Is it time to stop sweeping data cleaning under the carpet? A novel algorithm for outlier management in growth data

Protocol for improving equity in quantitative big data cleaning: lessons from longitudinal analysis of electronic health records from underrepresented and marginalized communities

BACKGROUND

Large biomedical datasets, including electronic health records (EHRs), are a significant source of epidemiologic data. To prepare an EHR for analysis, there are several data-cleaning approaches; here, we focus on data filtering. Common data-filtering methods employ rules that rely on data from socially constructed dominant populations but are inappropriate for marginalized populations, leading to the loss of valuable data and neglect of underrepresented communities. We propose a novel method based on a phenomenological framework that is more equitable and inclusive, leading to culturally responsive research and discoveries.

METHODS

EHRs from the Yukon-Kuskokwim Health Corporation (YKHC) containing 1 262 035 records from 12 402 unique individuals from 2002 to 2012 were cleaned by using the proposed phenomenological (individual) and common (cohort) data-filtering approach. Within the phenomenological framework, we (i) excluded values that were undeniably biologically impossible for any population, (ii) excludes values that fell outside three standard deviations from the mean value for each individual person, and (iii) used two forms of imputation methods for stable quantitative and qualitative values at the individual level when data were missing.

RESULTS

Compared with common data-filtering practices, the phenomenological approach retained more observations, participants, and a range of outcomes, allowing a truer representation of the priority population. In sensitivity analyses comparing the results of the raw data, the common approach implemented, and the phenomenological approach applied, we found that the phenomenological approach did not compromise the integrity of the results.

CONCLUSION

The phenomenological approach to filtering big data presents an opportunity to better advocate for marginalized communities even when using large datasets that require automated rules for data filtering. Our method may empower researchers who are partnering with communities to embrace large datasets without compromising their commitment to community benefit and respect.

Identifying biologically implausible values in big longitudinal data: an example applied to child growth data from the Brazilian food and nutrition surveillance system

BACKGROUND

Several strategies for identifying biologically implausible values in longitudinal anthropometric data have recently been proposed, but the suitability of these strategies for large population datasets needs to be better understood. This study evaluated the impact of removing population outliers and the additional value of identifying and removing longitudinal outliers on the trajectories of length/height and weight and on the prevalence of child growth indicators in a large longitudinal dataset of child growth data.

METHODS

Length/height and weight measurements of children aged 0 to 59 months from the Brazilian Food and Nutrition Surveillance System were analyzed. Population outliers were identified using z-scores from the World Health Organization (WHO) growth charts. After identifying and removing population outliers, residuals from linear mixed-effects models were used to flag longitudinal outliers. The following cutoffs for residuals were tested to flag those: -3/+3, -4/+4, -5/+5, -6/+6. The selected child growth indicators included length/height-for-age z-scores and weight-for-age z-scores, classified according to the WHO charts.

RESULTS

The dataset included 50,154,738 records from 10,775,496 children. Boys and girls had 5.74% and 5.31% of length/height and 5.19% and 4.74% of weight values flagged as population outliers, respectively. After removing those, the percentage of longitudinal outliers varied from 0.02% (<-6/>+6) to 1.47% (<-3/>+3) for length/height and from 0.07 to 1.44% for weight in boys. In girls, the percentage of longitudinal outliers varied from 0.01 to 1.50% for length/height and from 0.08 to 1.45% for weight. The initial removal of population outliers played the most substantial role in the growth trajectories as it was the first step in the cleaning process, while the additional removal of longitudinal outliers had lower influence on those, regardless of the cutoff adopted. The prevalence of the selected indicators were also affected by both population and longitudinal (to a lesser extent) outliers.

CONCLUSIONS

Although both population and longitudinal outliers can detect biologically implausible values in child growth data, removing population outliers seemed more relevant in this large administrative dataset, especially in calculating summary statistics. However, both types of outliers need to be identified and removed for the proper evaluation of trajectories.

