Administrative data for public health surveillance and planning

Barriers and facilitators for using administrative data for surveillance purpose: a narrative overview

Although administrative data are not originally intended for surveillance purposes, they are frequently used for monitoring public health and patient safety. This article provides a narrative overview of the barriers and facilitators for the use of administrative data for surveillance, with a focus on healthcare-associated infection (HAI) in Canada. In this case, only articles on administrative data in general or related to HAI were included. Validation study and meta-analyses on administrative data accuracy were excluded. Medline, Embase and Google Scholar were searched as well as references list of all included articles, for a total of 90 articles included. Our analysis identified 78 barriers at the individual, organizational and systemic levels and outlined 75 facilitators and solutions to improve administrative data utilization and quality. This narrative overview will help to understand barriers, facilitators and offer practical recommendations for optimizing the use of administrative data.

Patterns of prescription opioid use and opioid-related harms among adult patients with hematologic malignancies

INTRODUCTION

Treatment advances for hematologic malignancies (HM) have dramatically improved life expectancy, necessitating greater focus on long-term cancer pain management. This study explored real-world patterns of opioid use among patients with HM.

METHODS

This retrospective cohort study identified adults diagnosed with HM from January 1, 2013 through December 31, 2019 using the Truven MarketScan Commercial Claims and Encounters database. Across several HM types, we described rates of high-risk opioid use (based on Pharmacy Quality Alliance measures) and opioid-related harms, including incident opioid use disorder (OUD) diagnoses and opioid-related hospitalizations or emergency department (ED) visits. We used multivariable Cox regression to generate adjusted hazard ratios and 95% confidence intervals comparing the risk of opioid-related harms between patients with versus without high-risk opioid use.

RESULTS

Our sample included 43,190 patients with HM. Median age at HM diagnosis was 54 years (interquartile range  =  44-60). Most patients (61.9%) were diagnosed with lymphoma. Approximately half (49.2%) had an opioid dispensed in the follow-up period. Among all patients, 20.0% met criteria for high-risk opioid use, 0.9% had an OUD diagnosis, and 0.3% experienced an opioid-related hospitalization/ED visit in follow-up. High-risk opioid use increased the risk of an OUD diagnosis by 3.3 times (p < 0.0001) and an opioid-related hospitalization/ED visit 4.2 times (p < 0.0001).

CONCLUSION

High-risk opioid use was prevalent among patients with HM and significantly increased the risk of opioid-related harms. However, rates of opioid-related harms were low. These findings highlight the importance of continually monitoring pain and opioid use throughout HM survivorship to provide safe, effective HM pain management.

Monitoring Cardiovascular Disease in Métis Citizens Across Ontario, 2012-2020

Background

Population-based analyses of Métis-specific health outcomes in Canada are limited. This study aimed to address this gap and examine cardiovascular disease outcomes in citizens of the Métis Nation of Ontario (MNO) over a 9-year period.

Methods

Under a data governance and sharing agreement between the MNO and ICES, registered MNO citizens aged ≥ 20 years were linked to administrative health data in Ontario. Existing algorithms were used to determine the burden of heart failure and hypertension. In the most recent year, prevalence rates were compared for income quintiles, age, and sex.

Results

Age-adjusted prevalence rates of hypertension decreased, and age-adjusted prevalence rates of heart failure increased in MNO citizens from 2012 to 2020. A larger decrease in prevalence of hypertension was observed for female citizens, by 12% from 28.9 per 100 (confidence interval [CI]: 27.6-30.2) in 2012 to 25.4 per 100 (CI: 24.5-26.3) in 2020. As for heart failure, the age-adjusted prevalence rates for male citizens had the largest increase, by 47% from 2.6 per 100 (CI: 2.1-3.1) in 2012 to 3.8 per 100 (CI: 3.3-4.2) in 2020. Hypertension and heart failure were more prevalent in male citizens, those of advanced age, and those living in areas within the lowest income quintile.

Conclusions

This study is the first in nearly 10 years to investigate trends in cardiovascular outcomes among MNO citizens. Understanding this burden is critical to the MNO's ability to guide program and policy planning, as well as to advocate within and beyond the health system for Métis-specific needs.

Mycoplasma pneumoniae Outbreak in 2023: Post-pandemic Resurgence of an Atypical Bacterial Pathogen

The syndromic nature of infections caused by pneumonia-causing pathogens including Mycoplasma pneumoniae necessitates detection via multiplex PCR for accurate and timely diagnosis to control the infection spread. In this study, we demonstrate an increase in the detection of M. pneumoniae in the outpatient population, during 2023, as compared to the previous two years (2021-2022). In this aggregated survey, respiratory samples collected within the continental United States were tested for the presence of M. pneumoniae and other respiratory bacterial and viral pathogens using a multiplex PCR assay. Patient data was analyzed on the basis of age, gender and geographical location. The positive detection of M. pneumoniae in 2021 and 2022 was 0.004% and 0.006%, respectively. The positivity rate of M. pneumoniae in 2023 increased to 0.21%. The highest proportion of M. pneumoniae cases were detected from Georgia with the outbreak generally concentrated in large urban settings. Median age of the patients testing positive for M. pneumoniae was 10 (interquartile range [IQR] 8-18) years with an almost equal distribution between male and female patients. Other respiratory, viral and bacterial, pathogens detected in samples positive for M. pneumoniae were similar in proportion to the M. pneumonia-negative population. A survey of the ICD-10 codes submitted in conjunction with the samples suggests that the current outbreak is mostly associated with upper respiratory tract infections. The present study is the first detailed report in the United States that shows an unprecedented increase in the detection of M. pneumoniae in the outpatient population during 2023. Our analysis suggests that this outbreak was not associated with any other bacterial or viral respiratory pathogen. The outbreak of this atypical pathogen was concentrated in the pediatric population in large urban areas. The 2023 outbreak could be a return of the cyclical M. pneumoniae outbreaks witnessed prior to the COVID-19 pandemic. Our study highlights the importance of performing continuous surveillance of respiratory pathogens, especially in the altered epidemiological landscape of the post-COVID world.

Gaps in U.S. Public Health Monitoring and Surveillance Systems in Puerto Rico

INTRODUCTION

Puerto Rico (PR) has an ambiguous status within Centers for Disease Control and Prevention (CDC) public health monitoring and surveillance systems. However, as a U.S. territory, PR relies on federal health resource allocation processes that are informed by these data sources. This study sought to quantify the coverage of PR within CDC-supported systems compared with the 50 states, describe coverage across critical system-types, and assess the availability of data collection instruments in Spanish.

METHODS

This cross-sectional observational study employed the method of data curation by identifying, collecting, and categorizing data primarily from web-based sources maintained by the CDC. Data were originally collected and coded from August 2021 to March 2022 and reviewed by two additional coders from October 2022 to March 2023. Each curated system was assessed to determine coverage of PR across five system-types (probability-based, case-based, administrative, registry, and multiple-source) compared with the 50 states. The availability of data collection instruments in Spanish was also assessed.

RESULTS

Of 93 active CDC-supported systems assessed, results revealed that PR was not covered in 54% of CDC-supported systems. Comparatively, inclusive coverage of all 50 states was only lacking in 33% of CDC-supported systems. Of the 32 CDC-supported systems in PR that could be coded for language, only 53% had data collection instruments in Spanish.

CONCLUSIONS

There are significant gaps in the coverage of PR within CDC-supported systems. Future efforts must be made to identify the reasons for this exclusion and increase the territory's representation within these essential public health data systems.

Administrative Model for Profiling Hospital Performance on Coronary Artery Bypass Graft Surgery: Based on the Chinese Hospital Quality Monitoring System

BACKGROUND

We aimed to develop an administrative model to profile the performance on the outcomes of coronary artery bypass grafting across hospitals in China.

METHODS AND RESULTS

This retrospective study was based on the Chinese Hospital Quality Monitoring System (HQMS) from 2016 to 2020. The coronary artery bypass grafting cases were identified by procedure code, and those of 2016 to 2017 were randomly divided into modeling and validation cohorts, while those in other years were used to ensure the model stability across years. The outcome was discharge status as "death or withdrawal," and that withdrawal referred to discharge without medical advice when patients were in the terminal stage but reluctant to die in the hospital. Candidate covariates were mainly identified by diagnoses or procedures codes. Patient-level logistic models and hospital-level hierarchical models were established. A total of 203 010 coronary artery bypass grafts in 699 hospitals were included, with 60 704 and 20 233 cases in the modeling and validation cohorts and 40 423, 42 698, and 38 952 in the years 2018, 2019, and 2020, respectively. The death or withdrawal rate was 3.4%. The areas under the curve were 0.746 and 0.729 in the patient-level models of modeling and validation cohorts, respectively, with good calibration and stability across years. Hospital-specific risk-standardized death or withdrawal rates were 2.61% (interquartile range, 1.87%-3.99%) and 2.63% (interquartile range, 1.97%-3.44%) in the modeling and validation cohorts, which were highly correlated (correlation coefficient, 0.96; P<0.001). Between-hospital variations were distinguished among hospitals of different volumes and across years.

