 To determine the lifetime social value of using the guideline-recommended vaccines for children born in the United States in 2009. 

This study utilized an economic model with parameter values sourced from clinical and observational data, as well as the literature. 

The model quantified the health effects of routine vaccination for 14 diseases in terms of quality-adjusted life-years (QALYs) saved. The health effects were then valued by applying an economic value of a QALY. Producers’ profits were estimated using data on vaccine prices, profit margins, and the number of vaccines administrated in the 2009 US birth cohort. The costs of producing the vaccines were subtracted from the value of the health effects to yield the total social value of vaccination. The producers’ and consumers’ shares of this social value were calculated. Sensitivity analyses were conducted to determine how results depend on underlying parameter assumptions. 

Estimates indicated that vaccination of this cohort will save 1.2 million QALYs, relative to no vaccination. Of those health gains, 88% stemmed from reduced mortality and 12% from reduced morbidity. We estimated a social value of $184.1 billion from these gains, of which $3.4 billion accrues to manufacturers as profits, while $180.7 billion accrues to the rest of society. In sensitivity analysis, the total social value ranged from $40 billion to $675 billion, and the manufacturers’ share ranged from 0.3% to 11.5%. 

 Policy makers should account for this social value when considering policies affecting incentives to vaccinate and develop new vaccines.

		By preventing illness and premature deaths, vaccination of children born in the United States in 2009 will generate $184 billion in lifetime social value above the costs of the vaccines. 

		Because saving a child’s life yields many healthy life-years, the large majority (88%) of the health benefits of vaccines is due to avoided premature deaths rather than reduced morbidity (12%). 

		The high social value of vaccines has improved population health and provided economic benefit to multiple stakeholders, including patients, health plans, and vaccine manufacturers, whose profits in this cohort amount to approximately 2% ($3.2 billion) of the total social value.

The innovation of childhood vaccines has resulted in a decline in infectious disease, as well as gains in length and quality of life. Smallpox has been eradicated, poliomyelitis is nearly eliminated, and many other vaccine-preventable diseases have seen declines in incidence.

 Although adverse events (AEs) can occur with vaccines,

 and recent research has focused on their rising costs,

 the postvaccine era has seen life expectancy increase 15 to 25 years compared with the pre-vaccine era, and further gains are expected.

 Evidence suggests a large share of these survival gains is due to the control of infectious disease through vaccination.

When encouraged by public health policies, vaccination also provides a benefit to government and private payers by reducing overall costs and increasing population health. The CDC has cited evidence that common childhood vaccinations save over $5 in direct medical costs and effects on productivity for every $1 spent.

 Maciosek and colleagues found that preventative childhood immunization produced annual net medical savings of $267 per person.

 Vaccination also generates community (herd) immunity by reducing disease incidence and transmission, thus resulting in a healthier population.

Because vaccines have been successful at preventing disease, the public is no longer regularly confronted with many vaccine-preventable diseases, and the health and economic benefits of vaccination may be underappreciated.

 As childhood vaccines have reduced disease prevalence, real and perceived AEs of vaccination have become more salient to parents than the vaccine-targeted diseases.

 Consequently, vaccination rates in many US states have declined in recent years.

 As vaccination rates slip, the risk of new outbreaks increases.

Moreover, although consumers are insulated from the cost of many vaccines, vaccine cost is an important consideration for payers and providers and has been criticized.

 This focus on AEs and costs has obscured vaccines’ overall value to individuals and society. Previous research has yet to show how the total social value of vaccines is divided between innovators who develop these technologies and patients and the broader society who benefit from them. Therefore, in this study, we sought to measure the social value of childhood vaccines in the United States and the distribution of that value to manufacturers versus the rest of society.

The concept of social value of therapies and its distribution between manufacturers and patients has been described in other disease areas. For instance, Grabrowski et al found that statin usage resulted in a social value of $1.25 trillion, of which patients received 76%.

 Yin et al performed a similar analysis on tyrosine kinase inhibitors for the treatment of chronic myeloid leukemia and found a social value of $143 billion—90% of which was retained by patients.

 Recent gains in cancer survival have provided $1.9 trillion of additional social value, with 81% to 95% of that being retained by patients.

 Lastly, HIV/AIDS therapies have generated $1.38 trillion in social value, with 95% accruing to patients.

 Such analyses are net monetary benefit analyses, which is a common economic way of thinking about value which is distinct from cost-effectiveness analysis. The aim is to measure the total value a given health intervention generates for society, and how that value is distributed across patients and manufacturers.

This study applied similar methods to determine the social value of childhood vaccines for a birth cohort in the United States. Consistent with previous research, social value was defined from an economic perspective as the quantity of resources, in monetary terms, that society would be willing to give up in order to retain the health gains attributable to vaccines. Put another way, the overall social value of vaccines equals the aggregate value retained by consumers (above the actual payments for vaccines) plus the value retained by manufacturers (in the form of vaccine profits). We decomposed the social value into the shares accruing to manufacturers versus the rest of society. For infectious diseases, the social value includes not only those vaccinated, but also those not vaccinated who benefit from the reduction in disease incidence.

The study entailed constructing an economic model based on observational and clinical data. The model calculates the social value of the routine pediatric vaccination schedule used in the United States in 2009. We do so by quantifying the health effects of routine vaccination of children born in the United States in 2009. In particular, vaccines to prevent the following 14 diseases were considered: congenital rubella syndrome, diphtheria, haemophilus influenzae type b (Hib), hepatitis A, hepatitis B, measles, mumps, pertussis, pneumococcus-related diseases (including pneumococcal disease, otitis media, pneumonia, and meningitis), polio, rotavirus, rubella, tetanus, and varicella. The influenza vaccine was not included because its changing seasonal nature would have required different methods.

