Development and validation of an electronic health records-based opioid use disorder algorithm by expert clinical adjudication among patients with prescribed opioids.

BACKGROUND: In the US, over 200 lives are lost from opioid overdoses each day. Accurate and prompt diagnosis of opioid use disorders (OUD) may help prevent overdose deaths. However, international classification of disease (ICD) codes for OUD are known to underestimate prevalence, and their specificity and sensitivity are unknown. We developed and validated algorithms to identify OUD in electronic health records (EHR) and examined the validity of OUD ICD codes. METHODS: Through four iterations, we developed EHR-based OUD identification algorithms among patients who were prescribed opioids from 2014 to 2017. The algorithms and OUD ICD codes were validated against 169 independent "gold standard" EHR chart reviews conducted by an expert adjudication panel across four healthcare systems. After using 2014-2020 EHR for validating iteration 1, the experts were advised to use 2014-2017 EHR thereafter. RESULTS: Of the 169 EHR charts, 81 (48%) were reviewed by more than one expert and exhibited 85% expert agreement. The experts identified 54 OUD cases. The experts endorsed all 11 OUD criteria from the Diagnostic and Statistical Manual of Mental Disorders-5, including craving (72%), tolerance (65%), withdrawal (56%), and recurrent use in physically hazardous conditions (50%). The OUD ICD codes had 10% sensitivity and 99% specificity, underscoring large underestimation. In comparison our algorithm identified OUD with 23% sensitivity and 98% specificity. CONCLUSIONS AND RELEVANCE: This is the first study to estimate the validity of OUD ICD codes and develop validated EHR-based OUD identification algorithms. This work will inform future research on early intervention and prevention of OUD.

An Artificial Intelligence Model for Predicting Trauma Mortality Among Emergency Department Patients in South Korea: Retrospective Cohort Study.

BACKGROUND: Within the trauma system, the emergency department (ED) is the hospital's first contact and is vital for allocating medical resources. However, there is generally limited information about patients that die in the ED. OBJECTIVE: The aim of this study was to develop an artificial intelligence (AI) model to predict trauma mortality and analyze pertinent mortality factors for all patients visiting the ED. METHODS: We used the Korean National Emergency Department Information System (NEDIS) data set (N=6,536,306), incorporating over 400 hospitals between 2016 and 2019. We included the International Classification of Disease 10th Revision (ICD-10) codes and chose the following input features to predict ED patient mortality: age, sex, intentionality, injury, emergent symptom, Alert/Verbal/Painful/Unresponsive (AVPU) scale, Korean Triage and Acuity Scale (KTAS), and vital signs. We compared three different feature set performances for AI input: all features (n=921), ICD-10 features (n=878), and features excluding ICD-10 codes (n=43). We devised various machine learning models with an ensemble approach via 5-fold cross-validation and compared the performance of each model with that of traditional prediction models. Lastly, we investigated explainable AI feature effects and deployed our final AI model on a public website, providing access to our mortality prediction results among patients visiting the ED. RESULTS: Our proposed AI model with the all-feature set achieved the highest area under the receiver operating characteristic curve (AUROC) of 0.9974 (adaptive boosting [AdaBoost], AdaBoost + light gradient boosting machine [LightGBM]: Ensemble), outperforming other state-of-the-art machine learning and traditional prediction models, including extreme gradient boosting (AUROC=0.9972), LightGBM (AUROC=0.9973), ICD-based injury severity scores (AUC=0.9328 for the inclusive model and AUROC=0.9567 for the exclusive model), and KTAS (AUROC=0.9405). In addition, our proposed AI model outperformed a cutting-edge AI model designed for in-hospital mortality prediction (AUROC=0.7675) for all ED visitors. From the AI model, we also discovered that age and unresponsiveness (coma) were the top two mortality predictors among patients visiting the ED, followed by oxygen saturation, multiple rib fractures (ICD-10 code S224), painful response (stupor, semicoma), and lumbar vertebra fracture (ICD-10 code S320). CONCLUSIONS: Our proposed AI model exhibits remarkable accuracy in predicting ED mortality. Including the necessity for external validation, a large nationwide data set would provide a more accurate model and minimize overfitting. We anticipate that our AI-based risk calculator tool will substantially aid health care providers, particularly regarding triage and early diagnosis for trauma patients.

Deaths caused by osteoporotic fractures in Japan: An epidemiological study.

