Use and Utility of Administrative Health Data for Stroke Research and Surveillance

Association Between Time to Treatment With Endovascular Thrombectomy and Home-Time After Acute Ischemic Stroke

BACKGROUND AND OBJECTIVES

Home-time is a patient-prioritized stroke outcome that can be derived from administrative data linkages. The effect of faster time-to-treatment with endovascular thrombectomy (EVT) on home-time after acute stroke is unknown.

METHODS

We used the Quality Improvement and Clinical Research registry to identify a cohort of patients who received EVT for acute ischemic stroke between 2015 and 2022 in Alberta, Canada. We calculated days at home in the first 90 days after stroke. We used ordinal regression across 6 ordered categories of home-time to evaluate the association between onset-to-arterial puncture and higher home-time, adjusting for age, sex, rural residence, NIH Stroke Scale, comorbidities, intravenous thrombolysis, and year of treatment. We used restricted cubic splines to assess the nonlinear relationship between continuous variation in time metrics and higher home-time, and also reported the adjusted odds ratios within time categories. We additionally evaluated door-to-puncture and reperfusion times. Finally, we analyzed home-time with zero-inflated models to determine the minutes of earlier treatment required to gain 1 day of home-time.

RESULTS

We had 1,885 individuals in our final analytic sample. There was a nonlinear increase in home-time with faster treatment when EVT was within 4 hours of stroke onset or 2 hours of hospital arrival. There was a higher odds of achieving more days at home when onset-to-puncture time was <2 hours (adjusted odds ratio 2.36, 95% CI 1.77-3.16) and 2 to <4 hours (1.37, 95% CI 1.11-1.71) compared with ≥6 hours, and when door-to-puncture time was <1 hour (aOR 2.25, 95% CI 1.74-2.90), 1 to <1.5 hours (aOR 1.89, 95% CI 1.47-2.41), and 1.5 to <2 hours (1.35, 95% CI 1.04-1.76) compared with ≥2 hours. Results were consistent for reperfusion times. For every hour of faster treatment within 6 hours of stroke onset, there was an estimated increase in home-time of 4.7 days, meaning that approximately 1 day of home-time was gained for each 12.8 minutes of faster treatment.

DISCUSSION

Faster time-to-treatment with EVT for acute stroke was associated with greater home-time, particularly within 4 hours of onset-to-puncture and 2 hours of door-to-puncture time. Within 6 hours of stroke onset, each 13 minutes of faster treatment is associated with a gain of 1 day of home-time.

Validation of the Passive Surveillance Stroke Severity Score in Three Canadian Provinces

BACKGROUND

Stroke outcomes research requires risk-adjustment for stroke severity, but this measure is often unavailable. The Passive Surveillance Stroke SeVerity (PaSSV) score is an administrative data-based stroke severity measure that was developed in Ontario, Canada. We assessed the geographical and temporal external validity of PaSSV in British Columbia (BC), Nova Scotia (NS) and Ontario, Canada.

METHODS

We used linked administrative data in each province to identify adult patients with ischemic stroke or intracerebral hemorrhage between 2014-2019 and calculated their PaSSV score. We used Cox proportional hazards models to evaluate the association between the PaSSV score and the hazard of death over 30 days and the cause-specific hazard of admission to long-term care over 365 days. We assessed the models' discriminative values using Uno's c-statistic, comparing models with versus without PaSSV.

RESULTS

We included 86,142 patients (n = 18,387 in BC, n = 65,082 in Ontario, n = 2,673 in NS). The mean and median PaSSV were similar across provinces. A higher PaSSV score, representing lower stroke severity, was associated with a lower hazard of death (hazard ratio and 95% confidence intervals 0.70 [0.68, 0.71] in BC, 0.69 [0.68, 0.69] in Ontario, 0.72 [0.68, 0.75] in NS) and admission to long-term care (0.77 [0.76, 0.79] in BC, 0.84 [0.83, 0.85] in Ontario, 0.86 [0.79, 0.93] in NS). Including PaSSV in the multivariable models increased the c-statistics compared to models without this variable.

CONCLUSION

PaSSV has geographical and temporal validity, making it useful for risk-adjustment in stroke outcomes research, including in multi-jurisdiction analyses.

Synthetic data in cancer and cerebrovascular disease research: A novel approach to big data

OBJECTIVES

Synthetic datasets are artificially manufactured based on real health systems data but do not contain real patient information. We sought to validate the use of synthetic data in stroke and cancer research by conducting a comparison study of cancer patients with ischemic stroke to non-cancer patients with ischemic stroke.

DESIGN

retrospective cohort study.

SETTING

We used synthetic data generated by MDClone and compared it to its original source data (i.e. real patient data from the Ottawa Hospital Data Warehouse).

OUTCOME MEASURES

We compared key differences in demographics, treatment characteristics, length of stay, and costs between cancer patients with ischemic stroke and non-cancer patients with ischemic stroke. We used a binary, multivariable logistic regression model to identify risk factors for recurrent stroke in the cancer population.

RESULTS

Using synthetic data, we found cancer patients with ischemic stroke had a lower prevalence of hypertension (52.0% in the cancer cohort vs 57.7% in the non-cancer cohort, p<0.0001), and a higher prevalence of chronic obstructive pulmonary disease (COPD: 8.5% vs 4.7%, p<0.0001), prior ischemic stroke (1.7% vs 0.1%, p<0.001), and prior venous thromboembolism (VTE: 8.2% vs 1.5%, p<0.0001). They also had a longer length of stay (8 days [IQR 3-16] vs 6 days [IQR 3-13], p = 0.011), and higher costs associated with their stroke encounters: $11,498 (IQR $4,440 -$20,668) in the cancer cohort vs $8,084 (IQR $3,947 -$16,706) in the non-cancer cohort (p = 0.0061). A multivariable logistic regression model identified 5 predictors for recurrent ischemic stroke in the cancer cohort using synthetic data; 3 of the same predictors identified using real patient data with similar effect measures. Summary statistics between synthetic and original datasets did not significantly differ, other than slight differences in the distributions of frequencies for numeric data.

CONCLUSION

We demonstrated the utility of synthetic data in stroke and cancer research and provided key differences between cancer and non-cancer patients with ischemic stroke. Synthetic data is a powerful tool that can allow researchers to easily explore hypothesis generation, enable data sharing without privacy breaches, and ensure broad access to big data in a rapid, safe, and reliable fashion.

Use of machine learning techniques for identifying ischemic stroke instead of the rule-based methods: a nationwide population-based study

BACKGROUND

Many studies have evaluated stroke using claims data; most of these studies have defined ischemic stroke using an operational definition following the rule-based method. Rule-based methods tend to overestimate the number of patients with ischemic stroke.

OBJECTIVES

We aimed to identify an appropriate algorithm for identifying stroke by applying machine learning (ML) techniques to analyze the claims data.

METHODS

We obtained the data from the Korean National Health Insurance Service database, which is linked to the Ilsan Hospital database (n = 30,897). The performance of prediction models (extreme gradient boosting [XGBoost] or gated recurrent unit [GRU]) was evaluated using the area under the receiver operating characteristic curve (AUROC), the area under precision-recall curve (AUPRC), and calibration curve.

RESULTS

In total, 30,897 patients were enrolled in this study, 3145 of whom (10.18%) had ischemic stroke. XGBoost, a tree-based ML technique, had the AUROC was 94.46% and AUPRC was 92.80%. GRU showed the highest accuracy (99.81%), precision (99.92%) and recall (99.69%).

CONCLUSIONS

We proposed recurrent neural network-based deep learning techniques to improve stroke phenotyping. This can be expected to produce rapid and more accurate results than the rule-based methods.

Change in Hospital Risk-standardized Stroke Mortality Performance With and Without the Passive Surveillance Stroke Severity Score

BACKGROUND

Adjustment for baseline stroke severity is necessary for accurate assessment of hospital performance. We evaluated whether adjusting for the Passive Surveillance Stroke SeVerity (PaSSV) score, a measure of stroke severity derived using administrative data, changed hospital-specific estimated 30-day risk-standardized mortality rate (RSMR) after stroke.