The social value was estimated by applying an economic value to the health effects of vaccines, measured in terms of quality-adjusted life-years (QALYs) saved through vaccination. QALYs take into account both duration and quality of life. A year in perfect health would be measured as 1 QALY, whereas death counts as 0. From the value of the QALYs gained, the costs to produce vaccines were subtracted. This yielded the social value—or in economic terms, the total surplus&mdash;of vaccines, and represents the economic value of the health gains from vaccines minus the resources society spent to produce them. The shares of the total surplus accruing to manufacturers (producer surplus) versus the rest of society (consumer surplus) were also calculated.

It should be noted that vaccine-preventable illnesses impose additional costs on society beyond the utility loss infected individuals experience, including caregiver utility loss and the use of special services for persistent disability. Therefore our estimate of the health value of vaccination should be considered a lower bound.

In addition, one should exercise caution in interpreting the results of this framework for rotavirus, since in industrialized countries like the United States, the costs of rotavirus are mainly hospitalization and caregiver utility loss, as rotavirus mortality and morbidity are lower in the industrialized setting. In contrast, many of the other studied vaccines target illnesses that imposed a high mortality and morbidity burden in the pre-vaccine era.

According to the CDC, 4,130,665 children were born in the United States in 2009.

 The health effects of vaccination in this cohort were estimated by combining data from the literature with life tables from the Human Mortality Database.

 From the literature, we obtained for each disease data on cases of illness prevented, premature deaths avoided, average age of onset, average age at death from the disease, average duration of disease, and utility loss. Specific parameter values and sources are available in the eAppendix (

). The survival benefits of vaccination were net of adverse reactions to vaccination. To obtain the economic value of a QALY, we considered values generated by revealed and stated preference studies.

 A mid-range value of $150,000 was used and varied in sensitivity analysis.

Estimating manufacturers’ profits required 3 types of data: 1) data on vaccine prices, 2) data on manufacturers’ profit margins, and 3) data on the number of vaccines administered in the 2009 US birth cohort. We obtained data on prices (available in the eAppendix) from the CDC Vaccines for Children Program website, which contains archived data on public and private vaccine prices from 2008 to 2015.

We obtained data on vaccine manufacturers’ profit margins from annual reports and financial statements. When measuring manufacturers’ profits, we used the gross profit margin, which represents the sales volume minus production costs. Obtaining a companywide average across the top 5 vaccine manufacturers produced an average gross profit of 75%.

 This is a conservative approach, as gross profits do not subtract out manufacturer research and development (R&D) and marketing expenses. By using the gross profit margin, we can view vaccines’ social value as society’s benefit from vaccination, and society’s investment in R&D as the cost of inventing and developing the vaccines. (Subtracting R&D from profits would negate this framing.) This framing is useful because investments should be undertaken when the benefits (ie, the return on investment) exceeds the cost; social value is an important part of this equation. Moreover, R&D costs include the costs of many failures that the innovator encountered on the way to the given successful product; there is not an established method for measuring R&D costs for vaccines.

To estimate the number of vaccines administered, we required data on vaccine coverage rates, dosage schedules, wastage, and the cohort size. Following previous work, we assumed that 53% of vaccine doses were publicly (vs privately) administered and the wastage rate—the rate at which additional vaccines must be purchased beyond those needed for each vaccinated child because some vaccines will be unused&mdash;was 5%.

 from the CDC. Doses administered between ages 0 and 18 were included, but the costs of any adult booster doses were excluded. Given that any adult booster doses occur many years into the child’s life, whereas the lives saved and illnesses avoided from vaccination are realized mainly in early childhood, the effect of the focus on childhood doses should be minimal.

The 3 analytic steps are described broadly below. Additional detail is provided in the eAppendix. Throughout the analysis, monetary values were inflation-adjusted to 2014 US dollars using the Consumer Price Index,

 and an annual discount rate of 3% was applied.

Step 1: Value Health Effects of Vaccination

The health effects of vaccines were calculated by summing the changes in morbidity and mortality among the 2009 US birth cohort due to vaccination. The mortality effects were calculated as the number of deaths averted from vaccination multiplied by the QALYs the typical child would lose from dying of the given disease (calculated as average life expectancy minus average age at death from the given disease times aged-adjusted utility). The morbidity effects were calculated as the number of cases of illness prevented through vaccination times the typical duration of illness times the disutility from the given illness. The health effects of vaccines in QALYs were then converted to economic terms by valuing each QALY at $150,000.

 This yielded the economic value of the health effects of vaccination.

Step 2: Estimate Manufacturers’ Profits

Vaccine manufacturers’ profits from a given vaccine were estimated by multiplying the vaccine’s price by the number of vaccines sold by the profit margin. Although vaccines typically consist of multiple doses, and vaccination rates vary by dose, in the profit calculations we assumed that all children who received the first dose of a vaccine would also receive all subsequent doses. This assumption overestimates profits because some children will not receive all doses, and manufacturers’ profits will be lower than they would be had these children received all of their doses.

Using the results of the previous 2 steps, we calculated consumer, producer, and total surplus (ie, social value). Consumer surplus was calculated as the value of the health effects of vaccines (from Step 1) minus the cost of the vaccines (from Step 2). Producer surplus equaled the manufacturer profits (from Step 2). Consumer and producer surplus together yielded the total surplus.

We performed analyses to test the sensitivity of the model to the parameters. Specifically, we varied all parameters by 

10%, except for the disease-specific vaccination rates, which were varied by 

 5% to avoid specifying rates over 100%. In addition, we also varied the value of a QALY from $50,000 through $250,000 to reflect the wide range of values in the literature, and varied the discount rate from 0% to 6%. Lastly, we conducted an analysis, which included the parents’ time cost to take the children to receive the vaccines, taking into account that multiple vaccines may be given in the same visit.

