Imputing missing laboratory results may return erroneous values because they are not missing at random
Regression models incorporating laboratory tests treat unordered tests as missing and are often imputed. Imputation typically assumes that data are “missing at random” (MAR, test's order status is unrelated to its result after accounting for other variables). This study examined the validity of...
Gespeichert in:
Veröffentlicht in: | Journal of clinical epidemiology 2023-02, Vol.154, p.65-74 |
---|---|
Hauptverfasser: | , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
container_end_page | 74 |
---|---|
container_issue | |
container_start_page | 65 |
container_title | Journal of clinical epidemiology |
container_volume | 154 |
creator | van Walraven, Carl McCudden, Christopher Austin, Peter C. |
description | Regression models incorporating laboratory tests treat unordered tests as missing and are often imputed. Imputation typically assumes that data are “missing at random” (MAR, test's order status is unrelated to its result after accounting for other variables). This study examined the validity of this assumption.
We included 14 biochemistry tests. All tests were measured regardless of test order status. Test-stratified multiple linear regression determined the independent association between test result and order status after adjusting for patient age, sex, comorbidities, and patient location. Testing likelihood models were created for all tests using hospital-wide data.
Four hundred thirty-four patients were included (mean age [standard deviation] 60.7 [19.1], 50.5% female). In 9 of 14 tests (64.2%), test results were significantly associated with order status after adjustment. Results were significantly more abnormal when tests were ordered for 6 tests and significantly more normal for 3 tests. Test abnormality increased as testing likelihood decreased.
These data suggest that laboratory data are often not MAR. The direction and extent of differences in missing laboratory test values varies between tests. Overall the abnormality of ordered tests increased as testing likelihood decreased. These results suggest that imputating missing laboratory data may return biased values. |
doi_str_mv | 10.1016/j.jclinepi.2022.12.011 |
format | Article |
fullrecord | <record><control><sourceid>proquest_cross</sourceid><recordid>TN_cdi_proquest_miscellaneous_2755580930</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><els_id>S0895435622003298</els_id><sourcerecordid>2787615815</sourcerecordid><originalsourceid>FETCH-LOGICAL-c343t-83c7f5814b33f0646519435ee8f8c3c332025831bf3d59b06edb8439d92adf6a3</originalsourceid><addsrcrecordid>eNqFkU1v3CAQhlHUqNmm_QsRUi-92AXGYHxrFfUjUqRe0jPCeJxi2WYLONL--7DdJIdechoOzwzPzEvIFWc1Z1x9nurJzX7Fva8FE6Lmomacn5Ed162uZCf4G7JjupNVA1JdkHcpTYzxlrXyLbkAJYUWADsy3iz7Lfv1ni4-pWOdbR-izSEeaMS0zTnRxR7feYsrxRjDimFL9MHOGybao7NbQpr_4IHaiHQN-WWWzTTadQjLe3I-2jnhh6d6SX5__3Z3_bO6_fXj5vrrbeWggVxpcO0oNW96gJGpRkneFX9EPWoHDqDsKjXwfoRBdj1TOPS6gW7ohB1GZeGSfDrN3cfwt-hlU1QczrP9J21EK6XUrANW0I__oVMoGxa7QulW8eIhC6VOlIshpYij2Ue_2HgwnJljEmYyz0mYYxKGC1OSKI1XT-O3fsHhpe359AX4cgKw3OPBYzTJeVwdDj6iy2YI_rU_HgF1JJ4K</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2787615815</pqid></control><display><type>article</type><title>Imputing missing laboratory results may return erroneous values because they are not missing at random</title><source>MEDLINE</source><source>Access via ScienceDirect (Elsevier)</source><source>ProQuest Central UK/Ireland</source><creator>van Walraven, Carl ; McCudden, Christopher ; Austin, Peter C.</creator><creatorcontrib>van Walraven, Carl ; McCudden, Christopher ; Austin, Peter C.</creatorcontrib><description>Regression models incorporating laboratory tests treat unordered tests as missing and are often imputed. Imputation typically assumes that data are “missing at random” (MAR, test's order status is unrelated to its result after accounting for other variables). This study examined the validity of this assumption.
We included 14 biochemistry tests. All tests were measured regardless of test order status. Test-stratified multiple linear regression determined the independent association between test result and order status after adjusting for patient age, sex, comorbidities, and patient location. Testing likelihood models were created for all tests using hospital-wide data.
Four hundred thirty-four patients were included (mean age [standard deviation] 60.7 [19.1], 50.5% female). In 9 of 14 tests (64.2%), test results were significantly associated with order status after adjustment. Results were significantly more abnormal when tests were ordered for 6 tests and significantly more normal for 3 tests. Test abnormality increased as testing likelihood decreased.