METHODS

We used linked administrative data to identify adults who were hospitalized with ischemic stroke or intracerebral hemorrhage across 157 hospitals in Ontario, Canada between 2014 and 2019. We fitted a random effects logistic regression model using Markov Chain Monte Carlo methods to estimate hospital-specific 30-day RSMR and 95% credible intervals with adjustment for age, sex, Charlson comorbidity index, and stroke type. In a separate model, we additionally adjusted for stroke severity using PaSSV. Hospitals were defined as low-performing, average-performing, or high-performing depending on whether the RSMR and 95% credible interval were above, overlapping, or below the cohort's crude mortality rate.

RESULTS

We identified 65,082 patients [48.0% were female, the median age (25th,75th percentiles) was 76 years (65,84), and 86.4% had an ischemic stroke]. The crude 30-day all-cause mortality rate was 14.1%. The inclusion of PaSSV in the model reclassified 18.5% (n=29) of the hospitals. Of the 143 hospitals initially classified as average-performing, after adjustment for PaSSV, 20 were reclassified as high-performing and 8 were reclassified as low-performing. Of the 4 hospitals initially classified as low-performing, 1 was reclassified as high-performing. All 10 hospitals initially classified as high-performing remained unchanged.

CONCLUSION

PaSSV may be useful for risk-adjusting mortality when comparing hospital performance. External validation of our findings in other jurisdictions is needed.

Cerebrovascular disease case identification in inpatient electronic medical record data using natural language processing

BACKGROUND

Abstracting cerebrovascular disease (CeVD) from inpatient electronic medical records (EMRs) through natural language processing (NLP) is pivotal for automated disease surveillance and improving patient outcomes. Existing methods rely on coders' abstraction, which has time delays and under-coding issues. This study sought to develop an NLP-based method to detect CeVD using EMR clinical notes.

METHODS

CeVD status was confirmed through a chart review on randomly selected hospitalized patients who were 18 years or older and discharged from 3 hospitals in Calgary, Alberta, Canada, between January 1 and June 30, 2015. These patients' chart data were linked to administrative discharge abstract database (DAD) and Sunrise™ Clinical Manager (SCM) EMR database records by Personal Health Number (a unique lifetime identifier) and admission date. We trained multiple natural language processing (NLP) predictive models by combining two clinical concept extraction methods and two supervised machine learning (ML) methods: random forest and XGBoost. Using chart review as the reference standard, we compared the model performances with those of the commonly applied International Classification of Diseases (ICD-10-CA) codes, on the metrics of sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).

RESULT

Of the study sample (n = 3036), the prevalence of CeVD was 11.8% (n = 360); the median patient age was 63; and females accounted for 50.3% (n = 1528) based on chart data. Among 49 extracted clinical documents from the EMR, four document types were identified as the most influential text sources for identifying CeVD disease ("nursing transfer report," "discharge summary," "nursing notes," and "inpatient consultation."). The best performing NLP model was XGBoost, combining the Unified Medical Language System concepts extracted by cTAKES (e.g., top-ranked concepts, "Cerebrovascular accident" and "Transient ischemic attack"), and the term frequency-inverse document frequency vectorizer. Compared with ICD codes, the model achieved higher validity overall, such as sensitivity (25.0% vs 70.0%), specificity (99.3% vs 99.1%), PPV (82.6 vs. 87.8%), and NPV (90.8% vs 97.1%).

CONCLUSION

The NLP algorithm developed in this study performed better than the ICD code algorithm in detecting CeVD. The NLP models could result in an automated EMR tool for identifying CeVD cases and be applied for future studies such as surveillance, and longitudinal studies.

Reconstruction of pseudonymized patient-trajectories in Austria's stroke cohort using medical record-linkage of in-patient routine documentation to establish a nation-wide acute stroke cohort of 102,107 pseudonymized patients between 2015 and 2019

INTRODUCTION

Administrative health data are increasingly used for disease surveillance, quality assurance and research purposes. In Austria, reporting of a standardized dataset is mandatory for each patient.

PATIENTS AND METHODS

Routine documentation includes administrative and medical data, including admission and discharge characteristics, disease-diagnosis using ICD-10, medical procedure codes, and coding of involved hospital departments. Since 2015, a three-step pseudonymization on these data is provided including a pseudonym using secure hash algorithm 256, a non-recalculable record-ID, and age-groups of 5 years, allowing the reconstruction of individual patient-trajectories. We included persons aged ⩾20 years with an in-patient treatment in Austrian hospitals for acute stroke or transient ischemic attack (TIA) between 01.01.2015 and 31.12.2019 using medical record-linkage.

RESULTS

This totals 102,107 patients (49.3% women) with 107,055 treatment episodes. An ischemic stroke (IS) occurred in 60.9% (n = 65,133), 27.1% (n = 29,019) had a TIA, 3.3% (n = 3488) a subarachnoid hemorrhage, and 8.8% (n = 9415) an intracerebral hemorrhage (ICH). The study period covers 35.2 million person-years at risk, with a hospitalization rate for acute stroke of 221.8 per 100,000 person-years (95% CI 220.2-223.3), and 185.1 per 100,000 person-years (95% CI 183.7-186.5) for IS. Unscheduled re-admissions within 1 year occurred in 29.2% (95% CI 28.8-29.7) after IS, and 41.7% (95% CI 40.0-43.3) after ICH. Recurrent stroke occurred in 5.3% (95% CI 5.1-5.5) after IS, and 5.6% (95% CI 4.9-6.4) after ICH.

DISCUSSION

We present herein the details of a novel methodology to establish a nation-wide unselected Austrian stroke cohort, and to reconstruct pseudonymized individual longitudinal patient-trajectories on a national level. This approach shows potential applications in epidemiological research, quality assessment and outcome measurement.

CONCLUSION

This novel approach opens new research fields, facilitates international comparison, and is needed for national benchmarking to assess the achievement of goals according to the Stroke Action Plan for Europe and augment the quality of Austria's integrated stroke care.

Rule-based natural language processing for automation of stroke data extraction: a validation study

PURPOSE

Data extraction from radiology free-text reports is time consuming when performed manually. Recently, more automated extraction methods using natural language processing (NLP) are proposed. A previously developed rule-based NLP algorithm showed promise in its ability to extract stroke-related data from radiology reports. We aimed to externally validate the accuracy of CHARTextract, a rule-based NLP algorithm, to extract stroke-related data from free-text radiology reports.

METHODS

Free-text reports of CT angiography (CTA) and perfusion (CTP) studies of consecutive patients with acute ischemic stroke admitted to a regional stroke center for endovascular thrombectomy were analyzed from January 2015 to 2021. Stroke-related variables were manually extracted as reference standard from clinical reports, including proximal and distal anterior circulation occlusion, posterior circulation occlusion, presence of ischemia or hemorrhage, Alberta stroke program early CT score (ASPECTS), and collateral status. These variables were simultaneously extracted using a rule-based NLP algorithm. The NLP algorithm's accuracy, specificity, sensitivity, positive predictive value (PPV), and negative predictive value (NPV) were assessed.

RESULTS

The NLP algorithm's accuracy was > 90% for identifying distal anterior occlusion, posterior circulation occlusion, hemorrhage, and ASPECTS. Accuracy was 85%, 74%, and 79% for proximal anterior circulation occlusion, presence of ischemia, and collateral status respectively. The algorithm confirmed the absence of variables from radiology reports with an 87-100% accuracy.

CONCLUSIONS

Rule-based NLP has a moderate to good performance for stroke-related data extraction from free-text imaging reports. The algorithm's accuracy was affected by inconsistent report styles and lexicon among reporting radiologists.

External validation of the Passive Surveillance Stroke Severity Indicator

BACKGROUND

The Passive Surveillance Stroke Severity (PaSSV) Indicator was derived to estimate stroke severity from variables in administrative datasets but has not been externally validated.

