Using systematic data categorisation to quantify the types of data collected in clinical trials: the DataCat project.

BACKGROUND: Data collection consumes a large proportion of clinical trial resources. Each data item requires time and effort for collection, processing and quality control procedures. In general, more data equals a heavier burden for trial staff and participants. It is also likely to increase costs. Knowing the types of data being collected, and in what proportion, will be helpful to ensure that limited trial resources and participant goodwill are used wisely. AIM: The aim of this study is to categorise the types of data collected across a broad range of trials and assess what proportion of collected data each category represents. METHODS: We developed a standard operating procedure to categorise data into primary outcome, secondary outcome and 15 other categories. We categorised all variables collected on trial data collection forms from 18, mainly publicly funded, randomised superiority trials, including trials of an investigational medicinal product and complex interventions. Categorisation was done independently in pairs: one person having in-depth knowledge of the trial, the other independent of the trial. Disagreement was resolved through reference to the trial protocol and discussion, with the project team being consulted if necessary. KEY RESULTS: Primary outcome data accounted for 5.0% (median)/11.2% (mean) of all data items collected. Secondary outcomes accounted for 39.9% (median)/42.5% (mean) of all data items. Non-outcome data such as participant identifiers and demographic data represented 32.4% (median)/36.5% (mean) of all data items collected. CONCLUSION: A small proportion of the data collected in our sample of 18 trials was related to the primary outcome. Secondary outcomes accounted for eight times the volume of data as the primary outcome. A substantial amount of data collection is not related to trial outcomes. Trialists should work to make sure that the data they collect are only those essential to support the health and treatment decisions of those whom the trial is designed to inform.

The efficacy of using countermeasures in a model statement interview / La eficacia del uso de contramedidas en una entrevista de declaración modelo

In a countermeasures experiment, we examined to what extent liars who learn about the Model Statement tool and about the proportion of complications (complications/complications + common knowledge details + self-handicapping strategies) can successfully adjust their responses so that they sound like truth tellers. Truth tellers discussed a trip they had made; liars fabricated a story. Participants were of Lebanese, Mexican, and South-Korean origin. Prior to the interview they did or did not receive information about (I) the working of the Model statement and (II) three types of verbal detail: complications, common knowledge details and self-handicapping strategies. We found no evidence that liars sounded like truth tellers after being informed about the Model Statement and/or types of detail we examined. Actually, veracity differences were similar across experimental conditions, with truth tellers reporting more detail and more complications and obtaining a higher proportion of complications score than liars

En un experimento de contramedidas examinamos hasta qué punto los mentirosos que reciben información sobre la Declaración modelo y la proporción de complicaciones que presenta (complicaciones / complicaciones + detalles de conocimiento general + estrategias de autoobstaculización) pueden ajustar sus respuestas con éxito para que parezca que dicen la verdad. Los que dicen la verdad declararon sobre un viaje que habían hecho; los mentirosos inventaron una historia. Los participantes eran de origen libanés, mexicano y surcoreano. Antes de la entrevista habían recibido o no información sobre (I) el funcionamiento de la Declaración modelo y (II) tres tipos de detalles verbales: complicaciones, detalles de conocimiento general y estrategias de autoobstaculización. No encontramos evidencia de que los mentirosos se parecieran a los que dicen la verdad después de ser informados sobre la Declaración modelo y los tipos de detalle que examinamos. En realidad, las diferencias de veracidad fueron semejantes en todas las condiciones experimentales: los que decían la verdad informaron con mayor detalle y de más complicaciones y obtuvieron una mayor puntuación en la proporción de complicaciones que los mentirosos

Verbal baselining: within-subject consistency of CBCA scores across different truthful and fabricated accounts / La línea base verbal: la congruencia intrasujeto de las puntuaciones del CBCA en diferentes relatos veraces e inventados

