Daniel Polsky, PhD

Access to high-quality opioid use disorder (OUD) treatment remains insufficient. Alternative payment models offer opportunities to expand treatment by aligning payment with evidence-based care.

Medications that are effective in treating OUD remain underutilized.

Physicians face several major challenges in provision of buprenorphine, including fragmented and insufficient financing.

Currently, reimbursement for buprenorphine treatment is largely fee-for-service, which may not incentivize evidence-based care. Payments that “bundle” reimbursement for buprenorphine and wraparound social services would incentivize evidence-based care delivery, facilitate matching between patients in need of care and available clinicians, and promote entry of providers with innovative models of care.

Despite the ongoing opioid epidemic in the United States, the 3 FDA-approved medications that are effective in treating opioid use disorder (OUD)—methadone, buprenorphine, and naltrexone&mdash;remain significantly underutilized. Current data suggest that less than 35% of people with OUD received treatment in the previous year.

 Barriers to access include a lack of belief in agonist treatment, stigma associated with addiction, and insufficient reimbursement.

 Additionally, efforts to scale up access to these drugs face considerable regulatory barriers. In particular, methadone can be dispensed only in heavily regulated opioid treatment program settings; in part due to these regulations, such programs are not widely accessible.

Buprenorphine has expanded access to medication for OUD since it was approved by the FDA in 2012, primarily because it can be prescribed and managed in office settings so the existing workforce of geographically distributed physicians can be leveraged. Still, access to high-quality OUD treatment has failed to meet needs. Clinicians continue to face a number of challenges to offering treatment, including fragmented and insufficient financing, system misalignment (eg, between primary care and addiction treatment), and a lack of education and institutional and psychosocial support. Alternative payment models may offer opportunities to increase the provision of high-quality OUD buprenorphine treatment by aligning payment with evidence-based care and increasing resources for clinicians.

Currently, insurance payment to clinicians who prescribe and manage buprenorphine treatment relies on standard fee-for-service (FFS) evaluation and management (E&M) codes. This model of payment inhibits innovation in treatment delivery in at least 3 ways. First, the FFS system does not support integrated care teams that might provide higher value, and it offers limited reimbursement for evidence-based wraparound social services that are recommended to accompany OUD treatment. Second, FFS payment may create incentives to avoid patients who are less likely to stabilize (eg, because the payment rate for the time-intensive induction and stabilization phases, relative to payment for the stable maintenance phase, does not match the cost differential between these visits). Third, E&M codes are not tied to quality of services delivered. In fact, the per-service nature of FFS reimbursement may incentivize high volume of well-reimbursed services, even if they are of low value or even inappropriate.

Given its dominance in terms of physician revenue, FFS is likely to remain the primary reimbursement mechanism for buprenorphine treatment in many smaller practices. This payment model has some appealing features for incentivizing buprenorphine treatment in the community, including contingency on service delivery, which facilitates monitoring by policy makers, clinicians, and payers to ensure that reimbursable services do indeed get delivered and to prevent underprovision. Where appropriate, the FFS system should be supported at payment levels needed to encourage buprenorphine provision. Larger practices, however, may be well poised to transition to an alternative payment model that incentivizes higher-value care, including coordinated therapeutic and behavioral services.

Bundled payment is an alternative payment model that bundles payment for multiple services that an individual may receive during an episode of care into a single fixed payment. This model has been used in settings such as joint replacement to compensate for a central procedure (eg, joint replacement surgery) and all other associated care (eg, physician services, hospital readmissions) over a certain period post discharge. In the case of OUD, a bundled payment model should cover buprenorphine treatment and complementary services. To promote engagement of new patients, the rate would be higher in the initial phases of treatment to appropriately reimburse providers for the higher costs of intake and starting medication.

From the perspective of the practice, a bundled payment would shift the focus from being solely on administration of buprenorphine to a coordinated set of high-value services, including but not limited to medicine. A clinic would organize around team-based provision of evidence-based care (ie, rather than fragmented, high-volume services) and be rewarded for achieving preidentified patient-level outcomes. Importantly, bundled payment may also facilitate matching between patients in need of care and available clinicians by offering resources and incentives (eg, higher payment for early phases of treatment) to engage new patients in treatment.

Improving linkages between patients in need and available providers would address a major challenge that has emerged in efforts to scale up OUD treatment. Currently, most primary care

practices are not set up to see patients in acute need outside of business hours or use resources such as peer support services to match these patients with available providers; evidence suggests that where such specialists exist, they are often paid too little.

 To engage new patients effectively, practices may need to hire intake specialists and develop relationships with emergency departments where patients frequently seek overdose treatment. The alignment of a bundled payment would create the incentives for these and other necessary investments.

Beyond an individual practice, bundled payments are more likely than FFS to promote rapid entry of providers with innovative models of care and can be targeted to areas that are particularly in need due to high OUD rates and/or provider shortages. There are, however, several potential challenges with introducing bundled payment into the OUD setting. Episode-based payment can introduce incentives for underprovision, in part because it is difficult to track individual services.

 This concern highlights the critical role of quality measurement and tracking of patient outcomes to ensure appropriate care. To ensure quality and deter underprovision of care, triggering requirements could be employed. These requirements could be structural, requiring practices to have certain capacities (eg, for counseling services); process oriented, where documentation, screening, treatment plans, and coordination are required; and/or (over time) outcome oriented, where compliance and utilization are tracked.

Clinicians and payers have been slower to adopt value-based payment in the field of addiction treatment relative to other clinical areas.

 Bundled payment may play an important role in overcoming the limitations inherent in paying for OUD treatment services through FFS E&M codes and may offer promise as a supplement to existing models. Of course, whether bundles and other alternative payment models for OUD treatment result in increased capacity of evidence-based care remains to be seen. Rapid evaluation of ongoing efforts by public and private insurers will be valuable to optimize payment policies that support maximizing access to evidence-based OUD treatment.

 Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health (DP, APS), Baltimore, MD; Johns Hopkins Carey School of Business (DP), Baltimore, MD; Shatterproof (SA), New York, NY.

 Arnold Ventures and the Center for Health Economics of Treatment Interventions for Substance Use Disorder, HCV, and HIV, a National Institute on Drug Abuse Center of Excellence (P30DA040500).

 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 (DP, APS, SA); drafting of the manuscript (DP, APS); critical revision of the manuscript for important intellectual content (DP, APS, SA); and obtaining funding (DP, SA).

 Aditi P. Sen, PhD, Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, 624 N Broadway, Baltimore, MD 21205. Email: asen@jhu.edu.

