Tweedie distributions for fitting semicontinuous health care utilization cost data

Kurz, Christoph

doi:10.1186/s12874-017-0445-y

Cited by 64 publications

(53 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, more sophisticated generalized linear models using different error distributions could be examined (Duan et al, 1983;Kurz, 2017).…”

Section: Resultsmentioning

confidence: 99%

Association of Obesity with Health Care Costs: A Mendelian Randomization study

Kurz

Laxy

2019

Preprint

Self Cite

View full text Add to dashboard Cite

Causal effect estimates for the association of obesity with health care costs can be biased by reversed causation and omitted variables. In this study, we use genetic variants as instrumental variables to overcome these limitations. We estimate the effect of body mass index (BMI) and waist-to-hip ratio (WHR) on total health care costs using data from a German observational study. We find that the model using genetic instruments identifies additional annual costs of 189e for a one unit increase in BMI, and additional 1165e for a 0.1 unit increase in WHR. This is more than two times higher than estimates from linear regression without instrumental variables. We found little evidence of a non-linear relationship between BMI or WHR and health care costs. Our results imply that the use of genetic instruments can be a powerful tool for estimating causal effects in health economic evaluation that might be superior to other types of instruments where there is a strong association with a modifiable risk factor.

show abstract

“…However, more sophisticated generalized linear models using different error distributions could be examined (Duan et al, 1983;Kurz, 2017).…”

Section: Resultsmentioning

confidence: 99%

Association of Obesity with Health Care Costs: A Mendelian Randomization study

Kurz

Laxy

2019

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…All three ERG outcomes had a large proportion of zero responses and the b-wave amplitude had a skewed distribution; analyses for these outcomes were performed with generalized linear regression models for the Tweedie distribution and a log link function. 10 All analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC) and reported P values are two-sided.…”

Section: Methodsmentioning

confidence: 99%

The RUSH2A Study: Best-Corrected Visual Acuity, Full-Field Electroretinography Amplitudes, and Full-Field Stimulus Thresholds at Baseline

Birch

Cheng

Duncan

et al. 2020

Trans. Vis. Sci. Tech.

View full text Add to dashboard Cite

Purpose The purpose of this study was to evaluate baseline best corrected visual acuity (BCVA), full-field electroretinography (ERG), full-field stimulus thresholds (FST), and their relationship with baseline demographic and clinical characteristics in the Rate of Progression in Usher syndrome type 2 ( USH2A )-related Retinal Degeneration (RUSH2A) multicenter study. Methods Participants had Usher syndrome type 2 (USH2, N = 80) or autosomal recessive nonsyndromic retinitis pigmentosa (ARRP, N = 47) associated with biallelic variants in the USH2A gene. Associations of demographic and clinical characteristics with BCVA, ERG, and FST were assessed with regression models. Results In comparison to ARRP, USH2 had worse BCVA (median 79 vs. 82 letters; P < 0.001 adjusted for age), lower rod-mediated ERG b-wave amplitudes (median 0.0 vs. 6.6 µV; P < 0.001) and 30 Hz flicker cone-mediated ERG amplitudes (median 1.5 vs. 3.1 µV; P = 0.001), and higher (white, blue, and red) FST thresholds (means [−26, −31, −23 dB] vs. [−39, −45, −28 dB]; P < 0.001 for all stimuli). After adjusting for age, gender, and duration of vision loss, the difference in BCVA between diagnosis groups was attenuated ( P = 0.09). Only diagnosis was associated with rod- and cone-mediated ERG parameters, whereas both genders ( P = 0.04) and duration of visual loss ( P < 0.001) also were associated with FST white stimulus. Conclusions USH2 participants had worse BCVA, ERG, and FST than ARRP participants. FST was strongly associated with duration of disease; it remains to be determined whether it will be a sensitive measure of progression. Translational Relevance Using standardized research protocols in RUSH2A, measures have been identified to monitor disease progression and treatment response and differentiate features of prognostic relevance between USH2 and ARRP participants with USH2A mutations.

show abstract

“…Because the frequency distributions of medical costs and hospitalization days are often zero, and others are continuous, these data are mixed and include compound Poisson and Gamma distributions. Therefore, Tweedie distributions in a generalized linear model (GLM) with log link and robust error variance ware used to analyze the relationships of the number of teeth (categorical variable) to medical costs and hospitalization days (dependent variables) after adjusting for age, sex, BMI and smoking status . Multivariable medical costs ratio and multivariable hospitalization days ratio and each 95% confidence interval (CI) according to the categorical variable of the number of teeth were calculated in order to show how many times the average medical costs and the average hospitalization days in each category are compared with the reference categories using a GLM.…”

Section: Methodsmentioning

confidence: 99%

Associations of number of teeth with medical costs and hospitalization duration in an older Japanese population

Saito

Shimazaki

Nonoyama

et al. 2019

Geriatrics Gerontology Int

View full text Add to dashboard Cite

Aim Many studies have reported close relationships between oral and systemic health. We explored the association of the number of remaining teeth with medical costs and hospitalization duration in people aged 75 and 80 years. Methods Oral health examinations were carried out at dental clinics in 2014. Medical cost and hospitalization duration data for fiscal year 2015 were obtained from the Mie Prefecture health insurer. We analyzed the data of 4700 individuals who met our inclusion criteria: 2745 75‐year‐olds and 1955 80‐year‐olds. The effects of remaining tooth numbers on medical costs and hospitalization days were analyzed using a generalized linear model with log link adjustment for confounders. Results Total medical costs for all diseases were significantly higher in those with 20–27, 10–19 and 1–9 teeth, and in edentulous older individuals, compared with those with 28 teeth. Outpatient medical costs for diabetes were significantly higher in those with 20–27 and 1–9 teeth. Inpatient medical costs for digestive cancers were significantly higher in those with 10–19 and 1–9 teeth, and in edentulous older individuals. Hospitalization for digestive cancer was significantly longer in those with 20–27, 10–19 and 1–9 teeth, and in edentulous older individuals, than in those with 28 teeth. The number of teeth as a continuous variable was significantly inversely associated with medical costs for cerebrovascular disease and digestive cancer, and hospitalization days for digestive cancer. Conclusion Small numbers of teeth were associated with higher medical costs and longer hospital stays for older Japanese. Geriatr Gerontol Int 2019; 19: 335–341.

show abstract

Tweedie distributions for fitting semicontinuous health care utilization cost data

Cited by 64 publications

References 21 publications

Association of Obesity with Health Care Costs: A Mendelian Randomization study

Association of Obesity with Health Care Costs: A Mendelian Randomization study

The RUSH2A Study: Best-Corrected Visual Acuity, Full-Field Electroretinography Amplitudes, and Full-Field Stimulus Thresholds at Baseline

Associations of number of teeth with medical costs and hospitalization duration in an older Japanese population

Contact Info

Product

Resources

About