CONCLUSIONS

The administrative model based on Hospital Quality Monitoring System could profile hospital performance on coronary artery bypass grafting in China.

Identification of people with low prevalence diseases in administrative healthcare records: A case study of HIV in British Columbia, Canada

INTRODUCTION

Case-finding algorithms can be applied to administrative healthcare records to identify people with diseases, including people with HIV (PWH). When supplementing an existing registry of a low prevalence disease, near-perfect specificity helps minimize impacts of adding in algorithm-identified false positive cases. We evaluated the performance of algorithms applied to healthcare records to supplement an HIV registry in British Columbia (BC), Canada.

METHODS

We applied algorithms based on HIV-related diagnostic codes to healthcare practitioner and hospitalization records. We evaluated 28 algorithms in a validation sub-sample of 7,124 persons with positive HIV tests (2,817 with a prior negative test) from the STOP HIV/AIDS data linkage-a linkage of healthcare, clinical, and HIV test records for PWH in BC, resembling a disease registry (1996-2020). Algorithms were primarily assessed based on their specificity-derived from this validation sub-sample-and their impact on the estimate of the total number of PWH in BC as of 2020.

RESULTS

In the validation sub-sample, median age at positive HIV test was 37 years (Q1: 30, Q3: 46), 80.1% were men, and 48.9% resided in the Vancouver Coastal Health Authority. For all algorithms, specificity exceeded 97% and sensitivity ranged from 81% to 95%. To supplement the HIV registry, we selected an algorithm with 99.89% (95% CI: 99.76% - 100.00%) specificity and 82.21% (95% CI: 81.26% - 83.16%) sensitivity, requiring five HIV-related healthcare practitioner encounters or two HIV-related hospitalizations within a 12-month window, or one hospitalization with HIV as the most responsible diagnosis. Upon adding PWH identified by this highly-specific algorithm to the registry, 8,774 PWH were present in BC as of March 2020, of whom 333 (3.8%) were algorithm-identified.

DISCUSSION

In the context of an existing low prevalence disease registry, the results of our validation study demonstrate the value of highly-specific case-finding algorithms applied to administrative healthcare records to enhance our ability to estimate the number of PWH living in BC.

Positive Predictive Value of International Classification of Diseases, Ninth Revision, Clinical Modification, and International Classification of Diseases, Tenth Revision, Clinical Modification, Codes for Identification of Congenital Heart Defects

Background Administrative data permit analysis of large cohorts but rely on International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), and International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes that may not reflect true congenital heart defects (CHDs). Methods and Results CHDs in 1497 cases with at least 1 encounter between January 1, 2010 and December 31, 2019 in 2 health care systems, identified by at least 1 of 87 ICD-9-CM/ICD-10-CM CHD codes were validated through medical record review for the presence of CHD and CHD native anatomy. Interobserver and intraobserver reliability averaged >95%. Positive predictive value (PPV) of ICD-9-CM/ICD-10-CM codes for CHD was 68.1% (1020/1497) overall, 94.6% (123/130) for cases identified in both health care systems, 95.8% (249/260) for severe codes, 52.6% (370/703) for shunt codes, 75.9% (243/320) for valve codes, 73.5% (119/162) for shunt and valve codes, and 75.0% (39/52) for "other CHD" (7 ICD-9-CM/ICD-10-CM codes). PPV for cases with >1 unique CHD code was 85.4% (503/589) versus 56.3% (498/884) for 1 CHD code. Of cases with secundum atrial septal defect ICD-9-CM/ICD-10-CM codes 745.5/Q21.1 in isolation, PPV was 30.9% (123/398). Patent foramen ovale was present in 66.2% (316/477) of false positives. True positives had younger mean age at first encounter with a CHD code than false positives (22.4 versus 26.3 years; P=0.0017). Conclusions CHD ICD-9-CM/ICD-10-CM codes have modest PPV and may not represent true CHD cases. PPV was improved by selecting certain features, but most true cases did not have these characteristics. The development of algorithms to improve accuracy may improve accuracy of electronic health records for CHD surveillance.

Comparison of Administrative Database-Derived and Hospital-Derived Data for Monitoring Blood Culture Use in the Pediatric Intensive Care Unit

BACKGROUND

Optimizing blood culture practices requires monitoring of culture use. Collecting culture data from electronic medical records can be resource intensive. Our objective was to determine whether administrative data could serve as a data source to measure blood culture use in pediatric intensive care units (PICUs).

METHODS

Using data from a national diagnostic stewardship collaborative to reduce blood culture use in PICUs, we compared the monthly number of blood cultures and patient-days collected from sites (site-derived) and the Pediatric Health Information System (PHIS, administrative-derived), an administrative data warehouse, for 11 participating sites. The collaborative's reduction in blood culture use was compared using administrative-derived and site-derived data.

RESULTS

Across all sites and months, the median of the monthly relative blood culture rate (ratio of administrative- to site-derived data) was 0.96 (Q1: 0.77, Q3: 1.24). The administrative-derived data produced an estimate of blood culture reduction over time that was attenuated toward the null compared with site-derived data.

CONCLUSIONS

Administrative data on blood culture use from the PHIS database correlates unpredictably with hospital-derived PICU data. The limitations of administrative billing data should be carefully considered before use for ICU-specific data.

Maternal comorbidity and adverse perinatal outcomes in survivors of adolescent and young adult cancer: A cohort study

OBJECTIVE

To evaluate risks of preterm birth (PTB) and severe maternal morbidity (SMM) in female survivors of adolescent and young adult cancer and assess maternal comorbidity as a potential mechanism. To determine whether associations differ by use of assisted reproductive technology (ART).

DESIGN

Retrospective cohort.

SETTING

Commercially insured females in the USA.

SAMPLE

Females with live births from 2000-2019 within a de-identified US administrative health claims data set.

METHODS

Log-binomial regression models estimated relative risks of PTB and SMM by cancer status and tested for effect modification. Causal mediation analysis evaluated the proportions explained by maternal comorbidity.

MAIN OUTCOME MEASURES

PTB and SMM.

RESULTS

Among 46 064 cancer survivors, 2440 singleton births, 214 multiple births and 2590 linked newborns occurred after cancer diagnosis. In singleton births, the incidence of PTB was 14.8% in cancer survivors versus 12.4% in females without cancer (aRR 1.19, 95% CI 1.06-1.34); the incidence of SMM was 3.9% in cancer survivors versus 2.4% in females without cancer (aRR 1.44, 95% CI 1.13-1.83). Cancer survivors had more maternal comorbidities before and during pregnancy; 26% of the association between cancer and PTB and 30% of the association between cancer and SMM was mediated by maternal comorbidities. Tests for effect modification of cancer status on perinatal outcomes by ART were non-significant.

CONCLUSIONS

Preterm birth and SMM risks were modestly increased after cancer. Significant proportions of elevated risks may result from increased comorbidities. ART did not significantly modify the association between adolescent and young adult cancer and adverse perinatal outcomes. The prevention and treatment of comorbidities provides an opportunity to improve perinatal outcomes among cancer survivors.

Performance of French medico-administrative databases in epidemiology of infectious diseases: a scoping review

The development of medico-administrative databases over the last few decades has led to an evolution and to a significant production of epidemiological studies on infectious diseases based on retrospective medical data and consumption of care. This new form of epidemiological research faces numerous methodological challenges, among which the assessment of the validity of targeting algorithm. We conducted a scoping review of studies that undertook an estimation of the completeness and validity of French medico-administrative databases for infectious disease epidemiological research. Nineteen validation studies and nine capture-recapture studies were identified. These studies covered 20 infectious diseases and were mostly based on the evaluation of hospital claimed data. The evaluation of their methodological qualities highlighted the difficulties associated with these types of research, particularly those linked to the assessment of their underlying hypotheses. We recall several recommendations relating to the problems addressed, which should contribute to the quality of future evaluation studies based on medico-administrative data and consequently to the quality of the epidemiological indicators produced from these information systems.

Effect of Variation in Early Rehabilitation on Hospital Readmission After Hip Fracture

OBJECTIVE

Provision of early rehabilitation services during acute hospitalization after a hip fracture is vital for improving patient outcomes. The purpose of this study was to examine the association between the amount of rehabilitation services received during the acute care stay and hospital readmission in older patients after a hip fracture.

METHODS

Medicare claims data (2016-2017) for older adults admitted to acute hospitals for a hip fracture (n = 131,127) were used. Hospital-based rehabilitation (physical therapy, occupational therapy, or both) was categorized into tertiles by minutes per day as low (median = 17.5), middle (median = 30.0), and high (median = 48.8). The study outcome was risk-adjusted 7-day and 30-day all-cause hospital readmission.