These data suggest that laboratory data are often not MAR. The direction and extent of differences in missing laboratory test values varies between tests. Overall the abnormality of ordered tests increased as testing likelihood decreased. These results suggest that imputating missing laboratory data may return biased values.</description><identifier>ISSN: 0895-4356</identifier><identifier>EISSN: 1878-5921</identifier><identifier>DOI: 10.1016/j.jclinepi.2022.12.011</identifier><identifier>PMID: 36528233</identifier><language>eng</language><publisher>United States: Elsevier Inc</publisher><subject>Biochemistry ; Comorbidity ; Data Collection - methods ; Disease ; Emergency medical care ; Epidemiology ; Female ; Generalized estimating equations ; Hospitals ; Humans ; Imputation ; Laboratories ; Laboratory testing ; Laboratory tests ; Linear Models ; Male ; Missing at random ; Missing data ; Multiple linear regression ; Patients ; Phosphatase ; Physicians ; Potassium ; Regression analysis ; Regression models ; Research Design ; Variables</subject><ispartof>Journal of clinical epidemiology, 2023-02, Vol.154, p.65-74</ispartof><rights>2022 Elsevier Inc.</rights><rights>Copyright © 2022 Elsevier Inc. All rights reserved.</rights><rights>2022. Elsevier Inc.</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><cites>FETCH-LOGICAL-c343t-83c7f5814b33f0646519435ee8f8c3c332025831bf3d59b06edb8439d92adf6a3</cites><orcidid>0000-0002-8390-0930</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://www.proquest.com/docview/2787615815?pq-origsite=primo$$EHTML$$P50$$Gproquest$$H</linktohtml><link.rule.ids>315,781,785,3551,27929,27930,46000,64390,64392,64394,72474</link.rule.ids><backlink>$$Uhttps://www.ncbi.nlm.nih.gov/pubmed/36528233$$D View this record in MEDLINE/PubMed$$Hfree_for_read</backlink></links><search><creatorcontrib>van Walraven, Carl</creatorcontrib><creatorcontrib>McCudden, Christopher</creatorcontrib><creatorcontrib>Austin, Peter C.</creatorcontrib><title>Imputing missing laboratory results may return erroneous values because they are not missing at random</title><title>Journal of clinical epidemiology</title><addtitle>J Clin Epidemiol</addtitle><description>Regression models incorporating laboratory tests treat unordered tests as missing and are often imputed. Imputation typically assumes that data are “missing at random” (MAR, test's order status is unrelated to its result after accounting for other variables). This study examined the validity of this assumption.
We included 14 biochemistry tests. All tests were measured regardless of test order status. Test-stratified multiple linear regression determined the independent association between test result and order status after adjusting for patient age, sex, comorbidities, and patient location. Testing likelihood models were created for all tests using hospital-wide data.
Four hundred thirty-four patients were included (mean age [standard deviation] 60.7 [19.1], 50.5% female). In 9 of 14 tests (64.2%), test results were significantly associated with order status after adjustment. Results were significantly more abnormal when tests were ordered for 6 tests and significantly more normal for 3 tests. Test abnormality increased as testing likelihood decreased.
These data suggest that laboratory data are often not MAR. The direction and extent of differences in missing laboratory test values varies between tests. Overall the abnormality of ordered tests increased as testing likelihood decreased. These results suggest that imputating missing laboratory data may return biased values.</description><subject>Biochemistry</subject><subject>Comorbidity</subject><subject>Data Collection - methods</subject><subject>Disease</subject><subject>Emergency medical care</subject><subject>Epidemiology</subject><subject>Female</subject><subject>Generalized estimating equations</subject><subject>Hospitals</subject><subject>Humans</subject><subject>Imputation</subject><subject>Laboratories</subject><subject>Laboratory testing</subject><subject>Laboratory tests</subject><subject>Linear Models</subject><subject>Male</subject><subject>Missing at random</subject><subject>Missing data</subject><subject>Multiple linear regression</subject><subject>Patients</subject><subject>Phosphatase</subject><subject>Physicians</subject><subject>Potassium</subject><subject>Regression analysis</subject><subject>Regression models</subject><subject>Research