FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2801405364> ?p ?o ?g. }

• W2801405364 endingPage "663" @default.
• W2801405364 startingPage "660" @default.
• W2801405364 abstract "Robin N. Kamal MDAlthough randomized trials are at the top of the evidence pyramid, we will never be able to answer some questions using studies of that design. Whether antibiotics are needed prior to simple, clean, soft-tissue procedures is one such question; the event we are concerned about (infection) is just too rare. Powering a randomized study to get the answer would require tens of thousands of patients, and accounting for important differences among patients by stratified randomization would require hundreds of thousands. And yet we must make a prescribing choice about whether to use antibiotics every time we do one of these common procedures. Surgeons may feel as though this decision lacks consequence—it’s only a difference of a single dose of a single drug, one way or the other—but the downside risk of being wrong is severe. Infections following even seemingly minor hand surgery can develop into adjacent-joint pyarthrosis, leading to serious disability, fusion, and amputation [6]. And on the other side, in this space, we recently discussed a fatal complication associated with antibiotic use, Clostridium difficile colitis [3, 8], which can occur after even single-dose preoperative antibiotic prophylaxis [4]. There are selective [5] and even systematic [1] reviews on antibiotic use before clean surgery, but they are based on surprisingly sparse evidence. For that reason, we are excited to publish this month’s spotlight article, “Effectiveness of Preoperative Antibiotics in Preventing Surgical Site Infection After Common Soft Tissue Procedures of the Hand,” by Dr. Robin Kamal’s group at Stanford University [9]. Although its focus is hand surgery (particularly, soft-tissue surgery of the hand), its message may generalize to other similar procedures, and certainly its methods do. They found that antibiotics do not reduce the risk of surgical-site infection after these operations, and they did so convincingly. The group used a largescale, multistate commercial insurance database to perform a robust, amply powered analysis of more than half a million procedures, in which they controlled for more than a dozen infection-related confounding variables. Many no-difference large-database studies ultimately are unhelpful because they can be interpreted in two conflicting ways: Either the intervention in question was not effective, or it was effective in the patients in whom it was used (that is, had they not used the intervention in those particular patients, they would have fared worse rather than the same as patients in the control group). But since there are no inclusion criteria in these look-back studies, the result cannot be replicated in practice. And insufficient power—failure to detect a between-groups difference that was in fact present—should always be on our minds as we read no-difference studies [7]. Fortunately, the study in this month’s Clinical Orthopaedics and Related Research® by Dr. Kamal’s group addressed these concerns thoughtfully—the propensity-scoring approach used plus the vast number of patients analyzed provide as compelling an answer as we are likely to get on this important topic. This paper has a lot to recommend it, even if one is not a hand surgeon. The methods used here provide an aspirational standard for other researchers who wish to ask similarly large, important questions in the context of institutional databases and government registries. Readers can use this paper as a guide for what to look for when reading big-data studies on other subjects. Clinical practice guidelines—which I hope will be forthcoming on this key topic—should carefully incorporate the findings of this paper. It is that good. Please join me as I go behind the discovery with Dr. Kamal, senior author of “Effectiveness of Preoperative Antibiotics in Preventing Surgical Site Infection After Common Soft Tissue Procedures of the Hand” in the Take-5 interview that follows. Take Five Interview with Robin N. Kamal MD, senior author of “Effectiveness of Preoperative Antibiotics in Preventing Surgical Site Infection After Common Soft Tissue Procedures of the Hand” Seth S. Leopold MD:Congratulations on this important study, which is a perfect match of study methods to research problem, since we’ll never get a convincing answer to this question using randomized trials. But how can you reassure the skeptical reader that a study like this—which is necessarily retrospective—is sufficiently robust to guide a therapeutic decision about whether to use antibiotics prior to these kinds of procedures? Robin N. Kamal MD: Thank you, Dr. Leopold. Retrospective analysis of database data can be useful when a prospective, randomized design is not feasible, as in the case of antibiotic prophylaxis in reducing surgical-site infections. To ensure the validity of such retrospective analyses, study design should minimize the impact of bias and confounding that can result from loss to followup, misclassification bias, and confounding by indication. Carefully chosen selection criteria, precise and clinically relevant definitions of treatments and outcomes, propensity-score matching, and multivariable regression are among the strategies we employed to maximize the validity of our conclusions. Furthermore, our study achieved a large sample size from a variety of clinical settings across a geographically broad area, which supports broader generalizability of findings. Dr. Leopold:How should a study like this be replicated or confirmed? Repeat it in a different database? If so, which one(s)? Look at other clean surgical procedures in similar databases? As you know, replication is essential, but asking the same question of different databases can yield dramatically different answers[2], and not all are appropriate to the task at hand. Dr. Kamal: I would be excited to see a study like this