BACKGROUND: Osteoporosis is a global issue with a worldwide prevalence of 18.3%, and the presence of coexisting fragility fractures can reduce the survival rate by approximately 20%. In Japan, the prevalence of osteoporosis is estimated to be 12.8 million, and the annual occurrence of hip fractures is approximately 193,400. Remarkably, coexisting hip or spinal fragility fractures caused by slight external force meet the Japanese diagnostic criterion for osteoporosis regardless of bone mineral density. However, only 191 deaths due to osteoporosis were published in 2021 in Japan. With the concern that some cases of hip and spinal fragility fractures were assigned an underlying cause of death of traumatic fracture instead of osteoporosis, this study aimed to elucidate the actual number of deaths due to osteoporosis in Japan. METHODS: We used the data from Japan in 2018. First, the number of deaths due to osteoporosis and hip or spinal fractures was reviewed using published vital statistics. Second, we calculated the number of elderly deaths (age ≥80 years) resulting from hip or spinal fractures caused by falls on the same level using data from approximately 1.4 million annual individual death certificates. Combining the above data, the actual number of deaths due to osteoporosis was estimated. RESULTS: Only 190 deaths due to osteoporosis were reported in the published data. The individual certificate data revealed 3437 elderly deaths due to hip or spinal fractures caused by falls on the same level, which could meet the criteria of osteoporotic fragility fractures. Accordingly, the estimated number of deaths caused by osteoporosis was calculated as 3,627, approximately 19 times the published value. CONCLUSIONS: After researching the individual death certificate data focusing on the coexisting hip or spinal fragility fracture, it was implied that osteoporosis may have a higher mortality rate in Japan than what is published.

Validation of coding algorithms for identifying people with viral hepatitis using claims data according to different standard references.

BACKGROUND: To assess the performance of various coding algorithms for identifying people with hepatitis B virus (HBV) and hepatitis C virus (HCV) using claims data according to different reference standards (RSs) and study periods (SPs). METHODS: A proportional random sampling of 10,000 patients aged ≥ 20 years in a health care system in Southern Taiwan were enrolled as study participants. We used three hierarchical RSs (RS1: having positive results of laboratory tests; R2: having RS1 or having prescriptions of anti-HBV or anti-HCV medications; R3: having R1 or R2 or having textual diagnosis recorded in electrical medical records) with three SPs (4-, 8-, and 12-years) to calculate positive predictive value (PPV) and sensitivity (Sen) of 6 coding algorithms using HBV- and HCV-related International Classification of Disease Tenth Revision Clinical Modification (ICD-10-CM) codes in Taiwan National Health Insurance claims data for years 2016-2019. RESULTS: Of 10,000 enrolled participants, the number of participants had confirmed HBV and HCV was 146 and 165, respectively according to RS1 with 4-years SP and increased to 729 and 525, respectively according to RS3 with 12-years SP. For both HBV and HCV, the PPV was lowest according to RS1 and highest according to RS3. The longer the SP, the higher the PPV. However, the Sen was highest according to RS2 with 4-years SP. For both HBV and HCV, the coding algorithm with highest PPV and Sen was " ≥ 3 outpatient codes" and " ≥ 2 outpatient or ≥ 1 inpatients codes," respectively. CONCLUSIONS: In conclusion, using different RSs with different SPs would result in different estimation of PPV and Sen. To achieve the best yield of both PPV and Sen, the optimal coding algorithm is " ≥ 2 outpatients or ≥ 1 inpatients codes" for identifying people with HBV or HCV.

Validity of ICD-10 Codes for Cerebral Venous Thrombosis Depends on Clinical Context.

We examined the accuracy of International Classification of Disease 10th iteration (ICD-10) diagnosis codes within Canadian administrative data in identifying cerebral venous thrombosis (CVT). Of 289 confirmed cases of CVT admitted to our comprehensive stroke center between 2008 and 2018, 239/289 were new diagnoses and 204/239 were acute events with only 75/204 representing symptomatic CVTs not provoked by trauma or structural processes. Using ICD-10 codes in any position, sensitivity was 39.1% and positive predictive value was 94.2% for patients with a current or history of CVT and 84.0% and 52.5% for acute and symptomatic CVTs not provoked by trauma or structural processes.

Setting the record straight-Correcting uterine cancer incidence and mortality in the Nordic countries by reallocation of unspecified cases.

