Keith E. Willard

Probabilistic Analysis of Decision Trees Using Monte Carlo Simulation

The authors describe methods for modeling uncertainty in the specification of decision tree probabilities and utilities using Monte Carlo simulation techniques. Exact confidence levels based upon the underlying probabilistic structure are provided. Probabilistic measures of sensitivity are derived in terms of classical information theory. These measures identify which variables are probabilistically important components of the decision. These techniques are illustrated in terms of the clinical problem of anticoagulation versus observation in the setting of deep vein thrombosis during the first trimester of pregnancy. These methods provide the decision analyst with powerful yet simple tools which give quantitative insight into the structure and inherent limitations of decision models arising from specification uncertainty. The tech niques may be applied to complex decision models. Key words: decision analysis; sensitivity analysis; statistical analysis; statistical confidence; information theory. (Med Decis Making 6:85-92, 1986)

EFFECTIVENESS OF LIQUID SOAP VS. CHLORHEXIDINE GLUCONATE FOR THE REMOVAL OF CLOSTRIDIUM DIFFICILE FROM BARE HANDS AND GLOVED HANDS

OBJECTIVE To compare liquid soap versus 4% chlorhexidine gluconate in 4% alcohol for the decontamination of bare or gloved hands inoculated with an epidemic strain of Clostridium difficile. DESIGN C difficile (6.7 log10 colony-forming units [CFU], 47% spores), was seeded onto bare or latex gloved hands of ten volunteers and allowed to dry. Half the volunteers initially washed with soap and half with chlorhexidine, followed by the other agent 1 week later. Cultures were done with Rodac plates at three sites on the hand: finger/thumbtips, the palmar surfaces of the fingers, and the palm. Statistical comparison was by paired Students t test. RESULTS On bare hands, soap and chlorhexidine did not differ in residual bacterial counts on the finger/thumbtips (log10 CFU, 2.0 and 2.1, P = NS) and fingers (log10 CFU, 2.4 and 2.5, P = NS). Counts were too high on bare palms to quantitate. On gloved hands, soap was more effective than chlorhexidine on fingers (log10 CFU 1.3 and 1.7, P < .01) and palms (log10 CFU 1.5 and 2.0, P < .01), but not finger/thumbtips (log10 CFU 1.6 with each, P = NS). Residual C difficile counts were lower on gloved hands than bare hands (P < 0.01 to < 0.0001). CONCLUSIONS The two agents did not differ significantly in residual counts of C difficile on bare hands, but on gloved hands residual counts were lower following soap wash than following chlorhexidine wash. These observations support the use of either soap or chlorhexidine as a handwash for removal of C difficile, but efficacy in the prevention of C difficile transmission must be determined by prospective clinical trials.

Probabilistic sensitivity analysis methods for general decision models

Probabilistic sensitivity analysis has previously been described for the special case of dichotomous decision trees. We now generalize these techniques for a wider range of decision problems. These methods of sensitivity analysis allow the analyst to evaluate the impact of the multivariate uncertainty in the data used in the decision model and to gain insight into the probabilistic contribution of each of the variables to the decision outcome. The techniques are illustrated using Monte Carlo simulation on a trichotomous decision tree. Application of these powerful tools permits the decision analyst to investigate the structure and limitations of more complex decision problems with inherent uncertainties in the data upon which the decisions are based. Probabilistic sensitivity measures can provide guidance into the allocation of resources to resolve uncertainty about critical components of medical decisions.

Probabilistic analysis of decision trees using symbolic algebra.

