Development and validation of a deep-learning model to predict 10-year atherosclerotic cardiovascular disease risk from retinal images using the UK Biobank and EyePACS 10K datasets.

Background: Atherosclerotic cardiovascular disease (ASCVD) is a leading cause of death globally, and early detection of high-risk individuals is essential for initiating timely interventions. The authors aimed to develop and validate a deep learning (DL) model to predict an individual's elevated 10-year ASCVD risk score based on retinal images and limited demographic data. Methods: The study used 89,894 retinal fundus images from 44,176 UK Biobank participants (96% non-Hispanic White, 5% diabetic) to train and test the DL model. The DL model was developed using retinal images plus age, race/ethnicity, and sex at birth to predict an individual's 10-year ASCVD risk score using the pooled cohort equation (PCE) as the ground truth. This model was then tested on the US EyePACS 10K dataset (5.8% non-Hispanic White, 99.9% diabetic), composed of 18,900 images from 8969 diabetic individuals. Elevated ASCVD risk was defined as a PCE score of ≥7.5%. Results: In the UK Biobank internal validation dataset, the DL model achieved an area under the receiver operating characteristic curve of 0.89, sensitivity 84%, and specificity 90%, for detecting individuals with elevated ASCVD risk scores. In the EyePACS 10K and with the addition of a regression-derived diabetes modifier, it achieved sensitivity 94%, specificity 72%, mean error -0.2%, and mean absolute error 3.1%. Conclusion: This study demonstrates that DL models using retinal images can provide an additional approach to estimating ASCVD risk, as well as the value of applying DL models to different external datasets and opportunities about ASCVD risk assessment in patients living with diabetes.

Utilizing multimodal AI to improve genetic analyses of cardiovascular traits.

Electronic health records, biobanks, and wearable biosensors contain multiple high-dimensional clinical data (HDCD) modalities (e.g., ECG, Photoplethysmography (PPG), and MRI) for each individual. Access to multimodal HDCD provides a unique opportunity for genetic studies of complex traits because different modalities relevant to a single physiological system (e.g., circulatory system) encode complementary and overlapping information. We propose a novel multimodal deep learning method, M-REGLE, for discovering genetic associations from a joint representation of multiple complementary HDCD modalities. We showcase the effectiveness of this model by applying it to several cardiovascular modalities. M-REGLE jointly learns a lower representation (i.e., latent factors) of multimodal HDCD using a convolutional variational autoencoder, performs genome wide association studies (GWAS) on each latent factor, then combines the results to study the genetics of the underlying system. To validate the advantages of M-REGLE and multimodal learning, we apply it to common cardiovascular modalities (PPG and ECG), and compare its results to unimodal learning methods in which representations are learned from each data modality separately, but the downstream genetic analyses are performed on the combined unimodal representations. M-REGLE identifies 19.3% more loci on the 12-lead ECG dataset, 13.0% more loci on the ECG lead I + PPG dataset, and its genetic risk score significantly outperforms the unimodal risk score at predicting cardiac phenotypes, such as atrial fibrillation (Afib), in multiple biobanks.

Cost-Utility Analysis of Deep Learning and Trained Human Graders for Diabetic Retinopathy Screening in a Nationwide Program.

