Comparing SSRI Treatment Costs for Depression Using Retrospective Claims Data: The Role of Nonrandom Selection and Skewed Data

Background and Objectives: Since conventional randomized clinical trials often do not reflect the real world circumstances of prescribing behavior and patient outcomes, the use of retrospective administrative claims databases (RACD) has become more common in treatment cost comparisons among alternative pharmaceutical compounds. Several recent RACD studies have compared treatment costs for depressed patients prescribed SSRIs such as fluoxetine, sertraline and paroxetine. These cost comparisons have reached mixed conclusions. To begin to explain and reconcile the mixed SSRI cost comparison evidence, we undertake a variety of alternative multivariate analyses using a publicly available RACD.
Methods and Data: The 1995 to 1996 data encompasses a time period when all three SSRIs had become well-established agents. We report and compare results from multivariate linear regressions, logistic regressions, ordered probits and sample selectivity models, and examine robustness when adjustments are made for outlier observations and skewed distributions.
Results and Conclusions: While choice of initial SSRI is nonrandom, the effect of sample selectivity on total depression-related and total health care expenditure is neutral across SSRIs. Although most cost measures are numerically greatest for fluoxetine, depression-related outpatient and hospitalization costs do not significantly differ by choice of initial SSRI. These findings are robust to alternative assumptions, specifications, and procedures. Antidepressant medication costs, however, are significantly higher when fluoxetine is the initial SSRI rather than sertraline or paroxetine, reflecting the larger proportion of fluoxetine patients prescribed a daily dosage of two or more capsules. Both total depression-related and total health care log–transformed costs are significantly lower for sertraline than fluoxetine.     

Testing and Correcting for Non-Random Selection Bias due to Censoring: An Application to Medical Costs

Censoring is a common problem with medical cost data. Methods from traditional survival analysis are not directly applicable to estimate medical costs since patients accumulate costs with different rate functions over time, leading to negatively biased estimates. Heckman's two-step estimator results in large variances when identical explanatory variables that influence selection are included in the structural equation, i.e. when there are no exclusion restrictions. This paper provides a systematic treatment of the correction for nonrandom sample selection bias of medical cost data where the selection rule is described by a censored regression model. The proposed method first uses the duration of time a patient is tracked for the selection, rather than a binary variable, namely whether or not the duration is censored. Second, using Tobit residuals instead of the inverse Mills ratio in the structural equation allows us to decrease large variances introduced by the Heckman model when there are no exclusion restrictions. We show that the resulting estimators are consistent and asymptotically normal. Simulation studies confirmed our results. Moreover, we derive a simple test to determine possible sample selection bias due to censoring. Data from a study on the medical cost of breast, prostate, colon, and lung cancer is used as an application of the method.     

Depression in Public Community Long-Term Care: Implications for Intervention Development

The objective of this paper is to increase understanding of geriatric depression in the public community long-term care system to guide intervention development. Protocols included screening 1,170 new clients of a public community long-term care agency and interviewing all clients with major, dysthymia, or subthreshold depression (n = 299) and a randomly selected subset of nondepressed older adults (n = 315) at baseline, 6-month, and 1 year. Six percent had major depression, one-half of a percent had dysthymia only, and another 19% had subthreshold depression. Over the year observation period, 40% were persistently depressed; 32% were assessed as depressed only at the first observation; and the remainder was intermittently depressed. There were high levels of comorbid medical, functional, and psychosocial conditions. Mental health service use was low, and clients reported attitudinal and other barriers to depression treatment. Findings suggest the need for universal screening for depression with some strategies for triaging the most severely and persistently depressed for treatment. Although there will be challenges to the development of depression interventions, the public community long-term care system has high potential to assist vulnerable older adults receive help with depression. Presented at Improving Chronic Care Quality Conference, Columbia, Missouri, September. 2004.     