1. Jones CM, McCance-Katz EF. Co-occurring substance use and mental disorders among adults with opioid use disorder. 

. 2019;197:78-82. doi:10.1016/j.drugalcdep.2018.12.030

2. Huhn AS, Dunn KE. Why aren’t physicians prescribing more buprenorphine? 

. 2017;78:1-7. doi:10.1016/j.jsat.2017.04.005

3. National Academies of Sciences, Engineering, and Medicine. 

Medications for Opioid Use Disorder Save Lives

4. Myrick K, Del Vecchio P. Peer support services in the behavioral healthcare workforce: state of the field. 

. 2016;39(3):197-203. doi:10.1037/prj0000188

5. McGuire TG. Physician agency and payment for primary medical care. In: Glied S, Smith PC, eds. 

6. Schulman M, O’Brien J, Pierre-Wright M, Thomas-Henkel C. Exploring value-based payment to encourage substance use disorder treatment in primary care. Center for Health Care Strategies, Inc. June 2018. Accessed June 1, 2020. https://www.chcs.org/resource/exploring-value-based-payment-to-encourage-substance-use-disorder-treatment-in-primary-care/

7. McDowell MJ, Busch AB, Sen AP, et al. Participation in accountable care organizations among hospitals offering substance use disorder and mental health services. 

. 2018;69(11):1131-1134. doi:10.1176/appi.ps.201800248

Articles

Bundled payments offer promise as a supplement to existing payment models to promote scaling up of opioid use disorder treatment using buprenorphine.

Innovative Payment to Scale Up Access to Medications for Opioid Use Disorder

 To examine inaccuracies in health plan provider directories and consider whether the machine-readable (MR) formats required of provider directories in the health insurance exchanges are more accurate than conventional directories and have the potential to improve directory accuracy in the future.

 The descriptive study design included qualitative data collection through stakeholder interviews and quantitative data analysis and verification of provider data source accuracy from multiple sources.

 Four separate sources of provider data from 5 counties were captured and aggregated into an analytic database. Provider data were analyzed through text matching techniques and provider practice phone interviews. Additionally, we interviewed 21 stakeholders.

 In quantitative analysis, we found widespread inaccuracy in provider information across directory types. Provider directory phone numbers were more likely to align with Google data than with the directory for the same company’s health plans in other markets. It is vastly less expensive to aggregate data from MR files than from conventional directories, which suggests that MR files have potential to be cost-effectively leveraged for data quality improvements. In qualitative analysis, we found that interviewees perceived provider directories as inaccurate, but they differed in their perceptions of the severity of the problem. Interviewees who were familiar with MR directories understood their advantages over conventional directories.

 The MR provider directories are not more accurate than the conventional provider directories. However, there is strong reason to believe that MR technology can be leveraged to increase accuracy. Promising state- and vendor-led initiatives also have the potential to correct widespread provider directory inaccuracy.

Provider directories are widely inaccurate. Conventional Medicare Advantage directories are currently slightly more accurate than machine-readable (MR) health insurance exchange directories, and Google is more accurate than either directory type. Although MR directories are not more accurate than conventional directories, they can be leveraged to improve directory accuracy. However, this has not yet occurred.

The problems with provider information are more complex than the problems specific to directories.

Promising state- and vendor-led initiatives are underway and deserve further attention.

Without accurate provider information, plan members cannot navigate their plans successfully, regulators cannot ensure plans meet requirements, and researchers have no accurate source of provider information.

Enrollees in managed care plans expect reasonable access to healthcare providers, yet “reasonable access”—a standard mandated by the Affordable Care Act&mdash;is a vague term never defined in federal regulation. Regulators measure access differently, from national standards in Medicare Advantage (MA) plans to individualistic network access plans in several states. Once a provider network is judged adequate by the regulator, the plan’s provider directory is the document that enrollees use to find providers. However, information in provider directories is often incorrect. There are 3 types of errors: a listed “in-network” provider has errant information, a provider listed as in network is not, and a provider that is in network is omitted.

A recent report from CMS found that 52% of providers in MA provider directories included at least 1 inaccuracy.

 This MA error rate is consistent with those in exchange and California Medicaid directories.

 In the last few years, significant regulatory actions have been taken against a handful of health plans with patterns of inaccurate provider information. In the last year, lawsuits were filed against 3 health plans by consumers who allege that they selected their health plan based on false provider network information.

Health plans struggle with provider network accuracy for a number of reasons. Providers are frequently indifferent to keeping directories current: The unglamorous task of notifying health plans of changes is often delegated to junior office staff or no one at all. The patchwork of regulatory definitions and standards for directories is another problem. Although some promising state- and vendor-led initiatives are now underway (see 

]), regulators are uncoordinated on definitions, requirements, and oversight approaches.

Provider directories are commonly posted on the internet in PDF or other “flat file” formats that defy easy downloading, aggregation, or analysis. In 2013, the state of California began requiring machine-readable (MR) health information. A number of newly established state-run health insurance exchanges required the use of MR provider directories in 2014, and CMS picked up the requirement for the federally facilitated exchanges in 2016.

 By requiring health plans to post their provider directories in a common MR format, directories can be easily downloaded to assess network adequacy against regulatory standards or network breadth among competing plans. In addition, MR directories can also populate physician finder consumer tools and improve provider data accuracy by flagging cases in which a provider is inconsistently listed across data sources. Here, we present data on the accuracy of MR directories in comparison with other sources of provider information.

The study was conducted using quantitative and qualitative analyses. Data were collected and analyzed through quantitative research to compare accuracy among MR provider directories on the exchanges, traditional flat file provider directories in MA plans, and additional sources of provider information. Interviews were simultaneously conducted with relevant federal and state officials and key industry stakeholders to capture their knowledge of and experience using both types of provider directories. This study did not require institutional review board review and approval.

Our quantitative research questions were (1) “How do electronically available data sources vary? Are the same data reported across sources?” and (2) “How consistent are the provider data found in electronic sources with information gathered from phone validations conducted with provider offices?”

A secondary research question (“What are the differences in level of effort [time and cost] to aggregate MR files versus other data types to conduct analyses of provider networks?”) is discussed in 

 We selected 5 US counties across the country with insurance carriers that offer both MA and exchange health plans. We selected these counties based on geography, market penetration, and the presence of an insurer serving both the MA and exchange markets.

We downloaded and aggregated 4 separate sources of provider data—CMS’ National Plan and Provider Enumeration System (NPPES) file, conventional MA directories found on health plan websites and other online sources, exchange MR directory files, and Google Places&mdash;into a single analysis database (

MA file compilation required 5 manual processes to download into the analysis database. Data quality issues such as misspellings and small inconsistencies in addresses were solved manually. The exchange data were in a uniform and standard format, but we still experienced some data challenges, which were solved via code and automation with little manual intervention.

