Validation of the Jarman's method of calculation of hospital standardised mortality ratios / Sunil Kumar Bhat

2011

[Truncated abstract] PURPOSE: The main purpose of this study was to evaluate the validity of Jarman-derived hospital standardised mortality ratios (HSMR) against Linkage-derived cumulative mortality ratios (CMR) in Western Australia. METHODS: Patients receiving inpatient care in Western Australia in 1980, 1985, 1990 and 1995 were identified as four single calendar year cohorts. Deaths reported on the basis of the ICD-9 classification were identified as those occurring during inpatient stay under the Jarman method and linked deaths occurring within 30 days of initial hospital admission under the Linked method. The latter were identified using the Western Australia Data Linkage System, which systematically links, within the State, all available administrative health data with regular updates. The System uses probabilistic matching of patient names and other partial identifiers to assign a project-specific, encrypted, individual identifier to maintain confidentiality. Under the Linked method, the System also allowed for the identification of inter-hospital transfers, thus enabling deaths within 30 days to be ascribed to the hospital of first admission, which was not necessarily the hospital in which a patient died. For both methods, observed numbers of hospital-related deaths were divided by the numbers expected on the basis of age and sex-specific mortality risks across the entire study population.

The resultant ratios, the Hospital Standardised Mortality Ratio (Jarman’s HSMR) and the Cumulative Mortality Ratio (the alternative Linked CMR), were compared across Metropolitan Teaching, other Metropolitan Public, Metropolitan Private and Rural hospital groups so as to observe any differences in relativities in hospital mortality outcomes according to the two methods RESULTS: Just seventy eight disease conditions comprising 15% of all 518 ICD-9 disease conditions in the study cohort caused 80% of the cohort deaths. Nearly two-thirds of the total deaths under both the methods of analysis occurred in association with the Metro Teaching and Metro Public hospital groups, albeit that the exact numbers of deaths in these hospital groups differed between the two methods (p<0.001). For the Metro Teaching group of hospitals Jarman’s HSMR was 0.98 (95% CI 0.96-1.01) compared with the all-hospital average, a result almost identical to the Linked CMR of 0.99 (0.97-1.01). By way of contrast, the largest difference between the two methods was observed in the mortality outcomes for other Metro Public hospitals - HSMR 1.09 (1.04-1.14); CMR 0.81 (0.77-0.85). There were less pronounced but similar differences in mortality outcome results for Metro Private hospitals - HSMR 0.82 (0.78-0.86); CMR 0.72 (0.69-0.76); and Rural hospitals HSMR 1.14 (1.10-1.18); CMR 1.04 (0.99-1.07)...

Thesis (M.P.H.)--University of Western Australia, 2011

http://repository.uwa.edu.au:80/R/-?func=dbin-jump-full&amp;object_id=30453&amp;silo_library=GEN01