TABLE 7.2 Summary of research techniques used in surgical health services research. HSR uses a variety of both quantitative and qualitative techniques.This table provides an example of basic techniques utilized throughout the three major domains of HSR

by combining data generated from multiple studies in a scientifically rigorous way.

Intellectual Disciplines and Research Tools

Comparative Effectiveness Research

The Federal Coordinating Council for Comparative Effectiveness Research in 2009 defined comparative effectiveness research (CER) as the conduct and synthesis of research comparing the benefits and harms of different interventions and strategies to prevent, diagnose, treat, and monitor health conditions in “realworld” settings.1 This type of research seeks to understand which treatment approach is the most beneficial to patients outside of clinical trials. CER must compare at least two different approaches to care, either of which has the potential to be the optimal choice. CER can be either retrospective or prospective and can evaluate any type of intervention including drugs, operations, or approaches to health care delivery.

96 C.C. Greenberg and J.B. Dimick

Retrospective CER primarily uses large population-based datasets, including patient registries, administrative data, or health insurance claims data, to evaluate outcomes based on several different approaches to care. As an example, the treatment of prostate cancer tops most lists of the most critical questions to address in CER given the similar outcomes observed with very divergent approaches to care including radiation, surgery, and clinical observation. A 2009 study published in

Journal of the American Medical Association compared the effectiveness of open versus minimally invasive prostatectomy.2 By combining the disease-specific variables in the Surveillance Epidemiology and End Results (SEER) registry data with longitudinal Medicare claims data, the authors were able to compare both shortand long-term outcomes following each surgical approach. Open prostatectomy was associated with longer length of stay, higher rates of in-hospital complications, and higher rates of stricture, but lower rates of incontinence and erectile dysfunction. Importantly, between 2003 and 2007, the rate of minimally invasive surgery increased from 9% to 43%, emphasizing the importance of a timely evaluation of new surgical techniques. The rapid adoption of this minimally invasive prostatectomy occurred before a randomized clinical trial could have been performed, and its remarkable utilization rate precludes the performance of such a trial at this point. CER offers an important approach to providing much needed data on the comparative effectiveness of such new surgical techniques as illustrated by this example.

The major issue in retrospective CER is accounting for observed and unobserved confounders. In traditional clinical trials, imbalances in baseline characteristics between the different treatment arms are minimized by randomization. In CER, investigators use a variety of tools to adjust for confounders (or characteristics that may differ between the 2 groups of patients and lead to observed differences in outcome) and approximate randomization as closely as possible. Examples of such techniques include multivariable models, propensity scores, and instrumental variables.

Prospective CER can be randomized or non-randomized, such as a prospective observational cohort study (Table 7.1).

Randomized CER may be termed “pragmatic clinical trials,” to reflect the fact that they are conducted under “real-world” conditions and practice. Inclusion and exclusion criteria are minimal if they exist at all. Randomization may take place at any level – meaning patients may be randomized to one treatment or another individually, or randomization may take place at the level of the physician, clinic, or institution. This later type of randomization is referred to as a “cluster randomized” trial.

Patient-Centered Outcomes Research

The Institute of Medicine defined the six components of high-quality care in its 2001 “Crossing the Quality Chasm” and highlighted the importance of patient-centered care.3 Health-related quality of life (HRQOL) is the most obvious and developed of these patient-centered outcomes. A number of different validated instruments exist to measure general (such as the SF-36) and disease-specific (such as for cancer or inflammatory bowel disease) HRQOL. There are advantages and disadvantages to each type of measure that must be considered. Generic measures are comprehensive measures which assess overall effects on health, independent of disease or treatment, while condition-specific measures can be more sensitive. They are designed to capture symptoms that are specific to a given medical condition. While it is often ideal to include components of both a generic and diseasespecific tool, this is often not practical. One can modify a generic measure for a specific condition, or add a diseasespecific supplement to simplify the tool and expedite administration. It is always preferable to use a previously validated instrument if one exists. If not, rigorous methodology exists for the development and validation of an instrument prior to implementation. Other patient-centered outcomes to be considered include patient satisfaction, decision regret, and patient preference. The primary tools used to assess patientcentered outcomes include survey instruments, structured or semi-structured interviews, or focus groups, depending on the particular question at hand (Table 7.2).