Validity of ICD-10-based algorithms to identify patients with influenza in inpatient and outpatient settings.

PURPOSE: To evaluate the validity of ICD-10-CM code-based algorithms as proxies for influenza in inpatient and outpatient settings in the USA. METHODS: Administrative claims data (2015-2018) from the largest commercial insurer in New Jersey (NJ), USA, were probabilistically linked to outpatient and inpatient electronic health record (EHR) data containing influenza test results from a large NJ health system. The primary claims-based algorithms defined influenza as presence of an ICD-10-CM code for influenza, stratified by setting (inpatient/outpatient) and code position for inpatient encounters. Test characteristics and 95% confidence intervals (CIs) were calculated using test-positive influenza as a reference standard. Test characteristics of alternative outpatient algorithms incorporating CPT/HCPCS testing codes and anti-influenza medication pharmacy claims were also calculated. RESULTS: There were 430 documented influenza test results within the study period (295 inpatient, 135 outpatient). The claims-based influenza definition had a sensitivity of 84.9% (95% CI 72.9%-92.1%), specificity of 96.3% (95% CI 93.1%-98.0%), and PPV of 83.3% (95% CI 71.3%-91.0%) in the inpatient setting, and a sensitivity of 76.7% (95% CI 59.1%-88.2%), specificity of 96.2% (95% CI 90.6%-98.5%), PPV of 85.2% (95% CI 67.5%-94.1%) in the outpatient setting. Primary inpatient discharge diagnoses had a sensitivity of 54.7% (95% CI 41.5%-67.3%), specificity of 99.6% (95% CI 97.7%-99.9%), and PPV of 96.7% (95% CI 83.3%-99.4%). CPT/HCPCS codes and anti-influenza medication claims were present for few outpatient encounters (sensitivity 3%-10%). CONCLUSIONS: In a large US healthcare system, inpatient ICD-10-CM codes for influenza, particularly primary inpatient diagnoses, had high predictive value for test-positive influenza. Outpatient ICD-10-CM codes were moderately predictive of test-positive influenza.

Validation of new medication use algorithms as proxies for worsening disease activity in patients with juvenile idiopathic arthritis.

PURPOSE: To facilitate claims-based research on populations with juvenile idiopathic arthritis (JIA), we sought to validate an algorithm of new medication use as a proxy for worsening JIA disease activity. METHODS: Using electronic health record data from three pediatric centers, we defined new JIA medication use as (re)initiation of disease-modifying antirheumatic drugs or glucocorticoids (oral or intra-articular). Data were collected from 201 randomly selected subjects with (101) or without (100) new medication use. We assessed the positive predictive value (PPV) and negative predictive value (NPV) based on a reference standard of documented worsening of JIA disease activity. The algorithm was refined to optimize test characteristics. RESULTS: Overall, the medication-based algorithm had suboptimal performance in representing worsening JIA disease activity (PPV 69.3%, NPV 77.1%). However, algorithm performance improved for definitions specifying longer times after JIA diagnosis (≥1-year post-diagnosis: PPV 82.9%, NPV 80.0%) or after initiation of prior JIA treatment (≥1-year post-treatment: PPV 89.7%, NPV 80.0%). CONCLUSION: An algorithm for new JIA medication use appears to be a reasonable proxy for worsening JIA disease activity, particularly when specifying new use ≥1 year since initiating a prior JIA medication. This algorithm will be valuable for conducting research on JIA populations within administrative claims databases.

Childhood asthma diagnoses declined during the COVID-19 pandemic in the United States.

