Prevalence of prescription opioid use during pregnancy in eight US health plans during 2001-2014.

PURPOSE: To estimate prevalence of prescription opioid use during pregnancy in eight US health plans during 2001-2014. METHODS: We conducted a cohort study of singleton live birth deliveries. Maternal characteristics were ascertained from health plan and/or birth certificate data and opioids dispensed during pregnancy from health plan pharmacy records. Prevalence of prescription opioid use during pregnancy was calculated for any use, cumulative days of use, and number of dispensings. RESULTS: We examined prevalence of prescription opioid use during pregnancy in each health plan. Tennessee Medicaid had appreciably greater prevalence of use compared to the seven other health plans. Thus, results for the two groups were reported separately. In the seven health plans (n = 587 093 deliveries), prevalence of use during pregnancy was relatively stable at 9%-11% throughout 2001-2014. In Tennessee Medicaid (n = 256 724 deliveries), prevalence increased from 29% in 2001 to a peak of 36%-37% in 2004-2010, and then declined to 28% in 2014. Use for ≥30 days during pregnancy was stable at 1% in the seven health plans and increased from 2% to 7% in Tennessee Medicaid during 2001-2014. Receipt of ≥5 opioid dispensings during pregnancy increased in the seven health plans (0.3%-0.6%) and Tennessee Medicaid (3%-5%) during 2001-2014. CONCLUSION: During 2001-2014, prescription opioid use during pregnancy was more common in Tennessee Medicaid (peak prevalence in late 2000s) compared to the seven health plans (relatively stable prevalence). Although a small percentage of women had opioid use during pregnancy for ≥30 days or ≥ 5 dispensings, they represent thousands of women during 2001-2014.

Can digital communication technology reduce health system personnel time? An evaluation of personnel requirements and costs in a randomized controlled trial.

Use of digital communication technologies (DCT) shows promise for enhancing outcomes and efficiencies in asthma care management. However, little is known about the impact of DCT interventions on healthcare personnel requirements and costs, thus making it difficult for providers and health systems to understand the value of these interventions. This study evaluated the differences in healthcare personnel requirements and costs between usual asthma care (UC) and a DCT intervention (Breathewell) aimed at maintaining guidelines-based asthma care while reducing health care staffing requirements. We used data from a pragmatic, randomized controlled trial conducted in a large integrated health system involving 14,978 patients diagnosed with asthma. To evaluate differences in staffing requirements and cost between Breathewell and UC needed to deliver guideline-based care we used electronic health record (EHR) events, provider time tracking surveys, and invoicing. Differences in cost were reported at the patient and health system level. The Breathewell intervention significantly reduced personnel requirements with a larger percentage of participants requiring no personnel time (45% vs. 5%, p < .001) and smaller percentage of participants requiring follow-up outreach (44% vs. 68%, p < .001). Extrapolated to the total health system, cost for the Breathewell intervention was $16,278 less than usual care. The intervention became cost savings at a sample size of at least 957 patients diagnosed with asthma. At the population level, using DCT to compliment current asthma care practice presents an opportunity to reduce healthcare personnel requirements while maintaining population-based asthma control measures.

Functional clustering methods for longitudinal data with application to electronic health records.

We develop a method to estimate subject-level trajectory functions from longitudinal data. The approach can be used for patient phenotyping, feature extraction, or, as in our motivating example, outcome identification, which refers to the process of identifying disease status through patient laboratory tests rather than through diagnosis codes or prescription information. We model the joint distribution of a continuous longitudinal outcome and baseline covariates using an enriched Dirichlet process prior. This joint model decomposes into (local) semiparametric linear mixed models for the outcome given the covariates and simple (local) marginals for the covariates. The nonparametric enriched Dirichlet process prior is placed on the regression and spline coefficients, the error variance, and the parameters governing the predictor space. This leads to clustering of patients based on their outcomes and covariates. We predict the outcome at unobserved time points for subjects with data at other time points as well as for new subjects with only baseline covariates. We find improved prediction over mixed models with Dirichlet process priors when there are a large number of covariates. Our method is demonstrated with electronic health records consisting of initiators of second-generation antipsychotic medications, which are known to increase the risk of diabetes. We use our model to predict laboratory values indicative of diabetes for each individual and assess incidence of suspected diabetes from the predicted dataset.

