FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W1507106212> ?p ?o ?g. }

• W1507106212 endingPage "1019" @default.
• W1507106212 startingPage "1018" @default.
• W1507106212 abstract "Suboptimal drug prescribing to older adults has been well documented,1, 2 and many approaches result in short-term improvements in prescribing behavior.3, 4 Administrative databases provide an inexpensive, rapid, and reasonably accurate account of drug-prescribing patterns in large populations, and they can be useful in assessing the effect of regulations aimed at improving drug use conducted outside an experimental setting.5 Some of the strengths and limitations of this approach are demonstrated in this issue of the Journal of the American Geriatrics Society.6 This study compares prescription of potentially inappropriate medications for older adults in Ontario nursing homes, in which clinical pharmacists review medications at least every 3 months, with that of older adults living in the community.6 The study endpoint is the use of one of a list of drugs developed by Beers et al.7, 8 thought to be inappropriate for use in older adults. Drugs were included on this list because they were ineffective or because safer alternatives were available within a drug class. Using drug data identified from healthcare claims, the authors found that nursing home residents were close to 50% less likely to be dispensed one of these potentially inappropriate drug therapies than community-dwelling older adults (odds ratio=0.52, 95% confidence interval=0.49–0.55, P<.001). One potential interpretation of this finding is that the pharmacist program has resulted in a large (approximately 50%) and highly significant (P<.001) decrease in inappropriate drug use in older adults, and one conclusion, reached also by the authors, is that similar models may be useful in the community setting. I will comment briefly on the unconfirmed relationship between inappropriate drug use and adverse drug events (ADEs) and the distortion of effect size and significance when extractions from large databases are presented in relative, rather than absolute, terms. Although people aged 65 and older constitute only about 12% of the U.S. population, they consume one-third of all prescribed drugs and more than one-half of over-the-counter medicines. ADEs are a significant public health problem, which has been estimated to cost $76.6 billion in the ambulatory setting in the United States, with the largest component of this total cost being associated with drug-related hospitalizations.9 Although it is not clear whether age, per se, is an independent risk factor for ADEs, these events occur much more commonly in older adults.10 An important goal of any intervention to improve drug prescribing is to reduce risk of ADEs. Typically, administrative databases contain information on drug, dose, number of pills administered, and the date the prescription was filled.11 They are therefore well suited to identify use of specific medications or classes of medications in large populations. Although the validity of explicit lists of inappropriate drugs developed by Beers and other investigators has been discussed elsewhere,12 there is growing evidence that other aspects of drug prescribing, such as improper dosing, and failures in therapeutic monitoring account for the vast majority of ADEs. For example, recent studies of ADEs in older adults reveal that these events are most often caused by cardiovascular drugs, antibiotics, diuretics, lipid-lowering agents, and opiates in outpatients13 and by psychotropic drugs, antibiotics, and anticoagulants in nursing homes.14 Although it is not clear how many ADEs are caused by “inappropriate drugs,”7, 8 this evidence suggests that drugs that no criteria would identify as “inappropriate” cause the overwhelming majority of ADEs in older adults. Expressing data in relative terms can distort the perception of effect size. For example, investigators15 have compared the willingness of general practitioners to prescribe a lipid-lowering drug when trial results were presented in absolute (an absolute reduction of 1.4% cardiac events was demonstrated) and relative terms (a 34% cardiac event reduction was demonstrated). Each physician rated their likelihood of prescribing the drug on a scale from 0 to 100. The mean agreement to prescribe was 77% when efficacy data were expressed as a relative risk reduction, whereas the mean agreement to prescribe was only 24% when efficacy was expressed as an absolute risk reduction. This has been demonstrated for screening and therapeutic choice as well.16, 18 Lane et al. report a 48% relative reduction in the use of potentially inappropriate medications by nursing home residents, but the absolute difference between the two groups is 1% (3.3% vs 2.3%).6 Although the information needed to perform this calculation is in the abstract, it is not explicitly stated in this or many other studies.19 Absolute risk is easy to calculate, and careful readers should make a practice of it. From this information, one can calculate the number needed to treat and begin conceptualizing what risks (or costs) should be borne to achieve the benefit. There is no criterion standard for what constitutes a clinically significant improvement in drug prescribing for a given population. Most interventions related to improving drug prescribing are powered to measure an absolute improvement of 15% to 20%, and it is unlikely that an intervention would ever be undertaken if at baseline only 3% of prescriptions were considered suboptimal. Since the landmark publication 25 years ago on the importance of sample size,20 much attention has been paid to type II error in the design and interpretation of research studies. The issue of false negative results due to the failure to include enough participants in clinical trials is now taught to all medical students, and astute trainees seem to relish the opportunity to pounce on discussing possibility of type II error each time a negative study is discussed in academic settings. Sample size calculations must be included as a part of any research study. In studies using administrative data, obtaining adequate sample size is rarely a problem, and the number of subjects may be in the millions. However, a potential downside to working with large data sets is that statistical significance (P<.05) can be achieved when the magnitude of the difference between two groups is small (i.e., statistical significance is achieved in the absence of clinical significance). This is not a problem related to error or bias but nonetheless requires a thoughtful approach to how results in such studies should be presented and interpreted. My suggestion is not to be swayed by highly significant P-values and focus instead on the absolute magnitude of change. It is unfortunate that P<.05 has become the gatekeeper for publication—or “truth”—because it distracts attention from more important information. If the observed difference in drug prescribing is entirely attributable to the ongoing review by pharmacists, then this program is responsible for 12,000 older adults not receiving potentially inappropriate medications. Although it is unlikely that such a modest effect would achieve statistical significance without the pharmacy records of more than 1.2 million people, nonetheless it is important. However, what is not known is the number of ADEs the program prevented and what the program costs. This is the information I would want before deciding whether the program were to be continued or expanded." @default.