The health effects of vaccination are reported in 

. Compared with no vaccination, an estimated 1.2 million QALYs will be saved due to vaccination among children born in the United States in 2009, for a value of $185.2 billion. Because vaccines typically prevent deaths that would have occurred in childhood, these avoided deaths save a large number of QALYs, whereas QALY gains from avoided illness are more modest. Consequently, 88% of the health value of vaccines is due to avoided death compared with 12% due to avoided illness. Among vaccines, diphtheria, tetanus, and pertussis (DTaP/Tdap) will have the largest health value by an order of magnitude, at nearly 800,000 QALYs saved for a value of $119 billion. Other vaccines with large health values are pneumococcus-related diseases (154,000 QALYs, $23 billion); measles, mumps, and rubella (MMR) (135,000 QALYs, $19 billion); and hepatitis B (79,000 QALYs, $12 billion). The smallest health value will be for rotavirus, at 5264 QALYs and $790 million.

 reports estimates of manufacturers’ revenues, costs, gross and net profits by vaccines given to the 2009 US birth cohort. Estimated profits are the lowest for the hepatitis A vaccine ($109 million gross, $36 million net), and the highest for pneumococcus-related diseases ($1.0 billion gross, $350 million net). Across all 14 diseases, vaccines generate an estimated $4.5 billion in revenues, $1.1 billion in costs, $3.4 billion in gross profits, and $1.1 billion in net profits.

The total social value by vaccine and its distribution across manufacturers and consumers are reported in 

. Consumers’ share of value, or surplus, ranges from $10 million for rotavirus to $119 billion for DTaP/Tdap. Vaccines producing the greatest consumer surplus do not necessarily provide the largest profit, or producer surplus, to manufacturers. Total social value ranges from $595 million from rotavirus to $119 billion from DTaP/Tdap, with the full vaccination schedule generating $184 billion in social value (

), or $45,000 per child. Of that total social value, 1.8% accrues to manufacturers, whereas 98.2% accrues to the rest of society. The manufacturers’ share ranges from 0.3% for DTaP/Tdap to 98.3% for rotavirus, and is less than 15% for 7 out of the 9 vaccines.

In sensitivity analyses, we found that the model is most sensitive to the economic value of a QALY, the premature deaths prevented from vaccination, and the discount rate, which together contributed 98.1% of the variance in results (

). In simulations, the total social value ranged from $153 billion to $227 billion, and the manufacturers’ share ranged from 1.3% to 2.6%. In other words, varying parameters by 

10% maintained a social value of at least $37,172 per child, with at least 97% of the value retained as consumer surplus. In addition, varying the value of a QALY from $50,000 to $250,000 led to social value ranging from $61 billion to $308 billion, and varying the discount rate from 0% to 6% led to social value ranging from $447 billion to $107 billion. Subtracting parents’ time cost led to a social value of $182 billion and manufacture share of 1.8%.

Our estimates suggest that routine childhood vaccination produces a large social value. In particular, we found that vaccination will save 1.2 million QALYs among children born in the United States in 2009, relative to a counterfactual of no vaccination. Vaccines treating common diseases, such as DTaP/Tdap, pneumococcus, and MMR, generate a particularly large share of the health benefits.

Childhood illness is often transitory, but many childhood diseases have nontrivial fatality rates, and a death in childhood costs many years of healthy life. Therefore, it is not surprising that the bulk of these health gains (88%) stems from reduced mortality, with the remainder (12%) from reduced morbidity. Valuing each life-year at $150,000 and subtracting the production costs of vaccines, we found that vaccination of the 2009 birth cohort generated $184 billion in social value, or $45,000 per child. Of this, $3.4 billion (2%) accrues to manufacturers in the form of profits, while $180.7 billion (98%) is retained by the rest of society. Sensitivity analysis showed that the large social value and large share retained by society are robust to a wide range of plausible parameters.

The 2% manufacturers’ share of social value we found is smaller than that found in similar analyses of other health interventions. The manufacturers’ share of social value was between 5% to 24% in HIV/AIDS, chronic myeloid leukemia, and heart disease.

 The distribution of gains between manufacturers and the rest of society is determined by the cost of the intervention and the health gains it produces. Health gains will generally be larger in the context of diseases that are prevalent or take life early. Therefore, it is not surprising that childhood vaccines come out favorably in these comparisons, as they are among the most cost-effective health interventions

 and they preserve the health of children, who, in the absence of a devastating childhood illness, typically go on to lead a long, healthy life. To put the health results in context, in other industries, the producers’ share of total social value has ranged from 4% for minivans

The fact that 98% of the social value from existing childhood vaccines goes to children and their families is good news for access to current vaccines; however, it has longer-term implications for innovation. A 2% share of value suggests relatively weak incentives for the development of new and improved vaccines, relative to the incentives to develop cancer drugs, cholesterol-lowering drugs, or broadband Internet. However, there is debate in the literature about the exact manufacturer share of surplus required to optimally incentivize innovation, and therefore, although this study informs the debate, it does not determine optimal vaccine pricing.

Because the model relies on inputs from the literature, it is limited by their availability and quality. When exact parameters were not available, conservative assumptions were made. For example, costs and profits were overestimated by assuming children who receive 1 dose of a vaccine receive all subsequent doses, thus underestimating social value and overestimating manufacturers’ share. In cases in which illness is usually short and mild but can be severe and long-lasting (eg, polio), we assumed a short and mild case in our utility calculations. In calculating profit margins, we assumed the companywide profit margin applied to a given vaccine, although vaccines may comprise a less profitable division of pharmaceutical firms.

 We also used gross margin, which does not take R&D and marketing costs into account, rather than net margin, so that we could consider the social value as the return on society’s investment in developing the vaccine.

Additionally, while the 2009 birth cohort was selected for this analysis due to the availability of data on the health effects of vaccination,

 it has the drawback of including in the analysis a year of shortage of the Hib vaccine.

 However, given the scale of the findings of this study, this consideration likely had minimal impact.

Articles

POLICY

Vaccination of children born in the United States in 2009 will save 1.2 million quality-adjusted life-years, generating $184 billion in social value net of vaccination costs.

The Social Value of Childhood Vaccination in the United States

Although patient-centered care (PCC) was proclaimed a core health system aim in a 2001 Institute of Medicine report, it remains one of the most-used and least-understood terms in healthcare. We interviewed leaders at 15 Medicare accountable care organizations (ACOs) across the country that have been the most successful in putting patient-centeredness into actual practice to develop an operational definition. The ACOs we spoke with had a 3-pronged practical approach of: 1) patients as partners, 2) proactive customer-service orientation, and 3) care coordination with a whole-person approach. We believe this framework can serve as a guide as the healthcare system moves “from volume to value” and a true partnership becomes increasingly critical both to patients and the healthcare system as a whole.

