Outcomes analysis is fundamental to the modern practice of medicine. As physicians, we use outcomes analysis in numerous situations, such as to counsel patients on the risks and benefits of treatments (eg, the risk of endophthalmitis after cataract surgery), to develop new treatments and measure their efficacy (eg, comparing phacoemulsification with modern small-incision manual extracapsular cataract surgery), and to individually stay up to date and improve our practice (eg, using customized A-constants for IOL calculations).

Although not a new concept, outcomes analysis is increasingly linking patient care and research in unprecedented ways, driven by the opportunities presented by big data. Substantially more data are available, analytic tools are more widespread, and analysis itself is much faster. These developments now allow practitioners to use data from their own practices, or aggregated data from a whole country or large insurance system, to increase their efficiency, improve their results, and benchmark themselves against the performances of their colleagues.

INTERPRETING OUTCOMES ANALYSIS

As a comprehensive term, outcomes analysis can encompass anything involving investigation of clinical efficacy or effectiveness—randomized clinical trials, retrospective single- and multicenter studies, system-generated physician report cards, and individual physicians' efforts to evaluate their own results. However, the real evolution in outcomes analysis has been in scale and scope, with new technologies and methods of organization enabling us to follow unprecedented numbers of patients over unparalleled lengths of time, with minimal burden on physicians and practices.

Whether outcomes analyses are derived from peer-reviewed literature, personal data, or reports generated by hospitals, health systems, or insurers, interpretation of their results requires careful consideration of data sources and methods. Particularly, modern outcomes analysis often relies on aggregated empirical data obtained through clinical data registries, which provide the opportunity to answer clinical questions without the time, expense, or inherent constraints of randomized controlled trials. Randomized clinical trials have long been the gold standard, but they are expensive, limited in duration simply due to feasibility, and subject to concerns regarding generalizability. Further, retrospective reviews are usually conducted at one or a few sites, often academic centers, and are fundamentally limited by feasibility and cohort size for less common or rare diseases.

Ideally, sophisticated scrutiny will use both the results of randomized controlled trials and large-scale empirical data, recognizing and balancing the relative strengths and weaknesses of each—length and generalizability concerns for controlled trials, and variability and uncontrolled populations for empirical data analysis. Looking at a spectrum of data sources provides an opportunity to best meet the ultimate goal of counseling patients and practicing better medicine.

Organized clinical data registries are increasing in both number and scope; examples include the European Registry of Quality Outcomes for Cataract and Refractive Surgery (EUREQUO), the American Academy of Ophthalmology (AAO) Intelligent Research in Sight (IRIS) Registry, EyeNet Sweden, the Malaysian Cataract Surgery Registry, the Fight Retinal Blindness! (FRB!) Project in Australia, data collection initiatives by the UK National Health Service, and efforts in India by Aravind Eye Hospitals, among others (see Organized Clinical Data Registries and Resources).

Fundamentally, analyzing our outcomes implies that we recognize we must understand how we are doing before we can improve. Ideally, these analyses look at the big picture, beyond the level of individual provider. Aggregate data on clinical factors and outcomes are highly variable internationally. National health care systems or initiatives arguably have the best opportunity for data collection and the benefit of integrating data from multiple sources. Outcomes data for privatized and even some state health care systems are, ironically, derived mostly from insurance claims data.

IMPROVing CLINICAL PRACTICE

New technology for organized data collection allows organizations or individual clinicians to answer clinical questions more rapidly than previously possible. We no longer have to wait years through trial conception, funding, execution, analysis, and publication, with the risk of the analysis becoming obsolete in the interim. Collection of large aggregate data can now be automated in registries and presented in near real-time analysis, allowing us to answer clinical questions rapidly, although still to some extent bound by time for analysis and publication to disseminate results. This permits a new way of thinking about evidence-based medicine. We can, for example, find out who is implanting the iStent (Glaukos) and whether these devices are working as expected. Or we can measure the relative benefits and impact of laser-assisted cataract surgery compared with standard phacoemulsification.

Participation in data registries enables physicians to use benchmarking as a tool to increase their efficiency, improve their day-to-day work, or comply with mandated regulations. With these research tools, individual physicians can answer specific clinical questions, performing risk adjustment or demonstrating quality for value-based reimbursement arrangements. They can seamlessly collate their postoperative refractions to customize A-constants for cataract surgery, predict complication rates based on specific patient factors, or figure out whether a patient is likely to do better with one treatment approach versus another in their hands.

One thing that has been lacking from almost all outcomes data collection to date is formal assessment of patient-reported outcomes. Large-scale collection of structured patient-reported outcome measures offers a way to show the value of ophthalmic care to providers and policymakers, even beyond ophthalmology. International consistency in condition-specific outcome measures is being furthered by the International Consortium for Health Outcomes Measurement (ICHOM), founded in 2012 with the stated goal of “defining global Standard Sets of outcome measures that really matter to patients for the most relevant medical conditions and by driving adoption and reporting of these measures worldwide.” Thus far, guidelines have been released for cataract and macular degeneration, with intent to expand to additional visual domains.

CONCLUSION

Achieving consistency across registries is a goal that will ultimately provide reliable, meaningful value and opportunities to mine data on a large scale, accounting for variations in comorbidities, demographic differences, and population mobility and allowing long-term monitoring of trends.

However, the greatest potential value of outcomes analysis lies in the vision of what it can be—a new technology enabling us to extract information from medical records and process it in ways previously impossible. We are now able to do more with less, using data that already exist. By automating the identification and processing of information directly from patients' medical records, the burden on physicians and practices can be minimized. A system with a dashboard that can deliver near real-time processing and customized data extraction will allow outcomes analysis to transition from retrospective to prospective and enable customized predictions—for example, providing patients (and physicians) with a quantitative estimate of their own risk for specific complications from cataract surgery and the likely impact on visual function if a complication occurs. These results will also be useful as teaching tools.

This vision for outcomes analysis will not be realized immediately, but big data is the next frontier in medicine, complemented by the expanding use of electronic health records. Ophthalmology has been a pioneer in registry development internationally, compared with other specialties. Most data are currently obtained through customized structured data entry fields, but data extraction via natural language processing of free-text visit notes is already a reality, likely only to spread.

We can realistically predict true large-scale empirical databases, offering insights into rare diseases, patient perspectives, surveillance of postmarket drug and device performance, and real-world outcomes—balancing the weaknesses of randomized controlled clinical trials with the variability of clinical practice. Real-time feedback will provide a continuous quality improvement loop for internal and external benchmarking. In this not-too-distant future, ophthalmologists will have aggregate empiric clinical outcomes data as a powerful tool to revolutionize the practice of evidence-based medicine. n