Design</subject><subject>Variables</subject><issn>0895-4356</issn><issn>1878-5921</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>EIF</sourceid><sourceid>8G5</sourceid><sourceid>ABUWG</sourceid><sourceid>AFKRA</sourceid><sourceid>AZQEC</sourceid><sourceid>BENPR</sourceid><sourceid>CCPQU</sourceid><sourceid>DWQXO</sourceid><sourceid>GNUQQ</sourceid><sourceid>GUQSH</sourceid><sourceid>M2O</sourceid><recordid>eNqFkU1v3CAQhlHUqNmm_QsRUi-92AXGYHxrFfUjUqRe0jPCeJxi2WYLONL--7DdJIdechoOzwzPzEvIFWc1Z1x9nurJzX7Fva8FE6Lmomacn5Ed162uZCf4G7JjupNVA1JdkHcpTYzxlrXyLbkAJYUWADsy3iz7Lfv1ni4-pWOdbR-izSEeaMS0zTnRxR7feYsrxRjDimFL9MHOGybao7NbQpr_4IHaiHQN-WWWzTTadQjLe3I-2jnhh6d6SX5__3Z3_bO6_fXj5vrrbeWggVxpcO0oNW96gJGpRkneFX9EPWoHDqDsKjXwfoRBdj1TOPS6gW7ohB1GZeGSfDrN3cfwt-hlU1QczrP9J21EK6XUrANW0I__oVMoGxa7QulW8eIhC6VOlIshpYij2Ue_2HgwnJljEmYyz0mYYxKGC1OSKI1XT-O3fsHhpe359AX4cgKw3OPBYzTJeVwdDj6iy2YI_rU_HgF1JJ4K</recordid><startdate>202302</startdate><enddate>202302</enddate><creator>van Walraven, Carl</creator><creator>McCudden, Christopher</creator><creator>Austin, Peter C.</creator><general>Elsevier Inc</general><general>Elsevier Limited</general><scope>CGR</scope><scope>CUY</scope><scope>CVF</scope><scope>ECM</scope><scope>EIF</scope><scope>NPM</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>3V.</scope><scope>7QL</scope><scope>7QP</scope><scope>7RV</scope><scope>7T2</scope><scope>7T7</scope><scope>7TK</scope><scope>7U7</scope><scope>7U9</scope><scope>7X7</scope><scope>7XB</scope><scope>88C</scope><scope>88E</scope><scope>8AO</scope><scope>8C1</scope><scope>8FD</scope><scope>8FI</scope><scope>8FJ</scope><scope>8FK</scope><scope>8G5</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>C1K</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>FR3</scope><scope>FYUFA</scope><scope>GHDGH</scope><scope>GNUQQ</scope><scope>GUQSH</scope><scope>H94</scope><scope>K9.</scope><scope>KB0</scope><scope>M0S</scope><scope>M0T</scope><scope>M1P</scope><scope>M2O</scope><scope>M7N</scope><scope>MBDVC</scope><scope>NAPCQ</scope><scope>P64</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>Q9U</scope><scope>7X8</scope><orcidid>https://orcid.org/0000-0002-8390-0930</orcidid></search><sort><creationdate>202302</creationdate><title>Imputing missing laboratory results may return erroneous values because they are not missing at random</title><author>van Walraven, Carl ; McCudden, Christopher ; Austin, Peter C.</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c343t-83c7f5814b33f0646519435ee8f8c3c332025831bf3d59b06edb8439d92adf6a3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Biochemistry</topic><topic>Comorbidity</topic><topic>Data Collection - methods</topic><topic>Disease</topic><topic>Emergency medical care</topic><topic>Epidemiology</topic><topic>Female</topic><topic>Generalized estimating equations</topic><topic>Hospitals</topic><topic>Humans</topic><topic>Imputation</topic><topic>Laboratories</topic><topic>Laboratory testing</topic><topic>Laboratory tests</topic><topic>Linear Models</topic><topic>Male</topic><topic>Missing at random</topic><topic>Missing data</topic><topic>Multiple linear regression</topic><topic>Patients</topic><topic>Phosphatase</topic><topic>Physicians</topic><topic>Potassium</topic><topic>Regression analysis</topic><topic>Regression models</topic><topic>Research Design</topic><topic>Variables</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>van Walraven, Carl</creatorcontrib><creatorcontrib>McCudden, Christopher</creatorcontrib><creatorcontrib>Austin, Peter C.</creatorcontrib><collection>Medline</collection><collection>MEDLINE</collection><collection>MEDLINE (Ovid)</collection><collection>MEDLINE</collection><collection>MEDLINE</collection><collection>PubMed</collection><collection>CrossRef</collection><collection>ProQuest Central (Corporate)</collection><collection>Bacteriology Abstracts (Microbiology B)</collection><collection>Calcium & Calcified Tissue Abstracts</collection><collection>Nursing & Allied Health Database</collection><collection>Health and Safety Science Abstracts (Full archive)</collection><collection>Industrial and Applied Microbiology Abstracts (Microbiology A)</collection><collection>Neurosciences Abstracts</collection><collection>Toxicology Abstracts</collection><collection>Virology and AIDS Abstracts</collection><collection>Health & Medical Collection</collection><collection>ProQuest Central (purchase pre-March 2016)</collection><collection>Healthcare Administration Database (Alumni)</collection><collection>Medical Database (Alumni Edition)</collection><collection>ProQuest Pharma Collection</collection><collection>Public Health Database</collection><collection>Technology Research Database</collection><collection>Hospital Premium Collection</collection><collection>Hospital Premium Collection (Alumni Edition)</collection><collection>ProQuest Central (Alumni) (purchase pre-March 2016)</collection><collection>Research Library (Alumni Edition)</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>Proquest Central</collection><collection>Environmental Sciences and Pollution Management</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>Engineering