Value of quantitative airspace disease measured on chest CT and chest radiography at initial diagnosis compared to clinical variables for prediction of severe COVID-19.

Purpose: Rapid prognostication of COVID-19 patients is important for efficient resource allocation. We evaluated the relative prognostic value of baseline clinical variables (CVs), quantitative human-read chest CT (qCT), and AI-read chest radiograph (qCXR) airspace disease (AD) in predicting severe COVID-19. Approach: We retrospectively selected 131 COVID-19 patients (SARS-CoV-2 positive, March to October, 2020) at a tertiary hospital in the United States, who underwent chest CT and CXR within 48 hr of initial presentation. CVs included patient demographics and laboratory values; imaging variables included qCT volumetric percentage AD (POv) and qCXR area-based percentage AD (POa), assessed by a deep convolutional neural network. Our prognostic outcome was need for ICU admission. We compared the performance of three logistic regression models: using CVs known to be associated with prognosis (model I), using a dimension-reduced set of best predictor variables (model II), and using only age and AD (model III). Results: 60/131 patients required ICU admission, whereas 71/131 did not. Model I performed the poorest ( AUC = 0.67 [0.58 to 0.76]; accuracy = 77 % ). Model II performed the best ( AUC = 0.78 [0.71 to 0.86]; accuracy = 81 % ). Model III was equivalent ( AUC = 0.75 [0.67 to 0.84]; accuracy = 80 % ). Both models II and III outperformed model I ( AUC difference = 0.11 [0.02 to 0.19], p = 0.01 ; AUC difference = 0.08 [0.01 to 0.15], p = 0.04 , respectively). Model II and III results did not change significantly when POv was replaced by POa. Conclusions: Severe COVID-19 can be predicted using only age and quantitative AD imaging metrics at initial diagnosis, which outperform the set of CVs. Moreover, AI-read qCXR can replace qCT metrics without loss of prognostic performance, promising more resource-efficient prognostication.

Real-World Lung Cancer CT Screening Performance, Smoking Behavior, and Adherence to Recommendations: Lung-RADS Category and Smoking Status Predict Adherence.

BACKGROUND. Low-dose CT (LDCT) lung cancer screening (LCS) has been shown to decrease mortality in persons with a significant smoking history. However, adherence in real-world LCS programs is significantly lower than in randomized controlled trials. OBJECTIVE. The purpose of this article is to assess real-world LDCT LCS performance and factors predictive of adherence to LCS recommendations. METHODS. We retrospectively identified all persons who underwent at least two LCS examinations from 2014 to 2019. Patient demographics, smoking history and behavior changes, Lung-RADS category, PPV, NPV, and adherence to screening recommendations were recorded. Predictors of adherence were assessed via univariate comparisons and multivariate logistic regression. RESULTS. A total of 260 persons returned for follow-up LDCT (57.7% had two, 34.2% had three, 7.7% had four, and 0.4% had five LDCT examinations). A total of 43 of 260 (16.5%) had positive (Lung-RADS category 3 or above) scans, of which 27 of 260 persons (10.3%) were graded as Lung-RADS category 3, eight of 260 (3.1%) were category 4A, six of 260 (2.3%) were category 4B, and two of 260 (0.8%) were category 4X. Cancer was diagnosed in four of the 260 (three with lung cancer and one with metastatic melanoma). A total of 143 of 260 (55.0%) persons were current smokers at baseline and 121 of 260 (46.5%) were current smokers at the last round of LCS. LCS had sensitivity of 100.0%, specificity of 84.8%, PPV of 9.3%, and NPV of 100%. Overall adherence was 43.0% but increased progressively with higher Lung-RADS category (Lung-RADS 1: 33.2%; Lung-RADS 2: 46.3%; Lung-RADS 3: 53.8%; Lung-RADS 4A: 77.8%; Lung-RADS 4B: 83.3%; Lung-RADS 4X: 100%; p < .001). was also higher in former versus current smokers (50.0% vs 36.2%; p < .001). Being a former smoker and having a nodule that is Lung-RADS category 3 or greater were the only significant independent predictors of adherence. CONCLUSION. Our real-world LCS program showed very high sensitivity and NPV, but moderate specificity and very low PPV. Adherence to LCS recommendations increased with former versus current smokers and in those with positive (Lung-RADS categories 3, 4A, 4B, or 4X) LCS examinations. Adherence was less than 50.0% in current smokers and persons with negative (Lung-RADS categories 1 or 2) LCS examinations. CLINICAL IMPACT. Our results offer a road map for targeted performance improvement by focusing on LCS subjects less likely to remain in the program, such as persons with negative LCS examinations and persons who continue to smoke, potentially improving LCS cost effectiveness and maximizing its societal benefits.

