Natural Language Processing for Improved Characterization of COVID-19 Symptoms: Observational Study of 350,000 Patients in a Large Integrated Health Care System.

BACKGROUND: Natural language processing (NLP) of unstructured text from electronic medical records (EMR) can improve the characterization of COVID-19 signs and symptoms, but large-scale studies demonstrating the real-world application and validation of NLP for this purpose are limited. OBJECTIVE: The aim of this paper is to assess the contribution of NLP when identifying COVID-19 signs and symptoms from EMR. METHODS: This study was conducted in Kaiser Permanente Southern California, a large integrated health care system using data from all patients with positive SARS-CoV-2 laboratory tests from March 2020 to May 2021. An NLP algorithm was developed to extract free text from EMR on 12 established signs and symptoms of COVID-19, including fever, cough, headache, fatigue, dyspnea, chills, sore throat, myalgia, anosmia, diarrhea, vomiting or nausea, and abdominal pain. The proportion of patients reporting each symptom and the corresponding onset dates were described before and after supplementing structured EMR data with NLP-extracted signs and symptoms. A random sample of 100 chart-reviewed and adjudicated SARS-CoV-2-positive cases were used to validate the algorithm performance. RESULTS: A total of 359,938 patients (mean age 40.4 [SD 19.2] years; 191,630/359,938, 53% female) with confirmed SARS-CoV-2 infection were identified over the study period. The most common signs and symptoms identified through NLP-supplemented analyses were cough (220,631/359,938, 61%), fever (185,618/359,938, 52%), myalgia (153,042/359,938, 43%), and headache (144,705/359,938, 40%). The NLP algorithm identified an additional 55,568 (15%) symptomatic cases that were previously defined as asymptomatic using structured data alone. The proportion of additional cases with each selected symptom identified in NLP-supplemented analysis varied across the selected symptoms, from 29% (63,742/220,631) of all records for cough to 64% (38,884/60,865) of all records with nausea or vomiting. Of the 295,305 symptomatic patients, the median time from symptom onset to testing was 3 days using structured data alone, whereas the NLP algorithm identified signs or symptoms approximately 1 day earlier. When validated against chart-reviewed cases, the NLP algorithm successfully identified signs and symptoms with consistently high sensitivity (ranging from 87% to 100%) and specificity (94% to 100%). CONCLUSIONS: These findings demonstrate that NLP can identify and characterize a broad set of COVID-19 signs and symptoms from unstructured EMR data with enhanced detail and timeliness compared with structured data alone.

The Natural History of NAFLD, a Community-Based Study at a Large Health Care Delivery System in the United States.

Nonalcoholic fatty liver disease (NAFLD) is a global public health problem. However, the natural history of NAFLD is incomplete. This is a retrospective cohort study of patients identified with NAFLD by diagnosis codes in a large, community-based health care delivery system. The objectives were (1) to follow patients from initial NAFLD presentation through progression to cirrhosis and/or decompensated cirrhosis to liver cancer, liver transplant, and death for up to 10 years; and (2) to conduct disease progression analysis restricted to patients with NAFLD identified as having diabetes at baseline. A total of 98,164 patients with full NAFLD and 26,488 with diabetes were divided into three baseline prevalent states: (1) no cirrhosis, (2) compensated cirrhosis, and (3) decompensated cirrhosis. In baseline patients without cirrhosis, annual rates of compensated cirrhosis, decompensated cirrhosis, and death were 0.28%, 0.31%, and 0.63% per year, respectively. With baseline compensated cirrhosis, the annual rates of decompensation and death were 2.4% and 6.7% per year. Finally, in those with decompensated cirrhosis at baseline, the death rate was 8.0% per year. In those without cirrhosis and with cirrhosis at baseline, the rates of liver cancer and death were increased approximately 2-fold in the diabetic subpopulation compared with the full NAFLD cohort. Age and comorbidities increased with increasing disease severity. Cox proportional hazards regression analysis showed that cirrhosis was strongly associated with death and liver cancer, and that diabetes was associated with a significant increase in the hazard of both liver cancer and death (2.56 [2.04-3.20] and 1.43 [1.35-1.52]), respectively. Conclusion: The findings of this community-based study further our understanding of the natural history of NAFLD and demonstrate that diabetes is a major factor in the progression of this disease.