BACKGROUND: Prior studies have documented declines in pediatric asthma exacerbations and asthma-related health care utilization during the COVID-19 pandemic, but less is known about the incidence of asthma during the pandemic. METHODS: We conducted a retrospective cohort study of children under age 18 without a prior diagnosis of asthma within a large US commercial claims database. Incident asthma was defined using a combination of diagnosis codes, location of services, and medication dispensing. Crude quarterly rates of asthma diagnosis per 1000 children were calculated, and the incidence rate ratio and 95% confidence interval were estimated for newly diagnosed asthma during versus before the pandemic using negative binomial regression, adjusted for age, sex, region, and season. RESULTS: Compared with 3 years prior to the pandemic, crude incident diagnosis rates of asthma decreased by 52% across the first four quarters of the US pandemic. The covariate-adjusted pandemic-associated incidence rate ratio was 0.47 (95% confidence interval 0.43, 0.51). CONCLUSIONS: New diagnoses of childhood asthma in the US declined by half during the first year of the pandemic. These findings raise important questions whether pandemic-related changes in infectious or other triggers truly altered the incidence of childhood asthma beyond the well-described disruptions in healthcare access.

Decreased Emergency Cholecystectomy and Case Fatality Rate, Not Explained by Expansion of Medicaid.

INTRODUCTION: Population data on longitudinal trends for cholecystectomies and their outcomes are scarce. We evaluated the incidence and case fatality rate of emergency and ambulatory cholecystectomies in New Jersey (NJ) and whether the Medicaid expansion changed trends. MATERIALS AND METHODS: A retrospective population cohort design was used to study the incidence of cholecystectomies and their case fatality rate from 2009 to 2018. Using linear and logistic regression we explored the trends of incidence and the odds of case fatality after versus before the January 1, 2014 Medicaid expansion. RESULTS: Overall, 93,423 emergency cholecystectomies were performed, with 644 fatalities; 87,239 ambulatory cholecystectomies were performed, with fewer than 10 fatalities. The 2009 to 2018 annual incidence of emergency cholecystectomies dropped markedly from 114.8 to 77.5 per 100,000 NJ population (P < 0.0001); ambulatory cholecystectomies increased from 93.5 to 95.6 per 100,000 (P = 0.053). The incidence of emergency cholecystectomies dropped more after than before Medicaid expansion (P < 0.0001). The odds ratio for case fatality among those undergoing emergency cholecystectomies after versus before expansion was 0.85 (95% CI, 0.72-0.99). This decrease in case fatality, apparent only in those over age 65, was not explained by the addition of Medicaid. CONCLUSIONS: A marked decrease in the incidence of emergency cholecystectomies occurred after Medicaid expansion, which was not accounted for by a minimal increase in the incidence of ambulatory cholecystectomies. Case fatality from emergency cholecystectomy decreased over time due to factors other than Medicaid. Further work is needed to reconcile these findings with the previously reported lack of decrease in overall gallstone disease mortality in NJ.

Impact of the COVID-19 Pandemic on the Management of Juvenile Idiopathic Arthritis: Analysis of United States Commercial Insurance Data.

BACKGROUND/OBJECTIVE: Given limited information on health care and treatment utilization for juvenile idiopathic arthritis (JIA) during the pandemic, we studied JIA-related health care and treatment utilization in a commercially insured retrospective US cohort. METHODS: We studied rates of outpatient visits, new disease-modifying antirheumatic drug (DMARD) initiations, intra-articular glucocorticoid injections (iaGC), dispensed oral glucocorticoids and opioids, DMARD adherence, and DMARD discontinuation by quarter in March 2018-February 2021 (Q1 started in March). Incident rate ratios (IRR, pandemic vs prepandemic) with 95% confidence intervals (CIs) were estimated using multivariable Poisson or Quasi-Poisson models stratified by diagnosis recency (incident JIA, <12 months ago; prevalent JIA, ≥12 months ago). RESULTS: Among 1294 children diagnosed with JIA, total and in-person outpatient visits for JIA declined during the pandemic (IRR, 0.88-0.90), most markedly in Q1 2020. Telemedicine visits, while higher during the pandemic, declined from 21% (Q1) to 13% (Q4) in 2020 to 2021. During the pandemic, children with prevalent JIA, but not incident JIA, had lower usage of iaGC (IRR, 0.60; 95% CI, 0.34-1.07), oral glucocorticoids (IRR, 0.47; 95% CI, 0.33-0.67), and opioids (IRR, 0.44; 95% CI, 0.26-0.75). Adherence to and discontinuation of DMARDs was similar before and during the pandemic. CONCLUSIONS: In the first year of the pandemic, visits for JIA dropped by 10% to 12% in commercially insured children in the United States, declines partly mitigated by use of telemedicine. Pandemic-related declines in intra-articular glucocorticoids, oral glucocorticoids, and opioids were observed for children with prevalent, but not incident, JIA. These changes may have important implications for disease control and quality of life.

