Handling Missing Data in Health Economics and Outcomes Research (HEOR): A Systematic Review and Practical Recommendations.

BACKGROUND: Missing data in costs and/or health outcomes and in confounding variables can create bias in the inference of health economics and outcomes research studies, which in turn can lead to inappropriate policies. Most of the literature focuses on handling missing data in randomized controlled trials, which are not necessarily always the data used in health economics and outcomes research. OBJECTIVES: We aimed to provide an overview on missing data issues and how to address incomplete data and report the findings of a systematic literature review of methods used to deal with missing data in health economics and outcomes research studies that focused on cost, utility, and patient-reported outcomes. METHODS: A systematic search of papers published in English language until the end of the year 2020 was carried out in PubMed. Studies using statistical methods to handle missing data for analyses of cost, utility, or patient-reported outcome data were included, as were reviews and guidance papers on handling missing data for those outcomes. The data extraction was conducted with a focus on the context of the study, the type of missing data, and the methods used to tackle missing data. RESULTS: From 1433 identified records, 40 papers were included. Thirteen studies were economic evaluations. Thirty studies used multiple imputation with 17 studies using multiple imputation by chained equation, while 15 studies used a complete-case analysis. Seventeen studies addressed missing cost data and 23 studies dealt with missing outcome data. Eleven studies reported a single method while 20 studies used multiple methods to address missing data. CONCLUSIONS: Several health economics and outcomes research studies did not offer a justification of their approach of handling missing data and some used only a single method without a sensitivity analysis. This systematic literature review highlights the importance of considering the missingness mechanism and including sensitivity analyses when planning, analyzing, and reporting health economics and outcomes research studies.

Estimating Drug Costs: How do Manufacturer Net Prices Compare with Other Common US Price References?

BACKGROUND: Drug costs are frequently estimated in economic analyses using wholesale acquisition cost (WAC), but what is the best approach to develop these estimates? Pharmaceutical manufacturers recently released transparency reports disclosing net price increases after accounting for rebates and other discounts. OBJECTIVE: Our objective was to determine whether manufacturer net prices (MNPs) could approximate the discounted prices observed by the U.S. Department of Veterans Affairs (VA). METHODS: We compared the annual, average price discounts voluntarily reported by three pharmaceutical manufacturers with the VA price for specific products from each company. The top 10 drugs by total sales reported from company tax filings for 2016 were included. The discount observed by the VA was determined from each drug's list price, reported as WAC, in 2016. Descriptive statistics were calculated for the VA discount observed and a weighted price index was calculated using the lowest price to the VA (Weighted VA Index), which was compared with the manufacturer index. RESULTS: The discounted price as a percentage of the WAC ranged from 9 to 74%. All three indexes estimated by the average discount to the VA were at or below the manufacturer indexes (42 vs. 50% for Eli Lilly, 56 vs. 65% for Johnson & Johnson, and 59 vs. 59% for Merck). CONCLUSIONS: Manufacturer-reported average net prices may provide a close approximation of the average discounted price granted to the VA, suggesting they may be a useful proxy for the true pharmacy benefits manager (PBM) or payer cost. However, individual discounts for products have wide variation, making a standard discount adjustment across multiple products less acceptable.

Cardiovascular event-free survival after adjuvant radiation therapy in breast cancer patients stratified by cardiovascular risk.

The objective of this study was to estimate the risk of a cardiovascular event or death associated with modern radiation in a population of elderly female breast cancer patients with varying baseline cardiovascular risk. The data used for this analysis are from the linked Surveillance, Epidemiology, and End-Results (SEER)-Medicare database. The retrospective cohort study included women aged 66 years and older with stage 0-III breast cancer diagnosed between 2000 and 2005. Women were grouped as low, intermediate, or high cardiovascular risk based on the presence of certain clinical diagnoses. The risk for the combined outcome of a hospitalization for a cardiovascular event or death within 6 months and 24 months of diagnosis was estimated using a multivariable Cox model. The median follow-up time was 24 months. Among the 91,612 women with American Joint Committee on Cancer (AJCC) stage 0-III breast cancer: 39,555 (43.2%) were treated with radiation therapy and 52,057 (56.8%) were not. The receipt of radiation therapy in the first 6 months was associated with a statistically significant increased risk for the combined outcome in women categorized as high risk (HR = 1.510; 95% CI, 1.396-1.634) or intermediate risk (HR = 1.415; 95% CI, 1.188-1.686) but not low risk (HR = 1.027; 95% CI, 0.798-1.321). Women with a prior medical history of cardiovascular disease treated with radiation therapy are at increased risk for an event and should be monitored for at least 6 months following treatment with radiation therapy.

Guidance Document: Global Pharmacoeconomic Model Adaption Strategies.

