Search

Background and Objectives: All publicly funded hospital discharges in Aotearoa New Zealand are recorded in the National Minimum Dataset (NMDS). Movement of patients between hospitals (and occasionally within the same hospital) results in separate records (discharge events) within the NMDS and if these consecutive health records are not accounted for hospitalization (encounters) rates might be overestimated. The aim of this study was to determine the impact of four different methods to bundle multiple discharge events in the NMDS into encounters on the relative comparison of rural and urban Ambulatory Sensitive Hospitalization (ASH) rates., Methods: NMDS discharge events with an admission date between July 1, 2015, and December 31, 2019, were bundled into encounters using either using a) no method, b) an "admission flag", c) a "discharge flag", or d) a date-based method. ASH incidence rate ratios (IRRs), the mean total length of stay and the percentage of interhospital transfers were estimated for each bundling method. These outcomes were compared across 4 categories of the Geographic Classification for Health., Results: Compared with no bundling, using the date-based method resulted in an 8.3% reduction (150 less hospitalizations per 100,000 person years) in the estimated incidence rate for ASH in the most rural (R2-3) regions. There was no difference in the interpretation of the rural-urban IRR for any bundling methodology. Length of stay was longer for all bundling methods used. For patients that live in the most rural regions, using a date-based method identified up to twice as many interhospital transfers (5.7% vs 12.4%) compared to using admission flags., Conclusion: Consecutive events within hospital discharge datasets should be bundled into encounters to estimate incidence. This reduces the overestimation of incidence rates and the undercounting of interhospital transfers and total length of stay., Competing Interests: Declaration of competing interest There are no competing interests for any author., (Copyright © 2024 The Author(s). Published by Elsevier Inc. All rights reserved.)

Objectives: The Charlson score is a commonly used measure of comorbidity; however, there is little empirical research into the optimal implementation when studying cancer surgery outcomes using administrative data. We compared four alternative Charlson score implementations, including and excluding metastatic cancer and varying the look-back periods. Study Design and Setting: Nine years of linked administrative data were used to identify patients undergoing surgery for cancer of the colon, rectum, or lung in New South Wales, Australia. Four binary outcomes of 30- and 365-day mortality, length of stay greater than 21 days, and emergency readmission within 28 days were compared between groups of similar hospitals. Hospital risk adjustment models were compared for alternative Charlson score implementations. Results: Excluding metastatic cancer from the Charlson score improved model performance for short-term outcomes, but there was no implementation that was consistently optimal. Incorporating a look-back period reduced the number of patients for analysis but did not improve hospital risk adjustment. Conclusion: Charlson scores for hospital risk adjustment of short-term outcomes of cancer surgery should be calculated excluding metastatic cancer as a separate comorbidity. We found no clear best performing implementation and found no benefit in incorporating any look-back period. [ABSTRACT FROM AUTHOR]

Objectives: To conduct a systematic review of studies reporting on the development or validation of comorbidity indices using administrative health data and compare their ability to predict outcomes related to comorbidity (ie, construct validity). Study Design and Setting: We conducted a comprehensive literature search of MEDLINE and EMBASE, until September 2012. After title and abstract screen, relevant articles were selected for review by two independent investigators. Predictive validity and model fit were measured using c-statistic for dichotomous outcomes and R² for continuous outcomes. Results: Our review includes 76 articles. Two categories of comorbidity indices were identified: those identifying comorbidities based on diagnoses, using International Classification of Disease codes from hospitalization or outpatient data, and based on medications, using pharmacy data. The ability of indices studied to predict morbidity-related outcomes ranged from poor (C statistic ≤0.69) to excellent (C statistic >0.80) depending on the specific index, outcome measured, and study population. Diagnosis-based measures, particularly the Elixhauser Index and the Romano adaptation of the Charlson Index, resulted in higher ability to predict mortality outcomes. Medication-based indices, such as the Chronic Disease Score, demonstrated better performance for predicting health care utilization. Conclusion: A number of valid comorbidity indices derived from administrative data are available. Selection of an appropriate index should take into account the type of data available, study population, and specific outcome of interest. © 2015 Elsevier Inc. All rights reserved. [ABSTRACT FROM AUTHOR]

Objectives: To develop algorithms of locating patients with primary aldosteronism (PA) using insurance reimbursement data and to validate the algorithms using medical charts. Study Design and Setting: We extracted National Health Insurance (NHI) reimbursement data and medical charts in seven enrolled hospitals and analyzed diagnosis-related information for 1999-2010. The NHI codes PA as 255.1x, using the International Classification of Diseases, Ninth Revision, Clinical Modification. Confirmation of PA was based on suppression tests. Results: We reviewed medical charts for 1,094 cases with at least one PA diagnosis. PA was confirmed for 563 cases. Compared with patients with essential hypertension, PA patients had higher systolic blood pressure, higher aldosterone, lower renin activity, and lower potassium level (all P-values <0.05). An algorithm based on PA diagnosis reported in at least one hospital stay or three outpatient visits had modest performance (sensitivity 5 0.94 and specificity 5 0.20). The best additional condition for the algorithm was use of mineralocorticoid receptor antagonist (MRA; sensitivity 5 0.89 and specificity 5 0.88). Conclusion: Using information on PA diagnosis and MRA prescription reported in insurance claims data can precisely locate PA patients in high-risk groups. This algorithm can construct a reliable PA sample for conducting research in various fields, including epidemiology and clinical practice. [ABSTRACT FROM AUTHOR]

