Preoperative Depression is Associated With Increased Complications Following Ankle Fracture Surgery

Background The effects of preoperative depression following ankle fracture surgery remains unknown. The purpose of this study is to investigate the relationship between preoperative depression and outcomes following ankle fracture surgery. Methods This retrospective study used the Truven MarketScan database to identify patients who underwent ankle fracture surgery from January 2009 to December 2018. Patients with and without a diagnosis of preoperative depression were identified based on International Classification of Diseases (ICD) codes. Chi-squared and multivariate analyses were performed to determine the association between preoperative depression and postoperative complications following ankle fracture surgery. Results In total, 107,897 patients were identified for analysis, 13,981 of whom were diagnosed with depression (13%). Preoperative depression was associated with the increased odds for postoperative infection (odds ratio [OR]: 1.33, confidence interval [CI]: 1.20-1.46), wound complications (OR: 1.13, CI: 1.00-1.28), pain-related postoperative emergency department visits (OR: 1.58, CI: 1.30-19.1), 30-day and 90-day readmissions (OR: 1.08, CI: 1.03-1.21 and OR: 1.13, CI: 1.07-1.18), sepsis (OR: 1.39, CI: 1.12-1.72), and postoperative development of complex regional pain syndrome (OR: 1.46, CI: 1.18-1.81). Conclusion Preoperative depression is associated with increased complications following ankle fracture surgery. Further studies are warranted to investigate the degree to which depression is a modifiable risk factor. Level of Evidence: 3

Comparison of different feature extraction methods for applicable automated ICD coding

Background Automated ICD coding on medical texts via machine learning has been a hot topic. Related studies from medical field heavily relies on conventional bag-of-words (BoW) as the feature extraction method, and do not commonly use more complicated methods, such as word2vec (W2V) and large pretrained models like BERT. This study aimed at uncovering the most effective feature extraction methods for coding models by comparing BoW, W2V and BERT variants. Methods We experimented with a Chinese dataset from Fuwai Hospital, which contains 6947 records and 1532 unique ICD codes, and a public Spanish dataset, which contains 1000 records and 2557 unique ICD codes. We designed coding tasks with different code frequency thresholds (denoted as $$f_s$$ f s ), with a lower threshold indicating a more complex task. Using traditional classifiers, we compared BoW, W2V and BERT variants on accomplishing these coding tasks. Results When $$f_s$$ f s was equal to or greater than 140 for Fuwai dataset, and 60 for the Spanish dataset, the BERT variants with the whole network fine-tuned was the best method, leading to a Micro-F1 of 93.9% for Fuwai data when $$f_s=200$$ f s = 200 , and a Micro-F1 of 85.41% for the Spanish dataset when $$f_s=180$$ f s = 180 . When $$f_s$$ f s fell below 140 for Fuwai dataset, and 60 for the Spanish dataset, BoW turned out to be the best, leading to a Micro-F1 of 83% for Fuwai dataset when $$f_s=20$$ f s = 20 , and a Micro-F1 of 39.1% for the Spanish dataset when $$f_s=20$$ f s = 20 . Our experiments also showed that both the BERT variants and BoW possessed good interpretability, which is important for medical applications of coding models. Conclusions This study shed light on building promising machine learning models for automated ICD coding by revealing the most effective feature extraction methods. Concretely, our results indicated that fine-tuning the whole network of the BERT variants was the optimal method for tasks covering only frequent codes, especially codes that represented unspecified diseases, while BoW was the best for tasks involving both frequent and infrequent codes. The frequency threshold where the best-performing method varied differed between different datasets due to factors like language and codeset.

Comparison of two different frailty scales in the longitudinal Swedish Adoption/Twin Study of Aging (SATSA)