Validity of diagnosis and procedure codes for identifying neural tube defects in infants.

PURPOSE: The use of validated criteria to identify birth defects in electronic healthcare databases can avoid the cost and time-intensive efforts required to conduct chart reviews to confirm outcomes. This study evaluated the validity of various case-finding methodologies to identify neural tube defects (NTDs) in infants using an electronic healthcare database. METHODS: This analysis used data generated from a study whose primary aim was to evaluate the association between first-trimester maternal prescription opioid use and NTDs. The study was conducted within the Medication Exposure in Pregnancy Risk Evaluation Program. A broad approach was used to identify potential NTDs including diagnosis and procedure codes from inpatient and outpatient settings, death certificates and birth defect flags in birth certificates. Potential NTD cases were chart abstracted and confirmed by clinical experts. Positive predictive values (PPVs) and 95% confidence intervals (95% CI) are reported. RESULTS: The cohort included 113 168 singleton live-born infants: 55 960 infants with opioid exposure in pregnancy and 57 208 infants unexposed in pregnancy. Seventy-three potential NTD cases were available for the validation analysis. The overall PPV was 41% using all diagnosis and procedure codes plus birth certificates. Restricting approaches to codes recorded in the infants' medical record or to birth certificate flags increased the PPVs (72% and 80%, respectively) but missed a substantial proportion of confirmed NTDs. CONCLUSIONS: Codes in electronic healthcare data did not accurately identify confirmed NTDs. These results indicate that chart review with adjudication of outcomes is important when conducting observational studies of NTDs using electronic healthcare data.

Distributed Regression Analysis Application in Large Distributed Data Networks: Analysis of Precision and Operational Performance.

BACKGROUND: A distributed data network approach combined with distributed regression analysis (DRA) can reduce the risk of disclosing sensitive individual and institutional information in multicenter studies. However, software that facilitates large-scale and efficient implementation of DRA is limited. OBJECTIVE: This study aimed to assess the precision and operational performance of a DRA application comprising a SAS-based DRA package and a file transfer workflow developed within the open-source distributed networking software PopMedNet in a horizontally partitioned distributed data network. METHODS: We executed the SAS-based DRA package to perform distributed linear, logistic, and Cox proportional hazards regression analysis on a real-world test case with 3 data partners. We used PopMedNet to iteratively and automatically transfer highly summarized information between the data partners and the analysis center. We compared the DRA results with the results from standard SAS procedures executed on the pooled individual-level dataset to evaluate the precision of the SAS-based DRA package. We computed the execution time of each step in the workflow to evaluate the operational performance of the PopMedNet-driven file transfer workflow. RESULTS: All DRA results were precise (<10-12), and DRA model fit curves were identical or similar to those obtained from the corresponding pooled individual-level data analyses. All regression models required less than 20 min for full end-to-end execution. CONCLUSIONS: We integrated a SAS-based DRA package with PopMedNet and successfully tested the new capability within an active distributed data network. The study demonstrated the validity and feasibility of using DRA to enable more privacy-protecting analysis in multicenter studies.

Planning for Implementation Success Using RE-AIM and CFIR Frameworks: A Qualitative Study.