Objective: Prognostication is difficult in a diverse patient population or when outcomes depend on multiple factors. This study derived and internally validated a model to predict risk of death from any cause within 1 year of admission to hospital. Study Design and Setting: The study included all adult Ontarians admitted to nonpsychiatric hospital services in 2011 (n = 640,022) and deterministically linked administrative data to identify 20 patient and admission factors. A split-sample approach was used to derive and internally validate the model. Results: A total of 75,082 people (11.7%) died within 1 year of admission to hospital. The final model included one dozen patient factors (age, sex, living status, comorbidities, home oxygen status, and number of emergency room visits and hospital admissions by ambulance in previous year) and hospitalization factors (admission service and urgency, admission to intensive care unit, whether current hospitalization was a readmission, and admission diagnostic risk score). The model in the validation cohort was highly discriminative (c-statistic 92.3), well calibrated, and used to create the Hospital-patient One-year Mortality Risk score that accurately predicted 1-year risk of death. Conclusion: Routinely collected administrative data can be used to accurately predict 1-year death risk in adults admitted to nonpsychiatric hospital services. [ABSTRACT FROM AUTHOR]

Objectives: Understanding patterns of maternity care requires knowing which women have given birth previously, but this information is typically unavailable in administrative hospital data sets. We assessed how well parity can be derived using linked historical records. Study Design and Setting: Using Hospital Episode Statistics data, we identified records of women who gave birth between April 2009 and March 2010 in English National Health Service hospitals. The parity coded in these records was compared with an estimate derived from deliveries identified in previous hospital admissions between April 2000 and March 2009. Results: We identified 358,849 eligible deliveries with complete parity data in the 2009e10 birth records. The historical data classified 168,041 women as multiparous; of whom, 98% were coded as multiparous in their birth record. Among 190,798 women classified as primiparous using historical data, 72% were coded as primiparous in their birth record. The proportion of accurate predictions about primiparous status from historical data varied with age, ranging from 89% for 15e18 year olds to 50% for women aged more than 35 years. Conclusion: Historical records in administrative hospital data sets give accurate information on multiparous status of women. There is some misclassification of primiparous status, and error rates differ among subgroups of women. [ABSTRACT FROM AUTHOR]

Objective: To determine the validity of the diagnostic algorithms for osteoporosis and fractures in administrative data. Study Design and Setting: A systematic search was conducted to identify studies that reported the validity of a diagnostic algorithm for osteoporosis and/or fractures using administrative data. Results: Twelve studies were reviewed. The validity of the diagnosis of osteoporosis in administrative data was fair when at least 3 years of data from hospital and physician visit claims were used (area under the receiver operating characteristic [ROC] curve [AUC]50.70) or when pharmacy data were used (with or without the use of hospital and physician visit claims data, AUCO0.70). Nonetheless, the positive predictive values (PPVs) were low (!0.60). There was good evidence to support the use of hospital data to identify hip fractures (sensitivity: 69e97%; PPV: 63e96%) and the addition of physician claims diagnostic and procedural codes to hospitalization diagnostic codes improved these characteristics (sensitivity: 83e97%; PPV: 86e98%). Vertebral fractures were difficult to identify using administrative data. There was some evidence to support the use of administrative data to define other fractures that do not require hospitalization. Conclusions: Administrative data can be used to identify hip fractures. Existing diagnostic algorithms to identify osteoporosis and vertebral fractures in administrative data are suboptimal. [ABSTRACT FROM AUTHOR]

Abstract: Objective: Most of the urgent readmissions are unavoidable. This study developed a method that used observed urgent readmission rates to compare the latent avoidable readmission rates between the two hospitals. Study Design and Setting: To compare two hospitals, we identified all proportions of urgent readmissions deemed avoidable at each hospital making their avoidable readmission rates significantly different. We then calculated the probability that any of these conditions occurred. We applied this method to 25 randomly selected Ontario acute-care hospitals in 2008. Results: The hospitals had a median 30-day urgent readmission rate of 10.8% (interquartile [IQR] 9.7–12.8%). The median P-value of the 300 hospital–hospital comparisons for 30-day urgent readmission rate was 0.05 (interquartile range [IQR] 0.0005–0.31). In contrast, the median probability that hospitals with the lower urgent 30-day readmission rate outperformed their comparator hospital with respect to avoidable readmissions was only 0.161 (IQR 0.079–0.274). Conclusion: Urgent readmission rates can be used to estimate the probability that avoidable readmission rates differ significantly between the two hospitals. The probability that avoidable readmission rates differ significantly between hospitals is small even when significant differences in urgent 30-day readmission rates exist. Our results show that 30-day urgent readmission rates should be used very cautiously to compare hospital quality of care. [Copyright &y& Elsevier]

Abstract: Objective: The Charlson and Elixhauser indices are the most commonly used comorbidity indices with risk prediction models using administrative data. Our objective was to compare the original Charlson index, a modified set of Charlson codes after advice from clinical coders, and a published modified Elixhauser index in predicting in-hospital mortality. Study Design and Setting: Logistic regression using two separate years of administrative hospital data for all acute nonspecialist public hospitals in England. Results: For all admissions combined, discrimination was similar for the Charlson index using the original codes and weights and the Charlson index using the original codes but England-calibrated weights (c =0.73), although model fit was superior for the latter. The new Charlson codes improved discrimination (c =0.76), model fit, and consistency of recording between admissions. The modified Elixhauser had the best performance (c =0.80). For admissions for acute myocardial infarction and chronic obstructive pulmonary disease, the weights often differed, although the patterns were broadly similar. Conclusion: Recalibration of the original Charlson index yielded only modest benefits overall. The modified Charlson codes and weights offer better fit and discrimination for English data over the original version. The modified Elixhauser performed best of all, but its weights were perhaps less consistent across the different patient groups considered here. [Copyright &y& Elsevier]

