Association of initial opioid prescription duration and an opioid refill by pain diagnosis: Evidence from outpatient settings in ten US health systems.

OBJECTIVE: The Centers for Disease Control and Prevention's 2022 Clinical Practice Guideline for Prescribing Opioids for Pain cautioned that inflexible opioid prescription duration limits may harm patients. Information about the relationship between initial opioid prescription duration and a subsequent refill could inform prescribing policies and practices to optimize patient outcomes. We assessed the association between initial opioid duration and an opioid refill prescription. METHODS: We conducted a retrospective cohort study of adults ≥19 years of age in 10 US health systems between 2013 and 2018 from outpatient care with a diagnosis for back pain without radiculopathy, back pain with radiculopathy, neck pain, joint pain, tendonitis/bursitis, mild musculoskeletal pain, severe musculoskeletal pain, urinary calculus, or headache. Generalized additive models were used to estimate the association between opioid days' supply and a refill prescription. RESULTS: Overall, 220,797 patients were prescribed opioid analgesics upon an outpatient visit for pain. Nearly a quarter (23.5%) of the cohort received an opioid refill prescription during follow-up. The likelihood of a refill generally increased with initial duration for most pain diagnoses. About 1 to 3 fewer patients would receive a refill within 3 months for every 100 patients initially prescribed 3 vs. 7 days of opioids for most pain diagnoses. The lowest likelihood of refill was for a 1-day supply for all pain diagnoses, except for severe musculoskeletal pain (9 days' supply) and headache (3-4 days' supply). CONCLUSIONS: Long-term prescription opioid use increased modestly with initial opioid prescription duration for most but not all pain diagnoses examined.

Prescription Opioid Dose Reductions and Potential Adverse Events: a Multi-site Observational Cohort Study in Diverse US Health Systems.

BACKGROUND: In response to the opioid crisis in the United States, population-level prescribing of opioids has been decreasing; there are concerns, however, that dose reductions are related to potential adverse events. OBJECTIVE: Examine associations between opioid dose reductions and risk of 1-month potential adverse events (emergency department (ED) visits, opioid overdose, benzodiazepine prescription fill, all-cause mortality). DESIGN: This observational cohort study used electronic health record and claims data from eight United States health systems in a prescription opioid registry (Clinical Trials Network-0084). All opioid fills (excluding buprenorphine) between 1/1/2012 and 12/31/2018 were used to identify baseline periods with mean morphine milligram equivalents daily dose of  ≥ 50 during six consecutive months. PATIENTS: We identified 60,040 non-cancer patients with  ≥ one 2-month dose reduction period (600,234 unique dose reduction periods). MAIN MEASURES: Analyses examined associations between dose reduction levels (1- < 15%, 15- < 30%, 30- < 100%, 100% over 2 months) and potential adverse events in the month following a dose reduction using logistic regression analysis, adjusting for patient characteristics. KEY RESULTS: Overall, dose reduction periods involved mean reductions of 18.7%. Compared to reductions of 1- < 15%, dose reductions of 30- < 100% were associated with higher odds of ED visits (OR 1.14, 95% CI 1.10, 1.17), opioid overdose (OR 1.41, 95% CI 1.09-1.81), and all-cause mortality (OR 1.39, 95% CI 1.16-1.67), but lower odds of a benzodiazepine fill (OR 0.83, 95% CI 0.81-0.85). Dose reductions of 15- < 30%, compared to 1- < 15%, were associated with higher odds of ED visits (OR 1.08, 95% CI 1.05-1.11) and lower odds of a benzodiazepine fill (OR 0.93, 95% CI 0.92-0.95), but were not associated with opioid overdose and all-cause mortality. CONCLUSIONS: Larger reductions for patients on opioid therapy may raise risk of potential adverse events in the month after reduction and should be carefully monitored.

Incidence and Survival for Patients Diagnosed With Breast, Colorectal, and Lung Cancer in an Integrated System.

