Population-wide analysis of hospital laboratory tests to assess seasonal variation and temporal reference interval modification

•Environmental- and dietary-driven seasonality in laboratory test data is well known•Population-wide Danish lab were related to mortality, sex, age trends•A seasonal reference interval modification method proposed•Seasonal improved association rates with diagnoses In health care, tests are used help determine a diagnosis follow patients over time. Test results typically calculated from sampled bodily fluids, taken millions of individuals every day. Reference intervals by clinicians quickly separate what considered normal possibly could be disease related. However, influenced external factors such as diet other exposures, including sunlight, which varies the year. This study demonstrates relevance application seasonally adjusted for purpose better interpreting at both physician science researcher levels. Using patients, we present re-classified showing long-term impact potentially reducing under- overdiagnoses advancing research real-world clinical whole. We identified mortality-, age-, sex-associated differences relation (RIs) population-wide nearly 2 million hospital Denmark comprising more than 300 measurements. A low-parameter mathematical wave-based was developed adjust dietary environment influences during The resulting fit allowed between abnormal tests, patient diagnoses, mortality. highlights need modified RIs presents an approach that has potential reduce over- underdiagnosis, affecting physician-patient interactions electronic record Seasonal variation (e.g., values, prescription use) documented within many countries throughout hemispheres.1Marti-Soler H. Gonseth S. Gubelmann C. Stringhini Bovet P. Chen P.-C. Wojtyniak B. Paccaud F. Tsai D.-H. Zdrojewski T. et al.Seasonal overall cardiovascular mortality: 19 different geographic locations.PLoS One. 2014; 9: e113500https://doi.org/10.1371/journal.pone.0113500Crossref PubMed Scopus (84) Google Scholar,2Guinsburg A.M. Usvyat L.A. Etter M. Xu X. Thijssen Marcelli D. Canaud Marelli Barth Wang Y. variations mortality indicators international hemodialysis populations MONDO registry.BMC Nephrol. 2015; 16: 139https://doi.org/10.1186/s12882-015-0129-yCrossref (25) Scholar Most investigations, however, have focused on specific areas or their variable rates.2Guinsburg Scholar,3Seretakis Lagiou Lipworth L. Signorello L.B. Rothman K.J. Trichopoulos Changing coronary heart disease.JAMA. 1997; 278: 1012-1014https://doi.org/10.1001/jama.1997.03550120072036Crossref Scholar,4Li S.-Y. J.-Y. Chuang C.-L. T.-W. serum sodium levels biochemical parameters among peritoneal dialysis patients.Nephrol. Dial. Transplant. 2008; 23: 687-692https://doi.org/10.1093/ndt/gfm618Crossref (13) Scholar,5Wyse O’Malley G. Coogan A.N. McConkey Smith D.J. daytime multiple immune humans: evidence 329,261 participants UK Biobank cohort.iScience. 2021; 24: 102255https://doi.org/10.1016/j.isci.2021.102255Abstract Full Text PDF Scholar,6Kershenbaum A. Kershenbaum Tarabeia J. Stein N. Lavi I. Rennert Unraveling population averages: examination glucose diabetes using large population-based set.Chronobiol. Int. 2011; 28: 352-360https://doi.org/10.3109/07420528.2011.560315Crossref (23) Investigations correspondingly these areas, thereby limiting breadth findings naturally occurring seasonality.2Guinsburg population-wide, comprehensive across diseases corresponding sets yet conducted. healthy universally care following International Federation Clinical Chemistry (IFCC) guidelines.7Horn P.S. Pesce A.J. intervals: update.Clin. Chim. Acta. 2003; 334: 5-23https://doi.org/10.1016/S0009-8981(03)00133-5Crossref (279) Scholar,8Solberg H.E. IFCC recommendation estimation intervals. RefVal program.Clin. Chem. Lab. Med. 2004; 42: 710-714https://doi.org/10.1515/CCLM.2004.121Crossref (199) crucial because they enable interpret accurately combine information action. Modern do not normally factor changes. limitation, turn, ill-defined defining points action patient-physician interactions.9Özçürümez M.K. Haeckel R. Biological variables influencing limits.Scand. Clin. Invest. 