Sports analytics in the NFL: classifying the winner of the superbowl.

Sport teams' managers, coaches and players are always looking for new ways to win and stay competitive. The sports analytics field can help teams in gaining a competitive advantage by analyzing historical data and formulating strategies and making data driven decisions regarding game plans, play selection and player recruitment. This work focuses on the application of sports analytics in the National Football League. We compare the classification performance of several methods (C4.5, Neural Network and Random Forest) in classifying the winner of the Superbowl using data collected during the regular season. We split the data into a training set and test set and use the synthetic minority oversampling technique to address the data imbalance issue in the training set. The classification performance is compared on the test set using several measures. According to the findings, the Random Forest classifier had the highest recall, AUC, accuracy and specificity as the oversampling percentage was increased. Our results can be used to develop a decision support tool to assist team managers and coaches in developing strategies that would increase the team's chances of winning.

Classifying highly imbalanced ICU data.

Highly imbalanced data sets are those where the class of interest is rare. In this paper, we compare the performance of several common data mining methods, logistic regression, discriminant analysis, Classification and Regression Tree (CART) models, C5, and Support Vector Machines (SVM) in predicting the discharge status (alive or deceased, with "deceased" being the class of interest) of patients from an Intensive Care Unit (ICU). Using a variety of misclassification cost ratio (MCR) values and using specificity, recall, precision, the F-measure, and confusion entropy (CEN) as criteria for evaluating each method's performance, C5 and SVM performed better than the other methods. At a MCR of 100, C5 had the highest recall and SVM the highest specificity and lowest CEN. We also used Hand's measure to compare the five methods. According to Hand's measure, logistic regression performed the best. This article makes several contributions. We show how the use of MCR for analyzing imbalanced medical data significantly improves the method's classification performance. We also found that the F-measure and precision did not improve as the MCR was increased.

Central nervous system medication use and incident mobility limitation in community elders: the Health, Aging, and Body Composition study.

PURPOSE: To evaluate whether CNS medication use in older adults was associated with a higher risk of future incident mobility limitation. METHODS: This 5-year longitudinal cohort study included 3055 participants from the health, aging and body composition (Health ABC) study who were well-functioning at baseline. CNS medication use (benzodiazepine and opioid receptor agonists, antipsychotics, and antidepressants) was determined yearly (except year 4) during in-home or in-clinic interviews. Summated standardized daily doses (low, medium, and high) and duration of CNS drug use were computed. Incident mobility limitation was operationalized as two consecutive self-reports of having any difficulty walking 1/4 mile or climbing 10 steps without resting every 6 months after baseline. Multivariable Cox proportional hazard analyses were conducted adjusting for demographics, health behaviors, health status, and common indications for CNS medications. RESULTS: Each year at least 13.9% of participants used a CNS medication. By year 6, overall 49% had developed incident mobility limitation. In multivariable models, CNS medication users compared to never users showed a higher risk for incident mobility limitation (adjusted hazard ratio (Adj. HR) 1.28; 95% confidence interval (CI) 1.12-1.47). Similar findings of increased risk were seen in analyses examining dose- and duration-response relationships. CONCLUSIONS: CNS medication use is independently associated with an increased risk of future incident mobility limitation in community dwelling elderly. Further studies are needed to determine the impact of reducing CNS medication exposure on mobility problems.

Effect of central nervous system medication use on decline in cognition in community-dwelling older adults: findings from the Health, Aging And Body Composition Study.