Aims: Although up to 25% of older adults are frail, assessing frailty can be difficult, especially in registry data. This study evaluated the utility of a code-based frailty score in registry data by comparing it to a gold-standard frailty score to understand how frailty can be quantified in population data and perhaps better addressed in healthcare. Methods: We compared the Hospital Frailty Risk Score (HFRS), a frailty measure based on 109 ICD codes, to a modified version of the Frailty Index (FI) Frailty Index (FI), a self-report frailty measure, and their associations with all-cause mortality both cross-sectionally and longitudinally (follow-up = 36 years) in a Swedish cohort study ( n = 1368). Results: The FI and HFRS were weakly correlated (rho = 0.11, p < 0.001). Twenty-two percent ( n = 297) of participants were considered frail based on published cut-offs of either measure. Only 3% ( n = 35) of participants were classified as frail by both measures; 4% ( n = 60) of participants were considered frail by only the HFRS; and 15% ( n = 202) of participants were considered frail based only on the FI. Frailty as measured by the HFRS showed greater variance and no clear increase or decrease with age, while frailty as measured by the FI increased steadily with age. In adjusted Cox proportional hazard models, baseline HFRS frailty (HR = 1.17, 95% CI 0.92, 1.49) was not statistically significantly associated with mortality, while FI frailty was (HR = 2.89, 95% CI 1.61, 2.23). These associations were modified by age and sex. Conclusions: The HFRS may not capture the full spectrum of frailty among community-dwelling individuals, particularly at younger ages, in Swedish registry data.

Prevalence and Incidence Rates of Dementia: A Nationwide Population-Based Study of Electronic Health Records in Israel1

Background: Data on the rate of dementia is essential for planning and developing appropriate services at the national level. Objective: We report the prevalence and incidence of dementia, based on electronic health records available for the whole population. Methods: This national dementia dataset was established as a part of the National Program to Address Alzheimer’s and Other Types of Dementia. Data from medical health records for all persons aged 45+ in Israel, for 2016, were extracted from the databases of the four health maintenance organizations. Dementia cases were identified based on either recorded dementia diagnosis, through International Classification of Diseases (ICD-9 and ICD-10) or dispensation of anti-dementia drugs. The date of first diagnosis was determined by the earliest recording. Results: A total of 65,951 persons with dementia, aged 45+, were identified from electronic health data. Based on both ICD codes and anti-dementia drugs, the prevalence rates of dementia among individuals aged 45+ and 65+ in 2016 were 2.5%and 6.4%, respectively, and the incidence rates were 0.49%and 1.3%, respectively. Based on ICD codes alone, the prevalence rates of dementia among individuals aged 45+ and 65+ in 2016 were 2.1%and 5.4%respectively, and the incidence rates were 0.36%and 0.96%respectively. The rates were higher among females compared to males and paradoxically lower in lower socioeconomic status compared to higher statuses. Conclusion: This data collection reflects the present access of dementia patients to medical care resources and provides the basis for service planning and future dementia policies.

Development of a Natural Language Processing Machine to Enhance Identification of Opioid Use Disorder in Electronic Health Records

Background/Objective: Real-world data, including electronic health records (EHRs), has shown tremendous utility in research relating to opioid use disorder (OUD). Traditional analysis of EHR data relies on explicit diagnostic codes and results in incomplete capture of cases and therefore underrepresentation of OUD rates. Machine learning can rectify this by surveying free clinical notes in addition to structured codes. This study aimed to address disparities between true OUD rates and cases identified using traditional ICD codes by developing a natural language processing (NLP) machine for identifying affected patients from EHRs. Methods: Patients (≥12 years old) who had received an opioid prescription from IU Health or Eskenazi Health between 1/1/2009 and 12/31/2015 were identified by the Regenstrief Institute. Exclusion criteria included any cancer, sickle cell anemia, or palliative care diagnoses. Cases of OUD were identified through ICD codes and NLP. The NLP machine was developed using a dictionary of key OUD terms and a training corpus of 300 patient notes. A testing corpus of 148 patient notes was constructed and validated by manual review. The NLP machine and ICD 9/10 codes were independently tested against this corpus. Results: Although ICD codes identified OUD cases with high specificity (98.08%), this method demonstrated moderate sensitivity (53.13%), accuracy (68.92%), and F1 score (68.92%). Testing using the NLP method demonstrated increased sensitivity (93.75%), increased accuracy (89.19%), and increased F1 score (91.84%); specificity mildly decreased (80.77%). Conclusion: Our revised NLP machine was more effective at capturing OUD cases in EHRs than traditional identification using ICD codes. This illustrates NLP’s enhanced capability of identifying OUD cases compared to structured data. Potential Impacts: These findings establish a role for NLP in OUD research involving large datasets. Ultimately, this is intended to improve identification of risk factors for OUD, which is of significant clinical importance during a public health crisis.