Selection and bias—Two hostile brothers

We consider the situation where in a first stage of a clinical trial several treatments are compared with a single control and the ‘best’ treatment(s) are selected in an interim analysis to be carried on to the second stage. We quantify the mean bias and mean square error of the conventional estimates after selection depending on the number of treatments and the selection time during the trial. The cases without or with reshuffling the planned sample size of the dropped treatments to the selected ones are investigated. The mean bias shows very different patterns depending on the selection rule and the unknown parameter values. We stress the fact that the quantification of the bias is possible only in designs with planned adaptivity where the design allows reacting to new evidence, but the decision rules are laid down in advance. Finally, we calculate the mean bias which arises in a simple but influential regulatory selection rule, to register a new medical therapy only when two pivotal trials have both proven an effect by a statistical test. Copyright © 2009 John Wiley & Sons, Ltd.     

Comparing Methods for Estimating Direct Costs of Adverse Drug Events

Objectives

To estimate how direct health care costs resulting from adverse drug events (ADEs) and cost distribution are affected by methodological decisions regarding identification of ADEs, assigning relevant resource use to ADEs, and estimating costs for the assigned resources.

Good Research Practices for Measuring Drug Costs in Cost-Effectiveness Analyses: A Managed Care Perspective: The ISPOR Drug Cost Task Force Report—Part III

Objectives:  The objective of this report is to provide guidance and recommendations on how drug costs should be measured for cost-effectiveness analyses conducted from the perspective of a managed care organization (MCO).
Methods:  The International Society for Pharmacoeconomics and Outcomes Research (ISPOR) Task Force on Good Research Practices—Use of Drug Costs for Cost Effectiveness Analysis (DCTF) was appointed by the ISPOR Board of Directors. Members were experienced developers or users of CEA models. The DCTF met to develop core assumptions and an outline before preparing a draft report. They solicited comments on drafts from external reviewers and from the ISPOR membership at ISPOR meetings and via the ISPOR Web site.
Results:  The cost of a drug to an MCO equals the amount it pays to the dispenser for the drug's ingredient cost and dispensing fee minus the patient copay and any rebates paid by the drug's manufacturer. The amount that an MCO reimburses for each of these components can differ substantially across a number of factors that include type of drug (single vs. multisource), dispensing site (retail vs. mail order), and site of administration (self-administered vs. physician's office). Accurately estimating the value of cost components is difficult because they are determined by proprietary and confidential contracts.
Conclusion:  Estimates of drug cost from the MCO perspective should include amounts paid for medication ingredients and dispensing fees, and net out copays, rebates, and other drug price reductions. Because of the evolving nature of drug pricing, ISPOR should publish a Web site where current DCTF costing recommendations are updated as new information becomes available.     

Self-Rated Health and Long-Term Prognosis of Depression

PURPOSE

Indicators of prognosis should be considered to fully inform clinical decision making in the treatment of depression. This study examines whether self-rated health predicts long-term depression outcomes in primary care.

METHODS

Our analysis was based on the first 5 years of a prospective 10-year cohort study underway since January 2005 conducted in 30 randomly selected Australian primary care practices. Participants were 789 adult patients with a history of depressive symptoms. Main outcome measures include risks, risk differences, and risk ratios of major depressive syndrome (MDS) on the Patient Health Questionnaire.

RESULTS

Retention rates during the 5 years were 660 (84%), 586 (74%), 560 (71%), 533 (68%), and 517 (66%). At baseline, MDS was present in 27% (95% CI, 23%–30%). Cross-sectional analysis of baseline data showed participants reporting poor or fair self-rated health had greater odds of chronic illness, MDS, and lower socioeconomic status than those reporting good to excellent self-rated health. For participants rating their health as poor to fair compared with those rating it good to excellent, risk ratios of MDS were 2.10 (95% CI, 1.60–2.76), 2.38 (95% CI, 1.77–3.20), 2.22 (95% CI, 1.70–2.89), 1.73 (95% CI, 1.30–2.28), and 2.15 (95% CI, 1.59–2.90) at 1, 2, 3, 4, and 5 years, after accounting for missing data using multiple imputation. After adjusting for age, sex, multimorbidity, and depression status and severity, self-rated health remained a predictor of MDS up to 5 years.