We compiled the data into an analysis database, removing inconsistencies (eg, spacing and capitalization) and assigning each reported element to a category (ie, address or phone number) for validation and analysis across data sources. The analysis database contained all known information.

Throughout the data aggregation process, we tracked time and effort in order to assess the differences among the data sources by recording the time it took staff to complete each step to move the data into the analysis database (see eAppendix B).

After all of the data were compiled into a single database, data were analyzed via queries and text matching. For the text matching, human coders examined text similarity and assigned them into categories of like and unlike data. Because providers can practice in multiple locations, our analysis focused on whether 2 data sources reported at least 1 similar value.

Finally, we validated the accuracy of the electronically reported elements based on phone interviews with provider offices. We drew a random sample of more than 50 unique providers per specialty based on the primary specialty reported in NPPES. We contacted the provider’s practice address to validate each reported element in the analysis database.

We conducted phone interviews with representatives of CMS, state regulators, health plans and trade associations, and other experts to determine whether these stakeholders believe that MR directories facilitate more accessible and reliable information than conventional directories. Interviews were conducted with 4 CMS officials from 2 different program components, 5 senior state regulators from 4 states in different regions of the United States, 6 staff from 3 health plans that offer plans in the health insurance exchanges and MA, 3 health plan trade association staff from 2 associations, and 3 vendors of provider data accuracy solutions. These numbers do not include 2 individuals who declined to be interviewed.

Participants were provided with discussion topics in advance of the interview. Interviewees were offered confidentiality in exchange for their candor but no other compensation for their time.

We found widespread inaccuracy in provider information across directory types. We also found that provider directories from insurers with MA and exchange plans did not report the same phone number 50% of the time and did not list the same address 31% of the time. As displayed in 

, provider directory addresses have a 30% inconsistency rate when a common provider is in the same company’s health plan networks across markets.

We also analyzed provider information accuracy through phone validation by calling provider offices. During these calls, we successfully validated 80% (2850 of 3562 calls) of information attributes attempted. As summarized in 

, although all data sources contained inaccuracies and differences were sometimes slight, 2 findings can be implied: (1) Google is more accurate than provider directories or the federal NPPES file for name, address, and phone number (statistically significant [

 <.01]); and (2) despite the advantages of MR directories, exchange provider directories are less accurate than conventional MA directories.

The higher inaccuracy rate of exchange provider directories was unexpected given the advantages of MR technology to improve accuracy. This is not necessarily an indictment of MR. The high inaccuracy rate of exchange provider directories is likely because MA plan sponsors (facing oversight from CMS) are investing resources in raising the accuracy of MA directories, whereas there is no equivalent pressure to improve directory accuracy in the exchanges.

Interviewees—whether they were from health plans, government agencies, or provider data vendors&mdash;all understood that provider directories are frequently inaccurate, but they differed in their perceptions of the problem and potential solutions. Interviewees who were familiar with MR directories understood their advantages over conventional directories, but only some acknowledged their potential value in improving accuracy. Contrasting responses are offered in 

The information contained in provider directories is inaccurate across information types and markets (ie, the inaccuracy rate of provider addresses ranges from 27% to 35%, and the inaccuracy rate of provider phone numbers ranges from 25% to 48%). Although the facts are straightforward, the reasons behind them are complex: As noted by interviewees, providers often treat maintaining current directory information as a low priority; there is a lack of consistent standards or a common data dictionary for provider information; there is no central, reliable information source (“source of truth”) against which to assert accuracy; and there is no harmonized federal strategy to address the problem. As argued by CMS in its 2020 Call Letter, health plans cannot solve this inaccuracy problem on their own.

As noted, promising initiatives are now underway in a few states, and a few vendors are now offering promising accuracy tools (see eAppendix A), but we will not know their results for years. In the interim, MR directories offer great advantages over conventional directories, including crowdsourcing to identify information that is likely erroneous and quick data aggregation for ongoing network analyses. MR directories are mandated in the health insurance exchanges and Medicaid, but not federally enforced. They are recommended as a best practice in MA, but there is no evidence that they are widely used by MA health plans.

 The loose regulation of MA directories contrasts with CMS’ affirmative regulation of this market in most other respects. CMS Deputy Administrator Demetrios Kouzoukas warned MA organizations about the need for directory accuracy at the May 10, 2018, CMS Medicare Advantage and Part D conference, but without recommending the use of MR directories.

 The lack of a harmonized position on MR provider directories across markets merits further consideration. CMS’ recent requirement for hospitals to post MR hospital pricing information “to further improve the public accessibility of charge information” demonstrates that it values the technology.

Our analysis focused on whether MR directories can result in more accurate information on network providers. It did not focus on whether MR directories can be used to lessen the instances in which in-network providers are omitted or out-of-network providers are included. More research is needed on these topics.

 For CMS and other entities requiring MR directories, it is incumbent to utilize the advantages of machine readability. CMS required exchange plans to invest in machine readability and endure a bumpy rollout, but it has not yet leveraged all of the considerable benefits of machine readability. The advantage of the technology can be utilized in a nonpunitive manner by having CMS and health plans partner to improve directory accuracy.

 Regulators and researchers should analyze state and vendor initiatives to improve provider data accuracy. California’s statewide provider network utility and New Hampshire’s use of claims data to determine actual provider network are particularly interesting initiatives. These and other potential long-term accuracy solutions are summarized in eAppendix A. They may subsume and surpass the advantages provided by MR directories.

 Federal policy makers should consider benefits of federal leadership in correcting provider directory inaccuracy. Our analysis suggests that Medicare’s NPPES file is less accurate than health plan provider directories. This required Medicare data source could be reimagined to become a source of provider information accuracy. More broadly, national and transmarket efficiencies could be realized by establishing a national data dictionary and requirements across markets. In this regard, initiatives by HHS’ Office of the National Coordinator for Health Information Technology and CMS merit watching, including a recent proposal to require MA and Medicaid health plans to make provider directories available in a common electronic format equivalent to MR.

MR directories offer significant advantages over conventional directories. As noted in eAppendix B, data can be downloaded at roughly $0.01 per provider from an MR directory compared with $2.15 per provider from a conventional directory. This efficiency makes provider network aggregation and comparisons feasible for the first time. This, in turn, powers the potential of MR directories to improve the transparency and accuracy of provider information. However, machine readability does not correct inaccuracies by itself. We found that MR exchange plan directories are slightly less accurate than conventional MA directories. This is likely because MA plans (facing oversight from CMS) are working to raise the accuracy of MA directories, whereas there is no equivalent pressure to improve directory accuracy in the exchanges.

The authors thank Timothy Riddle, MSMIS, MBA, of NORC; Kacey Stotler, MSW, of Faegre Baker Daniels Consulting; and Tricia Beckmann, JD, of Faegre Baker Daniels Consulting, for their work on this project.