Association between infant breastfeeding practices and timing of peak height velocity: A nationwide longitudinal survey in Japan

BACKGROUND

Several studies have discovered an association between infant feeding practices and puberty timing; however, most have involved female cohorts. We investigated the association between infant feeding practices and the timing of peak height velocity in boys and girls.

METHODS

Data on infant feeding methods and anthropometric measurements were collected from a nationwide Japanese birth cohort study. The age at peak height velocity (APV, years) was estimated and compared. Subsequently, the effects of breastfeeding duration were analyzed.

RESULTS

Of the 13,074 eligible participants, 650, 9455, and 2969 were formula-, mixed-, and exclusively breastfed, respectively. Among girls, the mean APV was significantly later in the mixed-fed (standardized regression coefficient (β): 0.094, 95% confidence interval (CI): 0.004-0.180) and exclusively breastfed (β: 0.150, 95% CI: 0.056-0.250) groups than in the formula-fed group. Among boys, the mean APV was not significantly different among the three groups; however, a sensitivity analysis that excluded preterm birth revealed more significantly delayed APV in the breastfed-only group compared to the formula-fed group. Furthermore, a multiple linear regression model revealed that a longer breastfeeding period was associated with later APV.

CONCLUSIONS

Infant breastfeeding practices can affect the timing of peak height velocity in both boys and girls.

IMPACT

Several studies have discovered an association between infant feeding practices and puberty timing; however, most have involved female cohorts. Age at peak height velocity, derived from longitudinal height measurements, is a useful marker of secondary sexual maturity milestones in boys and girls. A Japanese birth cohort study revealed that breastfed children had a later age at peak height velocity than their formula-fed counterparts; this was more prominent among girls than boys. Furthermore, a duration-effect relationship was observed, where longer breastfeeding duration was associated with a later age at peak height velocity.

New approaches and technical considerations in detecting outlier measurements and trajectories in longitudinal children growth data

BACKGROUND

Growth studies rely on longitudinal measurements, typically represented as trajectories. However, anthropometry is prone to errors that can generate outliers. While various methods are available for detecting outlier measurements, a gold standard has yet to be identified, and there is no established method for outlying trajectories. Thus, outlier types and their effects on growth pattern detection still need to be investigated. This work aimed to assess the performance of six methods at detecting different types of outliers, propose two novel methods for outlier trajectory detection and evaluate how outliers affect growth pattern detection.

METHODS

We included 393 healthy infants from The Applied Research Group for Kids (TARGet Kids!) cohort and 1651 children with severe malnutrition from the co-trimoxazole prophylaxis clinical trial. We injected outliers of three types and six intensities and applied four outlier detection methods for measurements (model-based and World Health Organization cut-offs-based) and two for trajectories. We also assessed growth pattern detection before and after outlier injection using time series clustering and latent class mixed models. Error type, intensity, and population affected method performance.

RESULTS

Model-based outlier detection methods performed best for measurements with precision between 5.72-99.89%, especially for low and moderate error intensities. The clustering-based outlier trajectory method had high precision of 14.93-99.12%. Combining methods improved the detection rate to 21.82% in outlier measurements. Finally, when comparing growth groups with and without outliers, the outliers were shown to alter group membership by 57.9 -79.04%.

CONCLUSIONS

World Health Organization cut-off-based techniques were shown to perform well in few very particular cases (extreme errors of high intensity), while model-based techniques performed well, especially for moderate errors of low intensity. Clustering-based outlier trajectory detection performed exceptionally well across all types and intensities of errors, indicating a potential strategic change in how outliers in growth data are viewed. Finally, the importance of detecting outliers was shown, given its impact on children growth studies, as demonstrated by comparing results of growth group detection.