METHODS

We used linked administrative datasets to identify patients with first hospitalization for acute stroke between 2007-2018 in Alberta, Canada. We used the PaSSV indicator to estimate stroke severity. We used Cox proportional hazard models and evaluated the change in hazard ratios and model discrimination for 30-day and 1-year case fatality with and without PaSSV. Similar comparisons were made for 90-day home time thresholds using logistic regression. We also linked with a clinical registry to obtain National Institutes of Health Stroke Scale (NIHSS) and compared estimates from models without stroke severity, with PaSSV, and with NIHSS.

RESULTS

There were 28,672 patients with acute stroke in the full sample. In comparison to no stroke severity, addition of PaSSV to the 30-day case fatality models resulted in improvement in model discrimination (C-statistic 0.72 [95%CI 0.71-0.73] to 0.80 [0.79-0.80]). After adjustment for PaSSV, admission to a comprehensive stroke center was associated with lower 30-day case fatality (adjusted hazard ratio changed from 1.03 [0.96-1.10] to 0.72 [0.67-0.77]). In the registry sample (N = 1328), model discrimination for 30-day case fatality improved with the inclusion of stroke severity. Results were similar for 1-year case fatality and home time outcomes.

CONCLUSION

Addition of PaSSV improved model discrimination for case fatality and home time outcomes. The validity of PASSV in two Canadian provinces suggests that it is a useful tool for baseline risk adjustment in acute stroke.

Validation of Stroke Risk Factors in Patients with Acute Ischemic Stroke, Transient Ischemic Attack, or Intracerebral Hemorrhage on Taiwan’s National Health Insurance Claims Data

Purpose

Taiwan has changed the coding system to the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) coding since 2016. This study aimed to determine the optimal algorithms for identifying stroke risk factors in Taiwan’s National Health Insurance (NHI) claims data.

Patients and Methods

We retrospectively enrolled 4538 patients hospitalized for acute ischemic stroke (AIS), transient ischemic attack (TIA), or intracerebral hemorrhage (ICH) from two hospitals’ stroke registries, which were linked to NHI claims data. We developed several algorithms based on ICD-10-CM diagnosis codes and prescription claims data to identify hypertension, diabetes, hyperlipidemia, atrial fibrillation (AF), and ischemic heart disease (IHD) using registry data as the reference standard. The agreement of risk factor status between claims and registry data was quantified by calculating the kappa statistic.

Results

According to the registry data, the prevalence of hypertension, diabetes, hyperlipidemia, AF, and IHD among all patients was 77.5%, 41.5%, 47.9%, 12.1%, and 7.1%, respectively. In general, including diagnosis codes from prior inpatient or outpatient claims to those from the stroke hospitalization claims improved the agreement. Incorporating prescription data could improve the agreement for hypertension, diabetes, hyperlipidemia, and AF, but not for IHD. The kappa values of the optimal algorithms were 0.552 (95% confidence interval 0.524–0.580) for hypertension, 0.802 (0.784–0.820) for diabetes, 0.514 (0.490–0.539) for hyperlipidemia, 0.765 (0.734–0.795) for AF, and 0.518 (0.473–0.564) for IHD.

Conclusion

Algorithms using diagnosis codes alone are sufficient to identify hypertension, AF, and IHD whereas algorithms combining both diagnosis codes and prescription data are more suitable for identifying diabetes and hyperlipidemia. The study results may provide a reference for future studies using Taiwan’s NHI claims data.

Chronic disease surveillance in Alberta's tomorrow project using administrative health data

Introduction

Alberta’s Tomorrow Project (ATP) is the largest population-based prospective cohort study of cancer and chronic diseases in Alberta, Canada. The ATP cohort data were primarily self-reported by participants on lifestyle behaviors and disease risk factors at the enrollment, which lacks sufficient and accurate data on chronic disease diagnosis for longer-term follow-up.

Objectives

To characterize the occurrence rate and trend of chronic diseases in the ATP cohort by linking with administrative healthcare data.

Methods

A set of validated algorithms using ICD codes were applied to Alberta Health (AH) administrative data (October 2000-March 2018) linked to the ATP cohort to determine the prevalence and incidence of common chronic diseases.

Results

There were 52,770 ATP participants (51.2±9.4 years old at enrollment and 63.7% females) linked to the AH data with average follow-up of 10.1±4.4 years. In the ATP cohort, hypertension (18.5%), depression (18.1%), chronic pain (12.8%), osteoarthritis (10.1%) and cardiovascular diseases (8.7%) were the most prevalent chronic conditions. The incidence rates varied across diseases, with the highest rates for hypertension (22.1 per 1000 person-year), osteoarthritis (16.2 per 1000 person-year) and ischemic heart diseases (13.0 per 1000 person-year). All chronic conditions had increased prevalence over time (p < for trend tests), while incidence rates were relatively stable. The proportion of participants with two or more of these conditions (multi-morbidity) increased from 3.9% in 2001 to 40.3% in 2017.

Conclusions

This study shows an increasing trend of chronic diseases in the ATP cohort, particularly related to cardiovascular diseases and multi-morbidity. Using administrative health data to monitor chronic diseases for large population-based prospective cohort studies is feasible in Alberta, and our approach could be further applied in a broader research area, including health services research, to enhance research capacity of these population-based studies in Canada.

Incidence and Outcomes of Hemorrhagic Stroke among Adults in Spain (2016-2018) According to Sex: A Retrospective, Cohort, Observational, Propensity Score Matched Study

(1) Background: We aim to analyze sex differences in the incidence, clinical characteristics and in-hospital outcomes of hemorrhagic stroke (HS) in Spain (2016–2018) using the National Hospital Discharge Database. (2) Methods: Retrospective, cohort, observational study. We estimated the incidence of HS in men and women. We analyzed comorbidity, treatments, procedures, and hospital outcomes. We matched each woman with a man by age, type of HS and medical conditions using propensity score matching. (3) Results: HS was coded in 57,227 patients aged ≥18 years (44.3% women). Overall, men showed higher incidence rates (57.3/10⁵ vs. 43.0/10⁵; p < 0.001; IRR = 1.60; 95% CI: 1.38–1.83). Women suffered more subarachnoid hemorrhages (25.2% vs. 14.6%), whereas men more often had intracerebral hemorrhages (55.7% vs. 54.1%). In-hospital mortality (IHM) was higher for intracerebral hemorrhage in both men and women. Women underwent decompressive craniectomy less often than men (5.0% vs. 6.2%; p < 0.001). After matching, IHM among women was higher (29.0% vs. 23.7%; p < 0.001). Increments in age, comorbidity and use of anticoagulants and antiplatelet agents prior to hospitalization were associated were higher IHM, and decompressive craniectomy was associated with lower IHM in both sexes. After multivariable adjustment, women had higher IHM (OR = 1.23; 95% CI: 1.18–1.28). (4) Conclusion: Men had higher incidence rates of HS than women. Women less often underwent decompressive craniectomy. IHM was higher among women admitted for HS than among men.

Sex-related disparities in the incidence and outcomes of hemorrhagic stroke among type 2 diabetes patients: a propensity score matching analysis using the Spanish National Hospital Discharge Database for the period 2016-18

Background

To analyze incidence, use of therapeutic procedures, use of oral anticoagulants (OACs) and antiplatelet agents prior to hospitalization, and in-hospital outcomes among patients who were hospitalized with hemorrhagic stroke (HS) according to the presence of type 2 diabetes mellitus (T2DM) in Spain (2016–2018) and to assess the role of sex differences among those with T2DM.

Methods

Using the Spanish National Hospital Discharge Database we estimated the incidence of HS hospitalizations in men and women aged ≥ 35 years with and without T2DM. Propensity score matching (PSM) was used to compare population subgroups according to sex and the presence of T2DM.

Results

HS was coded in 31,425 men and 24,975 women, of whom 11,915 (21.12%) had T2DM. The adjusted incidence of HS was significantly higher in patients with T2DM (both sexes) than in non-T2DM individuals (IRR 1.15; 95% CI 1.12–1.17). The incidence of HS was higher in men with T2DM than in T2DM women (adjusted IRR 1.60; 95% CI 1.57–1.63). After PSM, men and women with T2DM have significantly less frequently received decompressive craniectomy than those without T2DM. In-hospital mortality (IHM) was higher among T2DM women than matched non-T2DM women (32.89% vs 30.83%; p = 0.037), with no differences among men. Decompressive craniectomy was significantly more common in men than in matched women with T2DM (5.81% vs. 3.33%; p < 0.001). IHM was higher among T2DM women than T2DM men (32.89% vs. 28.28%; p < 0.001). After adjusting for confounders with multivariable logistic regression, women with T2DM had a 18% higher mortality risk than T2DM men (OR 1.18; 95% CI 1.07–1.29). Use of OACs and antiplatelet agents prior to hospitalization were associated to higher IHM in men and women with and without T2DM.