	Patient-centered care (PCC), endorsed by the Institute of Medicine and emphasized in the Affordable Care Act, has become a reimbursement-linked priority for government and private payers. Despite CMS’ regulations on PCC, the concrete steps needed to implement it remain unclear and existing measures fall short. Clinicians and administrators at accountable care organizations (ACOs), which at the leading edge of healthcare innovation, are uniquely positioned to offer insight regarding lessons learned. In our interviews, leading ACOs shared valuable lessons on implementing PCC, which we have grouped into 3 categories: patients as partners, proactive customer service, and care coordination. This study helps to:

		Differentiate patient-centered from provider-centered activities.

Patient-centered care (PCC) is perhaps one of the most-used and least-understood terms in healthcare today. Although it was proclaimed a core health system aim in a 2001 Institute of Medicine (IOM) report,

 the term too often involves a series of aspirational statements about respect for patient values that give rise to widely different interpretations that are more “turn of phrase” than “term of art.”

Despite that ambiguity, Medicare accountable care organizations (ACOs) must find a way to turn ideal into real. The program includes a long list of regulatory patient-centeredness requirements that range from structural and clear (eg, place a beneficiary on the ACO governance board) to vaguely aspirational (eg, “patient-centeredness integrated into practice”). In the frontlines of care, however, a more practical viewpoint prevails.

Based on a literature review conducted by the authors and recommendations from an expert advisory panel, we interviewed clinicians and administrative leaders at 15 Medicare ACOs across the country that have been among the most successful in putting patient-centeredness into actual practice. Our primary purpose was to find 4 candidates for in-depth case studies; however, as we listened to these organizations closely, a disconnect between the patient-centered care theorists and the ACO implementers emerged.

The IOM, for instance, talks about responding to individual “preferences, needs, and values.”

 The ACOs we spoke to had a 3-pronged, practical approach: 1) engaging patients, as partners, with ACOs, thus ensuring that patients are activated, engaged in their care, and share decision-making power; 2) a proactive, customer-service orientation, making care convenient and accessible, and reaching out to patients—especially to those with chronic illness; and 3) individualized care coordination, with a whole-person orientation, tailored to the patient’s unique needs across settings and over time, with family and caregivers included, as appropriate. 

ACO leaders also added a caveat that cannot be found in any of the numerous definitions of patient-centered care, patient engagement, or the many other terms now in use: implementation requires an organization. That organizational context is important. Physicians acting alone will likely struggle to satisfy the PCC goals that are now routinely included in many value-based payment models. Organizational efforts to establish and enable such change are crucial.

For instance, the larger ACOs we interviewed made extensive use of sophisticated data analytics from the electronic health record system to identify chronic disease patients at risk, whereas smaller ACOs deployed nurse care coordinators and telephoned doctors and patients to accomplish the same task. Some ACOs trained physicians in motivational interviewing, and others directly measured patients’ motivation through the Patient Activation Measure, using that information to tailor messaging.

 Overall, turning a provider-centered organization into a patient-centered one means linking high-level leadership and goal-setting

 to practical actions, cumulatively meant to ensure that patient-centeredness would take root.

With the federal government pushing for 30% of all traditional Medicare payments to come from ACOs or similar alternative payment models by the end of this year, it is time to get practical. The first step is for governmental and private payers to stop constantly piling new requirements upon the old. Then, together, payers and providers need to examine far more closely what works in the frontlines of care. We believe learning from the approach of successful ACOs—having a framework of patients as partners, having a proactive customer-service orientation, and utilizing care coordination with a whole-person orientation&mdash;can guide the healthcare system as it moves “from volume to value” (

). During this transition period, a true partnership becomes increasingly critical to patients, doctors and to the healthcare system as a whole.

 Harvard Medical School (JSW, RSH), Boston, MA; Center for Surgery and Public Health, Brigham and Women’s Hospital (JSW, RSH), Boston, MA; Health Quality Advisors, LLC (MLM), Highland Park, IL; Northwestern University Feinberg School of Medicine (MLM), Chicago, IL; Institute of Public Communication, Università della Svizzera Italiana (RSH), Lugano, Switzerland; Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health (RSH), Baltimore, MD.

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

 Concept and design (JSW, MLM, RSH); acquisition of data (JSW, MLM, RSH); analysis and interpretation of data (JSW, MLM, RSH); drafting of the manuscript (MLM); critical revision of the manuscript for important intellectual content (JSW, MLM, RSH); obtaining funding (JSW, MLM).

 Joel S. Weissman, PhD, Center for Surgery and Public Health, Harvard Medical School, Brigham and Women’s Hospital, 1620 Tremont St, Suite 4-020, Boston, MA 02120. E-mail: Jweissman@partners.org.

1. Institute of Medicine (US) Committee on Quality of Health Care in America. 

Crossing the Quality Chasm: A New Health System for the 21st Century

. Washington, DC: National Academies Press; 2001.

2. Grob R. The heart of patient-centered care. 

. 2013;38(2):457-465. doi: 10.1215/03616878-1966406.

3. Hibbard JH, Stockard J, Mahoney ER, Tusler M. Development of the Patient Activation Measure (PAM): conceptualizing and measuring activation in patients and consumers. 

. 2004;39(4, pt 1):1005-1026. doi: 10.1111/j.1475-6773.2004.00269.x.

4. Shortell SM, Sehgal NJ, Bibi S, et al. An early assessment of accountable care organizations’ efforts to engage patients and their families. 

. 2015;72(5):580-604. doi: 10.1177/1077558715588874.&ensp;

WEB EXCLUSIVE

Understanding the perspectives of clinicians and administrators who are leaders in implementing patient-centeredness in accountable care organizations can help point the way toward implementation by others.

