Outliers Research Articles

The Accuracy of Hamilton–Perry Population Projections for Census Tracts in the United States

In a first-ever nation-wide census tract evaluation, we assess the accuracy of the Hamilton–Perry population projection method for 65,221 census tracts. We started with 73,607 census tracts but eliminated those for which zeros appeared in age/sex groups. The test uses 1990 and 2000 census tract data by age and gender to construct cohort change ratios, which are then applied to 2000 census tract data to generate 2010 Hamilton–Perry projections that are evaluated in an ex post facto test against the reported 2010 census tract data by age and gender. The projections include: (1) uncontrolled age and gender projections; and (2) age and gender projections controlled to a projection of the population total by census tract. Mean Absolute Percent Error (MAPE) is used to evaluate precision and Mean Algebraic Percent Error (MALPE) is used to evaluate bias. We find that controlling the Hamilton–Perry projections by age for each tract to the linearly projected total population of each tract reduces both MAPE and MALPE within age groups by gender and for total females and total males. As this result suggests, simple linear extrapolation provides more accurate projections of the total population than does the Hamilton–Perry Method. However, even with controlling we find the Hamilton–Perry projections by age to be biased upward. Finally, we use MAPE-R (MAPE- Revised) to evaluate the effect of extreme outliers and find that high MAPEs in the uncontrolled projections are largely driven by extreme errors (outliers) found in less than 1 percent of the 65,221 census tracts used in the study.

Thermal tolerance and the importance of microhabitats for Andean frogs in the context of land use and climate change.

Global warming is having impacts across the Tree of Life. Understanding species' physiological sensitivity to temperature change and how they relate to local temperature variation in their habitats is crucial to determining vulnerability to global warming. We ask how species' vulnerability varies across habitats and elevations, and how climatically buffered microhabitats can contribute to reduce their vulnerability. We measured thermal sensitivity (critical thermal maximum-CTmax ) of 14 species of Pristimantis frogs inhabiting young and old secondary, and primary forests in the Colombian Andes. Exposure to temperature stress was measured by recording temperature in the understorey and across five microhabitats. We determined frogs' current vulnerability across habitats, elevations and microhabitats accounting for phylogeny and then ask how vulnerability varies under four warming scenarios: +1.5, +2, +3 and +5°C. We found that CTmax was constant across species regardless of habitat and elevation. However, species in young secondary forests are expected to become more vulnerable because of increased exposure to higher temperatures. Microhabitat variation could enable species to persist within their thermal temperature range as long as regional temperatures do not surpass +2°C. The effectiveness of microhabitat buffering decreases with a 2-3°C increase, and is almost null under a 5°C temperature increase. Microhabitats will provide thermal protection to Andean frog communities from climate change by enabling tracking of suitable climates through short distance movement. Conservation strategies, such as managing landscapes by preserving primary forests and allowing regrowth and reconnection of secondary forest would offer thermally buffered microhabitats and aid in the survival of this group.

Non-weightbearing imaging and standard knee radiographs are inferior to formal alignment radiographs for calculating coronal alignment of the knee

IntroductionRealignment knee osteotomy relies on accurate preoperative assessment of coronal alignment. Weightbearing (WB) ‘long-leg’ (LL) radiographs are the accepted gold-standard investigation, though in practice standard knee radiographs (short leg; SL) and non-weightbearing (NWB) cross-sectional imaging such as computed tomography (CT) scanograms have been used. We compare the accuracy of SL and NWB radiographs to formal LL alignment radiographs. MethodsA prospectively maintained osteotomy database was reviewed to identify the study population. All patients underwent standardised weightbearing long-leg alignment radiographs. The series was screened consecutively until 30 patients who also underwent WB SL radiographs (‘WB cohort’), and 30 with NWB SL (‘NWB cohort’) radiographs, were identified. Anatomic tibiofemoral angle was calculated by independent reviewers using a validated technique from both radiographs and contrasted. Results60 patients were identified as outlined in the study protocol. There were no differences in baseline demographics. Coronal alignment calculated from SL and LL radiographs differed significantly (median difference 2.1°, p < 0.001). Alignment values from weightbearing SL radiographs demonstrated markedly greater agreement with LL values than those from NWB radiographs (intraclass correlation coefficient 0.878 vs 0.657), with the NWB cohort also exhibiting greater outlier and extreme outlier incidence. ConclusionOur data adds to the growing evidence that SL radiographs are inadequate in the interpretation of knee alignment. In addition, we demonstrate that NWB radiographs (and by extension other NWB modalities such CT scanograms) demonstrate poorer agreement to gold-standard than WB methods. Coronal alignment of the knee cannot be reliably measured from non-weightbearing imaging modalities. Implications for practiceThough potentially useful as an adjunct, non-weightbearing cross-sectional imaging and standard knee radiographs should not be used as a proxy for formal weightbearing long-leg radiographs in osteotomy planning.