Objective: To determine the positive predictive value of Medicare claims for identifying revision of total hip replacement (THR), a frequent marker of THR quality and outcome.Study Design and Setting: We obtained Medicare Part A (Hospital) claims from seven states on patients that had primary THR from July 1995 through June 1996. We searched claims to determine whether these THR recipients had a subsequent revision THR through December 2006. We selected a sample of subjects with codes indicating both index primary and subsequent revision THR. We obtained medical records for both procedures to establish whether the revision occurred on the same side as index primary THR.Results: Three hundred seventy-four subjects had codes indicating primary THR in 1995-96 and subsequent revision. Seventy-one percent (95% confidence interval: 66, 76) of the revisions were performed on the index joint and would be correctly attributed as revisions of the index THR, using Medicare claims data.Conclusion: Claims data on revision THR that do not contain information on the side that was operated on are ambiguous with respect to whether the revision was performed on the index or contralateral side. Claims-based analyses of revisions after an index THR should acknowledge and adjust for this source of potential misclassification. [ABSTRACT FROM AUTHOR]

Abstract: Objective: With the increasing use of electronic medical records (EMRs) comes the potential to efficiently evaluate and improve quality of care. We set out to determine if diabetics could be accurately identified using structured data contained within an EMR. Study Design and Setting: We used a 5% random sample of adult patients (969 patients) within a convenience sample of 17 primary care physicians using Practices Solutions EMR in Ontario. A reference standard of diabetes status was manually confirmed by reviewing each patient''s record. Accuracy for identifying people with diabetes was assessed using various combinations of laboratory tests and prescriptions. EMR data was also compared with administrative data. Results: A rule of one elevated blood sugar or a prescription for an antidiabetic medication had a 83.1% sensitivity, 98.2% specificity, 80.0% positive predictive value (PPV) and 98.5% negative predictive value (NPV) compared with the reference standard of diabetes status. Conclusion: We found that the use of structured data within an EMR could be used to identify patients with diabetes. Our results have positive implications for policy makers, researchers, and clinicians as they develop registries of diabetic patients to examine quality of care using EMR data. [Copyright &y& Elsevier]

Objective: To determine the magnitude and importance of declines in model performance associated with altering the data source and time frame from which comorbid conditions were identified in claims-based risk adjustment among persons with hip fracture.Study Design and Setting: Medicare claims data were used to identify incident hip fracture cases in 1999. Three risk-adjustment instruments were evaluated: one by Iezzoni, the Charlson Index (Romano adaptation), and the Clinical Classification Software (CCS). Several implementation strategies, defined by altering data source (MedPar and/or Part B claims) and time frame (index hospitalization and/or 1-year preperiod), were assessed for each instrument. Logistic regression was used to predict 1-year mortality, and model performance was compared.Results: Each instrument had modest ability to predict 1-year mortality after hip fracture. The CCS performed best overall (c=0.76), followed by the Iezzoni (c=0.73) and Charlson models (c=0.72). Although each instrument performed most favorably when applied to both inpatient and outpatient claims and when comorbidities were considered during the preperiod, varying data source and time frame had trivial effects on model performance.Conclusion: The similar predictive ability of the three risk-adjustment instruments suggests that ease of implementation be a key consideration in choosing an approach for hip fracture populations. [ABSTRACT FROM AUTHOR]

Objective: Secondary databases are used in descriptive studies of patient subgroups; evaluation of associations between individual characteristics and diagnosis, prognosis, and/or service utilization rates; and studies of the quality of health care delivered. This article identifies sources of bias for health state characteristics stored in secondary databases that arise from patients' encounters with health systems, highlighting sources of bias that arise from organizational and environmental factors.Study Design and Setting: Potential sources of bias, from patient access of services and diagnosis, through encoding and filing of patient information in secondary databases, are discussed. A patient presenting with acute myocardial infarction is used as an illustrative example.Results: The accuracy of health state characteristics derived from secondary databases is a function of both the quality and quantity of information collected before data entry and is dependent on complex interactions between patients, clinicians, and the structures and systems surrounding them.Conclusion: The use of health state information included in secondary databases requires that estimates of potential bias from all sources be included in the analysis and presentation of results. By making this common practice in the field, greater value can be achieved from secondary database analyses. [ABSTRACT FROM AUTHOR]

Abstract: Background and Objectives: Adjustment for comorbidity is an essential component of any observational study comparing outcomes. We evaluated the validity of the Charlson comorbidity score based on ICD-10 codes in patients undergoing urological cancer surgery within an English administrative database. Study Design and Setting: Patients who underwent radical urological cancer surgery between 1998 and 2002 in the English National Health Service were identified from the Hospital Episode Statistics database (N = 20,138). ICD-9-CM codes defining comorbid diseases according to the Deyo and Dartmouth–Manitoba adaptations of the Charlson comorbidity score were translated into ICD-10 codes. Results: Charlson scores derived by the ICD-10 translation of the Deyo and Dartmouth–Manitoba adaptations were identical in 16,623 patients (83%; κ = .63). For both adaptations, ICD-10 scores increased with age, were higher in patients admitted on an emergency basis, and predicted short-term outcome. Addition of either the ICD-10 Charlson Deyo or Dartmouth–Manitoba score to risk models containing age and sex to predict in-hospital mortality resulted in a better model fit but only in small improvements of the predictive power. Conclusion: The ICD-10 translations of the Deyo and Dartmouth–Manitoba adaptations performed similarly in risk models predicting hospital mortality following urological cancer surgery. Adjustment for comorbidity over and above age and sex alone does not seem to provide a large improvement. [Copyright &y& Elsevier]

