Abstract
The purpose of this dissertation is to assess Part D prescription waste in hospice patients, in three related studies. This dissertation seeks to characterize Part D prescription waste among linked Surveillance, Epidemiology, and End Results (SEER) Medicare and a random 5% sample of Medicare fee-for-service hospice patients by examining policy intervention impacts, assess the quantity and type of prescription medication at time of death, use a novel methodologies such as random forests to identify factors that influence the likelihood of such prescriptions on hand at time of death, and assess life expectancy as a factor in determining prescription lengths that reduce prescriptions on hand at death. Overview of Hospice and Part D ProgramsHospice was created with the goal of providing medical care that focuses on optimizing quality of life and mitigating suffering among people with terminal illness. For eligible Medicare patients, hospice is covered under the Medicare Part A insurance benefits and includes care for an individual’s terminal illness and related conditions. Hospice uses teams to provide doctor services, nursing care, medical supplies, prescription drugs, therapy (physical, occupational, and speech-language), social worker support, dietary counseling, grief counseling, and short-term inpatient or respite care. Medicare Part D is a voluntary prescription drug benefit provided by private insurance sponsors for a monthly premium. The program aims to make prescription medications more affordable and accessible to Medicare recipients. The benefit covers patient’s prescription drugs in most cases, but there are circumstances where drugs are covered instead under either Medicare Part A or Part B. One exception is when a patient elects hospice, and the prescription drugs related to the care of the terminal illness and conditions are covered under the Medicare Part A benefit. Medications unrelated to the patient’s terminal illness may still be obtained through the Part D benefits. In the past decade, hospice care has prioritized quality while aiming to reduce unnecessary waste. The Center for Medicare and Medicaid Services (CMS) defines waste as practices leading to unnecessary costs for the Medicare program. Since October 2010, CMS has issued several communications to Part D Sponsors and Hospice Providers highlighting the problem of Medicare paying for drugs under Part D that should be covered by hospice Medicare Part A. This inappropriate billing has led to significant costs and waste for the Medicare program. The following reviews the laws, practices, and/or guidelines CMS has published from 2008 to present providing historical context for understanding Part D prescription waste among hospice patients. Years 2008-2013In June 2008, the CMS through the Federal Registrar released the first major revision of Medicare Hospice Conditions of Participation (CoPs) since the Medicare Hospice Benefit was established in 1983. The major revision related to hospice patient’s prescriptions (including Part D medications) included: CFR-Explanation of Revision§ 418.106(e) Added clarification that reiterates the requirement that hospices must provide all drugs and supplies related to a patient’s terminal illness and related conditions and not expect patients to obtain drugs related to the terminal illness and related conditions through Medicare Part D. And that longstanding, preexisting conditions and comorbidities are included in the hospice bundle of services as written in the original implementing regulations of the Medicare hospice benefit. However, if a patient necessitates drugs unrelated to the terminal illness, they may seek coverage through Medicare Part D. CFR-Explanation of Revision § 418.54(c) Clarified the term “unnecessary drugs” as part of the content of the comprehensive assessment and reiterated that all medications should be included in the review in order to develop a plan of care. The ruling went on to clarify that as part of the drug profile review, the assessment should include a patient’s prescription and over-the-counter drugs in use, drug effectiveness, side effects, drug interactions, duplicate therapies, and under or overdosing. Following these changes and clarifications it wasn’t until October 2010 that CMS released a Memorandum entitled Preventing Part D Payment for Hospice Drugs. The memorandum indicated there were concerns that Part D sponsors were paying for drugs that should be the responsibility of the Medicare hospice provider. Guidelines were released directing Part D sponsors to communicate with their network pharmacies to ensure Medicare hospice drugs were not billed to Part D. CMS indicated they would provide best practices for doing so by late 2011. However, following an initial proposal (in February 2011), by April 2011 CMS issued as part of the Announcement of CY 2012 Medicare