Statistical methods for cost-effectiveness analyses

[Cost-effectiveness of diagnostic strategies of severe bacterial infection in infants with fever without a source]

INTRODUCTION

Serious bacterial infections in infants under 2-years-of-age with fever without a source remains, despite advances in vaccination, a matter of concern for both physicians and parents. Having cost-effectiveness information is relevant to guide decision making in clinical practice in this scenario. Objective: To determine the cost-effectiveness of four different strategies of screening in Argentina for serious bacterial infection in children presenting with fever without a source. Materials and methods: We designed a decision tree to model a hypothetical cohort of 10,000 children with fever without a source. We compared the incremental cost-effectiveness of four strategies to detect serious bacterial infection: Rochester criteria + C reactive protein test, Rochester criteria + procalcitonin test, Rochester criteria, and expectant observation. Results: Rochester criteria + C reactive protein test was the most cost-effective strategy with USD$ 784 for each correctly diagnosed case versus USD$ 839 of Rochester criteria + procalcitonin test, USD$ 1,116 of expectant observation or USD$ 1,193 of Rochester criteria. When the probability of serious bacterial infections was equal or less than 14%, the strategy of choice was expectant observation. Conclusions: The Rochester criteria + C reactive protein test was the most cost-effective strategy to detect serious bacterial infection in one to three months old children with fever without a source. However, in low risk settings for such infection, the strategy of choice is expectant observation.

Quality of Well-being profiles followed paths of health status change at micro- and meso-levels in trauma patients

OBJECTIVE

The objective of this study is to analyze Quality of Well-being Scale scores and profiles tracing Trauma Recovery Project (TRP) patient scores over time.

STUDY DESIGN AND SETTING

A total of 787 TRP patients had complete preinjury and injury day data. Of these 787, 574 patients were followed up 6 months after hospital release. Analyses include persons with head injury vs. long bone and pelvic injury.

RESULTS

Paired t-tests found significant differences for scores between each measurement point. Means analyses found significant variation on first day of hospitalization vs. 6-month recovery scores by injury site--head worse off than long bone and pelvic injury at first, but becoming better off 6 months after release from hospital. These effects were traced to specific symptom/problem complexes and functional limitations.

CONCLUSION

Examination of such profiles can add significant information about health implications not obvious from overall scores. The size and direction of such contributions to overall scores may be reliably traced over time.

Assessing physicians' use of treatment algorithms: Project IMPACTS study design and rationale

Effective treatments for major depressive disorder have been available for 35 years, yet inadequate pharmacotherapy continues to be widespread leading to suboptimal outcomes. Evidence-based medication algorithms have the potential to bring much-needed improvement in effectiveness of antidepressant treatment in "real-world" clinical settings. Project IMPACTS (Implementation of Algorithms using Computerized Treatment Systems) addresses the critical question of how best to facilitate integration of depression treatment algorithms into routine care. It tests an algorithm implemented through a computerized decision support system using a measurement-based care approach for depression against a paper-and-pencil version of the same algorithm and non-algorithm-based, specialist-delivered usual care. This paper reviews issues related to the Project IMPACTS study rationale, design, and procedures. Patient outcomes include symptom severity, social and work function, and quality of life. The economic impact of treatment is assessed in terms of health care utilization and cost. Data collected on physician behavior include degree of adherence to guidelines and physician attitudes about the perceived utility, ease of use, and self-reported effect of the use of algorithms on workload. Novel features of the design include a two-tiered study enrollment procedure, which initially enroll physicians as subjects, and then following recruitment of physicians, enrollment of subjects takes place based initially on an independent assessment by study staff to determine study eligibility. The study utilizes brief, easy-to-use symptom severity measures that facilitate physician decision making, and it employs a validated, phone-based, follow-up assessment protocol in order to minimize missing data, a problem common in public sector and longitudinal mental health studies. IMPACTS will assess the success of algorithm implementation and subsequent physician adherence using study-developed criteria and related statistical approaches. These new procedures and data points will also allow a more refined assessment of algorithm-driven treatment in the future.

Service costs for women with co-occurring disorders and trauma

Several aspects of costs related to health care and other service use at 6-month follow-up are presented for women with co-occurring mental health and substance abuse disorders with histories of physical and/or sexual abuse receiving comprehensive, integrated, trauma-informed and consumer/survivor/recovering person-involved interventions (n = 1023) or usual care (n = 983) in a nine-site quasi-experimental study. Results show that, controlling for pre-baseline use, there are no significant differences in total costs between participants in the intervention condition and those in the usual care comparison condition, either from a governmental (avg. US dollars 13,500) or Medicaid reimbursement perspectives (avg. just over US dollars 10,000). When combined with clinical outcomes analyzed in other works in this issue by Cocozza et al. (2005) and Morrissey et al. (2005), which favored the intervention sites, these cost findings indicate that the treatment intervention services are cost-effective as compared with the usual care received by women at the comparison sites.