Abstract: Background and Objectives: To develop a comorbidity scoring system that out-performs the Charlson index. Methods: Population-based cohorts of medical (n =326,456), procedural (n =349,686), and psychiatric (n =16,895) inpatients in Western Australia were followed for 1-year mortality, 30-day readmissions, and length of stay (LOS) using data linkage. Conditions were identified at index admission and over the preceding 12 months. A Multipurpose Australian Comorbidity Scoring System (MACSS) was developed, based on the most frequent 102 comorbid conditions associated with a rate ratio (RR) ≥ 1.1 of death or readmission or a LOS difference ≥0.5 days. The performance of MACSS and the Charlson index in predicting mortality, readmission, and LOS, and in controlling confounding by comorbidity, was compared in five test scenarios involving asthma, myocardial infarction, mastectomy, transurethral prostatectomy, and major depressive illness. Results: MACSS performed better than the Charlson index on all three outcomes in all five clinical groups. It reduced the failure of the Charlson index to discriminate on mortality and readmission outcomes by 5–40%, improved R 2 in LOS models by up to fourfold and often doubled the correction of originally confounded effect measures. Conclusion: The use of the MACSS and similar alternatives to the Charlson index are a new methodologic standard for adjustment of comorbidity risk. [Copyright &y& Elsevier]

Abstract: Background and objective: The accuracy of comorbidity data within the Western Australian Data Linkage System was evaluated by means of comparison with hospital charts and a general practitioner (GP) survey. Methods: Patients (n=2,037) with a hospital admission from 1991 to 1996 were selected. Linked data were extracted for 100 comorbidities, categorized into 16 diagnostic chapters, for each hospital admission within a 5-year period. Clinical chart review and a GP survey were performed. Comorbidity occurrence in each data source and false-positive and false-negative diagnoses were ascertained. Results: Administrative data contained 45.5% of comorbidity recorded in hospital charts and underascertained secondary conditions for all 16 diagnostic chapters. False-positive diagnoses were low for most conditions (range: 0–1.5%); however, a high occurrence of false negatives existed for all comorbidity chapters (range: 16.3–91.3%). GP-identified comorbidity was 20.0% greater than that found using administrative data but, with the exceptions of injury–poisoning and cutaneous–subcutaneous diseases, was less (42.0%) than that observed from hospital charts. Conclusion: Our results indicate that when accurate comorbidity data are crucial to health outcome research, hospital chart review (as opposed to using administrative data) may be required. Furthermore, surveying GPs, at least in Australia, appears an unsatisfactory alternative to hospital charts for obtaining retrospective comorbidity information. [Copyright &y& Elsevier]

Abstract: Background and objective: The ICD-9-CM adaptation of the Charlson comorbidity score has been a valuable resource for health services researchers. With the transition into ICD-10 coding worldwide, an ICD-10 version of the Deyo adaptation was developed and validated using population-based hospital data from Victoria, Australia. Methods: The algorithm was translated from ICD-9-CM into ICD-10-AM (Australian modification) in a multistep process. After a mapping algorithm was used to develop an initial translation, these codes were manually examined by the coding experts and a general physician for face validity. Because the ICD-10 system is country specific, our goal was to keep many of the translated code at the three-digit level for generalizability of the new index. Results: There appears to be little difference in the distribution of the Charlson Index score between the two versions. A strong association between increasing index scores and mortality exists: the area under the ROC curve is 0.865 for the last year using the ICD-9-CM version and remains high, at 0.855, for the ICD-10 version. Conclusion: This work represents the first rigorous adaptation of the Charlson comorbidity index for use with ICD-10 data. In comparison with a well-established ICD-9-CM coding algorithm, it yields closely similar prevalence and prognosis information by comorbidity category. [Copyright &y& Elsevier]

Objective: There is little data regarding the accuracy of pregnancy-specific ICD-9-CM codes used to identify patients with venous thromboembolism (VTE).Study design and setting: We identified a large cohort of pregnant patients in whom there were one or more pregnancy-specific (600 codes) or standard ICD-9-CM codes (400 codes) for VTE. Charts of these cases were abstracted to determine the presence of objectively documented VTE.Results: A total of 214 cases had a code for VTE either during pregnancy or the 6-week postpartum period; 82% had a pregnancy-specific code and 18% a standard code. Overall, 84 (39%, 95% CI = 33–46%) had objectively documented VTE. A pregnancy-specific ICD-9-CM for VTE had a positive predictive value (PPV) of 54/174 = 31% (95% CI = 24–38%), whereas standard VTE codes had a PPV of 30 of 38 = 80% (95% CI = 63–99%). A PPV in the range of 95–100% could be attained using other criteria, at the expense of detecting only 28 to 45% of all VTE cases.Conclusions: Pregnancy-specific ICD-9-CM codes for VTE have low PPV. Other criteria must be applied to select cases with a high probability of having objectively documented VTE. [Copyright &y& Elsevier]