INTRODUCTION: Deep learning (DL) for screening diabetic retinopathy (DR) has the potential to address limited healthcare resources by enabling expanded access to healthcare. However, there is still limited health economic evaluation, particularly in low- and middle-income countries, on this subject to aid decision-making for DL adoption. METHODS: In the context of a middle-income country (MIC), using Thailand as a model, we constructed a decision tree-Markov hybrid model to estimate lifetime costs and outcomes of Thailand's national DR screening program via DL and trained human graders (HG). We calculated the incremental cost-effectiveness ratio (ICER) between the two strategies. Sensitivity analyses were performed to probe the influence of modeling parameters. RESULTS: From a societal perspective, screening with DL was associated with a reduction in costs of ~ US$ 2.70, similar quality-adjusted life-years (QALY) of + 0.0043, and an incremental net monetary benefit of ~ US$ 24.10 in the base case. In sensitivity analysis, DL remained cost-effective even with a price increase from US$ 1.00 to US$ 4.00 per patient at a Thai willingness-to-pay threshold of ~ US$ 4.997 per QALY gained. When further incorporating recent findings suggesting improved compliance to treatment referral with DL, our analysis models effectiveness benefits of ~ US$ 20 to US$ 50 depending on compliance. CONCLUSION: DR screening using DL in an MIC using Thailand as a model may result in societal cost-savings and similar health outcomes compared with HG. This study may provide an economic rationale to expand DL-based DR screening in MICs as an alternative solution for limited availability of skilled human resources for primary screening, particularly in MICs with similar prevalence of diabetes and low compliance to referrals for treatment.

Artificial Intelligence-Enabled Assessment of the Heart Rate Corrected QT Interval Using a Mobile Electrocardiogram Device.

BACKGROUND: Heart rate-corrected QT interval (QTc) prolongation, whether secondary to drugs, genetics including congenital long QT syndrome, and/or systemic diseases including SARS-CoV-2-mediated coronavirus disease 2019 (COVID-19), can predispose to ventricular arrhythmias and sudden cardiac death. Currently, QTc assessment and monitoring relies largely on 12-lead electrocardiography. As such, we sought to train and validate an artificial intelligence (AI)-enabled 12-lead ECG algorithm to determine the QTc, and then prospectively test this algorithm on tracings acquired from a mobile ECG (mECG) device in a population enriched for repolarization abnormalities. METHODS: Using >1.6 million 12-lead ECGs from 538 200 patients, a deep neural network (DNN) was derived (patients for training, n = 250 767; patients for testing, n = 107 920) and validated (n = 179 513 patients) to predict the QTc using cardiologist-overread QTc values as the "gold standard". The ability of this DNN to detect clinically-relevant QTc prolongation (eg, QTc ≥500 ms) was then tested prospectively on 686 patients with genetic heart disease (50% with long QT syndrome) with QTc values obtained from both a 12-lead ECG and a prototype mECG device equivalent to the commercially-available AliveCor KardiaMobile 6L. RESULTS: In the validation sample, strong agreement was observed between human over-read and DNN-predicted QTc values (-1.76±23.14 ms). Similarly, within the prospective, genetic heart disease-enriched dataset, the difference between DNN-predicted QTc values derived from mECG tracings and those annotated from 12-lead ECGs by a QT expert (-0.45±24.73 ms) and a commercial core ECG laboratory [10.52±25.64 ms] was nominal. When applied to mECG tracings, the DNN's ability to detect a QTc value ≥500 ms yielded an area under the curve, sensitivity, and specificity of 0.97, 80.0%, and 94.4%, respectively. CONCLUSIONS: Using smartphone-enabled electrodes, an AI DNN can predict accurately the QTc of a standard 12-lead ECG. QTc estimation from an AI-enabled mECG device may provide a cost-effective means of screening for both acquired and congenital long QT syndrome in a variety of clinical settings where standard 12-lead electrocardiography is not accessible or cost-effective.

Development and Validation of a Deep-Learning Model to Screen for Hyperkalemia From the Electrocardiogram.

