Temporal electronic phenotyping by mining careflows of breast cancer patients.

In this work we present a careflow mining approach designed to analyze heterogeneous longitudinal data and to identify phenotypes in a patient cohort. The main idea underlying our approach is to combine methods derived from sequential pattern mining and temporal data mining to derive frequent healthcare histories (careflows) in a population of patients. This approach was applied to an integrated data repository containing clinical and administrative data of more than 4000 breast cancer patients. We used the mined histories to identify sub-cohorts of patients grouped according to healthcare activities pathways, then we characterized these sub-cohorts with clinical data. In this way, we were able to perform temporal electronic phenotyping of electronic health records (EHR) data.

medpie: an information extraction package for medical message board posts.

SUMMARY: We have developed medpie, a software package for preparing medical message board corpora and extracting patient mentions and statistics for drugs, herbs and adverse effects experienced from them. The package is divided into web-crawling, HTML-cleaning, de-identification and information extraction modules. It also includes a sample controlled vocabulary of drugs, herbs and adverse effect terms. AVAILABILITY: http://www.cis.upenn.edu/~ungar/medpie.zip. DEPENDENCIES: Python 2.6 or 2.7.

Demonstration of the safety and effectiveness of the RECELL® System combined with split-thickness meshed autografts for the reduction of donor skin to treat mixed-depth burn injuries.

INTRODUCTION: Split-thickness skin grafts (STSG) are the standard of care (SOC) for burns undergoing autografting but are associated with donor skin site morbidity and limited by the availability of uninjured skin. The RECELL® Autologous Cell Harvesting Device (RECELL® System, or RECELL) was developed for point-of-care preparation and application of a suspension of non-cultured, disaggregated, autologous skin cells, using 1cm2 of the patient's skin to treat up to 80cm2 of excised burn. METHODS: A multi-center, prospective, within-subject controlled, randomized, clinical trial was conducted with 30 subjects to evaluate RECELL in combination with a more widely meshed STSG than a pre-defined SOC meshed STSG (RECELL treatment) for the treatment of mixed-depth burns, including full-thickness. Treatment areas were randomized to receive standard meshed STSG (Control treatment) or RECELL treatment, such that each subject had 1 Control and 1 RECELL treatment area. Effectiveness measures were assessed and included complete wound closure, donor skin use, subject satisfaction, and scarring outcomes out to one year following treatment. RESULTS: At 8 weeks, 85% of the Control-treated wounds were healed compared with 92% of the RECELL-treated wounds, establishing the non-inferiority of RECELL treatment for wound healing. Control-treated and RECELL-treated wounds were similar in mean size; however, mean donor skin use was significantly reduced by 32% with the use of RECELL (p<0.001), establishing the superiority of RECELL treatment for reducing donor skin requirements. Secondary effectiveness and safety outcomes were similar between the treatments. CONCLUSIONS: In combination with widely meshed STSG, RECELL is a safe and effective point-of-care treatment for mixed-depth burns without confluent dermis, achieving short- and long-term healing comparable to standard STSG, while significantly decreasing donor skin use.

Transforming Data into Knowledge: How to Improve the Efficiency of Clinical Care?

Objectives: To provide an introduction to the 2017 IMIA Yearbook of Medical Informatics by the editors. Methods: We present a brief overview of the 2017 special topic "Learning from experience: Secondary use of patient data". We review our choice of special topic section editors, present the new section "Health Information Management", and discuss transitions in the editorial team. Results: In this edition of the Yearbook, we focused on one of the most important issues for the medical informatics community: The secondary use of clinical data. With the ubiquitous adoption of electronic health records (EHRs) and the increasing availability of genomic and environmental data, as well as the accessibility of unstructured data in social media, issues related to data integration, storage, and management, as well as the need for novel analytic approaches are clear challenges. The paradigm of Learning Health Systems (LHSs) is presented in the keynote paper and survey papers review the significant developments in allied fields such as clinical research, clinical systems, translational informatics, and public health over the past two years. IMIA Working Groups also contributed to this topic. Conclusion: The 2017 issue of the IMIA yearbook focuses on the secondary use of patient data and presents the difficulties that still need to be solved before witnessing the actual development of LHSs.