INTRODUCTION: Documenting trends in cancer incidence and survival is a national priority. This study estimated age- and sex-adjusted incidence and 5-year relative survival among patients with cancer diagnosed within Kaiser Permanente compared to Surveillance, Epidemiology, and End Results (SEER) estimates. METHODS: The cohort included Kaiser Permanente health plan members diagnosed with breast (BC), colorectal (CRC), or lung cancer (LC) between January 1, 1999 and December 31, 2018. Incidence was computed as age-adjusted rates per 100,000 member-years. SEER*Stat was used to compute 5-year relative survival. RESULTS: Kaiser Permanente BC incidence rates were persistently higher than SEER from 2004 (126.5 [95% confidence interval (CI) = 123.2-129.9] vs 122.6 [95% CI = 121.3-123.2]) through 2013 (132.06 [95% CI = 129.5-135.7] vs 126.7 [95% CI = 125.9-127.5]). Kaiser Permanente CRC and LC incidence rates were lower than SEER for all years except 2008, showing a spike in CRC incidence (51.5 [95% CI = 49.9-53.0] vs 46.1 [95% CI = 45.7-46.4]). Kaiser Permanente BC, CRC, and LC survival estimates for all stages were higher than SEER. CONCLUSIONS: Incidence rates for all-stage and localized-stage BC were consistently higher for Kaiser Permanente than for SEER. CRC and LC rates were lower. Kaiser Permanente survival rates were consistently higher than for SEER. The strengths of these findings are associated with the ability to capture "gold-standard" cancer registry data on defined Kaiser Permanente populations. However, findings should be interpreted cautiously given differences in the underlying populations and secular and regional differences between Kaiser Permanente and SEER. The Kaiser Permanente population is younger and more racially diverse than SEER aggregate populations, and Kaiser Permanente members are insured with access to preventive care (eg, smoking cessation programs, cancer screening).

Factor Analysis in Distinguishing Coronavirus Disease 2019 From Other Influenza-like Illness Using a Validated Patient-Reported Outcome Instrument FLU-PRO Plus: A Prospective Real-world Cohort Study.

BACKGROUND/OBJECTIVE: InFLUenza Patient-reported Outcome (FLU-PRO Plus) is a 34-item patient-reported outcome instrument designed to capture the intensity and frequency of viral respiratory symptoms. This study evaluates whether FLU-PRO Plus responses could discriminate between symptoms of coronavirus disease 2019 (COVID-19) and influenza-like illness (ILI) with no COVID diagnosis, as well as forecast disease progression. METHODS: FLU-PRO Plus was administered daily for 14 days. Exploratory factor analysis was used to reduce the FLU-PRO Plus responses on the first day to 3 factors interpreted as "symptom clusters." The 3 clusters were used to predict COVID-19 versus ILI diagnosis in logistic regression. Correlation between the clusters and quality of life (QoL) measures was used to assess concurrent validity. The timing of self-reported return to usual health in the 14-day period was estimated as a function of the clusters within COVID-19 and ILI groups. RESULTS: Three hundred fourteen patients completed day 1 FLU-PRO Plus, of which 65% had a COVID-19 diagnosis. Exploratory factor analysis identified 3 symptom clusters: (1)general Body, (2) tracheal/bronchial, and (3) nasopharyngeal. Higher nasopharyngeal scores were associated with higher odds of COVID-19 compared with ILI diagnosis [adjusted odds ratio = 1.61 (1.21, 2.12)]. Higher tracheal/bronchial scores were associated with lower odds of COVID-19 [0.58 (0.44, 0.77)]. The 3 symptom clusters were correlated with multiple QoL measures ( r = 0.14-0.56). Higher scores on the general body and tracheal/bronchial symptom clusters were associated with prolonged time to return to usual health [adjusted hazard ratios: 0.76 (0.64, 0.91), 0.80 (0.67, 0.96)]. CONCLUSION: Three symptom clusters identified from FLU-PRO Plus responses successfully discriminated patients with COVID-19 from non-COVID ILI and were associated with QoL and predicted symptom duration.

Impact of COVID-19 on Trends in Outpatient Clinic Utilization: A Tale of 2 Departments.