RESULTS

The median hospital stay was 5 days (interquartile range [IQR] = 4-6 days). The median rehabilitation minutes per day was 30 (IQR = 21-42.5 minutes), with 17 (IQR = 12.6-20.6 minutes) in the low tertile, 30 (IQR = 12.6-20.6 minutes) in the middle tertile, and 48.8 (IQR = 42.8-60.0 minutes) in the high tertile. Compared with high therapy minutes groups, those in the low and middle tertiles had higher odds of a 30-day readmission (low tertile: odds ratio [OR] = 1.11, 95% CI = 1.06-1.17; middle tertile: OR = 1.07, 95% CI = 1.02-1.12). In addition, patients who received low rehabilitation volume had higher odds of a 7-day readmission (OR = 1.20; 95% CI = 1.10-1.30) compared with high volume.

CONCLUSION

Elderly patients with hip fractures who received less rehabilitation were at higher risk of readmission within 7 and 30 days.

IMPACT

These findings confirm the need to update clinical guidelines in the provision of early rehabilitation services to improve patient outcomes during acute hospital stays for individuals with hip fracture.

LAY SUMMARY

There is significant individual- and hospital-level variation in the amount of hospital-based rehabilitation delivered to older adults during hip fracture hospitalization. Higher intensity of hospital-based rehabilitation care was associated with a lower risk of hospital readmission within 7 and 30 days.

Validity of diagnoses of respiratory diseases recorded in a Japanese administrative database

BACKGROUND

Validating the information recorded in administrative databases is essential. However, no study has comprehensively validated the accuracy of Japanese Diagnosis Procedure Combination (DPC) data on various respiratory diseases. Therefore, this study aimed to evaluate the validity of diagnoses of respiratory diseases in the DPC database.

METHODS

We conducted chart reviews of 400 patients hospitalized in the departments of respiratory medicine in two acute-care hospitals in Tokyo, between April 1, 2019 and March 31, 2021, and used them as reference standards. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of DPC data on 25 respiratory diseases were determined.

RESULTS

Sensitivity ranged from 22.2% (aspiration pneumonia) to 100% (chronic eosinophilic pneumonia and malignant pleural mesothelioma) and was <50% for eight diseases, while specificity was >90% for all diseases. PPV ranged from 40.0% (aspiration pneumonia) to 100% (coronavirus disease 2019, bronchiectasis, chronic eosinophilic pneumonia, pulmonary hypertension, squamous cell carcinoma, small cell carcinoma, lung cancer of other histological types, and malignant pleural mesothelioma) and was >80% for 16 diseases. Except for chronic obstructive pulmonary disease (82.9%) and interstitial pneumonia (other than idiopathic pulmonary fibrosis) (85.4%), NPV was >90% for all diseases. These validity indices were similar in both hospitals.

CONCLUSIONS

The validity of diagnoses of respiratory diseases in the DPC database was high in general, thereby providing an important basis for future studies.

Wide Geographic Variation in Resource Utilization after Shoulder Arthroplasty

Background

There is growing interest in optimizing cost and resource utilization after shoulder arthroplasty, but little data to guide improvement efforts.

Objective

The purpose of this study was to evaluate geographic variation in length of stay and home discharge disposition after shoulder arthroplasty across the United States.

Methods

The Centers for Medicare and Medicaid Services database was used to identify Medicare discharges following shoulder arthroplasties performed from April 2019 through March 2020. National, regional (Northeast, Midwest, South, West), and state-level variation in length of stay and home discharge disposition rates were examined. The degree of variation was assessed using the coefficient of variation, with a value greater than 0.15 being considered as "substantial" variation. Geographic maps were created for visual representation of the data.

Results

There was substantial state-level variation in home discharge disposition rates (64% in Connecticut to 96% in West Virginia) and length of stay (1.01 days in Delaware to 1.86 days in Kansas). There was wide regional variation in length of stay (1.35 days in the West to 1.50 days in the Northeast) and home discharge disposition rates (73% in the Northeast to 85% in the West).

Conclusions

There is wide variation in resource utilization after shoulder arthroplasty across the United States. Certain patterns emerge from our data; for instance, the Northeast has the longest hospital stays with the lowest home discharge rates. This study provides important information for the implementation of targeted strategies to effectively reduce geographic variation in healthcare resource utilization.

Identification of asthma cases in Japan using health insurance claims data: Positive and negative predictive values of proposed discrimination criteria: A single-center study

BACKGROUND

Asthma cases have been increasingly investigated using claims data. However, the validity of defining asthma cases using health insurance claims in Japan is unclear. This study aims to assess the positive and negative predictive values of our proposed discrimination criteria for asthma.

METHODS

We developed discrimination criteria for asthma based on both the International Statistical Classification of Diseases and Related Health Problems (ICD)-10 disease codes for asthma and health insurance claims data for prescriptions and the treatment of asthma. Inclusion criteria were patients aged ≥16 years with at least one health insurance claim from April 2018 to March 2019 in all departments of our hospital. Physician-diagnosed asthma documented in the charts was used as the reference standard. Positive and negative predictive values of the discrimination criteria for physician-diagnosed asthma were estimated and compared with those estimated from discrimination criteria based solely on ICD-10 codes.

RESULTS

The new discrimination criteria had a high positive predictive value (PPV) of 86.0%, which was significantly higher than the PPV for the criteria defined solely by the ICD-10 codes (61.5%) (P < 0.01). The negative predictive values for both criteria were 100%. Allergic rhinitis and chronic cough were frequently misclassified as asthma using the discrimination criteria based solely on ICD-10 codes but were more likely to be appropriately classified using our proposed criteria.

CONCLUSIONS

Our proposed criteria adequately identified asthma subjects using health insurance claims data in Japan with a high PPV. Further studies are needed for external validation of these criteria.

Cohort profile: the South African National Health Laboratory Service (NHLS) National HIV Cohort

PURPOSE

South Africa's National Health Laboratory Service (NHLS) National HIV Cohort was established in 2015 to facilitate monitoring, evaluation and research on South Africa's National HIV Treatment Programme. In South Africa, 84.8% of people living with HIV know their HIV status; 70.7% who know their status are on ART; and 87.4% on ART are virologically suppressed.

PARTICIPANTS

The NHLS National HIV Cohort includes the laboratory data of nearly all patients receiving HIV care in the public sector since April 2004. Patients are included in the cohort if they have received a CD4 count or HIV RNA viral load (VL) test. Using an anonymised unique patient identifier that we have developed and validated to linked test results, we observe patients prospectively through their laboratory results as they receive HIV care and treatment. Patients in HIV care are seen for laboratory monitoring every 6-12 months. Data collected include age, sex, facility location and test results for CD4 counts, VLs and laboratory tests used to screen for potential treatment complications.

FINDINGS TO DATE

From April 2004 to April 2018, 63 million CD4 count and VL tests were conducted at 5483 facilities. 12.6 million unique patients had at least one CD4 count or VL, indicating they had accessed HIV care, and 7.1 million patients had a VL test indicating they had started antiretroviral therapy. The creation of NHLS National HIV Cohort has enabled longitudinal research on all lab-monitored patients in South Africa's national HIV programme, including analyses of (1) patient health at presentation; (2) care outcomes such as 'CD4 recovery', 'retention in care' and 'viral resuppression'; (3) patterns of transfer and re-entry into care; (4) facility-level variation in care outcomes; and (5) impacts of policies and guideline changes.

FUTURE PLANS

Continuous updating of the cohort, integration with available clinical data, and expansion to include tuberculosis and other lab-monitored comorbidities.

Validity of operative information in Japanese administrative data: a chart review-based analysis of 1221 cases at a single institution

PURPOSE

To evaluate the validity of operative information recorded in the Diagnosis Procedure Combination (DPC) database, a national inpatient database including administrative claims data.

METHODS

We reviewed the medical charts of 1221 patients who underwent one of six surgeries (breast, esophageal, gastric, thyroid cancer surgery, appendectomy, or inguinal hernia repair) at a surgery department of a university hospital from April 2016 to March 2019. We compared operative information (type, date, laterality of procedure; type of anesthesia; transfusion; and duration of anesthesia) recorded in the DPC database with the information recorded in the medical charts.

RESULTS

The DPC data for type, date, laterality of surgery, and type of anesthesia were accurate in 99% of the reviewed patients. The sensitivity and specificity for identifying whether a patient received a transfusion procedure were 97.5% and 99.6%, respectively. Data regarding the duration of anesthesia in the DPC database were identical to those in medical chart records in 1114 of 1216 cases that received general or spinal anesthesia (91.5%). The duration of anesthesia in the DPC data was 53 min longer on average than the recorded operative time in the medical charts.

