Cause of death coding in asthma.

BACKGROUND: While clinical coding is intended to be an objective and standardized practice, it is important to recognize that it is not entirely the case. The clinical and bureaucratic practices from event of death to a case being entered into a research dataset are important context for analysing and interpreting this data. Variation in practices can influence the accuracy of the final coded record in two different stages: the reporting of the death certificate, and the International Classification of Diseases (Version 10; ICD-10) coding of that certificate. METHODS: This study investigated 91,022 deaths recorded in the Scottish Asthma Learning Healthcare System dataset between 2000 and 2017. Asthma-related deaths were identified by the presence of any of ICD-10 codes J45 or J46, in any position. These codes were categorized either as relating to asthma attacks specifically (status asthmatic; J46) or generally to asthma diagnosis (J45). RESULTS: We found that one in every 200 deaths in this were coded as being asthma related. Less than 1% of asthma-related mortality records used both J45 and J46 ICD-10 codes as causes. Infection (predominantly pneumonia) was more commonly reported as a contributing cause of death when J45 was the primary coded cause, compared to J46, which specifically denotes asthma attacks. CONCLUSION: Further inspection of patient history can be essential to validate deaths recorded as caused by asthma, and to identify potentially mis-recorded non-asthma deaths, particularly in those with complex comorbidities.

Use of open-text responses to recode categorical survey data on postpartum contraception use among women in the United States: A mixed-methods inquiry of Pregnancy Risk Assessment Monitoring System data.

BACKGROUND: Postpartum contraception prevents unintended pregnancies and short interpregnancy intervals. The Pregnancy Risk Assessment Monitoring System (PRAMS) collects population-based data on postpartum contraception nonuse and reasons for not using postpartum contraception. In addition to quantitative questions, PRAMS collects open-text responses that are typically left unused by secondary quantitative analyses. However, abundant preexisting open-text data can serve as a resource for improving quantitative measurement accuracy and qualitatively uncovering unexpected responses. We used PRAMS survey questions to explore unprompted reasons for not using postpartum contraception and offer insight into the validity of categorical responses. METHODS AND FINDINGS: We used 31,208 categorical 2012 PRAMS survey responses from postpartum women in the US to calculate original prevalences of postpartum contraception use and nonuse and reasons for contraception nonuse. A content analysis of open-text responses systematically recoded data to mitigate survey bias and ensure consistency, resulting in adjusted prevalence calculations and identification of other nonuse themes. Recoded contraception nonuse slightly differed from original reports (21.5% versus 19.4%). Both calculations showed that many respondents reporting nonuse may be at a low risk for pregnancy due to factors like tubal ligation or abstinence. Most frequent nonuse reasons were not wanting to use birth control (27.1%) and side effect concerns (25.0%). Other open-text responses showed common themes of infertility, and breastfeeding as contraception. Comparing quantitative and qualitative responses revealed contradicting information, suggesting respondent misinterpretation and confusion surrounding the term "pregnancy prevention." Though this analysis may be limited by manual coding error and researcher biases, we avoided coding exhaustion via 1-hour coding periods and validated reliability through intercoder kappa scores. CONCLUSIONS: In this study, we observed that respondents reporting contraception nonuse often described other methods of pregnancy prevention and contraception barriers that were not included in categorical response options. Open-text responses shed light on a more comprehensive list of pregnancy prevention methods and nonuse options. Our findings contribute to survey questions that can lead to more accurate depiction of postpartum contraceptive behavior. Additionally, future use of these qualitative methods may be used to improve other health behavior survey development and resulting data.

Explainable automated coding of clinical notes using hierarchical label-wise attention networks and label embedding initialisation.