BACKGROUND: The COVID-19 pandemic forced many US health care organizations to shift from mostly in-person care to a hybrid of virtual visits (VV) and in-person visits (IPV). While there was an expected and immediate shift to virtual care (VC) early in the pandemic, little is known about trends in VC use after restrictions eased. METHODS: This is a retrospective study using data from 3 health care systems. All completed visits from adult primary care (APC) and behavioral health (BH) were extracted from the electronic health record of adults aged 19 years and older from January 1, 2019 to June 30, 2021. Standardized weekly visit rates were calculated by department and site and analyzed using time series analysis. RESULTS: There was an immediate decrease in APC visits following the onset of the pandemic. IPV were quickly replaced by VV such that VV accounted for most APC visits early in the pandemic. By 2021, VV rates declined, and VC visits accounted for <50% of all APC visits. By Spring 2021, all 3 health care systems saw a resumption of APC visits as rates neared or returned to prepandemic levels. In contrast, BH visit rates remained constant or slightly increased. By April 2020, almost all BH visits were being delivered virtually at each of the 3 sites and continue to do so without changes to utilization. CONCLUSIONS: VC use peaked during the early pandemic period. While rates of VC are higher than prepandemic levels, IPV are the predominant visit type in APC. In contrast, VC use has sustained in BH, even after restrictions eased.

Patient Factors Associated With Use of Adult Primary Care and Virtual Visits During the COVID-19 Pandemic.

BACKGROUND: The delivery of adult primary care (APC) shifted from predominately in-person to modes of virtual care during the COVID-19 pandemic. It is unclear how these shifts impacted the likelihood of APC use during the pandemic, or how patient characteristics may be associated with the use of virtual care. METHODS: A retrospective cohort study using person-month level datasets from 3 geographically disparate integrated health care systems was conducted for the observation period of January 1, 2020, through June 30, 2021. We estimated a 2-stage model, first adjusting for patient-level sociodemographic, clinical, and cost-sharing factors, using generalized estimating equations with a logit distribution, along with a second-stage multinomial generalized estimating equations model that included an inverse propensity score treatment weight to adjust for the likelihood of APC use. Factors associated with APC use and virtual care use were separately assessed for the 3 sites. RESULTS: Included in the first-stage models were datasets with total person-months of 7,055,549, 11,014,430, and 4,176,934, respectively. Older age, female sex, greater comorbidity, and Black race and Hispanic ethnicity were associated with higher likelihood of any APC use in any month; measures of greater patient cost-sharing were associated with a lower likelihood. Conditional on APC use, older age, and adults identifying as Black, Asian, or Hispanic were less likely to use virtual care. CONCLUSIONS: As the transition in health care continues to evolve, our findings suggest that to ensure vulnerable patient groups receive high quality health care, outreach interventions to reduce barriers to virtual care use may be warranted.

Virtual Care and Urinary Tract Infection Management: Comparing Ancillary Service Orders and Patient Order Fulfillments Between Virtual and In-person Encounters During the COVID-19 Pandemic in the United States.

BACKGROUND: During the COVID-19 pandemic, more health care issues were being managed remotely. Urinary tract infections (UTIs) are being managed more often using telehealth although few reports compare the rate of UTI ancillary service orders placed and fulfilled during these visits. OBJECTIVES: We aimed to evaluate and compare the rate of ancillary service orders and order fulfillments in incident UTI diagnoses between virtual and in-person encounters. RESEARCH DESIGN: The retrospective cohort study involved 3 integrated health care systems: Kaiser Permanente (KP) Colorado, KP Georgia, and KP Mid-Atlantic States. SUBJECTS: We included incident UTI encounters from adult primary care data from January 2019 to June 2021. MEASURES: Data were categorized as: prepandemic (January 2019-March 2020), COVID-19 Era 1 (April 2020-June 2020), and COVID-19 Era 2 (July 2020-June 2021). UTI-specific ancillary services included medication, laboratory, and imaging. Orders and order fulfillments were dichotomized for analyses. Weighted percentages for orders and fulfillments were calculated using inverse probability treatment weighting from logistic regression and compared between virtual and in-person encounters using χ2 tests. RESULTS: We identified 123,907 incident encounters. Virtual encounters increased from 13.4% prepandemic to 39.1% in COVID-19 Era 2. Ancillary service orders from virtual encounters were not placed as often as in-person encounters. However, the weighted percentage for ancillary service order fulfillment across all services remained above 65.3% across sites and eras, with many fulfillment percentages above 90%. CONCLUSIONS: Our study reported a high rate of order fulfillment for both virtual and in-person encounters. Health care systems should encourage providers to place ancillary service orders for uncomplicated diagnoses, such as UTI, to provide enhanced access to patient-centered care.

Management of Neck or Back Pain in Ambulatory Care: Did Visit Mode or the COVID-19 Pandemic Affect Provider Practice or Patient Adherence?