Split sample skewness

The shape of a statistical distribution of data is important in such diverse areas as descriptive analysis, risk analysis, portfolio optimization and strategic decision making. If it is known (or believed) that the probability density function of a random variable is not symmetric, the question of its skewness becomes important. Various methods of assessing skewness have been formulated, but none are totally satisfactory. The classical measurement of skewness is based on higher moments of the random variable about its mean. However, it is well known that those measurements are sensitive to extreme outliers. The implication of the sensitivity is that mean-based metrics of skewness are inefficient, especially in small or medium size sample data. Those mean-based metrics are known as Pearson skewness, Quartile skewness and Octile skewness. All have been devised to try to accommodate for the presence of outliers. However, those test statistics are demonstrably inefficient in the presence of outliers. That inefficiency motivates the approach in this paper; the development of an efficient and more robust skewness metric we call Split Sample Skewness, hereafter referred to as SSS. In this context, efficiency means that the SSS metric requires fewer sample observations than the less efficient metrics cited above to achieve the same level of statistical robustness. The name reflects a methodology that partitions the sample into two subgroups at the median. This paper displays the findings of multiple simulation studies adducing evidence bearing on the efficiency of split ample skewness relative to other measures of skewness.

Abstract NG03: Spatial RNA-seq reveals intratumor heterogeneity and transcriptional substructure in high-grade ovarian cancers

Abstract NG03: Spatial RNA-seq reveals intratumor heterogeneity and transcriptional substructure in high-grade ovarian cancers

An assessment of the quality of vaccination data produced through smart paper technology in The Gambia

IntroductionMyChild Solution is an innovative Electronic Immunisation Register (EIR) reliant on Smart Paper Technology, thereby eliminating the need for electronic devices and internet connectivity at the point-of-care. The goal of this study is to characterise the quality of routine immunisation data generated using MyChild Solution compared to data obtained through the conventional health management information system (HMIS) used in The Gambia. MethodWe used the World Health Organization’s (WHO) Data Quality Review (DQR) Toolkit to evaluate MyChild Solution’s data quality in the 19 health facilities across two regions implementing MyChild Solution in The Gambia at the time of the evaluation. We evaluated all applicable data quality metrics as well as additional metrics of interest, including the incidence of recording errors, the incidence of incomplete indicator level data, and implausible dates. Where possible, we compared results to those of the conventional HMIS. ResultsBoth MyChild Solution and the conventional HMIS produced 100% complete and timely data in their reference years. Both systems had no moderate or extreme outliers and showed the expected Penta 1 to Penta 3 dropout direction. However, the proportion of verification factors that are not acceptable was higher in the conventional HMIS. MyChild Solution was found to near perfectly (99.98%) digitise scanned documents. These and other data quality indicators evaluated demonstrate that MyChild Solution produces high quality data with high completeness, timeliness, and consistency compared to the conventional HMIS system. ConclusionMyChild Solution produces high quality data as per the DQR Toolkit metrics and other metrics of interest of interest. The more internally consitent data produced through MyChild Solution compared to the conventional HMIS demonstrates its potential for supporting data-driven decision-making in immunisation.

Outlier by Necessity

Outlier by Necessity

A reliable skin toxicity predictor in permanent breast seed implant brachytherapy