BACKGROUND: Diagnostic or procedural coding of clinical notes aims to derive a coded summary of disease-related information about patients. Such coding is usually done manually in hospitals but could potentially be automated to improve the efficiency and accuracy of medical coding. Recent studies on deep learning for automated medical coding achieved promising performances. However, the explainability of these models is usually poor, preventing them to be used confidently in supporting clinical practice. Another limitation is that these models mostly assume independence among labels, ignoring the complex correlations among medical codes which can potentially be exploited to improve the performance. METHODS: To address the issues of model explainability and label correlations, we propose a Hierarchical Label-wise Attention Network (HLAN), which aimed to interpret the model by quantifying importance (as attention weights) of words and sentences related to each of the labels. Secondly, we propose to enhance the major deep learning models with a label embedding (LE) initialisation approach, which learns a dense, continuous vector representation and then injects the representation into the final layers and the label-wise attention layers in the models. We evaluated the methods using three settings on the MIMIC-III discharge summaries: full codes, top-50 codes, and the UK NHS (National Health Service) COVID-19 (Coronavirus disease 2019) shielding codes. Experiments were conducted to compare the HLAN model and label embedding initialisation to the state-of-the-art neural network based methods, including variants of Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs). RESULTS: HLAN achieved the best Micro-level AUC and F1 on the top-50 code prediction, 91.9% and 64.1%, respectively; and comparable results on the NHS COVID-19 shielding code prediction to other models: around 97% Micro-level AUC. More importantly, in the analysis of model explanations, by highlighting the most salient words and sentences for each label, HLAN showed more meaningful and comprehensive model interpretation compared to the CNN-based models and its downgraded baselines, HAN and HA-GRU. Label embedding (LE) initialisation significantly boosted the previous state-of-the-art model, CNN with attention mechanisms, on the full code prediction to 52.5% Micro-level F1. The analysis of the layers initialised with label embeddings further explains the effect of this initialisation approach. The source code of the implementation and the results are openly available at https://github.com/acadTags/Explainable-Automated-Medical-Coding. CONCLUSION: We draw the conclusion from the evaluation results and analyses. First, with hierarchical label-wise attention mechanisms, HLAN can provide better or comparable results for automated coding to the state-of-the-art, CNN-based models. Second, HLAN can provide more comprehensive explanations for each label by highlighting key words and sentences in the discharge summaries, compared to the n-grams in the CNN-based models and the downgraded baselines, HAN and HA-GRU. Third, the performance of deep learning based multi-label classification for automated coding can be consistently boosted by initialising label embeddings that captures the correlations among labels. We further discuss the advantages and drawbacks of the overall method regarding its potential to be deployed to a hospital and suggest areas for future studies.

Gender Disparities in Stroke Code Activation in Patients with Intracerebral Hemorrhage.

OBJECTIVES: Routine implementation of protocol-driven stroke "codes" results in timelier and more effective acute stroke management. However, it is unclear if patient demographics contribute to disparities in stroke code activation. We aimed to explore these demographic factors in a retrospective cohort study of patients with intracerebral hemorrhage (ICH). MATERIALS AND METHODS: We identified consecutive patients with non-traumatic ICH who presented directly to our Comprehensive Stroke Center over 2 years and collected data on demographics, clinical features, and stroke code activation. We used multivariable logistic regression to examine differences in stroke code activation based on patient demographics while adjusting for initial clinical features (NIH Stroke Scale, FAST [facial drooping, arm weakness, speech difficulties] vs. non-FAST symptoms, time from last-known-well [LKW], and systolic blood pressure [SBP]). RESULTS: Among 265 patients, 68% (n=179) had a stroke code activation. Stroke codes occurred less frequently in women (62%) than men (72%) and in non-white (57%) vs. white patients (70%). Non-stroke code patients were less likely to have FAST symptoms (37% vs. 87%) and had lower initial SBP (mean±SD 159.3±34.2 vs. 176.0±31.9 mmHg) than stroke code patients. In our primary multivariable models, neither age nor race were associated with stroke code activation. However, women were significantly less likely to have stroke codes than men (OR 0.49 [95% CI 0.24-0.98]), as were non-FAST symptoms (OR 0.11 [95% CI 0.05-0.22]). CONCLUSIONS: Our data suggest gender disparities in emergency stroke care that should prompt further investigations into potential systemic biases. Increased awareness of atypical stroke symptoms is also warranted.

Patient Coded Severity and Payment Penalties Under the Hospital Readmissions Reduction Program: A Machine Learning Approach.

