Why does the adverse effect of inappropriate MRI for LBP vary by geographic location? An exploratory analysis.

BACKGROUND: Early magnetic resonance imaging (eMRI) for nonspecific low back pain (LBP) not adherent to clinical guidelines is linked with prolonged work disability. Although the prevalence of eMRI for occupational LBP varies substantially among states, it is unknown whether the risk of prolonged disability associated with eMRI varies according to individual and area-level characteristics. The aim was to explore whether the known risk of increased length of disability (LOD) associated with eMRI scanning not adherent to guidelines for occupational LBP varies according to patient and area-level characteristics, and the potential reasons for any observed variations. METHODS: A retrospective cohort of 59,360 LBP cases from 49 states, filed between 2002 and 2008, and examined LOD as the outcome. LBP cases with at least 1 day of work disability were identified by reviewing indemnity service records and medical bills using a comprehensive list of codes from the International Classification of Diseases, Ninth Edition (ICD-9) indicating LBP or nonspecific back pain, excluding medically complicated cases. RESULTS: We found significant between-state variations in the negative impact of eMRI on LOD ranging from 3.4 days in Tennessee to 14.8 days in New Hampshire. Higher negative impact of eMRI on LOD was mainly associated with female gender, state workers' compensation (WC) policy not limiting initial treating provider choice, higher state orthopedic surgeon density, and lower state MRI facility density. CONCLUSION: State WC policies regulating selection of healthcare provider and structural factors affecting quality of medical care modify the impact of eMRI not adherent to guidelines. Targeted healthcare and work disability prevention interventions may improve work disability outcomes in patients with occupational LBP.

Regional socioeconomic disparities in outcomes for workers with low back pain in the United States.

BACKGROUND: Although regional socioeconomic (SE) factors have been associated with worse health outcomes, prior studies have not addressed important confounders or work disability. METHODS: A national sample of 59 360 workers' compensation (WC) cases to evaluate impact of regional SE factors on medical costs and length of disability (LOD) in occupational low back pain (LBP). RESULTS: Lower neighborhood median household incomes (MHI) and higher state unemployment rates were associated with longer LOD. Medical costs were lower in states with more workers receiving Social Security Disability, and in areas with lower MHI, but this varied in magnitude and direction among neighborhoods. Medical costs were higher in more urban, more racially diverse, and lower education neighborhoods. CONCLUSIONS: Regional SE disparities in medical costs and LOD occur even when health insurance, health care availability, and indemnity benefits are similar. Results suggest opportunities to improve care and disability outcomes through targeted health care and disability interventions.

Harnessing information from injury narratives in the 'big data' era: understanding and applying machine learning for injury surveillance.

OBJECTIVE: Vast amounts of injury narratives are collected daily and are available electronically in real time and have great potential for use in injury surveillance and evaluation. Machine learning algorithms have been developed to assist in identifying cases and classifying mechanisms leading to injury in a much timelier manner than is possible when relying on manual coding of narratives. The aim of this paper is to describe the background, growth, value, challenges and future directions of machine learning as applied to injury surveillance. METHODS: This paper reviews key aspects of machine learning using injury narratives, providing a case study to demonstrate an application to an established human-machine learning approach. RESULTS: The range of applications and utility of narrative text has increased greatly with advancements in computing techniques over time. Practical and feasible methods exist for semiautomatic classification of injury narratives which are accurate, efficient and meaningful. The human-machine learning approach described in the case study achieved high sensitivity and PPV and reduced the need for human coding to less than a third of cases in one large occupational injury database. CONCLUSIONS: The last 20 years have seen a dramatic change in the potential for technological advancements in injury surveillance. Machine learning of 'big injury narrative data' opens up many possibilities for expanded sources of data which can provide more comprehensive, ongoing and timely surveillance to inform future injury prevention policy and practice.

Combining statistics from two national complex surveys to estimate injury rates per hour exposed and variance by activity in the USA.

