Non-redundant association rules between diseases and medications: an automated method for knowledge base construction.

BACKGROUND: The widespread use of electronic health records (EHRs) has generated massive clinical data storage. Association rules mining is a feasible technique to convert this large amount of data into usable knowledge for clinical decision making, research or billing. We present a data driven method to create a knowledge base linking medications to pathological conditions through their therapeutic indications from elements within the EHRs. METHODS: Association rules were created from the data of patients hospitalised between May 2012 and May 2013 in the department of Cardiology at the University Hospital of Strasbourg. Medications were extracted from the medication list, and the pathological conditions were extracted from the discharge summaries using a natural language processing tool. Association rules were generated along with different interestingness measures: chi square, lift, conviction, dependency, novelty and satisfaction. All medication-disease pairs were compared to the Summary of Product Characteristics, which is the gold standard. A score based on the other interestingness measures was created to filter the best rules, and the indices were calculated for the different interestingness measures. RESULTS: After the evaluation against the gold standard, a list of accurate association rules was successfully retrieved. Dependency represents the best recall (0.76). Our score exhibited higher exactness (0.84) and precision (0.27) than all of the others interestingness measures. Further reductions in noise produced by this method must be performed to improve the classification precision. CONCLUSIONS: Association rules mining using the unstructured elements of the EHR is a feasible technique to identify clinically accurate associations between medications and pathological conditions.

Do Age-Related Variations Of Sagittal Alignment Rely On SpinoPelvic Organization? An Observational Study Of 1540 Subjects.

STUDY DESIGN: Descriptive radiographic analysis of a prospective multi-center database. OBJECTIVE: This study aims to provide normative values of spinopelvic parameters and their correlations according to age and pelvic incidence (PI) of subjects without spinal deformity. METHODS: After Institutional Review Board (IRB) approval, 1540 full spine radiographs were analyzed. Subjects were divided into 3 groups of PI: low PI < 45°, intermediate PI 45-60°, high PI > 60°, and then stratified by age (20-34, 35-49, 50-64, > 65 Y.O). Pelvic and spinal parameters were measured. Statistical analysis between parameters was performed using Bayesian inference and correlation. RESULTS: Mean age was 53.5 years (845 females, 695 males, range 20-93 years).In low PI group, lumbar lordosis (LL) decrease was mainly observed in the 2 younger age groups.In medium and high PI groups, loss of lordosis was linear during aging and occurred mainly on the distal arch of lordosis. Moderate PI group had a stable lordosis apex and thoracolumbar inflection point. High PI group had a stable thoracolumbar inflection point and a more distal lordosis apex in elderly subjects.For all subjects, kyphosis and pelvic tilt (PT) increased with age.There was a constant chain of correlation between PI and age groups. Proximal lumbar lordosis (PLL) was correlated with kyphosis and sagittal vertical axis (SVA C7), while the distal lumbar lordosis (DLL) was correlated with PI and PT. CONCLUSION: This study provides a detailed repository of sagittal spinopelvic parameters normative values with detailed analysis of segmental kyphosis and lordosis distribution according to gender, age, and PI.

Comparison of Spinopelvic Configuration and Roussouly Alignment Types Between Pediatric and Adult Populations.

STUDY DESIGN: Retrospective cross-sectional study. OBJECTIVE: The aim was to describe spinopelvic alignment types by pelvic incidence (PI) and age to compare the Roussouly classification between pediatric and adult populations. SUMMARY OF BACKGROUND DATA: The Roussouly classification was validated for adults. Alignment types may vary during growth. MATERIALS AND METHODS: Radiographs of 1706 non pathologic individuals (5-49 yr) were analyzed. Individuals ≤19 years were stratified by chronological age and skeletal maturity (triradiate cartilage, Risser), and compared with adults. Global and spinopelvic alignment parameters were assessed. Roussouly Types 1, 2, 3, 3A (anteverted pelvis), and 4 were determined. The distribution of parameters was analyzed by Bayesian inference. The relationship between PI and age by Roussouly type was modeled by linear regression. RESULTS: The Sagittal Vertical Axis C7 decreased during growth and was significantly smaller in adults (20-34 yr) (Pr>0.99). Thoracic kyphosis and lumbar lordosis increased during growth and were larger in adults (Pr<0.025). Lordosis increased mainly in the cranial arch (Pr<0.025). PI and pelvic tilt increased during growth and were larger in adults (Pr<0.025). In children and adolescents, PI<45° represented the largest proportion, significantly larger compared with adults (Pr>0.99). Proportions of Roussouly Types 1 and 2 were similar throughout ages. Types 3 and 4 were rarer during the prepubertal period (Pr<0.025). The proportion of Type 3A was significantly higher in children and adolescents (Pr>0.99). Linear regression showed that Type 4 had the largest PI increase with age, with significantly higher curve slope compared with other types (Pr>0.9999). Types 3, 3A and 2 had similar slopes and lowest PI increase with age. CONCLUSION: Global and spinopelvic alignment changed during childhood and adolescence, leading to different kyphosis and lordosis distribution compared with adults. Growth-related PI increase influenced Roussouly types with typical predominance of Type 3A in the pediatric population and larger PI increase in Type 4. LEVEL OF EVIDENCE: Level III.

