Comparison of methods for auto-coding causation of injury narratives.

Manually reading free-text narratives in large databases to identify the cause of an injury can be very time consuming and recently, there has been much work in automating this process. In particular, the variations of the naïve Bayes model have been used to successfully auto-code free text narratives describing the event/exposure leading to the injury of a workers' compensation claim. This paper compares the naïve Bayes model with an alternative logistic model and found that this new model outperformed the naïve Bayesian model. Further modest improvements were found through the addition of sequences of keywords in the models as opposed to consideration of only single keywords. The programs and weights used in this paper are available upon request to researchers without a training set wishing to automatically assign event codes to large data-sets of text narratives. The utility of sharing this program was tested on an outside set of injury narratives provided by the Bureau of Labor Statistics with promising results.

Development and evaluation of a Naïve Bayesian model for coding causation of workers' compensation claims.

INTRODUCTION: Tracking and trending rates of injuries and illnesses classified as musculoskeletal disorders caused by ergonomic risk factors such as overexertion and repetitive motion (MSDs) and slips, trips, or falls (STFs) in different industry sectors is of high interest to many researchers. Unfortunately, identifying the cause of injuries and illnesses in large datasets such as workers' compensation systems often requires reading and coding the free form accident text narrative for potentially millions of records. METHOD: To alleviate the need for manual coding, this paper describes and evaluates a computer auto-coding algorithm that demonstrated the ability to code millions of claims quickly and accurately by learning from a set of previously manually coded claims. CONCLUSIONS: The auto-coding program was able to code claims as a musculoskeletal disorders, STF or other with approximately 90% accuracy. IMPACT ON INDUSTRY: The program developed and discussed in this paper provides an accurate and efficient method for identifying the causation of workers' compensation claims as a STF or MSD in a large database based on the unstructured text narrative and resulting injury diagnoses. The program coded thousands of claims in minutes. The method described in this paper can be used by researchers and practitioners to relieve the manual burden of reading and identifying the causation of claims as a STF or MSD. Furthermore, the method can be easily generalized to code/classify other unstructured text narratives.

Disability outcomes research: why this supplement, on this topic, at this time?

The objective of this supplement is to define disability outcomes research and to describe the issues facing this new discipline. Drawing on the scientific, English-language literature, this group of articles reviews and comments on the current methods and measures of disability outcomes research, with an eye on providing insights into future directions for disability outcomes research. The future includes expansion of the framework that defined disability and evolution of the methods for the research itself.

Health-related quality of life outcomes measures.

OBJECTIVE: To review critically the features of measures of generic health-related quality of life (HRQOL) for disability outcomes research. DATA SOURCES: A search of electronic databases, summary reviews, books, and government documents was performed. Comment and experiences from participants of a conference on outcomes research were also incorporated. STUDY SELECTION: English language literature from scientists from a broad range of disciplines and research settings, including medicine, nursing, social science, and public health, and health services research and practice. DATA EXTRACTION: A critical review of measures that have been or might be used to measure disability outcomes. DATA SYNTHESIS: Commonly used generic measures of HRQOL can be applied to disability outcomes research with some caveats. Three common tools are the Medical Outcomes Study Short-Form Health Survey (SF-36), Sickness Impact Profile (SIP), and Quality of Well-Being (QWB) scale. The SF-36 and SIP have been used with some success in research with people with disability. The QWB scale has been used less frequently. CONCLUSION: Most studies using generic HRQOL tools are of groups with specific impairments rather than heterogeneous groups of people with disability. None of the tools appears to measure HRQOL without some potential biases (eg, inappropriate wording) for people with disability, but more specific testing of these problems is needed. Also needed are studies to determine whether these tools can measure meaningful longitudinal changes.

Enabling our instruments: accommodation, universal design, and access to participation in research.

The objective of this article is to discuss problems related to full participation of people with disabilities in health services and health outcomes research. To show the problems and to suggest solutions, we offer examples from personal research experiences (ours and colleagues'), as well as from published literature, requirements of research agencies, web and news sources, and research participants' feedback. A combination of formal and informal processes can be used to enable future instruments and methods. There are ethical, legal, and methodologic imperatives for research participation enablement.

A model of outcomes research: spinal cord injury.

