Accounting for misclassification in electronic health records-derived exposures using generalized linear finite mixture models

Informed consent for national registration of COVID-19 vaccination caused information bias of vaccine effectiveness estimates mostly in older adults: a bias correction study

OBJECTIVES

Registration in the Dutch national COVID-19 vaccination register requires consent from the vaccinee. This causes misclassification of nonconsenting vaccinated persons as being unvaccinated. We quantified and corrected the resulting information bias in vaccine effectiveness (VE) estimates.

STUDY DESIGN AND SETTING

National data were used for the period dominated by the SARS-CoV-2 Delta variant (July 11 to November 15, 2021). VE ((1-relative risk)∗100%) against COVID-19 hospitalization and intensive care unit (ICU) admission was estimated for individuals 12 to 49, 50 to 69, and ≥70 years of age using negative binomial regression. Anonymous data on vaccinations administered by the Municipal Health Services were used to determine informed consent percentages and estimate corrected VEs by iteratively imputing corrected vaccination status. Absolute bias was calculated as the absolute change in VE; relative bias as uncorrected/corrected relative risk.

RESULTS

A total of 8804 COVID-19 hospitalizations and 1692 COVID-19 ICU admissions were observed. The bias was largest in the 70+ age group where the nonconsent proportion was 7.0% and observed vaccination coverage was 87%: VE of primary vaccination against hospitalization changed from 75.5% (95% CI 73.5-77.4) before to 85.9% (95% CI 84.7-87.1) after correction (absolute bias -10.4 percentage point, relative bias 1.74). VE against ICU admission in this group was 88.7% (95% CI 86.2-90.8) before and 93.7% (95% CI 92.2-94.9) after correction (absolute bias -5.0 percentage point, relative bias 1.79).

CONCLUSION

VE estimates can be substantially biased with modest nonconsent percentages for vaccination data registration. Data on covariate-specific nonconsent percentages should be available to correct this bias.

What health inequalities exist in access to, outcomes from and experience of treatment for lung cancer? A scoping review

OBJECTIVES

Lung cancer (LC) continues to be the leading cause of cancer-related deaths and while there have been significant improvements in overall survival, this gain is not equally distributed. To address health inequalities (HIs), it is vital to identify whether and where they exist. This paper reviews existing literature on what HIs impact LC care and where these manifest on the care pathway.

DESIGN

A systematic scoping review based on Arksey and O'Malley's five-stage framework.

DATA SOURCES

Multiple databases (EMBASE, HMIC, Medline, PsycINFO, PubMed) were used to retrieve articles.

ELIGIBILITY CRITERIA

Search limits were set to retrieve articles published between January 2012 and April 2022. Papers examining LC along with domains of HI were included. Two authors screened papers and independently assessed full texts.

DATA EXTRACTION AND SYNTHESIS

HIs were categorised according to: (a) HI domains: Protected Characteristics (PC); Socioeconomic and Deprivation Factors (SDF); Geographical Region (GR); Vulnerable or Socially Excluded Groups (VSG); and (b) where on the LC pathway (access to, outcomes from, experience of care) inequalities manifest. Data were extracted by two authors and collated in a spreadsheet for structured analysis and interpretation.

RESULTS

41 papers were included. The most studied domain was PC (32/41), followed by SDF (19/41), GR (18/41) and VSG (13/41). Most studies investigated differences in access (31/41) or outcomes (27/41), with few (4/41) exploring experience inequalities. Evidence showed race, rural residence and being part of a VSG impacted the access to LC diagnosis, treatment and supportive care. Additionally, rural residence, older age or male sex negatively impacted survival and mortality. The relationship between outcomes and other factors (eg, race, deprivation) showed mixed results.

CONCLUSIONS

Findings offer an opportunity to reflect on the understanding of HIs in LC care and provide a platform to consider targeted efforts to improve equity of access, outcomes and experience for patients.

Unsupervised probabilistic models for sequential Electronic Health Records

We develop an unsupervised probabilistic model for heterogeneous Electronic Health Record (EHR) data. Utilizing a mixture model formulation, our approach directly models sequences of arbitrary length, such as medications and laboratory results. This allows for subgrouping and incorporation of the dynamics underlying heterogeneous data types. The model consists of a layered set of latent variables that encode underlying structure in the data. These variables represent subject subgroups at the top layer, and unobserved states for sequences in the second layer. We train this model on episodic data from subjects receiving medical care in the Kaiser Permanente Northern California integrated healthcare delivery system. The resulting properties of the trained model generate novel insight from these complex and multifaceted data. In addition, we show how the model can be used to analyze sequences that contribute to assessment of mortality likelihood.

