Observed number (%) of areas including all records Acknowledgement: This project was funded by the National Heaith Priorities and Quality Branch of the Commonwealth Department of Health and Aged Care. Distribution of hysterectomy rates Expected number (%) of areas under the null hypothesis 1. SchofieldMJ. HennrikusOJ.RedmanS, Sanson-Fisher RW. Prevalence and characteristics of women who have had a hysterectomy in a community survey. Aust N Z J Obstet Gynaeco/1991; 31: 153. 2. Carlson KJ. Outcomes of hysterectomy. Clin Obstet Gyneco/1997; 40: 939. 3. Reid BA, Aisbett CW, Jones LM, et al. Relative utilisation ratesof hysterectomy and linksto diagnosis.Canberra:Pub- lication Production Unit, Commonwealth Department of Health and Aged Care, 2000. Q 3 (2%) 18 (11%) 118 (73%) 18(11%) 6 (4%) 163 Observed number (%) of areas after removal of records with menstrual disorders codes 92 (56%) 26 (16%) 12 (7%) 163 15 (9%) 18 (11%) To THE EDITOR: Skin and tissue infections by Aeromonas hydrophila predominantly involve the lower extremities and are as- sociated with trauma with exposure to contaminated water or soil.1Aeromonas sep- ticaemia is rare, usually associated with immunosuppression resulting from hepati- tis, cirrhosis, biliary disease, pancreatic dis- ease or malignancy,' and has a fatality rate of between 30% and 70%.1 Recently, we cared for a 73-year-old Abo- riginal woman with a subcapital fractured neck of femur, which occurred when she slipped on the wet floor while stepping out of the bath. During the fall, she had grasped a live electrical wire with her left hand and received an electrical burn on the palm. Her wounds were cleaned, dressed and blood cultures were taken. An Austin- Moore hemiarthroplasty was performed. A few days after surgery, the patient developed fever with leukocytosis, and gradually became comatose. A large collection of pus was noted in her palm wound, which had enlarged and developed extensive necrosis, Cultures from the wound grew A, hydrophile. By contrast, her blood cultures were microbiologically unremarkable. Antibiotic treatment was commenced immediately with a single dose of gen- tamycin, followed by a seven-day course of ciprofloxacin. Within 48 hours of com- mencement of antibiotic therapy she Aeromonas hydrophila infection secondary to an electrical burn Robert A Wilcox,· George K Chln,t Murugasu Segasothy* * Graduate Medical Student; * Associate Professor of Medicine, and corresponding author, NT Clinical School of Flinders University, PO Box 2234, Alice Springs, NT 0871 (m.segasothy@nl.gov,au); tRegistrar, Alice Springs Hospital, NT. 0.025 (0,02%) 4,05 (2,5%) 154.85 (95%) 4,05 (2,5%) 0,025 (0.02%) 163 Extremely high rates High rates Rates within the expected range Low rates Extremely low rates Total records that contained anyone of these four codes were removed from the data, We found geographical areas with hys- terectomy rates higher and lower than expected, as based on the Poisson distribu- tion (Box). There were only a third as many areas with extremely high or low rates after records with the diagnosis of menstrual dis- orders were removed (Box). In contrast, there were more areas with extreme rates once the records with the other three diag- noses were removed in turn from the data, This indicates that the extremely high and low rates were largely driven by the subset of hysterectomies associated with menstrual disorders, Our findings raise the possibility that the management of menstrual disorders in areas with extreme rates differs from that in the areas where the rates are within the expected range. A study of the management of menstrual disorders across geographical areas would answer this question, Our pilot study shows that geographical variations in the rates of surgical procedures could be linked to their associated diagno- sis codes, Our methods could be replicated for other surgical procedures such as chole- cystectomy and prostatectomy, and admis- sion rates for medical conditions such as asthma and congestive cardiac failure, Fur- ther details of our project are available from our final report to the Commonwealth." To THE EDITOR: We conducted a pilot study to determine if atypical rates for a sur- gical procedure could be linked to diagno- sis codes recorded in hospital separation data. We chose hysterectomy because it is a common major operation with a recognised risk of postoperative complications and other adverse consequences.P We looked at whether variations between geographical areas in hysterectomy rates are associated with variations in the diagnosis codes (eg, do areas with higher than expected rates use certain diagnosis codes more often than areas where rates are within the expected range?). Our null hypothesis was that the age-adjusted hysterectomy rates were simi- lar (P> 0,05) across all geographical areas. We obtained separation data for hys- terectomy from acute care hospitals in New South Wales, Victoria and the Australian Capital Territory for the 1996/97 financial year. Records with cancer diagnoses were excluded. The data were age-standardised by the indirect method for the area of resi- dence based on 163 Statistical Local Areas (small ones amalgamated). It was only possible to identify a single indication for the hysterectomy in 30% of records because the remainder had more than one diagnosis code and it could not be assumed that the "principal diagnosis" was the indication for hysterectomy. The four commonest diagnosis codes (present in 93% of records, singly or in combination with other diagnosis codes) were for fibroids, endometriosis, prolapse, and men- strual disorders. We tested whether the number of geographical areas with high or low hysterectomy rates changed after the Beth A Reid,· Christopher W Aisbett,t Lauren M Jones,* George R Palmer,' Michael Mira,' L1selott M Muhlen-Schulte,·· Leslie Reti,tt Rosemary F Roberts** * Professor and Head, School of Health Information Management, Faculty of Health Sciences, The University of Sydney, PO Box 170, Lidcombe, NSW 1825 (b.reid@cchs.usyd,edu,au); tDirector of Laeta Pty Ltd, Coogee, and Honorary Visiting Fellow, Centre for Hospital Management and Information Systems Research, University of New South Wales; § Professor, School of Health Services Management, and Director, Centre for Hospital Management and Information Systems Research, University of New South Wales; *Consultant, Health Information Management; Director, Central Sydney Division of General Practice, Sydney, NSW; ** Lecturer, Department of Behavioural and Community Health Sciences, The University of Sydney, NSW; ttObstetrician and Gynaecologist, Sydney, NSW; H Director, National Centre for Classification in Health, The University of Sydney, NSW. Use of diagnosis codes to understand variations in hysterectomy rates: a pilot study MJA Vol 173 21 August 2000 219