Background: RE-AIM (Reach, Effectiveness, Adoption, Implementation, Maintenance) and CFIR (Consolidated Framework for Implementation Research) dissemination and implementation frameworks define theory-based domains associated with the adoption, implementation and maintenance of evidence-based interventions. Used together, the two frameworks identify metrics for evaluating implementation success, i.e., high reach and effectiveness resulting in sustained practice change (RE-AIM), and modifiable factors that explain and enhance implementation outcomes (CFIR). We applied both frameworks to study the implementation planning process for a technology-delivered asthma care intervention called Breathewell within an integrated care organization. The goal of the Breathewell intervention is to increase the efficiency of delivering resource-intensive asthma care services. Methods: We reviewed historical documents (i.e., meeting agendas; minutes) from 14 months of planning to evaluate alignment of implementation team priorities with RE-AIM domains. Key content was extracted and analyzed on topics, frequency and amount of discussion within each RE-AIM domain. Implementation team members were interviewed using questions adapted from the CFIR Interview Guide Tool to focus their reflection on the process and contextual factors considered during pre-implementation planning. Documents and transcripts were initially coded using RE-AIM domain definitions, and recoded using CFIR constructs, with intent to help explain how team decisions and actions can contribute to adoption, implementation and maintenance outcomes. Results: Qualitative analysis of team documents and interviews demonstrated strong alignment with the RE-AIM domains: Reach, Effectiveness, and Implementation; and with the CFIR constructs: formal inclusion of provider and staff stakeholders in implementation planning, compatibility of the intervention with workflows and systems, and alignment of the intervention with organizational culture. Focus on these factors likely contributed to RE-AIM outcomes of high implementation fidelity. However, team members expressed low confidence that Breathewell would be adopted and maintained post-trial. A potential explanation was weak alignment with several CFIR constructs, including tension for change, relative priority, and leadership engagement that contribute to organizational receptivity and motivation to sustain change. Conclusions: While RE-AIM provides a practical framework for planning and evaluating practice change interventions to assure their external validity, CFIR explains why implementation succeeded or failed, and when used proactively, identifies relevant modifiable factors that can promote or undermine adoption, implementation, and maintenance.

Adults With Asthma Experience No Increase in Asthma-related Exacerbations When Digital Communication Technology Tools Are Employed to Offset Provider Workload: A Pragmatic Randomized Trial.

BACKGROUND: Challenges to health care efficiency are increasingly addressed with the help of digital communication technology tools (DCTs). OBJECTIVE: The objective of this study was to test whether DCT, compared with Usual Care, can reduce health care clinician burden without increasing asthma-related exacerbations among patients with asthma in a large integrated health care system. RESEARCH DESIGN: The (Breathewell) program was a pragmatic, randomized trial at (Kaiser Permanente Colorado), where asthma nurses screen patients for poor symptom control when beta2-agonist refill requests came within 60 days of previous fill or in the absence of a controller medication fill within 4 months (beta2-agonist overfill). A total of 14,978 adults with asthma were randomized to Usual Care or 1 of 2 DCT intervention groups (Text/Phone call or Email). SUBJECTS: Participants included adults 18 and older with an asthma diagnosis at the time of randomization and no history of chronic obstructive pulmonary disease. MEASURES: Primary outcome measures included asthma-related health care resource utilization (eg, asthma nurse contacts), medication use, and exacerbations. RESULTS: A total of 1933 patients had 4337 events which met beta2-agonist overfill criteria. Of the 2874 events in the intervention arm, 1188 (41%) were resolved by DCT contact and did not require additional clinician contact. Asthma medication use and exacerbations over 12 months did not differ among the 3 groups. CONCLUSIONS: DCT tools can successfully contact adult asthma patients to screen for symptoms and facilitate intervention. The absence of differences in medication fills and health care utilization indicates that the strategic replacement of nursing interventions by digital outreach did not reduce treatment adherence or compromise health care outcomes.

Refill Reminder Preference and Inhaled Corticosteroid Adherence Among Patients with Asthma.