Pandemic vaccine testing: Combining conventional and challenge studies.

Early into COVID, human challenge trials were considered, but usually as alternatives to conventional randomized controlled trials. Instead, assessment of authorized COVID vaccines, of further COVID vaccines, and of vaccines against future pandemics should combine both designs, in five different ways, including a wholly novel one that we elaborate, Viz., combining data from both designs to answer a single question.

Oral Glucocorticoids and Incident Treatment of Diabetes Mellitus, Hypertension, and Venous Thromboembolism in Children.

Rates of incident treatment were quantified in this study for diabetes mellitus, hypertension, and venous thromboembolism (VTE) associated with oral glucocorticoid exposure in children aged 1-18 years. The retrospective cohort included more than 930,000 children diagnosed with autoimmune diseases (namely, inflammatory bowel disease, juvenile idiopathic arthritis, or psoriasis) or a nonimmune comparator condition (attention-deficit/hyperactivity disorder) identified using US Medicaid claims (2000-2010). Associations of glucocorticoid dose per age- and sex-imputed weight with incident treated diabetes, hypertension, and VTE were estimated using Cox regression models. Crude rates were lowest for VTE (unexposed: 0.5/million person-days (95% confidence interval (CI): 0.4, 0.6); currently exposed: 15.6/million person-days (95% CI: 11.8, 20.1)) and highest for hypertension (unexposed: 6.7/million person-days (95% CI: 6.5, 7.0); currently exposed: 74.4/million person-days (95% CI: 65.7, 83.9)). Absolute rates for all outcomes were higher in unexposed and exposed children with autoimmune diseases compared with those with attention-deficit/hyperactivity disorder. Strong dose-dependent relationships were found between current glucocorticoid exposure and all outcomes (adjusted hazard ratios for high-dose glucocorticoids: for diabetes mellitus, 5.93 (95% CI: 3.94, 8.91); for hypertension, 19.13 (95% CI: 15.43, 23.73); for VTE, 16.16 (95% CI: 8.94, 29.22)). These results suggest strong relative risks, but low absolute risks, of newly treated VTE, diabetes, and especially hypertension in children taking high-dose oral glucocorticoids.

Validation of a claims-based algorithm for identifying non-infectious pneumonitis in patients diagnosed with lung cancer.

PURPOSE: Non-infectious pneumonitis (NIP) is a common complication of treatments for lung cancer. We know of no existing validated algorithm for identifying NIP in claims databases, limiting our ability to understand the morbidity and mortality of this toxicity in real-world data. METHODS: Electronic health records (EHR), cancer registry, and administrative data from a National Cancer Institute-designated comprehensive cancer center were queried for patients diagnosed with lung cancer between 10/01/2015-12/31/2020. Health insurance claims were searched for ICD-10-CM codes that indicate an inpatient or outpatient diagnosis with possible NIP. A 20-code (Algorithm A) and 11-code (Algorithm B) algorithm were tested with and without requiring prescription with corticosteroids. Cases with a diagnosis of possible NIP in the 6 months before their first lung cancer diagnosis were excluded. The algorithms were validated by reviewing the EHR. The positive predictive value (PPV) for each algorithm was computed with 95% confidence intervals (CI). RESULTS: Seventy patients with lung cancer had a diagnosis code compatible with NIP: 36 (51.4%) inpatients and 34 (48.6%) outpatients. The PPV of Algorithm A was 77.1% (95% CI: 65.6-86.3). The PPV of Algorithm B was 86.9% (95% CI: 75.8-94.2). Requiring a documented prescription for a systemic corticosteroid improved the PPV of both Algorithm A and Algorithm B: 92.5% (95% CI: 79.6-98.4) and 100.0% (95% CI: 90.0-100.0), respectively. CONCLUSIONS: This study validated ICD-10-CM and prescription-claims-based definitions of NIP in lung cancer patients. All algorithms have at least reasonable performance. Enriching the algorithm with corticosteroid prescription records results in excellent performance.