Progress in Biomedical Knowledge Discovery: A 25-year Retrospective.

OBJECTIVES: We sought to explore, via a systematic review of the literature, the state of the art of knowledge discovery in biomedical databases as it existed in 1992, and then now, 25 years later, mainly focused on supervised learning. METHODS: We performed a rigorous systematic search of PubMed and latent Dirichlet allocation to identify themes in the literature and trends in the science of knowledge discovery in and between time periods and compare these trends. We restricted the result set using a bracket of five years previous, such that the 1992 result set was restricted to articles published between 1987 and 1992, and the 2015 set between 2011 and 2015. This was to reflect the current literature available at the time to researchers and others at the target dates of 1992 and 2015. The search term was framed as: Knowledge Discovery OR Data Mining OR Pattern Discovery OR Pattern Recognition, Automated. RESULTS: A total 538 and 18,172 documents were retrieved for 1992 and 2015, respectively. The number and type of data sources increased dramatically over the observation period, primarily due to the advent of electronic clinical systems. The period 1992- 2015 saw the emergence of new areas of research in knowledge discovery, and the refinement and application of machine learning approaches that were nascent or unknown in 1992. CONCLUSIONS: Over the 25 years of the observation period, we identified numerous developments that impacted the science of knowledge discovery, including the availability of new forms of data, new machine learning algorithms, and new application domains. Through a bibliometric analysis we examine the striking changes in the availability of highly heterogeneous data resources, the evolution of new algorithmic approaches to knowledge discovery, and we consider from legal, social, and political perspectives possible explanations of the growth of the field. Finally, we reflect on the achievements of the past 25 years to consider what the next 25 years will bring with regard to the availability of even more complex data and to the methods that could be, and are being now developed for the discovery of new knowledge in biomedical data.

Effect of testosterone treatment on bone mineral density in men over 65 years of age.

As men age, their serum testosterone concentrations decrease, as do their bone densities. Because bone density is also low in hypogonadal men, we hypothesized that increasing the serum testosterone concentrations of men over 65 yr to those found in young men would increase their bone densities. We randomized 108 men over 65 yr of age to wear either a testosterone patch or a placebo patch double blindly for 36 months. We measured bone mineral density by dual energy x-ray absorptiometry before and during treatment. Ninety-six men completed the entire 36-month protocol. The mean serum testosterone concentration in the men treated with testosterone increased from 367 +/- 79 ng/dL (+/-SD; 12.7 +/- 2.7 nmol/L) before treatment to 625 +/- 249 ng/dL (21.7 +/- 8.6 nmol/L; P < 0.001) at 6 months of treatment and remained at that level for the duration of the study. The mean bone mineral density of the lumbar spine increased (P < 0.001) in both the placebo-treated (2.5 +/- 0.6%) and testosterone-treated (4.2 +/- 0.8%) groups, but the mean changes did not differ between the groups. Linear regression analysis, however, demonstrated that the lower the pretreatment serum testosterone concentration, the greater the effect of testosterone treatment on lumbar spine bone density from 0-36 months (P = 0.02). This analysis showed a minimal effect (0.9 +/- 1.0%) of testosterone treatment on bone mineral density for a pretreatment serum testosterone concentration of 400 ng/dL (13.9 nmol/L), but an increase of 5.9 +/- 2.2% for a pretreatment testosterone concentration of 200 ng/dL (6.9 nmol/L). Increasing the serum testosterone concentrations of normal men over 65 yr of age to the midnormal range for young men did not increase lumbar spine bone density overall, but did increase it in those men with low pretreatment serum testosterone concentrations.

