RESUMO
OBJECTIVE: To enhance the value of the Pennsylvania Patient Safety Reporting System (PA-PSRS) falls reports by developing a falls reporting program that standardizes falls reporting and provides timely benchmarked falls rates and process measurement reports hospitals can use to identify areas of improvement in their falls program. METHODS: The new PA-PSRS falls reporting program requires adherence to standardized definitions of falls to generate standardized, customizable analytic reports. An advisory committee and statewide survey guided the development of the program, data definitions, system features and functionality, and methods for stratifying reporting criteria. RESULTS: Three real-time falls outcomes and 2 falls process reports with multiple configurable options were created. A falls dashboard was developed based on gaps in falls risk assessment processes identified in PA-PSRS falls event reports. Six months after launching the program, 41.3% of Pennsylvania hospitals enrolled. The Authority's annual survey indicated 82.9% of participating hospitals found that the new falls analytic reports were somewhat useful to very useful. Preliminary impact on the falls with harm rate has been limited, a less than 1% reduction, but the ability to identify specific organizational and patient risk factors in a timely manner provides hospitals with opportunities to target falls prevention resources more effectively. CONCLUSIONS: The PA-PSRS falls reporting program's standardized definition of falls offers new analytic reports that include falls rates with benchmarking data and a falls dashboard. The benchmarking data allow hospitals to compare themselves to peer hospitals statewide. The newly expanded PA-PSRS falls reporting program has turned an adverse event-reporting program into a quality improvement tool.
Assuntos
Acidentes por Quedas/estatística & dados numéricos , Sistemas de Notificação de Reações Adversas a Medicamentos/normas , Análise de Dados , Humanos , Melhoria de Qualidade , Fatores de Risco , Inquéritos e QuestionáriosRESUMO
INTRODUCTION: The objective of this study was to develop a semiautomated approach to screening cases that describe hazards associated with the electronic health record (EHR) from a mandatory, population-based patient safety reporting system. METHODS: Potentially relevant cases were identified through a query of the Pennsylvania Patient Safety Reporting System. A random sample of cases were manually screened for relevance and divided into training, testing, and validation data sets to develop a machine learning model. This model was used to automate screening of remaining potentially relevant cases. RESULTS: Of the 4 algorithms tested, a naive Bayes kernel performed best, with an area under the receiver operating characteristic curve of 0.927 ± 0.023, accuracy of 0.855 ± 0.033, and F score of 0.877 ± 0.027. DISCUSSION: The machine learning model and text mining approach described here are useful tools for identifying and analyzing adverse event and near-miss reports. Although reporting systems are beginning to incorporate structured fields on health information technology and the EHR, these methods can identify related events that reporters classify in other ways. These methods can facilitate analysis of legacy safety reports by retrieving health information technology-related and EHR-related events from databases without fields and controlled values focused on this subject and distinguishing them from reports in which the EHR is mentioned only in passing. CONCLUSIONS: Machine learning and text mining are useful additions to the patient safety toolkit and can be used to semiautomate screening and analysis of unstructured text in safety reports from frontline staff.
Assuntos
Automação , Mineração de Dados , Registros Eletrônicos de Saúde , Aprendizado de Máquina , Notificação de Abuso , Informática Médica , Segurança do Paciente , Algoritmos , Teorema de Bayes , Bases de Dados Factuais , Humanos , PennsylvaniaRESUMO
BACKGROUND: In 2009, researchers from Johns Hopkins University's Armstrong Institute for Patient Safety and Quality; public agencies, including the FDA; and private partners, including the Emergency Care Research Institute and the University HealthSystem Consortium (UHC) Safety Intelligence Patient Safety Organization, sought to form a public-private partnership for the promotion of patient safety (P5S) to advance patient safety through voluntary partnerships. The study objective was to test the concept of the P5S to advance our understanding of safety issues related to ventilator events, to develop a common classification system for categorizing adverse events related to mechanical ventilators, and to perform a comparison of adverse events across different adverse event reporting systems. METHODS: We performed a cross-sectional analysis of ventilator-related adverse events reported in 2012 from the following incident reporting systems: the Pennsylvania Patient Safety Authority's Patient Safety Reporting System, UHC's Safety Intelligence Patient Safety Organization database, and the FDA's Manufacturer and User Facility Device Experience database. Once each organization had its dataset of ventilator-related adverse events, reviewers read the narrative descriptions of each event and classified it according to the developed common taxonomy. RESULTS: A Pennsylvania Patient Safety Authority, FDA, and UHC search provided 252, 274, and 700 relevant reports, respectively. The 3 event types most commonly reported to the UHC and the Pennsylvania Patient Safety Authority's Patient Safety Reporting System databases were airway/breathing circuit issue, human factor issues, and ventilator malfunction events. The top 3 event types reported to the FDA were ventilator malfunction, power source issue, and alarm failure. CONCLUSIONS: Overall, we found that (1) through the development of a common taxonomy, adverse events from 3 reporting systems can be evaluated, (2) the types of events reported in each database were related to the purpose of the database and the source of the reports, resulting in significant differences in reported event categories across the 3 systems, and (3) a public-private collaboration for investigating ventilator-related adverse events under the P5S model is feasible.
Assuntos
Segurança do Paciente/estatística & dados numéricos , Gestão de Riscos/estatística & dados numéricos , Ventiladores Mecânicos/efeitos adversos , Estudos Transversais , Bases de Dados Factuais , HumanosRESUMO
BACKGROUND: External reporting of medical errors a adverse events enables learning from the errors of others in the pursuit of systems-level improvements that can prevent future errors. It is logical to presume that medication errors involving the use of anticoagulants, among the most frequently cited product classes involved in harmful medication errors, would be captured in a variety of patient safety reporting programs. METHODS: Data on reported errors involving the anticoagulant heparin were reviewed, compared, and aggregated from the databases of three large patient safety reporting programs-MEDMARX, the Pennsylvania Patient Safety Authority's Patient Safety Reporting System, and the University Health System Consortium, together representing more than 1,000 reporting organizations for 2005 RESULTS: Approximately 300,000 medication errors and near misses were reported to the programs, and 10,359-a mean of 3.6% (range, 3.1%-5.5%)-involved heparin products. The proportion of heparin-related reports that involved patient harm ranged from 1.4% to 4.9%. The phase of the medication use process cited most frequently in harmful events was the administration phase (56% of errors leading to harm), followed by the prescribing phase (19% of errors leading to harm). DISCUSSION: This study represents the first attempt by these three large reporting systems to combine data on a single clinical process. The consistent patterns evident in the reports, such as the percentage of all medication errors that involved heparin, suggests that reporting programs, at least for common events such as medication errors, may reach a point of diminishing returns in which aggregating more reports of a certain type yields no additional insight once a large volume of similar events is captured and analyzed.