Validation of a case definition for osteoporosis disease surveillance.

UNLABELLED: A simple case definition for osteoporosis case diagnosis is feasible based upon administrative health data. This may facilitate implementation of a population-based osteoporosis surveillance program, providing information that could help to inform and guide screening, prevention, and treatment resources. INTRODUCTION: Our aim was to construct and validate a simplified algorithm for osteoporosis case ascertainment from administrative databases that would be suitable for disease surveillance. METHODS: Multiple classification rules were applied to different sets of hospital diagnosis, physician claims diagnosis, and prescription drug variables from Manitoba, Canada. Algorithms were validated against results from a regional bone mineral density testing program that identified bone mineral density (BMD) measurements in 4,015 women age 50 years and older with at least one BMD test between April 1, 2000 and March 31, 2001. RESULTS: Sensitivity as high as 93.3% was achieved with 3 years of data. Specificity ranged from 50.8% to 91.4% overall, and from 81.2% to 99.1% for discriminating osteoporotic from normal BMD. Sensitivity and overall accuracy were generally lower for algorithms based on diagnosis codes alone than for algorithms that included osteoporosis prescriptions. In the subgroup without prior osteoporotic fractures or chronic corticosteroid use, one simple algorithm (one hospital diagnosis, physician claims diagnosis, or osteoporosis prescription within 1 year) gave accuracy measures exceeding 90% for discriminating osteoporosis from normal BMD across a wide range of disease prevalence. CONCLUSIONS: A relatively simple case definition for osteoporosis surveillance based upon administrative health data can achieve an acceptable level of sensitivity, specificity, and accuracy. Performance is enhanced when the case definition includes osteoporosis medication use in the formulation.

Stroke surveillance in Manitoba, Canada: estimates from administrative databases.

This study investigated the use of population-based administrative databases for stroke surveillance. First, a meta-analysis was conducted of four studies, identified via a PubMed search, which estimated the sensitivity and specificity of hospital data for ascertaining cases of stroke when clinical registries or medical charts were the gold standard. Subsequently, case-ascertainment algorithms based on hospital, physician and prescription drug records were developed and applied to Manitoba's administrative data, and prevalence estimates were obtained for fiscal years 1995/96 to 2003/04 by age group, sex, region of residence and income quintile. The meta-analysis results revealed some over-ascertainment of stroke cases from hospital data when the algorithm was based on diagnosis codes for any type of cerebrovascular disease (Mantel-Haenszel Odds-Ratio [OR] - 1.70 [95% confidence interval (CI): 1.53 - 1.88]). Analyses of Manitoba administrative data revealed that while the total number of stroke cases varied substantially across the algorithms, the trend in prevalence was stable regardless of the algorithm adopted.

Population-based data sources for chronic disease surveillance.

This study estimated agreement between population-based administrative and survey data for ascertaining cases of arthritis, asthma, diabetes, heart disease, hypertension and stroke. Chronic disease case definitions that varied by data source, number of years and number of diagnosis or prescription drug codes were constructed from Manitoba's administrative data. These data were linked to the Canadian Community Health Survey. Agreement between the two data sources, estimated by the kappa coefficient, was calculated for each case definition, and differences were tested. Socio-demographic and comorbidity variables associated with agreement were tested using weighted logistic regression. Agreement was strongest for diabetes and hypertension and lowest for arthritis. The case definition elements that contributed to the highest agreement between the two population-based data sources varied across the chronic diseases. Low agreement between administrative and survey data is likely to occur for conditions that are difficult to diagnose, but will be mediated by individual socio-demographic and health status characteristics. Construction of a chronic disease case definition from administrative data should be accompanied by a justification for the choice of each of its elements.

Attention-deficit hyperactivity disorder in Manitoba children: medical diagnosis and psychostimulant treatment rates.

OBJECTIVE: To describe physicians' diagnosis rates for attention-deficit hyperactivity disorder (ADHD) for children in the province of Manitoba and to describe the rate of psychostimulant medication use by these children. METHODS: This descriptive study reviewed the computerized administrative records of physician visits and prescriptions dispensed to examine a population-based, cross-sectional cohort of children diagnosed with ADHD or prescribed stimulant medication, or both. We found 4787 children with a diagnosis of ADHD over a 24-month period or a prescription for stimulant medication over a 12-month period, or both. Rates were calculated by age, sex, region of residence, neighbourhood income level, and physician specialty. RESULTS: Among Manitoba children, 1.52% received a medical diagnosis of ADHD and 0.89% received stimulant medication. Regionally, diagnosis rates for ADHD varied almost 4-fold, and over 8-fold for medications prescribed. Urban areas had higher rates than did rural areas, regardless of physician specialty. Diagnosis and prescription rates varied according to physician specialty, with the highest rates found among pediatricians. An income gradient was evident in rural areas, with rates of ADHD diagnosis and medication prescribed increasing with increased neighbourhood income level. CONCLUSIONS: The pattern of regional variation found in this study suggests that the diagnosis and treatment of ADHD in Manitoba are influenced strongly by the practice styles of local physicians.