Conclusions

T2DM is associated with a higher incidence of HS and with less frequent use of decompressive craniectomy in both sexes, but with higher IHM only among women. Sex differences were detected in T2DM patients who had experienced HS, with higher incidence rates, more frequent decompressive craniectomy, and lower IHM in men than in women.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-021-01334-2.

Automating Stroke Data Extraction From Free-Text Radiology Reports Using Natural Language Processing: Instrument Validation Study

BACKGROUND

Diagnostic neurovascular imaging data are important in stroke research, but obtaining these data typically requires laborious manual chart reviews.

OBJECTIVE

We aimed to determine the accuracy of a natural language processing (NLP) approach to extract information on the presence and location of vascular occlusions as well as other stroke-related attributes based on free-text reports.

METHODS

From the full reports of 1320 consecutive computed tomography (CT), CT angiography, and CT perfusion scans of the head and neck performed at a tertiary stroke center between October 2017 and January 2019, we manually extracted data on the presence of proximal large vessel occlusion (primary outcome), as well as distal vessel occlusion, ischemia, hemorrhage, Alberta stroke program early CT score (ASPECTS), and collateral status (secondary outcomes). Reports were randomly split into training (n=921) and validation (n=399) sets, and attributes were extracted using rule-based NLP. We reported the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the overall accuracy of the NLP approach relative to the manually extracted data.

RESULTS

The overall prevalence of large vessel occlusion was 12.2%. In the training sample, the NLP approach identified this attribute with an overall accuracy of 97.3% (95.5% sensitivity, 98.1% specificity, 84.1% PPV, and 99.4% NPV). In the validation set, the overall accuracy was 95.2% (90.0% sensitivity, 97.4% specificity, 76.3% PPV, and 98.5% NPV). The accuracy of identifying distal or basilar occlusion as well as hemorrhage was also high, but there were limitations in identifying cerebral ischemia, ASPECTS, and collateral status.

CONCLUSIONS

NLP may improve the efficiency of large-scale imaging data collection for stroke surveillance and research.

Characteristics and Outcomes Among US Patients Hospitalized for Ischemic Stroke Before vs During the COVID-19 Pandemic

IMPORTANCE

After the emergence of COVID-19, studies reported a decrease in hospitalizations of patients with ischemic stroke (IS), but there are little to no data regarding hospitalizations for the remainder of 2020, including outcome data from a large cohort of patients with IS and comorbid COVID-19.

OBJECTIVE

To assess hospital discharge rates, demographic factors, and outcomes of hospitalization associated with the COVID-19 pandemic among US patients with IS before vs during the COVID-19 pandemic.

DESIGN, SETTING, AND PARTICIPANTS

This retrospective cohort study used data from the Vizient Clinical Data Base on 324 013 patients with IS at 478 nonfederal hospitals in 43 US states between January 1, 2019, and December 31, 2020. Patients were eligible if they were admitted to the hospital on a nonelective basis and were not receiving hospice care at the time of admission. A total of 41 166 discharged between January and March 2020 were excluded from the analysis because they had unreliable data on COVID-19 status, leaving 282 847 patients for the study.

EXPOSURE

Ischemic stroke and laboratory-confirmed COVID-19.

MAIN OUTCOMES AND MEASURES

Monthly counts of discharges among patients with IS in 2020. Demographic characteristics and outcomes, including in-hospital death, among patients with IS who were discharged in 2019 (control group) were compared with those of patients with IS with or without comorbid COVID-19 (COVID-19 and non-COVID-19 groups, respectively) who were discharged between April and December 2020.

RESULTS

Of the 282 847 patients included in the study, 165 912 (50.7% male; 63.4% White; 26.3% aged ≥80 years) were allocated to the control group; 111 418 of 116 935 patients (95.3%; 51.9% male; 62.8% White; 24.6% aged ≥80 years) were allocated to the non-COVID-19 group and 5517 of 116 935 patients (4.7%; 58.0% male; 42.5% White; 21.3% aged ≥80 years) to the COVID-19 group. A mean (SD) of 13 846 (553) discharges per month among patients with IS was reported in 2019. Discharges began decreasing in February 2020, reaching a low of 10 846 patients in April 2020 before returning to a prepandemic level of 13 639 patients by July 2020. A mean (SD) of 13 492 (554) discharges per month was recorded for the remainder of 2020. Black and Hispanic patients accounted for 21.4% and 7.0% of IS discharges in 2019, respectively, but accounted for 27.5% and 16.0% of those discharged with IS and comorbid COVID-19 in 2020. Compared with patients in the control and non-COVID-19 groups, those in the COVID-19 group were less likely to smoke (16.0% vs 17.2% vs 6.4%, respectively) and to have hypertension (73.0% vs 73.1% vs 68.2%) or dyslipidemia (61.2% vs 63.2% vs 56.6%) but were more likely to have diabetes (39.8% vs 40.5% vs 53.0%), obesity (16.2% vs 18.4% vs 24.5%), acute coronary syndrome (8.0% vs 9.2% vs 15.8%), or pulmonary embolus (1.9% vs 2.4% vs 6.8%) and to require intubation (11.3% vs 12.3% vs 37.6%). After adjusting for baseline factors, patients with IS and COVID-19 were more likely to die in the hospital than were patients with IS in 2019 (adjusted odds ratio, 5.17; 95% CI, 4.83-5.53; National Institutes of Health Stroke Scale adjusted odds ratio, 3.57; 95% CI, 3.15-4.05).

CONCLUSIONS AND RELEVANCE

In this cohort study, after the emergence of COVID-19, hospital discharges of patients with IS decreased in the US but returned to prepandemic levels by July 2020. Among patients with IS between April and December 2020, comorbid COVID-19 was relatively common, particularly among Black and Hispanic populations, and morbidity was high.

How adequate is measles surveillance in the United States? Investigations of measles-like illness, 2010-2017

Given the availability of an effective and safe vaccine, the World Health Organization (WHO) declared that global measles eradication is achievable, and measles elimination goals have since been established as interim steps toward eradication. As part of a strategy to maintain elimination, the Pan American Health Organization (PAHO) and WHO stipulate a minimum annual reporting rate of discarded non-measles cases of ≥2 per 100,000 population, in order to ensure sensitive surveillance and adequate investigative effort. With its effective vaccination program, the United States in 2000 was among the first countries to verify elimination, although subsequently, it has not routinely reported discarded rates. We estimated MLI investigation rates among insured individuals during 2010-2017, using data from the MarketScan® databases. We defined "MLI investigations" as measles serologic testing within 5 days following diagnostic codes for measles-compatible symptoms and conditions. We provide a rationale for pre-specifying three subgroups for analysis: children aged ≤15 years; males aged 16-22 years excluding data from summer months; and males aged ≥23 years. MLI investigation rates ranged from 6.6─26.4 per 100,000, remaining stable over time except during the 2015 measles outbreaks when rates increased, particularly among young children. In addition to high vaccine uptake, measles elimination requires ongoing vigilance by clinicians and high-quality, case-based surveillance. Estimated rates of MLI investigations in this U.S. population suggesting that the quality of measles surveillance is sufficiently sensitive to detect endemic measles circulation if it were to be occurring.

Factors Associated with Stroke Coding Quality: A Comparison of Registry and Administrative Data

BACKGROUND

The International Statistical Classification of Diseases and Related Health Problems, 10th Revision, Australian Modification (ICD-10-AM) codes are commonly used to identify patients with diseases or clinical conditions for epidemiological research. We aimed to determine the diagnostic agreement and factors associated with a clinician-assigned stroke diagnosis in a national registry and the ICD-10-AM codes recorded in government-held administrative data.

