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


Autoři: Charlotte S. C. Woolley aff001;  Ian G. Handel aff002;  B. Mark Bronsvoort aff001;  Jeffrey J. Schoenebeck aff001;  Dylan N. Clements aff001
Působiště autorů: The Roslin Institute, The University of Edinburgh, Easter Bush Campus, Midlothian, Edinburgh, United Kingdom aff001;  The Royal (Dick) School of Veterinary Studies, The University of Edinburgh, Easter Bush Campus, Midlothian, Edinburgh, United Kingdom aff002
Vyšlo v časopise: PLoS ONE 15(1)
Kategorie: Research Article
doi: 10.1371/journal.pone.0228154

Souhrn

All data are prone to error and require data cleaning prior to analysis. An important example is longitudinal growth data, for which there are no universally agreed standard methods for identifying and removing implausible values and many existing methods have limitations that restrict their usage across different domains. A decision-making algorithm that modified or deleted growth measurements based on a combination of pre-defined cut-offs and logic rules was designed. Five data cleaning methods for growth were tested with and without the addition of the algorithm and applied to five different longitudinal growth datasets: four uncleaned canine weight or height datasets and one pre-cleaned human weight dataset with randomly simulated errors. Prior to the addition of the algorithm, data cleaning based on non-linear mixed effects models was the most effective in all datasets and had on average a minimum of 26.00% higher sensitivity and 0.12% higher specificity than other methods. Data cleaning methods using the algorithm had improved data preservation and were capable of correcting simulated errors according to the gold standard; returning a value to its original state prior to error simulation. The algorithm improved the performance of all data cleaning methods and increased the average sensitivity and specificity of the non-linear mixed effects model method by 7.68% and 0.42% respectively. Using non-linear mixed effects models combined with the algorithm to clean data allows individual growth trajectories to vary from the population by using repeated longitudinal measurements, identifies consecutive errors or those within the first data entry, avoids the requirement for a minimum number of data entries, preserves data where possible by correcting errors rather than deleting them and removes duplications intelligently. This algorithm is broadly applicable to data cleaning anthropometric data in different mammalian species and could be adapted for use in a range of other domains.

Klíčová slova:

Algorithms – Cohort studies – Data visualization – Dogs – Pets and companion animals – Reproducibility – Simulation and modeling – Statistical data


Zdroje

1. Pritzker L, Ogus J, and Hansen MH. Computer Editing Methods—Some Applications and Results. Bulletin of the International Statistical Institute, Proceedings of the 35th Session. Belgrade, Serbia. 1965;41:442–72

2. Horn PS, Feng L, Li Y, Pesce AJ. Effect of outliers and nonhealthy individuals on reference interval estimation. Clin Chem. 2001;47(12):2137–45. doi: 10.1.1.523.4943 11719478

3. Osborne JW. Data Cleaning Basics: Best Practices in Dealing with Extreme Scores. Newborn and Infant Nursing Reviews. 2010;10(1):37–43. doi: 10.1053/j.nainr.2009.12.009

4. Osborne JW. Is data cleaning and the testing of assumptions relevant in the 21st century? Front Psychol. 2013;4(370):5–7. doi: 10.3389/fpsyg.2013.00370 23805118

5. Clarke R, Shipley M, Lewington S, Youngman L, Collins R, Marmot M, et al. Underestimation of risk associations due to regression dilution in long-term follow-up of prospective studies. Am J Epidemiol. 1999;150(4):341–53. doi: 10.1093/oxfordjournals.aje.a010013 10453810

6. Van Den Broeck J, Cunningham SA, Eeckels R, Herbst K. Data cleaning: Detecting, diagnosing, and editing data abnormalities. PLoS Medicine. 2005;2(10):0966–70. doi: 10.1371/journal.pmed.0020267 16138788

7. Zhang A, Song S, Wang J, Yu PS. Time Series Data Cleaning: From Anomaly Detection to Anomaly Repairing. Proceedings of the VLDB Endowment. 2017;10(10):1046–57. doi: 10.14778/3115404.3115410