INTRODUCTION: The incidence of and mortality from cancers of the cervix uteri and corpus uteri are underestimated if the presence of uterine cancers, where the exact topography (site of origin) is not specified, is omitted. In this paper we present the corrected figures on mortality from and incidence of cervix and corpus uteri cancers in the Nordic countries by reallocating unspecified uterine cancer deaths and cases to originate either from the corpus uteri or cervix uteri. To further validate the accuracy of reallocation, we also analyzed how well the reallocation captures the changes occurring as the result of a transition in cause of death coding in Norway that took place in 2005. MATERIAL AND METHODS: This study uses data available in the NORDCAN database, which contains aggregated cancer data from all the Nordic countries for the years 1960-2016. The unspecified uterine cancer cases and deaths were reallocated to either cervix uteri or corpus uteri based on the estimated probability that follows the distribution of cases and deaths with verified topography. The estimated proportions of cases and deaths for both cancers were calculated for each combination of age group, year, and country as a proportion of cases (and deaths, respectively) with known topography. Annual age-standardized rates were calculated by direct age-adjustment. RESULTS: The proportions of unspecified uterine cancers were higher in the mortality data than in incidence data, with mean values for 1960-2016 ranging between 5.1% and 26.6% and between 0.2% and 6.8% by country, respectively. In the Nordic countries combined, the reallocation increased the number of cases by 4% and deaths by approximately 20% for both cancers. Finland was the only Nordic country where the mortality rate did not increase substantially after reallocation. CONCLUSIONS: The reallocation procedure had a significant impact on mortality from cancers of the cervix and corpus uteri for countries where the proportion of cancer deaths coded as uterus, not otherwise specified, is substantial. More effort to validate cause of death data with incidence data from cancer registries is warranted to avoid erroneous conclusions of temporal trends based on uncorrected cancer burden.

Changes in the rankings of leading causes of death in Japan, Korea, and Taiwan from 1998 to 2018: a comparison of three ranking lists.

BACKGROUND: The ranking lists used by most countries for leading causes of death (CODs) comprise broad category such as cancer, heart disease, and accidents. To provide more specific information, the World Health Organization (WHO) and the Institute of Health Metrics and Evaluation (IHME) proposed lists that splitting broad categories into specific categories. We examined the changes in rankings of leading CODs according to different lists in Japan, Korea, and Taiwan from 1998 to 2018. METHODS: We obtained the number of deaths for three countries from the WHO mortality database for 1998, 2008, and 2018. Age-standardized death rates were calculated for rankings 10 leading CODs using WHO 2000 age structure as standard. RESULTS: The first leading COD was cancer in Japan, Korea, and Taiwan from 1998 to 2018 based on government list; nevertheless, became stroke based on WHO list, and was stroke and ischemic heart disease based on IHME list. In the WHO and IHME lists, cancer is categorized based on cancer site. The number of cancer sites included in the 10 leading CODs in 2018 was 4, 4, and 3 in Japan, Korea, and Taiwan, respectively according to the WHO list and was 4, 4, and 2, respectively according to IHME list. The only difference was the rank of liver cancer in Taiwan, which was 6th according to WHO list and was 18th according to IHME list. The ranking and number of deaths for some CODs differed greatly between the WHO and IHME lists due to the reallocation of "garbage codes" into relevant specific COD in IHME list. CONCLUSIONS: Through the use of WHO and the IHME lists, the relative importance of several specific and avoidable causes could be revealed in 10 leading CODs, which could not be discerned if the government lists were used. The information is more relevant for health policy decision making.

Which ICD-9 codes were assigned for malignant mesothelioma in the mortality data in the United States before the ICD-10 was introduced?

BACKGROUND: Malignant mesothelioma (MM) is rare and fatal; survival in most cases is only about one year. Mortality rate is, therefore, a good proxy measure of incidence rate. However, the specific International Classification of Diseases (ICD) code for MM was not available until the Tenth Revision ICD (ICD-10). Little is known on which Ninth Revision ICD (ICD-9) codes were assigned for MM in the ICD-9 era. METHODS: We used a 1996 double-coded mortality file compiled by the National Center for Health Statistics to calculate the detection rate (DR) and confirmation rate (CR) of selected ICD-9 codes. RESULTS: Of 2386 decedents whose underlying cause of death was MM (ICD-10 code C45), the DR (deaths) of corresponding ICD-9 code was 57% (1365) for code 199 "malignant neoplasm without specification of site;" 19% (448) for code 162.9 "malignant neoplasm of trachea, bronchus, and lung, unspecified;" 13% (310) for code 163 "malignant neoplasm of pleura;" and 11% (271) for other codes. The CR (deaths) for the aforementioned three ICD-9 codes were 4.0% (1365/33,942), 0.3% (448/150,342), and 70.8% (310/438), respectively. CONCLUSIONS: The three ICD-9 codes (199, 162.9, and 163) were the most commonly used codes for MM and composed nine-tenths of all MM deaths in the years before the ICD-10 was introduced. Using only ICD-9 code 163, the code most often used as the surrogate measure of MM in mortality studies in the ICD-9 era, capture may have been only 13% of all MM deaths in the US, and the estimated number of MM deaths missed in 1996 would be 2086.