Obesity and Mortality Among Patients Diagnosed With COVID-19: Results From an Integrated Health Care Organization.

BACKGROUND: Obesity, race/ethnicity, and other correlated characteristics have emerged as high-profile risk factors for adverse coronavirus disease 2019 (COVID-19)-associated outcomes, yet studies have not adequately disentangled their effects. OBJECTIVE: To determine the adjusted effect of body mass index (BMI), associated comorbidities, time, neighborhood-level sociodemographic factors, and other factors on risk for death due to COVID-19. DESIGN: Retrospective cohort study. SETTING: Kaiser Permanente Southern California, a large integrated health care organization. PATIENTS: Kaiser Permanente Southern California members diagnosed with COVID-19 from 13 February to 2 May 2020. MEASUREMENTS: Multivariable Poisson regression estimated the adjusted effect of BMI and other factors on risk for death at 21 days; models were also stratified by age and sex. RESULTS: Among 6916 patients with COVID-19, there was a J-shaped association between BMI and risk for death, even after adjustment for obesity-related comorbidities. Compared with patients with a BMI of 18.5 to 24 kg/m2, those with BMIs of 40 to 44 kg/m2 and greater than 45 kg/m2 had relative risks of 2.68 (95% CI, 1.43 to 5.04) and 4.18 (CI, 2.12 to 8.26), respectively. This risk was most striking among those aged 60 years or younger and men. Increased risk for death associated with Black or Latino race/ethnicity or other sociodemographic characteristics was not detected. LIMITATION: Deaths occurring outside a health care setting and not captured in membership files may have been missed. CONCLUSION: Obesity plays a profound role in risk for death from COVID-19, particularly in male patients and younger populations. Our capitated system with more equalized health care access may explain the absence of effect of racial/ethnic and socioeconomic disparities on death. Our data highlight the leading role of severe obesity over correlated risk factors, providing a target for early intervention. PRIMARY FUNDING SOURCE: Roche-Genentech.

Increased cancer rates in patients with chronic hepatitis C.

AIMS: As previous reports show an association of chronic hepatitis C (HCV) with hepatocellular carcinoma (HCC) and non-liver cancers, we examine the association of HCV with liver cancer and non-liver cancers. METHODS: Retrospective cross-sectional study at Kaiser Permanente Southern California (KPSC) evaluating HCV and non-HCV patients from 1 January 2008 to 12 December 2012. Cancer diagnoses were obtained from the KPSC-SEER-affiliated registry. Logistic regression analyses were used for rate ratios and time-to-event analyses were performed using Cox proportional hazards models, adjusted for age, gender, race, smoking and cirrhosis. Cancer rate ratios were stratified by tobacco, alcohol abuse, diabetes and body mass index (BMI). RESULTS: The initial population and final population of multivariable analysis were N = 5 332 903 and N = 2 080 335 respectively. Cancer burden (all sites) was significantly higher in HCV than in non-HCV patients and HCV patients had a high rate of liver cancer. When liver cancer was excluded, cancer rates remained significantly increased in HCV. Unadjusted cancer rates were significantly higher in HCV compared to non-HCV for oesophageal, stomach, colorectal, pancreas, myeloma, non-Hodgkin's lymphoma, head/neck, lung, renal and prostate cancer. After stratification for alcohol abuse, tobacco, diabetes and BMI, increased cancer rates remained significant for all cancer sites, liver cancer and non-Hodgkin's lymphoma. Multivariable analyses demonstrated a strong correlation between cirrhosis and cancer. Tobacco use and diabetes were also associated with cancer. In the absence of cirrhosis, HCV, tobacco use and diabetes significantly increased the cancer risk. Mediation analyses showed that cirrhosis was responsible for a large proportion on the effect of HCV on cancer risk. CONCLUSION: This study supports the concept of HCV as a systemic illness and treating HCV regardless of disease severity and prior to progression to cirrhosis.

Understanding the Capacity for Exercise in Post-Bariatric Patients.