Objective: To evaluate trends in prognosis after acute myocardial infarction (AMI) between Denmark and Sweden using routinely collected data and different case–fatality measures.Study Design and Setting: We compared three case–fatality measures during 1987–1999 using national registries in Denmark and Sweden, and extended these measures with underlying deaths of ischemic heart disease and sudden deaths of unknown cause.Results: Changed coding practice distorted trends of case fatality rates during the day of the event. In general, Denmark had higher case–fatality rates, but trends in hospital-based rates were very similar, except for men 35–64 years old; Denmark declined more steeply. Short- and long-term prognosis improved considerably: the odds ratios for case fatality during days 1–28 for 1999 vs. 1987 were 0.48 among men in Denmark (women 0.58) and 0.53 among men in Sweden (women 0.55) and the odds ratios for case fatality during days 29–365 for 1999 vs. 1987 were 0.56 among men in Denmark (women 0.65) and 0.66 among men in Sweden (women 0.67).Conclusion: Short- and long-term prognosis improved considerably during 1987–1999 in Denmark and Sweden. Case fatality during the day of the event is epidemiologically important, but less certain than case–fatality measures defined after the day of the event when comparing countries. [Copyright &y& Elsevier]

: Background and ObjectivesThe purpose of this study was to develop a comorbidity index specific to Black Men with prostate cancer, because certain comorbidities and prostate cancer are particularly prevalent among this racial group.: MethodsThis research used the Surveillance, Epidemiology, and End Results (SEER)–Medicare-linked database to develop an index of comorbidity burden based on survival, and the presence/absence of comorbid illness in 2,931 Black males diagnosed with prostate cancer. Comorbidity burden was recognized using inpatient, outpatient, and physician claims for a 2-year period prior to the diagnosis of prostate cancer. We compared five different statistical models, each with two-way, three-way, and/or four-way interactions among the comorbidities, and selected the model with only two-way interactions as the optimal choice. We demonstrated the utility of refining the simplest model, with 27 comorbidity categories only, by adjusting for the number of different diagnoses within statistically significant categories. [Copyright &y& Elsevier]

This analysis was performed to examine whether Medicare claims accurately document underlying rheumatologic diagnoses in total hip replacement (THR) recipients. We obtained data on rheumatologic diagnoses including rheumatoid arthritis (RA), avascular necrosis (AVN), and osteoarthritis (OA) from medical records and from Medicare claims data. To examine the accuracy of claims data we calculated sensitivity and positive predictive value using medical records data as the “gold standard” and assessed bias due to misclassification of claims-based diagnoses. The sensitivities of claims-based diagnoses of RA, AVN, and OA were 0.65, 0.54, and 0.96, respectively; the positive predictive values were all in the 0.86–0.89 range. The sensitivities of RA and AVN varied substantially across hospital volume strata, but in different directions for the two diagnoses. We conclude that inaccuracies in claims coding of diagnoses are frequent, and are potential sources of bias. More studies are needed to examine the magnitude and direction of bias in health outcomes research due to inaccuracy of claims coding for specific diagnoses. [Copyright &y& Elsevier]

The problem of missing data is frequently encountered in observational studies. We compared approaches to dealing with missing data. Three multiple imputation methods were compared with a method of enhancing a clinical database through merging with administrative data. The clinical database used for comparison contained information collected from 6,065 cardiac care patients in 1995 in the province of Alberta, Canada. The effectiveness of the different strategies was evaluated using measures of discrimination and goodness of fit for the 1995 data. The strategies were further evaluated by examining how well the models predicted outcomes in data collected from patients in 1996. In general, the different methods produced similar results, with one of the multiple imputation methods demonstrating a slight advantage. It is concluded that the choice of missing data strategy should be guided by statistical expertise and data resources. [Copyright &y& Elsevier]

Objective: We evaluated reporting completeness and transparency in randomized controlled trials (RCTs) conducted using administrative data based on 2021 CONSORT Extension for Trials Conducted Using Cohorts and Routinely Collected Data (CONSORT-ROUTINE) criteria., Study Design and Setting: MEDLINE and the Cochrane Methodology Register were searched (2011 and 2018). Eligible RCTs used administrative databases for identifying eligible participants or collecting outcomes. We evaluated reporting based on CONSORT-ROUTINE, which modified eight items from CONSORT 2010 and added five new items., Results: Of 33 included trials (76% used administrative databases for outcomes, 3% for identifying participants, 21% both), most were conducted in the United States (55%), Canada (18%), or the United Kingdom (12%). Of eight items modified in the extension; six were adequately reported in a majority (>50%) of trials. For the CONSORT-ROUTINE modification portion of those items, three items were reported adequately in >50% of trials, two in <50%, two only applied to some trials, and one only had wording modifications and was not evaluated. For five new items, four that address use of routine data in trials were reported inadequately in most trials., Conclusion: How administrative data are used in trials is often sub-optimally reported. CONSORT-ROUTINE uptake may improve reporting., Competing Interests: Declaration of Competing Interests All authors have completed the ICJME uniform disclosure form and declare: no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years, (Copyright © 2021. Published by Elsevier Inc.)

