Rapid approximation of confidence intervals for Markov process decision models: applications in decision support systems.

OBJECTIVE: Develop the methodological foundation for interactive use of Markov process decision models by patients and physicians at the bedside. DESIGN: Monte Carlo simulation studies of a decision model comparing two treatments for benign prostatic hypertrophy: watchful waiting (WW) and transurethral prostatectomy (TUR). MEASUREMENTS: The 95% confidence interval (CI) for the mean of the Markov model; the correlation of a linear approximation with the full Markov model; the predictive performance of the approximation; the information index of specific utilities in the model. RESULTS: The 95% CI for the gain in utility with initial TUR was -1.4 to 19.0 quality-adjusted life-months. A multivariate linear model had an excellent fit to the predictions of the Markov model (R2 = 0.966). In an independent data set, the linear model also had a high correlation with the full Markov model (R2 = 0.967); its predictions were unbiased (p = 0.597, paired t-test); and, in 96.4% of simulated cases, its treatment recommendation was the same. CONCLUSION: Using the linear model, it was possible to efficiently compute which health state had the largest contribution to the variance of the decision model. This is the most informative utility value to elicit next. The most informative utility at any point in a sequence changed depending on utilities previously entered into the model. A linear model can be used to approximate the predictions of a Markov process decision model.

Use of meta-analytic results to facilitate shared decision making.

OBJECTIVES: Describe and evaluate an Internet-based approach to patient decision support using mathematical models that predict the probability of successful treatment on the basis of meta-analytic summaries of the mean and standard deviation of symptom response. DESIGN: An Internet-based decision support tool was developed to help patients with benign prostatic hypertrophy (BPH) determine whether they wanted to use alpha blockers. The Internet site incorporates a meta-analytic model of the results of randomized trials of the alpha blocker terazosin. The site describes alternative treatments for BPH and potential adverse effects of alpha blockers. The site then measures patients' current symptoms and desired level of symptom reduction. In response, the site computes and displays the probability of a patient's achieving his objective by means of terazosin or placebo treatment. SETTING: Self-identified BPH patients accessing the site over the Internet. MAIN OUTCOME MEASURES: Patients' perceptions of the usefulness of information. RESULTS: Over a three-month period, 191 patients who were over 50 years of age and who reported that they have BPH used the decision support tool. Respondents had a mean American Urological Association (AUA) score of 18.8 and a desired drop in symptoms of 10.1 AUA points. Patients had a 40 percent chance of achieving treatment goals with terazosin and a 20 percent chance with placebo. Patients found the information useful (93 percent), and most (71 percent) believed this type of information should be discussed before prescribing medications. CONCLUSIONS: Interactive meta-analytic summary models of the effects of pharmacologic treatments can help patients determine whether a treatment offers sufficient benefits to offset its risks.

Incorporating risk attitude into Markov-process decision models: importance for individual decision making.

Most decision models published in the medical literature take a risk-neutral perspective. Under risk neutrality, the utility of a gamble is equivalent to its expected value and the marginal utility of living a given unit of time is the same regardless of when it occurs. Most patients, however, are not risk-neutral. Not only does risk aversion affect decision analyses when tradeoffs between short- and long-term survival are involved, it also affects the interpretation of time-tradeoff measures of health-state utility. The proportional time tradeoff under- or overestimates the disutility of an inferior health state, depending on whether the patient is risk-seeking or risk-averse (it is unbiased if the patient is risk-neutral). The authors review how risk attitude with respect to gambles for survival duration can be incorporated into decision models using the framework of risk-adjusted quality-adjusted life years (RA-QALYs). They present a simple extension of this framework that allows RA-QALYs to be calculated for Markov-process decision models. Using a previously published Markov-process model of surgical vs expectant treatment for benign prostatic hypertrophy (BPH), they show how attitude towards risk affects the expected number of QALYs calculated by the model. In this model, under risk neutrality, surgery was the preferred option. Under mild risk aversion, expectant treatment was the preferred option. Risk attitude is an important aspect of preferences that should be incorporated into decision models where one treatment option has upfront risks of morbidity or mortality.

The effect of search procedures on utility elicitations.

OBJECTIVE: Elicited preferences for health states vary among scaling methods, manners of describing health states, and other features of the elicitation process. The authors examined the effects of changing the search procedure for a subject's utility on mean utility values. METHODS: A randomized controlled trial of two search procedures (titration and "ping-pong") using two otherwise identical computer programs that describe health states related to Gaucher's disease, then measuring subjects' preferences. SETTING: Paid, healthy volunteers recruited from the community through advertisements. RESULTS: The mean time tradeoff (TTO) and standard gamble (SG) utility values for life with severe anemia and splenomegaly and life with chronic bone pain from Gaucher's disease were between 0.10 and 0.15 higher with the titration search procedure than with the ping-pong procedure. Effects of the search procedure were additive with variability due to scaling methods, resulting in mean differences in utility ratings for the same health state of as much as 0.28 among procedures and scaling methods. Effects of search procedures on utility values persisted on repeated testing at week 2 and week 3; there was no evidence of convergence to a single "true" utility value over time. CONCLUSIONS: The procedure used to search for subjects' utility values strongly influences the results of preference-assessment experiments. Effects of search procedures persist on repeated testing. The results suggest that utility values are heavily influenced by, if not created during, the process of elicitation. Thus, utility values elicited using different search procedures may not be directly comparable.