A Bayesian approach to simultaneous adjustment of misclassification and missingness in categorical covariates

This study considers concurrent adjustment of misclassification and missingness in categorical covariates in regression models. Under various misclassification and missingness mechanisms, we derive a general mixture regression structure for regression models that can incorporate multiple surrogates of categorical covariates that are subject to misclassification and missingness. In simulation studies, we demonstrate that including observations with missingness and/or multiple surrogates of the covariate helps alleviate the efficiency loss caused by misclassification. In addition, we study the efficacy of misclassification adjustment when the number of categories increases for the covariate of interest. Using data from the Longitudinal Studies of HIV-Associated Lung Infections and Complications, we perform simultaneous adjustment of misclassification and missingness in the self-reported cocaine and heroin use variable when assessing its association with lung density measures.

ANALYSES OF PREVENTIVE CARE MEASURES WITH INCOMPLETE HISTORICAL DATA IN ELECTRONIC MEDICAL RECORDS: AN EXAMPLE FROM COLORECTAL CANCER SCREENING

The calculation of quality of care measures based on electronic medical records (EMRs) may be inaccurate because of incomplete capture of past services. We evaluate the influence of different statistical approaches for calculating the proportion of patients who are up-to-date for a preventive service, using the example of colorectal cancer (CRC) screening. We propose an extension of traditional mixture models to account for the uncertainty in compliance, which is further complicated by the choice of various screening modalities with different recommended screening intervals. We conducted simulation studies to compare various statistical approaches and demonstrated that the proposed method can alleviate bias when individuals with complete prior medical history information were not representative of the targeted population. The method is motivated by and applied to data from the National Cancer Institute-funded consortium Population-Based Research Optimizing Screening through Personalized Regiments (PROSPR). Findings from the application are important for the evaluation of appropriate use of preventive care and provide a novel tool for dealing with similar analytical challenges with EMR data in broad settings.

Case-control versus case-only estimates of gene-environment interactions with common and misclassified clinical diagnosis

Genetic studies provide valuable information to assess if the effect of genetic variants varies by the nongenetic "environmental" variables, what is traditionally defined to be gene-environment interaction (GxE). A common complication is that multiple disease states present with the same set of symptoms, and hence share the clinical diagnosis. Because (a) disease states might have distinct genetic bases; and (b) frequencies of the disease states within the clinical diagnosis vary by the environmental variables, analyses of association with the clinical diagnosis as an outcome variable might result in false positive or false negative findings. We develop estimates for this setting to be able to assess GxE in a case-only study and we compare the case-control and case-only estimates. We report extensive simulation studies that evaluate empirical properties of the estimates and show the application to a study of Alzheimer's disease.

Colonoscopy Indication Algorithm Performance Across Diverse Health Care Systems in the PROSPR Consortium

BACKGROUND

Despite the importance of characterizing colonoscopy indication for quality monitoring and cancer screening program evaluation, there is no standard approach to documenting colonoscopy indication in medical records.

METHODS

We applied two algorithms in three health care systems to assign colonoscopy indication to persons 50-89 years old who received a colonoscopy during 2010-2013. Both algorithms used standard procedure, diagnostic, and laboratory codes. One algorithm, the KPNC algorithm, used a hierarchical approach to classify exam indication into: diagnostic, surveillance, or screening; whereas the other, the SEARCH algorithm, used a logistic regression-based algorithm to provide the probability that colonoscopy was performed for screening. Gold standard assessment of indication was from medical records abstraction.

RESULTS

There were 1,796 colonoscopy exams included in analyses; age and racial/ethnic distributions of participants differed across health care systems. The KPNC algorithm's sensitivities and specificities for screening indication ranged from 0.78-0.82 and 0.78-0.91, respectively; sensitivities and specificities for diagnostic indication ranged from 0.78-0.89 and 0.74-0.82, respectively. The KPNC algorithm had poor sensitivities (ranging from 0.11-0.67) and high specificities for surveillance exams. The Area Under the Curve (AUC) of the SEARCH algorithm for screening indication ranged from 0.76-0.84 across health care systems. For screening indication, the KPNC algorithm obtained higher specificities than the SEARCH algorithm at the same sensitivity.

CONCLUSION

Despite standardized implementation of these indication algorithms across three health care systems, the capture of colonoscopy indication data was imperfect. Thus, we recommend that standard, systematic documentation of colonoscopy indication should be added to medical records to ensure efficient and accurate data capture.