BACKGROUND/OBJECTIVE: In recent years, 2 circumstances have changed provider-patient interactions in ambulatory care: (1) the replacement of virtual for in-person visits and (2) the COVID-19 pandemic. We studied the potential impact of each event on provider practice and patient adherence by comparing the frequency of the association of provider orders, and patient fulfillment of those orders, by visit mode and pandemic period, for incident neck or back pain (NBP) visits in ambulatory care. METHODS: Data were extracted from the electronic health records of 3 Kaiser Permanente regions (Colorado, Georgia, and Mid-Atlantic States) from January 2017 to June 2021. Incident NBP visits were defined from ICD-10 coded as primary or first listed diagnoses on adult, family medicine, or urgent care visits separated by at least 180 days. Visit modes were classified as virtual or in-person. Periods were classified as prepandemic (before April 2020 or the beginning of the national emergency) or recovery (after June 2020). Percentages of provider orders for, and patient fulfillment of orders, were measured for 5 service classes and compared on: virtual versus in-person visits, and prepandemic versus recovery periods. Comparisons were balanced on patient case-mix using inverse probability of treatment weighting. RESULTS: Ancillary services in all 5 categories at each of the 3 Kaiser Permanente regions were substantially ordered less frequently on virtual compared with in-person visits in both the prepandemic and recovery periods (both P ≤ 0.001). Conditional on an order, patient fulfillment within 30 days was high (typically ≥70%) and not likely meaningfully different between visit modes or pandemic periods. CONCLUSIONS: Ancillary services for incident NBP visits were ordered less frequently during virtual than in-person visits in both prepandemic and recovery periods. Patient fulfillment of orders was high, and not significantly different by mode or period.

Did Access to Ambulatory Care Moderate the Associations Between Visit Mode and Ancillary Services Utilization Across the COVID-19 Pandemic Period?

BACKGROUND/OBJECTIVE: In recent years, 2 circumstances changed provider-patient interactions in primary care: the substitution of virtual (eg, video) for in-person visits and the COVID-19 pandemic. We studied whether access to care might affect patient fulfillment of ancillary services orders for ambulatory diagnosis and management of incident neck or back pain (NBP) and incident urinary tract infection (UTI) for virtual versus in-person visits. METHODS: Data were extracted from the electronic health records of 3 Kaiser Permanente Regions to identify incident NBP and UTI visits from January 2016 through June 2021. Visit modes were classified as virtual (Internet-mediated synchronous chats, telephone visits, or video visits) or in-person. Periods were classified as prepandemic [before the beginning of the national emergency (April 2020)] or recovery (after June 2020). Percentages of patient fulfillment of ancillary services orders were measured for 5 service classes each for NBP and UTI. Differences in percentages of fulfillments were compared between modes within periods and between periods within the mode to assess the possible impact of 3 moderators: distance from residence to primary care clinic, high deductible health plan (HDHP) enrollment, and prior use of a mail-order pharmacy program. RESULTS: For diagnostic radiology, laboratory, and pharmacy services, percentages of fulfilled orders were generally >70-80%. Given an incident NBP or UTI visit, longer distance to the clinic and higher cost-sharing due to HDHP enrollment did not significantly suppress patients' fulfillment of ancillary services orders. Prior use of mail-order prescriptions significantly promoted medication order fulfillments on virtual NBP visits compared with in-person NBP visits in the prepandemic period (5.9% vs. 2.0%, P=0.01) and in the recovery period (5.2% vs. 1.6%, P=0.02). CONCLUSIONS: Distance to the clinic or HDHP enrollment had minimal impact on the fulfillment of diagnostic or prescribed medication services associated with incident NBP or UTI visits delivered virtually or in-person; however, prior use of mail-order pharmacy option promoted fulfillment of prescribed medication orders associated with NBP visits.

Health-care-Related Practices in Virtual Behavioral Health Treatment for Major Depression Before and During the COVID-19 Pandemic.