Importance: For patients with chronic kidney disease (CKD), hyperkalemia is common, associated with fatal arrhythmias, and often asymptomatic, while guideline-directed monitoring of serum potassium is underused. A deep-learning model that enables noninvasive hyperkalemia screening from the electrocardiogram (ECG) may improve detection of this life-threatening condition. Objective: To evaluate the performance of a deep-learning model in detection of hyperkalemia from the ECG in patients with CKD. Design, Setting, and Participants: A deep convolutional neural network (DNN) was trained using 1 576 581 ECGs from 449 380 patients seen at Mayo Clinic, Rochester, Minnesota, from 1994 to 2017. The DNN was trained using 2 (leads I and II) or 4 (leads I, II, V3, and V5) ECG leads to detect serum potassium levels of 5.5 mEq/L or less (to convert to millimoles per liter, multiply by 1) and was validated using retrospective data from the Mayo Clinic in Minnesota, Florida, and Arizona. The validation included 61 965 patients with stage 3 or greater CKD. Each patient had a serum potassium count drawn within 4 hours after their ECG was recorded. Data were analyzed between April 12, 2018, and June 25, 2018. Exposures: Use of a deep-learning model. Main Outcomes and Measures: Area under the receiver operating characteristic curve (AUC) and sensitivity and specificity, with serum potassium level as the reference standard. The model was evaluated at 2 operating points, 1 for equal specificity and sensitivity and another for high (90%) sensitivity. Results: Of the total 1 638 546 ECGs, 908 000 (55%) were from men. The prevalence of hyperkalemia in the 3 validation data sets ranged from 2.6% (n = 1282 of 50 099; Minnesota) to 4.8% (n = 287 of 6011; Florida). Using ECG leads I and II, the AUC of the deep-learning model was 0.883 (95% CI, 0.873-0.893) for Minnesota, 0.860 (95% CI, 0.837-0.883) for Florida, and 0.853 (95% CI, 0.830-0.877) for Arizona. Using a 90% sensitivity operating point, the sensitivity was 90.2% (95% CI, 88.4%-91.7%) and specificity was 63.2% (95% CI, 62.7%-63.6%) for Minnesota; the sensitivity was 91.3% (95% CI, 87.4%-94.3%) and specificity was 54.7% (95% CI, 53.4%-56.0%) for Florida; and the sensitivity was 88.9% (95% CI, 84.5%-92.4%) and specificity was 55.0% (95% CI, 53.7%-56.3%) for Arizona. Conclusions and Relevance: In this study, using only 2 ECG leads, a deep-learning model detected hyperkalemia in patients with renal disease with an AUC of 0.853 to 0.883. The application of artificial intelligence to the ECG may enable screening for hyperkalemia. Prospective studies are warranted.

Impact of hormone therapy on Medicare spending in the Women's Health Initiative randomized clinical trials.

BACKGROUND: Randomized trials can compare economic as well as clinical outcomes, but economic data are difficult to collect. Linking clinical trial data with Medicare claims could provide novel information on health care utilization and cost. METHODS: We linked data from Medicare claims of women ≥65 years old who had Medicare fee-for-service coverage with their clinical data from the Women's Health Initiative trials of conjugated equine estrogens plus medroxyprogesterone acetate (CEE+MPA) versus placebo and of CEE-alone versus placebo. The primary outcome was total Medicare spending during the intervention phase of the trial, and the secondary outcomes were spending on diseases hypothesized a priori to be sensitive to the effects of hormone therapy. RESULTS: In the CEE+MPA trial, 4,557 participants ≥65 years old were included. Women randomly assigned to CEE+MPA had 4% higher mean Medicare spending overall ($45,690 vs $43,920, P = .08) but 0.5% lower spending for hormone-sensitive diseases ($3,526 vs $3,547, P = .07), with 73% higher spending for coronary heart disease (P = .045) and 122% higher spending for pulmonary embolism (P = .026). In the CEE-alone trial, 3,107 participants were included. Total spending among women randomly assigned to CEE was 3.3% higher ($75,411 vs $72,997, P = .16), and 1.7% higher spending for hormone-sensitive diseases ($5,213 vs $5,127, P = .57), but with 39% lower spending for hip fracture (p<0.03). CONCLUSIONS: Menopausal hormone therapy increased spending for some diseases, but decreased spending for others. These offsetting effects led to modest (3%-4%), nonsignificant increases in overall spending among women aged 65 years and older.

Functional Testing or Coronary Computed Tomography Angiography in Patients With Stable Coronary Artery Disease.