Construction and validation of a morbidity index based on the International Classification of Primary Care.

OBJECTIVES: In epidemiological studies it is often necessary to describe morbidity. The aim of the present study is to construct and validate a morbidity index based on the International Classification of Primary Care (ICPC-2). DESIGN AND SETTING: This is a cohort study based on linked data from national registries. An ICPC morbidity index was constructed based on a list of longstanding health problems in earlier published Scottish data from general practice and adapted to diagnostic ICPC-2 codes recorded in Norwegian general practice 2015 - 2017. SUBJECTS: The index was constructed among Norwegian born people only (N = 4 509 382) and validated in a different population, foreign-born people living in Norway (N = 959 496). MAIN OUTCOME MEASURES: Predictive ability for death in 2018 in these populations was compared with the Charlson index. Multiple logistic regression was used to identify morbidities with the highest odds ratios (OR) for death and predictive ability for different combinations of morbidities was estimated by the area under receiver operating characteristic curves (AUC). RESULTS: An index based on 18 morbidities was found to be optimal, predicting mortality with an AUC of 0.78, slightly better than the Charlson index (AUC 0.77). External validation in a foreign-born population yielded an AUC of 0.76 for the ICPC morbidity index and 0.77 for the Charlson index. CONCLUSIONS: The ICPC morbidity index performs equal to the Charlson index and can be recommended for use in data materials collected in primary health care.Key pointsThis is the first morbidity index based on the International Classification of Primary Care, 2nd edition (ICPC-2)It predicted mortality equal to the Charlson index and validated acceptably in a different populationThe ICPC morbidity index can be used as an adjustment variable in epidemiological research in primary care databases.

Model for Predicting In-Hospital Mortality of Physical Trauma Patients Using Artificial Intelligence Techniques: Nationwide Population-Based Study in Korea.

BACKGROUND: Physical trauma-related mortality places a heavy burden on society. Estimating the mortality risk in physical trauma patients is crucial to enhance treatment efficiency and reduce this burden. The most popular and accurate model is the Injury Severity Score (ISS), which is based on the Abbreviated Injury Scale (AIS), an anatomical injury severity scoring system. However, the AIS requires specialists to code the injury scale by reviewing a patient's medical record; therefore, applying the model to every hospital is impossible. OBJECTIVE: We aimed to develop an artificial intelligence (AI) model to predict in-hospital mortality in physical trauma patients using the International Classification of Disease 10th Revision (ICD-10), triage scale, procedure codes, and other clinical features. METHODS: We used the Korean National Emergency Department Information System (NEDIS) data set (N=778,111) compiled from over 400 hospitals between 2016 and 2019. To predict in-hospital mortality, we used the following as input features: ICD-10, patient age, gender, intentionality, injury mechanism, and emergent symptom, Alert/Verbal/Painful/Unresponsive (AVPU) scale, Korean Triage and Acuity Scale (KTAS), and procedure codes. We proposed the ensemble of deep neural networks (EDNN) via 5-fold cross-validation and compared them with other state-of-the-art machine learning models, including traditional prediction models. We further investigated the effect of the features. RESULTS: Our proposed EDNN with all features provided the highest area under the receiver operating characteristic (AUROC) curve of 0.9507, outperforming other state-of-the-art models, including the following traditional prediction models: Adaptive Boosting (AdaBoost; AUROC of 0.9433), Extreme Gradient Boosting (XGBoost; AUROC of 0.9331), ICD-based ISS (AUROC of 0.8699 for an inclusive model and AUROC of 0.8224 for an exclusive model), and KTAS (AUROC of 0.1841). In addition, using all features yielded a higher AUROC than any other partial features, namely, EDNN with the features of ICD-10 only (AUROC of 0.8964) and EDNN with the features excluding ICD-10 (AUROC of 0.9383). CONCLUSIONS: Our proposed EDNN with all features outperforms other state-of-the-art models, including the traditional diagnostic code-based prediction model and triage scale.