CONCLUSIONS

Self-rated health offers family physicians an efficient and simple way to identify patients at risk of poor long-term depression outcomes and to inform therapeutic decision making.     

Long-Term Medical Costs and Life Expectancy of Acute Myeloid Leukemia: A Probabilistic Decision Model

BackgroundAcute myeloid leukemia (AML) can be diagnosed at any age and treatment, which can be given with supportive and/or curative intent, is considered expensive compared with that for other cancers. Despite this, no long-term predictive models have been developed for AML, mainly because of the complexities associated with this disease.ObjectiveThe objective of the current study was to develop a model (based on a UK cohort) to predict cost and life expectancy at a population level.MethodsThe model developed in this study combined a decision tree with several Markov models to reflect the complexity of the prognostic factors and treatments of AML. The model was simulated with a cycle length of 1 month for a time period of 5 years and further simulated until age 100 years or death. Results were compared for two age groups and five different initial treatment intents and responses. Transition probabilities, life expectancies, and costs were derived from a UK population-based specialist registry—the Haematological Malignancy Research Network (www.hmrn.org).ResultsOverall, expected 5-year medical costs and life expectancy ranged from £8,170 to £81,636 and 3.03 to 34.74 months, respectively. The economic and health outcomes varied with initial treatment intent, age at diagnosis, trial participation, and study time horizon. The model was validated by using face, internal, and external validation methods. The results show that the model captured more than 90% of the empirical costs, and it demonstrated good fit with the empirical overall survival.ConclusionsCosts and life expectancy of AML varied with patient characteristics and initial treatment intent. The robust AML model developed in this study could be used to evaluate new diagnostic tools/treatments, as well as enable policy makers to make informed decisions.     

Price elasticity and pharmaceutical selection: the influence of managed care

State Medicaid programs are turning increasingly to managed care to control expenditures, although the types of managed care programs in use have changed dramatically. Little is known about the influence of the shifting Medicaid managed care arena on treatment decisions. This paper investigates factors affecting the selection of treatments for depression by providers participating in either of two Medicaid managed care programs. Of particular interest is the influence of medication price on the choice of treatment, since one vehicle through which managed care organizations can reduce total expenditures is by increasing the price sensitivity of participating providers. We take a new approach by phrasing the problem as a discrete choice, using a nested multinomial logit model for the analyses. Contrary to earlier literature, we find some evidence that physicians in both programs do take price into consideration when selecting among treatment options. HMO providers in particular demonstrate increased price sensitivity in the two most commonly prescribed categories of antidepressants.     

用于治疗抑郁症的医疗器械的现状和前景

抑郁症是目前发病率最高的精神心理疾病,是重要的公共卫生问题,其病因涉及方面多,具体病理生理机制仍未明确。此阶段抑郁症的主要治疗方法为药物治疗、心理治疗和物理治疗三种方式,临床应用时通常两种或三种方法相结合。其中,物理治疗因其副作用低,患者依从性好,无需大量医疗资源等特点具有十分重要的意义,各类医疗器械应运而生,文章主要介绍了目前较为成熟的几种器械治疗方法,并对其进行了展望。     

The Long-Term Effects of Maternal Depression: Early Childhood Physical Health as a Pathway to Offspring Depression