8. Eyto E De, Pierson D. Data handling: cleaning and quality control. In Obrador, B, Jones, ID and Jennings, E (Eds) NETLAKE toolbox for the analysis of high-frequency data from lakes (Factsheet 1) Technical report NETLAKE COST Action ES1201. 2016;2–6. Available from; https://www.dkit.ie/system/files/files_with_detail/netlake_toolbox_01_data_handling_and_qaqc.pdf

9. Lo JC, Maring B, Chandra M, Daniels SR, Sinaiko A, Daley MF, et al. Prevalence of obesity and extreme obesity in children aged 3–5 years. Pediatric Obesity 2014;9(3):167–75. doi: 10.1111/j.2047-6310.2013.00154.x 23677690

10. Carsley S, Birken CS, Parkin P, Pullenayegum E, Tu K. Completeness and accuracy of anthropometric measurements in electronic medical records for children attending primary care. J Innov Health Inform. 2018; 25(1):963. doi: 10.14236/jhi.v25i1.963 29717951

11. Yang S, Hutcheon JA. Identifying outliers and implausible values in growth trajectory data. Ann Epidemiol. 2016;26(1):77–80. doi: 10.1016/j.annepidem.2015.10.002 26590476

12. Daymont C, Ross ME, Russell Localio A, Fiks AG, Wasserman RC, Grundmeier RW. Automated identification of implausible values in growth data from pediatricelectronic health records. J Am Med Inform Assoc. 2017;24:1080e7. doi: https://doi.org/10.1093/jamia/ocx037

13. Boone-Heinonen J, Tillotson C, Omalley J, Marino M, Andrea S, Brickman A, et al. Not so implausible: impact of longitudinal assessment of implausible anthropometric measures on obesity prevalence and weight change in children and adolescents. Ann Epidemiol. 2019;31(5):69–74. doi: 10.1016/j.annepidem.2019.01.006 30799202

14. Goldstein H. Data Processing for Longitudinal Studies. Applied Statistics. 1970;19(145):145–51. doi: 10.2307/2346544

15. Lawman HG, Ogden CL, Hassink S, Mallya G, Vander Veur S, Foster GD. Comparing Methods for Identifying Biologically Implausible Values in Height, Weight, and Body Mass Index Among Youth. Am J Epidemiol. 2015;182(4):359–65. doi: 10.1093/aje/kwv057 26182944

16. World Health Organization 1995. Physical status: the use and interpretation of anthropometry. Report of a WHO Expert Committee. World Health Organ Tech Rep Ser. 1995; 854:1–452. 8594834

17. Kuczmarski RJ, Ogden CL, Guo SS, Grummer-Strawn LM, Flegal KM, Mei Z, et al. 2000 CDC Growth Charts for the United States: methods and development. Vital Health Stat. 2002;11:1–190. 15

18. Kim J, Must A, Fitzmaurice GM, Gillman MW, Chomitz V, Kramer E, et al. Incidence and remission rates of overweight among children aged 5 to 13 years in a district-wide school surveillance system. Am J Public Health. 2005;95:1588–94. doi: 10.2105/AJPH.2004.054015 16051932

19. Gundersen C, Lohman BJ, Eisenmann JC, Garasky S, Stewart SD. Child-specific food insecurity and overweight are not associated in a sample of 10- to 15-year-old low-income youth. J Nutr. 2008;138:371–8. doi: 10.1093/jn/138.2.371 18203906

20. Youth Risk Behavior Surveillance System. 2013 YRBS data user’s guide. YRBS, 2012. [Cited 1 November 2019]. Available from: ftp://ftp.cdc.gov/pub/data/yrbs/2011/YRBS_2011_National_User_Guide.pdf

21. Hardy R, Johnson J, Park A. CLOSER work package 1: Harmonised height, weight and BMI user guide. UK Data Service, 2016. [Cited 6 December 2018]. Available from: http://doc.ukdataservice.ac.uk/doc/8207/mrdoc/pdf/closer_wp1_user_guide_v3_new_edition.pdf