Identifying new-onset conditions and pre-existing conditions using lookback periods in Australian health administrative datasets.

BACKGROUND: The condition onset flag (COF) variable was introduced into the hospitalization coding practice in 2008 to help distinguish between the new and pre-existing conditions. However, Australian datasets collected prior to 2008 lack the COF, potentially leading to data waste. The aim of this study was to determine if an algorithm to lookback across the previous admissions could make this distinction. METHODS: All patients requiring kidney replacement therapy (KRT) identified in the Australia and New Zealand Dialysis and Transplant Registry in New South Wales, South Australia and Tasmania between July 2008 and December 2015 were linked with hospital admission datasets using probabilistic linkage. Three different lookback periods entailing either one, two or three admissions prior to the index admission were investigated. Conditions identified in an index admission but not in the lookback periods were classified as a new-onset condition. Conditions identified in both the index admission and the lookback period were deemed to be pre-existing. The degrees of agreement were determined using the kappa statistic. Conditions examined for new onset were myocardial infarction, pulmonary embolism and pneumonia. Conditions examined for prior existence were diabetes mellitus, hypertension and kidney failure. Secondary analyses evaluated whether the conditions identified as pre-existing using COF were captured consistently in the subsequent admissions. RESULTS: 11 140 patients on KRT with 69 403 admissions were analysed. Lookback over a single admission interval (Period 1) provided the highest rates of true positives with COF for all three new-onset conditions, ranging from 89% to 100%. The levels of agreement were almost perfect for all conditions (k = 0.94-1.00). This was consistent across the different time eras. All lookback periods identified additional new-onset conditions that were not classified by COF: Lookback Period 1 picked up a further 474 myocardial infarction, 84 pulmonary embolism and 1092 pneumonia episodes. Lookback Period 1 had the highest percentage of true positives when identifying the pre-existing conditions (64-80%). The level of agreement was moderate to strong and was similar across the time eras. Secondary analysis showed that not all pre-existing conditions identified using COF carried forward to the subsequent admission (61-82%) but increased when looking forward across >1 admission (87-95%). CONCLUSION: The described algorithm using a lookback period is a pragmatic, reliable and robust means of identifying the new-onset and pre-existing patient conditions, thereby enriching the existing datasets predating the availability of the COF. The findings also highlight the value of concatenating a series of hospital patient admissions to more comprehensively adjudicate the pre-existing conditions, rather than assessing the index admission alone.

Classifying stillbirths in a tertiary care hospital of India: International Classification of Disease-perinatal Mortality (ICD-PM) versus cause of death-associated condition (CODAC) system.

A number of classification system are available to classify stillbirths, but there remains a lack of a uniform global system of classification. This study evaluated the feasibility of the ICD-PM classification system and COD-AC to classify the stillbirths and to discuss the interpretation of "the newer" classification system (ICD-PM) over the COD-AC system. Over a period of one year, out of 5776 total births 314 were stillborns with a stillbirth rate of 54 per 1000 total births. As per ICD PM Classification System, 69.1% of stillbirths were ante partum and rest intrapartum. The associated maternal conditions at the time of foetal death were also classified into five groups and maximum mothers (44.3%) were grouped under M4-medical/surgical disorders. According to COD-AC system of classification 90% of cases were assigned the cause of death, rest 10% remained unexplained. The ICD-PM and CODAC classification both seem to be feasible but ICD-PM clearly defines the time of foetal death and correlates feto-maternal dyad together.IMPACT STATEMENTWhat is already known on this subject? Classifying stillbirths is crucial to recognise the actual cause of foetal death and to gather the relevant information for planning the preventive strategies especially in low middle-income countries (LMICs) which contribute to 98% of total global burden of 2.6 million stillbirths annually. In literature CODAC system was found most suitable for low middle-income countries. In 2016, WHO proposed a newer system, i.e., ICD-PM: WHO application of ICD-10 to deaths during the perinatal period.What do the results of this study add? With ICD-PM classification stillbirths were categorised more clearly in different groups and feto-maternal condition were linked together along with both intrapartum and ante partum stillbirth which can help to set the priorities and future planning for prevention. The proportion of unexplained stillbirth has also reduced significantly compared to CODAC system.What are the implications of these findings for clinical practice and/or further research? CD-PM system of classification seems feasible and would facilitate the uniform and consistent stillbirth data even from LMICs for global comparison although more number of studies are needed for conclusion. The system has been changed to ICD-PM in our institute.