Delayed Colonoscopy Following a Positive Fecal Test Result and Cancer Mortality

Background

A fecal test followed by diagnostic colonoscopy for a positive result is a widely endorsed screening strategy for colorectal cancer (CRC). However, the relationship between the time delay from the positive test to the follow-up colonoscopy and CRC mortality has not been established.

Methods

From a population-based screening program, we identified CRC patients newly diagnosed from 2005 through 2015 by a positive fecal occult test followed by a colonoscopy. The primary outcome measure was CRC-specific mortality according to four categories for the time elapsed between the positive result and the subsequent colonoscopy.

Results

The 1749 patients underwent colonoscopies within 0–3 months (n = 981, 56.1%), 4–6 months (n = 307, 17.5%), 7–12 months (n = 157, 9.0%), and later than 12 months (n = 304, 17.4%). CRC-specific deaths according to exposure groups were: 13.8% (135 of 981) for 0–3 months, 10.7% (33 of 307) for 4–6 months (crude hazards ratio [HR] = 0.74, 95% confidence interval [CI] = 0.51 to 1.14), 11.5% (18 of 157) for 7–12 months (crude HR = 0.83, 95% CI = 0.51 to 1.42), and 22.7% (69 of 304) for longer than 12 months (crude HR = 1.40, 95% CI = 1.04 to 1.90). The only variable that was associated with mortality risk was the number of positive slides (P = .003). High positivity was twice the value in the 0–3 as the longer-than-12 months group: 51.9% vs 25.0% and similar for the 4–6 and 7–12 months groups (38.1% and 36.5%), respectively. The adjusted HRs for CRC mortality were 0.81 (95% CI = 0.55 to 1.19); 0.83 (95% CI = 0.50 to 1.41), and 1.53 (95% CI = 1.13 to 2.12, P = .006) for the 4–12, 7–12, and longer-than-12-months groups, respectively, compared with the shortest delay group.

Conclusions

Among screen-diagnosed CRC patients, performance of colonoscopy more than 12 months after the initial positive fecal occult blood test was associated with more advanced disease and higher mortality due to CRC.

Measurement Error and Misclassification in Electronic Medical Records: Methods to Mitigate Bias

PURPOSE OF REVIEW

We sought to: 1) examine common sources of measurement error in research using data from electronic medical records (EMR), 2) discuss methods to assess the extent and type of measurement error, and 3) describe recent developments in methods to address this source of bias.

RECENT FINDINGS

We identified eight sources of measurement error frequently encountered in EMR studies, the most prominent being that EMR data usually reflect only the health services and medications delivered within the specific health facility/system contributing to the EMR data. Methods for assessing measurement error in EMR data usually require gold standard or validation data, which may be possible using data linkage. Recent methodological developments to address the impact of measurement error in EMR analyses were particularly rich in the multiple imputation literature.

SUMMARY

Presently, sources of measurement error impacting EMR studies are still being elucidated, as are methods for assessing and addressing them. Given the magnitude of measurement error that has been reported, investigators are urged to carefully evaluate and rigorously address this potential source of bias in studies based in EMR data.

A Bayesian latent class approach for EHR-based phenotyping

Phenotyping, ie, identification of patients possessing a characteristic of interest, is a fundamental task for research conducted using electronic health records. However, challenges to this task include imperfect sensitivity and specificity of clinical codes and inconsistent availability of more detailed data such as laboratory test results. Despite these challenges, most existing electronic health records-derived phenotypes are rule-based, consisting of a series of Boolean arguments informed by expert knowledge of the disease of interest and its coding. The objective of this paper is to introduce a Bayesian latent phenotyping approach that accounts for imperfect data elements and missing not at random missingness patterns that can be used when no gold-standard data are available. We conducted simulation studies to compare alternative phenotyping methods under different patterns of missingness and applied these approaches to a cohort of 68 265 children at elevated risk for type 2 diabetes mellitus (T2DM). In simulation studies, the latent class approach had similar sensitivity to a rule-based approach (95.9% vs 91.9%) while substantially improving specificity (99.7% vs 90.8%). In the PEDSnet cohort, we found that biomarkers and clinical codes were strongly associated with latent T2DM status. The latent T2DM class was also strongly predictive of missingness in biomarkers. Glucose was missing in 83.4% of patients (odds ratio for latent T2DM status = 0.52) while hemoglobin A1c was missing in 91.2% (odds ratio for latent T2DM status = 0.03 ), suggesting missing not at random missingness. The latent phenotype approach may substantially improve on rule-based phenotyping.