22. Freedman DS, Lawman HG, Skinner AC, McGuire LC, Allison DB, Ogden CL.Validity of the WHO cutoffs for biologically implausible values of weight, height,and BMI in children and adolescents in NHANES from 1999 through 2012. Am J Clin Nutr. 2015;102:1000–1006. doi: 10.3945/ajcn.115.115576 26377160

23. Calle EE, Thun MJ, Petrelli JM, Rodriguez C, Heath CW Jr. Body-mass index and mortality in a prospective cohort of U.S. adults. N Engl J Med. 1999;341(15):1097–105. doi: 10.1056/NEJM199910073411501 10511607

24. Surkan PJ, Ettinger AK, Hock RS, Ahmed S, Strobino DM, Minkovitz CS. Early maternal depressive symptoms and child growth trajectories: A longitudinal analysis of a nationally representative US birth cohort. BMC Pediatrics. 2014;14(185):1–8. doi: 10.1186/1471-2431-14-185 25047367

25. Poon WB, Fook-Chong SMC, Ler GYL, Loh ZW, Yeo CL. Creation and validation of the Singapore birth nomograms for birth weight, length and head circumference based on a 12-year birth cohort. Ann Acad Med Singapore. 2014;43(6):296–304 25028138

26. Salt C, Morris PJ, German AJ, Wilson D, Lund EM, Cole TJ, et al. Growth Reference Charts for Dogs of Different Sizes. PLoS ONE. 2017;12(9):e0182064. doi: 10.1371/journal.pone.0182064 28873413

27. Muthalagu A, Pacheco JA, Aufox S, Peissig PL, Fuehrer JT, Tromp G, et al. A rigorous algorithm to detect and clean inaccurate adult height records within EHR systems. Appl Clin Inform. 2014;5(1):118–26. doi: 10.4338/ACI-2013-09-RA-0074 24734128

28. Cole TJ, Donaldson MDC, Ben-shlomo Y. SITAR-a useful instrument for growth curve analysis. Int J Epidemiol. 2010;39(6):1558–66. doi: 10.1093/ije/dyq115 20647267

29. Arribas-Gil A, Romo J. Shape outlier detection and visualization for functional data: the outliergram. Biostatistics. 2014;15(4):603–19. doi: 10.1093/biostatistics/kxu006 24622037

30. Chen S, Banks WA, Sheffrin M, Bryson W, Black M, Thielke SM. Identifying and categorizing spurious weight data in electronic medical records. Am J Clin Nutr. 2018:107(3):420–426. doi: 10.1093/ajcn/nqx056 29566188

31. Spooner S, Shields S, Dexheimer J, Mahdi C, Hagedorn P, Minich T. Weight Entry Error Detection: A Web Service for Real-time Statistical Analysis. AAP Council on Clinical Information Technology Scientific Abstract Session, San Francisco, CA; 2016. https://doi.org/10.1542/peds.141.1_MeetingAbstract.21

32. Shawe-Taylor J, Cristianini N. Kernel methods for pattern analysis. Cambridge, UK; New York: Cambridge University Press; 2004. 462 p

33. Wu DTY, Meganathan K, Newcomb M, Ni Y, Dexheimer JW, Kirkendall ES, et al. Comparison of Existing Methods to Detect Weight Data Errors in a Pediatric Academic Medical Center. AMIA Annu Symp Proc. 2018 Dec 5;2018:1103–1109 30815152

34. Shi J, Korsiak J, Roth DE. New approach for the identification of implausible values and outliers in longitudinal childhood anthropometric data. Ann Epidemiol. 2018;28(3):204–11. doi: 10.1016/j.annepidem.2018.01.007 29398298