Advantage Capitation Rates and Medicare Advantage and Part D Payment Policies a section detailing the best practices for “Preventing Part D Payment for Hospice Drugs”. The practices recommended Part D sponsors utilize patient-level transaction reply reports (TRR) they had previously been receiving from CMS. These reports contained patient enrollment information and hospice election information. The best practices detailed how to utilize the included hospice indicators and data to ensure the claims processor is notified of an enrollee’s hospice election and that processes are in place to prevent Part D payment for hospice drugs. Then in June 2012, the Department of Health and Human Service, Office of the Inspect General (DHHS OIG) released a report titled “Medicare Could Be Paying Twice for Prescription Drugs for Hospice Patients” (which examined data from 2009). CMS concurred with two recommendations DHHS OIG made with regard to preventing the Part D benefit paying for medications already covered under the hospice Part A per diem payments. The accepted recommendations included: 1) Educating Part D sponsors, hospices, and pharmacies that it is inappropriate for Medicare Part D to pay for drugs related to hospice patients’ terminal illnesses; and 2) Requiring Part D sponsors to develop controls that prevent Part D from paying for drugs that are already covered under the per diem payments. Following additional TRR report guidance was provided by CMS in April 2013, CMS released a final rule in August 2013 requiring all Part D sponsors to have in place “means” to prevent duplicate payment of hospice medications as well as provided additional clarifications and explanations to sponsors, hospices, and pharmacies. CMS strongly recommended the Part D sponsors use of the TRR reports and have in place controls to prevent the reimbursement for hospice medications. CMS indicated using prior authorizations (PA) for all hospice medications through Part D was best practice but wasn’t required. However, CMS gave specific instruction for sponsors to implement PAs (or other approaches) for four categories of prescription drugs in hospice patients: analgesics, antinauseants, laxatives, and antianxiety drugs. These were identified by the DHHS OIG as typically used to treat the symptoms generally experienced by hospice patients during the end of life. CMS also provided additional guidance and practices for Part D sponsors, hospices, and pharmacies detailing terminal diagnosis and interrelated conditions. CMS reiterated the original intent of the hospice benefit was to have a Medicare benefit available that provided virtually all-inclusive care for terminally ill individuals, provide pain relief and symptom management, and offered the opportunity to die with dignity and comfort in one’s own home rather than in an institutional setting. At the end of 2013 in December, CMS released a memorandum seeking comments on new expectation for stakeholders related to "Part D Payment for Drugs for Beneficiaries Enrolled in Hospice". The memorandum provided a condensed overview of prior regulatory directives pertaining to the eligibility criteria and extent of benefits applicable to Medicare hospice services under Part A. CMS reiterated that patients should only very rarely be taking drugs that are not covered under the hospice per diem. CMS further stated that for prescription drugs to be covered under Part D when the enrollee has elected hospice, the drug must be for treatment of a condition that is completely unrelated to the terminal condition(s) or related conditions. In other words, the drug is unrelated to the terminal prognosis of the individual. In addition, CMS communicated new expectations for Part D sponsors, aimed at preventing duplicate payments for medications covered within the hospice benefit or waived due to the beneficiary’s hospice election. CMS expected for drugs covered under Part D for hospice patients to be extremely rare, the Part D sponsors should place patient-level PA requirements on the following four categories of prescription drugs: analgesics, antinauseants, laxatives, and antianxiety drugs for hospice patients to determine whether the drugs are coverable under Part D. The memorandum also provided guidance to Part D sponsors on making retrospective determinations of payment responsibility for drugs within these categories during the hospice election. The guidance to sponsors was to conduct outreach to the hospice provider to determine whether the drug is for treatment of a completely unrelated condition. CMS stated they expected the hospice provider to coordinate with the plan sponsor regarding these claims and provide the necessary written information, as requested by the sponsor. Years 2014-PresentIn March 2014, CMS issued guidance and established a standard Part D PA form, required for use by Part D sponsors, hospices, and