BACKGROUND: A common issue in descriptive injury epidemiology is that in order to calculate injury rates that account for the time spent in an activity, both injury cases and exposure time of specific activities need to be collected. In reality, few national surveys have this capacity. To address this issue, we combined statistics from two different national complex surveys as inputs for the numerator and denominator to estimate injury rate, accounting for the time spent in specific activities and included a procedure to estimate variance using the combined surveys. METHODS: The 2010 National Health Interview Survey (NHIS) was used to quantify injuries, and the 2010 American Time Use Survey (ATUS) was used to quantify time of exposure to specific activities. The injury rate was estimated by dividing the average number of injuries (from NHIS) by average exposure hours (from ATUS), both measured for specific activities. The variance was calculated using the 'delta method', a general method for variance estimation with complex surveys. RESULTS: Among the five types of injuries examined, 'sport and exercise' had the highest rate (12.64 injuries per 100 000 h), followed by 'working around house/yard' (6.14), driving/riding a motor vehicle (2.98), working (1.45) and sleeping/resting/eating/drinking (0.23). The results show a ranking of injury rate by activity quite different from estimates using population as the denominator. CONCLUSIONS: Our approach produces an estimate of injury risk which includes activity exposure time and may more reliably reflect the underlying injury risks, offering an alternative method for injury surveillance and research.

Differences in time use and activity patterns when adding a second job: implications for health and safety in the United States.

OBJECTIVES: We compared work and lifestyle activities for workers who work in 1 job with those who work in multiple jobs during a 1-week period. METHODS: We used information from the 2003-2011 American Time Use Survey to classify workers into 6 work groups based on whether they were a single (SJH) or multiple (MJH) job holder and whether they worked their primary, other, multiple, or no job on the diary day. RESULTS: The MJHs often worked 2 part-time jobs (20%), long weekly hours (27% worked 60+ hours), and on weekends. The MJHs working multiple jobs on the diary day averaged more than 2 additional work hours (2.25 weekday, 2.75 weekend day; P < .05), odd hours (more often between 5 pm and 7 am), with more work travel time (10 minutes weekday, 9 minutes weekend day; P < .05) and less sleep (-45 minutes weekday, -62 minutes weekend day; P < .05) and time for other household (P < .05) and leisure (P < .05) activities than SJHs. CONCLUSIONS: Because of long work hours, long daily commutes, multiple shifts, and less sleep and leisure time, MJHs may be at heightened risk of fatigue and injury.

Work in multiple jobs and the risk of injury in the US working population.

OBJECTIVES: We compared the risk of injury for multiple job holders (MJHs) with that for single job holders (SJHs). METHODS: We used information from the National Health Interview Survey for the years 1997 through 2011 to estimate the rate of multiple job holding in the United States and compared characteristics and rates of self-reported injury (work and nonwork) for SJHs versus MJHs. RESULTS: Approximately 8.4% of those employed reported working more than 1 job in the week before the interview. The rate of work and nonwork injury episodes per 100 employed workers was higher for MJHs than for SJHs (4.2; 95% confidence interval [CI] = 3.5, 4.8; vs 3.3; 95% CI = 3.1, 3.5 work injuries and 9.9; 95% CI = 8.9, 10.9; vs 7.4; 95% CI = 7.1, 7.6 nonwork injuries per 100 workers, respectively). When calculated per 100 full-time equivalents (P < .05), the rate ratio remained higher for MJHs. CONCLUSIONS: Our findings suggest that working in multiple jobs is associated with an increased risk of an injury, both at work and not at work, and should be considered in injury surveillance.

Classifying injury narratives of large administrative databases for surveillance-A practical approach combining machine learning ensembles and human review.