OBJECTIVE: The objective of this study was to examine variation in hospital responses to the Centers for Medicare and Medicaid's expansion of allowable secondary diagnoses in January 2011 and its association with financial penalties under the Hospital Readmission Reduction Program (HRRP). DATA SOURCES/STUDY SETTING: Medicare administrative claims for discharges between July 2008 and June 2011 (N=3102 hospitals). RESEARCH DESIGN: We examined hospital variation in response to the expansion of secondary diagnoses by describing changes in comorbidity coding before and after the policy change. We used random forest machine learning regression to examine hospital characteristics associated with coded severity. We then used a 2-part model to assess whether variation in coded severity was associated with readmission penalties. RESULTS: Changes in severity coding varied considerably across hospitals. Random forest models indicated that greater baseline levels of condition categories, case-mix index, and hospital size were associated with larger changes in condition categories. Hospital coding of an additional condition category was associated with a nonsignificant 3.8 percentage point increase in the probability for penalties under the HRRP (SE=2.2) and a nonsignificant 0.016 percentage point increase in penalty amount (SE=0.016). CONCLUSION: Changes in patient coded severity did not affect readmission penalties.

Administrative data identify sickle cell disease: A critical review of approaches in U.S. health services research.

To identify people living with sickle cell disease (SCD) and study their healthcare utilization, researchers can either use clinical records linked to administrative data or use billing diagnosis codes in stand-alone administrative databases. Correct identification of individuals clinically managed for SCD using diagnosis codes in claims databases is limited by the accuracy of billing codes in outpatient encounters. In this critical review, we assess the strengths and limitations of claims-based SCD case-finding algorithms in stand-alone administrative databases that contain both inpatient and outpatient records. Validation studies conducted using clinical records and newborn screening for confirmation of SCD case status have found that algorithms that require three or more nonpharmacy claims or one inpatient claim plus two or more outpatient claims with SCD codes show acceptable accuracy (positive predictive value and sensitivity) in children and adolescents. Future studies might seek to assess the accuracy of case-finding algorithms over the lifespan.

A Study on the Status of Proton Pump Inhibitor Prescriptions Using Diagnosis Procedure Combination Data in Japan.

BACKGROUND/AIMS: Proton pump inhibitors (PPIs) are widely used for the management of acid-related diseases. This study aimed to clarify the status of PPI use in hospitalized patients. METHOD: A retrospective observational study was performed. We analyzed PPI prescription data for the past 8 years (2009-2016) using Diagnosis Procedure Combination survey data from approximately 10,000 patients per year at Saga University Hospital. We investigated the trend in the number of hospitalized patients who were prescribed PPI for the past 8 years and the changes in patient characteristics. RESULTS: We identified 11,009 patients using PPIs throughout the study period. PPI prescription proportions significantly increased over the study period. The use of PPIs increased steadily with increasing age. The proportion of PPIs prescribed was 1.3-times higher among men than compared with women. In most clinical departments, the number of patients prescribed PPIs was up to 3 times higher in the second half of the period (2013-2016) compared with the first half of the period (2009-2012). The number of patients taking concomitant PPIs and anticoagulants or dual antithrombotic combination therapy increased. CONCLUSION: PPI use has increased substantially in hospitalized patients. The prevalence of PPI prescription by doctors other than gastroenterologists also increased.

Physician code creep after the initiation of outpatient volume control program and implications for appropriate ICD-10-CM coding.

BACKGROUND: Most studies on the physician code creep (i.e., changes in case mix record-keeping practices to improve reimbursement) have focused on episodes (inpatient hospitalizations or outpatient procedures). Little is known regarding changes in diagnostic coding practices for better reimbursement among a fixed cohort of patients with chronic diseases. METHODS: To examine whether physicians in tertiary medical centers changed their coding practices after the initiation of the Outpatient Volume Control Program (OVCP) in Taiwan, we conducted a retrospective observational study of four patient cohorts (two interventions and two controls) from January 2016 to September 2017 in Taiwan. The main outcomes were the number of outpatient visits with four coding practices: 1) OVCP monitoring code recorded as primary diagnosis; 2) OVCP monitoring code recorded as secondary diagnosis; 3) non-OVCP monitoring code recorded as primary diagnosis; 4) non-OVCP monitoring code recorded as secondary diagnosis. RESULTS: The percentage change of the number of visits with coding practice 1 between 2016Q1 and 2017Q3 was - 74% for patients with hypertension and - 73% with diabetes in tertiary medical centers and - 23% and - 17% in clinics, respectively. By contrast, the percentage changes of coding practice 3 were + 73% for patients with hypertension and + 46% for patients with diabetes in tertiary medical centers and - 19% and - 2% in clinics, respectively. CONCLUSIONS: Physician code creep occurred after the initiation of the OVCP. Education regarding appropriate outpatient coding for physicians will be relatively effective when proper coding is related to reimbursement.