Myocardial infarction and acute cholecystitis: an application of sequence symmetry analysis.

Using a statewide hospital discharge database and a novel epidemiology method, sequence symmetry analysis (Epidemiology. 1996;7:478-84), I examined the relative risk for hospital admission for acute cholecystitis after admission for myocardial infarction. In sequence symmetry analysis, the ratio of the number of subjects in a fixed population who experienced two events in a "causal" vs "noncausal" temporal sequence estimates the incidence rate ratio (IRR). Of 514 patients admitted for both myocardial infarction and acute cholecystitis during a 3-year window period, 295 were admitted for myocardial infarction first and 219 for acute cholecystitis first, yielding a null sequence-adjusted IRR of 1.45 [95% confidence interval (CI) = 1.28-1.64]. A similar analysis for a known relation (myocardial infarction-->congestive heart failure, N = 27,850) showed the expected association [adjusted IRR = 1.92 (95% CI = 1.88-1.95)], whereas an analysis for a relation hypothesized not to be strong (congestive heart failure-->acute cholecystitis, N = 775) showed only a small association [adjusted IRR = 1.16 (95% CI = 1.05-1.28)]. Subgroup analysis revealed time courses that supported each relation as causal. Hospitalization for myocardial infarction may increase the risk for subsequent hospitalization for acute cholecystitis.

Two types of murine helper T cell clone. II. Delayed-type hypersensitivity is mediated by TH1 clones.

We have previously shown that at least two types of Lyt-1+, Lyt-2-, L3T4+ helper T cell clones can be distinguished in vitro by different patterns of lymphokine secretion and by different forms of B cell help. Evidence is presented here to show that one type of helper T cell clone (TH1) causes delayed-type hypersensitivity (DTH) when injected with the appropriate antigen into the footpads of naive mice. The antigen-specific, major histocompatability complex (MHC)-restricted footpad swelling reaction peaked at approximately 24 hr. Footpad swelling was induced by all TH1 clones tested so far, including clones specific for soluble, particulate, or allogeneic antigens. In contrast, local transfer of TH2 cells and antigen did not produce a DTH reaction, even when supplemented with syngeneic spleen accessory cells. Similarly, local transfer of an alloreactive cytotoxic T lymphocyte clone into appropriate recipients did not produce DTH. The requirements for the DTH reaction induced by TH1 cells were investigated further by using TH1 clones with dual specificity for both foreign antigens and M1s antigens. Although these clones responded in vitro to either antigen + syngeneic presenting cells, or M1s disparate spleen cells, they responded in vivo only to antigen + MHC and did not cause footpad swelling in an M1s-disparate mouse in the absence of antigen. Moreover, in vitro preactivation of TH1 or TH2 cells with the lectin concanavalin A was insufficient to induce DTH reactions upon subsequent injection into footpads. From these results, we conclude that the lack of DTH given by TH2 clones in vivo could be due to the inability of the TH2 cells to produce the correct mediators of DTH, or to a lack of stimulation of TH2 clones in the footpad environment.

Method of Medicare reimbursement and the rate of potentially ineffective care of critically ill patients.

CONTEXT: The worst outcome of critical care may not be death itself; rather, the worst may be an extended death process in which a patient's and his or her family's suffering has been prolonged by services that are ultimately impotent. We have previously used potentially ineffective care (PIC) as a proxy measure for this type of care. OBJECTIVE: To determine if PIC is delivered less often to Medicare patients enrolled in health maintenance organizations (HMOs) than those in traditional fee-for-service health plans. PATIENTS: All Medicare patients hospitalized in intensive care units in California during fiscal year 1994. OUTCOME: Potentially ineffective care was defined as the concurrence of in-hospital death or death within 100 days of hospital discharge and resource use (total hospital costs) above the 90th percentile. METHODS: Hospital costs were adjusted for institution-specific cost-to-charge ratios and local wage indices derived from Health Care Financing Administration cost reports. A multivariate regression model adjusted PIC rates for age, sex, race, elective admission to the hospital, Charlson index diseases, the 15 most common diagnosis related groups for death by 100 days, intensive care unit size, and number of residents at the hospital. RESULTS: A total of 3914 (4.8%) of 81 494 patients experienced PIC and used 21.6% of total intensive care unit resources. The occurrence of PIC was less common among HMO members (adjusted odds ratio, 0.75; 95% confidence interval, 0.65-0.87). However, HMO members were not more likely to experience in-hospital death (adjusted odds ratio, 0.99; 95% confidence interval, 0.91-1.07) and only slightly more likely to experience death by 100 days after hospital discharge (adjusted odds ratio, 1.08; 95% confidence interval, 1.01-1.15). CONCLUSIONS: Patients who experience PIC outcomes are not uncommon in the Medicare population, and patients experiencing this outcome consume a disproportionate amount of medical resources. Medicare beneficiaries in HMO practice settings had a lower risk of experiencing PIC outcomes after adjusting for age, sex, diagnosis, comorbid conditions, and characteristics of the treating hospital. This suggests that HMO practices may be better at limiting or avoiding injudicious use of critical care near the end of life.

