Cost of arthritis: a systematic review of methodologies used for direct costs

Abstract

A substantial amount of healthcare and costs are attributable to arthritis, which is a very common chronic disease. This paper presents the results of a systematic review of arthritis cost studies published from 2008 to 2013. MEDLINE, Embase, EconLit databases were searched, as well as governmental and nongovernmental organization websites. Seventy-one reports met the inclusion/exclusion criteria, and 24 studies were included in the review. Among these studies, common methods included the use of individual-level data, bottom-up costing approach, use of both an arthritis group and a control group to enable incremental cost computation of the disease, and use of regression methods such as generalized linear models and ordinary least squares regression to control for confounding variables. Estimates of the healthcare cost of arthritis varied considerably across the studies depending on the study methods, the form of arthritis and the population studied. In the USA, for example, the estimated healthcare cost of arthritis ranged from $1862 to $14,021 per person, per year. The reviewed study methods have strengths, weaknesses and potential improvements in relation to estimating the cost of disease, which are outlined in this paper. Caution must be exercised when these methods are applied to cost estimation and monitoring of the economic burden of arthritis.

Keywords:

• Aging
• arthritis
• economic burden
• generalized linear models
• healthcare cost
• incremental cost
• individual-level data
• quatile regression
• review

Financial & competing interests disclosure

TKT Lo was supported by the University of Newcastle through research scholarships. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Key issues

• Estimates of the cost of arthritis vary across studies, and are not only affected by differences in epidemiological case mix (e.g. forms of arthritis, localized vs. generalized arthritis, age and gender groups, and range of severity) of the sample but also the method(s) these studies employ.

• Common methods to measure the direct cost of arthritis include: use of individual-level data and bottom-up costing approach, health system data for healthcare utilization and cost information, include only arthritis-related costs in individuals with arthritis, and regression methods such as the generalized linear models to adjust for confounding variables.

• It is recommended that investigators: explore the wealth of information (e.g. healthcare utilization and adjusting/confounding variables) in health survey data; include a control group and all-cause costs to capture the full cost of arthritis; use a more versatile regression method (e.g. quantile regression) to estimate the cost at the median (and/or other percentiles); and monitor the cost of arthritis over time as it may change due to demographic and non-demographic factors.

• Traditional regression models (e.g. ordinary least squares regression and generalized linear models), that assume a single rate of change to accurately describe the relationship between healthcare costs and the explanatory variable across the entire cost distribution, may give biased results. It is recommended that the association at multiple quantiles of the cost (i.e. representing patients at different levels of healthcare consumption) be assessed using quantile regression.

ORCID

TKT Lo http://orcid.org/0000-0002-5925-9549 Lynne Parkinson http://orcid.org/0000-0001-9433-9555 Michelle Cunich http://orcid.org/0000-0002-0163-3665 Julie Byles http://orcid.org/0000-0002-3984-6877