Effect of testosterone treatment on body composition and muscle strength in men over 65 years of age.

As men age, serum testosterone concentrations decrease, the percentage of body mass that is fat increases, the percentage of lean body mass decreases, and muscle strength decreases. Because these changes are similar to those that occur in hypogonadal men, we hypothesized that increasing the serum testosterone concentration of men over 65 yr of age to that in young men would decrease their fat mass, increase their lean mass, and increase their muscle strength. We randomized 108 men over 65 yr of age to wear either a testosterone patch or a placebo patch in a double blind study for 36 months. We measured body composition by dual energy x-ray absorptiometry and muscle strength by dynamometer before and during treatment. Ninety-six men completed the entire 36-month protocol. Fat mass decreased (-3.0+/-0.5 kg) in the testosterone-treated men during the 36 months of treatment, which was significantly different (P = 0.001) from the decrease (-0.7+/-0.5 kg) in the placebo-treated men. Lean mass increased (1.9+/-0.3 kg) in the testosterone-treated men, which was significantly different (P < 0.001) from that (0.2+/-0.2 kg) in the placebo-treated men. The decrease in fat mass in the testosterone-treated men was principally in the arms (-0.7+/-0.1 kg; P < 0.001 compared to the placebo group) and legs (-1.1+/-0.2 kg; P < 0.001), and the increase in lean mass was principally in the trunk (1.9+/-0.3 kg; P < 0.001). The change in strength of knee extension and flexion at 60 degrees and 180 degrees angular velocity during treatment, however, was not significantly different between the two groups. We conclude that increasing the serum testosterone concentrations of normal men over 65 yr of age to the midnormal range for young men decreased fat mass, principally in the arms and legs, and increased lean mass, principally in the trunk, but did not increase the strength of knee extension and flexion, as measured by dynamometer.

Effects of testosterone replacement in hypogonadal men.

Treatment of hypogonadal men with testosterone has been shown to ameliorate the effects of testosterone deficiency on bone, muscle, erythropoiesis, and the prostate. Most previous studies, however, have employed somewhat pharmacological doses of testosterone esters, which could result in exaggerated effects, and/or have been of relatively short duration or employed previously treated men, which could result in dampened effects. The goal of this study was to determine the magnitude and time course of the effects of physiological testosterone replacement for 3 yr on bone density, muscle mass and strength, erythropoiesis, prostate volume, energy, sexual function, and lipids in previously untreated hypogonadal men. We selected 18 men who were hypogonadal (mean serum testosterone +/- SD, 78 +/- 77 ng/dL; 2.7 +/- 2.7 nmol/L) due to organic disease and had never previously been treated for hypogonadism. We treated them with testosterone transdermally for 3 yr. Sixteen men completed 12 months of the protocol, and 14 men completed 36 months. The mean serum testosterone concentration reached the normal range by 3 months of treatment and remained there for the duration of treatment. Bone mineral density of the lumbar spine (L2-L4) increased by 7.7 +/- 7.6% (P < 0.001), and that of the femoral trochanter increased by 4.0 +/- 5.4% (P = 0.02); both reached maximum values by 24 months. Fat-free mass increased 3.1 kg (P = 0.004), and fat-free mass of the arms and legs individually increased, principally within the first 6 months. The decrease in fat mass was not statistically significant. Strength of knee flexion and extension did not change. Hematocrit increased dramatically, from mildly anemic (38.0 +/- 3.0%) to midnormal (43.1 +/- 4.0%; P = 0.002) within 3 months, and remained at that level for the duration of treatment. Prostate volume also increased dramatically, from subnormal (12.0 +/- 6.0 mL) before treatment to normal (22.4 +/- 8.4 mL; P = 0.004), principally during the first 6 months. Self-reported sense of energy (49 +/- 19% to 66 +/- 24%; P = 0.01) and sexual function (24 +/- 20% to 66 +/- 24%; P < 0.001) also increased, principally within the first 3 months. Lipids did not change. We conclude from this study that replacing testosterone in hypogonadal men increases bone mineral density of the spine and hip, fat-free mass, prostate volume, erythropoiesis, energy, and sexual function. The full effect of testosterone on bone mineral density took 24 months, but the full effects on the other tissues took only 3-6 months. These results provide the basis for monitoring the magnitude and the time course of the effects of testosterone replacement in hypogonadal men.

