Trends in extent of surgical cytoreduction for patients with ovarian cancer.

PURPOSE: To identify patient and hospital characteristics associated with extended surgical cytoreduction in the treatment of ovarian cancer. METHODS: A retrospective analysis using the National Inpatient Sample (NIS) database identified women hospitalized for surgery to remove an ovarian malignancy between 2013 and 2017. Extended cytoreduction (ECR) was defined as surgery involving the bowel, liver, diaphragm, bladder, stomach, or spleen. Chi-square and logistic regression were used to analyze patient and hospital demographics related to ECR, and trends were assessed using the Cochran-Armitage test. RESULTS: Of the estimated 79,400 patients undergoing ovarian cancer surgery, 22% received ECR. Decreased adjusted odds of ECR were found in patients with lower Elixhauser Comorbidity Index (ECI) scores (OR 0.61, p<0.001 for ECI 2, versus ECI≥3) or residence outside the top income quartile (OR 0.71, p<0.001 for Q1, versus Q4), and increased odds were seen at hospitals with high ovarian cancer surgical volume (OR 1.25, p<0.001, versus low volume). From 2013 to 2017, there was a decrease in the proportion of cases with extended procedures (19% to 15%, p<0.001). There were significant decreases in the proportion of cases with small bowel, colon, and rectosigmoid resections (p<0.001). Patients who underwent ECR were more likely treated at a high surgical volume hospital (37% vs 31%, p<0.001) over the study period. For their hospital admission, patients who underwent ECR had increased mortality (1.6% vs. 0.5%, p<0.001), length of stay (9.6 days vs. 5.2 days, p<0.001), and mean cost ($32,132 vs. $17,363, p<0.001). CONCLUSIONS: Likelihood of ECR was associated with increased medical comorbidity complexity, higher income, and undergoing the procedure at high surgical volume hospitals. The proportion of ovarian cancer cases with ECR has decreased from 2013-17, with more cases performed at high surgical volume hospitals.

The obesity paradox: Underweight patients are at the greatest risk of mortality after cholecystectomy.

BACKGROUND: Elevated body mass index is a risk factor for gallstone disease and cholecystectomy, but outcomes for low body mass index patients remain uncharacterized. We examined the association of body mass index with morbidity, mortality, and resource use after cholecystectomy. METHODS: The 2005 to 2016 American College of Surgeons National Surgical Quality Improvement Program was retrospectively analyzed for adult patients undergoing laparoscopic and open cholecystectomy. Patients were stratified into 5 groups: body mass index <18.5 (underweight), body mass index 18.5 to 24.9 (normal weight), body mass index 25 to 29.9 (overweight), body mass index 30 to 34.9 (class I obesity), body mass index 35 to 39.9 (class II obesity), and body mass index ≥40 (class III obesity). Multivariable regressions identified independent associations of covariates with 30-day mortality, complications, and resource use. RESULTS: Of 327,473 cholecystectomy patients, 1.0% were underweight, 19.5% normal weight, 30.3% overweight, 24.0% class I obesity, 13.5% class II obesity, and 11.7% class III obesity. After multivariable analysis, underweight patients had a higher risk of mortality (adjusted odds ratio = 1.53; P = .029) and postoperative bleeding (adjusted odds ratio = 1.45; P = .011) relative to normal weight patients. Conversely, class III obesity patients had lower mortality (adjusted odds ratio = 0.66; P = .005) but increased operative time (ß = 10.2 minutes; P < .001), wound infection (adjusted odds ratio = 1.38; P < .001), and wound dehiscence (adjusted odds ratio = 2.20; P < .001). Hospital duration of stay and readmission rates were highest for underweight patients. CONCLUSION: Underweight patients experience increased risk of mortality and readmission, while class III obesity patients have higher rates of wound infection and dehiscence as well as prolonged operative time. These findings may guide choice of intervention.

Readmissions after ovarian cancer cytoreduction surgery: The first 30 days and beyond.