Capturing adverse events in elective orthopedic surgery: comparison of administrative, surgeon and reviewer reporting

Summary: Ensuring adverse event (AE) recording is standardized and accurate is paramount for patient safety. In this discussion, we outline our comparison of AE data collected by orthopedic surgeons and independent clinical reviewers using the Spine Adverse Events Severity System (SAVES) and Orthopedic Surgical Adverse Events Severity System (OrthoSAVES) against AE data recorded by hospital administrative discharge abstract coders. In 164 spine, hip, knee and shoulder patients, reviewers recorded significantly more AEs than coders, and coders recorded significantly more AEs than surgeons. The AEs were recorded similarly by reviewers using SAVES and OrthoSAVES in 48 spine patients. Despite our small sample size and use of different AE tools, we believe it is important to highlight that coders, surgeons and reviewers recorded AEs differently. While further investigations on its utility and cost-effectiveness are necessary, we assert that it is feasible to use Ortho-SAVES to prospectively record AEs across all orthopedic subspecialties.

Underestimation of Liver Transplantation for Alcoholic Hepatitis in the National Transplant Database.

Alcohol-associated liver disease (ALD) can be coded in United Network for Organ Sharing (UNOS) as either alcoholic cirrhosis or alcoholic hepatitis (AH), without having specific criteria to assign either diagnosis. In this multicenter American Consortium of Early Liver Transplantation for Alcoholic Hepatitis (ACCELERATE-AH) study, we sought to assess the concordance of the clinician diagnosis of AH at liver transplantation (LT) listing versus UNOS data entry of AH as listing diagnosis. In a prior study, consecutive early LT recipients transplanted for AH between 2012 and 2017 were identified by chart review at 10 ACCELERATE-AH sites. In this current study, these same LT recipients were identified in the UNOS database. The primary UNOS diagnostic code was evaluated for concordance with the chart-review assignment of AH. In cases where the primary listing diagnosis in UNOS was not AH, we determined the reason for alternate classification. Among 124 ACCELERATE-AH LT recipients with a chart-review diagnosis of AH, only 43/124 (35%) had AH as listing diagnosis in UNOS; 80 (64%) were listed as alcoholic cirrhosis, and 1 (1%) as fulminant hepatic necrosis. Of the 81 patients missing AH as a UNOS listing diagnosis code, the reasons for alternate classification were 44 (54%) due to a lack of awareness of a separate diagnosis code for AH; 13 (16%) due to concomitant clinical diagnosis of AH and alcoholic cirrhosis in the chart; 12 (15%) due to clinical uncertainty regarding the diagnosis of AH versus acute decompensated alcoholic cirrhosis; and 12 (15%) due to a data entry error. In conclusion, in a large cohort of LT recipients with AH, only 35% were documented as such in UNOS. Increased education and awareness for those performing UNOS data entry, the establishment of specific criteria to define AH in the UNOS database, and the ability to document dates of alcohol use would allow future research on ALD to be more informative.

A Bayesian latent class approach for EHR-based phenotyping.

Phenotyping, ie, identification of patients possessing a characteristic of interest, is a fundamental task for research conducted using electronic health records. However, challenges to this task include imperfect sensitivity and specificity of clinical codes and inconsistent availability of more detailed data such as laboratory test results. Despite these challenges, most existing electronic health records-derived phenotypes are rule-based, consisting of a series of Boolean arguments informed by expert knowledge of the disease of interest and its coding. The objective of this paper is to introduce a Bayesian latent phenotyping approach that accounts for imperfect data elements and missing not at random missingness patterns that can be used when no gold-standard data are available. We conducted simulation studies to compare alternative phenotyping methods under different patterns of missingness and applied these approaches to a cohort of 68 265 children at elevated risk for type 2 diabetes mellitus (T2DM). In simulation studies, the latent class approach had similar sensitivity to a rule-based approach (95.9% vs 91.9%) while substantially improving specificity (99.7% vs 90.8%). In the PEDSnet cohort, we found that biomarkers and clinical codes were strongly associated with latent T2DM status. The latent T2DM class was also strongly predictive of missingness in biomarkers. Glucose was missing in 83.4% of patients (odds ratio for latent T2DM status = 0.52) while hemoglobin A1c was missing in 91.2% (odds ratio for latent T2DM status = 0.03 ), suggesting missing not at random missingness. The latent phenotype approach may substantially improve on rule-based phenotyping.