PurposeCross-sectional and retrospective studies have highlighted the long-term negative effects of maternal depression on offspring physical, social, and emotional development, but longitudinal research is needed to clarify the pathways by which maternal depression during pregnancy and early childhood affects offspring outcomes. The current study tested one developmental pathway by which maternal depression during pregnancy might negatively impact offspring mental health in young adulthood, via poor physical health in early childhood.MethodsThe sample consisted of 815 Australian youth and their mothers who were followed for 20 years. Mothers reported on their own depressive symptoms during pregnancy and offspring early childhood. Youth completed interviews about health-related stress and social functioning at age 20 years, and completed a questionnaire about their own depressive symptoms 2 to 5 years later.ResultsPath analysis indicated that prenatal maternal depressive symptoms predicted worse physical health during early childhood for offspring, and this effect was partially explained by ongoing maternal depression in early childhood. Offspring poor physical health during childhood predicted increased health-related stress and poor social functioning at age 20. Finally, increased health-related stress and poor social functioning predicted increased levels of depressive symptoms later in young adulthood. Maternal depression had a significant total indirect effect on youth depression via early childhood health and its psychosocial consequences.ConclusionsPoor physical health in early childhood and its effects on young adults' social functioning and levels of health related stress is one important pathway by which maternal depression has long-term consequences for offspring mental health.     

Estimates of the Lifetime Direct Costs of Treatment for Metastatic Breast Cancer

OBJECTIVE: Breast cancer remains the highest incident cancer among females in the United States and previous research suggests that a considerable portion of patients will eventually progress to the metastatic phase of the disease. This paper provides the first estimate of the lifetime direct costs of treating metastatic disease for one annual diagnostic cohort of breast cancer patients. METHODS: Incidence rates were combined with US population counts to estimate the number of breast cancer cases diagnosed in 1994. Estimates of progression to metastatic disease (from Canadian provincial cancer registry data), costs of care (derived from patients' claims histories), survival (from SEER data), and national mortality rates (from US Census Bureau) were integrated, using Statistics Canada's Population Health Model (POHEM) to calculate lifetime costs. RESULTS: This study estimates that more than 40% of the women diagnosed with breast cancer will progress to metastatic disease. On average, women with metastatic disease are expected to live 3 years and to incur direct treatment costs of approximately dollar 60,000 per case, resulting in a total lifetime cost for the cohort of almost dollar 4.2 billion. CONCLUSIONS: The high rate of recurrence of breast cancer argues for the development of interventions that can prevent or delay the onset of metastatic disease. These estimates of lifetime costs and the methodology on which they are based can be used to evaluate the cost-effectiveness of such secondary prevention strategies. These estimates also can serve as a benchmark against which the lifetime costs of treating other diseases can be assessed.     

Managed Care Contracting and Medical Care for the Uninsured: Untangling Selection from Production

Capitated payment systems used by managed care plans potentially reduce the financial earnings that providers use to cross-subsidize care for the uninsured. Providers that value uninsured care highly, however, may improve production efficiency in response to capitation and thereby maintain or expand uninsured care. Measuring the effect of capitation is complicated by the endogenous selection and censoring processes that determine a provider's involvement in capitated payment systems. This study compares three alternative methods for modeling the effect of capitation—a single-equation generalized estimating equations (GEE) model, a two-stage tobit model, and a discrete factor model using full-information maximum likelihood estimation. Models are estimated using panel data on all U.S. federally-funded community health centers operating during 1992 through 1996 (3185 center-years). Single-equation estimates appear positively biased due to capitation selection and censoring. Estimates from two-stage and discrete factor models show no evidence that capitation adversely affects uninsured care after controlling for this bias. Discrete factor estimates are substantially more precise than two-stage estimates, and indicate that uninsured care actually increases modestly in response to capitation. Discrete factor models, though computationally intensive, offer the advantages of consistency and precision over other econometric models for studies involving censored endogenous variables and selection bias.     

Price elasticities of demand for curative health care with control for sample selectivity on endogenous illness: an analysis for Sri Lanka

Estimation of demand for health care with samples of only the ill may bias estimates. Additionally, the lack of exogenous information, especially distance, about the alternative care providers causes omitted variable problems. This paper alleviates both problems through geographic mapping of facility information to individuals, combined with joint estimation of illness (health production) and health care demand. The joint estimation full sample demand results are compared to those from one equation estimation for only the ill sample. The results indicate that the selectivity problem is significant, but that for this sample the magnitude of the bias on the price coefficient is small. © 1998 John Wiley & Sons, Ltd.