Cost has become an important outcome in health services research. It can be used not only as a measure for health care spending but also as a measure for a part of health care value. Given ever-increasing rising health care expenditure, the value of health care should include not only traditional measures, such as mortality and morbidity, but also the cost of health care. Due to a limited resource, a new treatment with a slightly better efficacy but much higher cost than an existing treatment may not be a choice of a treatment for a patient. Hence, it is important to be able to approximately analyze cost data. However, appropriately analyzing health care costs may be hindered by special distribution features of cost data, including skewness, zero values, clusters, heteroscedasticity, and multimodality.
Over the decades, various methods have been proposed to address these features. This chapter would be devoted in introducing methods that are able to provide relatively trustworthy results with acceptable efficiency, covering topics on mean inference, regression, and prediction.
This is a preview of subscription content, log in to check access.
Ai C, Norton EC. Standard errors for the retransformation problem with heteroscedasticity. J Health Econ. 2000;19(5):697–718.PubMedCrossRefGoogle Scholar
Aitchison J. On the distribution of a positive random variable having a discrete probability mass at the origin. J Am Stat Assoc. 1955;50(271):901–8.Google Scholar
Blough DK, Madden CW, Hornbrook MC. Modeling risk using generalized linear models. J Health Econ. 1999;18(2):153–71.PubMedCrossRefGoogle Scholar
Krishnamoorthy K, Mathew T. Inferences on the means of lognormal distributions using generalized p-values and generalized confidence intervals. J stat plann infer. 2003;115(1):103–21.CrossRefGoogle Scholar
Land CE. An evaluation of approximate confidence interval estimation methods for lognormal means. Technometrics. 1972;14(1):145–58.CrossRefGoogle Scholar
Manning WG, Mullahy J. Estimating log models: to transform or not to transform? J Health Econ. 2001;20(4):461–94.PubMedCrossRefGoogle Scholar
Manning WG, Basu A, Mullahy J. Generalized modeling approaches to risk adjustment of skewed outcomes data. J Health Econ. 2005;24(3):465–88.PubMedCrossRefGoogle Scholar
Owen WJ, DeRouen TA. Estimation of the mean for lognormal data containing zeroes and left-censored values, with applications to the measurement of worker exposure to air contaminants. Biometrics. 1980;36(4):707–19. ISSN 0006341X. URL http://www.jstor.org/stable/2556125CrossRefGoogle Scholar
Seber GAF, Lee AJ. Linear regression analysis, vol. volume 936. Hoboken: Wiley; 2012.Google Scholar
Zhou XH, Gao S. One-sided confidence intervals for means of positively skewed distributions. Am Stat. 2000:100–4.Google Scholar
Zhou XH, Tu W. Comparison of several independent population means when their samples contain log-normal and possibly zero observations. Biometrics. 1999;55(2):645–51.CrossRefGoogle Scholar
Zhou XH, Tu W. Confidence intervals for the mean of diagnostic test charge data containing zeros. Biometrics. 2000;56(4):1118–25.PubMedCrossRefGoogle Scholar
Zhou XH, Lin H, Johnson E. Non-parametric heteroscedastic transformation regression models for skewed data with an application to health care costs. J R Stat Soc Ser B Stat Methodol. 2008;70(5):1029–47.CrossRefGoogle Scholar