Can Lumbopelvic Parameters Be Used to Predict Thoracic Kyphosis at all Ages? A National Cross-Sectional Study.

STUDY DESIGN: National cross-sectional study. BACKGROUND: Thoracic kyphosis (TK) is related to sagittal parameters as pelvic tilt (PT), lumbar lordosis (LL) and pelvic incidence (PI). The equation TK =2 (PT+LL-PI) was validated for adolescents.Objective: to investigate if this equation correctly predicts TK regardless of age. METHODS: Sagittal alignment parameters were assessed on full spine radiographs of 2599 individuals without spine pathology (1488 females, 1111 males). Calculated TK (CTK) = 2 (PT+LL-PI) and measured TK (MTK) were compared by calculating the gap and using a linear regression between both parameters. Subgroup analyses were performed for gender, age, TK groups (≤20°, 21°-40°, 41°-60°, >60°), and PI groups (<45°, 45°-60°, >60°). RESULTS: Average values in the total population were: MTK 45.0°, CTK 36.9°. Average TK gap was 8.1°, 5.2° in females (intercept 11.7, slope .61) and 11.9° in males (intercept 7.1, slope .58). The mean gap was 3.6° for 15-34 years, 5.7° under 15 years and it increased progressively after 35 years with a maximum of 19.9° over 80 years. The gap also increased with the amount of MTK: -3.5° for TK<20° up to 17.3° for TK >60°. Differences in gaps were minor between PI groups. The intercept was smallest and slopes >.6 for PI <45° and TK ≤20°. CONCLUSION: The formula TK=2 (PT+LL-PI) yielded moderate accuracy for adolescents and young adults, but did not fit for over 35 years and under 15. The amount and variance in TK increased in elderly subjects, which made the formula less accurate.

Variation of Global Sagittal Alignment Parameters According to Gender, Pelvic Incidence, and Age.

STUDY DESIGN: Retrospective cross-sectional study. OBJECTIVE: The aim was to describe existing global sagittal alignment parameters across ages and to analyze differences according to gender and pelvic incidence (PI). SUMMARY OF BACKGROUND DATA: Variability with age has been reported. It remains unclear how gender and spinopelvic morphology could additionally influence global alignment parameters. MATERIALS AND METHODS: Radiographs of 2599 individuals (5-93 y) were analyzed. Translation parameters were: Sagittal Vertical Axis (SVA)-C7, SVA-C2, SVA-Center Acoustic Meatus (CAM), C7/Sacrofemoral Distance (SFD) ratio. Inclination parameters were: C7-Vertical Tilt (VT), T1-VT and T9-VT, Odontoid-Hip Axis (OD-HA), OD-CAM. Pelvic compensation parameters were: T1-Pelvic Angle (TPA), Global Tilt (GT), Spino-Sacral Angle (SSA). Global sagittal alignment (GSA) was considered among formulae. The distribution of parameters was analyzed using a Bayesian inference. Correlations with spinopelvic parameters were investigated. RESULTS: SVA-C7, SVA-C2, SVA-CAM were larger in males and high PI, and increased significantly after 50 years (Pr>0.9999). C7/SFD decreased during growth and was larger in low PI (Pr=0.951). There was no correlation with spinopelvic parameters. Age-related variations of inclination parameters were nonsignificant. T1-VT and T9-VT increased with PI and were significantly larger in high PI (Pr>0.95). C7-VT was significantly larger in low PI (Pr>0.9999). OD-HA and OD-CAM were constant and increased after 80 years. TPA and GT increased with PI (Pr>0.9999) and age after 35 years (Pr>0.9999). SSA decreased nonsignificantly after 50 years. TPA correlated with PI (ρ=0.6130) and pelvic tilt (PT) (ρ=0.8375). GT correlated with PI (ρ=0.5961) and PT (ρ=0.8996). SSA correlated with sacral slope (ρ=0.9026). GSA was larger in high PI (Pr>0.9999) and increased after 35 years (Pr>0.9999). GSA correlated with PT (ρ=0.7732). CONCLUSION: Translation parameters increase with age, more prominently in males and high PI. Variations of inclination parameters are smaller. Pelvic compensation parameters and GSA increase with age and are closely related to PT and spinopelvic morphology. LEVEL OF EVIDENCE: Level III.

Geographic variations of multiple sclerosis in France.

