RESEARCHERS USING ADMINISTRATIVE DATA BANKS ENCOUNTER LARGE AMOUNTS OF SUBSTANTIVE INFORMATION, ORGANIZED FOR PARTICULAR OPERATING PURPOSES. THE MANITOBA HEALTH SERVICES COMMISSION (MHSC) DATA BANK IN CANADA WAS DESIGNED PRIMARILY FOR PAYING PHYSICIAN CLAIMS, FOR STUDYING COMPLIANCE WITH MEDICAL STANDARDS, FOR EVALUATING THE EFFICIENCY OF INDIVIDUAL VERSUS GROUP PRACTICES, AND FOR ASCERTAINING THE EFFICACY OF PARTICULAR SURGICAL PROCEDURES. IN ORGANIZING MHSC DATA TO PERMIT BOTH CHECKS ON RELIABILITY AND SUBSTANTIVE ANALYSIS, A NUMBER OF DECISIONS HAD TO BE MADE. SINCE THE FOCUS OF RESEARCH WAS ON TONSIL/ADENOID SURGERY, DIAGNOSES REPORTED WITH EACH MEDICAL CLAIM WERE ORGANIZED ACCORDING TO THEIR BEARING UPON SUCH SURGERY. BASIC PROCESSES IN USING A DATA BANK ORGANIZED FOR ADMINISTRATIVE FUNCTIONS FOR RESEARCH PURPOSES INCLUDE TAKING NEEDED RECORDS OUT OF THE DATA BANK, SORTING THEM ACCORDING TO IDENTIFIERS, AGGREGATING RECORDS ACCORDING TO THE RESEARCHER'S PURPOSES, AND LINKING AND/OR MERGING TWO OR MORE FILES TO ASSEMBLE INFORMATION STORED IN A DIFFERENT PLACE. CHECKS OF VARIOUS TYPES, BOTH MANUAL AND COMPUTERIZED, HELP IN EVALUATING THE RELIABILITY OF RECORDED INFORMATION. A PROBLEM PROFILE FOR THE MHSC DATA BANK, PRESENTED IN TABULAR FORM, IDENTIFIES SPECIFIC PROBLEMS ASSOCIATED WITH THE USE OF A DATA BANK: CODING ERRORS, MISIDENTIFICATION OF INDIVIDUALS, ASSIGNMENT OF SEVERAL NUMBERS TO ONE INDIVIDUAL, NOT COVERING AN INDIVIDUAL FOR AN ENTIRE PERIOD, RELIABILITY OF DIAGNOSIS, OPERATIONALIZATION OF THE EPISODE OF ILLNESS CONCEPT, AND THREATS TO VALIDITY (INSTABILITY, INSTRUMENTATION, AND EXPERIMENTAL MORTALITY). MOST DATA BANKS SHARE ONE OR MORE OF THE CHARACTERISTICS OF THE MHSC DATA BANK, INCLUDING WIDE COVERAGE TO FACILITATE GENERALIZABILITY, A LARGE ENOUGH SAMPLE TO PERMIT SIMULTANEOUS CONTROLS IN DATA ANALYSIS, A LONG ENOUGH TIME SERIES OF DATA TO PERMIT ANALYSIS BEFORE AND AFTER A PARTICULAR INTERVENTION, AND THE POTENTIAL FOR COMBINING FILES. REFERENCES ARE INCLUDED. (DEP)