35. Welch C, Petersen I, Walters K, Morris RW, Nazareth I, Kalaitzaki E, et al. Two-stage method to remove population- and individual-level outliers from longitudinal data in a primary care database. Pharmacoepidemiol Drug Saf. 2012 Jul;21(7):725–732. doi: 10.1002/pds.2270 22052713

36. Monge AE, Elkan CP. An Efficient Domain-Independent Algorithm for Detecting Approximately Duplicate Database Records. In: Proc SIGMOD 1997 workshop on research issues on data mining and knowledge discovery. Tuscon, AZ; 1997. p. 23–9

37. Monge AE. Matching algorithms within a duplicate detection system. IEEE Techn Bulletin Data Engineering. 2000;23(4):14–20

38. Ripon KSN, Rahman A, Rahaman GMA. A domain-independent data cleaning algorithm for detecting similar-duplicates. Journal of Computers. 2010;5(12):1800–9. doi: 10.4304/jcp.5.12.1800–1809

39. Elm E von Altman DG, Egger M Pocock SJ, Gøtzsche PC Vandenbroucke JP, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: Guidelines for Reporting Observational Studies. PLOS Medicine. 2007;4(10):e296. doi: 10.1371/journal.pmed.0040296 17941714

40. Benchimol EI, Smeeth L, Guttmann A, Harron K, Moher D, Petersen I, et al. The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement. PLoS medicine. 2015;12(e1001885):1–22. doi: 10.1371/journal.pmed.1001885 26440803

41. Clements DN, Handel IG, Rose E, Querry D, Pugh CA, Ollier WER, et al. Dogslife: a web-based longitudinal study of Labrador Retriever health in the UK. BMC veterinary research. 2013;9(13):1–15. doi: 10.1186/1746-6148-9-13 23332044

42. Radford A, Tierney A, Coyne KP, Gaskell RM, Noble PJ, Dawson S, et al. Developing a network for small animal disease surveillance. Veterinary Record. 2010;167(13):472–4. doi: 10.1136/vr.c5180 20871079

43. Banfield Pet Hospital. About us. Banfield Pet Hospital, 2018 [Cited 2018 May 1]. Available from: https://www.banfield.com/about-us

44. Cohort and Longitudinal Studies Enhancement Resources. Harmonised Height, Weight and BMI in Five Longitudinal Cohort Studies: National Child Development Study, 1970 British Cohort Study and Millennium Cohort Study. [data collection]. UK Data Service, 2017. [Cited 6 December 2018]. Available from: http://doi.org/10.5255/UKDA-SN-8207-1

45. Power C, Elliott J. Cohort profile: 1958 British birth cohort (National Child Development Study). Int J Epidemiol. 2006;35(1):34–41. doi: 10.1093/ije/dyi183 16155052

46. Elliott J, Shepherd P. Cohort profile: 1970 British Birth Cohort (BCS70). Int J Epidemiol. 2006;35(4):836–43. doi: 10.1093/ije/dyl174 16931528

47. Hansen K. Millennium Cohort Study First, Second, Third and Fourth Surveys: A Guide to the Datasets (8th Edition). UK Data Service, 2014. [Cited 6 December 2018]. Available from: http://doc.ukdataservice.ac.uk/doc/7464/mrdoc/pdf/mcs_guide_to_the_datasets_020214.pdf

48. Johnson W, Li L, Kuh D, Hardy R. How Has the Age-Related Process of Overweight or Obesity Development Changed over Time? Co-ordinated Analyses of Individual Participant Data from Five United Kingdom Birth Cohorts. PLoS Med. 2015;12(5):e1001828. doi: 10.1371/journal.pmed.1001828 25993005

49. Bann D, Johnson W, Li L, Kuh D, Hardy R. Socioeconomic Inequalities in Body Mass Index across Adulthood: Coordinated Analyses of Individual Participant Data from Three British Birth Cohort Studies Initiated in 1946, 1958 and 1970. PLoS Med. 2017;14(1):e1002214. doi: 10.1371/journal.pmed.1002214 28072856