BACKGROUND AND OBJECTIVES: Postoperative readmissions are often used to assess quality of surgical care. This study compared 30-day vs 31- to 90-day readmission following surgery for ovarian, fallopian tube, or primary peritoneal cancer. METHODS: This retrospective study of the 2010-2015 Nationwide Readmissions Database characterized 90-day readmissions following cytoreductive surgery for these cancers. Each patient's first postoperative hospitalization was included. Univariate analysis compared patient demographics and reasons for readmission. Multivariable regression identified independent predictors of readmission. RESULTS: Of an estimated 76 652 patients, 10 264 (13.4%) were readmitted within 30 days, and 6942 (9.1%) between 31 and 90 days. The 30-day readmissions were more frequently associated with postoperative infection, while 31- to 90-day readmissions were more frequently associated with renal or hematologic diagnoses. Predictors of any 90-day readmission included index hospitalization longer than 7 days (adjusted odds ratio (AOR) 1.61 [1.48-1.75], P < .001), extended surgical procedure (AOR 1.41 [1.30-1.53], P < .001), pulmonary circulation disorder (AOR = 1.34 [1.13-1.60], P = .001), and diabetes mellitus (AOR = 1.12 [1.02-1.24], P = .020). CONCLUSIONS: Readmission rates remain high during the 31- to 90-day postoperative period in ovarian cancer patients, although these readmissions are less frequently related to postoperative complications. Prospective study is merited to optimize surveillance beyond the initial 30 days after ovarian cancer surgery.

Height mismatch: An overlooked component of adult heart transplant outcomes.

Heart transplantation guidelines recommend against matching donors with significant weight but not height discrepancies. This study analyzed the impact of donor-recipient height mismatch on mortality among heart transplant recipients. We retrospectively analyzed all adult patients in the United Network for Organ Sharing (UNOS) registry undergoing heart transplantation from 1990 to September 2016. Moderate and severe height mismatch were classified as >10% and >15% difference in donor height from recipient height, respectively. The primary outcome was 1-year mortality. Adjusted Cox hazards regression was performed, and Kaplan-Meier estimates illustrated 10-year survival. Of 44 877 transplants, 4822 (10.7%) were moderately height mismatched. Height-mismatched recipients were more frequently female (41.6% vs 21.8%, P < .001), sex mismatched (53.8% vs 24.9%, P < .001), and weight mismatched (4.9% vs 1.9%, P < .001). After adjustment, recipients of moderately (HR = 1.15 [1.02-1.30]) and severely (HR = 1.38 [1.10-1.74]) taller donor hearts were at increased risk of mortality at 1 year relative to height-matched recipients. Furthermore, of 1042 (21.6%) severe mismatches, recipients with taller (HR = 1.39 [1.11-1.74]) but not shorter (HR = 0.79 [0.44-1.43]) donors faced increased 10-year mortality. The effect was pronounced among re-transplant candidates (HR = 1.96 [1.07-3.59]). In conclusion, matching with moderately or severely taller donors is an independent predictor of mortality among primary and re-transplant candidates.

Acute kidney injury is independently associated with mortality and resource use after emergency general surgery operations.

INTRODUCTION: The incidence of severe perioperative renal dysfunction in high-acuity patients has not been well-explored at the national level. The present study aimed to evaluate the trends in the incidence of perioperative acute kidney injury and renal replacement therapy as well as associated mortality among patients undergoing an emergency general surgery operation. METHODS: This was a retrospective cohort study using the National Inpatient Sample to identify all adult patients (>18 y) without chronic kidney disease who underwent an emergency general surgery procedure from 2008 to 2016. The study cohort was stratified based on presence of acute kidney injury and need for renal replacement therapy postoperatively. A multivariable logistic regression model was developed to predict the odds of mortality and composite morbidity. Nonparametric trend analyses of acute kidney injury and renal replacement therapy incidence and associated mortality were performed. RESULTS: Of an estimated 5,862,657 patients who underwent an emergency general surgery procedure during the study period, 7.4% patients developed an acute kidney injury and 0.48% patients required renal replacement therapy. Overall, the incidence of acute kidney injury (5.3%-19.4%) and renal replacement therapy (0.43%-0.93%) increased (P < .0001) over the study period. Even without need for renal replacement therapy, acute kidney injury was associated with greater odds of mortality and composite morbidity (adjusted odds ratio 5.2, 95% confidence interval [CI] 5.1-5.3) and mortality (adjusted odds ratio = 2.20, 95% CI 2.3-2.4), as well as greater costs of hospitalization and duration of stay. CONCLUSION: In this national study, we found that the incidence of acute kidney injury and renal replacement therapy after an emergency general surgery operation has increased. Both acute renal failure and hemodialysis were associated with much greater odds of morbidity and mortality. The apparent increase in the rate of acute kidney injury and renal replacement therapy warrant further investigation of mechanisms for monitoring and limiting the impact of organ malperfusion associated with emergency general surgery operations.

Donor history of malignancy: A limited risk for heart transplant recipients.