Patient-Centered Care: Turning the Rhetoric Into Reality

 To address concerns that expanding insurance coverage without expanding provider supply can lead to worse access for the previously insured, we examined whether previously insured Medicaid beneficiaries faced greater difficulties accessing primary care after statewide insurance expansion in Massachusetts.

 We used the Medicaid Analytic eXtract databases for Massachusetts and 3 New England control states for 2006 and 2009. We calculated rates of overall, acute, and chronic preventable admissions (or Prevention Quality Indicators [PQIs]) and a composite of control conditions for adults aged 21 to 64 years.

 We used multivariate Poisson regression models, adjusting for age, race, gender, reason for Medicaid eligibility, and state-level physician supply, as well as a difference-in-differences (DID) approach to compare the change in the rate of PQIs and control admissions in Massachusetts versus control states before and after health reform.

 Massachusetts and control states had increases in unadjusted rates of overall, acute, and chronic PQIs. When adjusting for age, race, gender, reason for eligibility, and physician supply, this increase persisted for overall and chronic PQIs in control states, with no significant difference in the relative increase between the 2 groups for any of the PQI measures. For the within-Massachusetts analysis, low-uptake counties had a significant increase in admission for chronic PQIs that was greater than that observed for high-uptake counties (+148.0 vs +36.0; 

 = .045 for DID). There was no significant DID for acute or overall PQIs between the 2 groups.

 We found no evidence that insurance expansion in Massachusetts, compared with control states, reduced access to primary care for vulnerable Medicaid beneficiaries.

	We found no evidence that insurance expansion in Massachusetts, compared with control states, reduced access to primary care for vulnerable Medicaid beneficiaries. Our findings indicate that: 

		Massachusetts saw no increase in preventable admissions relative to control states following health insurance expansion, and no greater increase in Prevention Quality Indicators (PQIs) within Massachusetts among those counties with high insurance uptake compared with counties with low insurance uptake. 

		In looking at PQI rates for African American beneficiaries—who typically have had greater barriers to accessing care&mdash;we saw no differential change between Massachusetts and control states. 

		The lack of a negative spillover effect in this study of vulnerable Medicaid beneficiaries should be reassuring to policy makers.

The Affordable Care Act (ACA) remains controversial 6 years after its passage.

 Although the law has increased the number of insured Americans,

 its impact on the broader healthcare system remains unclear. One important concern regarding insurance expansions is how such efforts might impact those Americans who were already insured. The reason for concern is straightforward: as insurance expansion brings new individuals into the healthcare system, the existing supply of providers—physicians, physician assistants, and nurses&mdash;may not increase proportionately. Therefore, being able to access providers in a timely fashion can become a challenge,

 not just for the newly insured, but also for the previously insured who are now competing for the same providers.

The predicted “crowd-out” phenomenon has been a key concern raised by the ACA’s critics. Examining Massachusetts, which underwent similar healthcare reform a decade ago, may be instructive in helping us understand what is likely to happen in the rest of the nation. Although some early data suggested that in Massachusetts there were delays to see primary care physicians following health reform,

 more careful analysis found that, at least for Medicare patients, there were no detrimental effects on access to quality outpatient care.

 Critics have countered that examining the effects of health reform in Massachusetts on older Americans has limited applicability, however, because older Medicare beneficiaries likely receive care from a different set of providers than those sought out by newly insured patients. Therefore, understanding what happens to previously insured patients who live in the same communities as the newly insured, and are thus likely see the same providers, will be far more instructive.

We focused on a group of patients who are highly likely to experience a “crowd-out” effect: already-insured Medicaid beneficiaries. These patients generally have poor access to healthcare services at baseline.

 Further, both Medicaid beneficiaries and the uninsured

 tend to have low incomes, include a greater share of racial and ethnic minorities,

 and are considered medically and financially vulnerable.

 These vulnerable individuals tend to live in the same communities, underlying the important role of safety net providers, who care for a disproportionate share of the uninsured and Medicaid beneficiaries. Safety net providers have seen an increase in demand for their services following Massachusetts health reform,

 suggesting that concerns about a “crowd-out” effect for already-insured Medicaid beneficiaries are well founded.

Given these concerns and the ongoing debate about the potential impact of Medicaid expansion on states’ existing healthcare services, we sought to answer 3 questions. First, did insurance expansion in Massachusetts, compared with control states, decrease access to effective primary care for Medicaid beneficiaries? Second, did communities in Massachusetts with the greatest uptake of insurance experience greater challenges with access to primary care than communities with a much smaller uptake? And finally, were the effects of “crowding out” particularly pronounced among previously insured African American Medicaid beneficiaries, who typically have had greater barriers to accessing care and higher rates of preventable hospitalizations?

 With these 3 questions, we sought to more comprehensively examine the impact of insurance expansion on the ability of previously insured Medicaid beneficiaries to access high-quality primary care services.

We used the Medicaid Analytic eXtract (MAX) databases for Massachusetts and 3 New England control states (Vermont, New Hampshire, and Connecticut). MAX datasets, maintained by CMS, are extracted from the Medicaid Statistical Information System. Final-action claims are used to create beneficiary-level data on demographics and Medicaid eligibility, as well as on utilization and payment for medical services. We used the Small Area Health Insurance Estimates provided by the US Census Bureau to estimate insurance rates within counties for our within-Massachusetts analysis. We used the Area Resource File to obtain county-level physician supply. Consistent with prior research, we used 2006 as the pre-health reform period,

 as this was when health reform legislation was passed. By 2007, key features of health reform had been implemented, and it was not until 2008 that the individual mandate penalty was implemented. We used 2009 as the post health reform period to allow for a “wash-out period” after health reform had been fully implemented.