France is located in an area with a medium to high prevalence of multiple sclerosis, where its epidemiology is not well known. We estimated the national and regional prevalence of multiple sclerosis in France on 31 October 2004 and the incidence between 31 October 2003 and 31 October 2004 based on data from the main French health insurance system: the Caisse Nationale d'Assurance Maladie des Travailleurs Salariés. The Caisse Nationale d'Assurance Maladie des Travailleurs Salariés insures 87% of the French population. We analysed geographic variations in the prevalence and incidence of multiple sclerosis in France using the Bayesian approach. On the 31 October 2004, 49 417 people were registered with multiple sclerosis out of the 52 359 912 insured with the Caisse Nationale d'Assurance Maladie des Travailleurs Salariés. Among these, 4497 were new multiple sclerosis cases declared between 31 October 2003 and 31 October 2004. After standardization for age, total multiple sclerosis prevalence in France was 94.7 per 100,000 (94.3-95.1); 130.5 (129.8-131.2) in females and 54.8 (54.4-55.3) in males. The national incidence of multiple sclerosis between 31 October 2003 and 31 October 2004 was 7.5 per 100,000 (7.3-7.6); 10.4 (10.2-10.6) in females and 4.2 (4.0-4.3) in males. The prevalence and incidence of multiple sclerosis were higher in North-Eastern France, but there was no obvious North-South gradient. This study is the first performed among a representative population of France (87%) using the same method throughout. The Bayesian approach, which takes into account spatial heterogeneity among geographical units and spatial autocorrelation, did not confirm the existence of a prevalence gradient but only a higher prevalence of multiple sclerosis in North-Eastern France and a lower prevalence of multiple sclerosis in the Paris area and on the Mediterranean coast.

Multilocus sequence typing of clinical Borreliella afzelii strains: population structure and differential ability to disseminate in humans.

BACKGROUND: Lyme borreliosis in humans results in a range of clinical manifestations, thought to be partly due to differences in the pathogenicity of the infecting strain. This study compared European human clinical strains of Borreliella afzelii (previously named Borrelia afzelii) using multilocus sequence typing (MLST) to determine their spatial distribution across Europe and to establish whether there are associations between B. afzelii genotypes and specific clinical manifestations of Lyme borreliosis. For this purpose, typing was performed on 63 strains, and data on a further 245 strains were accessed from the literature. RESULTS: All 308 strains were categorized into 149 sequence types (STs), 27 of which are described here for the first time. Phylogenetic and goeBURST analyses showed short evolutionary distances between strains. Although the main STs differed among the countries with the largest number of strains of interest (Germany, the Netherlands, France and Slovenia), the B. afzelii clinical strains were less genetically structured than those previously observed in the European tick population. Two STs were found significantly more frequently in strains associated with clinical manifestations involving erythema migrans, whereas another ST was found significantly more frequently in strains associated with disseminated manifestations, especially neuroborreliosis. CONCLUSIONS: The MLST profiles showed low genetic differentiation between B. afzelii strains isolated from patients with Lyme borreliosis in Europe. Also, clinical data analysis suggests the existence of lineages with differential dissemination properties in humans.

Long-term cognitive outcome of Alzheimer's disease and dementia with Lewy bodies: dual disease is worse.

BACKGROUND: Longitudinal studies of dementia with Lewy bodies (DLB) are rare. Clinically, DLB is usually considered to worsen into Alzheimer's disease (AD). The aim of our study was to compare the rate of the cognitive decline in DLB, AD, and the association of the two diseases (AD + DLB). METHODS: Using the Regional Network for Diagnostic Aid and Management of Patients with Cognitive Impairment database, which includes all the patients seen at all memory clinics (medical consultation and day hospitals) in four French regions, and beta regression, we compared the longitudinal the Mini-Mental State Examination scores of 1159 patients with AD (n = 1000), DLB (n = 131) and AD + DLB (association of the two) (n = 28) during follow-up of at least 4 years. RESULTS: The mean follow-up of the patients was 5.88 years. Using beta regression without propensity scores, the comparison of the decline of patients with AD and patients with DLB did not show a significant difference, but the decline of patients with AD + DLB was worse than that of either patients with DLB (P = 0.006) or patients with AD (P < 0.001). Using beta regression weighted by a propensity score, comparison of patients with AD and patients with DLB showed a faster decline for patients with DLB (P < 0.001). The comparison of the decline of patients with AD + DLB with that of patients with DLB (P < 0.001) and patients with AD (P < 0.001) showed that the decline was clearly worse in the patients with dual disease. CONCLUSIONS: Whatever the analysis, the rate of decline is faster in patients with AD + DLB dual disease. The identification of such patients is important to enable clinicians to optimise treatment and care and to better inform and help patients and caregivers.

Medical record linkage in health information systems by approximate string matching and clustering.