OBJECTIVE: To use spinal cord injury (SCI) care and research as a paradigm to illustrate how the principles and practices of outcomes research have been and can be applied to the sequelae of a specific disability. DATA SOURCES: Review of data sources and literature pertaining to outcomes of SCI. STUDY SELECTION: English language literature, health status and health services research agencies, academic and governmental research, and surveillance settings. DATA EXTRACTION: A critical review of measures that have been and may be used to measure the outcomes of SCI. Special attention was paid to data sources; the need for methodologic accommodations: the research balance between generic and condition-specific methods; and the measurement outcomes that are highly relevant to people with SCIs. DATA SYNTHESIS: There is a substantial research record related to the natural history, rehabilitation, survival, and long-term social reintegration of people with SCI, but relatively less addressing widely used generic health outcomes. CONCLUSION: Contemporary outcomes research is relevant to people with SCI and those who provide treatment. One area of special attention is the occurrence of secondary conditions. To a large extent, SCI outcomes research can use conventional methods and generic instruments. There also is a need to modify research methods and to refine and apply some measures specific to people with SCI.

From function to felicitude: physical disability and the search for happiness in health services research.

Historically, research and practice in physical disability have emphasized the destructive and disabling consequences of illness and injury. More recently, we have begun to measure outcomes related to health, satisfaction, and quality-of-life. There has been little systematic attention to the study of happiness, felicitude, but data show that even for many of the people with most notable impairment, levels of happiness are high. The quest for felicitude requires methodological changes and, more significantly, greater professional humility to appreciate that people with both physical and intellectual impairments are able to experience and articulate their own satisfaction, pleasure, and joy. Fortified with this unpretentiousness, we can find happiness and begin to incorporate it as an outcome of clinical and public health programs.

Health-related quality of life in neurology.

CONTEXT: In a climate of growing concern about costs of health care, professional accountability, and the effectiveness and efficiency of treatment, neurologists should understand measures of health-related quality of life (HRQoL) and appreciate both their values and their limits. OBJECTIVE: To provide a critical review of current concepts in the measurement of HRQoL, with particular emphasis on those pertaining to neurologic conditions. DESIGN: A selective overview of measurement strategies and specific instruments. SETTING: Clinical and research settings. PARTICIPANTS: Special attention to people with such conditions as multiple sclerosis and Parkinson disease. MAIN OUTCOME MEASURES: The utility and relevance to neurologists of available measures of HRQoL. RESULTS: There are both generic and condition- or disease-specific measures of HRQoL. Though many measures do not elicit response or are even offensive to people with such conditions as multiple sclerosis and Parkinson disease, other measures pertain directly to their circumstances and needs. CONCLUSIONS: Measures of HRQoL are both meaningful and relevant to neurologists. However, they must be refined-enabled-and used in combinations to address the clinical and existential realities of many neurologic conditions. Arch Neurol. 2000;57:1224-1227

Reliability and validity of disability questions for US Census 2000.

OBJECTIVES: We investigated the validity and proxy reliability of 7 new disability questions from the 2000 US census ("Census 2000"). METHODS: A total of 131 people with disabilities and their proxies from St Louis, Mo, and Massachusetts were interviewed, and responses were compared for concordance. Responses also were compared with responses to questions from the Behavioral Risk Factor Surveillance System (BRFSS) and the Activities of Daily Living (ADL) instrument. RESULTS: Overall, proxies reported more impairment than did people with disabilities, and agreement was low (kappa = 0.24-0.55). Concordance was moderate between the census questions and their BRFSS and ADL counterparts. CONCLUSIONS: The Census 2000 questions may not provide an accurate profile of disability in America.

The "Boston model" of managed care and spinal cord injury: a cross-sectional study of the outcomes of risk-based, prepaid, managed care.

OBJECTIVE: To present preliminary data on the health-related consequences for people with spinal cord injury (SCI) of participation in a prepaid, risk-based, managed care plan. DESIGN: Baseline data from a longitudinal interview survey. SETTING: Massachusetts. PARTICIPANTS: Subjects were 114 independently living adults with SCI recruited from 6 independent living centers and 2 managed care plans; 45 received care from a risk-based prepaid managed care plan and 69 from other sources. MAIN OUTCOMES: Self-reports of use of health services; self-assessments of health, health at interview versus 12 months earlier; hospital admissions; numbers of secondary conditions, and experiences of specific secondary conditions (eg, pressure ulcers, depression, fatigue, and chronic pain). RESULTS: Persons in the managed care plan do not differ from their counterparts in terms of age, gender, education, level of SCI, number of comorbid conditions, activity of daily living profiles, household composition, and reliance upon health maintenance programs and routines (eg, bowel and bladder care). In terms of processes of care, they appear generally better-served. There also are a few differences in outcome that are statistically significant, or approach significance, and generally favor the managed care plan. CONCLUSION: Preliminary data suggest that thoughtfully and sensitively designed and implemented risk-based managed care may improve both access and outcomes.