BACKGROUND: The abrupt shift to virtual care at the onset of the COVID-19 pandemic had the potential to disrupt care practices in virtual behavioral health encounters. We examined changes over time in virtual behavioral health-care-related practices for patient encounters with diagnoses of major depression. METHODS: This retrospective cohort study utilized electronic health record data from 3 integrated health care systems. Inverse probability of treatment weighting was used to adjust for covariates across 3 time periods, prepandemic (January 2019-March 2020), peak-pandemic shift to virtual care (April 2020-June 2020), and recovery of health care operations (July 2020-June 2021). First virtual follow-up behavioral health department encounters after an incident diagnostic encounter were examined for differences across the time periods in rates of antidepressant medication orders and fulfillments, and completion of patient-reported symptoms screeners in service of measurement-based care. RESULTS: Antidepressant medication orders declined modestly but significantly in 2 of the 3 systems during the peak-pandemic period but rebounded during the recovery period. There were no significant changes in patient fulfillment of ordered antidepressant medications. Completion of symptom screeners increased significantly in all 3 systems during the peak-pandemic period and continued to increase significantly in the subsequent period. CONCLUSIONS: A rapid shift to virtual behavioral health care was possible without compromising health-care-related practices. The transition and subsequent adjustment period have instead been marked by improved adherence to measurement-based care practices in virtual visits, signaling a potential new capacity for virtual health care delivery.

Development of a Common Data Model for a Multisite and Multiyear Study of Virtual Visit Implementation: A Case Study.

BACKGROUND/OBJECTIVE: In multisite studies, a common data model (CDM) standardizes dataset organization, variable definitions, and variable code structures and can support distributed data processing. We describe the development of a CDM for a study of virtual visit implementation in 3 Kaiser Permanente (KP) regions. METHODS: We conducted several scoping reviews to inform our study's CDM design: (1) virtual visit mode, implementation timing, and scope (targeted clinical conditions and departments); and (2) extant sources of electronic health record data to specify study measures. Our study covered the period from 2017 through June 2021. Integrity of the CDM was assessed by a chart review of random samples of virtual and in-person visits, overall and by specific conditions of interest (neck or back pain, urinary tract infection, major depression). RESULTS: The scoping reviews identified a need to address differences in virtual visit programs across the 3 KP regionsto harmonize measurement specifications for our research analyses. The final CDM contained patient-level, provider-level, and system-level measures on 7,476,604 person-years for KP members aged 19 years and above. Utilization included 2,966,112 virtual visits (synchronous chats, telephone visits, video visits) and 10,004,195 in-person visits. Chart review indicated the CDM correctly identified visit mode on>96% (n=444) of visits, and presenting diagnosis on >91% (n=482) of visits. CONCLUSIONS: Upfront design and implementation of CDMs may be resource intensive. Once implemented, CDMs, like the one we developed for our study, provide downstream programming and analytic efficiencies by harmonizing, in a consistent framework, otherwise idiosyncratic temporal and study site differences in source data.

The Impact of Social Determinants of Health in Pediatric Shunted Hydrocephalus.

BACKGROUND: Social risk factors have been shown to negatively affect health outcomes in children. However, this has not been characterized regarding pediatric shunted hydrocephalus. OBJECTIVE: To explore the impact of social risk factors on outcomes in pediatric shunted hydrocephalus with the goal of identifying specific areas of intervention that might improve the outcomes of children undergoing cerebrospinal fluid (CSF) diversion. METHODS: In an all-payer administrative database, records between January 2010 and October 2020 were analyzed to identify children undergoing CSF shunting procedures. Children with social risk factors were compared with those without regarding rates of infection, shunt-related interventions, and mortality within 5 years of their shunting procedure. RESULTS: Among the 5420 children who underwent first-time shunting procedures, 263 (4.9%) were identified to have social risk factors. Children with identified social risk factors had increased odds of central nervous system infection (odds ratio [OR] 2.06, 95% CI 1.45-2.91), revision (OR 2.43, 95% I 1.89-3.12), and mortality (OR 2.86, 95% CI 1.23-5.72). The mean numbers of computed tomography studies (14.60 ± 17.78 vs 6.34 ± 6.10), MRI studies (18.76 ± 24.37 vs 7.88 ± 24.37), and shunt series X-rays (17.22 ± 19.04 vs 7.66 ± 8.54) were increased among children with social risk factors. CONCLUSION: Children with social risk factors had increased rates of central nervous system infection, shunt-related interventions, and mortality within 5 years of CSF shunting. We underscore the importance of characterizing the impact of social risk factors in specific conditions, such as pediatric shunted hydrocephalus, and look to future directions aimed to mitigate these risk factors with coordination and direction of individualized resources, encouragement of advocacy, and community partnership.