MATERIALS AND METHODS

Data from 39 hospitals (2009-2013) participating in the Australian Stroke Clinical Registry (AuSCR) were linked and merged with person-level administrative data. The AuSCR clinician-assigned stroke diagnosis was the reference standard. Concordance was defined as agreement between the clinician-assigned diagnosis and the ICD-10-AM codes for acute stroke or transient ischemic attack (TIA) (ICD-10-AM codes: I61-I64, G45.9). Multivariable logistic regression was undertaken to assess factors associated with coded diagnostic concordance.

RESULTS

A total of 14,716 patient admissions were included (46% female, 63% ischemic, 14% intracerebral hemorrhage [ICH], 18% TIA and 5% unspecified stroke based on the reference standard). Principal ICD-10-AM code concordance was ICH: 76.7%; ischemic stroke: 72.2%; TIA: 80.2%; unspecified stroke: 50.8%. Factors associated with a greater odds of ischemic stroke concordance included: treatment in a stroke unit (adjusted Odds Ratio, aOR:1.58; 95% confidence interval (CI) 1.37, 1.82); length of stay >4 days (aOR:1.30; 95% CI 1.17, 1.45); and discharge destination other than home (Residential care aOR:1.57; 95% CI 1.24, 1.96; Inpatient rehabilitation aOR:1.63; 95% CI 1.43, 1.86).

CONCLUSIONS

Diagnostic concordance varied based on stroke type. Future research to improve the quality of coding for stroke should focus on patients not treated in stroke units or with shorter lengths of stay where documentation in medical records may be limited.

Sex Differences in Endovascular Treatment for Stroke: A Population-based Analysis

BACKGROUND

Acute ischemic stroke may affect women and men differently. We aimed to evaluate sex differences in outcomes of endovascular treatment (EVT) for ischemic stroke due to large vessel occlusion in a population-based study in Alberta, Canada.

METHODS AND RESULTS

Over a 3-year period (April 2015-March 2018), 576 patients fit the inclusion criteria of our study and constituted the EVT group of our analysis. The medical treatment group of the ESCAPE trial had 150 patients. Thus, our total sample size was 726. We captured outcomes in clinical routine using administrative data and a linked database methodology. The primary outcome of our study was home-time. Home-time refers to the number of days that the patient was back at their premorbid living situation without an increase in the level of care within 90 days of the index stroke event. In adjusted analysis, EVT was associated with an increase of 90-day home-time by an average of 6.08 (95% CI -2.74-14.89, p-value 0.177) days in women compared to an average of 11.20 (95% CI 1.94-20.46, p-value 0.018) days in men. Further analysis revealed that the association between EVT and 90-day home-time in women was confounded by age and onset-to-treatment time.

CONCLUSIONS

We found a nonsignificant nominal reduction of 90-day home-time gain for women compared to men in this province-wide population-based study of EVT for large vessel occlusion, which was only partially explained by confounding.

Gastrointestinal Disorders and Risk of First-Ever Ischemic Stroke

BACKGROUND AND PURPOSE

Recent studies suggest that alteration of the normal gut microbiome contributes to atherosclerotic burden and cardiovascular disease. While many gastrointestinal diseases are known to cause disruption of the normal gut microbiome in humans, the clinical impact of gastrointestinal diseases on subsequent cerebrovascular disease remains unknown. We conducted an exploratory analysis evaluating the relationship between gastrointestinal diseases and ischemic stroke.

METHODS

We performed a retrospective cohort study using claims between 2008 and 2015 from a nationally representative 5% sample of Medicare beneficiaries. We included only beneficiaries ≥66 years of age. We used previously validated diagnosis codes to ascertain our primary outcome of ischemic stroke. In an exploratory manner, we categorized gastrointestinal disorders by anatomic location, disease chronicity, and disease mechanism. We used Cox proportional hazards models to examine associations of gastrointestinal disorder categories and ischemic stroke with adjustment for demographics and established vascular risk factors.

RESULTS

Among a mean of 1 725 246 beneficiaries in each analysis, several categories of gastrointestinal disorders were associated with an increased risk of ischemic stroke after adjustment for established stroke risk factors. The most notable positive associations included disorders of the stomach (hazard ratio, 1.17 [95% CI, 1.15-1.19]) and functional (1.16 [95% CI, 1.15-1.17]), inflammatory (1.13 [95% CI, 1.12-1.15]), and infectious gastrointestinal disorders (1.13 [95% CI, 1.12-1.15]). In contrast, we found no associations with stroke for diseases of the anus and rectum (0.97 [95% CI, 0.94-1.00]) or neoplastic gastrointestinal disorders (0.97 [95% CI, 0.94-1.00]).

CONCLUSIONS

In exploratory analyses, several categories of gastrointestinal disorders were associated with an increased risk of future ischemic stroke after adjustment for demographics and established stroke risk factors.

Performance of ICD-10-CM Diagnosis Codes for Identifying Acute Ischemic Stroke in a National Health Insurance Claims Database

Purpose

The validity of the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) coding for the identification of acute ischemic stroke (AIS) in Taiwan’s National Health Insurance claims database has not been investigated. This study aimed to construct and validate the case definition algorithms for AIS based on ICD-10-CM diagnostic codes.

Patients and Methods

This study identified all hospitalizations with ICD-10-CM code of I63* in any position of the discharge diagnoses from the inpatient claims database and all patients with a final diagnosis of AIS from the stroke registry between Jan 2018 and Dec 2019. Hospitalizations in the claims data that could be successfully linked to those in the registry data were regarded as true episodes of AIS. Otherwise, their electronic medical records and images were manually reviewed to ascertain whether they were true episodes of AIS. Using the true episodes of AIS as the reference standard, the positive predictive value (PPV) and sensitivity of various case definition algorithms for AIS were calculated.

Results

A total of 1227 hospitalizations were successfully linked. Among the 155 hospitalizations that could not be linked, 54 were determined to be true episodes of AIS. Using ICD-10-CM code of I63* in any position of the discharge diagnoses to identify AIS yielded a PPV and sensitivity of 92.7% and 99.4%, respectively. The PPV increased to 99.8% with >12% decrease in the sensitivity when AIS was restricted to those with I63* as the primary diagnosis. When AIS was defined to be I63* as the primary, first secondary, or second secondary diagnosis, both PPV and sensitivity were greater than 97%.

Conclusion

This study demonstrated the validity of various case definition algorithms for AIS based on ICD-10-CM coding and can provide a reference for future claims-based stroke research.

Home-Time as a Surrogate Measure for Functional Outcome After Stroke: A Validation Study

Purpose

Home-time has been found to correlate well with modified Rankin Scale (mRS) scores in patients with stroke. This study aimed to determine its correlations in patients with different types of stroke at various time points after stroke in a non-Western population.

Methods

This study used linked data from multi-center stroke registry databases and a nationwide claims database of health insurance. Functional outcomes as measured with the modified Rankin Scale were obtained from the registry databases and home-time was derived from the claims database. Spearman correlation coefficients were used to assess the correlation between home-time and mRS scores. The area under the receiver operating characteristic curve (AUC) was used to evaluate the performance of home-time in predicting good functional outcome.

Results

This study included 7959 patients hospitalized for stroke or transient ischemic attack (TIA), for whom mRS scores were available in 6809, 6694, and 4330 patients at 90, 180, and 365 days, respectively. Home-time was highly correlated with mRS scores at the three time-points in patients with ischemic (Spearman's rho -0.69 to -0.83) or hemorrhagic (Spearman's rho -0.86 to -0.88) stroke, but the correlation was only weak to moderate in those with TIA (Spearman's rho -0.32 to -0.58). Home-time predicted good functional outcome with excellent discrimination in patients with ischemic (AUCs >0.8) or hemorrhagic (AUCs >0.9) stroke but less so in those with TIA (AUCs >0.7).

Conclusion

Home-time was highly correlated with mRS scores and showed excellent discrimination in predicting good functional outcome in patients with ischemic or hemorrhagic stroke. Home-time could serve as a valid surrogate measure for functional outcome after stroke.

