Reconciling decision models with the real world. An application to anaemia of renal failure.

OBJECTIVE: The choice of evidence used in decision modelling of healthcare interventions divides analysts into 2 groups: (i) those who favour randomised clinical trial (RCT) data; and (ii) those who prefer 'real world' data. This preference may have serious consequences if the end result is to inform healthcare policy. This paper uses Medicare coverage of epoetin-alpha [erythropoietin (EPO)] as a case study to illustrate a technique which can be used to overcome some of the bias inherent in RCT data while avoiding some of the common pitfalls associated with the use of observational data. DESIGN AND SETTING: Cost analysis of 2 treatments for anaemia of renal failure primarily in an outpatient setting is modelled in a decision tree. This method can be used to analyse healthcare interventions or policies in any setting. PATIENTS AND PARTICIPANTS: Patients with nontransplanted end-stage renal disease (ESRD) who received either EPO or blood transfusion for treatment of anaemia at any time during the 1-year study period (July 1989 to June 1990) were included in the sample. METHODS: Outcome effects in the natural setting are decomposed into 2 parts: a treatment effect and a population effect. This is then extended to the special case of policy analysis. Logistic and multiple regression are used to estimate branch probabilities and payoffs, respectively, for 2 treatment options. MAIN OUTCOME MEASURES AND RESULTS: Under standard methods of decision analysis, an increase of $US7032 per patient following EPO coverage is observed. With the decomposition technique, the policy effect is estimated to be less, $US6172, the difference coming from the population effect. CONCLUSIONS: Failure to remove population effects from observed outcome effects may lead to biased decision-making. Although not directly observable, the population effect can be imputed from secondary data. The decomposition and imputting technique allows for a more meaningful interpretation of the results for the purpose of policy analysis.

Economic assessment of three antimicrobial therapies for uncomplicated urinary tract infection in women.

This retrospective cohort study used North Carolina Medicaid paid-claims data to assess clinical and economic outcomes of treatments for urinary tract infection (UTI). The study population comprised female Medicaid recipients, between 15 and 64 years of age, with a paid claim filed for a primary diagnosis of UTI or acute UTI from January 1 to June 30, 1994, who were treated with ciprofloxacin, nitrofurantoin, or trimethoprim/sulfamethoxazole (TMP/SMZ). Patients had follow-up for 6 months after the primary diagnosis. Patients who did not receive further treatment for UTI with 1 of the 3 drugs within 30 days after initial treatment were assumed to be cured. Costs were measured as the sum of reimbursements for UTI-related medical services and drug treatments. Outcomes for 409 patients were assessed. Cure rates of initial treatment with ciprofloxacin, nitrofurantoin, and TMP/SMZ were 81%, 88%, and 93%, respectively. Cost-effectiveness ratios of initial treatment with the 3 drugs were $150.80, $81.20, and $69.00, respectively. When efficacy rates generated from published randomized clinical studies were applied, cost-effectiveness ratios for the 3 drugs were $130.96, $86.17, and $72.00, respectively. A decision model of treatment pattern and associated costs is presented. Several patient variables indicate that the ciprofloxacin group included more severe cases of UTI than did the other groups. Study limitations, confounders, and future research suggestions are discussed. Our results show that treatment for >7 days results in a better cure rate regardless of the drug used than does treatment for < or =7 days and that TMP/SMZ is the most cost-effective of the 3 drugs for UTI or acute UTI.

Use of home health care by ESRD and Medicare beneficiaries.

The use of home health care (HHC) services among Medicare end stage renal disease (ESRD) enrollees remains an under-studied area. In this article, the authors report sociodemographic characteristics and patterns of HHC utilization by Medicare-covered ESRD patients. The authors found that those who were female, age 85 or over, diabetic, and residing in the New England or West South Central census divisions were more likely to use HHC services and were also more intensive users. Analysis of use patterns in such high-risk populations is necessary to ensure that health policy changes do not have unintended consequences for vulnerable patients.

Are state immunization programs effective? Implications for the children's immunization initiative.

Several states operate universal vaccine purchase (UVP) programs aimed at ensuring adequate immunization of children. Some of these programs have been in operation for many decades. Recently, there has been a great deal of interest in implementing a federal UVP program. It is not clear, however, that such programs can significantly increase immunization levels; many nonfinancial barriers to full immunization exist and would have to be addressed. This paper uses cross-sectional data at the state level to estimate the effect of state UVP programs on the immunization levels of preschool children. The results indicate that states with UVP programs do not have significantly higher immunization rates than do other states. Therefore, it is not likely that a federal UVP program would significantly affect immunization rates.