CONTEXT: Refill reminders can help patients improve adherence to inhaled corticosteroid (ICS) therapy. However, little is known about patient preferences for reminder type or whether patients who express a preference differ from patients who do not. OBJECTIVES: To describe patient preferences for ICS prescription refill reminder type and to compare baseline ICS therapy adherence, measured as proportion of days covered (PDC) 1 year before initiating preference-based reminders, between patients who did and did not express a preference. DESIGN: This substudy within a randomized multi-intervention study was conducted at Kaiser Permanente Colorado. Adults with asthma randomized to intervention were offered the opportunity to choose text, telephone, or email reminders. Patients who did and did not provide a preference were compared by baseline characteristics using log-binomial models. MAIN OUTCOME MEASURE(S): The primary outcomes were reminder preference and type. RESULTS: A total of 1497 of 4545 patients (32.9%) expressed a preference; 789 (52.7%) chose text. The adjusted relative risk (aRR) of not providing a preference increased with decreasing PDC (PDC of 0.50 to < 0.80: aRR, 1.14; 95% confidence interval [CI], 1.04-1.25; PDC < 0.5: aRR, 1.76; 95% CI, 1.59-1.95) compared with patients with a PDC of 0.80 or greater. CONCLUSION: Among patients who expressed a preference, text reminders were preferred. Patients who expressed a preference had higher baseline adherence. Further research is needed to determine whether expressing a preference for a refill reminder type is itself associated with adherence. Given that offering the opportunity to choose a reminder type only engaged a subset of patients, further work is needed to understand how best to leverage technology-enabled communication outreach to help patients optimize adherence.

Technology-Enabled Outreach to Patients Taking High-Risk Medications Reduces a Quality Gap in Completion of Clinical Laboratory Testing.

Clinical laboratory quality improvement (QI) efforts can include population test utilization. The authors used a health care organization's Medical Data Warehouse (MDW) to characterize a gap in guideline-concordant laboratory testing recommended for safe use of antirheumatic agents, then tested the effectiveness of laboratory-led, technology-enabled outreach to patients at reducing this gap. Data linkages available through the Kaiser Permanente Colorado MDW and electronic health record were used to identify ambulatory adults taking antirheumatic agents who were due/overdue for alanine aminotransferase (ALT), aspartate aminotransferase (AST), complete blood count (CBC), or serum creatinine (SCr) testing. Outreach was implemented using an interactive voice response system to send patients text or phone call reminders. Interrupted time series analysis was used to estimate reminder effectiveness. Rates of guideline-concordant testing and testing timeliness in baseline vs. intervention periods were determined using generalized linear models for repeated measures. Results revealed a decrease in percentage of 3763 patients taking antirheumatic agents due/overdue for testing at any given time: baseline 24.3% vs. intervention 17.5% (P < 0.001). Among 3205 patients taking conventional antirheumatic agents, concordance for all ALT testing was baseline 52.8% vs. intervention 65.4% (P < 0.001) among patients chronically using these agents and baseline 20.6% vs. intervention 26.1% (P < 0.001) among patients newly starting these agents. The 95th percentiles for days to ALT testing were baseline 149 vs. intervention 117 among chronic users and baseline 134 vs. intervention 92 among new starts. AST, CBC, and SCr findings were similar. Technology-enabled outreach reminding patients to obtain laboratory testing improves health care system outcomes.

Identifying Common Predictors of Multiple Adverse Outcomes Among Elderly Adults With Type-2 Diabetes.

OBJECTIVE: As part of a multidisciplinary team managing patients with type-2 diabetes, pharmacists need a consistent approach of identifying and prioritizing patients at highest risk of adverse outcomes. Our objective was to identify which predictors of adverse outcomes among type-2 diabetes patients were significant and common across 7 outcomes and whether these predictors improved the performance of risk prediction models. Identifying such predictors would allow pharmacists and other health care providers to prioritize their patient panels. RESEARCH DESIGN AND METHODS: Our study population included 120,256 adults aged 65 years or older with type-2 diabetes from a large integrated health system. Through an observational retrospective cohort study design, we assessed which risk factors were associated with 7 adverse outcomes (hypoglycemia, hip fractures, syncope, emergency department visit or hospital admission, death, and 2 combined outcomes). We split (50:50) our study cohort into a test and training set. We used logistic regression to model outcomes in the test set and performed k-fold validation (k=5) of the combined outcome (without death) within the validation set. RESULTS: The most significant predictors across the 7 outcomes were: age, number of medicines, prior history of outcome within the past 2 years, chronic kidney disease, depression, and retinopathy. Experiencing an adverse outcome within the prior 2 years was the strongest predictor of future adverse outcomes (odds ratio range: 4.15-7.42). The best performing models across all outcomes included: prior history of outcome, physiological characteristics, comorbidities and pharmacy-specific factors (c-statistic range: 0.71-0.80). CONCLUSIONS: Pharmacists and other health care providers can use models with prior history of adverse event, number of medicines, chronic kidney disease, depression and retinopathy to prioritize interventions for elderly patients with type-2 diabetes.