BACKGROUND: The choice of either anatomical or functional noninvasive testing to evaluate suspected coronary artery disease might affect subsequent clinical management and outcomes. OBJECTIVES: This study analyzed the association of initial noninvasive cardiac testing in outpatients with stable symptoms, with subsequent use of medications, invasive procedures, and clinical outcomes. METHODS: We studied patients enrolled in a Danish nationwide register who underwent initial noninvasive cardiac testing with either coronary computed tomography angiography (CTA) or functional testing (exercise electrocardiography or nuclear stress testing) from 2009 to 2015. Further use of noninvasive testing, invasive procedures, medications, and medical costs within 120 days were evaluated. Risks of long-term mortality and myocardial infarction (MI) were analyzed using adjusted Cox proportional hazard models. RESULTS: A total of 86,705 patients underwent either functional testing (n = 53,744, mean age 57.4 years, 49% males) or coronary CTA (n = 32,961, mean age 57.4 years, 45% males), and were followed for a median of 3.6 years. Compared with functional testing, there was significantly higher use of statins (15.9% vs. 9.1%), aspirin (12.7% vs. 8.5%), invasive coronary angiography (14.7% vs. 10.1%), and percutaneous coronary intervention (3.8% vs. 2.1%); all p < 0.001 after coronary CTA. The mean costs of subsequent testing, invasive procedures, and medications were higher after coronary CTA ($995 vs. $718; p < 0.001). Unadjusted rates of mortality (2.1% vs. 4.0%) and MI hospitalization (0.8% vs. 1.5%) were lower after coronary CTA than functional testing (both p < 0.001). After adjustment, coronary CTA was associated with a comparable all-cause mortality (hazard ratio: 0.96; 95% confidence interval: 0.88 to 1.05), and a lower risk of MI (hazard ratio: 0.71; 95% confidence interval: 0.61 to 0.82). CONCLUSIONS: In stable patients undergoing initial evaluation for suspected coronary artery disease, coronary CTA was associated with greater use of statins, aspirin, and invasive procedures, and higher costs than functional testing. Coronary CTA was associated with a lower risk of MI, but a similar risk of all-cause mortality.

Cost-Effectiveness of Left Ventricular Assist Devices in Ambulatory Patients With Advanced Heart Failure.

OBJECTIVES: This study assessed the cost-effectiveness of left ventricular assist devices (LVADs) as destination therapy in ambulatory patients with advanced heart failure. BACKGROUND: LVADs improve survival and quality of life in inotrope-dependent heart failure, but data are limited as to their value in less severely ill patients. METHODS: We determined costs of care among Medicare beneficiaries before and after LVAD implantation from 2009 to 2010. We used these costs and efficacy data from published studies in a Markov model to project the incremental cost-effectiveness ratio (ICER) of destination LVAD therapy compared with that of medical management. We discounted costs and benefits at 3% annually and report costs as 2016 U.S. dollars. RESULTS: The mean cost of LVAD implantation was $175,420. The mean cost of readmission was lower before LVAD than after ($12,377 vs. $19,465, respectively; p < 0.001), while monthly outpatient costs were similar ($3,364 vs. $2,974, respectively; p = 0.54). In the lifetime simulation model, LVAD increased quality-adjusted life-years (QALYs) (4.41 vs. 2.67, respectively), readmissions (13.03 vs. 6.35, respectively), and costs ($726,200 vs. $361,800, respectively) compared with medical management, yielding an ICER of $209,400 per QALY gained and $597,400 per life-year gained. These results were sensitive to LVAD readmission rates and outpatient care costs; the ICER would be $86,900 if these parameters were 50% lower. CONCLUSIONS: LVADs in non-inotrope-dependent heart failure patients improved quality of life but substantially increased lifetime costs because of frequent readmissions and costly follow-up care. LVADs may provide good value if outpatient costs and adverse events can be reduced.

Heart failure medications prescribed at discharge for patients with left ventricular assist devices.

