Health Services and Outcomes Research Methodology最新文献

While record linkage can expand analyses performable from survey microdata, it also incurs greater risk of privacy-encroaching disclosure. One way to mitigate this risk is to replace some of the information added through linkage with synthetic data elements. This paper describes a case study using the National Hospital Care Survey (NHCS), which collects patient records under a pledge of protecting patient privacy from a sample of U.S. hospitals for statistical analysis purposes. The NHCS data were linked to the National Death Index (NDI) to enhance the survey with mortality information. The added information from NDI linkage enables survival analyses related to hospitalization, but as the death information includes dates of death and detailed causes of death, having it joined with the patient records increases the risk of patient re-identification (albeit only for deceased persons). For this reason, an approach was tested to develop synthetic data that uses models from survival analysis to replace vital status and actual dates-of-death with synthetic values and uses classification tree analysis to replace actual causes of death with synthesized causes of death. The degree to which analyses performed on the synthetic data replicate results from analysis on the actual data is measured by comparing survival analysis parameter estimates from both data files. Because synthetic data only have value to the degree that they can be used to produce statistical estimates that are like those based on the actual data, this evaluation is an essential first step in assessing the potential utility of synthetic mortality data.

Noncommunicable diseases (NCDs) are the leading causes of mortality and morbidity across the world and factors influencing global poverty and slowing economic development. We summarize how the potential power of real-world data (RWD) and real-world evidence (RWE) can be harnessed to help address the disease burden of NCDs at global, national, regional and local levels. RWE is essential to understand the epidemiology of NCDs, quantify NCD burdens, assist with the early detection of vulnerable populations at high risk of NCDs by identifying the most influential risk factors, and evaluate the effectiveness and cost-benefits of treatments, programs, and public policies for NCDs. To realize the potential power of RWD and RWE, challenges related to data integration, access, interoperability, standardization of analytical methods, quality control, security, privacy protection, and ethical standards for data use must be addressed. Finally, partnerships between academic centers, governments, pharmaceutical companies, and other stakeholders aimed at improving the utilization of RWE can have a substantial beneficial impact in preventing and managing NCDs.

An interrupted time series with a parallel control group (ITS-CG) design is a powerful quasi-experimental design commonly used to evaluate the effectiveness of an intervention, on accelerating uptake of useful public health products, and can be used in the presence of regularly collected data. This paper illustrates how a segmented Poisson model that utilizes general estimating equations (GEE) can be used for the ITS-CG study design to evaluate the effectiveness of a complex social accountability intervention on the level and rate of uptake of modern contraception. The intervention was gradually rolled-out over time to targeted intervention communities in Ghana and Tanzania, with control communities receiving standard of care, as per national guidelines. Two ITS GEE segmented regression models are proposed for evaluating of the uptake. The first, a two-segmented model, fits the data collected during pre-intervention and post-intervention excluding that collected during intervention roll-out. The second, a three-segmented model, fits all data including that collected during the roll-out. A much simpler difference-in-difference (DID) GEE Poisson regression model is also illustrated. Mathematical formulation of both ITS-segmented Poisson models and that of the DID Poisson model, interpretation and significance of resulting regression parameters, and accounting for different sources of variation and lags in intervention effect are respectively discussed. Strengths and limitations of these models are highlighted. Segmented ITS modelling remains valuable for studying the effect of intervention interruptions whether gradual changes, over time, in the level or trend in uptake of public health practices are attributed by the introduced intervention. Trial Registration: The Australian New Zealand Clinical Trials registry. Trial registration number: ACTRN12619000378123. Trial Registration date: 11-March-2019.

As healthcare costs continue to increase, studies assessing costs are becoming increasingly common, but researchers planning for studies that measure costs differences (savings) encounter a lack of literature or consensus among researchers on what constitutes "small" or "large" cost savings for common measures of resource use.  Other fields of research have developed approaches to solve this type of problem. Researchers measuring improvement in quality of life or clinical assessments have defined minimally important differences (MID) which are then used to define magnitudes when planning studies. Also, studies that measure cost effectiveness use benchmarks, such as cost/QALY, but do not provide benchmarks for cost differences. In a review of the literature, we found no publications identifying indicators of magnitude for costs. However, the literature describes three approaches used to identify minimally important outcome differences: (1) anchor-based, (2) distribution-based, and (3) a consensus-based Delphi methods. In this exploratory study, we used these three approaches to derive MID for two types of resource measures common in costing studies for: (1) hospital admissions (high cost); and (2) clinic visits (low cost). We used data from two (unpublished) studies to implement the MID estimation. Because the distributional characteristics of cost measures may require substantial samples, we performed power analyses on all our estimates to illustrate the effect that the definitions of "small" and "large" costs may be expected to have on power and sample size requirements for studies. The anchor-based method, while logical and simple to implement, may be of limited value in cases where it is difficult to identify appropriate anchors. We observed some commonalities and differences for the distribution and consensus-based approaches, which require further examination. We recommend that in cases where acceptable anchors are not available, both the Delphi and the distribution-method of MID for costs be explored for convergence.