Pregnancy-Associated Cancer: A U.S. Population-Based Study.

BACKGROUND: The incidence of pregnancy-associated cancer (PAC) is expected to increase as more women delay childbearing until later ages. However, information on frequency and incidence of PAC is scarce in the United States. METHODS: We identified pregnancies among women aged 10-54 years during 2001-2013 from five U.S. health plans participating in the Cancer Research Network (CRN) and the Medication Exposure in Pregnancy Risk Evaluation Program (MEPREP). We extracted information from the health plans' administrative claims and electronic health record databases, tumor registries, and infants' birth certificate files to estimate the frequency and incidence of PAC, defined as cancer diagnosed during pregnancy and up to 1 year postpartum. RESULTS: We identified 846 PAC events among 775,709 pregnancies from 2001 to 2013. The overall incidence estimate was 109.1 (95% confidence interval [CI] = 101.8-116.7) per 100,000 pregnancies. There was an increase in the incidence between 2002 and 2012 (the period during which complete data were available), from 75.0 (95% CI = 54.9-100.0) per 100,000 pregnancies in 2002 to 138.5 (95% CI = 109.1-173.3) per 100,000 pregnancies in 2012. The most common invasive cancers diagnosed were breast (n = 208, 24.6%), thyroid (n = 168, 19.9%), melanoma (n = 93, 11.0%), hematologic (n = 87, 10.3%), and cervix/uterus (n = 74, 8.7%). CONCLUSIONS: Our study provides contemporary incidence estimates of PAC from a population-based cohort of U.S. women. These estimates provide the data needed to help develop clinical and public health policies aimed at diagnosing PAC at an early stage and initiating appropriate therapeutic interventions in a timely manner.

Validity of Privacy-Protecting Analytical Methods That Use Only Aggregate-Level Information to Conduct Multivariable-Adjusted Analysis in Distributed Data Networks.

Distributed data networks enable large-scale epidemiologic studies, but protecting privacy while adequately adjusting for a large number of covariates continues to pose methodological challenges. Using 2 empirical examples within a 3-site distributed data network, we tested combinations of 3 aggregate-level data-sharing approaches (risk-set, summary-table, and effect-estimate), 4 confounding adjustment methods (matching, stratification, inverse probability weighting, and matching weighting), and 2 summary scores (propensity score and disease risk score) for binary and time-to-event outcomes. We assessed the performance of combinations of these data-sharing and adjustment methods by comparing their results with results from the corresponding pooled individual-level data analysis (reference analysis). For both types of outcomes, the method combinations examined yielded results identical or comparable to the reference results in most scenarios. Within each data-sharing approach, comparability between aggregate- and individual-level data analysis depended on adjustment method; for example, risk-set data-sharing with matched or stratified analysis of summary scores produced identical results, while weighted analysis showed some discrepancies. Across the adjustment methods examined, risk-set data-sharing generally performed better, while summary-table and effect-estimate data-sharing more often produced discrepancies in settings with rare outcomes and small sample sizes. Valid multivariable-adjusted analysis can be performed in distributed data networks without sharing of individual-level data.

Identifying Preanalytic and Postanalytic Laboratory Quality Gaps Using a Data Warehouse and Structured Multidisciplinary Process.