BACKGROUND: Real-world use of traditional heart failure (HF) medications for patients with left ventricular assist devices (LVADs) is not well known. METHODS: We conducted a retrospective, observational analysis of 1,887 advanced HF patients with and without LVADs from 32 LVAD hospitals participating in the Get With The Guidelines-Heart Failure registry from January 2009 to March 2015. We examined HF medication prescription at discharge, temporal trends, and predictors of prescription among patients with an in-hospital (n = 258) or prior (n = 171) LVAD implant, and those with advanced HF but no LVAD, as defined by a left ventricular ejection fraction ≤25% and in-hospital receipt of intravenous inotropes or vasopressin receptor antagonists (n = 1,458). RESULTS: For ß-blocker and angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers (ACEI/ARB), discharge prescriptions were 58.9% and 53.5% for new LVAD patients, 53.8% and 42.9% for prior LVAD patients, and 73.4% and 63.2% for patients without LVAD support, respectively (both P < .0001). Aldosterone antagonist prescription quadrupled among LVAD patients during the study period (P < .0001), whereas ACEI/ARB use decreased nearly 20 percentage points (60.0% to 41.4%, P = .0003). In the multivariable analysis of LVAD patients, patient age was inversely associated with ß-blocker, ACEI/ARB, and aldosterone antagonist prescription. CONCLUSIONS: Traditional HF therapies were moderately prescribed at discharge to patients with LVADs and were more frequently prescribed to patients with advanced HF without LVAD support. Moderate prescription rates suggest clinical uncertainty in the use of antiadrenergic medication in this population. Further research is needed on the optimal medical regimen for patients with LVADs.

Outcomes after coronary artery calcium and other cardiovascular biomarker testing among asymptomatic medicare beneficiaries.

BACKGROUND: Biomarkers improve cardiovascular disease (CVD) risk prediction, but their comparative effectiveness in clinical practice is not known. We sought to compare the use, spending, and clinical outcomes in asymptomatic Medicare beneficiaries evaluated for CVD with coronary artery calcium (CAC) or other cardiovascular risk markers. METHODS AND RESULTS: We used a 20% sample of 2005 to 2011 Medicare claims to identify fee-for-service beneficiaries aged ≥65.5 years with no CVD claims in the previous 6 months. We matched patients with CAC with patients who received high-sensitivity C-reactive protein (hs-CRP; n=8358) or lipid screening (n=6250) using propensity-score methods. CAC was associated with increased noninvasive cardiac testing within 180 days (hazard ratio, 2.22, 95% confidence interval, 1.68-2.93, P<0.001, versus hs-CRP; hazard ratio, 4.30, 95% confidence interval, 3.04-6.06, P<0.001, versus lipid screening) and increased coronary angiography and revascularization. During 3-year follow-up, CAC was associated with higher CVD-related spending ($6525 versus $4432 for hs-CRP, P<0.001; and $6500 versus $3073 for lipid screening, P<0.001) and fewer CVD-related events when compared with hs-CRP (hazard ratio, 0.74, 95% confidence interval, 0.58-0.94, P=0.017) but not compared with lipid screening (hazard ratio, 0.84, 95% confidence interval, 0.64-1.11, P=0.23). CONCLUSIONS: CAC testing among asymptomatic Medicare beneficiaries was associated with increased use of cardiac tests and procedures, higher spending, and slightly improved clinical outcomes when compared with hs-CRP testing.

Late diagnosis of abdominal aortic aneurysms substantiates underutilization of abdominal aortic aneurysm screening for Medicare beneficiaries.