Objective: The objective of the study was to compare multimorbidity prevalence using self-reported and administrative data and identify factors associated with agreement between data sources., Study Design and Setting: Self-reported cross-sectional data from four Canadian Community Health Survey waves were linked to administrative data in Ontario, Canada. Multimorbidity prevalence was examined using two definitions, 2+ and 3+ chronic conditions (CCs). Agreement between data sources was assessed using Kappa and Phi statistics. Logistic regression was used to estimate associations between agreement and sociodemographic, health behavior, and health status variables for each multimorbidity definition., Results: Regardless of multimorbidity definition, prevalence was higher using administrative data (2+ CCs: 55.5% vs. 47.1%; 3+ CCs: 30.0% vs. 24.2%). Agreement between data sources was moderate (2+ CCs K = 0.482; 3+ CCs K = 0.442), and while associated with sociodemographic, health behavior, and health status factors, the magnitude and sometimes direction of association differed by multimorbidity definition., Conclusion: A better understanding is needed of what factors influence individuals' reporting of CCs and how they align with what is in administrative data as policy makers need a solid evidence base on which to make decisions for health planning. Our results suggest that data sources may need to be triangulated to provide accurate estimates of multimorbidity for health services planning and policy., (Copyright © 2020 Elsevier Inc. All rights reserved.)

Objectives: Although claims data are widely used in medical research, their ability to identify persons' health-related conditions has not been fully justified. We assessed the validity of claims-based algorithms (CBAs) for identifying people with common chronic conditions in a large population using annual health screening results as the gold standard., Study Design and Setting: Using a longitudinal claims database (n = 523,267) combined with annual health screening results, we defined the people with hypertension, diabetes, and/or dyslipidemia by applying health screening results as their gold standard and compared them against various CBAs., Results: By using diagnostic and medication code-based CBAs, sensitivity and specificity were 74.5% (95% confidence interval [CI], 74.2%-74.8%) and 98.2% (98.2%-98.3%) for hypertension, 78.6% (77.3%-79.8%) and 99.6% (99.5%-99.6%) for diabetes, and 34.5% (34.2%-34.7%) and 97.2% (97.2%-97.3%) for dyslipidemia, respectively. Sensitivity did not decrease substantially for hypertension (65.2% [95% CI, 64.9%-65.5%]) and diabetes (73.0% [71.7%-74.2%]) when we used the same CBAs without limiting to primary care settings., Conclusion: We used regularly collected data to obtain CBA association measures, which are applicable to a wide range of populations. Our framework can be a basis of the validity assessment of CBAs for identifying persons' health-related conditions with regularly collected data., (Copyright © 2018 Elsevier Inc. All rights reserved.)

Background and Objectives: Algorithms are often used to improve identification of Aboriginal Australians in linked data sets with inconsistent and incomplete recording of Aboriginal status. We compared how consistently some common algorithms identified family members, developed a new algorithm incorporating relatives' information, and assessed the effects of these algorithms on health estimates., Methods: The sample was people born 1980-2011 recorded as Aboriginal at least once (or a relative) in four Western Australian data sets and their relatives (N = 156,407). A very inclusive approach, ever-Aboriginal (EA/EA+, where + denotes children's records incorporated), and two more specific approaches, multistage median (MSM/MSM+) and last record (LR/LR+), were chosen, along with the new algorithm (MSM+Family)., Results: Ever-Aboriginal (EA) categorized relatives the least consistently; 25% of parent-child triads had incongruent Aboriginal statuses with EA+, compared with only 9% with MSM+. With EA+, 14% of full siblings had different statuses compared with 8% for MSM+. EA produced the lowest estimates of the proportion of Aboriginal people with poor health outcomes. Using relatives' records reduced the number of uncategorized people and categorized as Aboriginal more people who had few records (e.g., no hospital admissions)., Conclusion: When many data sets are linked, more specific algorithms select more representative Aboriginal samples and identify Aboriginality of relatives more consistently., (Copyright © 2017 Elsevier Inc. All rights reserved.)

Objectives: Previous mathematical analyses of diagnostic tests based on the categorization of a continuous measure have found that test sensitivity and specificity varies significantly by disease prevalence. This study determined if the accuracy of diagnostic codes varied by disease prevalence., Study Design and Setting: We used data from two previous studies in which the true status of renal disease and primary subarachnoid hemorrhage, respectively, had been determined. In multiple stratified random samples from the two previous studies having varying disease prevalence, we measured the accuracy of diagnostic codes for each disease using sensitivity, specificity, and positive and negative predictive value., Results: Diagnostic code sensitivity and specificity did not change notably within clinically sensible disease prevalence. In contrast, positive and negative predictive values changed significantly with disease prevalence., Conclusion: Disease prevalence had no important influence on the sensitivity and specificity of diagnostic codes in administrative databases., (Copyright © 2016 Elsevier Inc. All rights reserved.)

Objective: To compare confounding adjustment by high-dimensional propensity scores (hdPSs) and historically developed high-dimensional disease risk scores (hdDRSs) in three comparative study examples of newly marketed medications: (1) dabigatran vs. warfarin on major hemorrhage; (2) on death; and (3) cyclooxygenase-2 inhibitors vs. nonselective nonsteroidal anti-inflammatory drugs on gastrointestinal bleeds., Study Design and Setting: In each example, we constructed a concurrent cohort of new and old drug initiators using US claims databases. In historical cohorts of old drug initiators, we developed hdDRS models including investigator-specified plus empirically identified variables and using principal component analysis and lasso regression for dimension reduction. We applied the models to the concurrent cohorts to obtain predicted outcome probabilities, which we used for confounding adjustment. We compared the resulting estimates to those from hdPS., Results: The crude odds ratio (OR) comparing dabigatran to warfarin was 0.52 (95% confidence interval: 0.37-0.72) for hemorrhage and 0.38 (0.26-0.55) for death. Decile stratification yielded an OR of 0.64 (0.46-0.90) for hemorrhage using hdDRS vs. 0.70 (0.49-1.02) for hdPS. ORs for death were 0.69 (0.45-1.06) and 0.73 (0.48-1.10), respectively. The relative performance of hdDRS in the cyclooxygenase-2 inhibitors example was similar., Conclusion: hdDRS achieved similar or better confounding adjustment compared to conventional regression approach but worked slightly less well than hdPS., (Copyright © 2016 Elsevier Inc. All rights reserved.)