We included nonelderly adults aged 21 to 64 years. Because claims information for Medicaid comprehensive managed care plans was unavailable in the Massachusetts MAX database for the study years,

 we limited our sample to those beneficiaries with Medicaid fee-for-service or Primary Care Case Management (PCCM) plans. PCCM plans require beneficiaries to choose a primary care provider who receives a monthly payment to coordinate care in addition to fee-for-service payment for medical services rendered. To ensure we had complete claims for our sample, we excluded those who were not eligible for the entire year, as well as those with concomitant Medicare coverage or private insurance. Additionally, we excluded beneficiaries who had a claim related to childbirth that year. We felt that the barriers to accessing care for the general Medicaid population may not be generalizable to those who became Medicaid-eligible on the basis of pregnancy. Finally, we identified eligible beneficiaries in both 2006 and 2009, including only those beneficiaries who had continuous Medicaid coverage in both years, as we sought to determine the outcomes for beneficiaries who were already insured prior to Medicaid expansion. After identifying beneficiaries who were eligible in both years, we analyzed their claims. We also obtained beneficiaries’ age, race, gender, and reason for Medicaid eligibility from the MAX database personal summary file. We collapsed the 26 subgroups of Medicaid-eligible populations in our sample into 3 main categories: disabled, adult, and other.

To measure access to primary care, we used Prevention Quality Indicators (PQIs), developed by the Agency for Healthcare Research and Quality, to measure potentially avoidable hospitalizations.

 We used PQIs that might be particularly sensitive to effective primary care access—including acute PQIs (urinary tract infection, bacterial pneumonia, and dehydration) and chronic PQIs (diabetes, chronic obstructive pulmonary disease, hypertension, heart failure, angina)&mdash;and overall PQIs

both before and after reform. We also looked at a composite marker for control conditions (acute myocardial infarction, stroke, hip fracture, and gastrointestinal bleeding) that have been used in prior research

 to capture conditions that are less likely to be affected by changes in primary care access. Our predictor of interest was exposure to health insurance reform.

We used multiple different strategies to isolate the impact of the reform program. Initially, we chose a control group of 3 New England states (Vermont, New Hampshire, and Connecticut) in which healthcare would be most similar to that of Massachusetts. Because Medicaid eligibility criteria vary substantially by state, we calculated the proportion of total beneficiaries in each state that met our inclusion criteria. Next, we conducted an analysis comparing rates of PQIs and control conditions among counties in Massachusetts with high versus low insurance uptake after health reform. If health reform had a detrimental effect, then we expected to see a larger increase in PQIs in the high-uptake counties compared with the low-uptake counties. Finally, we repeated this analysis for the control condition composite (total number of admissions for control conditions per 100,000 beneficiaries) in order to see whether changes in Massachusetts for the control composite were comparable with changes in the control states.

We calculated our primary outcome—the absolute number of acute, chronic and overall PQIs&mdash;for the pre- and post periods in Massachusetts and the control states. In our first analysis, we compared PQI rates among Medicaid patients in Massachusetts in 2006 (pre-reform) with PQI rates for the same population in 2009 (post reform). Since this simple comparison could be confounded by a time trend, our second and primary analysis used a difference-in-differences (DID) approach to compare the change in PQIs in Massachusetts with the change in PQIs in the control states. The analyses were carried out at the beneficiary-year level using longitudinal Poisson regression models, allowing for correlated pre-post measurements within each beneficiary. The outcome for each beneficiary was the number of PQIs in a given year, and the primary predictors were indicators for time (pre vs post), condition (Massachusetts vs control), and the interaction between time and condition. To try to further balance the comparison between Massachusetts and the control states, we allowed the 3 control states to have differing initial PQI rates and we adjusted for the following available patient characteristics: age, gender, race, reason for Medicaid eligibility, and county-level physician supply. The models were implemented using the Glimmix procedure in the SAS version 9.4 statistical package (SAS Institute Inc, Cary, North Carolina).

 we identified counties with rates of baseline health insurance below the median for the state as high potential effect counties. We then calculated the change in insurance rate before (2005-2006) and after (2007-2009) health reform, and identified those counties with rates of insurance uptake that were above the median as “high-uptake” counties. We compared the change in rates of PQIs before and after health reform in high- versus low-uptake counties using the same longitudinal Poisson regression model as above. In order to address our specific interest in “crowding out” in the African American population, we repeated the analysis comparing Massachusetts with control states looking specifically at the potentially more vulnerable African Americans.

In addition to the primary analysis using Poisson regression modeling, we also conducted all analyses using linear regression to investigate if our findings were sensitive to modeling choice. Results were considered significant at a 2-sided 

 value of less than .05. The Office of Human Research Administration at the Harvard T.H. Chan School of Public Health approved this study.

Our analytic sample consisted of 127,532 beneficiaries in Massachusetts with 16,437 inpatient discharges in 2006 and 19,603 inpatient discharges in 2009. Control states had a total of 53,925 beneficiaries, with 9607 inpatient discharges in 2006 and 11,299 inpatient discharges in 2009. The proportion of beneficiaries who met each of the inclusion criteria in each state is presented in eAppendix Table A (

). The mean number of inpatient stays per beneficiary was similar across states, with the exception of New Hampshire, which had more than double the rate of admissions per beneficiary in both years in comparison with Massachusetts (0.35 admission per beneficiary in New Hampshire in 2009 vs 0.15 admission per beneficiary in Massachusetts). The demographic sample characteristics of our sample population of Medicaid beneficiaries in Massachusetts and control states are presented in 

Overall, Massachusetts had a higher percentage of white Medicaid beneficiaries relative to control states before health reform (53.8% vs 51.4%; 

 <.0001). The median age was slightly, but statistically, higher in Massachusetts than in control states (39.1 vs 38.2 years; 

 <.0001). The proportion of male beneficiaries was 30.3% in Massachusetts and 25.1% in control states (

 <.0001). Massachusetts and control states differed with respect to the proportion of beneficiaries in each eligibility category. Massachusetts had a greater proportion of disabled beneficiaries (40.4% vs 35.2%; 

 <.0001) and of beneficiaries in the “other” category (41.0% vs 9.8%), which included primarily adults gaining coverage through Medicaid expansion waivers. Massachusetts had a lower proportion of beneficiaries (18.6% vs 55.0%) gaining coverage through the traditional “adult” eligibility categories (adult, adult poverty, unemployed adult). Mean total physician supply was 4.55 per 1000 population in Massachusetts and 3.36 per 1000 population in control states.