OBJECTIVE: Abdominal aortic aneurysm (AAA) screening remains largely underutilized in the U.S., and it is likely that the proportion of patients with aneurysms requiring prompt treatment is much higher compared with well-screened populations. The goals of this study were to determine the proportion of AAAs that required prompt repair after diagnostic abdominal imaging for U.S. Medicare beneficiaries and to identify patient and hospital factors contributing to early vs late diagnosis of AAA. METHODS: Data were extracted from Medicare claims records for patients at least 65 years old with complete coverage for 2 years who underwent intact AAA repair from 2006 to 2009. Preoperative ultrasound and computed tomography was tabulated from 2002 to repair. We defined early diagnosis of AAA as a patient with a time interval of greater than 6 months between the first imaging examination and the index procedure, and late diagnosis as patients who underwent the index procedure within 6 months of the first imaging examination. RESULTS: Of 17,626 patients who underwent AAA repair, 14,948 met inclusion criteria. Mean age was 77.5 ± 6.1 years. Early diagnosis was identified for 60.6% of patients receiving AAA repair, whereas 39.4% were repaired after a late diagnosis. Early diagnosis rates increased from 2006 to 2009 (59.8% to 63.4%; P < .0001) and were more common for intact repair compared with repair after rupture (62.9% vs 35.1%; P < .0001) and for women compared with men (66.3% vs 59.0%; P < .0001). On multivariate analysis, repair of intact vs ruptured AAAs (odds ratio, 3.1; 95% confidence interval, 2.7-3.6) and female sex (odds ratio, 1.4; 95% confidence interval, 1.3-1.5) remained the strongest predictors of surveillance. Although intact repairs were more likely to be diagnosed early, over one-third of patients undergoing repair for ruptured AAAs received diagnostic abdominal imaging greater than 6 months prior to surgery. CONCLUSIONS: Despite advances in screening practices, significant missed opportunities remain in the U.S. Medicare population for improving AAA care. It remains common for AAAs to be diagnosed when they are already at risk for rupture. In addition, a significant proportion of patients with early imaging rupture prior to repair. Our findings suggest that improved mechanisms for observational management are needed to ensure optimal preoperative care for patients with AAAs.

Impact of the Screening Abdominal Aortic Aneurysms Very Efficiently (SAAAVE) Act on abdominal ultrasonography use among Medicare beneficiaries.

BACKGROUND: Since January 1, 2007, Medicare has covered abdominal aortic aneurysm (AAA) screening for new male enrollees with a history of smoking under the Screening Abdominal Aortic Aneurysms Very Efficiently (SAAAVE) Act. We examined the association between this program and abdominal ultrasonography for AAA screening, elective AAA repair, hospitalization for AAA rupture, and all-cause mortality. METHODS: We used a 20% sample of traditional Medicare enrollees from 2004 to 2008 to identify 65-year-old men eligible for screening and 3 control groups not eligible for screening (70-year-old men, 76-year-old men, and 65-year-old women). We used logistic regression to examine the change in outcomes at 365 days for eligible vs ineligible beneficiaries before and after SAAAVE Act implementation, adjusting for comorbidities, state-level smoking prevalence, geographic variation, and time trends. RESULTS: Fewer than 3% of abdominal ultrasonography claims after 2007 were for SAAAVE-specific AAA screening. There was a significantly greater increase in abdominal ultrasonography use among SAAAVE-eligible beneficiaries (2.0 percentage points among 65-year-old men, from 7.6% in 2004 to 9.6% in 2008; 0.7 points [8.9% to 9.6%] among 70-year-old men; 0.7 points [10.8% to 11.5%] among 76-year-old men; and 0.9 points [7.5% to 8.4%] among 65-year-old women) (P < .001 for all comparisons with 65-year-old men). The SAAAVE Act was associated with increased use of abdominal ultrasonography in 65-year-old men compared with 70-year-old men (adjusted odds ratio [AOR], 1.15; 95% CI, 1.11-1.19) (P < .001), and this increased use remained even when SAAAVE-specific AAA screening was excluded (AOR, 1.12; 95% CI, 1.08-1.16) (P < .001). Implementation of the SAAAVE Act was not associated with changes in rates of AAA repair, AAA rupture, or all-cause mortality. CONCLUSIONS: The impact of the SAAAVE Act on AAA screening was modest and was based on abdominal ultrasonography use that it did not directly reimburse. The SAAAVE Act had no discernable effect on AAA rupture or all-cause morality.

Association of coronary CT angiography or stress testing with subsequent utilization and spending among Medicare beneficiaries.