Organ donor contraindications are frequently reassessed for impact on recipient outcomes in attempt to meet demand for transplantation. This study retrospectively analyzed the United Network for Organ Sharing (UNOS) registry for adult heart transplants from 1987 to September 2016 to characterize the impact of donor malignancy history in heart transplantation. Kaplan-Meier estimates illustrated 10-year survival. Propensity score matching was utilized for 1:1 matching of donors with and without history of malignancy, and Cox proportional hazards and logistic regressions were used to analyze the matched population. Of 38 781 heart transplants, 622 (1.6%) had a donor history of malignancy. Cox regressions demonstrated that donor malignancy predicted increased 10-year mortality (HR = 1.16 [1.01-1.33]), but this difference did not persist when conditioned upon 1 year post-transplant survival (log-rank = 0.643). Cox regressions of the propensity score-matched population (455 pairs) found no association between donor malignancy and 10-year mortality (HR = 1.02 [0.84-1.24]). Older age and higher rates of hypertension were observed in donors with a history of malignancy whose recipients died within the first year post-transplant. Therefore, increased recipient mortality is likely due to donor characteristics beyond malignancy, creating the potential for expanded donor selection.

Predicting morbidity and mortality in laparoscopic cholecystectomy: Preoperative serum albumin still matters.

BACKGROUND: This study examined the association of preoperative serum albumin with outcomes for laparoscopic cholecystectomy. METHODS: The American College of Surgeons National Surgical Quality Improvement Program was retrospectively analyzed from 2005 to 2016 for adult patients undergoing laparoscopic cholecystectomy. Patients were stratified into four groups: <3.0 g/dL (Severe Malnutrition), 3.0-<3.5 (Moderate Malnutrition), 3.5-<4.0 (Mild Malnutrition), and ≥4.0 g/dL (Normal Nutrition). The primary outcome of 30-day mortality was evaluated with multivariable regression. RESULTS: Of 131,855 patients, 14.0% had Severe, 22.8% Moderate, and 29.7% Mild Malnutrition, with 33.5% classified as Normal Nutrition. Adjusted multivariable regressions demonstrated that relative to Normal Nutrition, mortality risk was increased for Severe (OR = 3.09 [95% Confidence Interval: 2.09-4.56]) and Moderate (OR = 1.83 [1.24-2.72]) Malnutrition. Severe (OR = 2.45 [1.67-3.61]) and Moderate (OR = 1.52 [1.04-2.24]) Malnutrition were also associated with increased risk of postoperative septic shock. CONCLUSIONS: Even in less invasive laparoscopic cholecystectomy, reduced preoperative serum albumin is strongly associated with increased morbidity and mortality.

International Normalized Ratio (INR) Is Comparable to MELD in Predicting Mortality after Cholecystectomy.

Guidelines suggest targeting a preoperative international normalized ratio (INR) < 1.5. We examined and compared the predictive value of INR relative to the Model for End-Stage Liver Disease (MELD). We reviewed the American College of Surgeons NSQIP from 2005 to 2016 for adult patients undergoing open or laparoscopic cholecystectomy. Patients with a preoperative INR were stratified into groups: ≤1, >1 to ≤1.5, >1.5 to ≤2, and >2. Thirty day postoperative mortality was the primary outcome. Multivariable logistic regressions controlled for baseline differences. Of 58,177 cholecystectomy patients, 15.2 per cent had INR ≤ 1, 80.4 per cent had INR > 1 to ≤1.5, 3.7 per cent had INR > 1.5 to ≤2, and 0.7 per cent had INR > 2. Patients with INR > 2 were older and more likely to have diabetes and hypertension (P < 0.001). Multivariable regression demonstrated a stepwise increase in mortality for INR > 1 to ≤1.5 (odds ratio (OR) = 1.50 [1.10-2.05]), INR > 1.5 to ≤2 (OR = 2.96 [1.97-4.45]), and INR > 2 (OR = 3.21 [1.64-6.31]) relative to INR ≤ 1. C-statistic for INR (0.910) and MELD (0.906) models indicated a similar value in predicting mortality. INR groups also faced an incremental, increased risk of bleeding. Although unable to track preoperative correction of INR, this analysis identifies that INR remains an excellent predictor of postoperative mortality and bleeding after both open and laparoscopic cholecystectomies and is comparable to MELD.

Fatal crashes in the 5 years after recreational marijuana legalization in Colorado and Washington.