 Faegre Baker Daniels Consulting (MA), Washington, DC; Department of Veterans Affairs (AF), West Roxbury, MA; Boston University (AF), Boston, MA; Harvard University (AF), Cambridge, MA; Leonard Davis Institute of Health Economics, University of Pennsylvania (DP), Philadelphia, PA; NORC at the University of Chicago (MKS), Bethesda, MD.

 Mr Adelberg received grant funding from The Commonwealth Fund and has spoken on this topic at meetings/conferences. The remaining 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 (MA, DP, MKS); acquisition of data (MA, MKS); analysis and interpretation of data (MA, AF, DP, MKS); drafting of the manuscript (MA, AF, MKS); critical revision of the manuscript for important intellectual content (MA, AF, DP, MKS); and obtaining funding (MA).

 Michael S. Adelberg, MA, MPP, Faegre Baker Daniels Consulting, 1050 K St, Washington, DC 20001. Email: Michael.Adelberg@faegrebd.com.

1. Online provider directory review report. CMS website. cms.gov/Medicare/Health-Plans/ManagedCareMarketing/Downloads/Provider_Directory_Review_Industry_Report_Year2_Final_1-19-18.pdf. Published January 19, 2018. Accessed May 1, 2018.

The authors examined the accuracy of provider directories and found widespread errors. Machine-readable directories are not more accurate than conventional directories, despite their advantages. A survey of promising initiatives to improve directory accuracy was also completed.

Improving Provider Directory Accuracy: Can Machine-Readable Directories Help?

The Affordable Care Act eliminated patient cost sharing for evidence-based preventive care, yet the impact of this policy on colonoscopy and mammography rates is unclear. We examined the elimination of cost sharing among small business beneficiaries of Humana, a large national insurer.

This was a retrospective interrupted time series analysis of whether the change in cost-sharing policy was associated with a change in screening utilization, using grandfathered plans as a comparison group.

 We compared beneficiaries in small business nongrandfathered plans that were required to eliminate cost sharing (intervention) with those in grandfathered plans that did not have to change cost sharing (control). There were 63,246 men and women aged 50 to 64 years eligible for colorectal cancer screening, and 30,802 women aged 50 to 64 years eligible for breast cancer screening. The primary outcome variables were rates of colonoscopy and mammography per person-month, with secondary analysis of colonoscopy rates coded as preventive only.

There was no significant change in the level or slope of colonoscopy and mammography utilization for intervention plans relative to the control plans. There was also no significant relevant change among those colonoscopies coded as preventive.

 The results suggest that the implementation of the policy is not having its intended effects, as cost sharing rates for colonoscopy and mammography did not change substantially, and utilization of colonoscopy and mammography changed little, following this new policy approach.

The Affordable Care Act eliminated patient cost sharing for evidence-based preventive care, yet the implementation of this policy in a large national insurer has not resulted in substantially higher utilization of colonoscopy and mammography.

Only those colonoscopies coded as preventive were eligible for the policy change, so many patients were still subject to cost sharing.

Many physicians and patients may have been unaware of cost sharing for this policy change prior to having the procedure done.

Additional measures are needed to increase screening rates, including waiving cost sharing for all eligible colonoscopies and making financial incentives more salient for patients.

The Affordable Care Act (ACA) of 2010 contains provisions for reducing financial barriers to receiving evidence-based preventive care. Specifically, it mandates that health insurance plans provide benefits for, and require no cost sharing for, services recommended with a rating of A or B by the US Preventive Services Task Force (USPSTF).

Despite aggressive efforts to increase breast and colorectal cancer screening in the past decade, screening rates remain suboptimal.

Across the United States, the rates of colorectal and breast cancer screening are 59% and 72%, respectively.

 The ACA policy attempts to increase these rates through first-dollar coverage of these screening tests.

Patient cost sharing has been shown in a variety of contexts to be an important driver of health service utilization,

 and higher cost sharing has been shown to reduce utilization of services, though it also has the potential downside of reducing both desirable and less desirable healthcare utilization.

 Concerns about the impact of higher cost sharing on cancer screening rates and preventive service utilization have been heightened by the proliferation of high-deductible insurance plans.

Per ACA policy, all new private insurance plans were required to eliminate cost sharing for such services starting on September 23, 2010, with Medicare plans facing similar requirements starting January 1, 2011. Commercial plans were designated as grandfathered if they existed in March 2010 and did not make any coverage changes. In this study, we examined the impact of the elimination of cost sharing for preventive services on colon and breast cancer screening rates among beneficiaries of Humana, a large national private health insurance plan, using grandfathered plans as a control group. We hypothesized that there would not be a dramatic impact of the policy on preventive health utilization due to the lack of knowledge about cost sharing and additional patient barriers to screening.

We analyzed de-identified data from Humana for all beneficiaries aged 50 to 64 years who were enrolled in small business plans at any time between October 2008 and May 2012. Member eligibility files included beneficiary coverage and benefit design files describing different characteristics of each plan, including grandfathered status.

Patient-level data obtained from the beneficiary file included age, gender, and plan type, along with census tract data for income, race, and education. Using Healthcare Effectiveness Data and Information Set (HEDIS) coding (

claims data were obtained for colorectal cancer screening (ie, colonoscopy, fecal occult blood testing, and flexible sigmoidoscopy) and breast cancer screening (ie, mammography). Colonoscopies were defined as “preventive” if associated with the designated diagnosis code (V16.0, V18.51, V76.41, V76.51; these codes are from a small subset of HEDIS codes as defined by Humana for compliance with the ACA policy).

Patients eligible for colorectal cancer screening included men and women aged 50 to 64 years, and those eligible for breast cancer screening included women aged 50 to 64 years, as per the USPSTF recommendations.

Patients 65 years or older were not included, as they would be eligible for Medicare.

The primary outcome variables were rates of colonoscopy and mammography per person-month, and the secondary outcome was rate of colonoscopy coded as preventive per person-month. We did not exclusively measure procedures used only for cancer screening. Likewise, we did not measure the proportion of the population that was up-to-date with screening, as the turnover of the insurer’s population limited ability to look back and fully capture prior use of these screening tests. Rather, we looked at the overall utilization rates without a lookback period to examine differential rates in screening as a result of the policy. Colonoscopy is the primary modality for colorectal cancer screening in this country,

and even when employed for the evaluation of symptoms, it serves as a diagnostic test for colorectal cancer. We also found predominant use of colonoscopy in our sample, with no increase in fecal occult blood tests over time and minimal flexible sigmoidoscopy use (data not shown).