A comparison of two algorithms to identify sudden cardiac deaths in computerized databases.

PURPOSE: Two previously validated algorithms to identify sudden cardiac death using administrative data showed high positive predictive value. We evaluated the agreement between the algorithms using data from a common source population. METHODS: We conducted a cross-sectional study to assess the percent agreement between deaths identified by two sudden cardiac death algorithms using Tennessee Medicaid and death certificate data from 2007 through 2014. The source population included all deceased patients aged 18 to 64 years with Medicaid enrollment in the 6 months prior to death. To identify sudden cardiac deaths, algorithm 1 used only hospital/emergency department (ED) claims from encounters at the time of death, and algorithm 2 required death certificates and used claims data for specific exclusion criteria. RESULTS: We identified 34 107 deaths in the source population over the study period. The two algorithms identified 4372 potential sudden cardiac deaths: Algorithm 1 identified 3117 (71.3%) and algorithm 2 identified 1715 (39.2%), with 460 (10.5%) deaths identified by both algorithms. Of the deaths identified by algorithm 1, 1943 (62.3%) had an underlying cause of death not specified in algorithm 2. Of the deaths identified by algorithm 2, 1053 (61.4%) had no record of a hospital or ED encounter at the time of death, and 202 (11.8%) had a discharge diagnosis code not specified in algorithm 1. CONCLUSIONS: We found low agreement between the two algorithms for identification of sudden cardiac deaths because of differences in sudden cardiac death definitions and data sources.

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.

Validity of hospital ICD-10-GM codes to identify acute liver injury in Germany.

PURPOSE: Acute liver injury (ALI) is an important adverse drug reaction. We estimated the positive predictive values (PPVs) of ICD-10-GM codes of ALI used in an international postauthorisation safety study (PASS). METHODS: Analyses used routine data (2007 to 2016, adults) from a German academic hospital in a cross-sectional design. Two algorithms from the PASS were applied to extract potential cases from the hospital information system: specific end point (A) (discharge diagnosis of liver disease-specific codes) and less specific end point (B) (discharge and outpatient-specific and nonspecific codes suggestive of liver injury). ALI cases were confirmed on the basis of plasma liver enzyme activity elevation. Secondary analysis was performed following exclusion of cases with known cancer, chronic liver, biliary and pancreatic disease, heart failure, and alcohol-related disorders, as applied in the PASS. RESULTS: On the basis of ICD codes: outcome A, 154 cases (143 with case notes and lab data for case verification); outcome B, 485 cases (357 with case notes and lab data). ALI was confirmed in 71 outcome A cases, PPV of 49.7% (95% confidence interval [CI], 41.2%-58.1%), and 100 outcome B cases, PPV of 28.0% (95% CI, 23.4%-33.0%). Applying exclusion criteria increased PPV (95% CI) to 62.7% (50.0%-74.2%) for outcome A and 45.7% (37.2%-54.3%) for outcome B. CONCLUSIONS: In safety studies on hepatotoxicity based on routine data using ICD-10-GM discharge codes and when validation of potential cases is not feasible, only the more specific codes should be used to describe ALI, and competing diagnoses for liver injury should be excluded to avoid substantial misclassification.

Validation of hip osteoarthritis diagnosis recording in the UK Clinical Practice Research Datalink.

PURPOSE: The diagnosis of hip osteoarthritis is subject to several uncertainties, especially in primary care. The aims of this study were to determine (i) the diagnostic accuracy of coding of hip osteoarthritis by primary care physicians in the UK Clinical Practice Research Datalink (CPRD), (ii) the relative influence of radiographic and clinical parameters on diagnostic accuracy, and (iii) the accuracy of the diagnosis date. METHODS: An extract of all patients aged over 65 years, with a Read code for hip osteoarthritis listed between January 1995 and December 2014, was obtained from CPRD. A random sample was selected of 170 participants. A questionnaire concerning data in medical records on relevant clinical and radiographic criteria used to establish the diagnosis of hip osteoarthritis was distributed to primary care physicians of participants. Using diagnostic criteria, we formulated thresholds for diagnosis based on clinical, radiographic, and combined grounds. RESULTS: One hundred nineteen completed questionnaires were returned (70% response rate). The positive predictive value (PPV) of hip osteoarthritis codes, based on radiological criteria, was 79.8%. The PPV, based on clinical criteria, was 79.0%, with substantial but not complete overlap. Overall 12% of diagnoses were not confirmed. In 42% of cases, there was disparity between date of diagnosis in CPRD and the medical record. Median difference in date was ±425 days (interquartile range, 18-1448 days). CONCLUSIONS: Despite the difficulties in reaching a diagnosis of hip osteoarthritis in primary care, CPRD Read codes have a sufficiently high PPV for most research uses. However, the accuracy of diagnosis date may not be as reliable.