Colorado and Washington legalized recreational marijuana in 2012, but the effects of legalization on motor vehicle crashes remains unknown. Using Fatality Analysis Reporting System data, we performed difference-in-differences (DD) analyses comparing changes in fatal crash rates in Washington, Colorado and nine control states with stable anti-marijuana laws or medical marijuana laws over the five years before and after recreational marijuana legalization. In separate analyses, we evaluated fatal crash rates before and after commercial marijuana dispensaries began operating in 2014. In the five years after legalization, fatal crash rates increased more in Colorado and Washington than would be expected had they continued to parallel crash rates in the control states (+1.2 crashes/billion vehicle miles traveled, CI: -0.6 to 2.1, p = 0.087), but not significantly so. The effect was more pronounced and statistically significant after the opening of commercial dispensaries (+1.8 crashes/billion vehicle miles traveled, CI: +0.4 to +3.7, p = 0.020). These data provide evidence of the need for policy strategies to mitigate increasing crash risks as more states legalize recreational marijuana.

Understanding lung transplant listing practices: Survival in lung transplant candidates who improve clinically to delisting.

BACKGROUND: Occasionally, lung transplant candidates improve to the point where they are removed from the transplant list. We sought to determine the characteristics and outcomes of lung transplant candidates who improved to delisting both before and after implementation of the lung allocation score. METHODS: Using the United Network for Organ Sharing database, we reviewed all adult patients listed for lung transplant between 1987 and 2012. The last permanent status change was classified into transplanted, improved to delisting (improved), or deteriorated to delisting (deteriorated). Survival time was calculated using the linked date of death from the Social Security Administration. Survival analysis was performed via the Kaplan-Meier method, and adjusted multivariable logistic regressions identified characteristics predicting improvement to delisting. RESULTS: Of 13,688 candidates, 12,188 (89.0%) were transplanted, 454 (3.3%) improved, and 1,046 (7.6%) deteriorated. The 5-year mortality was greater in improved (hazard ratio = 1.21 [1.07-1.38], P = .002) and deteriorated (hazard ratio = 3.36 [3.11-3.64], P < .001) candidates relative to those transplanted; however, 1-year survival was greater in improved versus transplanted candidates (75.9% vs 67.2%, log rank P < .001). Older, female patients listed for primary pulmonary hypertension and retransplantation were more likely to improve to delisting. The proportion of improved patients varied by hospital quartile volume (P < .001) and the United Network for Organ Sharing geographic region (P < .001). The number of patients improving to delisting decreased after implementation of the lung allocation score. CONCLUSION: Lung transplant candidates improving to delisting faced less short-term but greater long-term mortality relative to transplanted candidates. Given that the improved population decreased dramatically after implementation of the lung allocation score, redefining patient listing criteria appears to have improved patient appropriateness for transplant.

Impact of Donor Diabetes on Outcomes of Lung Transplantation in the United States.

BACKGROUND: Diabetes mellitus is among several factors considered when assessing the suitability of donated organs for transplantation. Lungs from diabetic donors (LDD) are not contraindicated for use as allografts, despite established evidence of diabetes-mediated parenchymal damage. The present study used a national database to assess the impact of donor diabetes on the longevity of lung transplant recipients. METHODS: This retrospective study of the United Network for Organ Sharing database analyzed all adult lung transplant recipients from June 2005 through September 2016. Donor and recipient demographics including the presence of diabetes were used to create a multivariable model. The primary outcome was 5-y mortality, with hazard ratios (HRs) assessed using multivariable Cox regression analysis. Survival curves were calculated using the Kaplan-Meier method. RESULTS: Of the 17,839 lung transplant recipients analyzed, 1203 (6.7%) received LDD. Recipients of LDD were more likely to be female (44.1% versus 40.2%, P < 0.01) and have mismatched race (47.5% versus 42.2%, P < 0.01). Diabetic donors were more likely to have hypertension (74.6% versus 19.0%, P < 0.01). Multivariable analysis revealed LDD to be an independent predictor of mortality at 5 y (HR 1.16 [1.04-1.29], P < 0.01). However, among the subgroup of diabetic recipients, transplantation of LDD showed no independent association with 5-y mortality (HR 0.81 [0.63-1.06], P = 0.12). CONCLUSIONS: Recipients of LDD had a lower 5-y post lung transplantation survival compared with recipients of lungs from nondiabetic donors. LDD allografts did not influence the survival of diabetic recipients.

Readmissions Following Isolated Coronary Artery Bypass Graft Surgery in the United States (from the Nationwide Readmissions Database 2010 to 2014).