Alcohol consumption, sensation seeking and ski injury: a case-control study.

OBJECTIVE: Alcohol consumption is known to be associated with both risk of accidental injury and with sensation seeking, and sensation seeking has been found to be common among those engaging in such high-risk activities as skiing. However, few studies have examined the joint association of alcohol consumption and sensation seeking on injury. METHOD: Alcohol consumption prior to injury and sensation-seeking disposition are analyzed on 389 injured skiers (clinic sample) and 899 randomly selected uninjured skiers (trailside sample) at a Northeastern ski resort. Cases and controls were asked questions pertaining to drinking within 24 hours, amount consumed, time lapsed between the last drink and the event, skiing ability, and sensation seeking. RESULTS: The clinic sample was more likely to be female, to have less skiing experience, to score lower on sensation seeking and to have been drinking within 24 hours compared to the trailside sample. However, they were less likely to have had six or more drinks within 24 hours and were more likely to report a greater time lapse between the last drink and injury or interview. A larger proportion of those who reported drinking in both samples scored high on sensation seeking compared to those who reported not drinking. Logistic regression analysis found the following variables predictive of ski injury: female, low on sensation seeking, amount of alcohol consumed prior to the event, a longer time lapsed between drinking and the event, time of day (later) and day of the week (weekend). CONCLUSIONS: The data suggest that, while drinking within 24 hours is positively associated with sensation seeking, drinking and not sensation seeking is positively predictive of injury. Drinking at least 12 hours prior to skiing, not drinking in close proximity to skiing, may increase risk for accidental injury, possibly due to a hangover or residual alcohol effect in which fatigue may play a part.

Primary care for persons with disabilities. The Boston, Massachusetts model program.

Boston's Community Medical Group (BCMG) was among the first primary care group practices to provide community-based continuous primary care to people with major disabling conditions, the first to rely on nurse-practitioners as primary care-givers, and one of the first to provide care on a prepaid capitated basis. With more than one thousand person-years' experience, BCMG has demonstrated that it is ethically and operationally feasible (1) to provide prepaid managed care to people with major disabling conditions; (2) to share financial risk for that care with providers; (3) to reinforce principles of independent living and consumer autonomy; (4) to assure high-quality clinical outcomes at reasonable costs.

Drinking by police officers, general drivers and late-night drivers.

OBJECTIVE: This study compares police officers' reported alcohol use with similar reports from age- and gender-matched samples from the general driving population (GDP) and drivers recruited at high-risk hours who are most likely to drink and drive (roadside drivers). METHOD: Data were derived from interviews with 239 police officers who spent at least 5% of their time on alcohol or traffic enforcement; 243 subjects from the GDP, selected by random-digit dialing; and 249 drivers recruited at roadside on Friday and Saturday nights between 10:00 PM and 3:00 AM. RESULTS: Officers report drinking less frequently and in smaller quantities than either of the driver samples. Differences between police and general drivers are modest: e.g., usual frequency of drinking (1 to < 3 vs 3 or 4 times/month; p < or = .001) and number of days/average month with > or = 2 drinks/day (mean = 2.5 [+/- 3.4] vs 4.9 [+/- 6.5]; p < or = .001). There are greater differences between police officers and roadside drivers: e.g., 14.2% and 2.4%, respectively, abstain (vs 15.6% of general drivers; police vs general drivers; NS; police and general drivers vs roadside drivers, p < or = .001). CONCLUSIONS: The data suggest that officers' drinking differs significantly from that of drivers most likely to be driving under the influence of alcohol (DUI). If this is so, police training programs pertaining to DUI and other alcohol-related offenses should pay explicit attention to the effects of alcohol and not assume that all officers have experienced them.

Predictors of medical care utilization by independently living adults with spinal cord injuries.