Uncertainty in medical decision making techniques occurs in the specification of both decision tree probabilities and utilities. Using a computer-based algebraic approach, methods for modeling this uncertainty have been formulated. This analytic procedure allows an exact calculation of the statistical variance at the final decision node using automated symbolic manipulation. Confidence and conditional confidence levels for the preferred decision are derived from gaussian theory, and the mutual information index that identifies probabilistically important tree variables is provided. The computer-based algebraic method is illustrated for a problem previously analyzed by Monte Carlo simulation. This methodology provides the decision analyst with a procedure to evaluate the outcome of specification uncertainty, in many decision problems, without resorting to Monte Carlo analysis. Key words: decision analysis; sensitivity analysis; statistical analysis; symbolic algebra; statistical confidence; information theory. (Med Decis Making 6:93-100, 1986)

GyrA sequence analysis of Staphylococcus aureus and methicillin-resistant S. aureus strains selected, in vitro, for high-level ciprofloxacin resistance

Four methicillin-susceptible Staphylococcus aureus and eight methicillin-resistant S. aureus (MRSA) isolates, all of which were ciprofloxacin susceptible (MIC < 2.0 micrograms/ml) were manipulated, in vitro, to achieve high-level ciprofloxacin resistance by means of up to 14 passages onto media containing increasing concentrations of ciprofloxacin. Resistance to ciprofloxacin at a concentration of at least 512 micrograms/ml was achieved in all 12 isolates tested. This resistance was continually detected during weekly passage on antibiotic-free media for 12 weeks. The parent and daughter cells from four strains had their gyrA sequenced from amino acid (aa) codons 70-100, the region of previous mutations in high level quinolone-resistant S. aureus. Mutations at aa codon 84 were seen in three of four strains, but appeared at varying levels of ciprofloxacin resistance. High-level resistance of S. aureus and MRSA to ciprofloxacin can be developed in vitro using multiple exposures to incremental concentrations of the drug. It is apparently due to multiple mechanisms and, once established, remains stable over time.

Nonparametric probability density estimation: improvements to the histogram for laboratory data

The histogram has long been used in the clinical laboratory for the depiction and manipulation of frequency data. We present recent results of refinements to the usual histogram procedures along with modern alternative methods of estimating frequency distributions, including the kernel and discrete maximum penalized likelihood estimation (DMPLE) approaches. We compared these nonparametric methods on 15 different types of simulated distributions, and on several sets (greater than 1000 subjects/set) of real data, including alanine aminotransferase, aspartate aminotransferase, and lactate dehydrogenase levels. Each frequency curve estimation technique was evaluated by measuring the integrated mean square error between each techniques prediction and the true underlying distribution, using Monte Carlo techniques on sample sets with size 49 and 119. The kernel method was the clear method of choice, both in performance (best in 22/36 cases) and in practical usage.

Activity of twenty-one antimicrobial agents including l-ofloxacin against quinolone-sensitive and -resistant, and methicillin-sensitive and -resistant Staphylococcus aureus.

There is a need to identify alternative agents to vancomycin for the treatment of infections with methicillin-resistant Staphylococcus aureus (MRSA). One candidate is the l isomer of ofloxacin (DR-3355). We tested 520 frozen MRSA isolates, 248 fresh MRSA isolates, and 375 fresh methicillin-susceptible S. aureus (MSSA) isolates from Minnesota, and 600 clinical isolates of S. aureus (150 MRSA and 450 MSSA) from Illinois. Over 90% of the MRSA strains were resistant to 32 micrograms/ml of oxacillin. Of the 520 frozen MRSA, 24% were susceptible to < or = 2 micrograms/ml ofloxacin, and an additional 74% were susceptible to ofloxacin between 8 and 16 micrograms/ml. More than 98% of all strains were susceptible to < or = 16 micrograms/ml ofloxacin or l-ofloxacin. All the quinolones had a bimodal distribution of in vitro activity, but for only ofloxacin and l-ofloxacin was activity confined to a very narrow range.

A general ampC active site oligonucleotide probe for Gram-negative rods

We have developed a 33 mer probe that hybridizes to the serine active site of the chromosomal ampC beta-lactamase gene of Pseudomonas aeruginosa and the Enterobacteriaceae. We tested this probe against a variety of Enterobacteriaceae, and a series of 23 P. aeruginosa by dot-blots and selected Southern blots. This probe is an alternative or supplement to enzyme studies for characterizing the class of a Gram-negative rods beta-lactamase and is a useful tool for studies of Pseudomonas beta-lactamase regulation.