Fatal anaphylaxis in Italy: Analysis of cause-of-death national data, 2004-2016.

BACKGROUND: Epidemiological data on fatal anaphylaxis are underestimated worldwide. Few Italian data do exist. The aims of the study are to determine the anaphylaxis mortality rate in Italy and its associations with demographic characteristics (gender, age, and geographical distribution), and to investigate which are the most common triggers of fatal anaphylaxis. MATERIAL AND METHODS: This is a descriptive study analyzing data reported to the National Register of Causes of Death database and managed by the Italian National Institute of Statistics for the years 2004-2016. An analytical method was developed to identify all the ICD-10 codes related to anaphylaxis deaths, which were divided into two classes: "Definite anaphylaxis deaths" and "Possible anaphylaxis deaths." RESULTS: From 2004 through 2016, 392 definite anaphylaxis deaths and 220 possible anaphylaxis deaths were recorded. The average mortality rate for definite anaphylaxis, from 2004 to 2016, was 0.51 per million population per year. Definite fatal anaphylaxis was mostly due to the use of medications (73.7%), followed by unspecified causes (20.7%) and hymenoptera stings (5.6%). Concerning possible anaphylaxis deaths, the most common cause was venom-stinging insect (51.4%). We did not find any data on food fatal anaphylaxis. Unspecified anaphylaxis accounted for 21%-28% of all cases, underlining the difficulty in accurately ascertaining the causes of fatal anaphylaxis and therefore in assigning the proper ICD-10 code. CONCLUSION: This is the first study of anaphylaxis-related mortality coming from an official database of the whole Italian population. However, the actual number of deaths by anaphylaxis, and their related triggers, is probably underreported, mostly due to limitations of the current recording system, and to a poor allergy education. Corrective actions should be undertaken for the benefit of the Health System.

Drug-related deaths: Statistics based on death certificates miss one-third of cases.

Aims: Statistics on drug-related deaths (DRD) provide crucial information on the drug situation. The European Monitoring Centre for Drug and Drug Addiction (EMCDDA) has published a specification for extracting DRD from national mortality registers to be used in international comparisons. However, surprisingly little is known of the accuracy of DRD statistics derived from national mortality registers. This study assesses the accuracy of Swedish data derived from national mortality registers by comparing it with other sources of data. Methods: We compared five Swedish datasets. Three were derived from national mortality registers, two according to a Swedish specification and one according to the EMCDDA specification. A fourth dataset was based on toxicological analyses. We used a fifth dataset, an inventory of DRD in Stockholm, to assess the completeness and coverage of the Swedish datasets. Results: All datasets were extracted from high-quality registers, but still did not capture all DRD, and both the numbers and demographic characteristics varied considerably. However, the time trends were consistent between the selections. In international comparisons, data completeness and investigation procedures may impact even more on stated numbers. Conclusions: Basing international comparisons on numbers or rates of DRDs gives misleading results, but comparing trends is still meaningful.

Analysis of causes of death among brought-in-dead cases in a third-level Hospital in Lusaka, Republic of Zambia, using the tariff method 2.0 for verbal autopsy: a cross-sectional study.

BACKGROUND: Over one third of deaths in Zambian health facilities involve someone who has already died before arrival (i.e., Brough in Dead), and in most BiD cases, the CoD have not been fully analyzed. Therefore, this study was designed to evaluate the function of automated VA based on the Tariff Method 2.0 to identify the CoD among the BiD cases and the usefulness by comparing the data on the death notification form. METHODS: The target site was one third-level hospital in the Republic of Zambia's capital city. All BiD cases who reached the target health facility from January to August 2017 were included. The deceased's closest relatives were interviewed using a structured VA questionnaire and the data were analyzed using the SmartVA to determine the CoD at the individual and population level. The CoD were compared with description on the death notification forms by using t-test and Cohen's kappa coefficient. RESULTS: One thousand three hundred seventy-eight and 209 cases were included for persons aged 13 years and older (Adult) and those aged 1 month to 13 years old (Child), respectively. The top CoD for Adults were infectious diseases followed by non-communicable diseases and that for Child were infectious diseases, followed by accidents. The proportion of cases with a determined CoD was significantly higher when using the SmartVA (75% for Adult and 67% for Child) than the death notification form (61%). A proportion (42.7% for Adult and 46% for Child) of the CoD-determined cases matched in both sources, with a low concordance rate for Adult (kappa coefficient = 0.1385) and a good for Child(kappa coefficient = 0.635). CONCLUSIONS: The CoD of the BiD cases were successfully analyzed using the SmartVA for the first time in Zambia. While there many erroneous descriptions on the death notification form, the SmartVA could determine the CoD among more BiD cases. Since the information on the death notification form is reflected in the national vital statistics, more accurate and complete CoD data are required. In order to strengthen the death registration system with accurate CoD, it will be useful to embed the SmartVA in Zambia's health information system.