Injury narratives are now available real time and include useful information for injury surveillance and prevention. However, manual classification of the cause or events leading to injury found in large batches of narratives, such as workers compensation claims databases, can be prohibitive. In this study we compare the utility of four machine learning algorithms (Naïve Bayes, Single word and Bi-gram models, Support Vector Machine and Logistic Regression) for classifying narratives into Bureau of Labor Statistics Occupational Injury and Illness event leading to injury classifications for a large workers compensation database. These algorithms are known to do well classifying narrative text and are fairly easy to implement with off-the-shelf software packages such as Python. We propose human-machine learning ensemble approaches which maximize the power and accuracy of the algorithms for machine-assigned codes and allow for strategic filtering of rare, emerging or ambiguous narratives for manual review. We compare human-machine approaches based on filtering on the prediction strength of the classifier vs. agreement between algorithms. Regularized Logistic Regression (LR) was the best performing algorithm alone. Using this algorithm and filtering out the bottom 30% of predictions for manual review resulted in high accuracy (overall sensitivity/positive predictive value of 0.89) of the final machine-human coded dataset. The best pairings of algorithms included Naïve Bayes with Support Vector Machine whereby the triple ensemble NBSW=NBBI-GRAM=SVM had very high performance (0.93 overall sensitivity/positive predictive value and high accuracy (i.e. high sensitivity and positive predictive values)) across both large and small categories leaving 41% of the narratives for manual review. Integrating LR into this ensemble mix improved performance only slightly. For large administrative datasets we propose incorporation of methods based on human-machine pairings such as we have done here, utilizing readily-available off-the-shelf machine learning techniques and resulting in only a fraction of narratives that require manual review. Human-machine ensemble methods are likely to improve performance over total manual coding.

Working multiple jobs over a day or a week: Short-term effects on sleep duration.

Approximately 10% of the employed population in the United States works in multiple jobs. They are more likely to work long hours and in nonstandard work schedules, factors known to impact sleep duration and quality, and increase the risk of injury. In this study we used multivariate regression models to compare the duration of sleep in a 24-hour period between workers working in multiple jobs (MJHs) with single job holders (SJHs) controlling for other work schedule and demographic factors. We used data from the Bureau of Labor Statistics US American Time Use Survey (ATUS) pooled over a 9-year period (2003-2011). We found that MJHs had significantly reduced sleep duration compared with SJHs due to a number of independent factors, such as working longer hours and more often late at night. Male MJHs, working in their primary job or more than one job on the diary day, also had significantly shorter sleep durations (up to 40 minutes less on a weekend day) than male SJHs, even after controlling for all other factors. Therefore, duration of work hours, time of day working and duration of travel for work may not be the only factors to consider when understanding if male MJHs are able to fit in enough recuperative rest from their busy schedule. Work at night had the greatest impact on sleep duration for females, reducing sleep time by almost an hour compared with females who did not work at night. We also hypothesize that the high frequency or fragmentation of non-leisure activities (e.g. work and travel for work) throughout the day and between jobs may have an additional impact on the duration and quality of sleep for MJHs.

Falls and Fall-Related Injuries among Community-Dwelling Adults in the United States.

INTRODUCTION: Falls are the leading cause of unintentional injuries in the U.S.; however, national estimates for all community-dwelling adults are lacking. This study estimated the national incidence of falls and fall-related injuries among community-dwelling U.S. adults by age and gender and the trends in fall-related injuries across the adult life span. METHODS: Nationally representative data from the National Health Interview Survey (NHIS) 2008 Balance and Dizziness supplement was used to develop national estimates of falls, and pooled data from the NHIS was used to calculate estimates of fall-related injuries in the U.S. and related trends from 2004-2013. Costs of unintentional fall-related injuries were extracted from the CDC's Web-based Injury Statistics Query and Reporting System. RESULTS: Twelve percent of community-dwelling U.S. adults reported falling in the previous year for a total estimate of 80 million falls at a rate of 37.2 falls per 100 person-years. On average, 9.9 million fall-related injuries occurred each year with a rate of 4.38 fall-related injuries per 100 person-years. In the previous three months, 2.0% of older adults (65+), 1.1% of middle-aged adults (45-64) and 0.7% of young adults (18-44) reported a fall-related injury. Of all fall-related injuries among community-dwelling adults, 32.3% occurred among older adults, 35.3% among middle-aged adults and 32.3% among younger adults. The age-adjusted rate of fall-related injuries increased 4% per year among older women (95% CI 1%-7%) from 2004 to 2013. Among U.S. adults, the total lifetime cost of annual unintentional fall-related injuries that resulted in a fatality, hospitalization or treatment in an emergency department was 111 billion U.S. dollars in 2010. CONCLUSIONS: Falls and fall-related injuries represent a significant health and safety problem for adults of all ages. The findings suggest that adult fall prevention efforts should consider the entire adult lifespan to ensure a greater public health benefit.