The association between buprenorphine treatment duration and mortality: a multi-site cohort study of people who discontinued treatment.

BACKGROUND AND AIMS: Buprenorphine is an effective medication for opioid use disorder that reduces mortality; however, many patients are not retained in buprenorphine treatment, and an optimal length of treatment after which patients can safely discontinue treatment has not been identified. This study measured the association between buprenorphine treatment duration and all-cause mortality among patients who discontinued treatment. Secondary objectives were to measure the association between treatment duration and drug overdose and opioid-related overdoses. DESIGN: Multi-site cohort study. SETTING: Eight US health systems. PARTICIPANTS: Patients who initiated and discontinued buprenorphine treatment between 1 January 2012 and 31 December 2018 (n = 6550). Outcomes occurring after patients discontinued buprenorphine treatment were compared between patients who initiated and discontinued treatment after 8-30, 31-90, 91-180, 181-365 and > 365 days. MEASUREMENTS: Covariate data were obtained from electronic health records (EHRs). Mortality outcomes were derived from EHRs and state vital statistics. Non-fatal opioid and drug overdoses were obtained from diagnostic codes. Four sites provided cause-of-death data to identify fatal drug and opioid-related overdoses. Adjusted frailty regression was conducted on a propensity-weighted cohort to assess associations between duration of the final treatment episode and outcomes. FINDINGS: The mortality rate after buprenorphine treatment was 1.82 per 100 person-years (n = 191 deaths). In regression analyses with > 365 days as the reference group, treatment duration was not associated with all-cause mortality and drug overdose (P > 0.05 for both). However, compared with > 365 days of treatment, 91-180 days of treatment was associated with increased opioid overdose risk (hazard ratio = 2.94, 95% confidence interval = 1.11-7.79). CONCLUSIONS: Among patients who discontinue buprenorphine treatment, there appears to be no treatment duration period associated with a reduced risk for all-cause mortality. Patients who discontinue buprenorphine treatment after 91-180 days appear to be at heightened risk for opioid overdose compared with patients who discontinue after > 365 days of treatment.

Development and implementation of a prescription opioid registry across diverse health systems.

Objective: Develop and implement a prescription opioid registry in 10 diverse health systems across the US and describe trends in prescribed opioids between 2012 and 2018. Materials and Methods: Using electronic health record and claims data, we identified patients who had an outpatient fill for any prescription opioid, and/or an opioid use disorder diagnosis, between January 1, 2012 and December 31, 2018. The registry contains distributed files of prescription opioids, benzodiazepines and other select medications, opioid antagonists, clinical diagnoses, procedures, health services utilization, and health plan membership. Rates of outpatient opioid fills over the study period, standardized to health system demographic distributions, are described by age, gender, and race/ethnicity among members without cancer. Results: The registry includes 6 249 710 patients and over 40 million outpatient opioid fills. For the combined registry population, opioid fills declined from a high of 0.718 per member-year in 2013 to 0.478 in 2018, and morphine milligram equivalents (MMEs) per fill declined from 985 MMEs per fill in 2012 to 758 MMEs in 2018. MMEs per member declined from 692 MMEs per member in 2012 to 362 MMEs per member in 2018. Conclusion: This study established a population-based opioid registry across 10 diverse health systems that can be used to address questions related to opioid use. Initial analyses showed large reductions in overall opioid use per member among the combined health systems. The registry will be used in future studies to answer a broad range of other critical public health issues relating to prescription opioid use.

Time-series analysis of recent antihyperglycemic medication prescribing trends for a diverse sample of Medicare enrollees with type 2 diabetes mellitus in an integrated health system.