Measuring uncertainty in complex decision analysis models

Prediction models used in support of clinical and health policy decision making often need to consider the course of a disease over an extended period of time, and draw evidence from a broad knowledge base, including epidemiologic cohort and case control studies, randomized clinical trials, expert opinions, and more. This paper is a brief introduction to these complex decision models, their relation to Bayesian decision theory, and the tools typically used to describe the uncertainties involved. Concepts are illustrated throughout via a simplified tutorial.

Affordability and cost-effectiveness: decision-making on the cost-effectiveness plane

Much recent research interest has focused on handling uncertainty in cost-effectiveness analysis and in particular the calculation of confidence intervals for incremental cost-effectiveness ratios (ICERs). Problems of interpretation when ICERs are negative have led to two important and related developments: the use of the net-benefit statistic and the presentation of uncertainty in cost-effectiveness analysis using acceptability curves. However, neither of these developments directly addresses the problem that decision-makers are constrained by a fixed-budget and may not be able to fund new, more expensive interventions, even if they have been shown to represent good value for money. In response to this limitation, the authors introduce the 'affordability curve' which reflects the probability that a programme is affordable for a wide range of threshold budgets. The authors argue that the joint probability an intervention is affordable and cost-effective is more useful for decision-making since it captures both dimensions of the decision problem faced by those responsible for health service budgets.

Statistical cost-effectiveness analysis of two treatments based on net health benefits

Statistical methods for cost-effectiveness analysis (CEA) for two treatments that mimic the deterministic optimal rules of CEA are presented. In these rules the objective is to determine the treatment with the maximal effectiveness whose unit cost is less than an amount, lambda, that a decision-maker is willing to pay (WTP). This is accomplished by identifying the treatment with the largest positive net health benefit (NHB), which is a function of lambda, while controlling the familywise error rate both when the WTP value is given and when it is unspecified. Fieller's theorem is used to determine a region of WTP values where the NHBs of the treatments are not distinguishable. For each lambda outside of the confidence region, the larger treatment is identified. A newly developed one-tailed analogue of Fieller's theorem is used to determine the WTP values where a treatment's NHB is positive. The situation in which both treatments are experimental is distinguished from the case where one of the treatments is usual care. The one-tailed confidence region is used in the latter case to obtain the lambda values where the NHBs are not different, and determining the region of positivity of the NHBs may be unnecessary. An example is presented in which the cost-effectiveness of two antipsychotic treatments is evaluated.

Covariation of physical and mental symptoms across illnesses: results of a factor analytic study

OBJECTIVE

Chronic illnesses are associated with reports of symptoms, problems, and dysfunction along multiple dimensions. To determine if the dimensionality is disease-specific and whether physical and emotional symptoms are concomitant and inseparable aspects of the illness experience, we present a factor analysis of symptom and problem reports from five different chronic conditions.

METHOD

People with five different conditions participated in this study: multiple sclerosis (MS) (n = 263), non-insulin-dependent diabetes mellitus (n = 420), nonhead nonneck injury trauma (n = 852), and a group of terminal patients comprised of acquired immune deficiency syndrome (AIDS) (n = 99) and cancer (n = 74) patients. Participants were asked to complete the Quality of Well-Being Scale (QWB) and symptom items from the QWB were factor analyzed.

RESULTS

Both within each condition and across conditions, two factors accounted for the majority of the explained variance and could be described as an Observable Limitations factor and a Subjective Symptoms factor.

CONCLUSIONS

Our factor analyses suggest that physical and emotional symptoms covary and are common to different types of illness.

Measuring costs of guideline-driven mental health care: the Texas Medication Algorithm Project