replicated using data procured from a clinical setting, rather than from an administrative context like the one we used. For example, there has been recent work on developing methods to mine free text (natural language processing) in electronic health records to assess clinical outcomes [10]. This approach could be applied to assess postoperative surgical-site infection rates across a sufficiently large number of patients in a relatively short amount of time. It would be interesting to see whether results derived from these data are similar to those obtained using insurance claims. To replicate our work using claims data, I would encourage other investigators to conduct similar analyses in other datasets. As each dataset is different with respect to the types of data fields available and the patient populations represented, a similar result would tend to confirm generalizability of findings. But a different result could suggest different treatment effects in different patients—which would be equally important. Dr. Leopold:As I noted in the commentary, retrospective, big-database studies that find no differences between treatment groups often can be interpreted several ways: Either that there is no difference between the groups, that there is a difference, but the study failed to detect it (insufficient power), or even that there would have been a difference, but the patients who received the treatment were well chosen and had they not received it, they would have been worse off. It can be complicated. I would contend that your study convincingly eliminated the latter two interpretations, and things are just as they seem here: Antibiotics are not needed for these patients. How do you see it? Dr. Kamal: Concluding that there is no decrease in surgical site infection rates following antibiotic prophylaxis requires a sample size large enough to detect a small treatment effect (this is the concept of statistical power that you mention). Studies using administrative claims data can achieve numbers required for adequate power, but care must be taken to minimize the impact of confounding by indication as you’ve suggested. Methods such as propensity-score matching or instrumental variable analysis can be used to help control for confounding by measured and, in the case of the latter method, unmeasured variables. Provided that a database investigation yields sufficiently large sample size and that variables related to both treatment and outcome can be captured from the data, I believe that these methods can help us feel confident in our conclusions. Dr. Leopold:There often is a gulf between evidence and practice, or at least a lag time between the emergence of new evidence and changes to practice. What steps should our specialty take to modify practice considering the evidence you’ve produced here? Dr. Kamal: Implementation of new practice is challenging and there are powerful barriers including our own prior experience, cognitive inertia, social influence (local culture), skepticism about or ignorance of new data, to name just a few. Facilitating adoption of new evidence can start with improving dissemination of new data, such as efforts like this Editor’s Spotlight/Take-5 that highlights high-impact, general-interest evidence for readers. On a local or health-system level, a focus on creating a culture for continuous improvement, or a so-called “learning health system,” can allow for implementation of new evidence to be embedded into the care-delivery process. Processes can also be created that facilitate physicians making the best decisions when possible, such as the use of decision-support tools in electronic health records. As our culture in healthcare embraces the idea that we always have room to improve and that it’s okay to question the status quo, the concept of taking new evidence and using it in practice becomes easier to implement. Dr. Leopold:Big-database papers are becoming more and more a part of our evidence base, but I think that many readers find them intimidating. What practical advice can you offer readers for how to make the most of this important new study design? Bear in mind that most are not methodologists, and remembering the particular elements of each database starts to look like alphabet soup to the typical reader. Dr. Kamal: Database studies are most effective when the question you’re asking is some version of “What happens in real practice when ...” For example, these studies can report practice patterns, and they can calculate incidence or prevalence of particular problems from a broad-based (national healthcare system) level that a single-center cohort or randomized controlled trial cannot. Another advantage is because of the number of samples included in the study, and the database’s national representation of patients, the results often are more generalizable to other physicians and their patients than might be a single-center study, where the reader of a study may feel that (s)he may not have the same support staff or systems in place to achieve the results reported in the new paper. A database study that represents a national sample of patients can demonstrate what happens in real life, across various geographic locations, and amongst different age groups, hospitals, and in some cases, different surgeons. However, like all other study methodologies, claims databases have limitations. They cannot answer “why” questions (for example, our paper cannot answer the question, “why do some surgeons use prophylactic antibiotics?”). Additionally, claims databases rely on accurate and specific coding, so the conclusions they can draw on are only as good as the codes physicians use to attribute to the diagnoses and procedures. Although large-database study designs are relatively new and perhaps daunting to read at first, the ability to track a national sample of patients longitudinally over time is powerful, and in some circumstances, allows the reader to see trends and associations that would otherwise go unnoticed." @default.