Comparative Effectiveness of Endovascular Treatment for Acute Ischemic Stroke: A Population-Based Analysis

Background A heterogeneous patient population receives endovascular treatment (EVT) for acute ischemic stroke caused by proximal large-vessel occlusion every day. We aimed to conduct a population-based study of EVT in the province of Alberta, Canada, to understand the effectiveness in a complete population and how the magnitude of effect differs from the artificial world of clinical trials. Methods and Results Within a 3-year period (April 2015 to March 2018), 576 patients fit the inclusion criteria of our study and constituted the EVT group of our analysis. The medical treatment group of the ESCAPE (Endovascular Treatment for Small Core and Anterior Circulation Proximal Occlusion With Emphasis on Minimizing CT [Computed Tomography] to Recanalization Times) trial had 150 patients. Thus, our total sample size was 726. We captured outcomes in clinical routine using administrative data and a linked database method. Primary outcome of our study was home-time. Home-time refers to the number of days that the patient was back at premorbid living situation without increase in level of care within 90 days of index stroke event. Median age of patients was 70 years (interquartile range, 59-81 years), and 47.8% were women. Median National Institutes of Health Stroke Scale score was 17 (interquartile range, 13-20). EVT was associated with an increased 90-day home-time by an average of 8.5 days compared with medical treatment alone using Cragg hurdle regression (P=0.009). Age and higher National Institutes of Health Stroke Scale score were associated with decreased 90-day home-time (both P<0.001). Multivariable logistic regression showed no association between EVT and mortality at 90 days (odds ratio, 0.76; 95% CI, 0.47-1.24). Conclusions EVT for acute ischemic stroke caused by proximal large-vessel occlusion was effective in our province-wide population-based study and results in an increase of 90-day home-time by ~8.5 days.

Development of stroke identification algorithm for claims data using the multicenter stroke registry database

BACKGROUND

Identifying acute ischemic stroke (AIS) among potential stroke cases is crucial for stroke research based on claims data. However, the accuracy of using the diagnostic codes of the International Classification of Diseases 10th revision was less than expected.

METHODS

From the National Health Insurance Service (NHIS) claims data, stroke cases admitted to the hospitals participating in the multicenter stroke registry (Clinical Research Collaboration for Stroke in Korea, CRCS-K) during the study period with principal or additional diagnosis codes of I60-I64 on the 10th revision of International Classification of Diseases were extracted. The datasets were randomly divided into development and validation sets with a ratio of 7:3. A stroke identification algorithm using the claims data was developed and validated through the linkage between the extracted datasets and the registry database.

RESULTS

Altogether, 40,443 potential cases were extracted from the NHIS claims data, of which 31.7% were certified as AIS through linkage with the CRCS-K database. We selected 17 key identifiers from the claims data and developed 37 conditions through combinations of those key identifiers. The key identifiers comprised brain CT, MRI, use of tissue plasminogen activator, endovascular treatment, carotid endarterectomy or stenting, antithrombotics, anticoagulants, etc. The sensitivity, specificity, and diagnostic accuracy of the algorithm were 81.2%, 82.9%, and 82.4% in the development set, and 80.2%, 82.0%, and 81.4% in the validation set, respectively.

CONCLUSIONS

Our stroke identification algorithm may be useful to grasp stroke burden in Korea. However, further efforts to refine the algorithm are necessary.

Deriving a Passive Surveillance Stroke Severity Indicator From Routinely Collected Administrative Data: The PaSSV Indicator

BACKGROUND

Adjusting for stroke severity is crucial for stroke outcomes research. However, this information is not available in administrative healthcare data. We aimed to derive an indicator of baseline stroke severity using these data.

METHODS AND RESULTS

We identified patients with stroke enrolled in a population-based registry in Ontario, Canada, and used the Canadian Neurological Scale (CNS), documented in the registry, as a measure of stroke severity. We derived an estimated CNS from a linear regression model in which we regressed the observed CNS on predictor variables: age, sex, arrival by ambulance, interhospital transfer, mechanical ventilation, and an emergency department triage score. The effect of stroke severity on the estimated hazard ratios for 30-day mortality was determined in 3 Cox-proportional hazards models with (1) no CNS, (2) observed CNS, and (3) estimated CNS, all adjusted for age, sex, Charlson index, and stroke type. We assessed model discrimination using C statistics. To assess for construct validity, we repeated these analyses in a subset of patients with documented National Institute of Health Stroke Scale and in a cohort of patients with stroke external to the registry. We derived the estimated stroke severity in 41 481 patients (48.7% female, median age of 75 years [interquartile range, 64- 83]). The magnitude of the association between stroke severity and mortality was similar for the observed and estimated CNS. The discriminative ability of the Cox-proportional hazards models to predict mortality was highest when the observed CNS was included (C statistic, 0.82 [95% CI, 0.81-0.82]), moderate with estimated CNS (0.76 [0.75-0.76]), and lowest without CNS (0.69 [0.69-0.70]. Our findings were replicated with the National Institute of Health Stroke Scale and in the external cohort.

CONCLUSIONS

We derived an estimated measure of stroke severity using administrative data. This can be applied for risk adjustment in population-based stroke outcomes research and in assessments of health system performance.

Sex Differences in Care Need and Survival in Patients Admitted to Nursing Home Poststroke

BACKGROUND

Women are more likely to be admitted to nursing home after stroke than men. Differences in patient characteristics and outcomes by sex after institutionalization are less understood. We examined sex differences in the characteristics and care needs of patients admitted to nursing home following stroke and their subsequent survival.

METHODS

We identified patients with stroke newly admitted to nursing home between April 2011 and March 2016 in Ontario, Canada, with follow-up until March 2018 using linked administrative data. We calculated prevalence ratios and 95% confidence intervals (CIs) for the primary outcomes of dependence for activities of daily living, cognitive impairment, frailty, health instability, and symptoms of depression or pain, comparing women to men. The secondary outcome was all-cause mortality.

RESULTS

Among 4831 patients, 60.9% were women. Compared to men, women were older (median age [interquartile range, IQR]: 84 [78, 89] vs. 80 [71, 86]), more likely to be frail (prevalence ratio 1.14, 95% CI [1.08, 1.19]), have unstable health (1.45 [1.28, 1.66]), and experience symptoms of depression (1.25 [1.11, 1.40]) or pain (1.21 [1.13, 1.30]), and less likely to have aggressive behaviors (0.87 [0.80, 0.94]). Overall median survival was 2.9 years. In a propensity-score-matched cohort, women had lower mortality than men (hazard ratio 0.85, 95% CI [0.77, 0.94]), but in the age-stratified survival analysis, the survival advantage in women was limited to those aged 75 years and older.

CONCLUSIONS

Despite lower subsequent mortality, women admitted to nursing home after stroke required more care than men. Pain and depression are two treatable symptoms that disproportionately affect women.

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.

Trends and Factors Associated With Concordance Between International Classification of Diseases, Ninth and Tenth Revision, Clinical Modification Codes and Stroke Clinical Diagnoses

Background and Purpose- International Classification of Diseases, Ninth and Tenth Revision, Clinical Modification (ICD-9-CM and ICD-10-CM) codes are often used for disease surveillance. We examined changes in concordance between ICD-CM codes and clinical diagnoses before and after the transition to ICD-10-CM in the United States (October 1, 2015), and determined if there were systematic variations in concordance by patient and hospital characteristics. Methods- We included Paul Coverdell National Acute Stroke Program patient discharges from 2014 to 2017. Concordance between ICD-CM codes and the clinical diagnosis documented by the physician (assumed as accurate) was calculated for each diagnosis category: ischemic stroke, transient ischemic attack, subarachnoid hemorrhage, and intracerebral hemorrhage. Results- In total, 314 857 patient records were included in the analysis (n=280 hospitals), 55.9% of which were obtained after the transition to ICD-10-CM. While concordance was generally high, a small, and temporary decline occurred from the last calendar quarter of ICD-9-CM (average unadjusted concordance =92.8%) to the first quarter of ICD-10-CM use (91.0%). Concordance differed by diagnosis category and was generally highest for ischemic stroke. In the analysis of ICD-10-CM records, disagreements often occurred between ischemic stroke and transient ischemic attack records and between subarachnoid and intracerebral hemorrhage records. Compared with the smallest hospitals (≤200 beds), larger hospitals had significantly higher odds of concordance (ischemic stroke adjusted odds ratio for ≥400 beds, 1.7; 95% CI, 1.5-1.9). Conclusions- This study identified a small and transient decline in concordance between ICD-CM codes and stroke clinical diagnoses during the coding transition, indicating no substantial impact on the overall identification of stroke patients. Researchers and policymakers should remain aware of potential changes in ICD-CM code accuracy over time, which may affect disease surveillance. Systematic variations in the accuracy of codes by hospital and patient characteristics have implications for quality-of-care studies and hospital comparative assessments.

