Dataset Information

Multi-site medical record review for validation of intentional self-harm coding in emergency departments.

ABSTRACT:

Background

Codes in the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM), are used for injury surveillance, including surveillance of intentional self-harm, as they appear in administrative billing records. This study estimated the positive predictive value of ICD-10-CM codes for intentional self-harm in emergency department (ED) billing records for patients aged 10 years and older who did not die and were not admitted to an inpatient medical service.

Methods

The study team in Maryland, Colorado, and Massachusetts selected all or a random sample of ED billing records with an ICD-10-CM code for intentional self-harm (specific codes that began with X71-X83, T36-T65, T71, T14.91). Positive predictive value (PPV) was determined by the number and percentage of records with a physician diagnosis of intentional self-harm, based on a retrospective review of the original medical record.

Results

The estimated PPV for the codes' capture of intentional self-harm based on physician diagnosis in the original medical record was 89.8% (95% CI 85.0-93.4) for Maryland records, 91.9% (95% CI 87.7-95.0) for Colorado records, and 97.3% (95% CI 95.1-98.7) for Massachusetts records.

Conclusion

Given the high PPV of the codes, epidemiologists can use the codes for public health surveillance of intentional self-harm treated in the ED using ICD-10-CM coded administrative billing records. However, these codes and related variables in the billing database cannot definitively distinguish between suicidal and non-suicidal intentional self-harm.

SUBMITTER: Gabella BA

PROVIDER: S-EPMC9175468 | biostudies-literature | 2022 Jun

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Multi-site medical record review for validation of intentional self-harm coding in emergency departments.

Gabella Barbara A BA Hume Beth B Li Linda L Mabida Marianne M Costich Julia J

Injury epidemiology 20220607 1

<h4>Background</h4>Codes in the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM), are used for injury surveillance, including surveillance of intentional self-harm, as they appear in administrative billing records. This study estimated the positive predictive value of ICD-10-CM codes for intentional self-harm in emergency department (ED) billing records for patients aged 10 years and older who did not die and were not admitted to an inpatient medical se ...[more]

PMID: 35672865

Similar Datasets

Project description:BackgroundSuicide is an important public health concern in the United States and around the world. There has been significant work examining machine learning approaches to identify and predict intentional self-harm and suicide using existing data sets. With recent advances in computing, deep learning applications in health care are gaining momentum.ObjectiveThis study aimed to leverage the information in clinical notes using deep neural networks (DNNs) to (1) improve the identification of patients treated for intentional self-harm and (2) predict future self-harm events.MethodsWe extracted clinical text notes from electronic health records (EHRs) of 835 patients with International Classification of Diseases (ICD) codes for intentional self-harm and 1670 matched controls who never had any intentional self-harm ICD codes. The data were divided into training and holdout test sets. We tested a number of algorithms on clinical notes associated with the intentional self-harm codes using the training set, including several traditional bag-of-words-based models and 2 DNN models: a convolutional neural network (CNN) and a long short-term memory model. We also evaluated the predictive performance of the DNNs on a subset of patients who had clinical notes 1 to 6 months before the first intentional self-harm event. Finally, we evaluated the impact of a pretrained model using Word2vec (W2V) on performance.ResultsThe area under the receiver operating characteristic curve (AUC) for the CNN on the phenotyping task, that is, the detection of intentional self-harm in clinical notes concurrent with the events was 0.999, with an F1 score of 0.985. In the predictive task, the CNN achieved the highest performance with an AUC of 0.882 and an F1 score of 0.769. Although pretraining with W2V shortened the DNN training time, it did not improve performance.ConclusionsThe strong performance on the first task, namely, phenotyping based on clinical notes, suggests that such models could be used effectively for surveillance of intentional self-harm in clinical text in an EHR. The modest performance on the predictive task notwithstanding, the results using DNN models on clinical text alone are competitive with other reports in the literature using risk factors from structured EHR data.

