty . Oregon demographics of low‐income children. 2017; http://www.nccp.org/profiles/OR_profile_6.html. Accessed March 15, 2018.
- 12. Oregon Health Authority . Oregon's health system transformation: annual update. Oregon Health Authority; Jan 2016. [Google Scholar]
- 13. John McConnell K, Marie Chang A, Cohen DJ, et al. Oregon's Medicaid transformation: an innovative approach to holding a health system accountable for spending growth. Healthcare. 2014;2(3):163‐167. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Howard SW, Bernell SL, Yoon J, Luck J, Ranit CM. Oregon's experiment in health care delivery and payment reform: coordinated care organizations replacing managed care. J Health Polit Policy Law. 2015;40(1):245‐255. [DOI] [PubMed] [Google Scholar]
- 15. Kushner J, Tracy K, Lind B, Renfro S, Rowland R, McConnell J. Evaluation on of Oregon's 2012–2017 Medicaid Waiver. Portland, OR: Oregon Health & Science University, Center for Health Systems Effectiveness; 2017. [Google Scholar]
- 16. Howard SW, Bernell SL, Jangho Y, Luck J. Oregon's coordinated care organizations: a promising and practical reform model. J Health Polit Policy Law. 2014;39(4):933‐940. [DOI] [PubMed] [Google Scholar]
- 17. Chang AM, Cohen DJ, McCarty D, Rieckmann T, Point‐Counterpoint M. Oregon's Medicaid transformation — observations on organizational structure and strategy. J Health Polit Policy Law. 2015;40(1):257‐264. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. McConnell KJ, Renfro S, Lindrooth RC, Cohen DJ, Wallace NT, Chernew ME. Oregon Medicaid reform and transition to global budgets were associated with reductions in expenditures. Health Aff (Millwood). 2017;36(3):451. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Oakley LP, Harvey SM, Yoon J, Luck J. Oregon's coordinated care organizations and their effect on prenatal care utilization among medicaid enrollees. Matern Child Health J. 2017;21(9):1784‐1789. [DOI] [PubMed] [Google Scholar]
- 20. McConnell KJ, Renfro S, Chan B, et al. Early performance in medicaid accountable care organizations: a comparison of Oregon and Colorado. JAMA Intern Med. 2017;177(4):538‐545. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Oregon Health Authority . Oregon health system transformation: CCO metrics 2017 final report. 2018.
- 22. Colorado Health Institute . The route to the RAEs: analyzing the next phase of Medicaid's Accountable Care Collaborative in Colorado. 2017.
- 23. Song PH, Xu WY, Chisolm DJ, et al. How does being part of a pediatric accountable care organization impact health service use for children with disabilities? Health Serv Res. 2019;54(5):1007‐1015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Kelleher KJ, Cooper J, Deans K, et al. Cost saving and quality of care in a pediatric accountable care organization. Pediatrics. 2015;135(3):E582‐E589. [DOI] [PubMed] [Google Scholar]
- 25. Christensen EW, Payne NR. Effect of attribution length on the use and cost of health care for a Pediatric Medicaid Accountable Care Organization. JAMA Pediatrics. 2016;170(2):148‐154. [DOI] [PubMed] [Google Scholar]
- 26. Perrin J, Zimmerman E, Hertz A, Johnson T, Merrill T, Smith D. Pediatric Accountable Care Organizations: insight from early adopters. Pediatrics. 2017;139(2):e20161840. [DOI] [PubMed] [Google Scholar]
- 27. Henke RM, Karaca Z, Gibson TB, et al. Medicaid Accountable Care Organizations and childbirth outcomes. Med Care Res Rev. 2019:1077558718823132. [DOI] [PubMed] [Google Scholar]
- 28. Harvey MS, Oakley LP, Yoon J, Luck J. Coordinated Care Organizations: neonatal and infant outcomes in Oregon. Med Care Res Rev. 2017;76(5):627‐642. [DOI] [PubMed] [Google Scholar]
- 29. Jarjour IT. Neurodevelopmental outcome after extreme prematurity: a review of the literature. Pediatr Neurol. 2015;52(2):143‐152. [DOI] [PubMed] [Google Scholar]
- 30. Mccourt MF, Griffin CM. Comprehensive primary care follow‐up for premature infants. J Pediatr Health Care. 2000;14(6):270‐279. [DOI] [PubMed] [Google Scholar]
- 31. Wade KC, Lorch SA, Bakewell‐Sachs S, Medoff‐Cooper B, Silber JH, Escobar GJ. Pediatric care for preterm infants after NICU discharge: high number of office visits and prescription medications. J Perinatol. 2008;28(10):696‐701. [DOI] [PubMed] [Google Scholar]
- 32. Oregon Health Authority . Oregon's health system transformation legislative update report. 2018.
- 33. Oregon Health Authority . Oregon health plan (MCO and CCO) administrative rulebook. In Programs HSDIH, ed 2017. [Google Scholar]
- 34. University of Washington . Rural‐urban community area codes (RUCAs). http://depts.washington.edu/uwruca/. Accessed March 15, 2018. [Google Scholar]
- 35. Ai C, Norton EC. Interaction terms in logit and probit models. Economics Letters. 2003;80(1):123‐129. [Google Scholar]
- 36. Kotelchuck M. An evaluation of the Kessner Adequacy of Prenatal Care Index and a proposed Adequacy of Prenatal Care Utilization Index. Am J Public Health. 1994;84(9):1414. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. American College of Obstetricians and Gynecologists. Committee on Professional Standards QACoO, Gynecologists . Standards for Obstetric‐Gynecologic Services, 6th edn Washington, DC: American College of Obstetricians and Gynecologists; 1985. [Google Scholar]
- 38. Stuart E, Huskamp H, Duckworth K, et al. Using propensity scores in difference‐in‐differences models to estimate the effects of a policy change. Health Serv Outcomes Res Method. 2014;14(4):166‐182. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Liu L, Hudgens MG, Becker‐Dreps S. On inverse probability‐weighted estimators in the presence of interference. Biometrika. 2016;103(4):829‐842. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Oregon Health Authority Health Policy & Analytics Division . CCO 2.0 Recommendations of the Oregon Health Policy Board. Portland, OR: Oregon Health Authority; 2018. [Google Scholar]
- 41. Smith J, Rissi J. Achieving the Triple Aim in Medicaid: Evaluating the Access, Quality, Health and Cost Impacts of Coordinated Care Organizations in Oregon. Oregon Health Authority; 2015. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials