Member risk adjustment for ambulatory episodes of care.

The purpose of this study was to model health-plan member risk based on member characteristics in order to separate member risk from other utilization determinants for the use of health care services across sites of care. The approach was to build episodes of care (EOCs) by sorting one year of encounter/claims data into Common Treatment Categories (CTCs). These data came from a variety of health plans, both capitated and non-capitated, covering over 2 million lives. The EOCs were characterized by an array of event and intensity measures. Episode-level risk for each of these measures was modeled by regressions based on member demographic and clinical characteristics. The results of this study show that member characteristics explain a substantial amount of event and intensity variation within episodes and that no single performance measure can summarize the care of health plan members. This method for evaluating member risk can be used both to stratify members according to their future risk and potentially to assess provider or health plan performance or to adjust reimbursement for performance or risk selection.

The impact of computer-assisted test interpretation on physician decision making: the case of electrocardiograms.

This research investigated the effect of computer-assisted test interpretation (CATI) on physicians' readings of electrocardiograms (ECGs). The authors used an experimental method based on direct observations of 22 cardiologists, each reading 80 ECGs, for a total of 1,760 (of which 1,745 were used in the study). There were 40 sets of clinically-matched pairs of ECGs, one with CATI and one without. Reading time was observed and interpretation accuracy was measured by criterion-referenced aggregate scoring. To control for potential biases, the findings were subjected to multivariate analyses using ordinary least-squares regressions. The impact of CATI on cardiologists' readings of ECGs is demonstrably beneficial: the main empirical conclusion of this study is that, compared with conventional interpretation, the use of computer-assisted interpretation of ECGs cuts physician time by an average of 28% and significantly improves the concordance of the physician's interpretation with the expert benchmark, without increasing the false-positive rate. Moreover, CATI is the most accurate and saves the most time when the ECGs have many unambiguous diagnoses. Given that computers alone cannot perform the task of cardiovascular diagnosis, and that cardiologists' ECG interpretations are greatly enhanced by ubiquitous CATI technology, it appears that the best approach is one that combines person and machine.

Comorbidity-adjusted complication risk: a new outcome quality measure.

The measurement of inpatient complications his received substantial attention in recent years because mortality rates and other outcome measures often appear unable to discriminate superior from inferior hospital care. Complication measurement holds out the promise of being more sensitive to variations in patient care because complications occur more frequently than do mortalities, and because complications are more direct consequences of the process of care. The authors developed a new measure of complications that seeks to give insight into the patient care given by different hospitals or physicians by using commonly available data. Specifically, this measure is based on a decision-theoretic model that estimates the probability of a complication for combinations of admitting and secondary International Classification of Diseases, 9th Revision, Clinical Modification diagnoses. The measure can be evaluated at the patient level, or aggregated and risk-adjusted for the population of a given care provider (eg, physician or hospital). When applied to a set of patient-level UB- 82/92 data, this measure estimates the risk of complication for any member of a population, controlling for comorbidity, and hence is designated comorbidity-adjusted complication risk (CACR). The authors describe the development of CACR and its testing and validation using data acquired from the states of Pennsylvania, California, and Florida, as well as facility data obtained directly from hospitals. The data set includes 480,000 patients from 50 Pennsylvania hospitals, 300,000 patients from 33 Florida hospitals, 370,000 patients from 35 California hospitals, and 37,000 patients from six validation hospitals. Comorbidity-adjusted complication risk is constructed from widely available data common to most patient cases. Comorbidity-adjusted complication risk can be adjusted for its case mix, but such risk adjustment has much less effect on CACR than on other adverse outcomes such as mortality and morbidity. Comorbidity-adjusted complication risk varies widely across the hospitals in this sample, yet it is stable across time and is correlated with other known quality outcomes, including such accepted "gold standards" as hospital-documented adverse event rates and chart review determinations of complications.

Time trends in peptic ulcer disease and in gastritis and duodenitis. Mortality, utilization, and disability in the United States.

This study compared the results of two analytic methods testing the effects of histamine H2 receptor antagonists on acid-related conditions. We examined the rates of peptic ulcer disease-related and of gastritis- and duodenitis-related mortality, hospitalizations, surgery, physician visits, work-loss, and disability retirements in the United States from 1970 to 1986. First, we performed a nonparametric epidemiologic analysis. For mortality, hospitalizations, and surgeries, age-specific rates continued their historic decline; there was an additional large one-time decline of operations in 1978. Trends were stronger for peptic ulcer than for gastritis and duodenitis. From pooled annual data, rates of physician visits and physician referral declined for peptic ulcer and for gastritis and duodenitis in the post-1977 period (p = 0.0001). Work-loss and other restrictions on normal daily activities also declined for persons with peptic ulcer and with gastritis and duodenitis (p = 0.0001). Second, we fit a parametric model by maximum likelihood to test specific population effects of H2 blockers. The model indicated that people > or = 65 years old had increasing peptic ulcer mortality rates after 1977 (p < 0.001), while people < 65 years old had a deceleration in rates of decline (p < 0.01). Hospitalization rates for peptic ulcer and for gastritis and duodenitis increased in the elderly after 1977 (p < 0.01) and decreased among those < 65 years old. Both age groups experienced similar declining trends of operations for peptic ulcer; these were not significantly different when pre- and post-1977 periods were compared. The rate of disability retirement declined sharply for workers > or = 50 years old (p < 0.01) and for those < 50 years of age (p < 0.001). The inconclusive results of the parametric analysis, plus only partial congruence between parametric and nonparametric analyses, emphasize the difficulty of relating diverse effects over time to a single, new, more effective treatment.

Substituting diagnostic services. New tests only partly replace older ones.

Hospitals' patterns of ancillary service use were examined to determine whether new technologies replace older, more outmoded technologies, and to explore the factors associated with adoption of newer services and abandonment of older services. Annual inpatient use of five pairs of ancillary services was measured for 1978 through 1980 at 63 hospitals in five regions. The diagnostic test pairs consisted of one well-established diagnostic test and one newer service that could largely substitute for the older one and included (1) oral cholecystogram and gallbladder ultrasound; (2) brain scan and computed tomographic head scan; (3) skull roentgenogram and brain scan; (4) bone survey and bone scan; and (5) blood type/cross and type/screen. Use of gallbladder ultrasound increased significantly after its adoption, with small decreases in the use of oral cholecystogram, its paired test. For the other newer tests examined, increased use was not accompanied by significantly decreased use of the paired older service. The strongest predictors of change in patterns of test use were hospital size, number of residencies, occupancy, urban location, and the proportion of specialists on staff. We conclude that diffusion of new diagnostic services occurs gradually and often without concomitant decrease in older, outmoded services; new services generally seem to complement rather than substitute for older ones. Larger hospitals with a greater teaching commitment make a faster transition to the use of new technologies and the abandonment of older ones.