COST ANALYSES AND CLINICAL PRACTICE
* Finalists for the Lee B. Lusted Student Prize
Method: We hand-collected data on all anticancer drugs approved by the US Food and Drug Administration from 1995 to 2013. We calculated the treatment episode price for each drug based on the dosing and duration of treatment for a typical patient and the Medicare payment rate. We obtained information on survival benefits from randomized controlled trials and, for drugs approved on the basis of single-arm trials, modeling studies. We calculated the price per year of life gained for each drug in 2013 dollars. Using least squares regression, we estimated the relationship between the price per year of life gained and approval year.
Result: The average drug price is $78,600, and the average survival benefit is 0.45 years. We find that benefit-adjusted prices increased by 5% a year, or about $7,800. Put another way, in 1995 patients and their insurers paid $101,000 for a year of life. By 2005, they had to pay $178,000 for the same benefit. The result is robust to the inclusion of various controls for drug attributes (e.g., side effects, oral or intravenous administration).
Conclusion: Our results are consistent with the conventional wisdom that new anticancer drugs are more expensive than older products. In the absence of a well-defined standard of value, oncologists may evaluate new products’ prices against a reference price. Because reference prices are malleable and influenced by the prices of existing products, manufacturers have some leeway to increase launch prices over time without reducing demand. Expansions in the 340B drug pricing program, which requires manufacturers to give discounts to eligible buyers, may also have contributed to price increases for non-eligible buyers.
Method: Older Medicare patients at high risk of hospitalization in an urban teaching hospital were enrolled based upon their likelihood of having an average of 10 days of hospital days in a given calendar year. Patients were surveyed to assess four different types of CRN behaviors: not filling or refilling a prescription, delaying getting a prescription filled, skipping doses to make the medicine last longer, and/or taking smaller doses to make the medicine last longer. We analyzed the rates of these behaviors and their combinations, and assessed if the addition of Medicaid coverage was associated with lower rates of CRN.
Result: 524 eligible patients were surveyed, among whom 148 (28.2%) reported one or more CRN behaviors. The mean age of the full sample was 61.9 years (s.d. 16.6), with 320 (61%) female, and 240 (47.6%) with Medicaid coverage (i.e., dual eligible). Among these dual eligibles, 28% reported CRN, compared to 28% of those without Medicaid (p=0.98). Delaying filling of a prescription (22%) was the most prevalent CRN behavior, followed by not filling a prescription (19%), skipping doses (13%), and splitting doses to make the medicine last longer (13%). Among those who reported CRN, delaying getting a prescription was the single most prevalent CRN behavior (76%), but 58% reported two or more CRN behaviors, and 26% reported all of the four CRN behaviors.
Conclusion: CRN is highly prevalent among older Medicare patients at high risk of hospitalization, even those with Medicaid coverage. Among patients reporting CRN, delay in filling prescriptions is the most frequent CRN behavior, but multiple CRN behaviors are common. The frequency and extent of CRN in this population suggest the potential for large adverse effects of CRN and the likelihood that physicians may be inadequately aware or responsive to CRN behaviors. Patient-centered efforts to identify and address CRN deserve attention.
In Mexico, around 60,000 individuals currently benefit from universal, free-of-charge antiretroviral treatment (ART) through MoH facilities. The financial constraints of a publicly funded system threaten the short-term sustainability of the HIV national program. Thus, facility-level performance data is required to increase value-for-money in HIV spending. Indicators of the treatment cascade provide relevant insight of quality-adjusted outcomes and its associated costs, which could have a managerial role in assisting decision-making.
Administrative databases of the Mexican MoH were analyzed to assemble a dashboard showing the distribution of three key indicators of facility-level performance. We calculated the proportion of people who initiated ART with CD4>200 cel/mm3 - a proxy for early initiation -, the proportion of those who were not lost to follow-up, and those who ever achieved viral suppression (viral load<50 copies/mm3) during treatment. We calculated total annual costs of antiretroviral drugs per patient, for they comprise nearly 80% of the total cost of treatment. We then compared costs per treated person along the steps of the service cascade and assessed inter-facility variation as a measure of performance.
In 2012, 55% of all new patients initiated ART with CD4>200 cel/mm3(IQR: 48.9;63.6). 91% were retained for at least 12 months (IQR: 84.6;96.9), 16 pp more than in 2008. 57% were below the undetectable viral load threshold (IQR: 48.4;66.7). Average annual cost per patient treated was 3,595 USD (IC95%: 4,155-3,035), and this figure increased steadily along the treatment cascade: 6,484 per early initiated patient , 7,163 per retained patient and 12,608 per virally suppressed patient. The difference in costs between the least and the most expensive facility was 3.4 times the cost per retained patient, and 94.8 times per virally suppressed patient. The gains in retention yielded a decrease of 20% in costs between 2008 and 2012.