Cost sharing refers to the out-of-pocket costs paid by the beneficiary, including copayment, deductible, and coinsurance for all the claims related to the procedure (physician and facility) billed on the same day. Given the nature of the policy reform, we examined the percentage of colonoscopies and mammograms with greater than zero cost sharing, but also looked at the mean cost sharing amount.

For the purposes of the analysis, time zero was the first renewal month for each plan after September 2010. Changes in benefit design for small business plans are typically made during the renewal month, which is plan-specific and is distributed throughout the year, so time zero ranged from October 2010 to September 2011 for the different plans (

). Small business plans were categorized as non-grandfathered or grandfathered, as denoted by the insurer, based on whether they were required by the policy to have no cost sharing for preventive procedures. Hereafter, we will refer to the non-grandfathered plans as the intervention group and the grandfathered plans as the control group.

We only included beneficiaries enrolled in plans that were in existence on March 2010, which was the time when grandfathered status was determined. To be included in the analysis, beneficiaries had to be continuously  enrolled for 2 years prior to the change in the policy (time zero) and for a minimum of 9 months afterward. This allowed us to analyze the impact of policy change on a consistent cohort over time. Members from 23,028 unique small business coverage plans from 48 states were enrolled. The 5 states with the most enrollees were Texas (20.6%), Illinois (15.9%), Kentucky (9.8%), Florida (8.5%), and Wisconsin (6.8%).

We assigned time zero to the renewal time during the policy change period to normalize the policy effect and assess overall trends for 24 months before and 9 months after time zero (ranging from month —24 to month 9). Thus, time zero and the months did not all occur at the same time for different plans. We first looked at the proportion of colonoscopy and mammography with greater than zero cost sharing and the degree to which this changed following implementation of the new policy. We then assessed trends in procedure utilization for the eligible cohort of beneficiaries who were continuously enrolled, demonstrating this graphically.

We used an aggregate-level “differenced” interrupted time series to examine whether the change in cost sharing policy was associated with a change in screening utilization for the small business population, since this was the group for which we could directly compare rates in plans that had benefit design changes (intervention group) and those that were grandfathered (control group). Using this design, we compared each intervention and control group with itself, before and after the policy change.

 This approach reduces potential biases from unmeasured variables and accounts for underlying changes in screening rates over time that are constant between both groups and might reflect general changes in clinical practice.

The intervention and control groups had similar characteristics (

), and individual designation to those groups was at the plan level. Education, race, and income data were obtained from the census tract data for the area of residence of the beneficiary. We assigned small business plans to the intervention or the control group based on designation by the insurer.

For this interrupted time series analysis, the outcome of interest was the aggregate difference between the rate of colonoscopy or mammography for the intervention and the control group. This was determined at each monthly time point by calculating the difference between the rates (per 1000 beneficiaries) of the intervention and the control group, which resulted in 33 data points. We utilized a generalized linear regression model, which included variables for month, post policy, and month post policy. A secondary analysis was performed looking only at colonoscopies coded as preventive. The “post policy” coefficient provides the level change (absolute difference) in percent screened between the study groups in the month immediately after the policy, and the coefficient of “month post policy” demonstrates any slope change after the policy. These 2 variables are the coefficients of interest, as they evaluate the changes in level and slope for colonoscopy and mammography rates, controlling for baseline trend. A positive coefficient represents a relative increase in the level or slope of monthly screening rate for members in the intervention plans compared with the control plans from pre- to post policy change.

All analyses were carried out using SAS software version 9.3 (SAS Institute, Cary, North Carolina). Approval for this study was obtained from the Institutional Review Board of the University of Pennsylvania. Humana was involved in the design and conduct of the study, in the collection and interpretation of the data, and in the review of the manuscript.

For colonoscopy and mammography, the intervention and control groups had similar demographic characteristics including age, gender, education, race, and income (Table 1). There were 25,926 members in the intervention group and 37,320 members in the control group for colonoscopy; both groups were 49% female. In the colonoscopy intervention and control groups, respectively, mean ages were 55.2 years and 55.3 years; percentages with greater than high school education were 73% and 74%; percentages living in areas with greater than 80% white population were 71% and 72%; and median incomes for census tract were $68,700 and $70,200. There were 12,603 members in the intervention group and 18,199 members in the control group for mammography. In the mammography intervention and control groups, respectively, mean ages were 55.2 and 55.3 years; the percentage with greater than high school education were 73% and 74%; the percentage living in areas with greater than 80% white population was 70% and 72%; and median incomes for census tract were $68,100 and $69,600.

Approximately 75% of small business plan colonoscopies required cost sharing, for an overall mean amount of about $650, before the policy change for both intervention and control groups. After the change, the percentage of intervention group procedures with non-zero cost sharing dropped from 75% to 45%, while the control group rate of procedures with non-zero cost sharing remained at 75% (

). Post policy, the mean cost sharing remained at $650 for the control and dropped to about $350 in the intervention group. Among colonoscopies that were coded as preventive, there was a larger reduction in the proportion of procedures with non-zero cost sharing (

), with a drop in intervention group plans from 65% to approximately 15%. Before the policy, preventive coded colonoscopies had a mean cost sharing of about $600 for intervention and control, which dropped to about $100 for the intervention group post policy.

Many insurance plans had already made efforts to reduce cost sharing for mammograms prior to the policy change, so only about 10% of procedures in the intervention and control group had cost sharing before and after the policy change (

There was not a consistent increase in colonoscopy utilization following the reduction in patient cost sharing for colonoscopies in 2010 (

). Interrupted time series regression analysis examining the aggregate difference in colonoscopy screening rates between the intervention and control groups showed no significant change in level immediately after the policy change (coefficient, 0.31; 

= .68) or in the slope change in utilization (coefficient, 0.054; 

). We also performed an analysis of colonoscopies coded as preventive only and found no significant relative change in the absolute difference (coefficient, 0.16; 

= .78) or slope change in utilization (coefficient, 0.091; 

= .32) after the policy change (Table 2).

The annual utilization rate of mammography also changed little after the policy change (

). For mammography, comparing the aggregate difference in small business cohort screening rates showed a nonsignificant change in the level immediately after the policy change (coefficient, 1.24; 

= .65) and in the slope change in utilization (coefficient, 0.093; P = .83) (Table 2). Thus,

similar to the limited impact on colonoscopy, there was no significant increase in mammogram utilization after implementation of the policy change in this population.

This is the first study to examine the impact of the ACA cost sharing elimination policy on colonoscopy and mammography use in an insured population. Utilization of colonoscopy and mammography changed little following the implementation of this new policy. A potential, and perhaps unanticipated, reason for the limited effect is that only preventive-coded colonoscopies were subject to the policy. About 60% of eligible colonoscopies in small business plans were coded as preventive following the policy implementation. Nevertheless, there was a significant reduction in the proportion of colonoscopies with cost sharing for the intervention group, while the proportion of those with cost sharing in the control group remained the same. Furthermore, even among the subset of colonoscopies coded as preventive, for which the policy is specifically directed, we did not find a significant increase in colonoscopy utilization.

