The information value of clinical data.

The study used an information-utility index to compare diagnostic contributions. Researchers analyzed three clinical databases with 1018 total cases. They examined 314 jaundice cases and 21 diseases with over 10 cases each. A Bayesian procedure was applied to calculate diagnostic probabilities. Patient-reported findings were compared to physician observations and lab tests. The team categorized diagnostic usefulness into four tiers based on data source. They calculated percentages of correct classifications for each data type. The analysis included prior probabilities and conditional probabilities for diseases.

Patient observations alone correctly classified 50% of cases in the jaundice database. Doctor observations added 16% more diagnostic accuracy in the same database. Patient age and sex contributed an additional 4% diagnostic value. Prior disease probabilities added another 4% in diagnostic certainty. These percentages varied significantly across different conditions. Five laboratory tests had lower average diagnostic scores than patient data. Sixteen of twenty-two high-value findings were patient-reported observations. The study found that patient data often outperformed physician and lab data in diagnostic utility.

The authors proposed that patient observations provide substantial diagnostic value. They suggested that patient-reported findings should be prioritized in diagnostic systems. The study showed that doctor observations add moderate diagnostic accuracy. The researchers noted that prior probabilities have limited but measurable impact. They emphasized that data source utility varies by disease type. The findings suggest that diagnostic systems should integrate multiple data types. The authors proposed that future databases should include patient-reported outcomes. They recommended further research on optimizing data collection strategies.