Amylase release from streptolysin O permeabilized fetal pancreatic acini.

Developmental regulation of Ca(2+)-dependent protein discharge was investigated in fetal and neonatal rat pancreatic acini permeabilized with streptolysin O. When incubated at 37 degrees C in a Ca(2+)-ethylene glycol-bis(beta-aminoethyl ether)-N,N,N',N'-tetraacetic acid/K glutamate buffer, permeabilized day 19 and 20 fetal acini demonstrated Ca(2+)-dependent release of amylase, whereas day 21 (term) fetal acini did not. Ca(2+)-dependent amylase release reappeared in day 1, 2, and 6 neonatal pancreases. ATP depletion completely inhibited Ca(2+)-stimulated amylase release from both day 20 fetal and adult acini. Ca(2+)-dependent amylase discharge from day 20 fetal acini was enhanced by the nonhydrolyzable GTP analogue, guanosine 5'-O-(3-thiotriphosphate) (GTP gamma S), and by the phorbol ester, 12-O-tetradecanoylphorbol 13-acetate (TPA). Ca(2+)-independent GTP gamma S-stimulated amylase release was observed from adult but not from day 20 fetal acini. In contrast to its stimulatory effects in permeabilized adult acini, adenosine 3',5'-cyclic monophosphate (cAMP) alone had little effect on release from permeabilized day 20 fetal acini. Our studies indicate that the fetal pancreas is competent to undergo Ca(2+)-dependent protein secretion but that this secretion is suppressed at birth. Our studies also suggest that the fetal gland is sensitive to modulators of exocytosis active in the adult pancreas, such as GTP gamma S, TPA, and cAMP but responds differently to these agents compared with responses in adult glands.

MRI for detecting silicone breast implant rupture: meta-analysis and implications.

The objective of this study was to assess the accuracy of magnetic resonance imaging (MRI) in detecting silicone breast implant rupture, and to explore implications of the use of MRI for screening and estimating the prevalence of rupture among asymptomatic women. The study consisted of a meta-analysis of published studies with the calculation of sensitivity and specificity as independent parameters, the summary receiver operating characteristic (ROC) curve, and other clinically important values such as positive predictive value (PPV) and negative predictive value. Participants included women with breast implants who underwent both MRI and subsequent implant removal. Eighteen studies, performed primarily in academic settings, with approximately 1,039 women and 2,036 implants met entry criteria. Most studies involved women with symptomatic implants. Assuming that sensitivity and specificity were independent parameters, the summary sensitivity was 78% (95% confidence interval [CI], 71-83) and the summary specificity was 91% (95% CI, 86-94). Using ROC meta-analysis methodology, the odds ratio describing the overall accuracy of the test was 40.1 (range, 18.8-85.4). There was substantial heterogeneity across studies. Studies using breast coils, those with convenience samples, larger studies, and studies of lower report quality tended to report higher levels of accuracy. The quality of study reports was generally poor. Among symptomatic women, PPV was fairly high. Among lower prevalence populations, PPV appeared to be insufficient to warrant use as a screening tool. MRI is moderately accurate in detecting silicone breast implant rupture. However, MRI should remain a confirmatory diagnostic test and should not be used to screen asymptomatic women.

SecondOpinion: interactive Web-based access to a decision model.

In this paper, we describe a computer architecture, which we call SecondOpinion, designed for automated, normative patient decision support over the World Wide Web. SecondOpinion custom tailors the discussion of therapy options for patients by eliciting their preferences for relevant health states via an interactive WWW interface and then integrating those results in a decision model. The SecondOpinion architecture uses a Finite State Machine representation to track the course of a patient's consultation and to choose the next action to take. The consultation has five distinct types of interactions: explanation of health states, assessment of preferences, detection and correction of errors in preference elicitations, and feedback on the implications of preference. A linear "summary model" speeds calculations of predictions from the decision model and makes it possible to dynamically calculate 95% confidence intervals for the marginal utility of each treatment option. Preferences for states are assessed in the order of their variance contribution to the models predictions in an iterative fashion. Only the states required to obtain a 95% Confidence Interval (CI) that excludes zero are assessed. In Monte Carlo simulation studies, the average number of utility assessments required for the 95% CI to exclude zero in an individual was 4.24 (SD = 1.97) out of 8 relevant health states. the SecondOpinion architecture provides an efficient, "discussion-like" experience leading to an individual-specific treatment recommendation. It may be a cost-effective approach to bring decision analytic advice to the bedside.