2018; 78: 337-345https://doi.org/10.1080/00365513.2018.1471617Crossref (19) It is, therefore, essential identify natural fluctuations values and, where indicated, intervention protocols accordingly. Several variation, e.g., vitamin D, folate, biomarkers C-reactive protein (CRP), neutrophils, monocytes.2Guinsburg Scholar,10Kasahara A.K. Singh R.J. Noymer Vitamin D (25OHD) United States.PLoS 2013; 8: e65785https://doi.org/10.1371/journal.pone.0065785Crossref (102) Scholar,11Khoo A.-L. Chai L.Y.A. Koenen H.J.P.M. Sweep F.C.G.J. Joosten Netea M.G. van der Ven A.J.A.M. Regulation cytokine responses status individuals.Clin. Exp. Immunol. 164: 72-79https://doi.org/10.1111/j.1365-2249.2010.04315.xCrossref (142) Scholar,12Barchetta Baroni Leonetti de Bernardinis Bertoccini Fontana Mazzei E. Fraioli Cavallo TSH associated adult euthyroid adults.Clin. 15: 389-396https://doi.org/10.1007/s10238-014-0290-9Crossref Scholar,13Maes Mommen K. Hendrickx Peeters D’Hondt Ranjan Meyer Scharpé Components biological seasonality, blood concentrations TSH, TT3, FT4, PRL, cortisol testosterone volunteers.Clin. Endocrinol. 46: 587-598https://doi.org/10.1046/j.1365-2265.1997.1881002.xCrossref (173) first two can changes daylight hours changes, respectively, while latter correlate allergen trends pneumonia flu).5Wyse Scholar,14McArthur J.O. Tang Petocz Samman variability oral contraceptives vitamins B6, B12 folate women reproductive age.Nutrients. 5: 3634-3645https://doi.org/10.3390/nu5093634Crossref (15) Scholar,15Nevo-Shor Kogan Joshua B.-Z. Bahat-Dinur Novack V. Fraenkel calcium, PTH primary hyperparathyroidism.Bone. 2016; 89: 59-63https://doi.org/10.1016/j.bone.2016.05.012Crossref (16) Scholar,16Hinz Seumois Gholami Greenbaum J.A. Lane White Broide D.H. Schulten Sidney Bakhru al.Lack allergy timothy grass pollen passive phenomenon but allergen-specific modulation reactivity.Clin. Allergy. 705-719https://doi.org/10.1111/cea.12692Crossref (28) Although this group been studied repeatedly, assessed only significance seasons; no modeling-based methods applied fluctuations.16Hinz accomplishes task its beneficial applications patient-survival outcomes. Thyroid-stimulating hormone (TSH) also shifts, shown driven temperature. recent al.18Wang Cheng Yu Qiu Lian Guo Hu Lu Yang Liu Data mining: temperature thyroid-stimulating hormone.Clin. Biochem. 60: 59-63https://doi.org/10.1016/j.clinbiochem.2018.08.008Crossref (34) emphasizes importance redefining since up 10% 15% identified.17Santi Spaggiari Brigante Setti Tagliavini Trenti Simoni Semi-annual pattern thyrotropin adults.Sci. Rep. 2019; 10786https://doi.org/10.1038/s41598-019-47349-4Crossref (14) Scholar,18Wang Scholar,19de Grande L.A.C. Goossens Uytfanghe Halsall Yoshimura Noh Hens Thienpont L.M. “big data” describe effect 2017; 55: e34-e36https://doi.org/10.1515/cclm-2016-0500Crossref (20) These discoveries aiding physicians correctly diagnosing thyroid function disorders opposed relying currently stagnant RIs, result underdiagnoses disorders.9Özçürümez Scholar,20Coiro Volpi Marchesi Ferri d’Amato Caffarri Davolio Rossi Caffarra Chiodera Lack secretion affective disorder.Biol. Psychiatry. 1994; 35: 36-41https://doi.org/10.1016/0006-3223(94)91165-7Abstract Another investigated (CVD) cancer Southern vs. Northern Hemisphere.1Marti-Soler As expected, CVD followed offset cyclical patterns year, did not, cancer—overall—is known highly disease.1Marti-Soler Conversely, affected cold weather, due correlation increased energy expenditure, triggering myocardial ischemia stroke.21Izzo J.L. Larrabee Sander Lillis Hemodynamics adaptation.Am. Hypertens. 1990; 3: 405-407https://doi.org/10.1093/ajh/3.5.405Crossref (78) Scholar,22Han M.-H. Yi H.-J. Kim Y.-S. Effect monthly weather air pollution stroke incidence Seoul, Korea.Stroke. 927-935https://doi.org/10.1161/STROKEAHA.114.007950Crossref (52) suggests patient’s home hemisphere, non-cyclical will remain constant year hemispheres, despite environmental shifts.1Marti-Soler Seasonality modeling important tool studying predicting new waves infection pandemics influenza now, SARS-CoV-2.23Chandra Christensen Likhtman Connectivity seasonality: 1918 COVID-19 global perspective.J. Glob. Hist. 2020; 408-420https://doi.org/10.1017/S1740022820000261Crossref (6) While noted above, usually RI indicate whether interval, consider hormonal therefore misclassify patients.24Tendler Bar Mendelsohn-Cohen Karin O. Korem Kohanim Maimon Milo Raz Mayo Tanay Alon U. Hormone medical