Racial and Ethnic Variations in Mortality Associated with Common Pediatric Gastrointestinal Disorders in Children's Hospitals in the US

Background: The objective of this study is to determine the mortality, risk factors, and disease associations of eight common pediatric gastrointestinal (GI) disorders: cystic fibrosis (CF), cirrhosis, gastrointestinal bleeding (GIB), inflammatory bowel disease (IBD), liver failure (LF), liver transplant, acute pancreatitis, and short bowel syndrome (SBS). Methods: Diagnoses were found using the International Classification of Disease (ICD) codes from 2004 through 2020. We performed a retrospective cohort study using the Pediatric Health Information System (PHIS) database from 50 children’s hospitals in the US. We analyzed all encounters with ICD codes for these disorders, then determined the per-encounter mortality rate for each. We performed a mixed-effects logistic regression modeling hospital as a random effect, mortality as the dependent variable, and patient demographics and medical history as independent variables. We hypothesized that demographic factors such as Black race, Hispanic ethnicity, and markers of socioeconomic status would be associated with increased mortality. Results: The per-encounter mortality for each diagnosis was: cirrhosis (2.19%), CF (0.66%), GIB (4.22%), IBD (0.21%), LF (7.03%), liver transplant (0.37%), acute pancreatitis (2.23%), and SBS (1.13%). There was a higher (p<0.05) mortality for those of Asian race and mixed-race populations in GIB (OR 1.76 and 1.37, respectively) and acute pancreatitis (OR 1.94 and 1.34, respectively). For those of Black race, there was a higher mortality in liver transplant and liver failure (OR 1.31 and 1.65 respectively). Additionally, mortality was increased in Hispanic/Latino patients with CF, GIB, and SBS (OR 2.34, 1.39, and 1.41, respectively). Coincident cardiovascular, renal/urologic, and neurologic/neuromuscular abnormalities were also associated with a significant higher mortality. Conclusion: The degree of variation associated with race and ethnicity is unlikely to be accounted for by variation in clinical features, thus the impact of social determinants of health should be the focus of future study. Cirrhosis CF GIB IBD LF Liver Transplant Acute Pancreatitis SBS Mortality Rate 2.19% 0.66% 4.22% 0.21% 7.03% 0.37% 2.23% 1.13% Asian Race OR 1.10 3.77 1.76* 1.71 1.08 0.89 1.94* 1.35 Black Race OR 0.90 0.48 1.03 1.55 1.31* 1.65* 1.15 1.14 Mixed Race OR 1.16 1.12 1.37* 1.44 1.23 0.85 1.34* 1.16 Hispanic/Latino OR 1.13 2.34* 1.39* 1.53 1.16 1.65 1.13 1.41* *Significant OR numbers with an associated p<0.05

Frailty prevalence and association with mortality across birth cohorts in Swedish registry data

Frailty is associated with poor health outcomes, reduced quality of life, and mortality. To understand how prevalence of frailty may have changed across birth cohorts, we investigated frailty prevalence at ages 75, 85, and 95 in people born in 1910, 1920, and 1930 in Swedish national registry data. Frailty was assessed with the Hospital Frailty Risk Score, a weighted sum of 109 ICD codes, which we calculated for each year leading up to the specified ages. We additionally investigated the association between frailty and mortality in these birth cohorts. We observed, at 75, a decrease in prevalence of frailty across birth cohorts (16.9%, 10.8%, and 8.8%, respectively). Interestingly, at 85, we found a U-shaped pattern, where those born in 1920 (14.1%) had lower prevalence of frailty than those born in either 1910 (27.7%) or 1930 (25.1%). At age 95, we saw a low prevalence of frailty in the 1910 (7.3%) and 1920 (3.8%) birth cohorts –potentially because of selective survival. There were not substantial differences in prevalence of frailty by sex or birth country. In Cox proportional hazard models adjusted for sex, frailty was consistently associated with mortality. We observed the greatest hazard ratios in the 1930 birth cohort at 75 (HR=2.79, 95% CI 2.62, 2.97) and 85 (HR=2.26, 95% CI 2.01, 2.53) and the 1920 birth cohort at 75 (HR=2.19, 95% CI 2.09, 2.29), where risk was double that of those who were not frail. Understanding changes in prevalence of frailty will help inform public health and intervention measures.