A practical tool for public health surveillance: Semi-automated coding of short injury narratives from large administrative databases using Naïve Bayes algorithms.

Public health surveillance programs in the U.S. are undergoing landmark changes with the availability of electronic health records and advancements in information technology. Injury narratives gathered from hospital records, workers compensation claims or national surveys can be very useful for identifying antecedents to injury or emerging risks. However, classifying narratives manually can become prohibitive for large datasets. The purpose of this study was to develop a human-machine system that could be relatively easily tailored to routinely and accurately classify injury narratives from large administrative databases such as workers compensation. We used a semi-automated approach based on two Naïve Bayesian algorithms to classify 15,000 workers compensation narratives into two-digit Bureau of Labor Statistics (BLS) event (leading to injury) codes. Narratives were filtered out for manual review if the algorithms disagreed or made weak predictions. This approach resulted in an overall accuracy of 87%, with consistently high positive predictive values across all two-digit BLS event categories including the very small categories (e.g., exposure to noise, needle sticks). The Naïve Bayes algorithms were able to identify and accurately machine code most narratives leaving only 32% (4853) for manual review. This strategy substantially reduces the need for resources compared with manual review alone.

Length of Disability and Medical Costs in Low Back Pain: Do State Workers' Compensation Policies Make a Difference?

OBJECTIVE: The aim of the study was to examine the impact of state workers' compensation (WC) policies regarding wage replacement and medical benefits on medical costs and length of disability (LOD) in workers with low back pain (LBP). METHODS: Retrospective cohort analysis of LBP claims from 49 states (n = 59,360) filed between 2002 and 2008, extracted from a large WC administrative database. RESULTS: Longer retroactive periods and state WC laws allowing treating provider choice were associated with higher medical costs and longer LOD. Limiting the option to change providers and having a fee schedule were associated with longer LOD, except that allowing a one-time treating provider change was associated with lower medical costs and shorter LOD. CONCLUSIONS: WC policies about wage replacement and medical treatment appear to be associated with WC LBP outcomes, and might represent opportunities to improve LOD and reduce medical costs in occupational LBP.

The direct cost burden of 13years of disabling workplace injuries in the U.S. (1998-2010): Findings from the Liberty Mutual Workplace Safety Index.

INTRODUCTION: Although occupational injuries are among the leading causes of death and disability around the world, the burden due to occupational injuries has historically been under-recognized, obscuring the need to address a major public health problem. METHODS: We established the Liberty Mutual Workplace Safety Index (LMWSI) to provide a reliable annual metric of the leading causes of the most serious workplace injuries in the United States based on direct workers compensation (WC) costs. RESULTS: More than $600 billion in direct WC costs were spent on the most disabling compensable non-fatal injuries and illnesses in the United States from 1998 to 2010. The burden in 2010 remained similar to the burden in 1998 in real terms. The categories of overexertion ($13.6B, 2010) and fall on same level ($8.6B, 2010) were consistently ranked 1st and 2nd. PRACTICAL APPLICATION: The LMWSI was created to establish the relative burdens of events leading to work-related injury so they could be better recognized and prioritized. Such a ranking might be used to develop research goals and interventions to reduce the burden of workplace injury in the United States.