Project description:ImportancePrescription opioids play a significant role in the ongoing opioid crisis. Guidelines and physician education have had mixed success in curbing opioid prescriptions, highlighting the need for other tools that can change prescriber behavior, including nudges based in behavioral economics.ObjectiveTo determine whether and to what extent changes in the default settings in the electronic medical record (EMR) are associated with opioid prescriptions for patients discharged from emergency departments (EDs).Design, setting, and participantsThis quality improvement study randomly altered, during a series of five 4-week blocks, the prepopulated dispense quantities of discharge prescriptions for commonly prescribed opioids at 2 large, urban EDs. These changes were made without announcement, and prescribers were not informed of the study itself. Participants included all health care professionals (physicians, nurse practitioners, and physician assistants) working clinically in either of the 2 EDs. Data were collected from November 28, 2016, through July 9, 2017, and analyzed from July 16, 2017, through May 14, 2018.InterventionsDefault quantities for opioids were changed from status quo quantities of 12 and 20 tablets to null, 5, 10, and 15 tablets according to a block randomization scheme. Regardless of the default quantity, each health care professional decided for whom to prescribe opioids and could modify the quantity prescribed without restriction.Main outcomes and measuresThe primary outcome was the number of tablets of opioid-containing medications prescribed under each default setting.ResultsA total of 104 health care professionals wrote 4320 prescriptions for opioids during the study period. Using linear regression, an increase of 0.19 tablets prescribed (95% CI, 0.15-0.22) was found for each tablet increase in default quantity. When evaluating each of the 15 pairwise comparisons of default quantities (eg, 5 vs 15 tablets), a lower default was associated with a lower number of pills prescribed in more than half (8 of the 15) of the pairwise comparisons; there was a higher quantity in 1 and no difference in 6 comparisons.Conclusions and relevanceThese findings suggest that default settings in the EMR may influence the quantity of opioids prescribed by health care professionals. This low-cost, easily implementable, EMR-based intervention could have far-reaching implications for opioid prescribing and could be used as a tool to help combat the opioid epidemic.Trial registrationClinicalTrials.gov identifier: NCT04155229.

Project description:BackgroundAbstractor training is a key element in creating valid and reliable data collection procedures. The choice between in-person vs. remote or simultaneous vs. sequential abstractor training has considerable consequences for time and resource utilization. We conducted a web-based (webinar) abstractor training session to standardize training across six individual Cancer Research Network (CRN) sites for a study of breast cancer treatment effects in older women (BOWII). The goals of this manuscript are to describe the training session, its participants and participants' evaluation of webinar technology for abstraction training.FindingsA webinar was held for all six sites with the primary purpose of simultaneously training staff and ensuring consistent abstraction across sites. The training session involved sequential review of over 600 data elements outlined in the coding manual in conjunction with the display of data entry fields in the study's electronic data collection system. Post-training evaluation was conducted via Survey Monkey©. Inter-rater reliability measures for abstractors within each site were conducted three months after the commencement of data collection.Ten of the 16 people who participated in the training completed the online survey. Almost all (90%) of the 10 trainees had previous medical record abstraction experience and nearly two-thirds reported over 10 years of experience. Half of the respondents had previously participated in a webinar, among which three had participated in a webinar for training purposes. All rated the knowledge and information delivered through the webinar as useful and reported it adequately prepared them for data collection. Moreover, all participants would recommend this platform for multi-site abstraction training. Consistent with participant-reported training effectiveness, results of data collection inter-rater agreement within sites ranged from 89 to 98%, with a weighted average of 95% agreement across sites.ConclusionsConducting training via web-based technology was an acceptable and effective approach to standardizing medical record review across multiple sites for this group of experienced abstractors. Given the substantial time and cost savings achieved with the webinar, coupled with participants' positive evaluation of the training session, researchers should consider this instructional method as part of training efforts to ensure high quality data collection in multi-site studies.

Dataset Information

Multi-site medical record review for validation of intentional self-harm coding in emergency departments.

Background

Methods

Results

Conclusion

Publications

Multi-site medical record review for validation of intentional self-harm coding in emergency departments.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets