A COVID-19 Inpatient Risk Calculator (CIRC) based on 24 variables known to be associated with COVID-19 found that four lab measurements at admission—absolute lymphocyte count, albumin, cardiac troponin, and C-reactive protein—were among the inputs most predictive of in-hospital disease progression (Ann Intern Med 2020; doi:10.7326/M20-3905). Other highly predictive variables included age, nursing home residence, comorbid conditions, obesity, respiratory systems, respiratory rate, and fever. This information could help hospitals make important decisions about planning and resource allocations for COVID-19 care, according to the researchers.

Researchers at Johns Hopkins University developed CIRC based on data from patients with confirmed SARS-CoV-2 infection at five Johns Hopkins Medicine hospitals in Maryland and Washington, D.C. The study involved 832 consecutive patients admitted from March 4, 2020, to April 24, 2020, with data on their hospitalizations fed into JH-CROWN, a COVID-19 registry, which utilizes the Johns Hopkins precision medicine analytics platform.

The authors sought to determine the factors at admission most predictive of severe disease or death from COVID-19, as categorized by the World Health Organization disease severity scale. CIRC incorporated demographic data, comorbid conditions, vital signs, presenting symptoms, and 20 laboratory values.

Overall, 16% of patients died, while 63% had mild-to-moderate disease, and 20% had severe disease. Of patients admitted with mild-to-moderate disease, 38% progressed to severe disease or death, 60% within 2 days and 79% within 4 days.

CIRC had an area under the receiver operating characteristic curve to predict in hospital admission of 0.85, 0.79, and 0.79 at day 2, 4, and 7, respectively. Different combinations of risk factors predicted disease or death probabilities ranging from more than 90% to 5%.

The authors noted several lab-related limitations about the study. First, collection of key lab values was not standardized across Johns Hopkins Medicine or at individual hospitals, contributing to missing data. In addition, testing challenges might have resulted in not all COVID-19 cases being captured. Finally, respiratory virus panel testing was not available on all patients, so the model does not account for co-viral infections that might have altered patients’ disease trajectories.

CIRC is available online to support providers in assessing their patients’ risk of worsening disease: https://rsconnect.biostat.jhsph.edu/covid_predict/.

Targeted Genetic Testing Found Cost-Effective in Newly Diagnosed GIST

Targeted genetic testing is cost-effective for patients newly diagnosed with metastatic gastrointestinal stromal tumors (GIST) (JAMA Network Open 2020;3:e2013565). This finding supports widespread adoption of genetic testing for GIST, according to the authors.

Though still rare, GIST is the most common sarcoma, according to the investigators. Oncologists tend to prescribe imatinib for all patients with metastatic GIST, but this small molecule therapy provokes variable responses depending on patients’ KIT variations. Patients also develop primary and acquired secondary resistance to imatinib. Guidelines recommend genetic testing for KIT variants in GIST to ensure imatinib therapy starts with the optimal dose; however, just 15% to 33% of patients actually undergo such testing, perhaps due to concerns about its cost and utility.

The authors developed a Markov model to compare the cost-effectiveness of targeted gene testing and variation-directed first-line therapy, versus empiric imatinib therapy. They explored outcomes for three genomic populations: KIT exon 11, KIT exon 9, and all other variations. The model also simulated treatment outcomes associated with first-line, second-line, and third-line therapies.

The authors determined the cost of targeted gene therapy based on Medicare claims data for multigene next-generation sequencing diagnostic tests.

Aside from cost, the other primary outcome was quality-adjusted life years (QALYs). The authors deducted 0.12 QALYs for each disease progression, represented by first-line, second-line, or third-line treatments, but did not deduct QALYs for any toxic effects from imatinib or sunitinib therapy. The model incorporated a willingness-to-pay threshold of $100,000 per QALY, with treatments less than this threshold deemed cost-effective.

The investigators determined that targeted gene therapy would increase QALYs by 0.10 at a cost increase of $9,513 compared with empiric imatinib therapy, for a $92,100 incremental cost-effectiveness ratio. A therapy-directed approach would remain cost-effective until genetic testing costs amounted to $3,730. A probabilistic sensitivity analysis found that this approach would be cost-effective 70% of the time.

1-time Cardiac Troponin Test Useful in Assessing Mortality, Cardiovascular Disease Risk in Patients With Diabetes

Subclinical levels of cardiac troponin I and T measured with high-sensitivity assays (hs-cTn I/T) in a middle-aged population of people with diabetes are “robustly associated” with long-term mortality and cardiovascular disease (CVD) risk (Diabetes Care 2020;43:e144-6). These findings suggest that a single measurement of either analyte in middle-aged patients could be used to risk-stratify these individuals to help guide their clinical management, according to the investigators.

The authors included data from 1,704 participants in the Atherosclerosis Risk in Communities (ARIC) study, who were between the ages of 54 and 75, attended ARIC visit 4 (between 1996 and 1998), and had diabetes, as reported by physician diagnosis, medication use, or blood glucose level ≥126 mg/dL (fasting) or ≥200 mg/dL (nonfasting).

The investigators measured cTn I or cTn T from stored plasma drawn during ARIC visit 4 and examined incident CVD events through the last date of follow-up in 2018. They found 1,102 deaths in this cohort, of which 443 were CVD-related. The cumulative mortality was 37.4% and 44.6% in patients with hs-cTn I ≥90th percentile and no CVD, and hs-cTn T ≥90th percentile and no CVD, respectively. Meanwhile, the cumulative mortality was 43.1% in participants who had preexisting CVD.

After they adjusted for traditional risk factors, the authors found both cTn I and cTn T were “independently associated with and significantly improved model discrimination” for all-cause and CVD-related mortality risk.