Development of an algorithm to detect methotrexate wrong frequency error using computerized health care data.

PURPOSE: We validated an algorithm to detect frequency errors in computerized healthcare data and estimated the incidence of these errors in an integrated healthcare system. METHODS: We applied Sentinel System analytic tools on the electronic health records of Kaiser Permanente, Northern California, January 1, 2010, through May 30, 2015,to identify rheumatoid arthritis (RA) patients with new use of methotrexate (365-day baseline period). We identified potential methotrexate frequency errors using ICD-9 code 995.20 (adverse drug event), Current Procedural Terminology (CPT) code 96409 for injection of leucovorin and prescription refill patterns. We performed chart review to confirm the frequency errors, assessed performance for detecting frequency errors, and estimated the incidence of chart-confirmed errors. RESULTS: The study included 24,529 methotrexate dispensings among 3,668 RA patients. Among these, 722 (3%) had one dispensing and 23,807 (97.1%) had ≥2 dispensings during 1-year follow-up period. We flagged 653 (2.7%) with a potential medication error (46 with one dispensing and 607 with ≥2 dispensings). We sampled 94 for chart review, and confirmed three methotrexate errors. All three confirmed frequency errors involved a first methotrexate dispensing followed by injected rescue therapy, leucovorin, (positive predictive value, 60%; 95% confidence interval [CI], 15-95%). No potential errors were found among patients with ≥2 dispensings. We estimated the frequency error incidence among one methotrexate dispensing to be 0.4% (95%CI, 0.1% to 1.2%). CONCLUSION: Rescue therapy is a specific indicator of methotrexate overdose among first methotrexate dispensings. This method is generalizable to other medications with serious adverse events treated with antidotes.

Algorithms to identify COPD in health systems with and without access to ICD coding: a systematic review.

BACKGROUND: Chronic obstructive pulmonary disease (COPD) causes significant morbidity and mortality worldwide. Estimation of incidence, prevalence and disease burden through routine insurance data is challenging because of under-diagnosis and under-treatment, particularly for early stage disease in health care systems where outpatient International Classification of Diseases (ICD) diagnoses are not collected. This poses the question of which criteria are commonly applied to identify COPD patients in claims datasets in the absence of ICD diagnoses, and which information can be used as a substitute. The aim of this systematic review is to summarize previously reported methodological approaches for the identification of COPD patients through routine data and to compile potential criteria for the identification of COPD patients if ICD codes are not available. METHODS: A systematic literature review was performed in Medline via PubMed and Google Scholar from January 2000 through October 2018, followed by a manual review of the included studies by at least two independent raters. Study characteristics and all identifying criteria used in the studies were systematically extracted from the publications, categorized, and compiled in evidence tables. RESULTS: In total, the systematic search yielded 151 publications. After title and abstract screening, 38 publications were included into the systematic assessment. In these studies, the most frequently used (22/38) criteria set to identify COPD patients included ICD codes, hospitalization, and ambulatory visits. Only four out of 38 studies used methods other than ICD coding. In a significant proportion of studies, the age range of the target population (33/38) and hospitalization (30/38) were provided. Ambulatory data were included in 24, physician claims in 22, and pharmaceutical data in 18 studies. Only five studies used spirometry, two used surgery and one used oxygen therapy. CONCLUSIONS: A variety of different criteria is used for the identification of COPD from routine data. The most promising criteria set in data environments where ambulatory diagnosis codes are lacking is the consideration of additional illness-related information with special attention to pharmacotherapy data. Further health services research should focus on the application of more systematic internal and/or external validation approaches.

Descriptive evaluation of methods for identifying work-related emergency department injury visits.

