A Comparative Study of the Psychiatric Care between Locum Tenens and Staff Physicians in a State Hospltal

Abstract

Artificial neural networks (ANNs) were used to mea sure the quality of care (Q) at two admission units in a state psychiatric hospital, each unit having two treat ment teams, one led by a permanent (PM) staff physi cian, and one led by various locum tenens (LT) physicians. An LT physician's tour of duty (TOD) averaged approx imately 30 days. Over nearly a 2½-year period the four treatment teams received 744 admissions. Our previous research has reported measuring Q using percent accu rate prediction of hospital length-of-stay (LOS), divided by a measure of severity of patient illness. We calculated Q for each treatment team's test set of patients using a trained ANN for each team. All the teams' test sets were run through each of the trained ANNs resulting in a set of four Q values for each ANN. We defined the standard deviation of Qs resulting from a single team's test set run through it own as well as the other three teams' ANNs as representative of the "diversity" of the patients in that test set. We defined the reciprocal of the standard deviation of the Qs resulting from each of the teams' test sets run through a single team's ANN as that team's "ro bustness." The product of "robustness" times "diver sity" was defined as the value (V) of the treatment team. The V of the PM physician-led teams was 1.9 times that of the LT physician-led teams. We normalized V for pa tient entropy (uncertainty) with a metric called the "risk ratio" (RR), derived from Boltzmann's law. This resulted in the V/RR of one PM physician-led team as superior, de spite treating patients with the highest risk. The LT physi cian-led teams, despite having fewer patients afflicted with the more problematic diagnosis of schizophrenia, were handicapped by not having preexisting therapeu tic relationships with their patients, giving both LT teams low robustness. There was no statistically significant dif ference in patient LOS between the teams. Because the greatest change in team composition was due to LT physi cians, we assumed that the differences in V/RR were due to the short (30-day) TOD and not to any skill deficits in the LT physicians. This article explores a new para digm which compares the value of patient care in sepa rate delivery systems despite differences in severity of illness, case-mix, and uncertainty associated with an im perfect therapeutic environment.