records circannual endocrine circuits.Proc. Natl. Acad. Sci. USA. 118e2003926118https://doi.org/10.1073/pnas.2003926118Crossref (21) Further, match fluctuation still lie indicative health, example, low, normal, summer spending ample time sun.15Nevo-Shor Scholar,25McKinley M.C. Strain J.J. McPartlin Scott J.M. McNulty Plasma homocysteine Subject variation.Clin. 2001; 47: 1430-1436https://doi.org/10.1093/clinchem/47.8.1430Crossref (55) Of course, question often enough experience abnormality given regardless reported RIs. On hand, (EHR) data-driven cannot—in same subjective way—account assessments must rely large-scale EHR investigations. concerns raised studies9Özçürümez Scholar,26Ma Li Yin Wu Zhang Zou W. al.Establishing thresholds effects gender, age, season thyroglobulin peroxidase antibodies mining big data.Clin. 74: 36-41https://doi.org/10.1016/j.clinbiochem.2019.08.011Crossref (11) Scholar; solution proposed. hormone-focused studies made smaller, limited investigations singular RIs.18Wang Scholar,27Wang Ma Z. al.Reference hormone, free thyroxine, triiodothyronine elderly Chinese persons.Clin. 57: 1044-1052https://doi.org/10.1515/cclm-2018-1099Crossref paper will, instead, make use all available presuming developing robust model. improvement aforementioned studies, able investigate several years define local level, facilitating accurate assisting differentiating disparate subtypes disease.26Ma big-data previously feasible restrictions access individual-level records, making it difficult models multitude tests. Precision medicine approaches may take type analysis even further re-define each individual.28Schüssler-Fiorenza Rose S.M. Contrepois Moneghetti Zhou Mishra Mataraso Dagan-Rosenfeld Ganz A.B. Dunn Hornburg al.A longitudinal precision health.Nat. 25: 792-804https://doi.org/10.1038/s41591-019-0414-6Crossref (223) Scholar,29Cohen N.M. Schwartzman Jaschek Lifshitz Hoichman Balicer Shlush L.I. Barbash Personalized quantify potentials individuals.Nat. 27: 1582-1591https://doi.org/10.1038/s41591-021-01468-6Crossref (5) summary, aims mass scale propose improve classification protocols.27Wang Scholar,30Wang Gender- age-specific common analytes – derivation real data.J. 39: 384-391https://doi.org/10.2478/jomb-2019-0046Crossref (9) By applying sinusoidal these, global, identifying changing undertaken. investigates systematically identifies classify patients’ undergoing workup future investigating EHRs. original dataset included 339,609,717 measurements derived 1,987,180 unique setting. cleaned outlined Figure 1A (see experimental procedures), 310,455,299 usable quantitative study. summarized 1B Table 1, see females, average, receive fewer per encounter males; there men represented database. addition, equal contribution Region Zealand Capital Region. Approximately 0.5% (1.6 observations) contained qualitative color specimen genotyping excluded analysis.Table 1Data availability overviewFemaleMaleOverallntot∼160M, 51.8%∼150M, 48.2%∼310M, 100%nind∼1.06M, 55.2%∼0.86M, 44.8%∼1.92M, 100%Age (years)Mean (SD)59.0 (21.8)59.9 (20.4)59.4 (21.1)Median (min, max)63.2 (0, 110)64.6 110)64.0 110)RegionCapitol Region∼114M (71.0%)∼106M (71.5%)221M (71.2%)Region Zealand∼46.7M (29.0%)∼42.5M (28.5%)89.3M (28.8%)Unique PIDMean (SD)31.5 (20.9)31.9 (21.2)31.7 max)30 (1, 261)29 232)29 261)Total (SD)151 (311)173 (378)161 (343)Median max)56 16,400)53 1,900)55.0 1,900)This table gives general statistical overview examined 1A. Results show fairly consistent sexes regions. PID stands identification. Open tab After filtering stratum requirements described procedures, 1,924,869 1,545 (defined analyte-specimen pair) indications 2. With robustness visually 2, total 404 eligible seasonality. Parameter fits coordinated false discovery rate (FDR)-corrected p S1. Minimum algorithm resulted 169 fitted 110 significant amplitude and/or shifts (p < 0.05 FDR corrected). 3 summarizes main heatmap, stratified level. k-means clustering us cluster degree value highest. Mann-Whitney-Wilcoxon performed confirm day deemed = 0.26).Figure 3Heatmaps (denoted analyte-source pair)Show full caption(A) (with least one parament fit, FDR-corrected 0.05) six groups, empirically labeled seen based amplitude/offset cluster.(B) Similarly, remaining, non-significant clustered into three tendencies seen. Bluer peak