BACKGROUND: Algorithms describe clinical choices to treat a specific disorder. To many, algorithms serve as important tools helping practitioners make informed choices about how best to treat patients, achieving better outcomes more quickly and at a lower cost. Appearing as flow charts and decision trees, algorithms are developed during consensus conferences by leading experts who explore the latest scientific evidence to describe optimal treatment for each disorder. Despite a focus on "optimal" care, there has been little discussion in the literature concerning how costs should be defined and measured in the context of algorithm-based practices. AIMS OF THE STUDY: This paper describes the strategy to measure costs for the Texas Medication Algorithm project, or TMAP. Launched by the Texas Department of Mental Health and Mental Retardation and the University of Texas Southwestern Medical Center at Dallas, this multi-site study investigates outcomes and costs of medication algorithms for bipolar disorder, schizophrenia and depression. METHODS: To balance costs with outcomes, we turned to cost-effectiveness analyses as a framework to define and measure costs. Alternative strategies (cost-benefit, cost-utility, cost-of-illness) were inappropriate since algorithms are not intended to guide resource allocation across different diseases or between health- and non-health-related commodities. "Costs" are operationalized consistent with the framework presented by the United States Public Health Service Panel on Cost Effectiveness in Medicine. Patient specific costs are calculated by multiplying patient units of use by a unit cost, and summing over all service categories. Outpatient services are counted by procedures. Inpatient services are counted by days classified into diagnosis groups. Utilization information is derived from patient self-reports, medical charts and administrative file sources. Unit costs are computed by payer source. Finally, hierarchical modeling is used to describe how costs and effectiveness differ between algorithm-based and treatment-as-usual practices. DISCUSSION: Cost estimates of algorithm-based practices should (i) measure opportunity costs, (ii) employ structured data collection methods, (iii) profile patient use of both mental health and general medical providers and (iv) reflect costs by payer status in different economic environments. IMPLICATION FOR HEALTH CARE PROVISION AND USE: Algorithms may help guide clinicians, their patients and third party payers to rely on the latest scientific evidence to make treatment choices that balance costs with outcomes. IMPLICATION FOR HEALTH POLICIES: Planners should consider consumer wants and economic costs when developing and testing new clinical algorithms. IMPLICATIONS FOR FURTHER RESEARCH: Future studies may wish to consider similar methods to estimate costs in evaluating algorithm-based practices.

Power and sample size in cost-effectiveness analysis

For resource allocation under a constrained budget, optimal decision rules for mutually exclusive programs require that the treatment with the highest incremental cost-effectiveness ratio (ICER) below a willingness-to-pay (WTP) criterion be funded. This is equivalent to determining the treatment with the smallest net health cost. The designer of a cost-effectiveness study needs to select a sample size so that the power to reject the null hypothesis, the equality of the net health costs of two treatments, is high. A recently published formula derived under normal distribution theory overstates sample-size requirements. Using net health costs, the authors present simple methods for power analysis based on conventional normal and on nonparametric statistical theory.

On the use of survival analysis techniques to estimate medical care costs

Measurement of treatment costs is important in the evaluation of medical interventions. Accurate cost estimation is problematic, when cost records are incomplete. Methods from the survival analysis literature have been proposed for estimating costs using available data. In this article, we clarify assumptions necessary for validity of these techniques. We demonstrate how assumptions needed for valid survival analysis may be violated when these methods are applied to cost estimation. Our observations are confirmed through simulations and empirical data analysis. We conclude that survival analysis approaches are not generally appropriate for the analysis of medical costs and review several valid alternatives.

Methodological and statistical issues of quality of life (QoL) and economic evaluation in cancer clinical trials: report of a workshop

In recent years, quality of life (QoL) and economic evaluations have become increasingly important as additional outcome measures in cancer clinical trials. However, both fields of research are relatively new and in need of finding solutions to a substantial number of specific methodological problems. This paper reports on the proceedings of a symposium aimed at summarising and discussing some of the most contentious methodological and statistical issues in QoL and economic evaluations. In addition, possible solutions are indicated and the most pertinent areas of research are identified. Issues specific to QoL evaluations that are addressed include clinically meaningful changes in QoL scores; how to analyse QoL data and to handle missing and censored data and integration of length of life and QoL outcomes. Issues specific to economic evaluations are the advantages and disadvantages of various outcome measures; statistical methods to analyse economic data and choice of decision criteria and analytical perspective. How to perform QoL and economic evaluations in large and simple trials and whether the gap between QoL and utility measures can be bridged are also discussed.

Net health benefits: a new framework for the analysis of uncertainty in cost-effectiveness analysis

In recent years, considerable attention has been devoted to the development of statistical methods for the analysis of uncertainty in cost-effectiveness (CE) analysis, with a focus on situations in which the analyst has patient-level data on the costs and health effects of alternative interventions. To date, discussions have focused almost exclusively on addressing the practical challenges involved in estimating confidence intervals for CE ratios. However, the general approach of using confidence intervals to convey information about uncertainty around CE ratio estimates suffers from theoretical limitations that render it inappropriate in many situations. The authors present an alternative framework for analyzing uncertainty in the economic evaluation of health interventions (the "net health benefits" approach) that is more broadly applicable and that avoids some problems of prior methods. This approach offers several practical and theoretical advantages over the analysis of CE ratios, is straightforward to apply, and highlights some important principles in the theoretical underpinnings of CE analysis.

Assessing uncertainty in cost-effectiveness analyses: application to a complex decision model

A framework for quantifying uncertainty about costs, effectiveness measures, and marginal cost-effectiveness ratios in complex decision models is presented. This type of application requires special techniques because of the multiple sources of information and the model-based combination of data. The authors discuss two alternative approaches, one based on Bayesian inference and the other on resampling. While computationally intensive, these are flexible in handling complex distributional assumptions and a variety of outcome measures of interest. These concepts are illustrated using a simplified model. Then the extension to a complex decision model using the stroke-prevention policy model is described.