• W2801405364 created "2018-05-17" @default.
• W2801405364 creator A5046851940 @default.
• W2801405364 date "2018-02-26" @default.
• W2801405364 modified "2023-09-26" @default.
• W2801405364 title "Editor’s Spotlight/Take 5: Effectiveness of Preoperative Antibiotics in Preventing Surgical Site Infection After Common Soft Tissue Procedures of the Hand" @default.
• W2801405364 cites W1796652980 @default.
• W2801405364 cites W2047143615 @default.
• W2801405364 cites W2087998258 @default.
• W2801405364 cites W2110364674 @default.
• W2801405364 cites W2161469103 @default.
• W2801405364 cites W2599569086 @default.
• W2801405364 cites W2737221302 @default.
• W2801405364 cites W2799664801 @default.
• W2801405364 cites W2800704417 @default.
• W2801405364 cites W2801194994 @default.
• W2801405364 doi "https://doi.org/10.1007/s11999.0000000000000231" @default.
• W2801405364 hasPubMedCentralId "https://www.ncbi.nlm.nih.gov/pmc/articles/6260092" @default.
• W2801405364 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/29481359" @default.
• W2801405364 hasPublicationYear "2018" @default.
• W2801405364 type Work @default.
• W2801405364 sameAs 2801405364 @default.
• W2801405364 citedByCount "0" @default.
• W2801405364 crossrefType "journal-article" @default.
• W2801405364 hasAuthorship W2801405364A5046851940 @default.
• W2801405364 hasBestOaLocation W28014053641 @default.
• W2801405364 hasConcept C136948725 @default.
• W2801405364 hasConcept C141071460 @default.
• W2801405364 hasConcept C177713679 @default.
• W2801405364 hasConcept C2992582194 @default.
• W2801405364 hasConcept C501593827 @default.
• W2801405364 hasConcept C71924100 @default.
• W2801405364 hasConcept C86803240 @default.
• W2801405364 hasConcept C89423630 @default.
• W2801405364 hasConceptScore W2801405364C136948725 @default.
• W2801405364 hasConceptScore W2801405364C141071460 @default.
• W2801405364 hasConceptScore W2801405364C177713679 @default.
• W2801405364 hasConceptScore W2801405364C2992582194 @default.
• W2801405364 hasConceptScore W2801405364C501593827 @default.
• W2801405364 hasConceptScore W2801405364C71924100 @default.
• W2801405364 hasConceptScore W2801405364C86803240 @default.
• W2801405364 hasConceptScore W2801405364C89423630 @default.
• W2801405364 hasIssue "4" @default.
• W2801405364 hasLocation W28014053641 @default.
• W2801405364 hasLocation W28014053642 @default.
• W2801405364 hasLocation W28014053643 @default.
• W2801405364 hasLocation W28014053644 @default.
• W2801405364 hasLocation W28014053645 @default.
• W2801405364 hasOpenAccess W2801405364 @default.
• W2801405364 hasPrimaryLocation W28014053641 @default.
• W2801405364 hasRelatedWork W1586374228 @default.
• W2801405364 hasRelatedWork W2002120878 @default.
• W2801405364 hasRelatedWork W2003938723 @default.
• W2801405364 hasRelatedWork W2047967234 @default.
• W2801405364 hasRelatedWork W2118496982 @default.
• W2801405364 hasRelatedWork W2364998975 @default.
• W2801405364 hasRelatedWork W2369162477 @default.
• W2801405364 hasRelatedWork W2439875401 @default.
• W2801405364 hasRelatedWork W4238867864 @default.
• W2801405364 hasRelatedWork W2525756941 @default.
• W2801405364 hasVolume "476" @default.
• W2801405364 isParatext "false" @default.
• W2801405364 isRetracted "false" @default.
• W2801405364 magId "2801405364" @default.
• W2801405364 workType "article" @default.