| Sub-title | |
| Authors | Isabel Stabile Ladislav J Kohout |
| Abstract | This paper describes the use of a novel statistical software system with data mining capabilities for analysis of medical data that is incomplete. The system is utilised for exploratory statistical analysis and generates interesting hypotheses that can be evaluated for clinical relevance by clinicians. Women with ectopic pregnancy may present with a variety of symptoms eg, amenorrhea, abdominal pain and vaginal bleeding. Unfortunately this combination of features is common to other gynaecological conditions. The choice of which additional test to use in the clinically stable women i.e., ultrasound scanning or biochemical tests, and at which cut off level, is often difficult and is dependent on the facilities available in the given institution. The aim of this work is to evaluate different cut off levels of various biochemical (hCG, SP1, PAPP-A) and ultrasonic parameters in varying combinations in order to improve their diagnostic efficacy in this life threatening condition. We hypothesised that the use of more than one test will provide increased sensitivity, and questioned which tests are relevant, and how many tests are essential. Three different tests are sufficient to capture all the relevant information in our sample of clinical data as well as all the logical relationships (including non symmetrical ones) between the different types of tests used in the study. No single test is sufficient to detect the condition in the whole affected population. Thus at least one more test must be added to increase diagnostic sensitivity and specificity to 100%. This is a minimum requirement for the diagnosis of a condition like ectopic pregnancy that has significant associated morbidity and mortality. It is important to look at the interrelationships and dependencies of several tests using GUHA software to process patient/clinical data. Typical results that will be discussed in greater detail in the presentation are as follows. If hCG is positive, SP1 or PAPP-A will also be positive. If PAPP-A is positive then SP1 or hCG are positive. If hCG and SP1 are positive, then PAPP-A will be positive. If hCG and PAPP-A are negative, then SP1 will be negative. In conclusion, this statistical tool has the important advantage over traditional statistical methods that it can be used with incomplete data, and that directional dependencies as well as those that are symmetrical can be analysed. |
Published in: | |
| Journal | Malta Medical Journal |
| Volume | 15 Issue 1-2/suppl. 2003 |
| Pages | - |
| Date | |
| Link to journal | |
| Key words | GUHA, hypotheses, ectopic pregnancy, statistics, symptom |