Rajib Paul PhD University of North Carolina at Charlotte November 6 2019 Joint work with Ahmed Arif Eric Delmelle Claudio Owusu Gabriela Brissette and Oluwaseun Adeyemi ID: 784760
Download The PPT/PDF document "Estimating Mortality from Pneumoconiosis..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Estimating Mortality from Pneumoconiosis Using Multilevel Spatial Binary Regression
Rajib Paul, PhDUniversity of North Carolina at CharlotteNovember 6, 2019Joint work with Ahmed Arif, Eric Delmelle, Claudio Owusu, Gabriela Brissette, and Oluwaseun Adeyemi
1
Slide2Presenter DisclosureRajib Paul
No relationships to disclose2
Slide3Research SupportThis project was supported by the Federal Office of Rural Health Policy (FORHP), Health Resources and Services Administration (HRSA), U.S. Department of Health and Human Services (HHS) under cooperative agreement # U1CRH30041. The information, conclusions and opinions expressed in this document are those of the authors and no endorsement by FORHP, HRSA, HHS, or the University of Kentucky is intended or should be inferred. ©2019, Rural & Underserved Health Research Center, University of Kentucky.
3
Slide4OutlineMotivationResearch Questions
Methods:DatasetDescriptive AnalysisInferential Statistical Analysis – Regression based approach with spatial covariance modelingResultsMajor FindingsStrengths and LimitationsFinal Remarks4
Slide5MotivationThe United States generates 30% of its electricity needs from coalCoal
mines are mostly located in rural counties and employ local workforceThe prevalence of Coal Workers Pneumoconiosis (CWP) is 9% (Blackley et al., 2016) among those who worked 25 years or more in coal minesIncreased prevalence is observed in central Appalachia (Kentucky, Virginia, West Virginia)5
Slide6PneumoconiosisPneumoconiosis represents a group of respiratory pathologic diagnosis due to inhalation and deposition of substances that damage the lung parenchyma It is a syndrome complex manifesting initially as anthracnosis, an asymptomatic discoloration, and occlusion of the lung which progresses to chronic bronchitis, ultimately culminating into progressive massive fibrosis
6
Slide7CWP and other related PneumoconiosisICD-9 500: CWP, Black Lung DiseaseICD-9 501: Asbestosis
ICD-9 502: SilicosisICD-9 503: Pneumoconiosis due to other inorganic dustICD-9 504: Pneumonopathy due to inhalation of other dustICD-9 505: Pneumoconiosis, unspecified7
Slide8Research QuestionsAggregate Level: Estimate the prevalence of county-level CWP and identify hotspots
Individual Level: Estimate the risk of mortality from Pneumoconiosis and identify factors that are indicative of increased risk for mortalityMultilevel: How individual level and aggregate level risk factors contribute to increased risk of mortality from CWP8
Slide9DatasetTime Period: 2011 – 2014Study Region: IL, IN, KY, OH, PA, VA, and WV
Medicare Limited Dataset (LDS): Administrative claims data that includes a random sample of 5% of the Medicare Fee For Service (FFS) populationDemographic characteristics including county of residence Inpatient claims file that contains ICD-9 diagnosis codes, dates of service, and hospital claims data Carrier file that contains claims data from non-institutional providers such as physicians, physician assistants, and nurse practitioners9
Slide10Study Region: Observed Locations10
Slide11Descriptive Analysis Our descriptive statistics include tables on counts, proportions, quartiles, mean, median, and standard deviationWe include maps of observed counts of county-wise black lung disease and crude prevalence rates
We also map county-wise mortality counts from CWP 11
Slide12Method: Aggregate Level Negative binomial regression on county-wise counts of CWPExplanatory Variables: % Population living below poverty level, Areal proportion within 10miles of mines, % Males, % Whites, % Rural, and No. of underground mines
We include the expected counts as offset variable in our regressionMoran’s I test was used for assessing spatial autocorrelation12
Slide13Offset Variable: Expected CountsThe expected counts were calculated from data for jth county using:
Model Outcome: Standardized Mortality Ratio (SMR) : observed rate/expected rate
13
Slide14Multilevel Logistic RegressionConsidering death as a binary outcome a multilevel logistic regression with county-level spatial random effects was used