The Data Error Criteria (DEC) for retrospective studies: development and preliminary application

Retrospective chart review (RCR) studies rely on the collection and analysis of documented clinical data, a process that can be prone to errors. The aim of this study was to develop a defined set of criteria to evaluate RCR datasets for potential data errors. The Data Error Criteria (DEC) were developed by identifying data coding and data entry errors via literature review and then classifying them based on error types. Three components comprise the DEC: general errors, numerical-specific errors, and categorical variable-specific errors. Two reviewers independently applied these criteria via a manual review process to an existing de-identified database. A total of 10,168 errors were identified out of a total of 28,656 data points. The total number of errors included redundancies as certain errors may be included in multiple categories. These included 2515 general errors, 39 numerical-specific errors, and 7614 categorical variable-specific errors. Input-related categorical variable-specific errors occurred most frequently, followed by errors secondary to blank cells. Inter-rater agreement was near perfect for all categories. Identifying errors outlined in the DEC can be crucial for the data analysis stage as they can lead to inaccurate calculations and delay study timelines. The DEC offers a framework to evaluate datasets while reducing time and efforts needed to create high-quality RCR-related databases.

Exploration of body weight in 115 000 young adult dogs of 72 breeds

High body weight (BW), due to large size or excess body fat, has been associated with developmental and metabolic alterations, and degenerative diseases in dogs. Study objectives were to determine mean BW in young adult dogs of different breeds, including changes over a 10-year period. Body weight data from the official Swedish hip dysplasia screening program were used, including data from dogs screened at 1-2.5 years of age, in breeds with ≥ 15 individual observations/year during 2007-2016. Mean BW per breed and sex was established from 114 568 dogs representing 72 breeds. Estimates of breed BW showed significant change in 33 (45%) breeds over the 10-year period. Body weight increased in five breeds (2-14% change) and decreased in 26 breeds (1-8% change). In two breeds, BW increased in male and decreased in female dogs. This observational study provides extensive breed BW data on young adult dogs. The change in breed BW, noted in almost half of the breeds, could be due to changes either in size or in body fat mass. In certain breeds, the change in BW over time might have an impact on overall health. Studies with simultaneous evaluation of BW and body condition over time are warranted.

Artificial Intelligence-Based Medical Data Mining

Understanding published unstructured textual data using traditional text mining approaches and tools is becoming a challenging issue due to the rapid increase in electronic open-source publications. The application of data mining techniques in the medical sciences is an emerging trend; however, traditional text-mining approaches are insufficient to cope with the current upsurge in the volume of published data. Therefore, artificial intelligence-based text mining tools are being developed and used to process large volumes of data and to explore the hidden features and correlations in the data. This review provides a clear-cut and insightful understanding of how artificial intelligence-based data-mining technology is being used to analyze medical data. We also describe a standard process of data mining based on CRISP-DM (Cross-Industry Standard Process for Data Mining) and the most common tools/libraries available for each step of medical data mining.

The impact of the COVID-19 pandemic on a cohort of Labrador retrievers in England

BACKGROUND

The COVID-19 pandemic is likely to have affected the welfare and health of dogs due to surges in adoptions and purchases, changes in the physical and mental health and financial status of dog owners, changes in dogs' lifestyle and routines and limited access to veterinary care. The aims of this study were to investigate whether COVID-19 restrictions were associated with differences in Labrador retrievers' lifestyle, routine care, insurance status, illness incidence or veterinary attendance with an illness, who were living in England and enrolled in Dogslife, an owner-based cohort study. Longitudinal questionnaire data from Dogslife that was relevant to the dates between the 23rd of March and the 4th of July 2020, during COVID-19 restrictions in England, were compared to data between the same dates in previous years from 2011 to 2019 using mixed regression models and adjusted chi-squared tests.

RESULTS

Compared with previous years (March 23rd to July 4th, 2010 to 2019), the COVID-19 restrictions study period (March 23rd to July 4th 2020) was associated with owners reporting increases in their dogs' exercise and worming and decreases in insurance, titbit-feeding and vaccination. Odds of owners reporting that their dogs had an episode of coughing (0.20, 95% CI: 0.04-0.92) and that they took their dogs to a veterinarian with an episode of any illness (0.58, 95% CI: 0.45-0.76) were lower during the COVID-19 restrictions compared to before. During the restrictions period, owners were less likely to report that they took their dogs to a veterinarian with certain other illnesses, compared to before this period.

