A multi-level log-linear model of childhood leukaemia mortality

Regions, hospitals and health outcomes over time: A multi-level analysis of repeat prevalence among a cohort of health-care workers

BACKGROUND

The relative importance of region, workplace, and individual determinants of health burden is debated.

OBJECTIVE

To model the contribution of hospital characteristics to employee mental and musculoskeletal disorders.

METHODS

We linked employment records of nurses and support services' staff with health records, neighbourhood census, and hospital administrative data. We conducted multi-level logistic regression analyses with three levels: year (I), employee characteristics (II), and hospital characteristics (III).

RESULTS

Northern region hospitals experienced lower disorder prevalences (odds ratios (OR) 0.58, 95% confidence intervals (0.40, 0.82) for mental and 0.56 (0.44, 0.73) for musculoskeletal disorders). Hospitals with yearly workloads of the highest versus lowest quintiles of inpatient days/1000 employee hours (>86.0 vs. <42.6) and surgical cases/1000 employee hours (>10.5 vs. <3.9) had greater odds of mental (1.29 (1.05, 1.57); 1.22 (1.05, 1.42)) and musculoskeletal (1.38 (1.21, 1.58); 1.21 (1.09, 1.34)) disorders.

CONCLUSION

Opportunities exist for reduction in burden with hospital workload reduction. Further exploration of regional effects is needed.

Mortality from non-Hodgkin lymphoma and UV exposure in the European Community

There has been a large, unexplained rise in the incidence of non-Hodgkin's lymphoma (NHL) in many countries. It has been hypothesised that increased exposure to solar ultraviolet (UV) radiation may have been a factor in this increase. The hypothesis that exposure to ultraviolet radiation is a factor in NHL can be tested by examining whether geographical variations in UV and in the incidence of the disease are positively correlated. Previous studies have given mixed results but some of these have failed to take into account confounding by socio-economic factors and the multilevel structure of data derived from several different countries. It was therefore decided to carry out a study using data on NHL mortality for the period 1971-80 for level II administrative units in 9 countries in the European Community. Estimated levels of solar UV and per capita GDP were also derived. Poisson regression models of the relationship between NHL mortality, UV and per capita GDP, taking into account the multilevel structure of the data, were fitted using the MLn package. Simple models that did not adjust for the effects of variations in per capita GDP or account adequately for the structure of the data produced apparent negative or quadratic associations between NHL and UV. However, further models show that there is a highly significant positive association between NHL and per capita GDP. Once this is included in the fixed and random parts of the multilevel model the association between NHL and UV becomes positive although non-significant (p = 0.081) at the conventional 0.05 level. These results underline the need to control for socio-economic factors and to take into account the multilevel structure of the data. Studies using international data that do not do this run the risk of producing misleading results.

Tuberculosis mortality in England and Wales during 1982-1992: its association with poverty, ethnicity and AIDS

This paper seeks to establish the strength of association between contemporary tuberculosis (TB) in England and Wales and several potential aetiological factors. It presents an ecological analysis of standardised annual TB mortality rates for the 403 local authority districts between 1982 and 1992, disaggregated by age and sex. Social, demographic and ethnicity measures from the 1981 and 1991 censuses and standardised annual AIDS-related mortality rates for young men are used to calculate Poisson regression models. A strong association was found between all TB mortality groups and overcrowding at the household level. For women, no other measures improved the explanatory power of the models. In multiple regressions, both poverty and AIDS-related mortality explained additional variation in the model for younger men. The link between ethnicity and tuberculosis notifications was not reflected in this analysis of mortality. For all groups no evidence of a positive relationship with ethnicity was found, once overcrowding had been accounted for. The significance of household as opposed to district level crowding suggests that prolonged contact is required for disease transmission. Regression analysis indicates that it is the overcrowding and poverty among ethnic populations that is significant for their tuberculosis mortality. The fact that the relationship between AIDS and TB is confined to the group most typical of AIDS patients provides evidence that AIDS has little influence on the level of tuberculosis mortality in the wider population. Explanations for the observed relationship include preferential certification, migration for treatment and shortcomings in health care provision.

Multilevel models: applications to health data

This paper presents an introductory account of multilevel models, highlighting the potential benefits that may be gained by the use of these methods. It draws on recent applications in health services research that have appeared in the literature. Methodological advances in these statistical techniques have taken place in the field of education, where empirical studies have mainly been concerned with comparing pupil achievement across different schools by exploring the relationship between individual and institutional factors. Although recent widespread availability of suitable software packages has enabled other disciplines to adopt these methods, to date they have received little attention in the health services research literature (the investigation of effects of geographical areas on health being a possible exception) despite their obvious application in many areas of current interest. Key areas that could benefit greatly from these techniques include the exploration of variations in clinical practice, comparisons of institutional performance and resource allocation.