Individual level explanatory variables: age, gender, race, disease type (CWP or not), and number of ER visitsCounty level explanatory variables: rural rates and poverty ratesData extraction was done using SAS 9.4 and analyses were done in RStudio/R version 3.6.114
Slide15Spatial Random Effects TermCounty level random effects term requires modeling spatial autocorrelationWe used nonparametric spatial covariance modeling approach based on the latitudes and longitudes of county centroids
A bivariate basis function of the following form is used for spatial smoothing: ||xk – x||2log ||xk – x||Where xk are selected knot locations over the region that control spatial smoothness15
Slide16Spatial Smoothness: Knot Selection16
Slide17Descriptive Statistics17
ER Use among CWP
Slide18Active Coal Mines and Rural Rates18
Slide19CWP: Prevalence Rates and SMR19
Slide20Aggregate Level CWP: Rates20
Slide21Individual Level: Mortality (Binary)21
OR2.50%
97.50%
OR (spatial)
2.5% (spatial)
97.5% (spatial)
(Intercept)
0.001667
0.000684
0.003965
8.00E-04
0
0.0155
Age
1.069471
1.059155
1.080109
1.0703
1.0595
1.0813
gender Male
1.126026
0.829468
1.544001
1.1221
0.815
1.5448
race White
1.245953
0.856595
1.848704
1.2283
0.8295
1.8188
BLD 500
1.317002
1.094909
1.584189
1.2843
1.0486
1.5729
Deviance
2936.8
2711.034
Slide22Multilevel: Mortality with ER22
Slide23Results: InterpretationsIt seems that under adjusted regression, those who have CWP, there is an increased odds of death among those who come to ER (about 2.3%)
It seems that under adjusted regression, those who do not have CWP, there is an increased odds of death among those who come to ER (about 4.8%, almost twice) Among CWP, those who never went to ER had 45% increased odds of death when adjusted for other variablesAmong CWP, those who went to ER once, there odds decreases by 2.4% compared to those who never went to ER23
Slide24Major FindingsThere are 411 (15.8%) cases below age 65, out of which 274 (66.7%) were diagnosed with CWP. Among them 41 (14.9%) died before there 65th birthday
There is an increased amount of ER use among CWPSpatial Cluster: Medicare beneficiaries living in counties in the central Appalachian region had relatively higher rates of health care utilization for CWP compared to other parts of the country24
Slide25LimitationsThis study is ecological, causal effect cannot be establishedOther comorbid conditions have not been considered in the current study
Analysis is based on claims data, other health conditions from electronic medical records are unavailableEmployment status and tenure of employment have strong influence on development of CWP and its severity, that information is unavailable 25
Slide26StrengthsHelps to identify regions with high utilization of Medicare services for PneumoconiosisIdentify hotspots where rates of CWP is higher so that targeted interventions can be planned
Identify which type of Pneumoconiosis has increased risk of mortalityIdentify patterns of ER use and MortalityAn example of secondary data analysis that fused claims data with American Community Survey (ACS) 26
Slide27Final RemarksProvides insights on rural occupational healthIdentifies regions needed attention by HRSA for future planning of black lungs clinic
Provides an assessment of Medicare utilization due to different PneumoconiosisEstimates death risks from CWP and counties with higher poverty rates have increased odds of death from Pneumoconiosis (about 2% increase in odds for 1% increase in poverty rate)27
Slide28Thank You
28
Slide29ReferencesWade WA, Petsonk EL, Young B, et al. Severe occupational pneumoconiosis among West Virginian coal miners: one hundred thirty-eight cases of progressive massive fibrosis compensated between 2000 and 2009. Chest. 2011;139(6):1458-1462.
Blackley DJ, Halldin CN, Wang ML, et al. Small mine size is associated with lung function abnormality and pneumoconiosis among underground coal miners in Kentucky, Virginia and West Virginia. Occupational and Environmental Medicine. 2014;71(10):690.Dos S Antao VC, Petsonk EL, Sokolow LZ, et al. Rapidly progressive coal workers’ pneumoconiosis in the United States: geographic clustering and other factors. Occupational and Environmental Medicine. 2005;62(10):670.Center for Disease Control and Prevention. Pneumoconiosis Prevalence Among Working Coal Miners Examined in Federal Chest Radiograph Surveillance Programs — United States, 1996–2002. Morbidity and Mortality Weekly Report. 2003;52(15):336-340.Getis, A. (2008). A history of the concept of spatial autocorrelation: A geographer's perspective. Geographical Analysis, 40(3), 297-30929