PURPOSE: Despite high type 2 diabetes mellitus (T2DM) prevalence in Medicare enrollees, newer therapeutic options, and revised treatment guidelines, little is known about US antihyperglycemic prescribing trends after 2015. This research describes recent monthly antihyperglycemic prescribing trends in a large, diverse population of Medicare enrollees from the US Mid-Atlantic region. METHODS: Encounter data (July 2018-July 2020) for Medicare enrollees 65 years of age or older with T2DM were extracted from electronic health records of a large integrated health system. Descriptive time-series regression models were estimated to describe monthly prescribing rates (ie, prescription orders per 100 eligible plan members with T2DM) overall and by medication subgroups for all-eligible and continuously-eligible samples. Trends in monthly prescription orders per 100 eligible plan members with T2DM were reported. RESULTS: The monthly all-eligible member sample (n > 22,000) exhibited an overall positive baseline monthly prescribing rate of 23.88 T2DM medication orders per 100 members with T2DM and a significant positive monthly prescribing rate trend (ie, change) of 0.12 T2DM medication orders per 100 members with T2DM (P < 0.05). Subgroup T2DM medication order rates per 100 members with T2DM at baseline were 16.28 for first-generation medications, 3.87 for human insulins, 3.04 for insulin analogs, 0.58 for second-generation medications, and 0.11 for combination medications. Human insulins, insulin analogs, and second-generation medications had positive monthly trends (P < 0.05). Among second-generation medications, sodium-glucose cotransporter 2 (SGLT-2) inhibitors and glucagon-like peptide 1 (GLP-1) agonists had positive monthly trends (P < 0.05). Continuously eligible members with T2DM (n = 19,185) had no significant overall monthly prescribing trend; however, human insulins, insulin analogs, and second-generation medications and the SGLT-2 inhibitor class had positive monthly prescribing trends (P < 0.05). CONCLUSION: In a diverse Medicare sample, this study observed increasing monthly trends for second-generation medications, human insulins, and insulin analogs consistent with emerging evidence. Among second-generation medications, SGLT-2 inhibitors became the most commonly prescribed over time.

Evaluation of a Natural Language Processing Approach to Identify Social Determinants of Health in Electronic Health Records in a Diverse Community Cohort.

BACKGROUND: Health care systems in the United States are increasingly interested in measuring and addressing social determinants of health (SDoH). Advances in electronic health record systems and Natural Language Processing (NLP) create a unique opportunity to systematically document patient SDoH from digitized free-text provider notes. METHODS: Patient SDoH status [recorded by Your Current Life Situation (YCLS) Survey] and associated provider notes recorded between March 2017 and June 2020 were extracted (32,261 beneficiaries; 50,722 YCLS surveys; 485,425 provider notes).NLP patterns were generated using a machine learning test statistic (Term Frequency-Inverse Document Frequency). Patterns were developed and assessed in a training, training validation, and final validation dataset (64%, 16%, and 20% of total data, respectively).NLP models analyzed SDoH-specific categories (housing, medical care, and transportation needs) and a combined SDoH metric. Model performance was assessed using sensitivity, specificity, and Cohen κ statistic, assuming the YCLS Survey to be the gold standard. RESULTS: Within the training validation dataset, NLP models showed strong sensitivity and specificity, with moderate agreement with the YCLS Survey (Housing: sensitivity=0.67, specificity=0.89, κ=0.51; Medical care: sensitivity=0.55, specificity=0.73, κ=0.20; Transportation: sensitivity=0.79, specificity=0.87, κ=0.58). Model performance in the training and training validation datasets were comparable.In the final validation dataset, a combined SDoH prediction metric showed sensitivity=0.77, specificity=0.69, κ=0.45. CONCLUSION: This NLP algorithm demonstrated moderate performance in identification of unmet patient social needs. This novel approach may enable improved targeting of interventions, allocation of limited resources and monitoring a health care system's addressing its patients' SDoH needs.

Physician adjudication of angioedema diagnosis codes in a population of patients with heart failure prescribed angiotensin-converting enzyme inhibitor therapy.

PURPOSE: Our objective was to calculate the positive predictive value (PPV) of the ICD-9 diagnosis code for angioedema when physicians adjudicate the events by electronic health record review. Our secondary objective was to evaluate the inter-rater reliability of physician adjudication. METHODS: Patients from the Cardiovascular Research Network previously diagnosed with heart failure who were started on angiotensin-converting enzyme inhibitors (ACEI) during the study period (July 1, 2006 through September 30, 2015) were included. A team of two physicians per participating site adjudicated possible events using electronic health records for all patients coded for angioedema for a total of five sites. The PPV was calculated as the number of physician-adjudicated cases divided by all cases with the diagnosis code of angioedema (ICD-9-CM code 995.1) meeting the inclusion criteria. The inter-rater reliability of physician teams, or kappa statistic, was also calculated. RESULTS: There were 38 061 adults with heart failure initiating ACEI in the study (21 489 patient-years). Of 114 coded events that were adjudicated by physicians, 98 angioedema events were confirmed for a PPV of 86% (95% CI: 80%, 92%). The kappa statistic based on physician inter-rater reliability was 0.65 (95% CI: 0.47, 0.82). CONCLUSIONS: ICD-9 diagnosis code of 995.1 (angioneurotic edema, not elsewhere classified) is highly predictive of angioedema in adults with heart failure exposed to ACEI.