PurposeTo establish skin dose–outcome relationships using a reliable metric in permanent breast seed implant (PBSI). MethodsSixty-seven consecutive patients who underwent PBSI at our institution were included. Skin doses were calculated using two skin dose indices: maximum point dose to the skin surface, Dmax, and D0.2cc for a 2-mm internal skin rind (a surrogate to the dose to 1 cm2 area of skin) from CT-based postoperative treatment plans. Toxicity data were extracted from patients’ charts and photographs. The associations between skin dose and skin toxicity were investigated using the analysis of variance, and the predictive performance of skin dose measures was evaluated using receiver operating characteristic curves. ResultsFor acute reactions, 49.3% of patients had Grade 1, 4.5% Grade 2, and 1.5% Grade 3 toxicity. For telangiectasia at 3 years, very minor and minimally apparent telangiectasia was observed in 25% of patients. Moderate but asymptomatic telangiectasia was observed in 9.1% of cases. Both metrics were significantly associated with the occurrence of acute toxicity and telangiectasia at 3 years (p < 0.01). The predictive values for Dmax and D0.2cc were 0.779 and 0.763, respectively, (p < 0.0001) for acute skin toxicity and 0.786 and 0.810 for telangiectasia (p < 0.0002). Extreme dose outliers (up to 878 Gy) and a high variability were observed for Dmax but not for D0.2cc, illustrating the superior reliability of D0.2cc. ConclusionD0.2cc, as an alternate skin dose measure to Dmax, is a robust metric for measuring skin dose that is simple to calculate, yet is clinically relevant and not prone to inaccuracies inherent to point dose measurement.

Human blood PIG-A mutation and micronucleated reticulocyte flow cytometric assays: Method optimization and evaluation of intra- and inter-subject variation.

We previously described flow cytometry-based methods for scoring the incidence of micronucleated reticulocytes (MN-RET) and PIG-A mutant phenotype reticulocytes (MUT RET) in rodent and human blood samples. The current report describes important methodological improvements for human blood analyses, including immunomagnetic enrichment of CD71-positive reticulocytes prior to MN-RET scoring, and procedures for storing frozen blood for later PIG-A analysis. Technical replicate variability in MN-RET and MUT RET frequencies based on blood specimens from 14 subjects, intra-subject variability based on serial blood draws from 6 subjects, and inter-subject variation based on up to 344 subjects age 0 to 73 years were quantified. Inter-subject variation explained most of the variability observed for both endpoints (≥77%), with much lower intra-subject and technical replicate variability. The relatively large degree of inter-subject variation is apparent from mean and standard deviation values for MN-RET (0.15 ± 0.10%) and MUT RET (4.7 ± 5.0 per million, after omission of two extreme outliers). The influences of age and sex on inter-subject variation were investigated, and neither factor affected MN-RET whereas both influenced MUT RET frequency. The lowest MUT RET values were observed for subjects <11 years old, and males had moderately higher frequencies than females. These results indicate that MN-RET and MUT RET are automation-compatible biomarkers of genotoxicity that bridge species of toxicological interest to include human populations. These data will be useful for appropriately designing future human studies that include these biomarkers of genotoxicity, and highlight the need for additional work aimed at identifying the sources of inter-individual variability reported herein.

TEASER: Fast and Certifiable Point Cloud Registration

We propose the first fast and certifiable algorithm for the registration of two sets of three-dimensional (3-D) points in the presence of large amounts of outlier correspondences. A <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">certifiable algorithm</i> is one that attempts to solve an intractable optimization problem (e.g., robust estimation with outliers) and provides readily checkable conditions to verify if the returned solution is optimal (e.g., if the algorithm produced the most accurate estimate in the face of outliers) or bound its suboptimality or accuracy. Toward this goal, we first reformulate the registration problem using a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">truncated least squares</i> (TLS) cost that makes the estimation insensitive to a large fraction of spurious correspondences. Then, we provide a general graph-theoretic framework to decouple scale, rotation, and translation estimation, which allows solving in cascade for the three transformations. Despite the fact that each subproblem (scale, rotation, and translation estimation) is still nonconvex and combinatorial in nature, we show that 1) TLS scale and (component-wise) translation estimation can be solved in polynomial time via an <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">adaptive voting</i> scheme, 2) TLS rotation estimation can be relaxed to a semidefinite program (SDP) and the relaxation is tight, even in the presence of extreme outlier rates, and 3) the graph-theoretic framework allows drastic pruning of outliers by finding the maximum clique. We name the resulting algorithm TEASER ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Truncated least squares Estimation And SEmidefinite Relaxation</i> ). While solving large SDP relaxations is typically slow, we develop a second fast and certifiable algorithm, named TEASER++, that uses <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">graduated nonconvexity</i> to solve the rotation subproblem and leverages <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Douglas-Rachford Splitting</i> to efficiently certify global optimality. For both algorithms, we provide theoretical bounds on the estimation errors, which are the first of their kind for robust registration problems. Moreover, we test their performance on standard benchmarks, object detection datasets, and the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"/> 3DMatch <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"/> scan matching dataset, and show that 1) both algorithms dominate the state-of-the-art (e.g., RANSAC, branch-&-bound, heuristics) and are robust to more than <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\text{99}\%$</tex-math></inline-formula> outliers when the scale is known, 2) TEASER++ can run in milliseconds and it is currently the fastest robust registration algorithm, and 3) TEASER++ is so robust it can also solve problems without correspondences (e.g., hypothesizing all-to-all correspondences), where it largely outperforms ICP and it is more accurate than Go-ICP while being orders of magnitude faster. We release a fast open-source C++ implementation of TEASER++.