CONCLUSION

The DPC database had high validity for operative information.

Predicting hospitalisations related to ambulatory care sensitive conditions with machine learning for population health planning: derivation and validation cohort study

OBJECTIVE

To predict older adults' risk of avoidable hospitalisation related to ambulatory care sensitive conditions (ACSC) using machine learning applied to administrative health data of Ontario, Canada.

DESIGN, SETTING AND PARTICIPANTS

A retrospective cohort study was conducted on a large cohort of all residents covered under a single-payer system in Ontario, Canada over the period of 10 years (2008-2017). The study included 1.85 million Ontario residents between 65 and 74 years old at any time throughout the study period.

DATA SOURCES

Administrative health data from Ontario, Canada obtained from the (ICES formely known as the Institute for Clinical Evaluative Sciences Data Repository.

MAIN OUTCOME MEASURES

Risk of hospitalisations due to ACSCs 1 year after the observation period.

RESULTS

The study used a total of 1 854 116 patients, split into train, validation and test sets. The ACSC incidence rates among the data points were 1.1% for all sets. The final XGBoost model achieved an area under the receiver operating curve of 80.5% and an area under precision-recall curve of 0.093 on the test set, and the predictions were well calibrated, including in key subgroups. When ranking the model predictions, those at the top 5% of risk as predicted by the model captured 37.4% of those presented with an ACSC-related hospitalisation. A variety of features such as the previous number of ambulatory care visits, presence of ACSC-related hospitalisations during the observation window, age, rural residence and prescription of certain medications were contributors to the prediction. Our model was also able to capture the geospatial heterogeneity of ACSC risk in Ontario, and especially the elevated risk in rural and marginalised regions.

CONCLUSIONS

This study aimed to predict the 1-year risk of hospitalisation from ambulatory-care sensitive conditions in seniors aged 65-74 years old with a single, large-scale machine learning model. The model shows the potential to inform population health planning and interventions to reduce the burden of ACSC-related hospitalisations.

Development of Deep Learning Models for Predicting In-Hospital Mortality Using an Administrative Claims Database: Retrospective Cohort Study

BACKGROUND

Administrative claims databases have been used widely in studies because they have large sample sizes and are easily available. However, studies using administrative databases lack information on disease severity, so a risk adjustment method needs to be developed.

OBJECTIVE

We aimed to develop and validate deep learning-based prediction models for in-hospital mortality of acute care patients.

METHODS

The main model was developed using only administrative claims data (age, sex, diagnoses, and procedures on the day of admission). We also constructed disease-specific models for acute myocardial infarction, heart failure, stroke, and pneumonia using common severity indices for these diseases. Using the Japanese Diagnosis Procedure Combination data from July 2010 to March 2017, we identified 46,665,933 inpatients and divided them into derivation and validation cohorts in a ratio of 95:5. The main model was developed using a 9-layer deep neural network with 4 hidden dense layers that had 1000 nodes and were fully connected to adjacent layers. We evaluated model discrimination ability by an area under the receiver operating characteristic curve (AUC) and calibration ability by calibration plot.

RESULTS

Among the eligible patients, 2,005,035 (4.3%) died. Discrimination and calibration of the models were satisfactory. The AUC of the main model in the validation cohort was 0.954 (95% CI 0.954-0.955). The main model had higher discrimination ability than the disease-specific models.

CONCLUSIONS

Our deep learning-based model using diagnoses and procedures produced valid predictions of in-hospital mortality.

Midlife Cardiorespiratory Fitness and the Development of Peripheral Artery Disease in Later Life

Background Data are sparse on the prospective associations between physical activity and incidence of lower extremity peripheral artery disease (PAD). Methods and Results Linking participant data from the CCLS (Cooper Center Longitudinal Study) to Medicare claims files, we studied 19 023 participants with objectively measured midlife cardiorespiratory fitness through maximal effort on the Balke protocol who survived to receive Medicare coverage between 1999 and 2009. The study aimed to determine the association between midlife cardiorespiratory fitness and incident PAD with proportional hazards intensity models, adjusted for age, sex, body mass index, and other covariates, to PAD failure time data. During 121 288 person-years of Medicare follow-up, we observed 805 PAD-related hospitalizations/procedures among 19 023 participants (21% women, median age 50 years). Lower midlife fitness was associated with a higher rate of incident PAD in patients aged 65 years and older (low fit [quintile 1]: 11.4, moderate fit [quintile 2 to 3]: 7.8, and high fit [quintile 4 to 5]: 5.7 per 1000 person years). After multivariable adjustment for common predictors of incident PAD such as age, body mass index, hypertension, and diabetes, these findings persisted. Lower risk for PAD per greater metabolic equivalent task of fitness was observed (hazard ratio [HR], 0.93 [95% CI, 0.90-0.97]; P<0.001). Among a subset of patients with an additional fitness assessment, each 1 metabolic equivalent task increase from baseline fitness was associated with decreased risk of incident PAD (HR, 0.90 [95% CI, 0.82-0.99]; P=0.03). Conclusions Cardiorespiratory fitness in healthy, middle-aged adults is associated with lower risk of incident PAD in later life, independent of other predictors of incident PAD.

Validity of a model using routinely collected data for identifying infections following gastric, colorectal, and liver cancer surgeries

PURPOSE

Validating outcome measures is a prerequisite for using administrative databases for comparative effectiveness research. Although the Japanese Diagnosis Procedure Combination database is widely used in surgical studies, the outcome measure for postsurgical infection has not been validated. We developed a model to identify postsurgical infections using the routinely-collected Diagnosis Procedure Combination data.

METHODS

We retrospectively identified inpatients who underwent surgery for gastric, colorectal, or liver cancer between April 2016 and March 2018 at four hospitals. Chart reviews were conducted to identify postsurgical infections. We used bootstrap analysis with backwards variable elimination to select independent variables from routinely-collected diagnosis and procedure data. Selected variables were used to create a score predicting the chart review-identified infections, and the performance of the score was tested.

RESULTS

Among the 756 eligible patients, 102 patients (13%) had postoperative infections. Three variables were identified as predictors: diagnosis of infectious disease recorded as a complication arising after admission, addition of an intravenous antibiotic, and bacterial microscopy or culture. The prediction model had a C-statistic of 0.891 and pseudo-R² of 0.380. A cut-off of 1 point of the score showed a sensitivity of 92% and specificity of 71%, and a cut-off of 2 points showed a sensitivity of 77% and specificity of 91%.

CONCLUSIONS

Our model using routinely-collected administrative data accurately identified postoperative infections. Further external validation would lead to the application of the model for research using administrative databases. This article is protected by copyright. All rights reserved.

Characterization of the Early Years of Bevacizumab Use for First-Line Treatment of Ovarian Cancer in the United States

PURPOSE

To quantify early dissemination patterns, factors influencing use, and costs of bevacizumab (BEV) for the treatment of newly diagnosed ovarian cancer (OC) in the United States before its regulatory approval for this indication (off-label use).

METHODS

We identified women 18-65 years of age with newly diagnosed OC treated with surgery and platinum-based chemotherapy from 2008 to 2016 through the MarketScan database (N = 8,109). The proportion of women receiving BEV over time was calculated, multivariate logistic regression used to determine factors associated with BEV use, and total costs per cycle of chemotherapy with and without BEV abstracted.

RESULTS

BEV utilization rose 1.8-fold during the study period, from 4.1% (2008) to 7.4 % (2016). BEV was used with non-platinum/taxane regimens over a third of the time (37.2%). Physician specialty (medical oncology v gyn oncology) and geography (southeast region) were significantly associated with higher rates of use. Clinical factors associated with BEV use were metastatic disease and presence of ascites. The median cost of one cycle of platinum/taxane chemotherapy plus BEV was $10,897 in US dollars (USD) (interquartile range $7,573-$18,133 USD), compared with $1,629 USD (interquartile range, $683.0-$4,461 USD) for platinum/taxane alone.

CONCLUSION

Off-label use of BEV for newly diagnosed OC was rare (< 10%), but doubled following presentation of phase II and III data at international meetings. Both clinical (ascites, metastatic disease, and age) and nonclinical (specialty and region) factors were associated with BEV use, and its use was accompanied by a six-fold increase in the cost of one cycle of treatment.