Readmission following cardiac surgery is associated with poor outcomes and increased healthcare expenditure. However, a nationwide understanding of the incidence, cost, causes, and predictors of 30-day readmission following coronary artery bypass grafting is limited. The Nationwide Readmissions Database was used to identify all adult patients who underwent isolated coronary artery bypass grafting (CABG) with no other concomitant surgery between 2010 and 2014. The primary outcome was all-cause readmission within 30 days of discharge after surgery. Risk-adjusted multivariable analyses were used to develop a model of readmission risk. Of 855,836 patients, 95,504 (11.2%) had an emergent 30-day readmission following CABG. The most common causes of readmission were related to respiratory complications (17.1%), infection (13.5%), and heart failure (11.9%). Readmission cost an average of $13,392 per patient, accounting for an estimated annual cost of over $250 million. Independent predictors of 30-day readmission encompassed female gender (odds ratio [OR] 1.27; 95% confidence interval [CI] 1.24 to 1.31), emergent index admission (OR 1.29; 95% CI 1.25 to 1.33), and preoperative co-morbidities, including atrial fibrillation (OR 1.24; 95% CI 1.21 to 1.28), liver disease (OR 1.29; 95% CI 1.17 to 1.41), renal failure (OR 1.38; 95% CI 1.34 to 1.43), among others. CABG performed at a high CABG volume hospital was protective of readmission (OR 0.95; 95% CI 0.91 to 0.99). In conclusion, we characterized using a national sample the incidence, causes, costs, and predictors of 30-day readmission following CABG. Targeting modifiable risk factors for readmission should be a priority to reduce rates of readmission and decrease healthcare expenditure.

Association of Autoimmune Connective Tissue Disease and Outcomes in Patients Undergoing Transcatheter Aortic Valve Implantation.

Patients with autoimmune connective tissue disease (CTD) are at higher risk for developing aortic valve pathology, but the safety and value of transcatheter aortic valve implantation (TAVI) in this population has not been investigated. This study evaluated mortality, complication, and readmission rates along with length of stay and total costs after TAVI in patients with CTD. We retrospectively reviewed 47,216 patients who underwent TAVI from the National Readmissions Database between January 2011 and September 2015. Patients with systemic lupus erythematosus, scleroderma, rheumatoid arthritis, and other autoimmune CTD comprised the cohort. The primary outcome was mortality at index hospitalization. The 2,557 CTD patients (5.4%) had a higher Elixhauser co-morbidity index (7.1 vs 6.1, p <0.001) than non-CTD patients. CTD and non-CTD patients had similar mortality (2.8 vs 4.1%, p = 0.052), 30-day readmission (19.3 vs 17.0%, p = 0.077), length of stay (8.2 vs 8.3 days, p = 0.615), and total adjusted costs ($57,202 vs $58,309, p = 0.196), respectively. However, CTD patients were more frequently readmitted for postoperative infection (9.4 vs 5.6%, p = 0.042) and septicemia (8.2 vs 4.5%, p = 0.019). After multivariable adjustment, CTD patients faced lower mortality at index hospitalization (odds ratio [OR] 0.56 [0.38 to 0.82], p = 0.003) but were more frequently readmitted for septicemia (OR = 1.95 [1.10 to 3.45], p = 0.023) and postoperative infection (OR = 3.10 [1.01 to 9.52], p = 0.048) relative to non-CTD patients. In conclusion, CTD is not a risk factor for in-hospital mortality but is an independent risk factor for infectious complications post-TAVI.

Pretransplant malignancy among lung transplant recipients in the modern era.

BACKGROUND: Malignancy is a relative contraindication in transplant candidates, given the increased neoplastic risk accompanying posttransplant immunosuppression. However, the number of patients receiving a lung transplant despite pretransplant malignancy is rising, and their outcomes remain unclear. Our purpose was to examine the outcomes of lung transplant recipients with pretransplant malignancy in the modern era. METHODS: We evaluated the United Network for Organ Sharing registry for adult lung transplants that were completed between June 2005 and September 2016. Transplant recipients were stratified by pretransplant malignancy, with subgroup analysis by sex and active malignancy. The primary outcome was 5-year survival and the secondary outcome was cause of death. Kaplan-Meier estimates illustrated 5-year survival and multivariable Cox proportional hazards regressions controlled for demographics and comorbidities. RESULTS: Of 18,032 transplant patients, 1,321 transplant recipients (7.3%) possessed a pretransplant malignancy. Patients with pretransplant malignancy faced significantly greater mortality within 5 years (36.0% vs 32.8%, P = .017), an effect greatest in men with pretransplant malignancy (39.2% vs 33.7%, P = .002). Patients with pretransplant malignancy also faced greater risk of death from posttransplant malignancy (15.6% vs 9.4%, P < .001), particularly for those with active malignancy at transplant (34.8% vs 9.8%, P < .001). Pretransplant malignancy remained a significant predictor of 5-year mortality in adjusted Cox regressions (hazard ratio: 1.16 [1.05-1.27], P = .003). CONCLUSION: Patients with pretransplant malignancy, and particularly men with pretransplant malignancy and those with active malignancy at transplant, are at an increased risk of 5-year mortality and posttransplant death from malignancy. Balancing individual risk of posttransplant malignancy with immunosuppressive care is necessary to optimize outcomes for pretransplant malignancy patients.