Proficiency test for determination of bromate in drinking water

To evaluate the capacity of national laboratories for determination of bromate in drinking water using Proficiency Testing Program. The preparation method of the Secondary Standard Materials were used as the reference for the sample preparation in this Proficiency Testing Program. The homogeneity and stability of the samples and short-term stability for simulated transportation were tested by Single Factor Analysis of Variance(ANOVA) and Linear Regression and Mean consistency test. The result provided by participant laboratories were analyzed by robust statistics and assessed using the Z-score. A total of 369 laboratories throughout the country participated in the Proficiency Testing Program. 341 laboratories, or 92. 4% of total participating laboratories, obtained satisfactory result. Results provided by 8 laboratories, or 2. 2% of total participating laboratories, were found to suggest doubts in their capacities. Finally, there were 20 laboratories, constituting 5. 4% of total participating laboratories, with result that were found to be outliers. The capacity of national laboratories for determination of bromate in drinking water has been ranked as satisfactory according to statistical analysis of the Proficiency Testing Program result. Only a small portion of the participants require further improvement in their capacities.

Proficiency test for determination of chlorate in drinking water

To evaluate the capacity of national laboratories for determination of chlorate in drinking water using Proficiency Testing Program. The preparation method of the Secondary Standard Materials were used as the reference for the sample preparation in this Proficiency Testing Program. The homogeneity and stability of the samples and short-term stability for simulated transportation were tested by Single Factor Analysis of Variance(ANOVA) and Linear Regression and Mean consistency test. The result provided by participant laboratories were analyzed by robust statistics and assessed using the Z-score. A total of 327 laboratories throughout the country participated in the Proficiency Testing Program. 307 laboratories, or 93. 9% of total participating laboratories, obtained satisfactory result. Results provided by 5 laboratories, or 1. 5% of total participating laboratories, were found to suggest doubts in their capacities. Finally, there were 15 laboratories, constituting 4. 6% of total participating laboratories, with result that were found to be outliers. The capacity of national laboratories for determination of chlorate in drinking water has been ranked as satisfactory according to statistical analysis of the Proficiency Testing Program result. Only a small portion of the participants require further improvement in their capacities.

Proficiency test for determination of iron in drinking water

To evaluate the capacity of national laboratories for determination of iron in drinking water using Proficiency Testing Program. The preparation method of the Secondary Standard Materials were used as the reference for the sample preparation in this Proficiency Testing Program. The homogeneity and stability of the samples and short-term stability for simulated transportation were tested by Single Factor Analysis of Variance(ANOVA) and Linear Regression and Mean consistency test. The result provided by participant laboratories were analyzed by robust statistics and assessed using the Z-score. A total of 1010 laboratories throughout the country participated in the Proficiency Testing Program. 928 laboratories, or 91. 9% of total participating laboratories, obtained satisfactory result. Results provided by 19 laboratories, or 1. 9% of total participating laboratories, were found to suggest doubts in their capacities. Finally, there were 63 laboratories, constituting 6. 2% of total participating laboratories, with result that were found to be outliers. The capacity of national laboratories for determination of iron in drinking water has been ranked as satisfactory according to statistical analysis of the Proficiency Testing Program result. Only a small portion of the participants require further improvement in their capacities.