Risk of clinically relevant hyperglycemia with metoprolol compared to carvedilol in older adults with heart failure and diabetes.

BACKGROUND: Although prior literature suggests that metoprolol may worsen glucose control compared to carvedilol, whether this has clinical relevance among older adults with diabetes and heart failure (HF) remains an open question. METHODS: This was a US retrospective cohort study utilizing data sourced from a 50% national sample of Medicare fee-for-service claims of patients with part D prescription drug coverage (2007-2017). Among patients with diabetes and HF, we identified initiators of metoprolol or carvedilol, which were 1:1 propensity score matched on >90 variables. The primary outcome was initiation of a new oral or injectable antidiabetic medication (proxy for uncontrolled diabetes); secondary outcomes included initiation of insulin and severe hyperglycemic event (composite of emergency room visits or hospitalizations related to hyperglycemia). RESULTS: Among 24 239 propensity score-matched pairs (mean [SD] age 77.7 [8.0] years; male [39.1%]), there were 8150 (incidence rate per 100 person-years [IR] = 33.5) episodes of antidiabetic medication initiation among metoprolol users (exposure arm) compared to 8576 (IR = 33.4) among carvedilol users (comparator arm) compared to corresponding to an adjusted hazard ratio (aHR) of 0.97 (95% confidence interval [CI]: 0.94, 1.01). Similarly, metoprolol was not associated with a significant increase in the risk of secondary outcomes including insulin initiation: aHR of 0.98 (95% CI: 0.93, 1.04) and severe hyperglycemic events: aHR of 0.98 (95% CI: 0.93, 1.02). CONCLUSIONS: In this large study of older adults with HF and diabetes, initiation of metoprolol compared to carvedilol was not associated with an increase in the risk of clinically relevant hyperglycemia.

Pre-diagnosis alcohol consumption and mortality risk among black women and white women with invasive breast cancer.

BACKGROUND: Alcohol consumption is associated with increased risk of breast cancer; however, its association with subsequent risk of breast cancer death is unclear. METHODS: We followed 4523 women with complete information on relevant risk factors for mortality; these women were 35 to 64 years of age when diagnosed with incident invasive breast cancer between 1994 and 1998. During follow up (median, 8.6 years), 1055 women died; 824 died from breast cancer. The information on alcohol consumption before diagnosis was collected shortly after breast cancer diagnosis (average: 5.1 months) during an in-person interview which used a structured questionnaire. Multivariable Cox proportional hazards regression models provided hazard ratios (HRs) and 95% confidence intervals (CIs) for breast cancer-specific mortality, mortality due to causes other than breast cancer, and all-cause mortality associated with alcohol consumption from age 15 years until breast cancer diagnosis and during recent periods of time prior to breast cancer diagnosis. RESULTS: Average weekly alcohol consumption from age 15 years until breast cancer diagnosis was inversely associated with breast cancer-specific mortality (Ptrend = 0.01). Compared to non-drinkers, women in the highest average weekly alcohol consumption category (≥7 drinks/week) had 25% lower risk of breast cancer-specific mortality (HR = 0.75, 95% CI = 0.56-1.00). Breast cancer mortality risk was also reduced among women in the highest average weekly alcohol consumption category in two recent time periods (5-year period ending 2-years prior to breast cancer diagnosis, HR = 0.74, 95% CI = 0.57-0.95; 2-year period immediately prior to breast cancer diagnosis: HR = 0.73, 95% CI = 0.56-0.95). Furthermore, analyses of average weekly alcohol consumption by beverage type from age 15 years until breast cancer diagnosis suggested that wine consumption was inversely associated with breast cancer-specific mortality risk (wine Ptrend = 0.06, beer Ptrend = 0.24, liquor Ptrend = 0.74). No association with any of these alcohol consumption variables was observed for mortality risk due to causes other than breast cancer. CONCLUSIONS: Overall, we found no evidence that alcohol consumption before breast cancer diagnosis increases subsequent risk of death from breast cancer.