Mammography rates were also minimally affected, since few mammograms required cost sharing before the policy change, so only a small proportion of beneficiaries were impacted. Meanwhile, the percentage of mammograms with cost sharing did not drop to zero because some were coded as diagnostic as opposed to preventive. This reflects aggressive efforts to improve breast cancer screening rates that predated the ACA policy and limited the potential impact of this legislation on mammography screening rates. There was also the possibility that the ACA policy might draw attention to preexisting policies that required no cost sharing for certain preventive services and drive an increase in behavior, yet we did not find evidence of this.

Another potential reason for the limited initial impact is that patients and their physicians were unaware of these co-payment reductions for a relatively infrequent screening test.

Patients are given feedback of medication cost sharing every time they refill, while cancer screening is a relatively rare event—particularly in the case of routine screening colonoscopy every 10 years. This plan policy change was communicated during annual renewal, which supplemented existing efforts by Humana to remind beneficiaries about screening through automated phone calls, e-mail, and direct mail. However, insurance benefit design is typically complicated, and patients can have difficulty accurately estimating out-of-pocket healthcare payments.

Additionally, if a polyp was removed or if a procedure is coded for any symptoms (eg, change in bowel habits, bleeding), such colonoscopies are typically billed as diagnostic rather than preventive, and are thereby subject to cost sharing. Even if patients were informed of the ACA policy change for preventive care, it would have been difficult for them to predict whether their procedure would be coded as preventive or diagnostic. After this study period, there was a clarification by the government in 2013 which stated that private insurers may not require cost sharing for colonoscopy with polyp removal performed as a screening procedure, but this does not apply to the Medicare population.

A third reason for the limited initial impact may be an asymmetry in response to cost sharing changes. Value-based insurance designs that reduce cost sharing for high-value medications have resulted in small increases in adherence relative to the decrements that have been observed following co-payment increases.

 We had expected that patients would respond less to changes in the degree of cost sharing for a colonoscopy than for other healthcare interventions, due to the invasive and unpleasant nature of the test. There also may be additional costs to colonoscopy, such as time off from work and the requirement for an escort, that could be calculated by the patient.

The main limitation of this study is that it looked at the post policy time period for 9 months, and it may take longer for intended effects to manifest. However, we did apply the same analysis to the smaller cohort of beneficiaries for whom we had 12 and 15 months of post policy data and still found the same nonsignificant results. Additionally, there was substantial room for improvement, as publicly reported Humana HEDIS rates in 2010 were 63.7% and 76.9% for colorectal cancer and breast cancer screening, respectively.

Another limitation is that there was no way to take into account the appropriateness of the test or prior screening utilization for both colonoscopy and mammography. Because of insurance company turnover, requiring an additional 1 year of prior enrollment included only 62%, and 2 years would have included only 39%, of the insured population. This is particularly challenging in assessing colorectal cancer screening rates, since screening prior colonoscopy may have occurred up to 10 years before. However, assessing utilization within different time windows should not be biased differentially by our approach, and an analysis of only colonoscopies coded as preventive showed similar results.

Another potential limitation is that there may have been changes in coding practice for these procedures, but it is likely to have been consistent for the 2 study groups. There also may have been additional changes in benefit design for the intervention group, such as reduction of cost sharing for primary care visits, but they would be likely to bias toward a significant finding. We also looked at whether there was a difference in distribution of plan renewal months for the intervention and control groups, and found a slight increase in intervention renewal in the later months (

). Assuming that it may take time for the policy to take hold, this difference would also bias toward a greater impact of the policy on the intervention group. Finally, we studied small business plans in 1 insurer, but with beneficiaries in 48 states, these results should be generalizable to other commercially insured populations. Additionally, this population offered the unique advantages of having a natural control group and relatively higher cost sharing than other plans before the policy change.

While the waiver of cost sharing for preventive services was introduced with the goal of increasing cancer screening rates, the implementation of the policy to eliminate cost sharing has not increased rates of colonoscopy and mammography to a significant degree. Consideration should be given to waiving cost sharing for all colonoscopies and mammograms in the eligible population as opposed to just those that are coded as preventive, as many of these contribute to the screening and surveillance process. Additional communication and education about the policy alone may not be sufficient. Interventions need to be developed and studied that could lower the price of a colonoscopy below zero to offset the time and hassle costs of the procedure, make the process of obtaining screening easier, or address other patient barriers to screening.

Feifei Yang contributed to the preliminary analysis of this study.

Department of Medicine (SJM, DP, JDL, KGV) and Center for Clinical Epidemiology and Biostatistics and Department of Biostatistics and Epidemiology (JDL), Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA; Center for Health Incentives and Behavioral Economics, Leonard Davis Institute of Health Economics (SJM, DP, JZ, JDL, JTK, GL, KGV), University of Pennsylvania, Philadelphia, PA; Penn Medicine Center for Health Care Innovation (SJM, KGV), Philadelphia, PA; Department of Health Care Management, Wharton School (JTK, KGV), University of Pennsylvania, Philadelphia, PA; Department of Social and Decision Sciences, Carnegie Mellon University (GL), Pittsburgh, PA; Center for Health Equity Research and Promotion, Philadelphia Veterans Affairs Medical Center (KGV), Philadelphia, PA.

Humana, National Institutes of Health (K24 DK078228, T32 DK007740).

 Drs Mehta and Volpp received a grant from Humana for this research study. Dr Volpp is a consultant for VAL Health and CVS Caremark, and has received grants from Humana, CVS Caremark, and Discovery. The remaining 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 (SJM DP, JDL, JTK, GL, KGV); acquisition of data (SJM, KGV); analysis and interpretation of data (SJM, DP, JZ, JDL, KGV); drafting of the manuscript (SJM, KGV); critical revision of the manuscript for important intellectual content (SJM, DP, JZ, JDL, JTK, GL, KGV); statistical analysis (SJM, JZ, KGV); obtaining funding (SJM, GL, JTK, KGV); administrative, technical, or logistic support (SJM); and supervision (SJM, KGV).

Shivan J. Mehta, MD, MBA, Perelman School of Medicine, University of Pennsylvania, 1318 Blockley Hall, 423 Guardian Dr, Philadelphia, PA 19104. E-mail: 

1. National Federation of Independent Business v Sebelius, 567 US (11th Cir 2012). HHS website. 

http://www.hhs.gov/healthcare/rights/law/patient-protection.pdf

2. Centers for Disease Control and Prevention. Vital signs: colorectal cancer screening among adults aged 50-75 years - United States, 2008. 