Development and Validation of a Healthcare Utilization-Based Algorithm to Identify Acute Exacerbations of Chronic Obstructive Pulmonary Disease.

Introduction: Acute exacerbations of chronic obstructive pulmonary disease (AECOPD) are important events that may precipitate other adverse outcomes. Accurate AECOPD event identification in electronic administrative data is essential for improving population health surveillance and practice management. Objective: Develop codified algorithms to identify moderate and severe AECOPD in two US healthcare systems using administrative data and electronic medical records, and validate their performance by calculating positive predictive value (PPV) and negative predictive value (NPV). Methods: Data from two large regional integrated health systems were used. Eligible patients were identified using International Classification of Diseases (Ninth Edition) COPD diagnosis codes. Two algorithms were developed: one to identify potential moderate AECOPD by selecting outpatient/emergency visits associated with AECOPD-related codes and antibiotic/systemic steroid prescriptions; the other to identify potential severe AECOPD by selecting inpatient visits associated with corresponding codes. Algorithms were validated via patient chart review, adjudicated by a pulmonologist. To estimate PPV, 300 potential moderate AECOPD and 250 potential severe AECOPD events underwent review. To estimate NPV, 200 patients without any AECOPD identified by the algorithms (100 patients each without moderate or severe AECOPD) during the two years following the index date underwent review to identify AECOPD missed by the algorithm (false negatives). Results: The PPVs (95% confidence interval [CI]) for both moderate and severe AECOPD were high: 293/298 (98.3% [96.1-99.5]) and 216/225 (96.0% [92.5-98.2]), respectively. NPV was lower for moderate AECOPD (75.0% [65.3-83.1]) than for severe AECOPD (95.0% [88.7-98.4]). Results were consistent across both healthcare systems. Conclusion: This study developed healthcare utilization-based algorithms to identify moderate and severe AECOPD in two separate healthcare systems. PPV for both algorithms was high; NPV was lower for the moderate algorithm. Replication and consistency of results across two healthcare systems support the external validity of these findings.

Pollen Exposure and Associated Healthcare Utilization: A Population-based Study Using Health Maintenance Organization Data in the Washington, DC, Area.

Rationale: Most studies of the healthcare utilization impact of pollen exposure have focused on emergency department visits or hospital admissions. However, other frequent but lower cost services-phone calls and e-mails to providers and office visits-may also be affected. Objectives: The objective of our study was to estimate the impact of tree and grass pollen exposures on respiratory-related healthcare utilization across a range of medical services, including calls and e-mails to providers, nonurgent face-to-face visits, urgent and emergent care visits, and hospitalizations. Methods: We conducted a retrospective observational study of daily tree and grass pollen counts linked to electronic health records of Kaiser Permanente beneficiaries in the metropolitan Washington, DC, area for 2013-2014. Results: The proportion of Kaiser Permanente beneficiaries with respiratory-related healthcare utilization was significantly greater (for P ⩽ 0.05) given a 1 standard deviation increase in same-day pollen exposure. For tree pollen, a 1 standard deviation increase in same-day pollen exposure was associated with relative increases in utilization ranging from 1.77% (95% confidence interval [CI], 0.07-4.17%) for urgent and emergent care visits to 12.84% (95% CI, 11.02-14.65%) for provider calls/e-mails. For grass pollen exposure, a 1 standard deviation increase in same-day pollen exposure was associated with relative increases in utilization ranging from 1.42% (95% CI, 0.39-2.46) for provider face-to-face visits to 11.09% (95% CI, 9.26-12.92) for provider calls/e-mails. Conclusions: Increased pollen exposure was associated with increases in healthcare utilization across a range of services, with relatively higher increases in provider calls/e-mails and lower increases in emergent or acute care. If climate change increases intensity and geographic scope of pollen exposure as predicted and if this study's estimates of association of peak pollen exposure on healthcare utilization are generalizable, then the impact of climate change on healthcare utilization may be significant.