BACKGROUND: Multiplication of data sources within heterogeneous healthcare information systems always results in redundant information, split among multiple databases. Our objective is to detect exact and approximate duplicates within identity records, in order to attain a better quality of information and to permit cross-linkage among stand-alone and clustered databases. Furthermore, we need to assist human decision making, by computing a value reflecting identity proximity. METHODS: The proposed method is in three steps. The first step is to standardise and to index elementary identity fields, using blocking variables, in order to speed up information analysis. The second is to match similar pair records, relying on a global similarity value taken from the Porter-Jaro-Winkler algorithm. And the third is to create clusters of coherent related records, using graph drawing, agglomerative clustering methods and partitioning methods. RESULTS: The batch analysis of 300,000 "supposedly" distinct identities isolates 240,000 true unique records, 24,000 duplicates (clusters composed of 2 records) and 3,000 clusters whose size is greater than or equal to 3 records. CONCLUSION: Duplicate-free databases, used in conjunction with relevant indexes and similarity values, allow immediate (i.e. real-time) proximity detection when inserting a new identity.

Resources allocation in healthcare for cancer: a case study using generalised additive mixed models.

Our aim is to develop a method for helping resources re-allocation in healthcare linked to cancer, in order to replan the allocation of providers. Ageing of the population has a considerable impact on the use of health resources because aged people require more specialised medical care due notably to cancer. We propose a method useful to monitor changes of cancer incidence in space and time taking into account two age categories, according to healthcar general organisation. We use generalised additive mixed models with a Poisson response, according to the methodology presented in Wood, Generalised additive models: an introduction with R. Chapman and Hall/CRC, 2006. Besides one-dimensional smooth functions accounting for non-linear effects of covariates, the space-time interaction can be modelled using scale invariant smoothers. Incidence data collected by a general cancer registry between 1992 and 2007 in a specific area of France is studied. Our best model exhibits a strong increase of the incidence of cancer along time and an obvious spatial pattern for people more than 70 years with a higher incidence in the central band of the region. This is a strong argument for re-allocating resources for old people cancer care in this sub-region.

Bayesian semi-parametric ZIP models with space-time interactions: an application to cancer registry data.

We analyse lymphoid leukemia incidence data collected between 1988 and 2002 from the cancer registry of Haut-Rhin, a region in north-east France. For each patient, sex, area of residence, date of birth and date of diagnosis are available. Incidence summaries in the registry are grouped by 3-year periods. A disproportionately large frequency of zeros in the data leads to a lack of fit for Poisson models of relative risk. The aim of our analysis was to model the spatio-temporal variations of the disease taking into account some non-standard requirements, such as count data with many zeros and space-time interactions. For this purpose, we consider a flexible zero-inflated Poisson model for semi-parametric regression which incorporates space-time interactions (modelled by means of varying coefficient model) using an extension of the methodology proposed in Fahrmeir & Osuna (2006, Structured additive regression for overdispersed and zero-inflated count data. Stoc. Models Bus. Ind., 22, 351-369). Inference is carried out from a Bayesian perspective using Markov chain Monte Carlo methods by means of the BayesX software. Our analysis of the geographical distribution of the disease and its evolution in time may be considered as a starting point for further studies.

Age, period and cohort effects in Bayesian smoothing of spatial cancer survival with geoadditive models.

Based on the example of data on breast cancer survival in a specific area in France, this paper describes a Bayesian approach to analysing individual continuous-time spatially referenced survival data. Starting from the well-known Cox model we develop a geoadditive survival model with a baseline effect, spatial effects, age, period and cohort effects as well as the effect of the number of metastases at the time of diagnosis. Furthermore, we also investigate temporal and spatial variations in the effect of the number of metastases. Our approach is particularly useful since we find clear hints for a violation of the proportional hazards assumption and the existence of different spatial patterns for patients with no, one and more than one metastasis. The reliability of our approach is attested by comparison with a parametric model and with several simulated data sets with known risk profiles.

Comparison of measurement strategies for prospective occupational epidemiology.

In the context of a prospective assessment of exposure for epidemiology, our objective is to obtain an optimal group-based design of allocation of a fixed total number of measurements. Such a design has been described by Ashford [Ashford JR. (1958) The design of a long-term sampling programme to measure the hazard associated with an industrial environment. J R Statist Soc A; 121: 331-47]. As this strategy is not operational, we developed three series of strategies: the first based on simplifications of Ashford's strategy; the second based on a pilot study; and the third on an iterative assessment of the group specific standard deviation of exposure. These strategies are compared by simulating a day-to-day individual exposure in several industrial sites and the resulting health effect. Our criteria for comparing strategies are the mean squared error of the estimated exposure in each group weighted by the number of subjects and the mean squared error of the estimated linear regression coefficient in the dose-response relationship. Strategies relying on an iterative approach have been found to perform best whatever the circumstances, nearly as well as Ashford's optimal strategy.