Effect of transdermal testosterone treatment on serum lipid and apolipoprotein levels in men more than 65 years of age.

PURPOSE: Because the effects of androgen replacement on lipoprotein levels are uncertain, we sought to determine the effect of transdermal testosterone treatment on serum lipid and apolipoprotein levels in elderly men. SUBJECTS AND METHODS: One hundred and eight healthy men more than 65 years of age who had serum testosterone concentrations >1 SD below the mean for young men were randomly assigned to receive either testosterone (54 men; 6 mg/day) or placebo (54 men) transdermally in a double-blind fashion for 36 months. Serum concentrations of lipids and apolipoproteins were measured, and cardiovascular events recorded. RESULTS: Serum total cholesterol concentrations decreased in both the testosterone-treated men and placebo-treated men, but the 3-year mean (+/- SD) decreases in the two groups (testosterone treated, -17 +/- 29 mg/dL; placebo treated, -12 +/- 38 mg/dL) were not significantly different from each other (P = 0.4). Similarly, serum low-density lipoprotein (LDL) cholesterol levels decreased in both treatment groups, but the decreases in the two groups (testosterone treated, -16 +/- 24 mg/dL; placebo treated, -16 +/- 33 mg/dL) were similar (P = 1.0). Levels of high-density lipoprotein (HDL) cholesterol, triglycerides, and apolipoproteins A-I and B did not change. Lipoprotein(a) levels increased in both groups by similar amounts (testosterone treated, 3 +/- 9 mg/dL; placebo treated, 4 +/- 6 mg/dL; P = 1.0). The number of cardiovascular events was small and did not differ significantly between the testosterone-treated men (9 events) and the placebo-treated men (5 events) during the 3-year study (relative risk = 1.8; 95% confidence interval: 0.7 to 5.0). CONCLUSIONS: As compared with placebo, transdermal testosterone treatment of healthy elderly men for 3 years did not affect any of the lipid or apolipoprotein parameters that we measured. The effect of testosterone treatment on cardiovascular events was unclear, because the number of events was small.

Effect of high-protein feed supplements on concentrations of growth hormone (GH), insulin-like growth factor-I (IGF-I) and IGF-binding protein-3 in plasma and on the amounts of GH and messenger RNA for GH in the pituitary glands of adult rams.

Three groups of mature rams were maintained on diets of hay, hay + 2% lupin or hay + 2% cowpea for 11 weeks. Serial blood samples were taken at 15-min intervals for 12 h for the determination of GH and IGF-I content by radioimmunoassay and for IGF-binding protein-3 (IGFBP-3) levels by Western blotting. The rams were killed after 77 days of supplementary feeding and their pituitary glands analysed for content of GH and GH mRNA. Mean plasma GH and baseline GH levels were significantly (P < 0.01) decreased in the rams fed lupin and cowpea compared with controls fed hay and GH pulse amplitude was significantly (P < 0.001) decreased in the group fed the cowpea diet. The frequency of GH pulses was not significantly altered by either treatment. Plasma concentrations of IGF-I were elevated in rams fed lupin (P < 0.001) or cowpea (P < 0.05). IGFBP-3 levels were not significantly (P > 0.05) altered by either treatment. There were no significant differences in pituitary content of GH mRNA but pituitary content of GH was increased in rams fed lupin (P < 0.05) and cowpea (P = 0.07). In conclusion, a high-protein diet decreases plasma GH levels and increases IGF-I without changing plasma IGFBP-3 levels in rams. Thus ongoing synthesis of GH, as indicated by the mRNA levels, may cause a build up of GH stores in the pituitary gland.

