The bias from heaping on risk estimation: effect of age at diagnosis of hypertension on risk of subsequent cardiovascular comorbidities

TY - JOUR

T1 - The bias from heaping on risk estimation: effect of age at diagnosis of hypertension on risk of subsequent cardiovascular comorbidities

AU - Jaeschke, Lina

AU - Becher, Matthias

AU - Velásquez, Ilais Moreno

AU - Ahrens, Wolfgang

AU - Bächle, Christina

AU - Baurecht, Hansjörg

AU - Fricke, Julia

AU - Greiser, Karin Halina

AU - Günther, Kathrin

AU - Heier, Margit

AU - Karch, André

AU - Kluttig, Alexander

AU - Krist, Lilian

AU - Leitzmann, Michael

AU - Michels, Karin

AU - Mikolajczyk, Rafael

AU - Peters, Annette

AU - Schipf, Sabine

AU - Völzke, Henry

AU - Pischon, Tobias

AU - Becher, Heiko

N1 - Copyright © 2022 Elsevier Inc. All rights reserved.

PY - 2022/10

Y1 - 2022/10

N2 - PURPOSE: To investigate (1) the bias in effect estimation due to heaping or digit preference, (2) the association between age at hypertension diagnosis and risk of cardiovascular comorbidities, and (3) the influence of heaping on risk estimates.METHODS: We performed a simulation study with various scenarios, binary outcome, and normal or lognormal distributed covariables. We calculated mean logistic coefficients under the original and heaped data and their relative deviation. The association of age at hypertension diagnosis and risk of ≥1 cardiovascular comorbidity was investigated using logistic regression among 50,858 participants in the NAKO Gesundheitsstudie (German National Cohort) who reported such diagnosis. We assessed the proportion of heaped observations and to what extent heaping may have influenced risk estimates.RESULTS: Based on the simulation study and assuming 50% of observations in the variable of interest to be heaped, relative bias was <6%. In NAKO, a 5-year younger age at hypertension diagnosis was associated with a 15% increased risk of having ≥1 cardiovascular comorbidity. Observed heaping in age at hypertension diagnosis was 12.6%, and bias of the risk estimate was 0.14%.CONCLUSIONS: Bias in effect estimation due to heaping is low in most common scenarios. Younger age at hypertension diagnosis is associated with a higher risk of cardiovascular comorbidities.

AB - PURPOSE: To investigate (1) the bias in effect estimation due to heaping or digit preference, (2) the association between age at hypertension diagnosis and risk of cardiovascular comorbidities, and (3) the influence of heaping on risk estimates.METHODS: We performed a simulation study with various scenarios, binary outcome, and normal or lognormal distributed covariables. We calculated mean logistic coefficients under the original and heaped data and their relative deviation. The association of age at hypertension diagnosis and risk of ≥1 cardiovascular comorbidity was investigated using logistic regression among 50,858 participants in the NAKO Gesundheitsstudie (German National Cohort) who reported such diagnosis. We assessed the proportion of heaped observations and to what extent heaping may have influenced risk estimates.RESULTS: Based on the simulation study and assuming 50% of observations in the variable of interest to be heaped, relative bias was <6%. In NAKO, a 5-year younger age at hypertension diagnosis was associated with a 15% increased risk of having ≥1 cardiovascular comorbidity. Observed heaping in age at hypertension diagnosis was 12.6%, and bias of the risk estimate was 0.14%.CONCLUSIONS: Bias in effect estimation due to heaping is low in most common scenarios. Younger age at hypertension diagnosis is associated with a higher risk of cardiovascular comorbidities.

U2 - 10.1016/j.annepidem.2022.07.012

DO - 10.1016/j.annepidem.2022.07.012

M3 - SCORING: Journal article

C2 - 35940393

VL - 74

SP - 84

EP - 96

JO - ANN EPIDEMIOL

JF - ANN EPIDEMIOL

SN - 1047-2797

ER -