Measuring diabetes guideline adherence with claims data: systematic construction of indicators and related challenges

OBJECTIVES

Indicators of guideline adherence are frequently used to examine the appropriateness of healthcare services. Only some potential indicators are actually usable for research with routine administrative claims data, potentially leading to a biased selection of research questions. This study aimed at developing a systematic approach to extract potential indicators from clinical practice guidelines (CPG), evaluate their feasibility for research with claims data and assess how the extracted set reflected different types of healthcare services. Diabetes mellitus (DM), Swiss national guidelines and health insurance claims data were analysed as a model case.

METHODS

CPG for diabetes patients were retrieved from the Swiss Endocrinology and Diabetes Society website. Recommendation statements involving a specific healthcare intervention for a defined patient population were translated into indicators of guideline adherence. Indicators were classified according to disease stage and healthcare service type. We assessed for all indicators whether they could be analysed with Swiss mandatory health insurance administrative claims data.

RESULTS

A total of 93 indicators were derived from 15 CPG, representing all sectors of diabetes care. For 63 indicators, the target population could not be identified using claims data only. For 67 indicators, the intervention could not be identified. Nine (10%) of all indicators were feasible for research with claims data (three addressed gestational diabetes and screening, five screening for complications and one glucose measurement). Some types of healthcare services, eg, management of risk factors, treatment of the disease and secondary prevention, lacked corresponding indicators feasible for research.

CONCLUSIONS

Our systematic approach could identify a number of indicators of healthcare service utilisation, feasible for DM research with Swiss claims data. Some areas of healthcare were covered less well. The approach could be applied to other diseases and countries, helping to identify the potential bias in the selection of indicators and optimise research.

Demand for neurological services in Central Eastern Europe: a 10-year national survey in Hungary

BACKGROUND AND PURPOSE

In order to plan neurological capacities at a national level for the next decade, the current use of neurological services should be evaluated. We analyzed the utilization of neurological services in Hungary, a country with a single-payer health insurance system covering the whole population.

METHODS

We created a database from medical reports submitted to the National Health Insurance Fund from all hospitals and outpatient services between 2004 and 2013. The number of subjects presenting to the neurological healthcare system and their major diagnoses by 10th International Classification of Diseases categories were analyzed. The overall healthcare service utilization of these patients was also estimated.

RESULTS

Of the 10 million inhabitants, 2.9 million people used an inpatient or outpatient neurological service at least once over the 10-year period. Annually, 1% of the population was admitted to neurological inpatient wards and 6% of the population used some neurological outpatient service. Major reasons for using neurological services were: cerebrovascular diseases (I60-I69; 1.2 million patients), episodic and paroxysmal disorders (G40-G47; 1.3 million patients) and general symptoms and signs (R50-R56; 1.3 million patients). The 2.9 million people had 12.7 million hospital admissions to any ward and 365.7 million outpatient visits to any specialist during the 10 years.

CONCLUSIONS

The demand for neurological services is high in Hungary; close to 30% of the population used an inpatient or outpatient neurological service at least once during this 10-year period. Results from this project provide data for international comparisons and help to ensure better informed and more focused resource allocation.

Outcomes in Hospitalized Ischemic Stroke Patients with Dementia on Admission: A Population-Based Cohort Study

Background

Dementia prevalence is rising, and it will double in the next 20 years. This study sought to understand the prevalence of dementia in hospitalized patients with ischemic stroke, and its impact on outcomes.

Methods

Using the Canadian Institute of Health Information’s (CIHI) Discharge Abstract Database (DAD), all acute ischemic stroke admissions from April 2003 to March 2015 in Canada (excluding Quebec) were analyzed. Concurrent dementia at the time of admission was assessed based on hospital diagnostic codes. Characteristics and in-hospital outcomes were compared in patients with and without dementia using χ2 and negative binomial, as well as Poisson regression analysis.

Results

During the observed period, 313,138 people were admitted to a hospital in Canada for an ischemic stroke. Of those, 21,788 (7.0%) had a concurrent diagnosis of dementia. People with dementia had older median age (84 vs. 76 years; p<0.0001), were more often female (59.6% vs. 48.4%; p<0.0001) and more often had Charlson-Deyo Comorbidity Index ≥2 (64.5% vs. 43.5%; p<0.0001). Patients with dementia were less likely to be discharged to a rehabilitation facility (adjusted risk ratio [RR] 3.089, 95% confidence interval [CI] 2.992-3.188, p<0.0001) or home independently (adjusted RR 0.756, 95% CI 0.737-0.776, p<0.0001).

Interpretation

Approximately 1 in 13 hospitalized ischemic stroke patients has coded dementia. Patients with ischemic stroke and concurrent dementia have higher mortality, face significantly more dependence after stroke and utilize greater healthcare resources than stroke patients without dementia. Causative conclusions are limited by the administrative data source. Early care planning and coordination could potentially optimize outcomes.

Administrative data underestimate acute ischemic stroke events and thrombolysis treatments: Data from a multicenter validation survey in Italy

Background

Informing health systems and monitoring hospital performances using administrative data sets, mainly hospital discharge data coded according to International-Classification-Diseases-9edition-Clinical-Modifiers (ICD9-CM), is now commonplace in several countries, but the reliability of diagnostic coding of acute ischemic stroke in the routine practice is uncertain. This study aimed at estimating accuracy of ICD9-CM codes for the identification of acute ischemic stroke and the use of thrombolysis treatment comparing hospital discharge data with medical record review in all the six hospitals of the Florence Area, Italy, through 2015.

Methods

We reviewed the medical records of all the 3915 potential acute stroke events during 2015 across the six hospitals of the Florence Area, Italy. We then estimated sensitivity and Positive Predictive Value of ICD9-CM code-groups 433*1, 434*1 and thrombolysis code 99.10 against medical record review with clinical adjudication. For each false-positive case we obtained the actual diagnosis. For each false-negative case we obtained the primary and secondary ICD9-CM diagnoses.

Results

The medical record review identified 1273 acute ischemic stroke events. The hospital discharge records identified 898 among those (true-positive cases),but missed 375 events (false-negative cases), and identified 104 events that were not eventually confirmed as acute ischemic events (false-positive cases). Code-group specific Positive Predictive Value was 85.7% (95%CI,74.6–93.3) for 433*1 and 89.9% (95%CI, 87.8–91.7) for 434*1 codes. Thrombolysis treatment, as identified by ICD9-CM code 99.10, was only documented in 6.0% of acute ischemic stroke events, but was 13.6% in medical record review.

Conclusions

Hospital discharge data were found to be fairly specific but insensitive in the reporting of acute ischemic stroke and thrombolysis, providing misleading indications about both quantity and quality of acute ischemic stroke hospital care. Efforts to improve coding accuracy should precede the use of hospital discharge data to measure hospital performances in acute ischemic stroke care.

Population-based study of home-time by stroke type and correlation with modified Rankin score

OBJECTIVE

To describe home-time, stratified by stroke type, in a complete population and to determine its correlation with modified Rankin Scale (mRS) scores.

METHODS

We used linked administrative data to derive home-time in all patients admitted for a cerebrovascular event in Alberta, Canada, between 2012 and 2016. Home-time is the number of days spent outside a health institution in the first 90 days after index hospitalization. We used negative binomial regression, adjusted for age, sex, Charlson comorbidity index, and hospital location, to determine the association between home-time and stroke type. In 552 patients enrolled in 4 acute ischemic stroke clinical trials, we used multivariable ordinal logistic regression analysis to determine the association between home-time and mRS score at 90 days.