Comparison of the rate of delayed/nonunion in fifth metatarsal fractures receiving anti-inflammatory medications

Purpose Controversy exists regarding the acute effect of non-steroidal anti-inflammatory drugs (NSAIDs) on early fracture healing. The purpose of this study was to analyze the rate of nonunion or delayed union in patients with fifth metatarsal (5th MT) fractures. We hypothesize that the use of NSAIDs would increase the rate of nonunion/delayed union in 5th MT fractures. Methods Using PearlDiver, a national insurance database was analyzed. ICD codes were used to identify patients diagnosed with 5th MT fracture from 2007-2018. Patients were grouped by initial management (nonoperative vs. open reduction and internal fixation (ORIF) or non/malunion repair within 60 days) and sub-grouped by whether they had been prescribed at least one pre-defined NSAID. Subsequent ORIF or nonunion/malunion repair operative intervention was used as a surrogate for fracture nonunion/delayed union. Results Of the 10,991 subjects with a diagnosis of 5th MT, 10,626 (96.7%) underwent initial nonoperative treatment, 1,409 of which (13.3%) received prescription NSAIDS within 60 days of diagnosis. 16/1,409 (1.14%) subjects who received anti-inflammatory prescriptions underwent ORIF or repair of non/malunion at least 60 days after diagnosis while 46/9,217 (0.50%; P=0.003483) subjects who did not receive anti-inflammatory prescriptions underwent ORIF or repair of non/malunion at least 60 days after diagnosis. In the 365 subjects who underwent early repair/ORIF (within 60 days), there was no significant difference in the rate of nonunion/delayed union. Conclusion The rate of nonunion/delayed union of 5th MT fractures was significantly higher in subjects receiving NSAIDs within 60 days of initial diagnosis in patients managed non-operatively. Level of evidence Level III

Exposure to toxic occupations and their association with Parkinson’s disease: a systematic review with meta-analysis

Objectives Earlier longitudinal reviews on environmental and occupational toxins and Parkinson’s disease (PD) risk have limitations. This study aimed to determine the strength of association between three types of toxic occupational exposures and the occurrence of PD by diagnostic methods. Methods A search was conducted of EMBASE, PubMed/Medline, Toxnet, LILACS, and Cochrane Library databases for longitudinal studies that assessed toxic occupational exposure, Parkinsonian, or related disorders, diagnosed by International Classification of Diseases (ICD) codes, medical records, or confirmation by a neurologist/nurse, and published in the English language from January 1990 to July 2021. Pooled risk ratios (RR) estimates were produced using random-effects models. Systematic review with meta-analysis synthesized the results. Study quality, heterogeneity, and publication bias were examined. High-quality articles that met the inclusion criteria were analyzed. Results Twenty-four articles were used in the analyses. The pooled RR for electromagnetic exposure and PD were (RR=1.03, 95% confidence interval [CI] 0.91–1.16) while the pooled RR between PD and metal and pesticide exposure were (RR=1.07, 95% CI 0.92–1.24) and (RR=1.41, 95% CI 1.20–1.65), respectively. Pooled RR for methods of diagnosis and their associations with PD were: confirmation by a neurologist or nurse (RR=2.17, 95% CI 1.32–3.54); ICD codes (RR=1.14, 95% CI 1.03–1.26), and medical records (RR=1.06, 95% CI 0.92–1.21). Conclusions Our systematic review provides robust evidence that toxic occupational exposures are significant risk factors for PD especially those diagnosed by neurologists or nurses using standardized methods.

The importance of social activity to risk of major depression in older adults

Background Several social determinants of health (SDoH) have been associated with the onset of major depressive disorder (MDD). However, prior studies largely focused on individual SDoH and thus less is known about the relative importance (RI) of SDoH variables, especially in older adults. Given that risk factors for MDD may differ across the lifespan, we aimed to identify the SDoH that was most strongly related to newly diagnosed MDD in a cohort of older adults. Methods We used self-reported health-related survey data from 41 174 older adults (50–89 years, median age = 67 years) who participated in the Mayo Clinic Biobank, and linked ICD codes for MDD in the participants' electronic health records. Participants with a history of clinically documented or self-reported MDD prior to survey completion were excluded from analysis (N = 10 938, 27%). We used Cox proportional hazards models with a gradient boosting machine approach to quantify the RI of 30 pre-selected SDoH variables on the risk of future MDD diagnosis. Results Following biobank enrollment, 2073 older participants were diagnosed with MDD during the follow-up period (median duration = 6.7 years). The most influential SDoH was perceived level of social activity (RI = 0.17). Lower level of social activity was associated with a higher risk of MDD [hazard ratio = 2.27 (95% CI 2.00–2.50) for highest v. lowest level]. Conclusion Across a range of SDoH variables, perceived level of social activity is most strongly related to MDD in older adults. Monitoring changes in the level of social activity may help identify older adults at an increased risk of MDD.