Concordance of hospitalizations between Clinical Practice Research Datalink and linked Hospital Episode Statistics among patients treated with oral antidiabetic therapies.

PURPOSE: The ability of the Clinical Practice Research Datalink (CPRD) to ascertain all-cause hospitalizations remains unknown. We determined the proportion of hospitalizations in CPRD that were also recorded in Hospital Episode Statistics (HES), and vice versa, among patients initiating oral antidiabetic (OAD) therapy. METHODS: We conducted a retrospective cohort study from October 2009 to September 2012 among OAD-treated patients registered with general practitioners who contribute to CPRD and consent to HES linkage. In CPRD, we identified initial hospitalizations for each calendar year by an Inpatient Referral, Consultation Type code, or Read code indicating an inpatient episode and determined if an admission date was recorded in HES within ±30 days. We then identified initial HES admission dates and determined if a hospitalization was documented in CPRD within ±30 days. Sensitivity analyses were conducted utilizing HES discharge, rather than admission, dates. RESULTS: Among 8574 OAD-treated HES-linked patients in CPRD, 6574 initial hospitalizations across the study period were identified in CPRD, and 5188 (78.9% [95% CI, 77.9%-79.9%]) were confirmed by a HES admission date within ±30 days (median difference, ±3 days [IQR, 1-7 days]). Among 8609 initial hospital admissions in HES, 4803 (55.7% [95% CI, 54.7%-56.8%]) hospitalizations were recorded in CPRD within ±30 days (median difference, ±4 days [IQR, 1-9 days]). Similar results were observed using HES discharge dates. CONCLUSION: A substantial minority of patient-level hospitalization data are nonconcordant between HES and CPRD. Pharmacoepidemiologic studies within CPRD that seek to identify hospitalizations should consider linkage with HES to ensure adequate ascertainment of inpatient events.

Improving postapproval drug safety surveillance: getting better information sooner.

Adverse drug events (ADEs) are an important public health concern, accounting for 5% of all hospital admissions and two-thirds of all complications occurring shortly after hospital discharge. There are often long delays between when a drug is approved and when serious ADEs are identified. Recent and ongoing advances in drug safety surveillance include the establishment of government-sponsored networks of population databases, the use of data mining approaches, and the formal integration of diverse sources of drug safety information. These advances promise to reduce delays in identifying drug-related risks and in providing reassurance about the absence of such risks.

Polymorphisms in cytochrome P450 are associated with extensive efavirenz pharmacokinetics and CNS toxicities in an HIV cohort in Botswana.

Inter-individual variability in efavirenz (EFV) pharmacokinetics and dynamics is dominantly driven by the polymorphism in cytochrome P450 (CYP) isoenzyme 2B6 516G>T. We hypothesized that additional CYP polymorphisms mediate the relationship between CYP2B6 516G>T, EFV metabolism, and clinical events. We investigated 21 SNPs in 814 HIV-infected adults initiating EFV-based therapy in Botswana for population pharmacokinetics, CNS toxicities, and treatment outcomes. Two SNPs (rs28399499 and rs28399433) showed reduced apparent oral EFV clearance. Four SNPs (rs2279345, rs4803417, rs4802101, and rs61663607) showed extensive clearance. Composite CYP2B-mediated EFV metabolism was significantly associated with CNS toxicity (p = 0.04), with extensive metabolizers reporting more and slow and very slow metabolizers reporting less toxicity after 1 month compared to intermediate metabolizers. Composite CYP2B6 metabolism was not associated with composite early treatment failure. In conclusion, our data suggest that CNS-related toxicities might not be solely the result of super-therapeutic parent EFV concentrations in HIV-infected individuals in patients of African ancestry.