Changes in Concurrent Opioid and Benzodiazepine Use Following a Low-Touch Prescriber Fax Intervention.

BACKGROUND: Concurrent use of opioids and benzodiazepines (COB) can lead to additive respiratory and central nervous system effects, putting patients at increased risk of fatal overdose. In 2016, the Centers for Disease Control and Prevention released an opioid-prescribing guideline recommending against COB, and the Pharmacy Quality Alliance (PQA) endorsed a COB measure in its core opioid set. From May 1, 2017, to December 4, 2017, a California Medicaid plan launched a COB-focused prescriber outreach intervention for members receiving recent opioid and benzodiazepine claims with the intent of decreasing concurrent use. OBJECTIVE: To assess the effect of a prescriber fax intervention by a Medicaid plan on COB. METHODS: Two retrospective analyses were conducted using administrative pharmacy claims data: a comparison of the PQA COB rate among selected California Medicaid plans for 2016 and 2017 and a cohort utilization analysis of members identified for the fax intervention compared with controls. Intervention and control members were matched based on 12 pre-index utilization characteristics. Outcomes assessed included proportion of members with resolution of COB in the post-index period, change in mean number of COB days before and after the index date, and proportion of members with decreased benzodiazepine daily dose after the index date. Analyses were also performed for the subgroups of members with < 30 days of COB and ≥ 30 days of COB in the pre-index period. RESULTS: All California Medicaid plans in the study saw an improvement in the PQA COB rate between 2016 and 2017. In the utilization analysis, 4,182 intervention members were eligible according to study criteria and matched to similar control members. Many differences in medication use existed between the subgroups with < 30 days and ≥ 30 days of COB in the pre-index period, with the latter group consisting of much more chronic, complex users. The intervention cohort had a statistically significant higher proportion of members with complete resolution of COB compared with the control cohort (43.8% vs. 40.0%; P < 0.01), which was also statistically significant for the 2 subgroups. The intervention cohort had a decrease in the mean number of COB days from pre- to post-index periods, but this was only statistically significant for the subgroup with < 30 COB days (-2.5 vs. -1.5; P = 0.0217). No statistically significant differences were detected between cohorts in proportion of members with decreased benzodiazepine dose. CONCLUSIONS: Our analyses demonstrated that this low-touch prescriber fax intervention produced statistically significant improvements in COB outcomes, despite the overall trend of declining COB among the other California Medicaid plans. Low-touch, targeted prescriber outreach can be an inexpensive yet effective tool to affect prescriber behavior, particularly before COB becomes chronic. DISCLOSURES: No outside funding was used to support this study. The authors do not have any financial relationships or potential conflicts of interest relevant to this article to disclose. At the time of conducting this research, all authors were employees of MedImpact Healthcare Systems. The results of this study were presented at the AMCP Managed Care & Specialty Pharmacy Annual Meeting 2019; March 25-28, 2019; San Diego, CA.

Selection bias and subject refusal in a cluster-randomized controlled trial.

BACKGROUND: Selection bias and non-participation bias are major methodological concerns which impact external validity. Cluster-randomized controlled trials are especially prone to selection bias as it is impractical to blind clusters to their allocation into intervention or control. This study assessed the impact of selection bias in a large cluster-randomized controlled trial. METHODS: The Improved Cardiovascular Risk Reduction to Enhance Rural Primary Care (ICARE) study examined the impact of a remote pharmacist-led intervention in twelve medical offices. To assess eligibility, a standardized form containing patient demographics and medical information was completed for each screened patient. Eligible patients were approached by the study coordinator for recruitment. Both the study coordinator and the patient were aware of the site's allocation prior to consent. Patients who consented or declined to participate were compared across control and intervention arms for differing characteristics. Statistical significance was determined using a two-tailed, equal variance t-test and a chi-square test with adjusted Bonferroni p-values. Results were adjusted for random cluster variation. RESULTS: There were 2749 completed screening forms returned to research staff with 461 subjects who had either consented or declined participation. Patients with poorly controlled diabetes were found to be significantly more likely to decline participation in intervention sites compared to those in control sites. A higher mean diastolic blood pressure was seen in patients with uncontrolled hypertension who declined in the control sites compared to those who declined in the intervention sites. However, these findings were no longer significant after adjustment for random variation among the sites. After this adjustment, females were now found to be significantly more likely to consent than males (odds ratio = 1.41; 95% confidence interval = 1.03, 1.92). CONCLUSIONS: Though there appeared to be a higher consent rate for females than for males, the overall impact of potential selection bias and refusal to participate was minimal. Without rigorous methodology, selection bias may be a threat to external validity in cluster-randomized trials. TRIAL REGISTRATION: NCT01983813 . Date of registration: Oct. 28, 2013.