Research Database</collection><collection>Health Research Premium Collection</collection><collection>Health Research Premium Collection (Alumni)</collection><collection>ProQuest Central Student</collection><collection>Research Library Prep</collection><collection>AIDS and Cancer Research Abstracts</collection><collection>ProQuest Health & Medical Complete (Alumni)</collection><collection>Nursing & Allied Health Database (Alumni Edition)</collection><collection>Health & Medical Collection (Alumni Edition)</collection><collection>Healthcare Administration Database</collection><collection>Medical Database</collection><collection>Research Library</collection><collection>Algology Mycology and Protozoology Abstracts (Microbiology C)</collection><collection>Research Library (Corporate)</collection><collection>Nursing & Allied Health Premium</collection><collection>Biotechnology and BioEngineering Abstracts</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central Basic</collection><collection>MEDLINE - Academic</collection><jtitle>Journal of clinical epidemiology</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>van Walraven, Carl</au><au>McCudden, Christopher</au><au>Austin, Peter C.</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Imputing missing laboratory results may return erroneous values because they are not missing at random</atitle><jtitle>Journal of clinical epidemiology</jtitle><addtitle>J Clin Epidemiol</addtitle><date>2023-02</date><risdate>2023</risdate><volume>154</volume><spage>65</spage><epage>74</epage><pages>65-74</pages><issn>0895-4356</issn><eissn>1878-5921</eissn><abstract>Regression models incorporating laboratory tests treat unordered tests as missing and are often imputed. Imputation typically assumes that data are “missing at random” (MAR, test's order status is unrelated to its result after accounting for other variables). This study examined the validity of this assumption.
We included 14 biochemistry tests. All tests were measured regardless of test order status. Test-stratified multiple linear regression determined the independent association between test result and order status after adjusting for patient age, sex, comorbidities, and patient location. Testing likelihood models were created for all tests using hospital-wide data.
Four hundred thirty-four patients were included (mean age [standard deviation] 60.7 [19.1], 50.5% female). In 9 of 14 tests (64.2%), test results were significantly associated with order status after adjustment. Results were significantly more abnormal when tests were ordered for 6 tests and significantly more normal for 3 tests. Test abnormality increased as testing likelihood decreased.
These data suggest that laboratory data are often not MAR. The direction and extent of differences in missing laboratory test values varies between tests. Overall the abnormality of ordered tests increased as testing likelihood decreased. These results suggest that imputating missing laboratory data may return biased values.</abstract><cop>United States</cop><pub>Elsevier Inc</pub><pmid>36528233</pmid><doi>10.1016/j.jclinepi.2022.12.011</doi><tpages>10</tpages><orcidid>https://orcid.org/0000-0002-8390-0930</orcidid></addata></record> |
fulltext | fulltext |
identifier | ISSN: 0895-4356 |
ispartof | Journal of clinical epidemiology, 2023-02, Vol.154, p.65-74 |
issn | 0895-4356 1878-5921 |
language | eng |
recordid | cdi_proquest_miscellaneous_2755580930 |
source | MEDLINE; Access via ScienceDirect (Elsevier); ProQuest Central UK/Ireland |
subjects | Biochemistry Comorbidity Data Collection - methods Disease Emergency medical care Epidemiology Female Generalized estimating equations Hospitals Humans Imputation Laboratories Laboratory testing Laboratory tests Linear Models Male Missing at random Missing data Multiple linear regression Patients Phosphatase Physicians Potassium Regression analysis Regression models Research Design Variables |
title | Imputing missing laboratory results may return erroneous values because they are not missing at random |
url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-14T17%3A20%3A58IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Imputing%20missing%20laboratory%20results%20may%20return%20erroneous%20values%20because%20they%20are%20not%20missing%20at%20random&rft.jtitle=Journal%20of%20clinical%20epidemiology&rft.au=van%20Walraven,%20Carl&rft.date=2023-02&rft.volume=154&rft.spage=65&rft.epage=74&rft.pages=65-74&rft.issn=0895-4356&rft.eissn=1878-5921&rft_id=info:doi/10.1016/j.jclinepi.2022.12.011&rft_dat=%3Cproquest_cross%3E2787615815%3C/proquest_cross%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2787615815&rft_id=info:pmid/36528233&rft_els_id=S0895435622003298&rfr_iscdi=true |