OBJECTIVE: The purpose of this guidance was to assist in the adaptation of pharmacoeconomic models originally developed in one country and intended for use in another. The intent was to produce user-friendly recommendations and a checklist for adapting a global model to treat a specific disease state. This guidance will allow model developers to tailor existing models so that they are "locally applicable," while maintaining the scientific integrity of the original pharmacoeconomic model and will benefit formulary decision makers and other stakeholders involved in evaluating pharmacoeconomic studies. METHODS: A working group of experts from various countries participated in the Global Pharmacoeconomic Model Guidance development to discuss the adaptation of pharmacoeconomic models. A systematic review of studies adapting pharmacoeconomic models and translation across countries was conducted and recommendations were made for adaptation. The working group interviewed internal and external stakeholders to solicit best practices for model adaptation and developed a draft set of key principles and general recommendations for global adaptation. RESULTS: The working group provided a set of 16 recommendations for adapting pharmacoeconomic models for local decision makers. The recommendations span various aspects of estimating or modeling both the costs and effectiveness of pharmacoeconomic models as well as guidance for ensuring local acceptability. CONCLUSIONS: These recommendations and the related principles not only will provide pharmacoeconomic models that are meaningful to local decision makers but also will improve the consistency and credibility of pharmacoeconomic model adaptations. The guidance may also help those who will build the original models to design them with the flexibility to allow pharmacoeconomic model adaptations as described in this document.

Impact of venous thromboembolism on mortality of elderly Medicare patients with stage III colon cancer.

BACKGROUND: The improvement in survival rates for patients with colon cancer has shifted the focus from examining cancer-specific mortality to exploring all-cause mortality. Adverse events such as venous thromboembolism (VTE) affect overall survival times and the net clinical benefit of cancer management strategies. METHODS: This retrospective study used Surveillance, Epidemiology and End Results (SEER) Medicare data to examine VTE incidence and mortality rates for elderly patients with stage III colon cancer who were diagnosed in 2004 or 2005 and followed through 2007. The impact of VTE on mortality was estimated using multivariable Cox proportional hazards regression. RESULTS: In all, 20.7% of 4,985 elderly patients with stage III colon cancer had clinically diagnosed VTE following diagnosis. All-cause mortality risk was higher for patients with a VTE diagnosis (hazard ratio [HR]: 1.15, 95% confidence interval [CI]: 1.04-1.27), greater comorbidity burden, more advanced tumor depth and nodal involvement within stage III, advanced age, and male sex; the risk was lower for patients treated with chemotherapy. VTE was associated with higher mortality hazards (HR: 1.41, 95% CI: 1.21-1.64) for patients treated with adjuvant chemotherapy but not for untreated patients. CONCLUSIONS: A new diagnosis of VTE significantly reduced survival rates for elderly patients with stage III colon cancer and further reduced survival rates for patients treated with chemotherapy. Improved prevention and management of VTE for elderly patients with stage III colon cancer who are at risk for VTE is warranted, particularly for patients treated with chemotherapy.

Barriers to self-management of diabetes: a qualitative study among low-income minority diabetics.

OBJECTIVES: Diabetes self-management is a key element in the overall management of diabetes. Identifying barriers to disease self-management is a critical step in achieving optimal health outcomes. Our goal was to explore patients' perceptions about barriers to self-management of diabetes that could possibly help explain poor health outcomes among minority patients. STUDY DESIGN: Four focus groups were conducted among 31 predominately African American patients with diabetes who were enrolled in the Baltimore Cardiovascular Partnership Study, a NIH-funded multiyear prospective partnership study. The topic guide consisted of a series of open-ended questions about knowledge of current health status, medication use, continuity of care, blood glucose level and nutrition. RESULTS: The focus groups confirmed that previously reported barriers to self-management persisted and identified new concerns that could be associated with poor health outcomes among minority patients with diabetes. Attitudes, perceptions and behaviors surrounding diabetes and self-management of the condition did vary across individuals, however, the variation appeared to reflect the individual's knowledge and opinions rather than patient's age, sex, or culture. The primary barrier to diabetes self-management resulted from lack of knowledge of target blood glucose and blood pressure. Several participants found some of the health information to be quite confusing. CONCLUSIONS: Diabetes is a major public health concern and the lack of awareness of target blood glucose and blood pressure further complicates the problem. The limited health literacy seen in this study could help explain several of the barriers to self-management. The barriers to self-management identified in this qualitative study are amenable to intervention that could improve health outcomes.

Obesity and medicare expenditure: accounting for age-related height loss.

To determine the relationship between BMI and Medicare expenditure for adults 65-years and older and determine whether this relationship changes after accounting for misclassification due to age-related height loss. Using a cross sectional study design, the relationship between BMI and fee-for-service Medicare expenditure was examined among beneficiaries who completed the Medicare Current Beneficiary Survey (MCBS) in 2002, were not enrolled in Medicare Health Maintenance Organization, had a self-reported height and weight, and were 65 and older (n = 7,706). Subjects were classified as underweight, normal weight, overweight, obese (obese I), and severely obese (obese II/III). To adjust BMI for the artifactual increase associated with age-related height loss, the reported height was transformed by adding the sex-specific age-associated height loss to the reported height in MCBS. The main outcome variable was total Medicare expenditure. There was a significant U-shaped pattern between unadjusted BMI and Medicare expenditure: underweight $4,581 (P < 0.0003), normal weight $3,744 (P < 0.0000), overweight $3,115 (reference), obese I $3,686 (P < 0.0039), and obese II/III $4,386 (P < 0.0000). This pattern persisted after accounting for height loss: underweight $4,640 (P < 0.0000), normal weight $3,451 (P < 0.0507), overweight $3,165 (reference), obese I $3,915 (P < 0.0010), and obese II/III $4,385 (P < 0.0004) compared to overweight. In older adults, minimal cost is not found at "normal" BMI, but rather in overweight subjects with higher spending in the obese and underweight categories. Adjusting for loss-of-height with aging had little affect on cost estimates.