RESULTS

Among 15,644 patients (n = 10,428 with ischemic stroke, n = 1,415 with intracerebral hemorrhage, n = 760 with subarachnoid hemorrhage, n = 3,041 with TIA), patients with TIA have the longest home-time, almost triple the number of days at home compared to patients with intracerebral hemorrhage (incidence rate ratio 2.85, 95% confidence interval [CI] 2.58-3.15). Among clinical trial ischemic stroke patients, longer home-time was associated with a lower mRS score at 90 days (adjusted common odds ratio 1.04, 95% CI 1.04-1.05).

CONCLUSIONS

We showed that home-time is an objective and graded indicator that is correlated with disability after stroke. It is obtainable from administrative data, applicable to different stroke types, and a valuable outcome indicator in population-based health services research.

Moderate sensitivity and high specificity of emergency department administrative data for transient ischemic attacks

Background

Validation of administrative data case definitions is key for accurate passive surveillance of disease. Transient ischemic attack (TIA) is a condition primarily managed in the emergency department. However, prior validation studies have focused on data after inpatient hospitalization. We aimed to determine the validity of the Canadian 10th International Classification of Diseases (ICD-10-CA) codes for TIA in the national ambulatory administrative database.

Methods

We performed a diagnostic accuracy study of four ICD-10-CA case definition algorithms for TIA in the emergency department setting. The study population was obtained from two ongoing studies on the diagnosis of TIA and minor stroke versus stroke mimic using serum biomarkers and neuroimaging. Two reference standards were used 1) the emergency department clinical diagnosis determined by chart abstractors and 2) the 90-day final diagnosis, both obtained by stroke neurologists, to calculate the sensitivity, specificity, positive and negative predictive values (PPV and NPV) of the ICD-10-CA algorithms for TIA.

Results

Among 417 patients, emergency department adjudication showed 163 (39.1%) TIA, 155 (37.2%) ischemic strokes, and 99 (23.7%) stroke mimics. The most restrictive algorithm, defined as a TIA code in the main position had the lowest sensitivity (36.8%), but highest specificity (92.5%) and PPV (76.0%). The most inclusive algorithm, defined as a TIA code in any position with and without query prefix had the highest sensitivity (63.8%), but lowest specificity (81.5%) and PPV (68.9%). Sensitivity, specificity, PPV, and NPV were overall lower when using the 90-day diagnosis as reference standard.

Conclusions

Emergency department administrative data reflect diagnosis of suspected TIA with high specificity, but underestimate the burden of disease. Future studies are necessary to understand the reasons for the low to moderate sensitivity.

Electronic supplementary material

The online version of this article (10.1186/s12913-017-2612-6) contains supplementary material, which is available to authorized users.

Association Between Cirrhosis and Stroke in a Nationally Representative Cohort

Importance

Cirrhosis is associated with hemorrhagic and thrombotic extrahepatic complications. The risk of cerebrovascular complications is less well understood.

Objective

To investigate the association between cirrhosis and various stroke types.

Design, Setting, and Participants

We performed a retrospective cohort study using inpatient and outpatient Medicare claims data from January 1, 2008, through December 31, 2014, for a random 5% sample of 1 618 059 Medicare beneficiaries older than 66 years.

Exposures

Cirrhosis, as defined by a validated diagnosis code algorithm.

Main Outcomes and Measures

The primary outcome was stroke, and secondary outcomes were ischemic stroke, intracerebral hemorrhage, and subarachnoid hemorrhage as defined by validated diagnosis code algorithms.

Results

Among 1 618 059 beneficiaries, 15 586 patients (1.0%) had cirrhosis (mean [SD] age, 74.1 [6.9] years; 7263 [46.6%] female). During a mean (SD) of 4.3 (1.9) years of follow-up, 77 268 patients were hospitalized with a stroke. The incidence of stroke was 2.17% (95% CI, 1.99%-2.36%) per year in patients with cirrhosis and 1.11% (95% CI, 1.10%-1.11%) per year in patients without cirrhosis. After adjustment for demographic characteristics and stroke risk factors, patients with cirrhosis had a higher risk of stroke (hazard ratio [HR], 1.4; 95% CI, 1.3-1.5). The magnitude of association appeared to be higher for intracerebral hemorrhage (HR, 1.9; 95% CI, 1.5-2.4) and subarachnoid hemorrhage (HR, 2.4; 95% CI, 1.7-3.5) than for ischemic stroke (HR, 1.3; 95% CI, 1.2-1.5).

Conclusions and Relevance

In a nationally representative sample of Medicare beneficiaries, cirrhosis was associated with an increased risk of stroke, particularly hemorrhagic stroke. A potential explanation of these findings implicates the mixed coagulopathy observed in cirrhosis.

A cohort study on physician documentation and the accuracy of administrative data coding to improve passive surveillance of transient ischaemic attacks

BACKGROUND

Administrative health data are valuable in health research and disease surveillance, but have low to moderate sensitivity in identifying transient ischaemic attacks (TIA) in the emergency department (ED). We aimed to identify the predictors of coding accuracy for TIA.

METHODS

The study population was obtained from two ongoing studies on the diagnosis of TIA, minor stroke and stroke mimic. ED charts were manually reviewed by a stroke neurologist to obtain the clinical diagnosis, patient characteristics and content of physician documentation. Administrative data codes were compared with the chart-adjudicated diagnosis to determine cases of misclassification by administrative data. Univariable regression was used to evaluate candidate predictors of disagreement, and the significant variables were tested in a multivariable model to obtain an adjusted estimate of effect.

RESULTS

Among 417 patients (39.1% TIA, 37.2% minor stroke and 23.7% stroke mimics), there were 122 cases of disagreement between adjudications and administrative data codes for the diagnosis of TIA. The majority of disagreement (n=103/122, 84.4%) arose from adjudicated TIA cases that were misclassified as non-TIA in administrative data coding. There were 78 (18.7%) charts with documented uncertain diagnosis, and 73 (17.5%) charts had no definite diagnosis. The relative risk of disagreement between chart adjudication and administrative data coding when the final diagnosis was uncertain or absent was 1.82(1.36, 2.44) and the risk difference was 18.5%. Multivariable logistic regression analyses confirmed this association using different case definition algorithms.

CONCLUSIONS

In suspected patients with TIA and minor stroke presenting to the ED, physician documentation was the dominant factor in coding accuracy, supporting the concept that physicians are active participants in administrative data coding. Strategies to improve chart documentation are predicted to have a positive effect on coding accuracy.

Trends in vascular risk factors, stroke performance measures, and outcomes in patients with first-ever ischemic stroke in Taiwan between 2000 and 2012

BACKGROUND

With the aging of the population in Taiwan, the financial burden of stroke on the healthcare system is expected to rise. We aimed to investigate the trends in vascular risk factors, adherence to stroke performance measures, and stroke outcomes based on a nationwide representative sample.

METHODS

Adult patients hospitalized for first-ever ischemic stroke between 2000 and 2012 were identified from a nationwide administrative database. The study period was divided into 1-year intervals. The Cuzick test and the Cochran-Armitage test were used to determine the significance of changes over time. Trends in stroke outcomes as a function of year were assessed using logistic regression, controlling for age, sex, comorbidity, and stroke severity.

RESULTS

A total of 11,462 patients (mean age 67.3years, female 40.9%) were hospitalized. Between 2000 and 2012, the prevalence of hypertension, diabetes mellitus, hyperlipidemia, and atrial fibrillation increased while the prevalence of coronary artery disease decreased. The proportion of patients taking antihypertensive or antidiabetic medication prior to stroke decreased, whereas the proportion of patients taking lipid lowering medication increased. Adherence to the five selected performance measures significantly improved. A significant decreasing trend in the proportion of recurrent stroke or all-cause death within one year was observed regardless of whether adjustment for age, sex, comorbidity, and stroke severity was made.

CONCLUSIONS

Despite the rising prevalence of vascular risk factors, improved adherence to stroke performance measures was accompanied by better stroke outcomes.