STATISTICAL DETECTION OF CANCER CLUSTERS

THE ACCURATE DETECTION OF DISEASE CLUSTERS IS A WELL-ACKNOWLEDGED PROBLEM. FALSE DETECTIONS MAY BE EXPECTEDUNDER REPETITIVE TESTING, BUT THE OPPOSITE SITUATION, THAT OF FAILING TO DETECT CLUSTERS WHEN THEY ACTUALLY EXIST,IS INCREASINGLY RECOGNIZED. BOTH FALSE DETECTIONS AND FAILURES TO DETECT HAVE STRIKING IMPLICATIONS FOR EPIDEMIOLOGICAL STUDIES, BECAUSE SPACE-TIME CLUSTERING IS OFTEN A FIRST STEP IN THE ANALYSIS OF SUSPECTED PUBLIC HEALTH PROBLEMS. THE FAILURE TO FIND EXISTING CLUSTERS MEANS THAT REAL DISEASE CLUSTERS WILL BE IGNORED, BUT FALSE DETECTIONS RESULT IN A LOSS OF PUBLIC CONFIDENCE AND WASTED RESOURCES. CLEARLY, STUDIES ARE REQUIRED TO ACCURATELY DETERMINE THE ABILITIES OF SPACE-TIME CLUSTERING METHODS UNDER DIFFERENT SPATIAL AND TEMPORAL PATTERNS OF DISEASE INCIDENCE. THE OBJECTIVES OF THIS PROJECT ARE (1) TO DETERMINE THE ABILITY OF EXISTING METHODS TO DETECT CANCER CLUSTERS, AND (2) TO DEVELOP NOVEL, SENSITIVE APPROACHES TO THE DETECTION OF DISEASE CLUSTERS IN SITUATIONS WHERE ENVIRONMENTAL AND GEOGRAPHIC CAUSES ARE SUSPECTED. A USER-FRIENDLY, INTERACTIVE, PC PROGRAM CALLED CLUSTER ANALYSIS IN SPACE AND TIME (CAST) WILL BE DEVELOPED IN PHASE I. THE PROGRAM SERVES TWO PURPOSES: (1) CLUSTER DESCRIPTION THROUGH INTERACTIVE MAPPING AND GRAPHICS, AND (2) CLUSTER DETECTION USING SOPHISTICATED STATISTICAL TECHNIQUES. CAST WILL SERVEAS A BASIS FOR THE PHASE II RESEARCH, WHICH WILL DETERMINE THE SENSITIVITY OF EXISTING CLUSTER DETECTION TECHNIQUES ANDDEVELOP NEW CLUSTER DETECTION METHODOLOGIES.