PURPOSE: The aim of this study is to conduct a pilot randomized trial testing an exercise program specifically adapted for post-bariatric patients. METHODS: A total of 51 post-bariatric patients, 6-24 months post-surgery, were randomly assigned to usual care control (n = 25) or the exercise intervention (n = 26). The intervention included twice weekly 60-min group exercise classes with functional strength, flexibility, and aerobic activities; at least 3 days per week of self-directed exercise; daily pedometer; recording of steps and activities; and weekly telephone counseling. There was also a 6-month maintenance period. RESULTS: Patients were 49 ± 12 years old, 84 % female, 59 % non-Hispanic white, with a BMI of 32.9 ± 5.7 kg/m2 and percent excess BMI loss since surgery of 56 ± 35 %. Patients were 14 ± 5 months post-surgery. A total of 44 patients (86 %) completed both phases of the program and all assessments. The following measures improved significantly for intervention participants with no significant change in control participants: yards walked in 6 min, seconds for 8-foot up-and-go, number of arm curls, and distance in inches for chair sit-and-reach. Intervention changes remained after 6 months of maintenance. CONCLUSIONS: When compared to patients in usual care, a specially adapted exercise program for post-bariatric patients resulted in significant improvements in objectively monitored health outcomes. This program was delivered in a clinical setting and could be implemented in a variety of settings to improve health outcomes for post-bariatric patients.

Postoperative Risk of Hepatic Decompensation after Orthopedic Surgery in Patients with Cirrhosis.

BACKGROUND AND AIMS: Previous studies have shown increased hepatic decompensation in patients with cirrhosis undergoing surgery. However, there are little data available in cirrhotics undergoing orthopedic surgery compared to cirrhotics who did not undergo surgery. The aim of this study was to examine the demographics, comorbid conditions, and clinical factors associated with hepatic decompensation within 90 days in cirrhotics who underwent orthopedic surgery. METHODS: This is a retrospective matched cohort study. Inclusion criteria were cirrhosis diagnosis, age > 18 years, ≥ 6 months continuous health plan membership, and a procedure code for orthopedic surgery. Up to five cirrhotic controls without orthopedic surgery were matched on age, gender, and cirrhosis diagnosis date. Data abstraction was performed for demographics, socioeconomics, clinical, and decompensation data. Chart review was performed for validation. Multivariable analysis estimated relative risk of decompensation. RESULTS: Eight hundred fifty-three orthopedic surgery cases in cirrhotics were matched with 4,263 cirrhotic controls. Among the cases and matched controls, the mean age was 60.5 years, and 52.2% were female. Within 90 days after surgery, cases had more decompensation compared to matched controls (12.8% vs 4.9%). Using multivariable analysis, orthopedic surgery, a 0.5 g/dL decrease in serum albumin, and a 1-unit increase in Charlson Comorbidity Index were associated with a significant increase in decompensation within 90 days of surgery. Diabetes, chronic obstructive pulmonary disease, and chronic kidney disease were seen with increased frequency in cases vs. matched controls. CONCLUSIONS: Cirrhotics who underwent orthopedic surgery had a significant increase in hepatic decompensation within 90 days of surgery compared to matched controls. An incremental decrease in serum albumin and an incremental increase in the Charlson Comorbidity Index were significantly associated with hepatic decompensation after surgery.

Fundraising, celebrations and classroom rewards are substantial sources of unhealthy foods and beverages on public school campuses.

OBJECTIVE: The emphasis in school nutrition policy has been on vending and competitive items. Our study was designed to characterize and quantify the amount and source of other foods and beverages on school campuses. DESIGN: A cross-sectional observational study was conducted using a specially designed objective nutrition observation system. SETTING: One low-income school district in southern California with six elementary and two middle schools. SUBJECTS: Data were not collected from individual children. A total of 4033 students, 42 % of whom were Hispanic/Latino, 26 % African American and 21 % non-Hispanic white, were observed across school settings. RESULTS: Data were collected continuously from 9 January 2008 to 16 June 2010. Healthy foods had, per serving, total energy ≤732 kJ (≤175 kcal), total fat content ≤35 %, total saturated fat ≤10 %, sugar less ≤15 g, sodium <200 mg and trans-fat ≤0·5 g. Healthy beverages were only 100 % juice or water, and unflavoured non-fat, 1 %, 2 % milk and soya or rice milk. The system had high inter-rater reliability (r = 0·78 to 0·99), percentage agreement (83 % to 100 %) and test-retest reliability (r = 0·81 to 0·98). Significantly more unhealthy foods and beverages than healthy items were observed on all campuses (P < 0·001). An average of 1·26 (sd 0·46) items per student per week was found with an average of 0·86 (sd 0·34) unhealthy items per child per week. CONCLUSIONS: There were substantial amounts of unhealthy foods and beverages brought onto campuses for classroom rewards, celebrations and fundraising that should be targeted for intervention.