BACKGROUND: Use of worker's compensation (WC) as payer underestimates work-related (WR) injuries. We evaluated three methods to identify WR injuries: WC as payer, ICD-9-CM work-status codes E000.0/E000.1, and other ICD-9-CM external cause codes. METHODS: We identified injury-related emergency department visits from North Carolina's syndromic surveillance system (2010-2013). Characteristics were compared by indicator. We manually reviewed 800 admission notes to confirm if the visit was WR or non-WR; WR keywords from the review were applied to all visits. RESULTS: 133 156 injury-related visits (age, 16 years or older) were identified: WC = 69%, work-status codes = 18%, other ICD-9-CM codes = 13%. Among manually reviewed visits: few visits identified by WC (0.3%) or work-status codes (2%) were non-WR, while 12% of other ICD-9-CM code identified visits were non-WR; 53%, 46%, and 31% of visits identified by WC, work-status codes, and other ICD-9-CM codes were WR, respectively. CONCLUSIONS: Findings support use of WC and work-status codes to capture WR injuries; other ICD-9-CM codes should be used with caution or in combination with other indicators.

Assessing the impact of the new ICD-10-CM coding system on pharmacoepidemiologic studies-An application to the known association between angiotensin-converting enzyme inhibitors and angioedema.

PURPOSE: To replicate the well-established association between angiotensin-converting enzyme inhibitors versus beta blockers and angioedema in the International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) era. METHODS: We conducted a retrospective, inception cohort study in a large insurance database formatted to the Sentinel Common Data Model. We defined study periods spanning the ICD-9-CM era only, ICD-10-CM era only, and ICD-9-CM and ICD-10-CM era and conducted simple-forward mapping (SFM), simple-backward mapping (SBM), and forward-backward mapping (FBM) referencing the General Equivalence Mappings to translate the outcome (angioedema) and covariates from ICD-9-CM to ICD-10-CM. We performed propensity score (PS)-matched and PS-stratified Cox proportional hazards regression to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). RESULTS: In the ICD-9-CM and ICD-10-CM eras spanning April 1 to September 30 of 2015 and 2016, there were 152 017 and 145 232 angiotensin-converting enzyme inhibitor initiators and 115 073 and 116 652 beta-blocker initiators, respectively. The PS-matched HR was 4.19 (95% CI, 2.82-6.23) in the ICD-9-CM era, 4.37 (2.92-6.52) in the ICD-10-CM era using SFM, and 4.64 (3.05-7.07) in the ICD-10-CM era using SBM and FBM. The PS-matched HRs from the mixed ICD-9-CM and ICD-10-CM eras ranged from 3.91 (2.69-5.68) to 4.35 (3.33-5.70). CONCLUSION: The adjusted HRs across different diagnostic coding eras and the use of SFM versus SBM and FBM produced numerically different but clinically similar results. Additional investigations as ICD-10-CM data accumulate are warranted.

Optimizing an algorithm for the identification and classification of pregnancy outcomes in German claims data.

PURPOSE: For studying drug utilization and safety in pregnancy based on administrative health care data, the reliable identification and classification of pregnancy outcomes in the data is essential. We aimed to optimize an existing algorithm for the identification and classification of pregnancy outcomes in the German Pharmacoepidemiological Research Database (GePaRD) with a particular focus on births. METHODS: We reconsidered all codes used by the original algorithm and applied it to data of GePaRD from 2006 to 2014. Longitudinal records of pregnancies were used to identify targets for enhancing the algorithm's specificity. We checked the plausibility of the results, eg, regarding the age distribution of persons with pregnancy outcomes. Based on 20 longitudinal records of pregnancies, we compared the outcome classification by clinical experts with the results of the modified algorithm. RESULTS: Our algorithm identified 1 235 261 pregnancy outcomes in the database, with the majority (94%) being live births, classified as preterm (10%), term (78%), and (12%) births after the expected delivery date. The median age of pregnant women was 32 years (Q1 28; Q3 35). Implausible sequence of outcomes (for example, an induced abortion within a pregnancy categorized as ending in a live birth) were rare (0.03%). The case profile review by clinical experts resulted in the same outcome type and date as the algorithm in 95%. CONCLUSIONS: Our algorithm led to plausible results regarding the identification and classification of pregnancy outcomes. It will be an important foundation for studies on drug utilization and drug safety during pregnancy based on GePaRD.