CONCLUSIONS

Dogslife provided a unique opportunity to study prospective questionnaire data from owners already enrolled on a longitudinal cohort study. This approach minimised bias associated with recalling events prior to the pandemic and allowed a wider population of dogs to be studied than is available from primary care data. Distinctive insights into owners' decision making about their dogs' healthcare were offered. There are clear implications of the COVID-19 pandemic and associated restrictions for the lifestyle, care and health of dogs.

Characterizing Undernourished Children Under-Five Years Old with Diarrhoea in Mozambique: A Hospital Based Cross-Sectional Study, 2015-2019

Diarrhoea is associated with undernutrition and this association is related to increased morbidity and mortality in children under-five. In this analysis we aimed to assess the frequency and associated factors of undernutrition in children under-five with diarrhoea. A hospital-based cross-sectional study was conducted from January 2015 to December 2019 through a surveillance system in five sentinel hospitals in Mozambique. Sociodemographic and clinical information was collected, including anthropometry. A total of 963 children were analysed. The overall undernutrition frequency was 54.1% (95% CI: 50.9−57.2), with 32.5% (95% CI: 29.6−35.5) stunting, 26.6% (95% CI: 23.9−29.6) wasting and 24.7% (95% CI: 22.1−27.5) underweight. Children from Nampula province had 4.7 (p = 0.016) higher odds for stunting compared with children from Maputo. Children whose caregiver was illiterate had higher odds of being underweight 5.24 (p < 0.001), and the wet season was associated with higher odds 1.70 (p = 0.012) of being wasted. Children born under 2500 g of weight had 2.8 (p = 0.001), 2.7 (p < 0.001) and 2.6 (p = 0.010) higher odds for being underweighted, wasted and stunted, respectively. The HIV positive status of the children was associated with higher odds of being underweight 2.6 (p = 0.006), and stunted 3.4 (p = 0.004). The province, caregiver education level, wet season, child’s birthweight and HIV status were factors associated with undernutrition in children with diarrhoea. These findings emphasise the need for additional caregiver’s education on the child’s nutrition and associated infectious diseases. More studies are needed to better understand the social context in which a child with diarrhoea and undernutrition is inserted.

Veterinary Big Data: When Data Goes to the Dogs

Simple Summary

Big data has created many opportunities to improve both preventive medicine and medical treatments. In the field of veterinary medical big data, information collected from companion animals, primarily dogs, can be used to inform healthcare decisions in both dogs and other species. Currently, veterinary medical datasets are an underused resource for translational research, but recent advances in data collection in this population have helped to make these data more accessible for use in translational studies. The largest open access dataset in the United States is part of the Dog Aging Project and includes detailed information about individual dog participant’s physical and chemical environments, diet, exercise, behavior, and comprehensive health history. These data are collected longitudinally and at regular intervals over the course of the dog’s lifespan. Large-scale datasets such as this can be used to inform our understanding of health, disease, and how to increase healthy lifespan.

Abstract

Dogs provide an ideal model for study as they have the most phenotypic diversity and known naturally occurring diseases of all non-human land mammals. Thus, data related to dog health present many opportunities to discover insights into health and disease outcomes. Here, we describe several sources of veterinary medical big data that can be used in research. These sources include medical records from primary medical care centers or referral hospitals, medical claims data from animal insurance companies, and datasets constructed specifically for research purposes. No data source provides information that is without limitations, but large-scale, prospective, longitudinally collected data from dog populations are ideal for further research as they offer many advantages over other data sources.