Background: Migraine is a common and important source of pain and disability in society. Accurately identifying such people using routinely collected health data would be beneficial for health services research., Objective: Externally validate a previously published method to identify migraineurs using health administrative data; and determine if a better model can be derived using data-mining techniques., Methods: Migraine status was determined for Ontarians participating in a population-based, cross-sectional survey. Consenting participants were linked to population-based health administrative data to identify age, sex, and coded diagnoses. Discrimination and calibration measures were used to appraise the models. A de novo technique we term "double threshold analysis" was used to determine optimal lower and upper expected probabilities to identify migraine status in the newly derived model., Results: A total of 1,01,114 people (mean age 46 years, 46% male) were included in the study, of which 11,314 (11.2%) had migraines. Using data-driven parameter estimates, the previous model to identify migraineurs had adequate discrimination (c-statistic 0.707 [95% CI 0.701-0.712]) and calibration (Hosmer-Lemeshow [H-L] statistic 20.8). A new model that included diagnostic code scores for physician visits, emergency visits, and hospitalizations with nonlinear terms for age and interactions significantly improved the model (c-statistic 0.724 [0.716-0.733], 16.4). Categorizing all people with a predicted migraine probability less than 10% or greater than 90% as without and having the disease, respectively, resulted in a sensitivity of 3.1%, a specificity of 99.96%, and a positive predictive value of 81.0% while capturing 57.0% of the cohort and 29.3% of migraineurs., Conclusion: A previously derived model to identify migraineurs was improved using data-mining techniques permitting accurate cohort identification using routinely collected health administrative data., (Copyright © 2016 Elsevier Inc. All rights reserved.)

Objective: We derived and validated a method to screen all hospital admissions for 1° subarachnoid hemorrhage (SAH) by retrospectively implementing recognized diagnostic criteria., Study Design and Setting: A screen for 1° SAH was developed using two previously created registries. Screen-positive cases underwent diagnosis confirmation with primary record review. A review of all patient hospital encounters with the diagnostic code for 1° SAH, and cross-referencing with an existing SAH registry was undertaken to identify missed cases., Results: Three subscreens were combined to form the 1° SAH screen (sensitivity: 98.4% [95% CI: 91.7-99.7%], specificity: 93.4% [95% CI: 90.4-95.4%], n = 455 patients in validation sample). From 1,699 screen-positive admissions between July 1, 2002 and June 30, 2011, we identified 831 true cases of SAH of which 632 patients had 1° SAH from ruptured aneurysm/arteriovenous malformation (sensitivity: 96.5% [95% CI: 94.8-97.8%], specificity: 40.3% [95% CI: 38.1-42.6%]). A review of all encounters with a diagnostic code for 1° SAH yielded additional 22 true cases., Conclusion: When positive, our 1° SAH screen significantly increases the probability of this diagnosis in a particular hospitalization. The addition of patient hospitalizations encoded with the diagnostic code for 1° SAH improved sensitivity. Together, these methods represent the best way to retrospectively identify all cases of 1° SAH within an extensive sampling frame., (Copyright © 2016 Elsevier Inc. All rights reserved.)

Objectives: Case-mix adjustment is difficult for stroke outcome studies using administrative data. However, relevant prescription, laboratory, procedure, and service claims might be surrogates for stroke severity. This study proposes a method for developing a stroke severity index (SSI) by using administrative data., Study Design and Setting: We identified 3,577 patients with acute ischemic stroke from a hospital-based registry and analyzed claims data with plenty of features. Stroke severity was measured using the National Institutes of Health Stroke Scale (NIHSS). We used two data mining methods and conventional multiple linear regression (MLR) to develop prediction models, comparing the model performance according to the Pearson correlation coefficient between the SSI and the NIHSS. We validated these models in four independent cohorts by using hospital-based registry data linked to a nationwide administrative database., Results: We identified seven predictive features and developed three models. The k-nearest neighbor model (correlation coefficient, 0.743; 95% confidence interval: 0.737, 0.749) performed slightly better than the MLR model (0.742; 0.736, 0.747), followed by the regression tree model (0.737; 0.731, 0.742). In the validation cohorts, the correlation coefficients were between 0.677 and 0.725 for all three models., Conclusion: The claims-based SSI enables adjusting for disease severity in stroke studies using administrative data., (Copyright © 2015 Elsevier Inc. All rights reserved.)

Objectives: Comorbidity measures are widely used in administrative databases to predict mortality. The Japanese Diagnosis Procedure Combination database is unique in that secondary diagnoses are recorded into subcategories, and procedures are precisely recorded. We investigated the influence of these features on the performance of mortality prediction models., Study Design and Setting: We obtained data of adult patients with main diagnosis of acute myocardial infarction, congestive heart failure, acute cerebrovascular disease, gastrointestinal hemorrhage, pneumonia, or septicemia during a 1-year period. Multiple models were constructed representing different subcategories from which Charlson and Elixhauser comorbidities were extracted. Prevalence of comorbidities and C statistics of logistic regression models predicting in-hospital mortality was compared. Associations between four procedures (computed tomography, oxygen administration, urinary catheter, and vasopressors) and mortality were also evaluated., Results: C statistics of the model using all secondary diagnoses (Charlson: 0.717; Elixhauser: 0.762) were greater than those using a limited subcategory to strictly specify comorbidities (Charlson: 0.708; Elixhauser: 0.744). However, misidentification of complications and main diagnoses as comorbidities was observed in the all-diagnosis model. The four procedures were associated with mortality., Conclusion: Subcategorized diagnoses allowed correct identification of comorbidities and procedures predicted mortality. Incorporation of these two features should be considered for other administrative databases., (Copyright © 2015 Elsevier Inc. All rights reserved.)