Automated grouping of medical codes via multiview banded spectral clustering.

OBJECTIVE: With its increasingly widespread adoption, electronic health records (EHR) have enabled phenotypic information extraction at an unprecedented granularity and scale. However, often a medical concept (e.g. diagnosis, prescription, symptom) is described in various synonyms across different EHR systems, hindering data integration for signal enhancement and complicating dimensionality reduction for knowledge discovery. Despite existing ontologies and hierarchies, tremendous human effort is needed for curation and maintenance - a process that is both unscalable and susceptible to subjective biases. This paper aims to develop a data-driven approach to automate grouping medical terms into clinically relevant concepts by combining multiple up-to-date data sources in an unbiased manner. METHODS: We present a novel data-driven grouping approach - multi-view banded spectral clustering (mvBSC) combining summary data from multiple healthcare systems. The proposed method consists of a banding step that leverages the prior knowledge from the existing coding hierarchy, and a combining step that performs spectral clustering on an optimally weighted matrix. RESULTS: We apply the proposed method to group ICD-9 and ICD-10-CM codes together by integrating data from two healthcare systems. We show grouping results and hierarchies for 13 representative disease categories. Individual grouping qualities were evaluated using normalized mutual information, adjusted Rand index, and F1-measure, and were found to consistently exhibit great similarity to the existing manual grouping counterpart. The resulting ICD groupings also enjoy comparable interpretability and are well aligned with the current ICD hierarchy. CONCLUSION: The proposed approach, by systematically leveraging multiple data sources, is able to overcome bias while maximizing consensus to achieve generalizability. It has the advantage of being efficient, scalable, and adaptive to the evolving human knowledge reflected in the data, showing a significant step toward automating medical knowledge integration.

Positive predictive value of ICD-10 codes to detect anaphylaxis due to vaccination: A validation study.

PURPOSE: To validate the use of selected International Classification of Disease Codes 10th revision (ICD-10) to predict (positive predictive value) anaphylaxis due to vaccination using emergency department (ED) data. METHODS: We conducted a retrospective study using ED encounter data from a large tertiary-care teaching hospital, Monash Medical Centre, Melbourne, Australia. We searched all ED encounters potentially due to anaphylaxis after vaccination, between 1 January 2010 and 31 December 2018, using ICD-10-CM codes T80.5, T80.6, T88.1, T88.6, and T78.2. Health records of potential cases were examined to determine if they met the Brighton Collaboration (BC) criteria for anaphylaxis. We calculated the PPV to evaluate the accuracy of the selected ICD-10-CM codes in predicting anaphylaxis due to vaccination. RESULTS: Of the 69 health records identified and reviewed, 29 (42.2%) met the criteria for anaphylaxis regardless of the cause, and 24.6% (17/69) of records were confirmed as anaphylaxis triggered by vaccination (low positive predictive value). However, of the 23 records identified using ICD-10-CM code T80.5, 22 were classified as anaphylaxis cases regardless of the cause, and 12 were anaphylaxis due to vaccination cases giving PPV of 95.7% and 52.2%, respectively. CONCLUSIONS: Given that there is no specific ICD-10-CM code for anaphylaxis due to vaccination, ICD-10-CM code T80.5 may be suitable to monitor anaphylaxis due to vaccination in the ED setting. The current study was conducted at a single centre and needs to be confirmed by future multicentre studies.

Calcium channel blockers as drug repurposing candidates for gestational diabetes: Mining large scale genomic and electronic health records data to repurpose medications.

