Modeling the effect of hospital charges and quality on choice.

The authors apply a conditional choice model to simulate the results of patient and physician choices of hospitals for a specific surgical procedure in response to improvements in quality or changes in charges. The model includes all zip code areas and relevant hospitals in a large metropolitan area and estimates the impact on admissions at each hospital. It can be used to estimate both the impact of decisions by a given hospital and the potential responses of competitors, as well as the effects of selective contracting with hospitals by certain payors.

Does quality influence choice of hospital?

In recent years, much information has been provided to the public and to physicians about hospital quality measured in terms of patient outcomes. To examine if, before these public data releases, quality influenced the attractiveness of a hospital to referring or admitting physicians and to patients, we estimated the influences of quality, charges, ownership, and distance on the choice of hospitals for patients with seven surgical procedures and five medical diagnoses in hospitals in three geographic areas in California in 1983. Greater distance and public or proprietary ownership consistently reduced the likelihood of selection while medical school affiliation increased the likelihood of selection. For five of seven surgical procedures and two of five medical diagnoses, hospitals with poorer than expected outcomes attracted significantly fewer admissions. The reverse held for two surgical procedures and one medical diagnosis. The results suggest that quality played an important role in choices among hospitals even before explicit data were widely available.

The sensitivity of conditional choice models for hospital care to estimation technique.

It is plausible that distance, quality, and hospital charges all influence which hospital patients (and their referring physicians) choose. Several researchers have estimated conditional choice models that explicitly incorporate the existence of competing hospitals. To be useful for hospital administrators, health planners and insurers, however, estimates must be made for specific types of patients and include entire market areas. Data sets meeting these requirements have many combinations of hospitals and locations with zero patients. This raises computational difficulties with the linear estimation techniques used previously. In this paper, we use data on patients undergoing cardiac catheterization in several market areas to assess alternative estimation techniques. First, we estimate the conditional choice model with the two techniques used previously to transform the non-linear choice model. These involve using as a reference (1) a single hospital, or (2) the geometric mean of all the hospitals in the market. When there are many zeros, these techniques require extensive adjustments to the data which may lead to biased estimators. We then compare these results with maximum likelihood estimates. The latter results are substantively and significantly different from those using traditional techniques. More importantly, the linear estimates are much more sensitive to the proportion of zeros. We thus conclude that maximum likelihood estimates are preferable when there are many zeros.

Neurometrics does not detect 'pure' dyslexics.

Thirty-eight severely dyslexic boys and 38 good readers were evaluated with neurometrics, a diagnostic procedure based on the application of numerical taxonomy to EEG spectra obtained during resting conditions, supplemented by selected evoked potential features. This procedure generates deviance scores for the EEG spectra by comparing each individual's values to those obtained from a normative population and has been reported to discriminate learning disabled children from normal controls (Ahn et al. 1980). In the present study, all subjects, dyslexic and control, passed stringent screening to assure normal intellectual, neurological, sensory and emotional status. The false positive rate obtained in our control group was comparable to that reported earlier. However, none of the deviance scores significantly discriminated dyslexics from controls; most subjects from both groups were classified as normal. Severe dyslexia per se is thus not associated with the specific neurometric abnormalities reported previously in more heterogeneous learning disabled populations.

Changes in human event related potentials with prolonged delta-9-tetrahydrocannabinol (THC) use.

Auditory event related potentials and contingent negative variation (CNV) were recorded in a series of behavioral tasks from 27 hospitalized human volunteers before, early during (1--3 days), late during (10--14 days) and after chronic THC intoxication at two dose levels (70--90 mg/day or 180--210 mg/day in divided doses). Twelve additional hospitalized volunteers served as a control group. The N1 component of the auditory event related potential in all tasks was depressed when tested early during 180--210 mg/day THC administration. No N1 changes were seen for the subjects receiving 70--90 mg/day or the control subjects. Ten to 14 days during the chronic THC perioe CNV was depressed in all tasks except the control task for both doses at the early THC testing. Only at the most difficult task at the higher of the doses was the CNV depressed at the test late in the THC period. The changes seen in the N1 and the CNV may be indicative of different aspects of a common attentional alteration during stimulus processing. The results on our more difficult tasks suggest the alteration may continue in complex stimulus processing tasks in heavy marijuana users.