prescribers. The following July, CMS issued a final rule memorandum regarding the “Determination of Payment Responsibility for Drugs for Hospice Patients”. In this final memorandum, CMS provided updates to the March 2014 PA form and explanatory documentation and communicated their expectation for its universal implementation of their guidance by October 1, 2014. In June 2015, the DHHS OIG issued a report titled "Ensuring the Integrity of Medicare Part D", providing a synthesis of investigations, audits, evaluations, and legal guidance related to weaknesses in the Part D program. It again identified highlights the 2012 DHHS OIG report discussing the inappropriate billing of hospice patient’s drugs in 2009 to Part D that should be covered by hospice Medicare Part A. The DHHS OIG followed that report with a March 2016 report titled “Hospices Inappropriately Billed Medicare Over $250 Million for General Inpatient Care”. CMS responded by concurring with the recommendation to increase its oversight of Part D payments for drugs for hospice patients. CMS began the process to procure a Hospice Recovery Audit contractor to conduct claim reviews and recoup payments as necessary. That November, CMS issued a memorandum titled "Update on Part D Payment Responsibility for Drugs for Beneficiaries Enrolled in Medicare Hospice", acknowledging, and thanking stakeholders for improvements in billing practices and noting the implementation of their Hospice Recovery Audit contractor for claim reviews. Since 2013 stakeholders have raised concerns about the promptness of communications concerning the entitlement status of hospice patients. In response CMS, in August 2017, took additional action by outlining their strategy for introducing an electronic notice of election form for hospices to communication a patients election (OMB No. 0938-1269) as a component of the FY 2018 Hospice Wage Index and Payment Rate Update. The form went through updates before it was finalized in October 2018. In July 2018, the DHHS OIG recommended to CMS, via a report that assessed the vulnerabilities in the Medicare hospice program, the need to execute a strategy to intervene with hospices to ensure they are providing the drugs covered under the hospice benefits and not inappropriately billed to Part D. CMS did not concur with this recommendation, which was similar to a recommendation from the June 2012 “Medicare Could Be Paying Twice for Prescription Drugs for Hospice Patients” report that they didn’t concur with then either. Then the following year, in August 2019, the DHHS OIG conducted a follow-up audit to their 2012 findings. The DHHS OIG released their report titled “Medicare Part D Is Still Paying Millions for Drugs Already Paid for Under the Part A Hospice Benefit” [15]. The report, which examined data from 2016, detailed the ongoing inappropriate billing of prescriptions for hospice patients that the DHHS OIG found previously in its 2012 Report. In response, CMS commented that they would continue to engage in meaningful activities to reduce duplicate payment in this area, such as ensuring hospice providers are proactively educating patients on covered services and items (including drugs) and Part D drug plan sponsors are appropriately applying PA criteria and coordinating with hospice providers on drug coverage issues. To further address the ongoing problem of inappropriate billing documented in the 2012 and 2019 DHHS OIG reports, CMS through the Federal Registrar issued as part of the FY 2020 Hospice Wage Index and Payment Rate Update requires hospices disclose in an extensive written addendum to patients (and other health care providers) any care that would be deemed unrelated to hospice care as part of a Patient Notification of Hospice Non-Covered Items, Services, and Drugs (OMB 0938-1153). The following year in August 2020, CMS made form OMB 0938-1153 a condition for payment for hospices. Unfortunately, prescription waste can occur in many ways, not just through inappropriate billing. Estimates suggest up to $2 billion annually, in unused prescription medication, is being wasted in Medicare Part A long-term care facilities alone. A 2013 report by Visante found that around 14 million (approximately 1%) of all Part D prescriptions are wasted yearly. The study reported that most of this waste stems from therapy discontinuation, medication switching, dosage adjustments, and death. Regarding waste due to patient mortality, the study unearthed that, on average, patients had 50% of each prescription on hand at the time of their death. Goals of this DissertationThis dissertation seeks to characterize Part D prescription waste among SEER Medicare hospice patients by examining policy intervention impacts, assess the quantity and type of prescription medication at time of death, use a novel methodology such as random forests to identify factors