Preventable Hospitalization Rates in Massachusetts and Control States

We report unadjusted rates of PQIs by state in the pre- and postreform periods (eAppendix Table B). Massachusetts had a lower unadjusted rate of PQIs relative to control states in both years. Both Massachusetts and control states had increases in unadjusted rates of overall, acute, and chronic PQIs during the study period (

 and eAppendix Table B). When we adjusted for age, race, gender, reason for eligibility, and physician supply, this increase persisted for overall and chronic PQIs in both Massachusetts and the control group of states, with no significant difference in the relative increase between the 2 groups for any of the PQI measures (

). For example, among overall PQIs, Massachusetts had an increase of 73.6 admissions per 100,000 beneficiaries (from 557.4 to 631.0; 

 = .0049) whereas control states had an increase of 182.1 admissions per 100,000 beneficiaries (from 812.6 to 994.7; 

= .0003). For chronic PQIs, the increases in Massachusetts and control states were 50.9 (from 314.6 to 365.6; 

 = .0002) admissions per 100,000 beneficiaries, respectively. There were no significant changes after health reform in the rate of acute PQIs in either Massachusetts or control states. There were no significant differences between Massachusetts and control states in the relative change between the 2 groups on any of the PQI measures using Poisson regression analysis (Table 2).

For control conditions, control states saw a trend toward an increase (59.3 admissions, from 359.7 to 419.1; 

= .06) whereas Massachusetts saw a decrease (—19.1 admissions, from 305.2 to 286.1; 

 = .25), but it was not significant. The DID (—78.4; 95% CI, –148.1 to –8.8) was significant at 

 = .02. When we repeated the analyses using linear regression, the relative increase in PQIs was greater in control states than in Massachusetts for overall and chronic PQIs. For control conditions, Massachusetts saw a significant decrease, whereas control states saw a significant increase, with a significant DID between the two (eAppendix Table C).

Preventable Hospitalization Rates Within Massachusetts

We performed an analysis comparing PQI rates before and after health reform among counties in Massachusetts with a baseline insurance rate that was below the median (high-uptake counties) with those with a pre-reform insurance rate above the median (low-uptake counties) (

). Prior studies have found that counties with the lowest rates of health insurance at baseline had, unsurprisingly, the greatest new uptake of health insurance; therefore, we hypothesized, they would have the largest crowding-out effect. We found that both high- and low-uptake Massachusetts counties experienced an increase in overall PQIs with no significant differences in the rate of change between the 2 groups (+97.3 vs +157.0 admissions per 100,000 beneficiaries, respectively; 

 = .42 for DID). Only high-uptake counties saw a significant increase in acute PQIs (+62.8 admissions per 100,000 beneficiaries; 

= .03), but there was again no significant difference in the trend observed between the 2 groups (

= .16 for DID). For chronic PQIs, low-uptake counties had a significant increase that was greater than that observed for high-uptake counties (+148.0 vs +36.0; 

When we examined admissions for control conditions in high- and low-uptake counties (Table 3), overall admissions for control conditions decreased only in high-uptake counties (—56.4 admissions per 100,000 beneficiaries), but there was no significant differential change between the 2 groups (

= .16 for DID). A linear regression model produced similar findings (eAppendix Table D).

Preventable Hospitalization Rates by Race

Finally, because PQI rates differ by race, we looked specifically at rates of PQIs for African American beneficiaries in Massachusetts and in control states (

). For overall and chronic PQIs, Massachusetts and control states saw a significant increase. For overall PQIs, Massachusetts had an increase of 156.1 PQI admissions per 100,000 beneficiaries (from 575.2 to 731.3; 

 = .002), and control states had an increase of 445.5 PQI admissions per 100,000 beneficiaries (from 963.5 to 1409.0; 

 =.002). For chronic PQIs, Massachusetts saw a significant increase of 117.4 PQIs per 100,000 beneficiaries (from 355.8 to 473.2; 

 = .025), and control states saw an increase of 339.1 PQIs per 100,000 beneficiaries (from 689.9 to 1029.1; 

 = .004). For acute PQIs, neither Massachusetts nor control states had a significant increase in PQIs for African American beneficiaries. We found no significant DID between Massachusetts and control states in any admission rates (PQI or control conditions) for African American beneficiaries following Massachusetts health reform, using either Poisson regression (Table 4) or linear regression.

In our study of Medicaid beneficiaries, Massachusetts, relative to control states, saw no increase in preventable admissions following health insurance expansion. When we stratified counties within Massachusetts by their rates of insurance uptake, high-uptake counties saw no greater increase in PQIs relative to low-uptake counties. Finally, when we looked at PQI rates for African Americans, we saw no differential change between Massachusetts and control New England states. Taken together, our findings suggest that there was no evidence of crowd-out in access to primary care for Medicaid beneficiaries as a result of Medicaid expansion from Massachusetts health reform.

Massachusetts health reform is a model for national health insurance expansion, a key component of which is Medicaid expansion in many states. Although insurance expansion has clear benefits for the newly insured, one persistent concern has been that reform efforts could jeopardize the care of individuals with existing health insurance and provider relationships via a crowd-out effect. Yet, we found no evidence of impaired primary care access among Medicaid beneficiaries who traditionally have had the greatest barriers to obtaining care. Our sample of beneficiaries in Massachusetts did have an increase in PQIs, but this trend was also seen in control states that did not undergo health reform during this time period.

We hypothesized that those counties with the lowest baseline rates of health insurance (ie, those with the greatest new uptake) would experience the largest impact of health insurance expansion and would be most likely to experience any negative spillover effect. Although these high-uptake counties did have increases in preventable admissions overall and for acute conditions, they did not experience a differential increase relative to low-uptake counties. In fact, for chronic PQIs, high-uptake counties saw less of an increase than low-uptake counties. These results suggest that the healthcare infrastructure in these communities was able to accommodate the increased demand for services associated with insurance expansion.