Validity of diagnostic codes to identify hospitalizations for infections among patients treated with oral anti-diabetic drugs.

PURPOSE: Identification of hospitalizations for infection is important for post-marketing surveillance of drugs, but the validity of using diagnosis codes to identify these events is unknown. Differentiating between hospitalization for and with infection is important, as the latter is common and less likely to arise from pre-admission exposure to drugs. We determined positive predictive values (PPVs) of diagnostic coding-based algorithms to identify hospitalization for infection among patients prescribed oral anti-diabetic drugs (OADs). METHODS: We identified patients initiating OADs within 2 United States claims databases (Medicare, HealthCore Integrated Research DatabaseSM [HIRDSM ]) and 2 United Kingdom electronic medical record databases (Clinical Practice Research Datalink [CPRD], The Health Improvement Network [THIN]) from 2009 to 2014. To identify potential hospitalizations for infection, we selected patients with a hospital diagnosis of infection and, within 7 days prior to hospitalization, either an outpatient/emergency department visit with an infection diagnosis or outpatient antimicrobial treatment. Hospital records were reviewed by infectious disease specialists to adjudicate hospital admissions for infection. PPVs for confirmed outcomes were determined for each database. RESULTS: Code-based algorithms to identify hospitalization for infection had PPVs exceeding 80% within Medicare (PPV, 83% [90/109]; 95% CI, 74-89%), HIRDSM (PPV, 89% [73/82]; 95% CI, 80-95%), and THIN (PPV, 86% [12/14]; 95% CI, 57-98%) but not within CPRD (PPV, 67% [14/21]; 95% CI, 43-85%). CONCLUSIONS: Algorithms identifying hospitalization for infection utilizing hospital diagnoses along with antecedent outpatient/emergency infection diagnoses or antimicrobial therapy had sufficiently high PPVs for confirmed events within Medicare, HIRDSM , and THIN to enable their use for pharmacoepidemiologic research.

Impact of the black triangle label on prescribing of new drugs in the United Kingdom: lessons for the United States at a time of deregulation.

PURPOSE: Newly approved novel drugs in Europe receive a black triangle label to promote pharmacovigilance. With growing momentum for earlier drug approvals and reliance on real-world evidence, we studied if the black triangle label promotes more judicious prescribing. METHODS: We examined whether general practitioners prescribed escitalopram, tadalafil, and vardenafil with a black triangle more cautiously than the same or similar drugs without a black triangle in The Health Improvement Network (UK). We performed interrupted time-series analyses to estimate changes in new prescription rates and nested case-control studies to compare characteristics of new users before and after removal of a black triangle. RESULTS: Prescribing rates to the 33 441 new users of these new drugs were highest shortly after initial approval and declined subsequently; there were no increases in rates of new prescriptions after a black triangle's removal (new prescriptions/million/month postlabel: escitalopram -1.5 [95% CI, -1.9 to -1.2]; tadalafil and vardenafil: -0.1 [95% CI, -0.6 to 0.4]). Among drugs in the same class, loss of a patent had more impact on prescribing rates than loss of a black triangle. People who began taking black triangle drugs were less likely to be young or to have multiple comorbidities or recent hospitalization compared with those starting the same drugs after the label's removal. However, these differences generally reflected secular trends seen also in similar, unlabeled medicines. CONCLUSIONS: Accelerated drug approvals could cause more uncertainty about drug effectiveness and safety, but specific labeling of newly approved medicines is unlikely to promote more judicious prescribing.