that influence the likelihood of such prescriptions on hand at time of death, and assess life expectancy as a factor in determining prescription lengths that reduce prescriptions on hand at death compared to traditional prescribing methods. This will be accomplished through three related studies in hospice care and prescriptions. The first study will utilize generalized estimating equations (GEE) with negative binomial regression analysis to understand the effects of hospice patient Part D billing policy guidance on linked SEER Medicare data of male hospice patients with prostate cancer and their Part D prescriptions. The second study seeks to examine Part D prescriptions waste in linked SEER Medicare data of hospice patients, with breast; lung; pancreas; prostate; and stomach cancer and identify any predictive characteristics. The methodology for this study consists of calculating the type and amount of medication on hand at time of death and the associated costs by year and then conducting predictive analyses of characteristics that influence Part D prescriptions waste using machine learning techniques. The third study will develop and test rule-based prescription durations for Medicare patients in hospice, with a particular focus on those with a survival of 90 days or less. This work will inform the development of a decision support tool that will describe Part D prescription durations that reduce potential waste related to the amount of prescription medication on hand at death compared to traditional prescribing methods. The methodology will use Random Survival Forest (RSF) calibrated with median trapezoidal rule to develop survival estimates, to simulate clinician predicted survival, which the rule-based prescription durations were applied to. Medication on hand at time of death was calculated and the resulting waste was compared between the rule-based prescription lengths and provider durations. By employing generalized estimating equations, the first study was able to assess the (1) total monthly average prescriptions of all medications and (2) four categories of commonly prescribed hospice medications in pre-and-post policy guidance. This study investigated the effects of guidance issued by CMS on April 4, 2011, targeting providers to prevent the improper billing of prescription drugs for hospice patients’ terminal illness and related conditions to the Part D benefit. Using linked SEER Medicare data for male hospice patients between April 2009 and March 2013, the analysis found that hospice patients’ monthly average total Part D prescriptions decreased from 7.3 pre-policy guidance to 6.5 medications following the issuing of the guidance, while the four categories of hospice-specific medications decreased from 0.57 to 0.49. The findings of this study show that CMS’s guidance issued to providers to prevent the inappropriate billing of hospice patients’ prescriptions to the Part D benefit may lead to decreases in improper billing as observed in this sample. Summary statistics were applied in the second study to examine the type and quantity of Medicare Part D medications on hand at time of death in hospice patients. This analysis utilized a 5% subset of Medicare fee-for-service patient claims and linked SEER Medicare patient claims spanning from January 2015 to December 2019. Results indicated that cardiovascular medications accounted for 25% of prescriptions, followed by central nervous system medications at 20%. The mean prescription length was 36.65 days’ supply with a mean of 62.18 quantity dispensed. Prescriptions resulting in medication on hand at time of death on average were dispensed 72.69 days after a patient’s admission to hospice and resulted in a mean of 20.02 days’ supply and 34.18 quantity wasted. Additionally, the study evaluated the predictive accuracy of four classifiers in forecasting prescription waste at time of death, with Random Forest achieving the highest performance, boasting an area under the curve (AUC) exceeding 93%. Feature importance analysis revealed prescription days’ supply and quantity dispensed as the most influential factors. Even after removing these predictive features, Random Forest still demonstrated a respectable AUC of 73.5%. The study demonstrates that medication on hand at time of death in hospice patients can be predicted and supports additional research should be done to identify ways to reduce the waste. In the final study, rule-based prescription durations were developed and applied to each patient based on their simulated survival days. RSF calibrated with median trapezoidal rule was used to simulate clinician estimated patient survival days. Medication on hand at time of death was then calculated for the rule-based prescriptions and compared to the amount caused by the traditional clinician prescription durations. Two scenarios were conducted that compared the