Our study complements the findings of previous work indicating that Massachusetts health reform had no adverse effects on preventable admissions among the Medicare population.

 This lack of observable deleterious effects among the elderly served as initial evidence that fears of a negative spillover effect of insurance expansion may be unfounded. However, Medicaid beneficiaries are thought to have even greater barriers to care because of historically low reimbursement rates. Now, the lack of an observable negative spillover effect in 2 separate vulnerable populations, Medicare and Medicaid beneficiaries, suggests that Massachusetts health reform did not result in clinically significant decrements in access to high-quality care for the previously insured. This should be reassuring for policy makers.

Although there may not have been a large, overall negative impact of health reform on access for those previously insured, one worries that certain particularly vulnerable populations, such as racial minorities, might have been affected. We found that Massachusetts—relative to other New England control states&mdash;did not have a differential change in preventable admissions for African Americans. However, African Americans continued to have higher PQIs than the overall population across both Massachusetts and control states, suggesting that we need to continue to focus on reducing disparities.

There are several possible explanations for our findings, which suggest that Massachusetts health reform did not impair access to care for already-insured Medicaid beneficiaries. First, Massachusetts was unique relative to other states in that it already had a relatively low number of uninsured, compared with the national average. Despite reports of physician shortages in Massachusetts, the per capita supply of physicians was higher in Massachusetts compared with control states, as well. These factors may have made the state optimally suited to implement near-universal coverage without significantly overburdening the existing system. Although there were anecdotal reports of increased wait times to see a physician,

 there have been no peer-reviewed studies, to our knowledge, investigating wait times directly. It is also conceivable that healthcare providers and practices were able to appropriately triage the needs of new and existing patients so that any delays in care did not have any observable adverse outcomes.

Our study has a number of limitations. It did not include claims for enrollees in Medicaid Managed Care, as their claims were not available in the MAX database for Massachusetts nor for 2 out of 3 control states during our study period. Our study sample was limited to fee-for-service beneficiaries and those in PCCM plans; in Massachusetts, managed care plans tend to reimburse at higher rates than fee-for-service plans.

 Also, Medicaid managed care beneficiaries tend to have lower medical complexity than those enrolled in PCCM plans.

 Given that higher reimbursement rates for Medicaid are associated with improved care access,

 the de facto exclusion of this group from our study leaves us with an even more vulnerable population for which a negative spillover effect would be even more likely. Yet, despite studying the most vulnerable and medically complex Medicaid beneficiaries, we saw no negative spillover effect in access to outpatient care in Massachusetts, and we actually saw improvement in several of these measures.

Second, after applying our exclusion criteria, our sample consisted of less than 10% of the total beneficiaries in a given year. Given that Medicaid beneficiaries are a heterogeneous group with significant turnover,

 we felt that these exclusion criteria were necessary to have complete claims for nonelderly, nonpregnant adults. Medicaid programs also vary substantially across states in terms of eligibility criteria. For example, New Hampshire had significantly less generous income eligibility criteria for parents of dependent children (51%-56% of the federal poverty level) relative to the remaining 4 states (133%-191% of the federal poverty level) in both study years.

 Our finding that New Hampshire Medicaid beneficiaries had more than double the rate of inpatient hospitalizations relative to other New England states likely reflects a more impoverished beneficiary pool with greater medical and social needs.

Additionally, although PQIs are a well-validated set of metrics used as indicators of adequate primary care access, they are still an indirect measure of access to outpatient care. We did not look specifically at changes in outpatient utilization or wait times to see a physician that would mediate any observed increases in preventable admissions. Finally, another limitation noted by previous investigators is that Massachusetts had a high baseline rate of insurance relative to many other states even before insurance expansion. Thus, the experience in Massachusetts may not be generalizable to other states with lower rates of health insurance and/or physician supply. Yet, the consistent lack of negative spillover effects in several studies should be reassuring to policy makers implementing health reform in their jurisdictions.

We examined the effect of the first statewide health insurance expansion on access to care for Medicaid beneficiaries. Despite fears that an increase in the number of insured could lead to delays in primary care access, we found no evidence of a negative spillover effect in this particularly vulnerable population.

 Department of Emergency Medicine, Beth Israel Deaconess Medical Center (LGB), Boston, MA; Department of Health Policy and Management (LGB, TCT, JZ, AKJ) and Department of Biostatistics (EJO), Harvard T.H. Chan School of Public Health, Boston, MA; Department of Surgery (TCT) and Department of Medicine (AKJ), Division of General Internal Medicine, Brigham and Women’s Hospital, Boston, MA.

 Rx Foundation grant: “The Impact of Insurance Expansion on Medicaid Patients.”

 Concept and design (LB, AKJ, EJO, TCT, JZ); acquisition of data (AKJ, TCT, JZ); analysis and interpretation of data (LB, AKJ, EJO, TCT, JZ); drafting of the manuscript (LB, AKJ, TCT); critical revision of the manuscript for important intellectual content (LB, AKJ, EJO, TCT); statistical analysis (LB, EJO, JZ); obtaining funding (AKJ); administrative, technical, or logistic support (LB, JZ); and supervision (AKJ, EJO).

Although concerns remain that expanding insurance coverage may have a “crowding-out” effect, we saw no evidence of this for Medicaid beneficiaries in Massachusetts following statewide health reform.

January 2017

The Social Value of Childhood Vaccination in the United States

Patient-Centered Care: Turning the Rhetoric Into Reality

The Effect of Massachusetts Health Reform on Access to Care for Medicaid Beneficiaries

The Value of Survival Gains in Myelodysplastic Syndromes

Electronic Health Records and the Frequency of Diagnostic Test Orders

An Assessment of the CHIP/Medicaid Quality Measure for ADHD

The Challenge of Paying for Cost-Effective Cures

Alignment of Breast Cancer Screening Guidelines, Accountability Metrics, and Practice Patterns

Value-Based Payment in Implementing Evidence-Based Care: The Mental Health Integration Program in Washington State

An Expanded Portfolio of Survival Metrics for Assessing Anticancer Agents