A prospective study of 87 independently living adults with spinal cord injury (SCI) as a major disabling condition showed the following average annual health care utilization rates: 1.3 hospital admissions, 16.8 days hospitalized, 1.7 emergency room (ER) visits, and 22.4 outpatient contacts (in person or by telephone). Those hospitalized (n = 66) experienced a mean of 22.2 days hospitalized per person per year. Mean length of stay (LOS) was 11.1 days per admission. Stepwise regression analysis indicated no statistically significant (p less than or equal to .05) predictors of hospital admissions. There were three independent predictors of days hospitalized (greater age, fewer years of education, and more days hospitalized during the previous year), three predictors of days hospitalized for those hospitalized only (greater age, fewer years of education, and longer hospital LOS during the previous year), one predictor of LOS (self-assessment of health), three of emergency room (ER) visits (more unmet instrumental activities of daily living needs, lack of organizational memberships, and more ER visits during the previous year), and five predictors of outpatient contacts (greater age, less satisfaction with health care providers' expressions of concern for their health, lower frequency of leaving apartments, lower levels of life satisfaction, and nonparticipation in a managed medical care demonstration project). Many predictors of health services utilization are immutable. However, changes which facilitate social interaction and changes in the organization of health services may reduce certain types of medical care utilization by people with SCI.

Alcohol, tobacco, and cannabis use by independently living adults with major disabling conditions.

Data from an 18-month prospective study of 205 adults with major disabling conditions revealed that there is a positive but not statistically significant association (p greater than .10) between respondents' scores on a scale which measures both numbers and amounts of substances (alcohol, tobacco, and cannabis) used and bed disability days. There are also nonsignificant positive relationships between scale scores and five of six measures of utilization of medical care.

The epidemiology of medical care utilization by severely-disabled independently-living adults.

A prospective study of the medical care utilization experience of 205 severely-disabled independently-living adults in Eastern Massachusetts shows that there was a mean of 0.83 +/- 1.26 hospital admissions, 9.9 +/- 22.7 hospital days, 1.5 +/- 2.31 emergency room (ER) visits, and 26.88 +/- 44.4 outpatient contacts per person per year. Among those hospitalized, the mean experience was 16.2 +/- 27.1 days per person per year; mean length-of-stay was 9.3 +/- 14.7 days per admission. Regression analysis indicates that those with spinal cord injuries as major disabling conditions were significantly more likely to be hospitalized. So were those with lower self-assessments of health, higher levels of depressions, and more baseline ER visits. Self-assessment of health is a significant predictor of hospital days for the total cohort (including those with no admissions); so are age at onset of disability (greater age; higher risk), and bed disability days in the month before the baseline survey (more disability days; higher risk). Among those hospitalized, the total number of days hospitalized is significantly related to both age at onset of disability (later onset; more days) and baseline days hospitalized (greater number; more days). Lengths-of-stay are significantly related to two factors; age and age at onset of disability (in both cases, greater age associated with longer stays). Prior ER visits are a significant predictor of subsequent ER visits (more baseline; more subsequent); so are respondents' reported satisfaction with their participation in their medical care (lower reported satisfaction; more ER visits), organizational affiliations, and frequencies of contacts with friends or relatives. Higher levels of social interaction (i.e. organizational affiliation and more frequent social contacts) were associated with more ER visits. Prior contacts with physicians, nurse-practitioners, or physician-assistants was the most powerful predictor of subsequent outpatient contacts (more baseline; more subsequent). There were also significant relationships between subsequent contacts and respondents' assessments of their health relative to others with similar disabilities (relatively worse health; more contacts), age (greater age; more contacts), and baseline ER visits (more visits; more contacts).

'Lumping it': the hidden denominator of the medical malpractice crisis.

In a recent article, Miller has reminded us that medical malpractice litigation is not simply an economic problem which inhibits medical practice and increases health care costs. She argues that it has three broader "societal objectives": reparation, emotional vindication, and deterrence. Viewed in the broader perspective of social values, the Maine data suggest that our current approach to medical malpractice does not perform well. Significant numbers of respondents believe that they have been neither vindicated nor compensated for their own or their relatives' illness, injury, or death; and that they have not had the opportunity to protect others from harm. As Miller suggests in her review of British alternatives to medical malpractice litigation, there may be more efficient and effective means of reparation. There may also be more direct and less costly means to deter incompetent practitioners and vindicate those who are harmed. We shall never discover these alternatives if we view the medical malpractice "crisis" as a simple or straightforward problem of costs of premiums, costs of settlements, and costs of judgments; numerators. Medical malpractice litigation is the expression of deep and highly complicated problems, which cannot be solved or even significantly alleviated by false solutions motivated only by concerns of costs and cost containment. They can be addressed only by careful, thoughtful, and comprehensive analysis.