Validity Of Body Composition Measures By Inbody 770 Bioelectrical Impedance Analyzer

PURPOSE: To test the validity of InBody 770 bioimpedance analyzer (IB770) against isotope dilution (D2O) for total body water (TBW) and against DXA for percent body fat (PBF). METHODS: Fifty-eight apparently healthy women (mean age 21.2 ±2.6 years, BMI 26.7 ±6.3) visited the Exercise Science Physiology Laboratory at KSU after an overnight fast. Subjects were given a 10g deuterium oxide solution (Cambridge Laboratories) to drink within 5 minutes and were asked to remain still in a reclining chair for 4 hours before providing a urine sample. Subjects were assessed via DXA for PBF and IB770 for both TBW and PBF. Frozen urine samples were shipped to an external laboratory for TBW assessment by mass spectroscopy. One extreme outlier was excluded from analysis due to probable D2O measurement error. Correlation analyses were used to assess the validity of TBW and PBF measures from IB770 compared to those from D20 and DXA. RESULTS: TBW estimates from IB770 were similar to those from D20 (32.6 ±5.9 vs. 32.3 ±6.0, respectively). PBF estimates from IB770 were also consistent with those from DXA (35.1 ±9.1 vs. 35.8 ±8.5, respectively). T-tests indicated that there were no significant differences among these measures (p>0.05 for both). IB770 measures for TBW and PBF were strongly correlated with D20 (r=.976, p<0.001) and DXA (r=.961, p<0.001), respectively. CONCLUSIONS: These data suggest that IB770 is a valid method for estimating TBW and PBF and may be appropriate for use as an alternative to D20 and DXA in a young female population.

Influence Of Resistance-exercise Training On Total Physical Activity In Gulf War Veterans With Chronic Pain

Nearly 30% of US military Veterans of the Persian Gulf War are currently suffering from debilitating multisymptom illnesses. A primary complaint among these individuals is chronic widespread musculoskeletal pain (CMP). Previously our lab has observed lower levels of total physical activity (PA) in civilian CMP patients compared to their healthy peers. In general, CMP patients with lower levels of PA are at risk for greater disability. While exercise-training interventions may have demonstrated benefits for fitness and health related outcomes, their influence on total PA, particularly in CMP patients, is an open question. PURPOSE: To quantify the influence of a 16-week resistance exercise training (RET) trial on self-reported and actigraphy measures of total PA in Gulf War Veterans (GV) with CMP. METHODS: Fifty-five GV with CMP were randomly assigned to 16 weeks of RET (n=28) or a wait-list control (n=27). The RET consisted of twice weekly sessions with a personal trainer, initiated at a low intensity (25-35% 1 RM) with progression as tolerated. At baseline, 6 weeks, 11 weeks, and 17 weeks, all participants completed the International Physical Activity Questionnaire (IPAQ) and wore a waist-mounted actigraphy monitor (ACTI) for 7 d during waking hours. Total PA was defined for the IPAQ as the summative total score (MET-min/week) and for the ACTI as total counts per day relative to wear time (counts/d/min). Analyses were limited to GV with valid measures at baseline and at least one additional time point. Data were log transformed and extreme outliers (> 3 SD) were excluded. Separate linear mixed models with group and time point as fixed effects were computed for the IPAQ and ACTI measures, using baseline values as a covariate to control for initial differences. RESULTS: GV assigned to RET completed 88% of training sessions and exhibited strength increases of >20% for 7 of 8 lifts. Estimates for the fixed effects and their interaction were not significant for either measure (IPAQ: FG =0.17, FTP = 0.02, FG×TP = 0.09, p < 0.05; ACTI: FG =0.47, FTP = 0.09, FG×TP = 0.22, p < 0.05). CONCLUSION: Although 16 weeks of RET were well attended and resulted in improvements in fitness for GV with CMP, total PA level, outside of the exercise sessions, did not appear to be impacted. Supported by US Department of Veterans Affairs grant #I01CX000383

Automated data cleaning of paediatric anthropometric data from longitudinal electronic health records: protocol and application to a large patient cohort