Factors associated with smoking in low-income pregnant women: relationship to birth weight, stressful life events, social support, health behaviors and mental distress.

Since low-income women are at increased risk of having low birth weight infants, factors associated with birth weight among such groups have special relevance. Cigarette-smoking has emerged as an important predictor of low birth weight due to intrauterine growth retardation and pre-term delivery. After confirming the relation of smoking with birth weight, we examined the association of smoking with sociodemographic factors, attitudes towards pregnancy, health behaviors, stressful life events, social support, and symptoms of mental distress in a cohort of 458 Central Harlem women. We found that social support, stress and mental health were associated with smoking behavior but not directly with birth weight. These findings suggest that programs designed to modify health behaviors such as smoking during pregnancy must also take into account such characteristics of the women and their environments which may make behavioral change difficult. Moreover, programs aimed at fostering better health behaviors to improve pregnancy outcome may have to extend beyond the current pregnancy, as indicated by an association between prior adverse pregnancy outcome and smoking in the current pregnancy.

Association of dialyzer reuse with hospitalization and survival rates among U.S. hemodialysis patients: do comorbidities matter?

The objective of this study was to determine whether the associations between reuse of hemodialyzers and higher rates of death and hospitalization persist after adjustment for comorbidity. This was a nonconcurrent cohort study of survival and hospitalization rates among 1491 U.S. chronic hemodialysis patients beginning treatment in 1986 and 1987. The impact of dialyzer reuse was compared across three survival models: an unadjusted model, a "base" model adjusted only for demographics and renal diagnosis, and an "augmented" model additionally adjusted for comorbidities. We found that reuse of dialyzers was associated with a similarly higher rate of death in analyses unadjusted for confounders (relative risk [RR] 1.25, 95% confidence interval [CI] 0.97-1.61), adjusted for demographics and renal diagnosis (RR 1.16, 95% CI 0.96-1.41), and analyses additionally adjusted for comorbidities (RR = 1.25, CI, 1.03, 1.52). Reusing dialyzers was also associated with a greater rate of hospitalization that was stable regardless of adjustment procedures. We conclude that higher rates of death and hospitalization associated with dialyzer reuse persist regardless of adjustment for demographic characteristics or baseline comorbidities. These findings amplify concerns that there exists elevated morbidity among hemodialysis patients treated in facilities that reuse hemodialyzers. Although the association we observed was not confounded by comorbidity, a cause-and-effect relationship between dialyzer reuse and morbidity could not be proved because of the inability to control for aspects of care other than dialyzer reuse.

Injuries in an elderly inner-city population.

Even though injuries are a leading cause of morbidity and mortality among the elderly in the United States, no comprehensive population-based study of nonfatal and fatal injuries has been carried out in an elderly minority inner-city population. To study injuries in this population, we developed an active surveillance system as part of a large injury prevention program in a poor urban black community. We report 577 cases of nonfatal and fatal injuries in a community of 12,139 persons 65 years of age and older that resulted in emergency room treatment or death between March 1, 1987, and February 29, 1988. Nearly 5% of the elderly population was treated at an emergency room for, or died as a result of, an injury during the study period; the overall injury rate was 48 injuries per 1,000 persons. Injury rates for older women exceeded those for older men and increased with advancing age in both sexes. Fall injuries accounted for 312 (54%) of all injuries and 75% of all hospitalizations for injury. Motor vehicle incidents and violence were the second and third most common injuries, accounting for 13% and 7% of injuries, respectively. Given the predominance of falls relative to other injuries, prevention of falls should receive major emphasis in injury prevention efforts in inner-city minority populations.

Risk factors of branch retinal vein occlusion.