Population-representative incidence of drug-induced acute liver failure based on an analysis of an integrated health care system.

BACKGROUND & AIMS: Medications are a major cause of acute liver failure (ALF) in the United States, but no population-based studies have evaluated the incidence of ALF from drug-induced liver injury. We aimed to determine the incidence and outcomes of drug-induced ALF in an integrated health care system that approximates a population-based cohort. METHODS: We performed a retrospective cohort study using data from the Kaiser Permanente Northern California (KPNC) health care system between January 1, 2004, and December 31, 2010. We included all KPNC members age 18 years and older with 6 months or more of membership and hospitalization for potential ALF. The primary outcome was drug-induced ALF (defined as coagulopathy and hepatic encephalopathy without underlying chronic liver disease), determined by hepatologists who reviewed medical records of all KPNC members with inpatient diagnostic and laboratory criteria suggesting potential ALF. RESULTS: Among 5,484,224 KPNC members between 2004 and 2010, 669 had inpatient diagnostic and laboratory criteria indicating potential ALF. After medical record review, 62 (9.3%) were categorized as having definite or possible ALF, and 32 (51.6%) had a drug-induced etiology (27 definite, 5 possible). Acetaminophen was implicated in 18 events (56.3%), dietary/herbal supplements in 6 events (18.8%), antimicrobials in 2 events (6.3%), and miscellaneous medications in 6 events (18.8%). One patient with acetaminophen-induced ALF died (5.6%; 0.06 events/1,000,000 person-years) compared with 3 patients with non-acetaminophen-induced ALF (21.4%; 0.18/1,000,000 person-years). Overall, 6 patients (18.8%) underwent liver transplantation, and 22 patients (68.8%) were discharged without transplantation. The incidence rates of any definite drug-induced ALF and acetaminophen-induced ALF were 1.61 events/1,000,000 person-years (95% confidence interval, 1.06-2.35) and 1.02 events/1,000,000 person-years (95% confidence interval, 0.59-1.63), respectively. CONCLUSIONS: Drug-induced ALF is uncommon, but over-the-counter products and dietary/herbal supplements are its most common causes.

Risk of Acute Liver Failure in Patients With Drug-Induced Liver Injury: Evaluation of Hy's Law and a New Prognostic Model.

BACKGROUND & AIMS: Few studies have evaluated the ability of laboratory tests to predict risk of acute liver failure (ALF) among patients with drug-induced liver injury (DILI). We aimed to develop a highly sensitive model to identify DILI patients at increased risk of ALF. We compared its performance with that of Hy's Law, which predicts severity of DILI based on levels of alanine aminotransferase or aspartate aminotransferase and total bilirubin, and validated the model in a separate sample. METHODS: We conducted a retrospective cohort study of 15,353 Kaiser Permanente Northern California members diagnosed with DILI from 2004 through 2010, liver aminotransferase levels above the upper limit of normal, and no pre-existing liver disease. Thirty ALF events were confirmed by medical record review. Logistic regression was used to develop prognostic models for ALF based on laboratory results measured at DILI diagnosis. External validation was performed in a sample of 76 patients with DILI at the University of Pennsylvania. RESULTS: Hy's Law identified patients that developed ALF with a high level of specificity (0.92) and negative predictive value (0.99), but low level of sensitivity (0.68) and positive predictive value (0.02). The model we developed, comprising data on platelet count and total bilirubin level, identified patients with ALF with a C statistic of 0.87 (95% confidence interval [CI], 0.76-0.96) and enabled calculation of a risk score (Drug-Induced Liver Toxicity ALF Score). We found a cut-off score that identified patients at high risk patients for ALF with a sensitivity value of 0.91 (95% CI, 0.71-0.99) and a specificity value of 0.76 (95% CI, 0.75-0.77). This cut-off score identified patients at high risk for ALF with a high level of sensitivity (0.89; 95% CI, 0.52-1.00) in the validation analysis. CONCLUSIONS: Hy's Law identifies patients with DILI at high risk for ALF with low sensitivity but high specificity. We developed a model (the Drug-Induced Liver Toxicity ALF Score) based on platelet count and total bilirubin level that identifies patients at increased risk for ALF with high sensitivity.