‘Big data’ in healthcare encompass measurements collated from multiple sources with various degrees of data quality. These data require quality control assessment to optimise quality for clinical management and for robust large-scale data analysis in healthcare research. Height and weight data represent one of the most abundantly recorded health statistics. The shift to electronic recording of anthropometric measurements in electronic healthcare records, has rapidly inflated the number of measurements. WHO guidelines inform removal of population-based extreme outliers but an absence of tools limits cleaning of longitudinal anthropometric measurements. We developed and optimised a protocol for cleaning paediatric height and weight data that incorporates outlier detection using robust linear regression methodology using a manually curated set of 6,279 patients’ longitudinal measurements. The protocol was then applied to a cohort of 200,000 patient records collected from 60,000 paediatric patients attending a regional teaching hospital in South England. WHO guidelines detected biologically implausible data in <1% of records. Additional error rates of 3% and 0.2% for height and weight respectively were detected using the protocol. Inflated error rates for height measurements were largely due to small but physiologically implausible decreases in height. Lowest error rates were observed when data was measured and digitally recorded by staff routinely required to do so. The protocol successfully automates the parsing of implausible and poor quality height and weight data from a voluminous longitudinal dataset and standardises the quality assessment of data for clinical and research applications.

How to Extract and Explore Big Data for Fraud Detection in the Healthcare Sector: The EOPYY Case Study.

Big Data technologies can contribute to medical fraud detection. The aim of this paper is to present the methodological approach of the Hellenic National Organization for the Provision of Health Services (EOPYY) in data analysis to detect financial or medical fraud. To analyze the data for fraud detection, a selection of prescription data from the year 2018 were examined. The Local Correlation Integral algorithm was applied to detect any outliers on the dataset. The results revealed that 7 out of 879 cases could be characterized as outliers. These outliers must be further investigated to determine if they have been associated with fraud. According to the results of this study, this outliers' detection approach can support and help the fraud detection process conducted by the auditing services in Healthcare sector.

Exploring spatial dependencies in the prevalence of childhood diarrhea in Mozambique using global and local measures of spatial autocorrelation.

Background: Diarrhea in children under 5 years is generally considered as an important public health problem and the major cause of childhood morbidity and mortality in Sub-Saharan Africa. The purpose of this study was to present exploratory spatial data analyses to identify spatial clusters and outliers in the prevalence of childhood diarrhea in Mozambique. Methods: Using data from 2011 Mozambique Demographic and Health Survey, we calculated the prevalence of childhood diarrhea on the district level. Two exploratory spatial data analyses methods were applied, namely, global and local Moran’s I statistics, providing spatial autocorrelation and spatial clusters/outlier in the prevalence of childhood diarrhea, respectively. Results: Choropleth mapping and global Moran’s I statistics showed that the prevalence of childhood diarrhea has clustered distribution across the study area. A local Moran’s I index revealed spatial clusters within the province of Tete, Gaza, Cabo Delgado, and Zambezia. However, spatial outliers were confined within Sofala province. Conclusion: The exploratory spatial data analyses showed various spatial clustering and outliers present in the prevalence of childhood diarrhea, indicating interventions needed in targeted regions.

Introducing the Percent, Number, Availability, and Capacity [PNAC] Spatial Approach to Identify Priority Rural Areas Requiring Targeted Health Support in Light of COVID-19: A Commentary and Application.

Introducing the Percent, Number, Availability, and Capacity [PNAC] Spatial Approach to Identify Priority Rural Areas Requiring Targeted Health Support in Light of COVID-19: A Commentary and Application.

Technology-Dependent Pediatric Inpatients at Children's Versus Nonchildren's Hospitals.

Technology-dependent children (TDC) are admitted to both children's hospitals (CHs) and nonchildren's hospitals (NCHs), where there may be fewer pediatric-specific specialists or resources. Our objective was to compare the characteristics of TDC admitted to CHs versus NCHs. This was a multicenter, retrospective study using the 2012 Kids' Inpatient Database. We included patients aged 0 to 18 years with a tracheostomy, gastrostomy, and/or ventricular shunt. We excluded those who died, were transferred into or out of the hospital, had a length of stay (LOS) that was an extreme outlier, or had missing data for key variables. We compared patient and hospital characteristics across CH versus NCH using χ2 tests and LOS and cost using generalized linear models. In the final sample of 64 521 discharges, 55% of discharges of TDC were from NCHs. A larger proportion of those from CHs had higher disease severity (55% vs 49%; P < .001) and a major surgical procedure during hospitalization (28% vs 24%; P < .001). In an adjusted generalized linear model, the mean LOS was 4 days at both hospital types, but discharge from a CH was associated with a higher adjusted mean cost ($16 754 vs $12 023; P < .001). Because the majority of TDC are hospitalized at NCHs, future research on TDC should incorporate NCH settings. Further studies should investigate if some may benefit from regionalization of care or earlier transfer to a CH.