CONTEXT.­: The laboratory total testing process includes preanalytic, analytic, and postanalytic phases, but most laboratory quality improvement efforts address the analytic phase. Expanding quality improvement to preanalytic and postanalytic phases via use of medical data warehouses, repositories that include clinical, utilization, and administrative data, can improve patient care by ensuring appropriate test utilization. Cross-department, multidisciplinary collaboration to address gaps and improve patient and system outcomes is beneficial. OBJECTIVE.­: To demonstrate medical data warehouse utility for characterizing laboratory-associated quality gaps amenable to preanalytic or postanalytic interventions. DESIGN.­: A multidisciplinary team identified quality gaps. Medical data warehouse data were queried to characterize gaps. Organizational leaders were interviewed about quality improvement priorities. A decision aid with elements including national guidelines, local and national importance, and measurable outcomes was completed for each gap. RESULTS.­: Gaps identified included (1) test ordering; (2) diagnosis, detection, and documentation, and (3) high-risk medication monitoring. After examination of medical data warehouse data including enrollment, diagnoses, laboratory, pharmacy, and procedures for baseline performance, high-risk medication monitoring was selected, specifically alanine aminotransferase, aspartate aminotransferase, complete blood count, and creatinine testing among patients receiving disease-modifying antirheumatic drugs. The test utilization gap was in monitoring timeliness (eg, >60% of patients had a monitoring gap exceeding the guideline recommended frequency). Other contributors to selecting this gap were organizational enthusiasm, regulatory labeling, and feasibility of a significant laboratory role in addressing the gap. CONCLUSIONS.­: A multidisciplinary process facilitated identification and selection of a laboratory medicine quality gap. Medical data warehouse data were instrumental in characterizing gaps.

Trends in Outpatient Antibiotic Use in 3 Health Plans.

OBJECTIVES: Previous analyses of data from 3 large health plans suggested that the substantial downward trend in antibiotic use among children appeared to have attenuated by 2010. Now, data through 2014 from these same plans allow us to assess whether antibiotic use has declined further or remained stable. METHODS: Population-based antibiotic-dispensing rates were calculated from the same health plans for each study year between 2000 and 2014. For each health plan and age group, we fit Poisson regression models allowing 2 inflection points. We calculated the change in dispensing rates (and 95% confidence intervals) in the periods before the first inflection point, between the first and second inflection points, and after the second inflection point. We also examined whether the relative contribution to overall dispensing rates of common diagnoses for which antibiotics were prescribed changed over the study period. RESULTS: We observed dramatic decreases in antibiotic dispensing over the 14 study years. Despite previous evidence of a plateau in rates, there were substantial additional decreases between 2010 and 2014. Whereas antibiotic use rates decreased overall, the fraction of prescribing associated with individual diagnoses was relatively stable. Prescribing for diagnoses for which antibiotics are clearly not indicated appears to have decreased. CONCLUSIONS: These data revealed another period of marked decline from 2010 to 2014 after a relative plateau for several years for most age groups. Efforts to decrease unnecessary prescribing continue to have an impact on antibiotic use in ambulatory practice.

The Effect of Age and Comorbidities on the Association Between the Medicare STAR Oral Antihyperglycemic Adherence Metric and Glycemic Control.

BACKGROUND: The Medicare STAR program for Medicare Advantage Plans that include drug benefits provides monetary incentives for health plans to achieve good adherence to oral antihyperglycemic (OAH) agents but does not account for differential case mix that could affect the ability of health plans to achieve the required quality metrics. OBJECTIVE: To determine whether OAH adherence varies by age and comorbidities among patients aged 65 years or older and the extent to which adherence affects glycemic control across age and comorbidity strata. METHODS: We studied 54,480 patients with diabetes aged > 65 years from the Colorado, Northwest, and Northern California regions of Kaiser Permanente who received OAH agents but not insulin in 2010. We calculated adherence using the proportion of days covered (PDC) method. Per the STAR program, hemoglobin A1c < 8% defined good glycemic control. We also defined poor control as A1c > 9%. We used modified Poisson regression to identify predictors of adherence and to determine its effects on A1c across age and comorbidity strata, adjusting for sociodemographics and medication-related variables. RESULTS: The risk of being adherent to OAH declined moderately with an increasing number of comorbidities (risk ratio [RR] = 0.99, 95% CI = 0.98-1.00 for 1 comorbidity and RR = 0.90, 95% CI = 0.88-0.91 for 4 or more comorbidities). Adherence to OAH agents was associated with a 0%-3% increased risk of A1c < 8% across age and comorbidity categories, as well as a large decreased risk (RR = 0.55-0.73) of A1c > 9% for patients aged < 80 years or with < 3 comorbidities. CONCLUSIONS: Among patients with diabetes aged > 65 years, having multiple comorbidities affects adherence. Adherence reduces the risk of poor A1c control among patients aged 65-79 years or with 2 or fewer comorbidities. Our results suggest that health plan case mix minimally influenced the Medicare STAR OAH adherence metric, but it may affect glycemic control quality measures, especially if a HEDIS-like measure of poor control were adopted. DISCLOSURES: This study was supported by grant number 1R21DK103146-01A1 from the National Institute of Diabetes and Digestive and Kidney Disorders. Nichols currently receives grant funding from Boehringer-Ingelheim, Sanofi, Amarin Pharma, and Janssen Pharmaceuticals for other unrelated research projects. The other authors declare no conflicts of interest. This study was presented at the American Diabetes Association's 77th Scientific Sessions; June 9-13, 2017; San Diego, CA.