Clinical Outcomes of Infective Endocarditis in Injection Drug Users.

BACKGROUND: Rising rates of hospitalization for infective endocarditis (IE) have been increasingly tied to rising injection drug use (IDU) associated with the opioid epidemic. OBJECTIVES: This study analyzed recent trends in IDU-IE hospitalization and characterized outcomes and readmissions for IDU-IE patients. METHODS: The authors evaluated the National Readmissions Database (NRD) for IE cases between January 2010 and September 2015. Patients were stratified by IDU status and surgical versus medical management. Primary outcome was 30-day readmission and cause, with secondary outcomes including mortality, length of stay (LOS), adjusted costs, and 180-day readmission. The Kruskal-Wallis and chi-square tests were used to analyze baseline differences by IDU status. Multivariable regressions were used to analyze mortality, readmissions, LOS, and adjusted costs. RESULTS: The survey-weighted sample contained 96,344 (77.8%) non-IDU-IE and 27,432 (22.2%) IDU-IE cases. IDU-IE increased from 15.3% to 29.1% of IE cases between 2010 and 2015 (p < 0.001). At index hospitalization, IDU-IE was associated with reduced mortality (6.8% vs. 9.6%; p < 0.001) but not 30-day readmission (23.8% vs. 22.9%; p = 0.077) relative to non-IDU-IE. Medically managed IDU-IE patients had higher LOS (ß = 1.36 days; 95% confidence interval [CI]: 0.71 to 2.01), reduced costs (ß = -$4,427; 95% CI: -$7,093 to -$1,761), and increased readmission for endocarditis (18.1% vs. 5.6%; p < 0.001), septicemia (14.0% vs. 7.3%; p < 0.001), and drug abuse (4.3% vs. 0.7%; p < 0.001) compared with medically managed non-IDU-IE. Surgically managed IDU-IE patients had increased LOS (ß = 4.26 days; 95% CI: 2.73 to 5.80) and readmission for septicemia (15.6% vs. 5.2%; p < 0.001) and drug abuse (7.3% vs. 0.9%; p < 0.001) compared with non-IDU-IE. CONCLUSIONS: The incidence of IDU-IE continues to rise nationally. Given the increased readmission for endocarditis, septicemia, and drug abuse, IDU-IE presents a serious challenge to current management of IE.

Crash Fatality Rates After Recreational Marijuana Legalization in Washington and Colorado.

OBJECTIVES: To evaluate motor vehicle crash fatality rates in the first 2 states with recreational marijuana legalization and compare them with motor vehicle crash fatality rates in similar states without recreational marijuana legalization. METHODS: We used the US Fatality Analysis Reporting System to determine the annual numbers of motor vehicle crash fatalities between 2009 and 2015 in Washington, Colorado, and 8 control states. We compared year-over-year changes in motor vehicle crash fatality rates (per billion vehicle miles traveled) before and after recreational marijuana legalization with a difference-in-differences approach that controlled for underlying time trends and state-specific population, economic, and traffic characteristics. RESULTS: Pre-recreational marijuana legalization annual changes in motor vehicle crash fatality rates for Washington and Colorado were similar to those for the control states. Post-recreational marijuana legalization changes in motor vehicle crash fatality rates for Washington and Colorado also did not significantly differ from those for the control states (adjusted difference-in-differences coefficient = +0.2 fatalities/billion vehicle miles traveled; 95% confidence interval = -0.4, +0.9). CONCLUSIONS: Three years after recreational marijuana legalization, changes in motor vehicle crash fatality rates for Washington and Colorado were not statistically different from those in similar states without recreational marijuana legalization. Future studies over a longer time remain warranted.