Association of chronic pain with comorbidities and health care utilization: a retrospective cohort study using health administrative data

ABSTRACT

Health administrative data provide a potentially robust information source regarding the substantial burden chronic pain exerts on individuals and the health care system. This study aimed to use health administrative data to estimate comorbidity prevalence and annual health care utilization associated with chronic pain in Newfoundland and Labrador, Canada. Applying the validated Chronic Pain Algorithm to provincial Fee-for-Service Physician Claims File data (1999-2009) established the Chronic Pain (n = 184,580) and No Chronic Pain (n = 320,113) comparator groups. Applying the Canadian Chronic Disease Surveillance System coding algorithms to Claims File and Provincial Discharge Abstract Data (1999-2009) determined the prevalence of 16 comorbidities. The 2009/2010 risk and person-year rate of physician and diagnostic imaging visits and hospital admissions were calculated and adjusted using the robust Poisson model with log link function (risks) and negative binomial model (rates). Results indicated a significantly higher prevalence of all comorbidities and up to 4 times the odds of multimorbidity in the Chronic Pain Group (P-value < 0.001). Chronic Pain Group members accounted for 58.8% of all physician visits, 57.6% of all diagnostic imaging visits, and 54.2% of all hospital admissions in 2009/2010, but only 12% to 16% of these were for pain-related conditions as per recorded diagnostic codes. The Chronic Pain Group had significantly higher rates of physician visits and high-cost hospital admission/diagnostic imaging visits (P-value < 0.001) when adjusted for demographics and comorbidities. Observations made using this methodology supported that people identified as having chronic pain have higher prevalence of comorbidities and use significantly more publicly funded health services.

Claims-based algorithms for common chronic conditions were efficiently constructed using machine learning methods

Identification of medical conditions using claims data is generally conducted with algorithms based on subject-matter knowledge. However, these claims-based algorithms (CBAs) are highly dependent on the knowledge level and not necessarily optimized for target conditions. We investigated whether machine learning methods can supplement researchers' knowledge of target conditions in building CBAs. Retrospective cohort study using a claims database combined with annual health check-up results of employees' health insurance programs for fiscal year 2016-17 in Japan (study population for hypertension, N = 631,289; diabetes, N = 152,368; dyslipidemia, N = 614,434). We constructed CBAs with logistic regression, k-nearest neighbor, support vector machine, penalized logistic regression, tree-based model, and neural network for identifying patients with three common chronic conditions: hypertension, diabetes, and dyslipidemia. We then compared their association measures using a completely hold-out test set (25% of the study population). Among the test cohorts of 157,822, 38,092, and 153,608 enrollees for hypertension, diabetes, and dyslipidemia, 25.4%, 8.4%, and 38.7% of them had a diagnosis of the corresponding condition. The areas under the receiver operating characteristic curve (AUCs) of the logistic regression with/without subject-matter knowledge about the target condition were .923/.921 for hypertension, .957/.938 for diabetes, and .739/.747 for dyslipidemia. The logistic lasso, logistic elastic-net, and tree-based methods yielded AUCs comparable to those of the logistic regression with subject-matter knowledge: .923-.931 for hypertension; .958-.966 for diabetes; .747-.773 for dyslipidemia. We found that machine learning methods can attain AUCs comparable to the conventional knowledge-based method in building CBAs.

Evidence-Based Decision Making 6: Administrative Databases as Secondary Data Source for Epidemiologic and Health Service Research

Health-care systems require reliable information on which to base health-care planning and make decisions, as well as to evaluate their policy impact. Administrative data, predominantly captured for non-research purposes, provide important information about health services use, expenditures, and clinical outcomes and may be used to assess quality of care. With increased digitalization and accessibility of administrative databases, this data is more readily available for health service research purposes, aiding evidence-based decision making. This chapter discusses the utility of administrative data for population-based studies of health and health care.

A population-based study to develop juvenile arthritis case definitions for administrative health data using model-based dynamic classification

BACKGROUND

Previous research has shown that chronic disease case definitions constructed using population-based administrative health data may have low accuracy for ascertaining cases of episodic diseases such as rheumatoid arthritis, which are characterized by periods of good health followed by periods of illness. No studies have considered a dynamic approach that uses statistical (i.e., probability) models for repeated measures data to classify individuals into disease, non-disease, and indeterminate categories as an alternative to deterministic (i.e., non-probability) methods that use summary data for case ascertainment. The research objectives were to validate a model-based dynamic classification approach for ascertaining cases of juvenile arthritis (JA) from administrative data, and compare its performance with a deterministic approach for case ascertainment.

METHODS

The study cohort was comprised of JA cases and non-JA controls 16 years or younger identified from a pediatric clinical registry in the Canadian province of Manitoba and born between 1980 and 2002. Registry data were linked to hospital records and physician billing claims up to 2018. Longitudinal discriminant analysis (LoDA) models and dynamic classification were applied to annual healthcare utilization measures. The deterministic case definition was based on JA diagnoses in healthcare use data anytime between birth and age 16 years; it required one hospitalization ever or two physician visits. Case definitions based on model-based dynamic classification and deterministic approaches were assessed on sensitivity, specificity, and positive and negative predictive values (PPV, NPV). Mean time to classification was also measured for the former.

RESULTS

The cohort included 797 individuals; 386 (48.4 %) were JA cases. A model-based dynamic classification approach using an annual measure of any JA-related healthcare contact had sensitivity = 0.70 and PPV = 0.82. Mean classification time was 9.21 years. The deterministic case definition had sensitivity = 0.91 and PPV = 0.92.

CONCLUSIONS

A model-based dynamic classification approach had lower accuracy for ascertaining JA cases than a deterministic approach. However, the dynamic approach required a shorter duration of time to produce a case definition with acceptable PPV. The choice of methods to construct case definitions and their performance may depend on the characteristics of the chronic disease under investigation.

Integrating Data To Advance Research, Operations, And Client-Centered Services In California

The value of using administrative records for operational and evaluation purposes has been well established in health and human services. However, these records typically reflect the reach of a single government agency or program and fail to capture the experiences of individuals as they engage with different agencies or programs over time. Thus, the potential for these data to improve everyday operations, coordinate services, develop targeted interventions, and advance the science behind broader social policies has yet to be fully realized. A first step toward realizing that potential is to transition from an agency-centered to a client- or person-centered organization of data. We systematically linked tens of millions of records across California's largest health and human services programs. Our results underscore how the integration of records can help shift discussions from the programs that administer services to the people who are served.

Disparities in neonatal abstinence syndrome and health insurance status: A statewide study using non-claims real-time surveillance data

BACKGROUND

Most studies examining the relationship between neonatal abstinence syndrome (NAS) and health insurance status in the United States (USA) have used administrative insurance claims data, which is subject to myriad limitations. We examined the association between NAS and health insurance status in a large geographically defined rural population in the United States, using non-claims data.

METHODS

We utilized data from a population-based cohort of all newborns born in 2017-2019 in the rural state of West Virginia (WV) and restricted analyses to WV residents' births (n = 46 213). NAS was defined as neonatal withdrawal from many substances, including opiates and not limited to those cases that require pharmacological treatment.

RESULTS

Medicaid covered more than half (52.6%) of all infants' births in the state of WV. The incidence of NAS was 85.8 and 12.7 per 1000 livebirths in the Medicaid and privately insured groups, respectively. Among all infants diagnosed with NAS, 86.1% were enrolled in the state's Medicaid programme. The risk of NAS in the Medicaid-insured newborns was higher than privately insured newborns in the unadjusted analysis (risk ratio (RR) 6.76, 95% confidence interval (CI) 5.95, 7.68) and the adjusted analysis RR 3.00, 95% CI 2.01, 4.49); adjusted risk difference 20.3 (95% CI 17.5, 23.1 cases per 1000 livebirths).

CONCLUSIONS

NAS is an important indicator of the immediate effect of the opioid crisis. This study shows the disparity in NAS by health insurance status for a large rural population in the United States, and its burden on the state's Medicaid programme. Providing timely and accurate estimates of NAS is important for public health policies and decision making.

Descriptive epidemiology and health resource utilization for status epilepticus in the emergency department in the United States of America

OBJECTIVE

To describe the epidemiology and health resource utilization for convulsive status epilepticus (SE) in the emergency department (ED).

METHODS

Retrospective descriptive study in the Nationwide Emergency Department Sample (NEDS). Primary SE and secondary SE (SE in a case who visited the ED for other primary reason) were compared with non-SE seizures. Secondary SE is expected to have worse outcomes and higher costs because of another primary cause for ED visit.

RESULTS

In the period 2010-2014, there were 149,750 ED visits with primary SE; 83,459 ED with secondary SE; and 5,359,103 ED visits with non-SE seizures. On multivariable analysis adjusting for potential confounders, the odds of hospital admission were 7 times higher for primary SE than for non-SE seizures, and 5 times higher for secondary SE than for non-SE seizures; the odds of transfer to another hospital were 9 times higher for primary SE than for non-SE seizures, and 3 times higher for secondary SE than for non-SE seizures; the odds of death were 2.5 times higher for primary SE than for non-SE seizures, and 12 times higher for secondary SE than for non-SE seizures; and the charges (in January 2020 USA dollars) were $9000 higher in primary SE than in non-SE seizures, and $35,000 higher in secondary SE than in non-SE seizures.