Diagnosis-based cohort augmentation using laboratory results data: The case of chronic kidney disease.

PURPOSE: In this report, we use data from FDA's Sentinel System to focus on how augmenting a diagnosis-based chronic kidney disease cohort with patients identified through laboratory results impacts cohort characteristics and outcomes. METHODS: We used data from 2 Data Partners. Patients were eligible if they were health plan members on January 1, 2012. We classified chronic kidney disease patients into mutually exclusive categories according to the hierarchy of (1) ICD-9-CM diagnosis (DXGroup), or (2) two estimated glomerular filtration rates <60 mL/min/1.73m2 , separated by at least 90 days (2-LabGroup), or (3) a single estimated glomerular filtration rates <60 mL/min/1.73m2 (1-LabGroup). We compared the groups on demographic, clinical, and health care utilization characteristics using pairwise standardized differences. We used Cox regression to compare the groups on mortality, adjusting for baseline covariates. RESULTS: We identified 209 864 patients: 107 607 in DxGroup (51%) and 102 257 (49%) from laboratory data alone. For every characteristic, the DxGroup was the sickest, followed by the 2-LabGroup and then the 1-LabGroup. The DxGroup was more likely to die than 2-LabGroup (hazard ratio [HR], 1.47; 95% CI, 1.22-1.77) at Site 1; that effect was observed, but attenuated, at Site 2 (HR, 1.16; 95% CI, 1.07-1.25). The DxGroup was more likely to die than the 1-LabGroup at Site 1 (HR, 1.36; 95% CI, 1.20-1.55), but not at Site 2 (HR, 0.94; 95% CI, 0.89-1.00). CONCLUSIONS: We suggest that drug safety researchers consider whether the method of cohort identification contributes to generalizability of safety findings.

Near Real-time Surveillance for Consequences of Health Policies Using Sequential Analysis.

BACKGROUND: New health policies may have intended and unintended consequences. Active surveillance of population-level data may provide initial signals of policy effects for further rigorous evaluation soon after policy implementation. OBJECTIVE: This study evaluated the utility of sequential analysis for prospectively assessing signals of health policy impacts. As a policy example, we studied the consequences of the widely publicized Food and Drug Administration's warnings cautioning that antidepressant use could increase suicidal risk in youth. METHOD: This was a retrospective, longitudinal study, modeling prospective surveillance, using the maximized sequential probability ratio test. We used historical data (2000-2010) from 11 health systems in the US Mental Health Research Network. The study cohort included adolescents (ages 10-17 y) and young adults (ages 18-29 y), who were targeted by the warnings, and adults (ages 30-64 y) as a comparison group. Outcome measures were observed and expected events of 2 possible unintended policy outcomes: psychotropic drug poisonings (as a proxy for suicide attempts) and completed suicides. RESULTS: We detected statistically significant (P<0.05) signals of excess risk for suicidal behavior in adolescents and young adults within 5-7 quarters of the warnings. The excess risk in psychotropic drug poisonings was consistent with results from a previous, more rigorous interrupted time series analysis but use of the maximized sequential probability ratio test method allows timely detection. While we also detected signals of increased risk of completed suicide in these younger age groups, on its own it should not be taken as conclusive evidence that the policy caused the signal. A statistical signal indicates the need for further scrutiny using rigorous quasi-experimental studies to investigate the possibility of a cause-and-effect relationship. CONCLUSIONS: This was a proof-of-concept study. Prospective, periodic evaluation of administrative health care data using sequential analysis can provide timely population-based signals of effects of health policies. This method may be useful to use as new policies are introduced.