CONTEXT: Coronary computed tomography angiography (CCTA) is a new noninvasive diagnostic test for coronary artery disease (CAD), but its association with subsequent clinical management has not been established. OBJECTIVE: To compare utilization and spending associated with functional (stress testing) and anatomical (CCTA) noninvasive cardiac testing in a Medicare population. DESIGN, SETTING, AND PATIENTS: Retrospective, observational cohort study using claims data from a 20% random sample of 2005-2008 Medicare fee-for-service beneficiaries 66 years or older with no claims for CAD in the preceding year, who received nonemergent, noninvasive testing for CAD (n = 282,830). MAIN OUTCOME MEASURES: Cardiac catheterization, coronary revascularization, acute myocardial infarction, all-cause mortality, and total and CAD-related Medicare spending over 180 days of follow-up. RESULTS: Compared with stress myocardial perfusion scintigraphy (MPS), CCTA was associated with an increased likelihood of subsequent cardiac catheterization (22.9% vs 12.1%; adjusted odds ratio [AOR], 2.19 [95% CI, 2.08 to 2.32]; P < .001), percutaneous coronary intervention (7.8% vs 3.4%; AOR, 2.49 [2.28 to 2.72]; P < .001), and coronary artery bypass graft surgery (3.7% vs 1.3%; AOR, 3.00 [2.63 to 3.41]; P < .001). CCTA was also associated with higher total health care spending ($4200 [$3193 to $5267]; P < .001), which was almost entirely attributable to payments for any claims for CAD ($4007 [$3256 to $4835]; P < .001). Compared with MPS, there was lower associated spending with stress echocardiography (-$4981 [-$4991 to -$4969]; P < .001) and exercise electrocardiography (-$7449 [-$7452 to -$7444]; P < .001). At 180 days, CCTA was associated with a similar likelihood of all-cause mortality (1.05% vs 1.28%; AOR, 1.11 [0.88 to 1.38]; P = .32) and a slightly lower likelihood of hospitalization for acute myocardial infarction (0.19% vs 0.43%; AOR, 0.60 [0.37 to 0.98]; P = .04). CONCLUSION: Medicare beneficiaries who underwent CCTA in a nonacute setting were more likely to undergo subsequent invasive cardiac procedures and have higher CAD-related spending than patients who underwent stress testing.

The relationship between low back magnetic resonance imaging, surgery, and spending: impact of physician self-referral status.

OBJECTIVE: To examine the relationship between use of magnetic resonance imaging (MRI) and receipt of surgery for patients with low back pain. DATA SOURCES: Medicare claims for a 20 percent sample of beneficiaries from 1998 to 2005. STUDY DESIGN: We identify nonradiologist physicians who appear to begin self-referral arrangements for MRI between 1999 and 2005, as well as their patients who have a new episode of low back pain care during this time. We focus on regression models that identify the relationship between receipt of MRI and subsequent use of back surgery and health care spending. Receipt of MRI may be endogenous, so we use physician acquisition of MRI as an instrument for receipt of MRI. The models adjust for demographic and socioeconomic covariates as well as month, year, and physician fixed effects. DATA COLLECTION/EXTRACTION METHODS: We include traditional, fee-for-service Medicare beneficiaries with a visit to an orthopedist or primary care physician for nonspecific low back pain, and no claims for low back pain in the year prior. PRINCIPAL FINDINGS: In the first stage, acquisition of MRI equipment is a strongly correlated with patients receiving MRI scans. Among patients of orthopedists, receipt of an MRI scan increases the probability of having surgery by 34 percentage points. Among patients of primary care physicians, receiving a low back MRI is not statistically significantly associated with subsequent surgery receipt. CONCLUSIONS: Orthopedists and primary care physicians who begin billing for the performance of MRI procedures, rather than referring patients outside of their practice for MRI, appear to change their practice patterns such that they use more MRI for their patients with low back pain. These increases in MRI use appear to lead to increases in low back surgery receipt and health care spending among patients of orthopedic surgeons, but not of primary care physicians.