OBJECTIVES: To evaluate whether combined use of multiple central nervous system (CNS) medications over time is associated with cognitive change. DESIGN: Longitudinal cohort study. SETTING: Pittsburgh, Pennsylvania, and Memphis, Tennessee. PARTICIPANTS: Two thousand seven hundred thirty-seven healthy adults (aged > or =65) enrolled in the Health, Aging and Body Composition study without baseline cognitive impairment (modified Mini-Mental State Examination (3MS) score > or =80). MEASUREMENTS: CNS medication (benzodiazepine- and opioid-receptor agonists, antipsychotics, antidepressants) use, duration, and dose were determined at baseline (Year 1) and Years 3 and 5. Cognitive function was measured using the 3MS at baseline and Years 3 and 5. The outcome variables were incident cognitive impairment (3MS score <80) and cognitive decline (> or =5-point decline on 3MS). Multivariable interval-censored survival analyses were conducted. RESULTS: By Year 5, 7.7% of subjects had incident cognitive impairment; 25.2% demonstrated cognitive decline. CNS medication use increased from 13.9% at baseline to 15.3% and 17.1% at Years 3 and 5, respectively. It was not associated with incident cognitive impairment (adjusted hazard ratio (adj HR)=1.11, 95% confidence interval (CI)=0.73-1.69) but was associated with cognitive decline (adj HR 1.37, 95% CI=1.11-1.70). Longer duration (adj HR=1.39, CI=1.08-1.79) and higher doses (>3 standardized daily doses) (adj HR=1.87, 95% CI=1.25-2.79) of CNS medications suggested greater risk of cognitive decline than with nonuse. CONCLUSION: Combined use of CNS medications, especially at higher doses, appears to be associated with cognitive decline in older adults. Future studies must explore the effect of combined CNS medication use on vulnerable older adults.

Provider management of and satisfaction with laboratory testing in the nursing home setting: results of a national internet-based survey.

OBJECTIVES: To describe the management of and satisfaction with laboratory testing, and desirability of laboratory health information technology in the nursing home setting. DESIGN: Cross-sectional study using an Internet-based survey. PARTICIPANTS AND SETTING: National sample of 426 nurse practitioners and 308 physicians who practice in the nursing home setting. MEASUREMENTS: Systems and processes available for ordering and reviewing laboratory tests, laboratory test result management satisfaction, self-reported delays in laboratory test result review, and desirability of computerized laboratory test result management features in the nursing home setting. RESULTS: A total of 96 participants (48 physicians and 48 nurse practitioners) completed the survey, for an overall response rate of 13.1% (96/734). Of the survey participants, 77.1% had worked in the nursing home setting for more than 5 years. Over half of clinicians (52.1%) reported 3 or more recent delays in receiving laboratory test results. Only 43.8% were satisfied with their laboratory test results management. Satisfaction was associated with keeping a list of laboratory orders and availability of computerized laboratory test order entry. In the nursing home, 35.4% of participants reported the ability to electronically review laboratory test results, 12.5% and 10.4% respectively had computerized ordering of chemistry/hematology and microbiology/pathology tests. The following 3 features were rated most desirable in a computerized laboratory test result management system: showing abnormal results first, warning if a test result was missed, and allowing electronic acknowledgment of test results. CONCLUSION: Delays in receiving laboratory test results and dissatisfaction with the management of laboratory test result information are commonly reported among physicians and nurse practitioners working in nursing homes. Test result management satisfaction was associated with computerized order entry and keeping track of ordered laboratory tests, suggesting that implementation of certain health information technology could potentially improve quality of care.

Number and dosage of central nervous system medications on recurrent falls in community elders: the Health, Aging and Body Composition study.

BACKGROUND: Few studies have examined the risk of multiple or high doses of combined central nervous system (CNS) medication use for recurrent falls in the elderly. The study objective was to evaluate whether multiple- or high-dose CNS medication use in older adults was associated with a higher risk of recurrent (>or=2) falls. METHODS: This longitudinal cohort study included 3,055 participants from the Health, Aging and Body Composition study who were well functioning at baseline. CNS medication use (benzodiazepine and opioid receptor agonists, antipsychotics, antidepressants) was determined annually (except Year 4) during in-person interviews. The number and summated standard daily doses (SDDs; low, medium, and high) of CNS medications were computed. Falls 1 year later were ascertained annually for 5 years. RESULTS: For a period of 5 years, as many as 24.1% of CNS medication users took 2+ agents annually, whereas as no more than 18.9% of CNS medication users took high doses annually (3+ SDDs). Yearly, as many as 9.7% of participants reported recurrent falls. Multivariable Generalized Estimating Equation analyses showed that multiple CNS medication users compared with never users had an increased risk of sustaining 2+ falls (adjusted odds ratio [OR] 1.95; 95% confidence interval [CI] 1.35-2.81). Those taking high (3+) CNS SDDs also exhibited an increased risk of 2+ falls (adjusted OR 2.89; 95% CI 1.96-4.25). CONCLUSIONS: Higher total daily doses of CNS medications were associated with recurrent falls. Further studies are needed to determine the impact of reducing the number of CNS medications and/or dosage on recurrent falls.