overage for (1) all prescriptions regardless of when the clinician determined prescription ended and (2) prescriptions where a threshold excluded prescriptions where either the clinician or rule-based prescriptions ended more than 3 days before the death date. In the initial scenario, the rule-based prescriptions reduced overage in 28% of cases, leading to a decrease of 29.1% to 36.1% in the amount of prescription medication on hand at the time of death. The second scenario saw similar success with the rule-based prescriptions reducing overage in 32% of cases, leading to a decrease of 32% to 45.5% in the amount of prescription waste. Overall, in this sample the rule-based initial and refill prescription durations were effective in reducing waste. The occurrence of Part D prescription waste in hospice has been well documented by CMS in memorandums and reports. However, few studies exist examining CMS implemented policy impacts to reduce waste, characteristics identification of Part D hospice prescription waste for predicting medication on hand at death, or novel methods to reduce medication waste at the source, i.e., the prescription. Each of these three related studies is significant in that it addresses a gap in hospice care and prescription waste using novel machine learning approaches. The first study is significant in it addresses how government policy guidance has impacted the inappropriate billing of Part D prescriptions in hospice patients. While CMS has documented decreases in billing of Part D prescriptions in hospice patients, no study or analysis exists that definitively ties the decrease and policy together. This study addresses this gap using a GEE with negative binomial regression will address this gap and has the potential to bolster the findings of CMS. The significance of the second study is twofold, as like the first paper, this study also addresses two gaps: 1) the review of Part D prescription waste patterns in hospice patients using individual hospice patient claim records, and 2) in its identification of characteristics that influence the likelihood of Part D prescription waste. Limited research exists in exploring the type, quantity, and costs of medications on hand at time of death in hospice patients and no research exists that examines characteristics that influence the likelihood of prescription waste. Currently this gap in research exists as prior studies have only assessed the amount of hospice Medicare Part A prescription waste at individual hospice organizations. Using novel methods, such as K Nearest Neighbor, the study will assess the effects of hospice patient’s characteristic on the likelihood of Part D medications on hand at time of death. This study is crucial as a first step to bring about further understanding and awareness of potential waste patterns in hospice patients’ Part D prescriptions in order to better the prescribing patterns of physicians and improved care coordination between teams. The use of average life expectancy to influence prescription lengths in hospice patients is a significant gap in the literature. While studies have explored the clinical and cost-effectiveness of longer and shorter (3-month vs 28 days) duration prescriptions to reduce waste, studies identifying prescription durations for reducing medication on hand at time of death in hospice patients do not. The use of machine learning methods in this study, e.g. RSF, to develop a rule-based decision support tool to assess prescription lengths that reduce waste in real time is novel. This study has the potential to influence how Part D prescriptions are prescribed to hospice patients in order to address ongoing issues of medical waste in terms of money and resources. This work provides critical evidence on the scope of problematic medication waste at the end of life. The research should ideally be used to promote increased efforts around appropriate billing of prescription drugs in conjunction with identifying predictors of medication on hand at time of death to developed tailored medication prescription duration strategies that minimize burden without impacting quality of life during the patient’s final weeks and months. Additionally, this work’s implementation of novel machine learning approaches, using individual patient data, provides further evidence of need for these advanced models to be incorporated into provider’s medical software to aide in decision making as well as providing personalized medicine to the patient. Until then, a more careful review of patients’ prescription prescribed through Part D and the prescription duration at the time of hospice enrollment is warranted.
Author and committee
dc:creator, dc:contributor.*- Author
-
- Irvin, Katherine Jean
Subjects
dc:subject × 6Identifiers
dc:identifier.*- Identifier
- hdl:1920/13810
- OAI identifier oai:identifier
- oai:MARS:1920/13810