New therapeutic approaches are needed for gestational diabetes mellitus (GDM), but must show safety and efficacy in a historically understudied population. We studied associations between electronic medical record (EMR) phenotypes and genetic variants to uncover drugs currently considered safe in pregnancy that could treat or prevent GDM. We identified 129 systemically active drugs considered safe in pregnancy targeting the proteins produced from 196 genes. We tested for associations between GDM and/or type 2 diabetes (DM2) and 306 SNPs in 130 genes represented on the Illumina Infinium Human Exome Bead Chip (DM2 was included due to shared pathophysiological features with GDM). In parallel, we tested the association between drugs and glucose tolerance during pregnancy as measured by the glucose recorded during a routine 50-g glucose tolerance test (GTT). We found an association between GDM/DM2 and the genes targeted by 11 drug classes. In the EMR analysis, 6 drug classes were associated with changes in GTT. Two classes were identified in both analyses. L-type calcium channel blocking antihypertensives (CCBs), were associated with a 3.18 mg/dL (95% CI -6.18 to -0.18) decrease in glucose during GTT, and serotonin receptor type 3 (5HT-3) antagonist antinausea medications were associated with a 3.54 mg/dL (95% CI 1.86-5.23) increase in glucose during GTT. CCBs were identified as a class of drugs considered safe in pregnancy could have efficacy in treating or preventing GDM. 5HT-3 antagonists may be associated with worse glucose tolerance.

Exploration of association rule mining for coding consistency and completeness assessment in inpatient administrative health data.

OBJECTIVE: Data quality assessment is a challenging facet for research using coded administrative health data. Current assessment approaches are time and resource intensive. We explored whether association rule mining (ARM) can be used to develop rules for assessing data quality. MATERIALS AND METHODS: We extracted 2013 and 2014 records from the hospital discharge abstract database (DAD) for patients between the ages of 55 and 65 from five acute care hospitals in Alberta, Canada. The ARM was conducted using the 2013 DAD to extract rules with support ≥0.0019 and confidence ≥0.5 using the bootstrap technique, and tested in the 2014 DAD. The rules were compared against the method of coding frequency and assessed for their ability to detect error introduced by two kinds of data manipulation: random permutation and random deletion. RESULTS: The association rules generally had clear clinical meanings. Comparing 2014 data to 2013 data (both original), there were 3 rules with a confidence difference >0.1, while coding frequency difference of codes in the right hand of rules was less than 0.004. After random permutation of 50% of codes in the 2014 data, average rule confidence dropped from 0.72 to 0.27 while coding frequency remained unchanged. Rule confidence decreased with the increase of coding deletion, as expected. Rule confidence was more sensitive to code deletion compared to coding frequency, with slope of change ranging from 1.7 to 184.9 with a median of 9.1. CONCLUSION: The ARM is a promising technique to assess data quality. It offers a systematic way to derive coding association rules hidden in data, and potentially provides a sensitive and efficient method of assessing data quality compared to standard methods.

Association between delivery of small-for-gestational age neonate and long-term pediatric ophthalmic morbidity.

OBJECTIVE: To investigate whether delivery of small for gestational age (SGA) neonate poses a risk for subsequent long-term ophthalmic morbidity. METHODS: In this population-based study, all deliveries between 1991 and 2014 were included. Congenital malformations and multiple gestations were excluded from the analysis. Offspring were defined as either SGA (weight below the 5th percentile for gestational age) or non-SGA. Comparison was performed regarding the incidence of long-term ophthalmic morbidity in a cohort of neonates who were born SGA and those who were not. Ophthalmic morbidity was documented during any encounter with the hospital for a period of up to 18 years after delivery. Ophthalmic morbidity included infections of the eye or the adnexa, inflammation of any cause requiring admission, visual disturbances, and other hospital admissions carrying an ICD-9 code of ophthalmic designation. A Cox proportional hazards model was used to estimate the adjusted hazards ratio (HR) for ophthalmic morbidity During the study period, 243,682 deliveries met the inclusion criteria, of which 11,290 (4.63%) were defined as SGA. RESULTS: During the follow-up period, SGA neonates had higher rates of ophthalmic-related hospitalizations (1.2% versus 1.0%; OR = 1.22, 95% CI 1.02-1.46; p = 0.026). In a Cox proportional hazards model, adjusted for confounders such as maternal age, gestational age at delivery, child birth year, low 5 min Apgar scores (< 7), gestational diabetes, maternal hypertensive disorders, placental abruption and placenta previa, SGA neonate was independently associated with subsequent long-term ophthalmic morbidity (adjusted HR = 1.22; 95% CI 1.02-1.47; p = 0.024). CONCLUSION: Delivery of an SGA neonate is an independent risk factor for long-term ophthalmic morbidity.