Assessing the performance characteristics of signals used by a clinical event monitor to detect adverse drug reactions in the nursing home.

Adverse drug reactions (ADRs) are a common cause of morbidity and mortality in the nursing home (NH) setting. Traditional non-automated mechanisms for ADR detection are time-consuming, costly, and fail to detect the majority of ADRs. We describe the implementation and pharmacist evaluation of a clinical event monitor using signals previously developed by our research team to detect potential ADRs in the NH. The overall positive predictive value (PPV) for all signals combined was 81% (54/67), with individual signal PPVs ranging from 0-100%. The PPVs were 53% (10/19) for the antidote signals category and 96% (44/46) for the laboratory/ medication combination signals category. The majority 75% (12/16) of the preventable ADRs were laboratory/medication combination signals. The results suggest that ADRs can be detected in the NH setting with a high degree of accuracy using a clinical event monitor that employs a set of signals derived by expert consensus.

Consensus list of signals to detect potential adverse drug reactions in nursing homes.

OBJECTIVES: To develop a consensus list of agreed-upon laboratory, pharmacy, and Minimum Data Set signals that a computer system can use in the nursing home to detect potential adverse drug reactions (ADRs). DESIGN: Literature search for potential ADR signals, followed by an internet-based, a two-round, modified Delphi survey. SETTING: A nationally representative survey of experts in geriatrics. PARTICIPANTS: Panel of 13 physicians, 10 pharmacists, and 13 advanced practitioners. MEASUREMENTS: Mean score and 95% confidence interval (CI) for each of 80 signals rated on a 5-point Likert scale (5=strong agreement with likelihood of indicating potential ADRs). Consensus agreement indicated by a lower-limit 95% CI of 4.0 or greater. RESULTS: Panelists reached consensus agreement on 40 signals: 15 laboratory and medication combinations, 12 medication concentrations, 10 antidotes, and three Resident Assessment Protocols (RAPs). Highest consensus scores (4.6, 95% CI=4.4-4.9 or 4.4-4.8) were for naloxone when taking opioid analgesics; phytonadione when taking warfarin; dextrose, glucagon, or liquid glucose when taking hypoglycemic agents; medication-induced hypoglycemia; supratherapeutic international normalized ratio when taking warfarin; and triggering the Falls RAP when taking certain medications. CONCLUSION: A multidisciplinary expert panel was able to reach consensus agreement on a list of signals to detect potential ADRs in nursing home residents. The results of this study can be used to prioritize an initial list of signals to be included in paper- or computer-based methods for potential ADR detection.

Improvement in usual gait speed predicts better survival in older adults.

OBJECTIVES: To estimate the relationship between 1-year improvement in measures of health and physical function and 8-year survival. DESIGN: Prospective cohort study. SETTING: Medicare health maintenance organization and Veterans Affairs primary care programs. PARTICIPANTS: Persons aged 65 and older (N=439). MEASUREMENTS: Six measures of health and function assessed at baseline and quarterly over 1 year. Participants were classified as improved at 1 year, transiently improved, or never improved for each measure using a priori definitions of meaningful change: gait speed (usual walking pace over 4 m), 0.1 m/s; Short Physical Performance Battery, 1 point; Medical Outcomes Study 36-item Short Form Health Survey physical function, 10 points; EuroQol, 0.1 point; National Health Interview activity of daily living scale, 2 points; and global health change, two levels or reaching the ceiling. Mortality was ascertained from the National Death Index. Covariates included demographics, comorbidity, cognitive function, and hospitalization. RESULTS: Of the six measures, only improved gait speed was associated with survival. Mortality after 8 years was 31.6%, 41.2%, and 49.3% for those with improved, transiently improved, and never improved gait speed, respectively. The survival benefit for improvement at 1 year persisted after adjustment for covariates (hazard ratio=0.42, 95% confidence interval=0.29-0.61, P<.001) and was consistent across subgroups based on age, sex, ethnicity, initial gait speed, healthcare system, and hospitalization. CONCLUSION: Improvement in usual gait speed predicts a substantial reduction in mortality. Because gait speed is easily measured, clinically interpretable, and potentially modifiable, it may be a useful "vital sign" for older adults. Further research is needed to determine whether interventions to improve gait speed affect survival.