3. Joseph DA, King JB, Miller JW, Richardson LC; CDC. Prevalence of colorectal cancer screening among adults--Behavioral Risk Factor Surveillance System, United States, 2010. 

4. Miller JW, King JB, Joseph DA, Richardson LC; CDC. Breast cancer screening among adult women--Behavioral Risk Factor Surveillance System, United States, 2010. 

5. Newhouse JP; Insurance Experiment Group. Free for All? Lessons From the Rand Health Insurance Experiment. Cambridge, MA: Harvard University Press; 1993.

The ACA eliminated patient cost sharing for evidence-based preventive care, yet this policy has not resulted in substantial increases in colonoscopy and mammography utilization.

ACA-Mandated Elimination of Cost Sharing for Preventive Screening Has Had Limited Early Impact

To estimate incidence rates and quantify excess medical and productivity cost of community-acquired pneumonia (CAP) in a commercially insured, working-age population.

 Using the Thomson Reuters MarketScan Commercial Claims and Encounters Database, we estimated the annual incidence of CAP from 2003 through 2007 and evaluated its excess direct medical and productivity costs due to short-term disability and workplace absenteeism for adult patients aged 18 to 64 years. A cohort of CAP patients was 1:3 propensity score matched to a control cohort without pneumonia. Both excess direct medical costs and excess productivity costs were estimated in multivariate ordinary least squares (OLS) and generalized linear model (GLM) adjusting for demographic and clinical characteristics.

 A total of 402,831 patients with CAP and 1,208,231 matched controls were included. Overall annual CAP incidence was 4.89 cases per 1000 person-years. Excess annual medical cost of CAP ranged from $7220 by OLS to $11,443 by GLM. Inpatient costs, outpatient costs, and pharmacy costs explained 34%, 51%, and 15% of this excess cost of the GLM models, respectively. CAP patients had an additional productivity cost of $2391. Among adults younger than 65 years, these estimates imply an incidence of CAP of 950,000 annual cases at a cost of $10.6 billion (range: $9.04-$13.1), of which 80% are direct medical costs and 20% are productivity costs.

 CAP is a frequent and costly event in a working-age population with a national cost of $10.6 billion. Interventions that could successfully prevent CAP could have a significant impact on healthcare costs and productivity.

This study estimated the burden and incidence of community-acquired pneumonia (CAP) among non-elderly adults (younger than 65 years) in the United States and determined the following:

		 CAP is common among non-elderly adults (4.89 cases per 1000 person-years). There are approximately 950,000 cases of CAP among non-elderly adults annually.

		 CAP costs approximately $10.6 billion annually.

		 CAP has high excess medical costs, both per case (range: $7220-$11,443) and projected nationally ($6.8-$10.8 billion).

		 Approximately 20% of the burden of CAP comes from indirect costs. Preventing CAP could have a substantial impact on healthcare costs and productivity.

Pneumonia ranks among the top causes of death and hospitalization and is a major driver of healthcare costs. Most cases are classified as community-acquired: pneumonia occurring in a patient who is not hospitalized or residing in a long-term facility. In the United States, 55,000 deaths were attributed to pneumonia in 2006, with 1.2 million hospital discharges.

 While the burden of this disease is highest among the elderly, adults younger than 65 years accounted for 6000 deaths and 325,000 hospitalizations in 2006.

Despite the obvious burden of this disease on the healthcare system, surprisingly few studies have attempted to estimate the cost burden. Available estimates come from the mid-1990s where the total cost burden of treating community-acquired pneumonia (CAP) was estimated at $12.2 billion,

excluding productivity costs. Productivity costs were estimated for a Fortune 100 company where the annual total employer costs for pneumonia patients were $14,837, with nearly 30% attributed to work absence and disability (disability claims were assigned the cost of 5 workdays in addition to imputed illness absence days).

 This current analysis addressees this limitation by evaluating the total cost burden of CAP among non-elderly adults including both direct medical costs and productivity costs in a large cohort of commercially insured employees and dependents, geographically dispersed throughout the United States.

This study used the Thomson Reuters MarketScan Commercial Claims (Commercial) and Encounters database to determine the incidence and medical costs of CAP. The Commercial database contains adjudicated medical and outpatient prescription drug claims of several million employees and their dependents (annually), covered under a variety of health plan structures. In 2008, there were over 34 million covered lives in the Commercial database. Productivity costs were assessed on a subset of patients with data available in the Thomson Reuters Health Productivity and Management (HPM) database. The HPM database contains employeelevel workplace absenteeism and short-term disability data for a subset of approximately 1.4 million patients in the Commercial database.

The Commercial and HPM databases have been de-identified and are in compliance with the Health Insurance Portability and Accountability Act (HIPAA) regulations; this analysis was exempt from Institutional Review Board approval.

We selected adults aged 18 to 64 years in the Commercial database between July 1, 2002, and December 31, 2007. We identified all claims for pneumonia based on the 

International Classification of Diseases, Ninth Revision, Clinical Modification

 diagnosis codes (codes: 480.xx-486.xx). The index date for any patient with pneumonia was the date of the first diagnosis of pneumonia following 6 months of continuous medical and prescription coverage. This means that while pre-index periods could go back into 2002, index dates had to occur between January 1, 2003, and December 31, 2007. Patients with a diagnosis of pneumonia in the 6-month pre-index period were excluded. There was no minimum requirement for post-index continuous medical and prescription coverage; patients were followed up to 12 months, depending on data availability.

Pneumonia was categorized as community-acquired if the patient had either a principal or non-principal diagnosis of pneumonia on an outpatient claim or a principal diagnosis of pneumonia on an inpatient claim with no evidence of an inpatient or long-term care stay in the previous 30 days. (For inpatient admissions, principal diagnosis is the condition established after study to be chiefly responsible for occasioning the admission of the patient to the hospital or for care, which may or may not be the admitting diagnosis for the inpatient admission. In the case of outpatient services, principal [or primary] diagnosis is the main condition treated or investigated during the relevant episode of outpatient [ambulatory] healthcare.) The remaining patients with a diagnosis of pneumonia were assumed to have healthcare-associated pneumonia or hospital-acquired pneumonia (HAP) and were excluded from this analysis. Patients in the database without a diagnosis of pneumonia were included in incidence calculations and as controls in cost analyses. Control patients were randomly assigned an index date between January 1, 2003, and December 31, 2007, and were also required to have 6 months of continuous medical and prescription coverage preceding the index date.