Surveillance of a vomiting outbreak in dogs in the UK using owner-derived and internet search data

BACKGROUND

In early 2020, the Small Animal Veterinary Surveillance Network reported evidence of an outbreak of acute prolific vomiting in dogs in the UK. The aims of this study were to investigate whether there was evidence for a vomiting outbreak in Dogslife and Google Trends data and to describe its characteristics.

METHODS

Incidence of Dogslife vomiting reports and the Google search index for 'dog vomiting' and 'puppy vomiting' between December 2019 and March 2020 was compared to the respective data from the same months in previous years. Risks for dogs vomiting and factors influencing veterinary attendance in Dogslife were identified using multivariable logistic regression.

RESULTS

This study confirmed a vomiting outbreak was evident in UK dogs between December 2019 and March 2020 using data from Dogslife and Google Trends. The odds of a vomiting incident being reported to Dogslife was 1.51 (95% CI: 1.24-1.84) in comparison to previous years. Dogslife data identified differences in owner-decision making when seeking veterinary attention and identified factors associated with dogs at higher odds of experiencing a vomiting episode.

CONCLUSION

Owner-derived data including questionnaires and internet search queries should be considered a valid, valuable source of information for veterinary population health surveillance.

Associations between Food Group Intake and Physical Frailty in Irish Community-Dwelling Older Adults

Background

Certain nutrients have shown protective effects against frailty, but less is known about the influence of individual food groups. Thus, this study aimed to investigate the relationship between the intake of different food groups and physical frailty in a cohort of community-dwelling older adults in Cork, Ireland.

Methods

One hundred and forty-two (n = 81 females, n = 61 males, age 74.1 ± 6.80 years) Irish community-dwelling volunteers aged ⩾65 years participated in this cross-sectional study. Dietary intake was assessed using a validated food frequency questionnaire (FFQ). Frailty was identified as having 3 or more of the following criteria: weight loss, exhaustion, weakness, slow walking speed and low physical activity. Relationships between intakes of food groups and frailty score were determined using Spearman's rank (and partial rank) correlations and ordinal logistic regression analysis.

Results

Negative Spearman's rank correlations were observed between frailty score and fish and fish products, fruit and vegetables and nuts and seeds, while positive correlations were found between frailty score and potatoes, fats and oils and sugars, preserves and snacks (P < .05). After adjustment for confounders, partial rank correlations remained statistically significant (P < .05) for all of the above dietary variables, with the exception of nuts and seeds (P > .05). Following ordinal logistic regression, the odds ratios (ORs) (95%CI) for frailty incidence for those in the lowest tertile of food group intake compared to the highest were; 3.04 (1.09-8.85) for fish and fish products, 4.34 (1.54-13.13) for fruit and vegetables, 1.52 (0.58-4.15) for nuts and seeds, 0.54 (0.19-1.51) for potatoes, 0.58 (0.17-1.95) for fats and oils and 0.49 (0.16-1.47) for sugars, preserves and snacks.

Conclusion

This study suggests that intakes of selected food groups are independently associated with frailty. These findings may hold significant relevance for the development of future frailty prevention strategies.

Methods for Cleaning and Managing a Nurse-Led Registry

BACKGROUND

Clinical registries provide insight on the quality of patient care by providing data to identify associations and patterns in diagnosis, disease, and treatment. This has led to a push toward using large data sets in healthcare research. Nurse researchers are developing data registries, but most are unaware of how to manage a data registry. This article examines a neuroscience nursing registry to describe a quality control and data management process.

DATA QUALITY PROCESS

Our registry contains more than 90 000 rows of data from almost 5000 patients at 4 US hospitals. Data management is a continuous process that consists of 5 phases: screening, data organization, diagnostic, treatment, and missing data. These phases are repeated with each registry update.

DISCUSSION

The interdisciplinary approach to data management resulted in high-quality data, which was confirmed by missing data analysis. Most technical errors could be systematically diagnosed and resolved using basic statistical outputs, and fixed in the source file.

CONCLUSION

The methods described provide a structured way for nurses and their collaborators to clean and manage registries.