CONCLUSION

Among all reasons for ED visits, SE, and in particular, secondary SE, are among the most resource-consuming conditions, being much more expensive than non-SE seizures in the ED.

Predicting adverse outcomes due to diabetes complications with machine learning using administrative health data

Across jurisdictions, government and health insurance providers hold a large amount of data from patient interactions with the healthcare system. We aimed to develop a machine learning-based model for predicting adverse outcomes due to diabetes complications using administrative health data from the single-payer health system in Ontario, Canada. A Gradient Boosting Decision Tree model was trained on data from 1,029,366 patients, validated on 272,864 patients, and tested on 265,406 patients. Discrimination was assessed using the AUC statistic and calibration was assessed visually using calibration plots overall and across population subgroups. Our model predicting three-year risk of adverse outcomes due to diabetes complications (hyper/hypoglycemia, tissue infection, retinopathy, cardiovascular events, amputation) included 700 features from multiple diverse data sources and had strong discrimination (average test AUC = 77.7, range 77.7-77.9). Through the design and validation of a high-performance model to predict diabetes complications adverse outcomes at the population level, we demonstrate the potential of machine learning and administrative health data to inform health planning and healthcare resource allocation for diabetes management.

Antipsychotic medication adherence and preventive diabetes screening in Medicaid enrollees with serious mental illness: an analysis of real-world administrative data

BACKGROUND

There is excess amenable mortality risk and evidence of healthcare quality deficits for persons with serious mental illness (SMI). We sought to identify sociodemographic and clinical characteristics associated with variations in two 2015 Healthcare Effectiveness Data and Information Set (HEDIS) measures, antipsychotic medication adherence and preventive diabetes screening, among Medicaid enrollees with serious mental illness (SMI).

METHODS

We retrospectively analyzed claims data from September 2014 to December 2015 from enrollees in a Medicaid specialty health plan in Florida. All plan enrollees had SMI; analyses included continuously enrolled adults with antipsychotic medication prescriptions and schizophrenia or bipolar disorder. Associations were identified using mixed effects logistic regression models.

RESULTS

Data for 5502 enrollees were analyzed. Substance use disorders, depression, and having both schizophrenia and bipolar disorder diagnoses were associated with both HEDIS measures but the direction of the associations differed; each was significantly associated with antipsychotic medication non-adherence (a marker of suboptimal care quality) but an increased likelihood of diabetes screening (a marker of quality care). Compared to whites, blacks and Hispanics had a significantly greater risk of medication non-adherence. Increasing age was significantly associated with increasing medication adherence, but the association between age and diabetes screening varied by sex. Other characteristics significantly associated with quality variations according to one or both measures were education (associated with antipsychotic medication adherence), urbanization (relative to urban locales, residing in suburban areas was associated with both adherence and diabetes screening), obesity (associated with both adherence and diabetes screening), language (non-English speakers had a greater likelihood of diabetes screening), and anxiety, asthma, and hypertension (each positively associated with diabetes screening).

CONCLUSIONS

The characteristics associated with variations in the quality of care provided to Medicaid enrollees with SMI as gauged by two HEDIS measures often differed, and at times associations were directionally opposite. The variations in the quality of healthcare received by persons with SMI that were identified in this study can guide quality improvement and delivery system reform efforts; however, given the sociodemographic and clinical characteristics' differing associations with different measures of care quality, multidimensional approaches are warranted.

Babesiosis Occurrence Among United States Medicare Beneficiaries, Ages 65 and Older, During 2006-2017: Overall and by State and County of Residence

Background

Human babesiosis is a mild-to-severe parasitic infection that poses health concerns especially in older and other at-risk populations. The study objective was to assess babesiosis occurrence among US Medicare beneficiaries, ages 65 and older, during 2006-2017.

Methods

Our retrospective claims-based study used Medicare databases. Babesiosis cases were identified using recorded diagnosis codes. The study estimated rates (per 100 000 beneficiary-years) overall, by year, diagnosis month, demographics, and state and county of residence.

Results

Nationwide, 19 469 beneficiaries had babesiosis recorded, at a rate of 6 per 100 000 person-years, ranging from 4 in 2006 to 9 in 2017 (P < .05). The highest babesiosis rates by state were in the following: Massachusetts (62), Rhode Island (61), Connecticut (51), New York (30), and New Jersey (19). The highest rates by county were in the following: Nantucket, Massachusetts (1089); Dukes, Massachusetts (236); Barnstable, Massachusetts (213); and Dutchess, New York (205). Increasing rates, from 2006 through 2017 (P < .05), were identified in multiple states, including states previously considered nonendemic. New Hampshire, Maine, Vermont, Pennsylvania, and Delaware saw rates increase by several times.

Conclusions

Our 12-year study shows substantially increasing babesiosis diagnosis trends, with highest rates in well established endemic states. It also suggests expansion of babesiosis infections in other states and highlights the utility of real-world evidence.

Identifying cases of chronic pain using health administrative data: A validation study

Background

Most prevalence estimates of chronic pain are derived from surveys and vary widely, both globally (2%–54%) and in Canada (6.5%–44%). Health administrative data are increasingly used for chronic disease surveillance, but their validity as a source to ascertain chronic pain cases is understudied.

Aim

The aim of this study was to derive and validate an algorithm to identify cases of chronic pain as a single chronic disease using provincial health administrative data.

Methods

A reference standard was developed and applied to the electronic medical records data of a Newfoundland and Labrador general population sample participating in the Canadian Primary Care Sentinel Surveillance Network. Chronic pain algorithms were created from the administrative data of patient populations with chronic pain, and their classification performance was compared to that of the reference standard via statistical tests of selection accuracy.

Results

The most performant algorithm for chronic pain case ascertainment from the Medical Care Plan Fee-for-Service Physicians Claims File was one anesthesiology encounter ever recording a chronic pain clinic procedure code OR five physician encounter dates recording any pain-related diagnostic code in 5 years with more than 183 days separating at least two encounters. The algorithm demonstrated 0.703 (95% confidence interval [CI], 0.685–0.722) sensitivity, 0.668 (95% CI, 0.657–0.678) specificity, and 0.408 (95% CI, 0.393–0.423) positive predictive value. The chronic pain algorithm selected 37.6% of a Newfoundland and Labrador provincial cohort.

Conclusions

A health administrative data algorithm was derived and validated to identify chronic pain cases and estimate disease burden in residents attending fee-for-service physician encounters in Newfoundland and Labrador.

Interferon gamma release assay tests are associated with persistence and completion of latent tuberculosis infection treatment in the United States: Evidence from commercial insurance data

Background

Risk-targeted testing and treatment of latent tuberculosis infection (LTBI) is a critical component of the United States’ (US) tuberculosis (TB) elimination strategy, but relatively low treatment completion rates remain a challenge. Both treatment persistence and completion may be facilitated by diagnosing LTBI using interferon gamma release assays (IGRA) rather than tuberculin skin tests (TST).

Methods

We used a national sample of administrative claims data to explore associations diagnostic test choice (TST, IGRA, TST with subsequent IGRA) and treatment persistence and completion in persons initiating a daily dose isoniazid LTBI treatment regimen in the US private healthcare sector between July 2011 and March 2014. Associations were analyzed with a generalized ordered logit model (completion) and a negative binomial regression model (persistence).

Results

Of 662 persons initiating treatment, 327 (49.4%) completed at least the 6-month regimen and 173 (26.1%) completed the 9-month regimen; 129 (19.5%) persisted in treatment one month or less. Six-month completion was least likely in persons receiving a TST (42.2%) relative to persons receiving an IGRA (55.0%) or TST then IGRA (67.2%; p = 0.001). Those receiving an IGRA or a TST followed by an IGRA had higher odds of completion compared to those receiving a TST (aOR = 1.59 and 2.50; p = 0.017 and 0.001, respectively). Receiving an IGRA or a TST and subsequent IGRA was associated with increased treatment persistence relative to TST (aIRR = 1.14 and 1.25; p = 0.027 and 0.009, respectively).

Conclusions

IGRA use is significantly associated with both higher levels of LTBI treatment completion and treatment persistence. These differences are apparent both when IGRAs alone were administered and when IGRAs were administered subsequent to a TST. Our results suggest that IGRAs contribute to more effective LTBI treatment and consequently individual and population protections against TB.

Utility of medical record diagnostic codes to ascertain attention-deficit/hyperactivity disorder and learning disabilities in populations of children

BACKGROUND

To develop and evaluate machine learning algorithms to ascertain attention-deficit/hyperactivity (ADHD) and learning disability (LD) using diagnostic codes in the medical record.

METHOD

Diagnoses of ADHD and LD were confirmed in cohorts of children in Olmsted County of Minnesota based on validated research criteria. Models to predict ADHD and LD were developed using ICD-9 codes in a derivation cohort of 1057 children before evaluated in a validation cohort of 536 children.

