Comparative evaluation of three continuous speech recognition software packages in the generation of medical reports.

OBJECTIVE: To compare out-of-box performance of three commercially available continuous speech recognition software packages: IBM ViaVoice 98 with General Medicine Vocabulary; Dragon Systems NaturallySpeaking Medical Suite, version 3.0; and L&H Voice Xpress for Medicine, General Medicine Edition, version 1.2. DESIGN: Twelve physicians completed minimal training with each software package and then dictated a medical progress note and discharge summary drawn from actual records. MEASUREMENTS: Errors in recognition of medical vocabulary, medical abbreviations, and general English vocabulary were compared across packages using a rigorous, standardized approach to scoring. RESULTS: The IBM software was found to have the lowest mean error rate for vocabulary recognition (7.0 to 9.1 percent) followed by the L&H software (13.4 to 15.1 percent) and then Dragon software (14.1 to 15.2 percent). The IBM software was found to perform better than both the Dragon and the L&H software in the recognition of general English vocabulary and medical abbreviations. CONCLUSION: This study is one of a few attempts at a robust evaluation of the performance of continuous speech recognition software. Results of this study suggest that with minimal training, the IBM software outperforms the other products in the domain of general medicine; however, results may vary with domain. Additional training is likely to improve the out-of-box performance of all three products. Although the IBM software was found to have the lowest overall error rate, successive generations of speech recognition software are likely to surpass the accuracy rates found in this investigation.

The patient care component: patient-centered horizontal integration in a vertical world.

This paper describes the structure and operational properties of the Patient Care Component, a patient care data system developed by the Indian Health Service to support primary care in a multi-site, decentralized, health care organization. Sharing the same technology base as the Department of Veterans Affairs Distributed Hospital Computer Program, the system requires a minimal level of investment in technology compared to alternative approaches and is in operation at 140 sites. The Indian Health Service and historical aspects of the system are described briefly; the paper focuses on the design objectives for the system and lessons learned from development and several years of operational experience.

A comparison of LISP and MUMPS as implementation languages for knowledge-based systems.

Major components of knowledge-based systems are summarized, along with the programming language features generally useful in their implementation. LISP and MUMPS are briefly described and compared as vehicles for building knowledge-based systems. The paper concludes with suggestions for extensions to MUMPS that might increase its usefulness in artificial intelligence applications without affecting the essential nature of the language.

Centennial paper: Beta dimethylaminoethyl benzhydryl ether hydrochloride (benadryl) in treatment of urticaria by Arthur C. Curtis and Betty Brechner Owens.

1. Beta dimethylaminoethyl benzhydryl ether hydrochloride (Benadryl) is a member of a new group of pharmacologically active antihistamine drugs. 2. Its use in amounts of 50 to 100 mg. given orally one to five times daily to a diverse group of 18 patients with both acute and chronic urticaria is reported. 3. Eleven patients experienced prompt relief of symptoms as long as the drug was taken. Three had definite and real improvement, and many of the wheals which did appear were not pruritic. Four patients were not benefited. 4. The drug has a wide margin of safety, and the only toxic manifestations noted were drowsiness and muscular aching (2 cases) and dizziness, weakness and vertigo (1 case). No cumulative toxic symptoms were noted in patients who ingested the drug as long as six or seven months. All toxic symptoms promptly disappeared when the drug was discontinued. 5. The effect is palliative, and in many patients the urticaria recurred when the administration of the drug was discontinued. 6. Since an antihistamine drug seems to be effective in controlling urticaria, it may be assumed that this is further indirect evidence that histamine is a factor in the production of urticaria.

Evaluation of varicella-zoster immune globulin: protection of immunosuppressed children after household exposure to varicella.

Varicella-zoster immune globulin (VZIG), an immunoglobulin prepared from normal donor plasma selected for high titer of antibody to varicella-zoster virus (VZV), and zoster immune globulin (ZIG), prepared from the plasma of donors convalescing from herpes zoster, were compared in a double-blind, randomized clinical trial to determine their relative efficacy in protecting immunosuppressed children from severe varicella. VZV infection occurred in 49 (60.4%) of 81 recipients of VZIG and in 57 (68.6%) of 83 recipients of ZIG. These rates and the clinical severity of varicella were not significantly different; however, the subclinical infection rate was significantly higher in ZIG recipients (31.3% vs. 16.0%). This difference was accounted for by a subgroup of patients receiving immunosuppressive therapy for nonneoplastic diseases. Doubling the dose of VZIG administered reduced the rate of subclinical infection. These data indicate that VZIG can be used to protect immunosuppressed children from severe chicken pox.

