Testing for Diabetes Reduces Healthcare Costs
Based on a presentation by Judy Y. Chen, MD, MSHS, and Josh Marehbian, MPH. Dr Chen is Director of Research and Clinical Development, and Mr Marehbian is a consulting manager, Health Benchmarks/IMS Health, Woodland Hills, CA.
Ateam of researchers has created a decision-tree analysis of a pay-for-performance (P4P) program in diabetes care that yields a savings of $25 per patient over 2 years. The total savings for the 25,605 patients from an administrative claims database who were used in the analysis was more than $765,000.
The decision-tree model was created by a team at Health Benchmarks, IMS Health, Woodland Hills, California. The model was based on 4 decision points for each diabetic patient, considering the following questions:
- Was the patient's health plan using a P4P program?
- Was the patient's physician participating in the program?
- Did the patient receive at least 2 glycosylated hemoglobin (HbA1C) tests and 1 lipid test yearly?
- Did the patient have to be hospitalized because of complications from the diabetes?
The team calculated the probabilities for each of the 2 possible outcomes at each decision point based on data from 25,605 individuals whose healthcare provision was administered by a preferred provider organization in Hawaii between March 31, 2003, and April 21, 2005.
The baseline characteristics were similar in patients who were cared for by physicians participating in a P4P program and those who were not. There was a 91% probability of each patient seeing only program-participating physicians in 2003. Investigators also found, based on multivariate regression analyses, that patients who saw only program-participating physicians had a 61.5% chance of being tested for HbA1C and lipids, whereas those who saw nonparticipating physicians had a 52.7% of receiving the quality care (P <.01). Individuals who received HbA1C and lipid testing had a 31.1% chance of being hospitalized in the subsequent 2 years compared with a 38.8% chance for those who did not. The difference in cost per patient was $2641 versus $2665, respectively, which yielded a total savings of more than $675,000 to the health plan over the 2 years.
"We compared the hospitalization rates of diabetic patients who received at least the minimum number of sugar or cholesterol tests recommended by the American Diabetes Association for diabetes patients to those of patients who did not receive the minimum number. Testing is the first step to ensuring that patients have the appropriate blood sugar and cholesterol levels," noted Judy Y. Chen, MD, MSHS. "And in our analyses, we found that appropriate testing is consistently associated with less hospitalization, and hence with lower healthcare costs."
Cost-Analysis of Bortezomib Use at M.D. Anderson Cancer Center
Based on a presentation by Lincy Lal, PharmD, PhD, and Dwight Kloth, PharmD. Dr Lal is a pharmacoeconomic research specialist, Drug Use Policy and Pharmacoeconomics, M.D. Anderson Cancer Center, Houston, TX, and Dr Kloth is Director of Pharmacy, Fox Chase Cancer Center, Philadelphia, PA.
AUniversity of Texas M.D. Anderson Cancer Cen ter analysis of 1 year of bortezomib (Velcade, Millennium Pharmaceuticals) use revealed that the hospital staff treated nearly 6 times as many patients with the monoclonal antibody as they had expected, and spent nearly 3 times as much money as they had planned. The overall reimbursement-to-charge ratio was 53%, slightly lower than the projected 55% ratio.
Bortezomib is only the second medication on which the team has performed a full budget-impact analysis. The US Food and Drug Administration approv ed this protease inhibitor in 2003 for the treatment of multiple myeloma in patients who have failed 2 previous drug trials and for mantle-cell lymphoma in 2006.
The investigators performed an initial budget analysis in 2003 based on the number of patients they estimated would be given bortezomib. The Pharmacy & Therapeutics Committee at the hospital subsequently added bortezomib to the formulary for the thenapproved indication, noting that physicians should use their clinical discretion for off-label use. A postapproval budget-impact analysis was subsequently performed based on the number of patients treated outside of clinical trials between June 2006 and May 2007.
"These analyses did not affect the committee's decision—we felt it was necessary to add bortezomib to the formulary because all chemotherapy products that show efficacy are included in our formulary, regardless of budget impact," said coinvestigator Lincy Lal, PharmD, PhD. "This is a difference from other countries, such as the UK, where NICE [National Institute for Health and Clinical Excellence] makes decisions based on both efficacy and cost considerations. However, we are taking the initial steps in including this information in our discussions in the formulary management process."
The M.D. Anderson team had estimated that the hospital would treat 25 patients annually, for a total purchase cost of $414,974 after adjustment to 2007 dollars. However, in the study year, 140 patients were treated at the hospital with bortezomib at a cost of $8,676, for a total cost of $1,214,640. Overall, the reimbursement to charge ratio was 53.3%. Furthermore, 87% of the use (122/140) was for multiple myeloma, and patients received an average of 2 cycles of treatment rather than the 4 cycles recommended by the manufacturer.
Dwight Kloth, PharmD, commented that although M.D. Anderson does not yet include cost in its decisions, it is farther along this path than most other centers in the United States.
"I applaud M.D. Anderson's rigorous approach to doing these analyses, because they provide valuable data for the rest of us," said Dr Kloth. "We don't have the resources—that is, a pharmacoeconomics study group—to conduct such thorough analyses. In general, what my institution is going to continue to do is follow the available evidence as it evolves, but with an eye on the cost to the healthcare system and to the patients in terms of copays."
Repeat Testing Needed to Estimate US Health Expenditures
Based on a presentation by Aniket Kawatkar, MS, BPharm, and Michael Nichol, PhD. Mr Kawatkar is a MERCK and QSAD Centurion Fellow, and Dr Nichol is QSAD Centurion Professor in Pharmaceutical Sciences and professor of clinical pharmacy, Department of Pharmaceutical Economics and Policy, School of Pharmacy, University of Southern California, Los Angeles.
Not all econometric models are created equal when it comes to estimating the costs to the American economy of serious medical conditions. A new study shows clear differences between some of the most frequently used econometric models.
Aniket Kawatkar, MS, BPharm, and Michael Nichol, PhD, compared the performance of 11 different models in estimating the total direct medical expenditures associated with 7 common and costly diseases—diabetes, arthritis, cardiac disease, asthma, hypertension, stroke, and emphysema. The models included the ordinary least squares (OLS) on logtransformed and raw-scale expenditure, the generalized linear model with log link and 2-part versions of these models. Each model performed better or worse than the others in the tests the investigators subjected them to.
"People ignore these considerations and say, 'Model X is the most popular model in the literature, so that's the one I'll use,'" Mr Kawatkar said. "But none of these models dominates the others, so you've got to be careful in picking the model to use. You have to go through a series of tests to determine which model is the best one for your particular application."
The research team ran each of the models through 4 standard tests for robustness, and each model failed at least 1 of the tests. They then determined how each model dealt with problems, such as skewed distribution of expenditure among the population and people with either zero or very high expenditures. They found that while, for example, the OLS can deal with skewed distribution to some extent, it is not designed to deal with the high proportion of people with zero expenditures, something that is frequently encountered in costdata distribution.
"The strong influence of model choice on the total medical expenditure estimate underscores the importance of understanding the data-generation procedure before selection of the appropriate estimator," concluded Mr Kawatkar.