Bisphosphonate Drug Holiday and Fracture Risk: A Population-Based Cohort Study.

Holidays from bisphosphonates (BPs) may help to prevent rare adverse events such as atypical femoral fractures, but may be appropriate only if risk of osteoporosis-related fractures does not increase. Our objective was to compare the incidence of osteoporosis-related fractures among women who had a BP holiday to those who continued to use BPs. This retrospective cohort study, conducted within four Kaiser Permanente integrated health system regions, included 39,502 women aged ≥45 years with ≥3 years exposure to BP. Participants with a BP holiday (≥12 months with no use) were compared to persistent (use with ≥50% adherence) and nonpersistent (use with <50% adherence) users for incident osteoporosis-related fractures. The BP holiday (n = 11,497), nonpersistent user (n = 10,882), and persistent user groups (n = 17,123) were observed for 156,657 person-years. A total of 5199 osteoporosis-related fractures (including 1515 hip fractures and 2147 vertebral fractures) were observed. Compared to the persistent use group, there was a slight difference in overall osteoporosis-related fracture risk (HR 0.92; 95% CI, 0.84 to 0.99)and no difference in hip fracture risk (HR 0.95; 95% CI, 0.83 to 1.10) for the BP holiday group. A slight reduction in risk of vertebral fracture was observed (HR 0.83; 95% CI, 0.74 to 0.95). Compared to the nonpersistent user group, the BP holiday group was at decreased risk for osteoporosis-related fractures (HR 0.71; 95% CI, 0.65 to 0.79), vertebral fractures (HR 0.68; 95% CI, 0.59 to 0.78), and hip fractures (HR 0.59; 95% CI, 0.50 to 0.70). Women who undertake a BP holiday from BP of ≥12 months duration for any reason after ≥3 years of BP use do not appear to be at greater risk of osteoporosis-related fragility fracture, hip, or vertebral fractures compared to ongoing BP users. In our cohort, BP holiday remains a viable strategy for balancing the benefits and potential harms associated with long-term BP use. © 2018 American Society for Bone and Mineral Research.

Validation of an algorithm for identifying type 1 diabetes in adults based on electronic health record data.

PURPOSE: Algorithms using information from electronic health records to identify adults with type 1 diabetes have not been well studied. Such algorithms would have applications in pharmacoepidemiology, drug safety research, clinical trials, surveillance, and quality improvement. Our main objectives were to determine the positive predictive value for identifying type 1 diabetes in adults using a published algorithm (developed by Klompas et al) and to compare it to a simple requirement that the majority of diabetes diagnosis codes be type 1. METHODS: We applied the Klompas algorithm and the diagnosis code criterion to a cohort of 66 690 adult Kaiser Permanente Colorado members with diabetes. We reviewed 220 charts of those identified as having type 1 diabetes and calculated positive predictive values. RESULTS: The Klompas algorithm identified 3286 (4.9% of 66 690) adults with diabetes as having type 1 diabetes. Based on chart reviews, the overall positive predictive value was 94.5%. The requirement that the majority of diabetes diagnosis codes be type 1 identified 3000 (4.5%) as having type 1 diabetes and had a positive predictive value of 96.4%. However, the algorithm criterion involving dispensing of urine acetone test strips performed poorly, with a positive predictive value of 20.0%. CONCLUSIONS: Data from electronic health records can be used to accurately identify adults with type 1 diabetes. When identifying adults with type 1 diabetes, we recommend either a modified version of the Klompas algorithm without the urine acetone test strips criterion or the requirement that the majority of diabetes diagnosis codes be type 1 codes.