A case-control study of 225 patients with branch retinal vein occlusion (BRVO) and 100 age-matched controls was conducted to assess potential clinical risk factors for BRVO. Male gender, hypertension, and hyperopia were significantly more prevalent in patients with BRVO. There was no significant association with race, diabetes, or chronic open-angle glaucoma.

Development and validation of the injury severity assessment survey/parent report: a new injury severity assessment survey.

OBJECTIVE: To develop and pilot test a telephone-based survey instrument that enables parents to identify and characterize the body region and severity of childhood injuries using the Abbreviated Injury Scale (AIS) scoring system. DESIGN: A prospective cross-sectional survey. SETTING: The emergency department of an urban, tertiary care, pediatric trauma center. PARTICIPANTS: One hundred forty-seven parents of children younger than 18 years and seen in the emergency department for acute treatment of an unintentional injury. INTERVENTIONS: None. MAIN OUTCOME MEASURE: The degree of agreement, measured as sensitivity, specificity, and kappa statistic, between medical record information and parents' responses to the telephone survey regarding the identification and characterization of clinically significant (AIS > or =2) injuries. RESULTS: The survey, known as the Injury Severity Assessment Survey/Parent Report, was developed via a systematic review of the AIS 1990 manual. Answers to questions were developed in a way that enabled automated coding of responses into AIS scores or ranges of scores. The sensitivity of the survey (its ability to detect injuries scoring 2 or more on the AIS that were documented in the medical record) varied somewhat by the body region of injury, ranging from 88% for head, face, neck, and spine injuries to 95% for extremity injuries. Intermediate sensitivity (92%) was noted for the detection of significant chest and abdomen injuries. The specificity of the survey (its ability to rule out the presence of a significant injury when one was not documented in the medical record) was more than 95% in each of the 3 body region groups. The kappa statistics for the 3 body region groups ranged from 0.89 to 0.92. CONCLUSIONS: A new telephone-based survey has been developed that enables parents to characterize their child's injuries by body region and to differentiate between minor injuries and more significant injuries using a well-established injury classification system. This survey has a significant advantage over previous telephone-based or written surveys of childhood injuries and may be particularly valuable in population-based (e.g., random-digit dial surveys) or multi-institutional studies of pediatric injuries.

Brain SPECT imaging and neuropsychological testing in coronary artery bypass patients: single photon emission computed tomography.

BACKGROUND: Cognitive deficits appear frequently after cardiac operation. While the etiology remains unclear, alterations in cerebral perfusion during cardiopulmonary bypass may be causative. Single photon emission computed tomography (SPECT) scanning utilizes a radiopharmaceutical to provide images of cerebral perfusion. We proposed to study the cerebral circulation of patients during coronary artery bypass operation employing cardiopulmonary bypass. METHODS: Thirty-five neurologically normal patients underwent preoperative SPECT brain scanning and neuropsychological testing. A second SPECT brain perfusion scan was obtained by administering the radioisotope during cardiopulmonary bypass, with subsequent scanning upon completion of the procedure. Postoperative neuropsychological testing was performed prior to discharge. RESULTS: Fourteen (40%) of patients demonstrated significant neuropsychological decline. Patients who suffered cognitive impairment were no different in demographic, general health, or surgical variables. Patients who demonstrated neuropsychological decline had significantly poorer cerebral perfusion both at baseline and during operation. CONCLUSIONS: Impaired cerebral perfusion at baseline may identify patients at risk for cognitive injury after cardiac operation. Alterations in cerebral perfusion during cardiopulmonary bypass is common, and may be a factor in neuropsychological deficits seen after cardiac operation.

Polycystic kidney disease: earlier diagnosis using ultrasound.

A unique family study is presented, representing 124 members in four generations. Sixty-four patients were studied clinically, and 31 were found to have sonographic evidence of polycystic kidney disease. Findings on intravenous urography were positive in only 15 of these 31. Ultrasound is particularly useful in younger patients in whom urinary symptoms and physical and urographic abnormalities are infrequent. The merits of early diagnosis of this disease in persons at risk are discussed. Ultrasound offers definite advantages over standard urographic and isotopic methods in the early diagnosis of polycystic kidney disease.