Abstract: Objective: To describe techniques used to address confounding in published observational studies. Study Design and Setting: A systematic literature review identified studies using administrative or registry data to investigate health effects of drug therapies. Studies published from January 2001 to December 2005 came from BMJ, New England Journal of Medicine, Lancet, Annals of Internal Medicine, and JAMA. A structured abstraction form was used to collect information about confounding. Results: The search identified 29 studies. Twenty-two studies (76%) had 10,000 or more subjects and 18 (62%) used a mortality outcome. None mentioned use of a literature search to identify confounders, however, 28 (97%) listed confounders included, and 26 (90%) listed confounders not included in the study. Eighteen (62.1%) discussed the validity of confounder data. Most (22, or 76%) studies included a table with the distribution of confounders but none used effect size to assess imbalance between comparison groups. Almost all studies used regression techniques (28, or 97%); fewer used stratification (16, or 55%) or matching (four, or 14%) to address confounding. Eleven (40%) studies discussed sensitivity analyses. Conclusion: Published cohort studies routinely include a list of potential confounders but there is room for improvement in confounder identification, measurement, and analysis. [Copyright &y& Elsevier]

Abstract: Objectives: Reoperations may reflect a suboptimal initial surgical treatment. The study aimed to develop a screening algorithm for those potentially avoidable, using only routinely collected hospital data and a prediction model to adjust rates for case-mix. Study Design and Setting: Data of a 3-year random sample of 7,370 therapeutic operations on inpatients, among which 833 were followed-up by a reoperation during the same stay. A review of medical records identified clearly avoidable and other potentially avoidable reoperations to develop and test the screening algorithm. A logistic prediction model of potentially avoidable reoperations was developed on one randomly chosen half of the data (about 9,000 interventions) and tested on the other half (cross-validation). Results: Two hundred thirty-seven interventions (3%) were followed by a potentially avoidable reoperation, among which 144 were clearly avoidable. The screening algorithm had a sensitivity of 75% and a specificity of 72%. Predictors of potentially avoidable reoperations were surgery categories, diagnosis related conditions, and experiencing prior surgery. The risk score, based on these variables, showed at once a satisfactory discriminative performance (C-statistic=0.76) and goodness-of-fit measure on the validation set. Conclusion: The adjusted rate of potentially avoidable reoperations should be included in internal reporting of hospital quality indicators, but further validated in various settings. [Copyright &y& Elsevier]

Objectives: To develop algorithms of locating patients with primary aldosteronism (PA) using insurance reimbursement data and to validate the algorithms using medical charts., Study Design and Setting: We extracted National Health Insurance (NHI) reimbursement data and medical charts in seven enrolled hospitals and analyzed diagnosis-related information for 1999-2010. The NHI codes PA as 255.1x, using the International Classification of Diseases, Ninth Revision, Clinical Modification. Confirmation of PA was based on suppression tests., Results: We reviewed medical charts for 1,094 cases with at least one PA diagnosis. PA was confirmed for 563 cases. Compared with patients with essential hypertension, PA patients had higher systolic blood pressure, higher aldosterone, lower renin activity, and lower potassium level (all P-values <0.05). An algorithm based on PA diagnosis reported in at least one hospital stay or three outpatient visits had modest performance (sensitivity = 0.94 and specificity = 0.20). The best additional condition for the algorithm was use of mineralocorticoid receptor antagonist (MRA; sensitivity = 0.89 and specificity = 0.88)., Conclusion: Using information on PA diagnosis and MRA prescription reported in insurance claims data can precisely locate PA patients in high-risk groups. This algorithm can construct a reliable PA sample for conducting research in various fields, including epidemiology and clinical practice., (Copyright © 2014 Elsevier Inc. All rights reserved.)

Objective: Prognostication is difficult in a diverse patient population or when outcomes depend on multiple factors. This study derived and internally validated a model to predict risk of death from any cause within 1 year of admission to hospital., Study Design and Setting: The study included all adult Ontarians admitted to nonpsychiatric hospital services in 2011 (n = 640,022) and deterministically linked administrative data to identify 20 patient and admission factors. A split-sample approach was used to derive and internally validate the model., Results: A total of 75,082 people (11.7%) died within 1 year of admission to hospital. The final model included one dozen patient factors (age, sex, living status, comorbidities, home oxygen status, and number of emergency room visits and hospital admissions by ambulance in previous year) and hospitalization factors (admission service and urgency, admission to intensive care unit, whether current hospitalization was a readmission, and admission diagnostic risk score). The model in the validation cohort was highly discriminative (c-statistic 92.3), well calibrated, and used to create the Hospital-patient One-year Mortality Risk score that accurately predicted 1-year risk of death., Conclusion: Routinely collected administrative data can be used to accurately predict 1-year death risk in adults admitted to nonpsychiatric hospital services., (Copyright © 2014 Elsevier Inc. All rights reserved.)