RESULTS

The ENET-MIN model using selected ICD-9 codes at prior probability of 0.25 has a sensitivity of 0.76, PPV of 0.85, specificity of 0.98, and NPV of 0.97 in the validation cohort. However, it does not offer significant advantage over a model using a single ICD-9 code of 314.X, which shows sensitivity of 0.81, PPV of 0.83, specificity of 0.98, and NPV of 0.97. None of the models developed for LD performed well in the validation cohort.

CONCLUSIONS

It is feasible to utilize diagnostic codes to ascertain cases of ADHD in a population of children. Machine learning approaches do not have advantage compared with simply using a single family of diagnostic codes for ADHD. The use of medical record diagnostic codes is not feasible to ascertain LD.

The effects of different lookback periods on the sociodemographic structure of the study population and on the estimation of incidence rates: analyses with German claims data

BACKGROUND

Defining incident cases has always been a challenging issue for researchers working with routine data. Lookback periods should enable researchers to identify and exclude recurrent cases and increase the accuracy of the incidence estimation. There are different recommendations for lookback periods depending on a disease entity of up to 10 years. Well-known drawbacks of the application of lookback periods are shorter remaining observation period in the dataset or smaller number of cases. The problem of selectivity of the remaining population after introducing lookback periods has not been considered in the literature until now.

METHODS

The analyses were performed with pseudonymized claims data of a German statutory health insurance fund with annual case numbers of about 2,1 million insured persons. Proportions of study population excluded due to the application of lookback periods are shown according to age, occupational qualification and income. Myocardial infarction and stroke were used to demonstrate changes in incidence rates after applying lookback periods of up to 5 years.

RESULTS

Younger individuals show substantial dropouts after the application of lookback periods. Furthermore, there are selectivities regarding occupational qualification and income, which cannot be handled by age standardization. Due to selective dropouts of younger individuals, crude incidence rates of myocardial infarction and stroke increase after applying lookback periods. Depending on the income group, age-standardized incidence rates changed differentially, leading to a decrease and possible underestimation of the social gradient after applying lookback periods.

CONCLUSIONS

Selectivity analyses regarding age and sociodemographic structure should be performed for the study population after applying lookback periods since the selectivity can affect the outcome especially in health care research. The selectivity effects might occur not only in claims data of one health insurance fund, but also in other longitudinal data with left- or right-censoring not covering the whole population. The effects may also apply to health care systems with a mix of public and private health insurance. A trade-off has to be considered between selectivity effects and eliminating recurrent events for more accuracy in the definition of incidence.

Identifying children with Cystic Fibrosis in population-scale routinely collected data in Wales: A Retrospective Review

INTRODUCTION

The challenges in identifying a cohort of people with a rare condition can be addressed by routinely collected, population-scale electronic health record (EHR) data, which provide large volumes of data at a national level. This paper describes the challenges of accurately identifying a cohort of children with Cystic Fibrosis (CF) using EHR and their validation against the UK CF Registry.

OBJECTIVES

To establish a proof of principle and provide insight into the merits of linked data in CF research; to identify the benefits of access to multiple data sources, in particular the UK CF Registry data, and to demonstrate the opportunity it represents as a resource for future CF research.

METHODS

Three EHR data sources were used to identify children with CF born in Wales between 1st January 1998 and 31st August 2015 within the Secure Anonymised Information Linkage (SAIL) Databank. The UK CF Registry was later acquired by SAIL and linked to the EHR cohort to validate the cases and explore the reasons for misclassifications.

RESULTS

We identified 352 children with CF in the three EHR data sources. This was greater than expected based on historical incidence rates in Wales. Subsequent validation using the UK CF Registry found that 257 (73%) of these were true cases. Approximately 98.7% (156/158) of individuals identified as CF cases in all three EHR data sources were confirmed as true cases; but this was only the case for 19.8% (20/101) of all those identified in just a single data source.

CONCLUSION

Identifying health conditions in EHR data can be challenging, so data quality assurance and validation is important or the merit of the research is undermined. This retrospective review identifies some of the challenges in identifying CF cases and demonstrates the benefits of linking cases across multiple data sources to improve quality.

The dawning of the digital era in the management of hypertension

Awareness, treatment, and control of hypertension are of the utmost importance in conquering stroke and cardiovascular disease. To reduce the global burden of hypertension, the Japanese Society of Hypertension (JSH) established the "JSH Future Plan" based on an increasing need to transform the strategy for combating hypertension. In addition to energizing conventional approaches in basic, translational, and clinical research, the application of rapidly evolving digital health technologies and artificial intelligence to hypertension healthcare and research (digital hypertension) holds promise for providing further insights into the pathophysiology and therapeutic targets and implementing predictive, personalized, and preemptive approaches in clinical practice. With great potential to revolutionize the landscape of hypertension, digital hypertension has some technical, legal, ethical, social, and financial issues to overcome. Given the multidisciplinary framework, digital hypertension requires comprehensive and strategic collaboration among industry, academia, and government to move forward toward the goal of "Future Medicine".

Identifying cardiovascular severe maternal morbidity in epidemiologic studies

BACKGROUND

Cardiovascular severe maternal morbidity (CSMM) is rising and has become the leading cause of maternal mortality. Research using administrative data sets may allow for better understanding of this critical group of diseases.

OBJECTIVE

To validate a composite variable of CSMM for use in epidemiologic studies.

METHODS

We analysed delivery hospitalisations at an obstetric teaching hospital from 2007 to 2017. We utilised a subset of indicators developed by the Centers for Disease Control and Prevention based on ICD codes to form the composite variable for CSMM. Two expert clinicians manually reviewed all qualifying events using a standardised tool to determine whether these represented true CSMM events. Additionally, we estimated the number of CSMM cases among delivery hospitalisations without qualifying ICD codes by manually reviewing all hospitalisations with severe preeclampsia, a population at high risk of CSMM, and a random sample of 1000 hospitalisations without severe preeclampsia. We estimated validity of the composite variable.

RESULTS

Among 91 355 admissions for delivery, we captured 113 potential CSMM cases using qualifying ICD codes. Of these, 65 (57.5%) were true CSMM cases. Indicators for acute myocardial infarction, cardiac arrest, and cardioversion had the highest true-positive rates (100% for all). We found an additional 70 CSMM cases in the 2102 admissions with severe preeclampsia and a single CSMM case in the random sample. Assuming a rate of 1 CSMM case per 1000 deliveries in the remaining cohort, the composite variable had a positive predictive value of 57.5% (95% CI 47,9, 66.8), a negative predictive value of 99.8% (95% CI 99.8, 99.9), a sensitivity of 29.0% (95% CI 23.2, 35.4), and a specificity of 100% (95% CI 99.9, 100.0).

CONCLUSION

A novel composite variable for CSMM had reasonable PPV but limited sensitivity. This composite variable may enable epidemiologic studies geared towards reducing maternal morbidity and mortality.

Developing a Regional Distributed Data Network for Surveillance of Chronic Health Conditions: The Colorado Health Observation Regional Data Service

Electronic health records (EHRs) provide an alternative to traditional public health surveillance surveys and administrative data for measuring the prevalence and impact of chronic health conditions in populations. As the infrastructure for secondary use of EHR data improves, many stakeholders are poised to benefit from data partnerships for regional access to information. Electronic health records can be transformed into a common data model that facilitates data sharing across multiple organizations and allows data to be used for surveillance. The Colorado Health Observation Regional Data Service, a regional distributed data network, has assembled diverse data partnerships, flexible infrastructure, and transparent governance practices to better understand the health of communities through EHR-based, public health surveillance. This article describes attributes of regional distributed data networks using EHR data and the history and design of Colorado Health Observation Regional Data Service as an emerging public health surveillance tool for chronic health conditions. Colorado Health Observation Regional Data Service and our experience may serve as a model for other regions interested in similar surveillance efforts. While benefits from EHR-based surveillance are described, a number of technology, partnership, and value proposition challenges remain.

Tuberculosis Prevention in the Private Sector: Using Claims-Based Methods to Identify and Evaluate Latent Tuberculosis Infection Treatment With Isoniazid Among the Commercially Insured

CONTEXT

Targeted identification and treatment of people with latent tuberculosis infection (LTBI) are key components of the US tuberculosis elimination strategy. Because of recent policy changes, some LTBI treatment may shift from public health departments to the private sector.

OBJECTIVES

To (1) develop methodology to estimate initiation and completion of treatment with isoniazid for LTBI using claims data, and (2) estimate treatment completion rates for isoniazid regimens from commercial insurance claims.

METHODS

Medical and pharmacy claims data representing insurance-paid services rendered and prescriptions filled between January 2011 and March 2015 were analyzed.

PARTICIPANTS

Four million commercially insured individuals 0 to 64 years of age.