• W1507106212 created "2016-06-24" @default.
• W1507106212 creator A5014379891 @default.
• W1507106212 date "2004-06-01" @default.
• W1507106212 modified "2023-09-23" @default.
• W1507106212 title "Interpreting Interventions to Improve Drug Use Using Administrative Databases" @default.
• W1507106212 cites W1969388541 @default.
• W1507106212 cites W1970046651 @default.
• W1507106212 cites W1997241558 @default.
• W1507106212 cites W2012947624 @default.
• W1507106212 cites W2022758008 @default.
• W1507106212 cites W2105814725 @default.
• W1507106212 cites W2109180102 @default.
• W1507106212 cites W2125669390 @default.
• W1507106212 cites W2141248662 @default.
• W1507106212 cites W4293237928 @default.
• W1507106212 cites W4293765665 @default.
• W1507106212 doi "https://doi.org/10.1111/j.1532-5415.2004.52275.x" @default.
• W1507106212 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/15161473" @default.
• W1507106212 hasPublicationYear "2004" @default.
• W1507106212 type Work @default.
• W1507106212 sameAs 1507106212 @default.
• W1507106212 citedByCount "1" @default.
• W1507106212 crossrefType "journal-article" @default.
• W1507106212 hasAuthorship W1507106212A5014379891 @default.
• W1507106212 hasConcept C159110408 @default.
• W1507106212 hasConcept C17744445 @default.
• W1507106212 hasConcept C199539241 @default.
• W1507106212 hasConcept C27415008 @default.
• W1507106212 hasConcept C2779473830 @default.
• W1507106212 hasConcept C2780035454 @default.
• W1507106212 hasConcept C41008148 @default.
• W1507106212 hasConcept C545542383 @default.
• W1507106212 hasConcept C71924100 @default.
• W1507106212 hasConcept C77088390 @default.
• W1507106212 hasConcept C98274493 @default.
• W1507106212 hasConceptScore W1507106212C159110408 @default.
• W1507106212 hasConceptScore W1507106212C17744445 @default.
• W1507106212 hasConceptScore W1507106212C199539241 @default.
• W1507106212 hasConceptScore W1507106212C27415008 @default.
• W1507106212 hasConceptScore W1507106212C2779473830 @default.
• W1507106212 hasConceptScore W1507106212C2780035454 @default.
• W1507106212 hasConceptScore W1507106212C41008148 @default.
• W1507106212 hasConceptScore W1507106212C545542383 @default.
• W1507106212 hasConceptScore W1507106212C71924100 @default.
• W1507106212 hasConceptScore W1507106212C77088390 @default.
• W1507106212 hasConceptScore W1507106212C98274493 @default.
• W1507106212 hasIssue "6" @default.
• W1507106212 hasLocation W15071062121 @default.
• W1507106212 hasLocation W15071062122 @default.
• W1507106212 hasOpenAccess W1507106212 @default.
• W1507106212 hasPrimaryLocation W15071062121 @default.
• W1507106212 hasRelatedWork W1571055461 @default.
• W1507106212 hasRelatedWork W1965095608 @default.
• W1507106212 hasRelatedWork W1992254244 @default.
• W1507106212 hasRelatedWork W2100669381 @default.
• W1507106212 hasRelatedWork W2376914980 @default.
• W1507106212 hasRelatedWork W2426261983 @default.
• W1507106212 hasRelatedWork W2901146940 @default.
• W1507106212 hasRelatedWork W2936211750 @default.
• W1507106212 hasRelatedWork W4251901137 @default.
• W1507106212 hasRelatedWork W4323926278 @default.
• W1507106212 hasVolume "52" @default.
• W1507106212 isParatext "false" @default.
• W1507106212 isRetracted "false" @default.
• W1507106212 magId "1507106212" @default.
• W1507106212 workType "article" @default.