The healthy options for nutrition environments in schools (Healthy ONES) group randomized trial: using implementation models to change nutrition policy and environments in low income schools.

BACKGROUND: The Healthy Options for Nutrition Environments in Schools (Healthy ONES) study was an evidence-based public health (EBPH) randomized group trial that adapted the Institute for Healthcare Improvement's (IHI) rapid improvement process model to implement school nutrition policy and environmental change. METHODS: A low-income school district volunteered for participation in the study. All schools in the district agreed to participate (elementary = 6, middle school = 2) and were randomly assigned within school type to intervention (n = 4) and control (n =4) conditions following a baseline environmental audit year. Intervention goals were to 1) eliminate unhealthy foods and beverages on campus, 2) develop nutrition services as the main source on campus for healthful eating (HE), and 3) promote school staff modeling of HE. Schools were followed across a baseline year and two intervention years. Longitudinal assessment of height and weight was conducted with second, third, and sixth grade children. Behavioral observation of the nutrition environment was used to index the amount of outside foods and beverages on campuses. Observations were made monthly in each targeted school environment and findings were presented as items per child per week. RESULTS: From an eligible 827 second, third, and sixth grade students, baseline height and weight were collected for 444 second and third grade and 135 sixth grade students (51% reach). Data were available for 73% of these enrolled students at the end of three years. Intervention school outside food and beverage items per child per week decreased over time and control school outside food and beverage items increased over time. The effects were especially pronounced for unhealthy foods and beverage items. Changes in rates of obesity for intervention school (28% baseline, 27% year 1, 30% year 2) were similar to those seen for control school (22% baseline, 22% year 1, 25% year 2) children. CONCLUSIONS: Healthy ONES adaptation of IHI's rapid improvement process provided a promising model for implementing nutrition policy and environmental changes that can be used in a variety of school settings. This approach may be especially effective in assisting schools to implement the current federally-mandated wellness policies.

Translating physical activity recommendations for overweight adolescents to steps per day.

BACKGROUND: Steps/day guidelines for children aged 12 years and under are 12,000 to 16,000. There are limited reports in the literature on how many steps/day adolescents need to meet the 60-minutes/day moderate-to-vigorous physical activity (MVPA) recommendation. This study aimed to determine the steps/day that correctly classifies adolescents as meeting the 60-minute recommendation using objective measures. METHODS: This analysis used data from 40 overweight (BMI M=31.5, SD=5.3) adolescents enrolled between 2005 and 2006 in a study to lose weight. Participants were aged 11 to 16 years (M=13.2, SD=1.2), 70.0% girls, and 52.6% non-Hispanic white, 23.7% Hispanic, and 23.7% other race/ethnicity. Participants wore an Actigraph accelerometer for up to 7 days that measured accelerations and step counts simultaneously. The age-adjusted Freedson equation was used to estimate physical activity-intensity minutes. Two criteria were used to define MVPA (3/6 and 4/7 METs). Analyses were conducted between 2008 and 2009. RESULTS: Based on accelerometer data, approximately 48% and 10% of adolescents met the MVPA recommendation depending on the MVPA criteria. Adolescents who met phyiscial activity recommendation had higher MVPA min/day (p<0.001) and higher step counts (p<0.001), after adjusting for gender, age, and ethnicity. Receiver operating characteristic (ROC) curve analyses found that 9930 steps produced 0.84 sensitivity and 0.81 specificity (area under the curve [AUC] =0.89) and 11,714 steps produced 0.75 and 0.81 specificity (AUC=0.94) for meeting the recommendation using the 3/6 and 4/7 criteria, respectively. CONCLUSIONS: Depending on the MVPA criteria used, these data suggest that overweight adolescents are likely to meet national MVPA recommendations if they accumulate between 10,000 and 11,700 steps per day.