MAIN OUTCOME MEASURES

Six-month and 9-month treatment completion rates for isoniazid LTBI regimens.

RESULTS

There was an annual isoniazid LTBI treatment initiation rate of 12.5/100 000 insured persons. Of 1074 unique courses of treatment with isoniazid for which treatment completion could be assessed, almost half (46.3%; confidence interval, 43.3-49.3) completed 6 or more months of therapy. Of those, approximately half (48.9%; confidence interval, 44.5-53.3) completed 9 months or more.

CONCLUSIONS

Claims data can be used to identify and evaluate LTBI treatment with isoniazid occurring in the commercial sector. Completion rates were in the range of those found in public health settings. These findings suggest that the commercial sector may be a valuable adjunct to more traditional venues for tuberculosis prevention. In addition, these newly developed claims-based methods offer a means to gain important insights and open new avenues to monitor, evaluate, and coordinate tuberculosis prevention.

Validation of the Use of Electronic Health Records for Classification of ADHD Status

OBJECTIVE

To validate an electronic health record (EHR)-based algorithm to classify ADHD status of pediatric patients.

METHOD

As part of an applied study, we identified all primary care patients of The Children's Hospital of Philadelphia [CHOP] health care network who were born 1987-1995 and residents of New Jersey. Patients were classified with ADHD if their EHR indicated an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis code of "314.x" at a clinical visit or on a list of known conditions. We manually reviewed EHRs for ADHD patients ( n = 2,030) and a random weighted sample of non-ADHD patients ( n = 807 of 13,579) to confirm the presence or absence of ADHD.

RESULTS

Depending on assumptions for inconclusive cases, sensitivity ranged from 0.96 to 0.97 (95% confidence interval [CI] = [0.95, 0.97]), specificity from 0.98 to 0.99 [0.97, 0.99], and positive predictive value from 0.83 to 0.98 [0.81, 0.99].

CONCLUSION

EHR-based diagnostic codes can accurately classify ADHD status among pediatric patients and can be used by large-scale epidemiologic and clinical studies with high sensitivity and specificity.

Using data linkage methodologies to augment healthcare-associated infection surveillance data

Background and objectives:

The landscape of antimicrobial resistance (AMR) surveillance is changing rapidly. The primary objective of this study was to assess the benefit of linking population-based infection prevention and control surveillance data on methicillin-resistant Staphylococcus aureus (MRSA) to hospital discharge abstract data (DAD). We assessed the value of this novel data linkage for the characterization of hospital-acquired (HA) and community-acquired MRSA (CA-MRSA) cases.

Methods:

Incident inpatient MRSA surveillance data for all adults (≥18 years) from 4 acute-care facilities in Calgary, Alberta, between April 1, 2011, and March 31, 2017, were linked to DAD. Personal health number (PHN) and gender were used to identify specific individuals, and specimen collection time-points were used to identify specific hospitalization records. A third common variable on admission date between these databases was used to validate the linkage process. Descriptive statistics were used to characterize HA-MRSA and CA-MRSA cases identified through the linkage process.

Results:

A total of 2,430 surveillance records (94.6%) were successfully linked to the correct hospitalization period. By linking surveillance and administrative data, we were able to identify key differences between patients with HA- and CA-MRSA. These differences are consistent with previously reported findings in the literature. Data linkage to DAD may be a novel tool to enhance and augment the details of base surveillance data.

Conclusion and recommendations:

This is the first Canadian study linking a frontline healthcare-associated infection AMR surveillance database to an administrative population database. This work represents an important methodological step toward complementing traditional AMR surveillance data practices. Data linkage to other data types, such as primary care, emergency, social, and biological data, may be the basis of achieving more precise data focused around AMR.

Examining postpartum depression screening effectiveness in well child clinics in Alberta, Canada: A study using the All Our Families cohort and administrative data

Affecting 10-15% of women, postpartum depression (PPD) can be debilitating and costly. While early identification has the potential to improve timely care, recommendations regarding the implementation of routine screening are inconsistent. In Alberta, screening is completed using the Edinburgh Postnatal Depression Scale during public health well child clinic visits. The objective of this study was to examine the effectiveness of screening in identifying, diagnosing and treating women at increased risk for PPD over the first year postpartum, compared to those unscreened. The All Our Families prospective pregnancy cohort was linked to public health, inpatient, outpatient, physician claims and community pharmaceutical data over the first year postpartum. Descriptive statistics and bivariate analyses examined differences in sample characteristics and PPD and non-PPD related utilization by screening category. Odds ratios and 95% confidence intervals for PPD diagnosis and mental health drugs dispensed were generated using crude and multivariable logistic regression models. Within our sample, 87% of the eligible population were screened, with 3% receiving a high-risk score, and 13% were unscreened. Compared to those unscreened, women screened high-risk had higher odds of being diagnosed with PPD (OR: 3.88, 95% CI: 2.18-6.92) and women screened low/moderate-risk had reduced odds of receiving a diagnosis (OR: 0.51, 95% CI: 0.35-0.74). High-risk women had an increased likelihood of diagnosis, higher PPD-related utilization and drugs dispensed compared to those unscreened. This information suggests that screening was effective at streamlining resources in Alberta. Future work should focus on evaluating the cost-effectiveness of PPD screening.

Changes in the health of mothers of children with neurodevelopmental disabilities: An administrative data study

BACKGROUND

Using linked administrative health data, this study compared the health and healthcare service utilization between mothers of children with and without neurodevelopmental disabilities (NDD), before, during, and after the birth of a child.

METHODS

The population (N = 25,388) was based on a cohort of children born in 2000 and who were, along with their mothers, continuously registered with the British Columbia's universal health insurance program between 2000 and 2007.

RESULTS

Compared to mothers of children without NDD, mothers of children with NDD were more likely to have chronic conditions and higher service utilization before child birth. Mothers of children with NDD showed a smaller increase in physician visits in the year before birth but a greater increase in different prescription drugs in the year after birth. There was no further divergence (or convergence) in health and service utilization between the groups in the 7-year period post-birth.

CONCLUSIONS

Differences in health and healthcare service utilization between mothers of children with and without NDD existed before the birth of the child and did not diverge in the 7 years post-birth. Replication of these findings is warranted as well as follow-up analyses examining longer term outcomes for mothers beyond 7 years post-birth.

Assessing the social determinants of health care costs for Medicaid-enrolled adolescents in Washington State using administrative data

OBJECTIVE

The study used administrative data to identify the social determinants that have the greatest impact on Medicaid expenditures in adolescence.

DATA SOURCES

Data were compiled using the Washington State Department of Social and Health Services Integrated Client Databases, which link data from state systems including Medicaid claims and social services receipt.

STUDY DESIGN

Medical system and behavioral health service costs of over 180 000 Medicaid-enrolled adolescents aged 12-17 were measured using integrated administrative data from Washington State. Social determinants of health, including child maltreatment and parent risk factors, were also measured. Two-stage regression models were used to identify factors associated with increased health care utilization and costs.

PRINCIPAL FINDINGS

Regression models revealed that the factors most predictive of higher health care costs were child abuse, child neglect, and instability in out-of-home placements related to foster care. Other social determinants of health, such as parent risk factors, were not associated with health care costs. Child maltreatment and placement instability impacted health care costs primarily through large increases in behavioral health utilization and costs.

CONCLUSIONS

Prevention and early interventions for children and families to decrease child maltreatment and increase foster care placement stability could reduce overall health care costs.

Individual- and area-level socioeconomic inequalities in diabetes mellitus in Saskatchewan between 2007 and 2012: a cross-sectional analysis

BACKGROUND

Improving our understanding of social inequalities may improve prevention and treatment efforts for diabetes mellitus. We examined the association between individual- and area-level socioeconomic measures and physician-diagnosed diabetes in Saskatchewan over time.

METHODS

In this cross-sectional study, we linked health administrative data with individual-level socioeconomic data from the Canadian Community Health Survey and area-level data from the 2006 Canadian census. We used general linear mixed-models regression to analyze the effect of each factor, controlling for geographic and demographic measures.

RESULTS

Area-level deprivation was associated with medically diagnosed type 2 diabetes mellitus after adjustment for the individual-level factors of age, sex, household income and education. Individuals residing in areas ranked in the least deprived quintile had a lower likelihood of diabetes than those in the most deprived quintile (odds ratio 0.40, 95% confidence interval 0.18-0.88). However, this disparity existed only in urban areas. This result may reflect less pronounced health inequalities in rural areas, greater socioeconomic heterogeneity, larger geographic units or some combination of these factors.

INTERPRETATION

Individual- and area-level socioeconomic factors were associated with the likelihood of medically diagnosed diabetes; however, the strength of this association varied between urban and rural communities. Acknowledgement of area-level deprivation as a modifiable risk factor related to the prevalence of diabetes is important in the development of effective interventions for urban, but not rural, areas.