This study employed the most recent HPM data available, which covered January 1, 2003, through December 31, 2006. Patients could contribute work absence, short-term disability, or both forms of data; work absence and short-term disability data were analyzed separately to maximize the sample sizes. While the direct costs analysis includes data on both employees and dependents, indirect cost data is only available for employees.

Healthcare expenditures were derived from all inpatient medical, outpatient medical, and outpatient pharmacy claims data captured in the Commercial database during the study period. (Emergency department [ED] visits that did not lead to an inpatient stay were classified as outpatient medical visits. ED visits were classifed as inpatient medical if they led to an inpatient stay.) Medical expenditures were defi ned as total gross payments to a provider, including deductibles, copayments, and coordination of benefits. Pharmaceutical expenditures were the actual reimbursement to the pharmacy, including the drug’s administrative dispensing fee, ingredient cost, patient portion of costs, and sales tax. Expenditures for services provided under capitation arrangements were estimated using payment proxy based on procedure code and region of care by the data provider. All expenditures were inflated to 2008 dollars by multiplying each year’s cost by the US Bureau of Labor Statistics Medical Care Specific Consumer Price Index.

Productivity costs were quantifi ed using the number of employer-recorded days of absence from work and short-term disability days during the post-index period. A wage constant of $41 per hour was used in this analysis. This was based on a previous benchmarking study of health- and productivityrelated costs for 32 companies across the United States and adjusted to 2008 US dollars.

 It was assumed a patient would only receive 70% of wages while on disability, thus a $28.70/hr wage constant was used to valuate short-term disability claims.

Patient demographic characteristics were measured on the index date and included age, gender, US Bureau of Census region of residence, health plan type, insurance plan capitation, population density status (urban vs rural), industry type, employee classification (hourly versus salary), and union classification (union vs non-union). Clinical variables calculated during the baseline period included the Deyo Charlson Comorbidity Index (CCI); comorbidities of chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), cancer, cardiovascular disease (CVD), other musculoskeletal surgery, diabetes, and asthma; pre-period use of antibiotics, long-acting beta-agonists, short-acting beta-agonists, and systemic corticosteroids; indicator of whether there was a hospitalization in the 1 month prior to index date; 6-month pre-period costs for inpatient, outpatient, and pharmacy; and time observed for follow-up.

The incidence of CAP for the 5-year study period was calculated based on all the eligible benefi ciaries in the Commercial Database between January 1, 2003, and December 31, 2007, that had at least 6 months of uninterrupted enrollment preceding the index date for pneumonia or controls. The numerator was the number of cases during the study period. The denominator was the total enrollment time during the study period in the Commercial Database among potentially eligible patients. Incidence rates were reported as CAP cases per 1000 person-years and stratifi ed by gender and age categories.

In order to estimate the direct medical cost and productivity burden of pneumonia, we generated a control group from all beneficiaries without a pneumonia diagnosis using nearest neighbor propensity score matching without replacement in a 3:1 control:case matching ratio. The propensity score was estimated by using a logistic regression modeling the presence of pneumonia as the dependent variable with the following covariates: index year; quarter; age; gender; geographic region; plan capitation status; industry; length of follow-up; baseline Deyo CCI; and baseline comorbid conditions, including asthma, cancer, CVD, CHF, COPD, diabetes, and HIV. Propensity score matching was performed and reported separately among the subset of patients (and controls) with work absence and short-term disability data. Despite the propensity score match, imbalances in risk will remain, and with the highly skewed distribution of costs, these imbalances may have an important infl uence on the estimate of excess costs. Therefore, in addition to the excess costs estimated from the observed differences between the pneumonia and non-pneumonia matched groups, we estimated excess costs in multivariate analyses by ordinary least squares (OLS) and generalized linear model (GLM). Separate models were run for total inpatient cost, ambulatory care excluding pharmacy cost, and pharmacy cost due to the distinct cost distribution for each of these categories. These predicted costs were then summed in order to assess total costs. Similarly, we used separate models to analyze absenteeism and short-term disability. Two-part models were used for inpatient costs, absenteeism, and short-term disability.

For GLM, the log link and gamma variance functions were used. The control variables for all multivariate models included: age; gender; insurance type; region; urbanicity; employee class (union vs non-union); employment status; employee relationship; Deyo CCI score in pre-period; COPD; CHF; cancer; CVD; other musculoskeletal surgery; diabetes; asthma; pre-period use of antibiotics, long-acting beta-agonists, short-acting beta-agonists, systemic steroids; indicator of whether there was a hospitalization in the 1 month prior to index date; total inpatient costs in pre-period; total outpatient costs in pre-period; total pharmacy costs in pre-period; and time observed for follow-up. All cost estimates are in 2008 US dollars.

To quantify the total annual cost of CAP that occurs in the US population aged 18 to 64 years, we multiplied our estimate of cost of an episode of CAP by our estimate of the incidence of CAP and by the number of people in the United States in this age group. We quantify total annual costs of CAP with and without productivity costs. We express our range of methodological uncertainly by producing estimates for each of our 3 methods of estimating medical and productivity costs. We quantify uncertainty to stochastic variation in our estimates through Monte Carlo simulation based on the means and standard errors of inputs with distributional assumptions. Because each input could be skewed to the right and had to be non-negative, we fi tted lognormal distributions to each major input and ran a simulation with 20,000 iterations. The simulation was done in Crystal Ball, Fusion Edition, version 11.1, and Excel 2007 software packages.

 contains a summary of the incidence of CAP among patients in the MarketScan Commercial Database, in total and stratified by age. Annualized overall incidence for this population was 4.89 cases per 1000 person-years. This incidence calculation was based on the 402,831 cases of CAP identified from 2003 to 2007 in our sample of 36.25 million commercially insured adults aged 18 to 64 years, with a total enrollment time of about 83 million years of risk. We project this annualized rate to the current US adult non-elderly population to arrive at an estimate of 947,000 cases of CAP in 2010 among adults aged 18 to 64 years. This analysis suggests increasing incidence with age; incidence among patients aged 55 to 64 years was 3.3 times the incidence of the group aged 18 to 34 years.

Demographics and Baseline Clinical Characteristics of CAP Sample

Approximately 5% of non-elderly adults have a community acquired pneumonia (CAP) annually, with an annual total of $10.6 billion in direct and indirect costs.

Daniel Polsky, PhD

Articles

Innovative Payment to Scale Up Access to Medications for Opioid Use Disorder

Improving Provider Directory Accuracy: Can Machine-Readable Directories Help?

ACA-Mandated Elimination of Cost Sharing for Preventive Screening Has Had Limited Early Impact

Incidence and Cost of CAP in a Large Working-Age Population

Are Primary Care Physicians Ready to Practice in a Consumer-Driven Environment?

Drug Benefit Generosity and Essential Medication Use Among Medicare-eligible Retirees