We found considerable room to improve early initiation and retention of HIV patients receiving ART. Breakdown of clinical information into useful metrics is imperative for managers and policy-makers. Late initiation and treatment withdrawal do matter on account of their impact on costs, and because they portray themselves a considerable effect on mortality. Our results encourage prompt and continuous surveillance of the treatment cascade indicators in Mexico and other settings, where administrative data is available.
Purpose: Randomized trials often find that costly, widely-used medical technologies are not effective. These trials may reduce costs, but only if physicians modify their beliefs accordingly. Since 1984, a number of studies have found that routine episiotomy during vaginal delivery does not benefit mothers or babies. We document “cognitive inertia” – the persistence of beliefs in the face of contradictory evidence – in physicians' use of episiotomy.
Methods: We measure physician-level episiotomy rates using Pennsylvania Inpatient Hospital Discharge Data for the period 1994 to 2010. The data are a census of hospital discharges in the state. The data include physician identifiers that can be linked across years. We identified spontaneous vaginal deliveries and episiotomy using ICD-9 diagnosis and procedure
Results: The sample includes over 1.6 million non-operative vaginal deliveries. The episiotomy rate decreased from 42% in 1994 to 10% in 2010. The figure shows trends in episiotomy rates by the year in which physicians entered practice. For example, the top line shows rates among physicians who entered practice before 1994. The next line shows rates among physicians who entered between 1995 and 1998. Older physicians reduced their use of episiotomy but continued to perform episiotomy at much higher rates than younger physicians. For example, in 2010 the episiotomy rate was 13% among physicians who entered practice before 1994. The episiotomy rate was only 5% among physicians who entered practice in 2009-2010.
Conclusion: Even as late as 2010, a woman treated by an older physician was substantially more likely to receive an episiotomy. We find that there is a clinically-significant degree of inertia in physicians' practice patterns in this context. There are no learning or switching costs that would have made it more difficult for older physicians to discontinue routine episiotomy, indicating that observed patterns are attributable to cognitive inertia. Results suggest caution in permitting technologies to diffuse into routine practice before they have been tested in trials.
Method: We updated and expanded existing systematic reviews of recommendations for the conduct and reporting of decision and simulation modeling with input from a multidisciplinary team of clinical, policy, and decision analysis experts. The results of the systematic review were discussed in-person with a panel of 28 stakeholders including patient representatives, providers of care, purchasers of care, payers, policy makers, and principal investigators. Stakeholders commented on existing recommendations from various sources and identified gaps, limitations and areas for elaboration. We subsequently reviewed the websites of 126 international health technology assessment organizations providing guidance on the conduct and reporting of decision and simulation models. We solicited further input from senior researchers with experience in decision and simulation modeling from AHRQ and its Evidence-based Practice Centers.
Result: We developed a list of principles and good practice recommendations for modeling conducted to enhance and contextualize findings of systematic reviews. The guidance applies to structural mathematical models, including declarative, functional, and spatial models. We categorized recommendations by whether they pertain to the model structure, model data, or consistency, and reporting. We provide the rationale for each recommendation, evidence supporting the recommendation, or best judgment where adequate evidence was lacking.
Conclusion: We used a systematic approach to develop guidance for decision and simulation modeling in the context of systematic reviews. We are optimistic that this work will contribute to increased use of modeling in systematic reviews.
Method: We developed a clinically informed framework for interventions that have direct and/or indirect effects on fertility and childbearing. We defined a set of health interventions that span the combinations in the framework (e.g., in vitrofertilization, selective embryo reduction, chlamydia screening and treatment, prenatal genetic testing). We performed a targeted literature review, searching for articles presenting CEAs of the interventions. For each article, we identified the health outcomes stemming from potential benefits and harms of the intervention that were considered, with particular focus on QALYs associated with current and future fertility and childbearing. We also reviewed relevant economic, legal, and ethical theoretical work.
Result: Our framework included interventions with direct and indirect effects on fertility, miscarriage, and birth defects as they related to women’s current pregnancies and women’s and men’s future fertility. We identified nearly 150 studies that illustrated current practice for CEAs of interventions in the framework including 12 papers providing theoretical guidance underpinning such analyses. We observed four patterns: 1) Studies examining interventions whose intended impact was to increase/decrease fertility tended to include/ignore QALYs that could accrue to offspring; 2) Studies examining interventions with indirect effects on fertility tended to ignore such QALYs; 3) Studies often avoided the issue by reporting incremental costs per outcome (e.g., birth defects avoided) as opposed to per QALY gained; 4) Even within categories (e.g., interventions intended to increase fertility), studies took heterogeneous approaches.
Conclusion: Current practice of considering QALYs from current pregnancies and future fertility is inconsistent and frequently appears biased towards the interventions considered. As the Panel on Cost-Effectiveness in Health and Medicine updates its guidelines, making the practice of CEA more consistent is a priority. Analysts risk having their work dismissed as biased if consistent and transparent standards are not followed. The framework we have developed contributes to harmonizing methods in this respect.