Use of demographic and pharmacy data to identify patients included within both the Clinical Practice Research Datalink (CPRD) and The Health Improvement Network (THIN).

PURPOSE: Pharmacoepidemiology researchers often utilize data from two UK electronic medical record databases, the Clinical Practice Research Datalink (CPRD) and The Health Improvement Network (THIN), and may choose to combine the two in an effort to increase sample size. To minimize duplication of data, previous studies examined the practice-level overlap between these databases. However, the proportion of overlapping patients remains unknown. We developed a method using demographic and pharmacy variables to identify patients included in both CPRD and THIN, and applied this method to measure the proportion of overlapping patients who initiated the oral anti-diabetic drug saxagliptin. METHODS: We conducted a cross-sectional study among patients initiating saxagliptin in CPRD and THIN between October 2009 and September 2012. Within both databases, we identified patients: (i) ≥18 years, (ii) newly prescribed saxagliptin, and (iii) with ≥180 days enrollment prior to saxagliptin initiation. Demographic data (birth year, sex, patient registration date, family number, and marital status) and prescriptions (including dates) for the first two oral anti-diabetic drugs prescribed within the study period were used to identify matching patients. RESULTS: Among 4202 CPRD and 3641 THIN patients initiating saxagliptin, 2574 overlapping patients (61% of CPRD saxagliptin initiators; 71% of THIN saxagliptin initiators) were identified. Among these patients, 2474 patients (96%) perfectly matched on all demographic and prescription data. CONCLUSIONS: Within each database, over 60% of patients initiating saxagliptin were included within both CPRD and THIN. Combined demographic and prescription data can be used to identify patients included in both CPRD and THIN.

Validity of diagnostic codes and laboratory tests of liver dysfunction to identify acute liver failure events.

PURPOSE: Identification of acute liver failure (ALF) is important for post-marketing surveillance of medications, but the validity of using ICD-9 diagnoses and laboratory data to identify these events within electronic health records is unknown. We examined positive predictive values (PPVs) of hospital ICD-9 diagnoses and laboratory tests of liver dysfunction for identifying ALF within a large, community-based integrated care organization. METHODS: We identified Kaiser Permanente Northern California health plan members (2004-2010) with a hospital diagnosis suggesting ALF (ICD-9 570, 572.2, 572.4, 572.8, 573.3, 573.8, or V42.7) plus an inpatient international normalized ratio ≥1.5 (off warfarin) and total bilirubin ≥5.0 mg/dL. Hospital records were reviewed by hepatologists to adjudicate ALF events. PPVs for confirmed outcomes were determined for individual ICD-9 diagnoses, diagnoses plus prescriptions for hepatic encephalopathy treatment, and combinations of diagnoses in the setting of coagulopathy and hyperbilirubinemia. RESULTS: Among 669 members with no pre-existing liver disease, chart review confirmed ALF in 62 (9%). Despite the presence of co-existing coagulopathy and hyperbilirubinemia, individual ICD-9 diagnoses had low PPVs (range, 5-15%); requiring prescriptions for encephalopathy treatment did not increase PPVs of these diagnoses (range, 2-23%). Hospital diagnoses of other liver disorders (ICD-9 573.8) plus hepatic coma (ICD-9 572.2) had high PPV (67%; 95%CI, 9-99%) but only identified two (3%) ALF events. CONCLUSIONS: Algorithms comprising relevant hospital diagnoses, laboratory evidence of liver dysfunction, and prescriptions for hepatic encephalopathy treatment had low PPVs for confirmed ALF events. Studies of ALF will need to rely on medical records to confirm this outcome.