50. Woolley C, Clements D, Summers K, Querry D, Rose E, Chamberlain K, et al. Dogslife height and weight data—the first 7 years of the cohort, 2010–2017 [dataset]. 2019. University of Edinburgh. The Roslin Institute and Royal (Dick) School of Veterinary Studies. https://doi.org/10.7488/ds/2569

51. American Kennel Club. Official Standard for the Labrador Retriever. American Kennel Club, 1994. [Cited 6 December 2018]. Available from: https://images.akc.org/pdf/breeds/standards/LabradorRetriever.pdf

52. Office for National Statistics. “Average” Briton highlighted on UN World Statistics Day. 2010. [Cited 10 December 2018]. Available from: https://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=12&ved=2ahUKEwjvgJjAnsHaAhXLQ8AKHZLMCfMQFjALegQIABBd&url=https%3A%2F%2Fwww.ons.gov.uk%2Fons%2Fabout-ons%2Fget-involved%2Fevents%2Fevents%2Fun-world-statictics-day%2F-average—briton-highlighted-on-un-world-statistics-day.pdf&usg=AOvVaw3XCQgMDZQsZPs00HLuiLyr

53. Tukey J. Exploratory Data Analysis. Reading, MA: Addison-Wesley; 1977

54. Andritsos P, Fuxman A, Miller RJ. Clean answers over dirty databases: A probabilistic approach. Proc 22nd Int Conf on Data Eng. 2006;30

55. Yorkin M, Spaccarotella K, Martin-Biggers J, Quick V, Byrd-Bredbenner C. Accuracy and consistency of weights provided by home bathroom scales. BMC Public Health. 2013;13(1194):1–5. doi: 10.1186/1471-2458-13-1194 24341761

56. Dubois L, Girad M. Accuracy of maternal reports of pre-schoolers’ weights and heights as estimates of BMI values. Int J Epidemiol. 2007;36:132–8. doi: 10.1093/ije/dyl281 17510077

57. Stein RJ, Haddock CK, Poston WSC, Catanese D, Spertus JA. Precision in weighing: A comparison of scales found in physician offices, fitness centers, and weight loss centers. Public Health Reports. 2005;120(3):266–70. doi: 10.1177/003335490512000308 16134566

58. Huybrechts I, Himes JH, Ottevaere C, De Vriendt T, De Keyzer W, Cox B, et al. Validity of parent-reported weight and height 7of preschool children measured at home or estimated without home measurement: A validation study. BMC Pediatrics. 2011;11(63):1–8. doi: 10.1186/1471-2431-11-63 21736757

59. Dekkers JC, Van Wier MF, Hendriksen IJM, Twisk JWR, Van Mechelen W. Accuracy of self-reported body weight, height and waist circumference in a Dutch overweight working population. BMC Med Res Methodol. 2008;8(69):1–13. doi: 10.1186/1471-2288-8-69 18957077

60. Okamoto N, Hosono A, Shibata K, Tsujimura S, Oka K, Fujita H, et al. Accuracy of self-reported height, weight and waist circumference in a Japanese sample. Obes Sci Pract. 2017;3(4):417–24. doi: 10.1002/osp4.122 29259800

61. Engstrom JL, Paterson SA, Doherty A, Trabulsi M, Speer KL. Accuracy of self-reported height and weight in women: an intergrative review of the literature. J Midwifery Womens Health. 2003;48(5):338–45. doi: 10.1016/s1526-9523(03)00281-2 14526347

62. Flegal KM, Ogden CL, Fryar C, Afful J, Klein R, Huang DT. Comparisons of Self-Reported and Measured Height and Weight, BMI, and Obesity Prevalence from National Surveys: 1999–2016. Obesity. 2019 Oct;27(10):1711–1719. doi: 10.1002/oby.22591 31544344

63. Eager CD, Roy J. Mixed Effects Models are Sometimes Terrible. 2017. [Cited 7 December 2018]. Available from: http://arXiv.org/abs/1701.04858.


Článek vyšel v časopise

PLOS One


2020 Číslo 1