Thermography in breast diagnosis.

The role of two types of thermography in the diagnosis of breast disease was studied in 502 women seen over a two-year period. Thirteen cancers were diagnosed in eleven women. The most significant finding was the large number of equivocal or abnormal thermograms in women with normal breasts of benign disease, while in patients with proven cancer, the thermogram was abnormal in less than half. Clinical diagnosis of breast cancer was not enhanced by either or both types of thermogram. Despite specific criteria, thermographic interpretation was inconsistent except in the thermograms reported as "normal." On the basis of the findings, the authors could not recommend that an abnormal thermogram be used as an indication for mammography, since this would result in an inordinate number of these studies, particularly in young women. The findings suggest that thermography is not a sufficiently precise modality for use in routine breast diagnosis.

Rubella seroepidemiology in Hawaii.

To determine rubella susceptibility levels 10 years after the introduction of the rubella vaccine in Hawaii, a large-scale serosurvey was conducted in conjunction with a campaign to raise the immunity levels of adolescent and adult women. Each woman tested for rubella antibody was asked her age, ethnic group, migration history, number of siblings, vaccination history, and the occupation of the head of the household. In the period from September 1977 through June 1979, serum specimens acceptable for analysis were collected from 3,852 women; 23.8% were susceptible (haemagglutination inhibition antibody titre less than 8) to rubella. A statistical analysis by fitting log-linear models revealed that rubella vaccination history, birthplace, ethnic group, number of siblings and island of residence appear to be factors related to rubella susceptibility. Although caution must be used in comparing this survey with previous surveys, the relatively low rubella susceptibility rate found in this survey may represent a true decrease in rubella susceptibility due to the rubella vaccination programme.

Outbreak of influenza A/USSR/77 at Marquette University.

An outbreak of influenzalike illness occurred at Marquette University, Milwaukee, Wisconsin, in February 1978. Epidemiologic and laboratory data indicate that the outbreak was due to influenza A/USSR/77 (N1H1) virus. A self-administered questionnaire inquiring about the presence and spectrum of illness was given to faculty members, dental students, and residents of an undergraduate dormitory. Age-specific attack rates were 61.5% for persons less than or equal to 22 years of age, 24.7% for those 23 or 24 years of age, and 9.7% for those greater than or equal to 25 years of age. This pattern of age-specific attack rates paralleled the age distribution of persons without hemagglutination-inhibiting antibody to influenza A/USSR/77 virus found in independent serosurveys. Prior swine influenza (H1N1) immunization provided no protection from illness in this outbreak. Past epidemic antibody titers in undergraduates were only weakly associated with clinical illness. The data on disease impact gathered in this study indicate that in a university setting influenza A/USSR/77 virus produced a short-lived outbreak of respiratory illness with a very high attack rate in young adults.

Illness after influenza vaccination reported through a nationwide surveillance system, 1976-1977.

In 1976, the Center for Disease Control coordinated nationwide surveillance for illnesses after influenza vaccination as part of an effort to vaccinate the nation against influenza A/New Jersey/76. For the 48,161,019 persons vaccinated in 1976, a total of 4733 reports of illness were received which included reports of 223 deaths. When Guillain-Barré syndrome was reported in vaccine recipients, an investigation was begun to examine this possible association. Other than the Guillain-Barré syndrome and rare cases of anaphylaxis, no serious illnesses were causally associated with influenza vaccination by this type of surveillance. Widespread underreporting of illness and death in the passive phase of this surveillance system, however, impaired the ability to draw conclusions about reactions to vaccine from the reports of illness received.

The epidemiology of influenza B in a rural setting in 1977.

An epidemic of influenza B began in January, 1977, in two rural parishes in northeastern Louisiana and quickly spread to involve 29% of their population with clinical influenza. This epidemic was investigated using a standard questionnaire and a random sample of 4.2% of the population. The clinical illness was typical of influenza, with predominant fever, cough, malaise and headache. Gastrointestinal symptoms were part of the clinical syndrome. Rhinitis and diarrhea were significantly more common in children aged five years or less. Clinical attack rates increased with larger household size. The youngest age groups had clinical attack rate of 40--55%, but the elderly had very low attack rates. The direct cost of influenza-like illness during the epidemic averaged almost $30.00 per case. Knowledge of the cost of influenza-like illness and age-specific attack rates should be useful in planning future control efforts for influenza B.