Management and long-term follow-up of hepatic cysts.

PURPOSE: To provide an algorithm for the management of hepatic cysts through an analysis of our series over 16 years. METHOD: We reviewed the surgical management and outcome of patients with hepatic cysts between 1984 and 2000 at a single institution. Data were collected by chart review, telephone interview, and follow-up hepatic ultrasonography. RESULTS: Forty-four patients (36 females, 8 males) underwent a total of 46 operations for hepatic cysts (mean size 12.0 +/- 5.2 cm) with a mean follow-up of 5.1 +/- 4.0 years. We treated 28 simple cysts, 4 polycystic liver disease (PCLD), 7 cystadenomas, 2 hydatid cysts, 1 cystadenocarcinoma, 1 endometrioma, and 1 hepatic foregut cyst. Operations included simple drainage, wide unroofing (open and laparoscopic), and hepatic resection. Four patients experienced a symptomatic recurrence after definitive treatment; 3 of these patients had PCLD. Four of the 7 patients with cystadenomas had undergone previous operations that required subsequent definitive resection without a recurrence. CONCLUSIONS: The preoperative distinction between simple cysts and cystadenomas/cystadenocarcinomas can be difficult, yet the management is different. Unroofing is a safe and effective operation for patients with simple cysts. Patients with PCLD frequently have recurrences. Cystadenomas should be completely resected owing to the likelihood of recurrence after partial excision and the risk of eventual cystadenocarcinoma. We present a treatment algorithm for the preoperative evaluation and management of hepatic cysts based on the largest number of patients with the longest follow-up reported to date.

Occupational injuries in a poor inner-city population.

This study aims to characterize occupational injuries in a defined poor inner-city population in terms of demographic features, types, and circumstances of injuries, and medical and financial consequences. It is a case series drawn from a larger population-based injury registry in emergency departments that serve 17 poor census tracts in Philadelphia. Of 335 patients from the study area who had been treated at the emergency departments under study for occupational injuries, 107 could be contacted by telephone 2 to 3 years after their injuries. Interviews sought information on the patients, their employment, their injuries, and the consequences. Respondents were almost all African-American, approximately 50% male, and had a median age of 32. Approximately one third were employed in the health care industry, one fourth in the service sector (including conventional service firms, restaurants, and hotels), and the remainder in construction, retail and wholesale trade, education, transportation, and manufacturing. Major causes of injuries included overexertion, contact with sharp objects, and falls. Major types on injuries included sprain/strains and lacerations. Approximately half the respondents had missed more than 3 days of work, with 15% missing more than 1 month. Almost 40% of respondents reported persistent health problems after their injuries. Only about one quarter had received workers' compensation. We conclude that poor and minority workers are at risk of a wide range of occupational injuries, which may result in considerable lost work time and have serious medical and economic consequences. More, attention to the workplace risks of these relatively marginalized workers and more vigorous preventive interventions are needed.

The learning classifier system: an evolutionary computation approach to knowledge discovery in epidemiologic surveillance.

The learning classifier system (LCS) integrates a rule-based system with reinforcement learning and genetic algorithm-based rule discovery. This investigation reports on the design, implementation, and evaluation of EpiCS, a LCS adapted for knowledge discovery in epidemiologic surveillance. Using data from a large, national child automobile passenger protection program, EpiCS was compared with C4. 5 and logistic regression to evaluate its ability to induce rules from data that could be used to classify cases and to derive estimates of outcome risk, respectively. The rules induced by EpiCS were less parsimonious than those induced by C4.5, but were potentially more useful to investigators in hypothesis generation. Classification performance of C4.5 was superior to that of EpiCS (P<0.05). However, risk estimates derived by EpiCS were significantly more accurate than those derived by logistic regression (P<0.05).