Statement Validity Assessment (SVA) proposes that baseline statements on different events can serve as a within-subject measure of a witness' individual verbal capabilities when evaluating scores from Criteria-based Content Analysis (CBCA). This assumes that CBCA scores will generally be consistent across two accounts by the same witness. We present a first pilot study on this assumption. In two sessions, we asked 29 participants to produce one experience-based and one fabricated baseline account as well as one experience-based and one fabricated target account (each on different events), resulting in a total of 116 accounts. We hypothesized at least moderate correlations between target and baseline indicating a consistency across both experience-based and fabricated CBCA scores, and that fabricated CBCA scores would be more consistent because truth-telling has to consider random event characteristics, whereas lies must be constructed completely by the individual witness. Results showed that differences in correlations between experience-based CBCA scores and between fabricated CBCA scores took the predicted direction (cexperience-based = .44 versus cfabricated =.61) but this difference was not statistically significant. As predicted, a subgroup of event-related CBCA criteria were significantly less consistent than CBCA total scores, but only in experience-based accounts. The discussion considers methodological issues regarding the usage of total CBCA scores and whether to measure consistency with correlation coefficients. It is concluded that more studies are needed with larger samples

El Statement Validity Assessment (SVA) propone que las declaraciones sobre diferentes eventos pueden servir como una línea base intrasujeto de la medida de las capacidades verbales individuales de un testigo al evaluar las puntuaciones del Criteria Based Content Analysis (CBCA). Esto implica que las puntuaciones del CBCA serán congruentes en dos relatos del mismo testigo. Presentamos un primer estudio piloto sobre este supuesto. Se pidió a 29 participantes en dos sesiones que elaboraran un relato verdadero (línea base) y otro inventado, además de un relato verdadero y otro inventado (cada uno en situaciones diferentes), arrojando un total de 116 relatos. Se planteó la hipótesis de una correlación al menos moderada entre la declaración fabricada y la verdadera, que indicaría una consistencia entre las puntuaciones en el CBCA de relatos inventados y experimentados y que las puntuaciones en el CBCA inventadas serían más consistentes porque la verdad incluye las características aleatorias de los hechos, mientras que las mentiras las construye totalmente el testigo. Los resultados mostraron que las diferencias en las correlaciones entre las puntuaciones en el CBCA de relatos experimentados y fabricados iban en la dirección predicha (cvivido = .44 frente a cinventado = .61), pero esta diferencia no fue significativa. Como se predijo, un subgrupo de criterios de CBCA relacionados con los hechos fue menos congruente que las puntuaciones totales de CBCA, pero sólo en los relatos de hechos experimentados. Se discuten las implicaciones metodológicas relacionadas con el uso de las puntuaciones totales del CBCA y si se debe medir la consistencia mediante el coeficiente de correlación. Se concluye que se necesitan otros estudios con muestras más grandes

Named entity recognition in electronic health records using transfer learning bootstrapped Neural Networks.

Neural networks (NNs) have become the state of the art in many machine learning applications, such as image, sound (LeCun et al., 2015) and natural language processing (Young et al., 2017; Linggard et al., 2012). However, the success of NNs remains dependent on the availability of large labelled datasets, such as in the case of electronic health records (EHRs). With scarce data, NNs are unlikely to be able to extract this hidden information with practical accuracy. In this study, we develop an approach that solves these problems for named entity recognition, obtaining 94.6 F1 score in I2B2 2009 Medical Extraction Challenge (Uzuner et al., 2010), 4.3 above the architecture that won the competition. To achieve this, we bootstrap our NN models through transfer learning by pretraining word embeddings on a secondary task performed on a large pool of unannotated EHRs and using the output embeddings as a foundation of a range of NN architectures. Beyond the official I2B2 challenge, we further achieve 82.4 F1 on extracting relationships between medical terms using attention-based seq2seq models bootstrapped in the same manner.

Evaluation of Educational intervention on Standard Precautions among Healthcare Provider Based on Health Belief Model

Background and objective: Occupational exposure makes healthcare provider at risk of a variety of infections such as AIDS, Hepatitis B, and Hepatitis C. This study investigated the effect of educational intervention on standardized precautionary behaviors in healthcare provider based on health belief model, in Jam city, Iran during 2016. Methods: This experimental study was carried out on Tohid hospital staff and health care provider of Jam`s health center. Random stratified sampling based on different occupation designated into two groups, intervention (n=50) and control (n=50). After confirming the validity and reliability of the data collection tool, the educational intervention was examined before and after the intervention. Data were analyzed using descriptive statistical methods, independent t-test and one-way ANOVA (SPSS 20). Results: The results revealed that the healthcare provider did not have any previous educational background on standardized precautionary (34.3%). Furthermore, the history of needle stick injuries (42.5%) and contact with patients' body fluids (17.5%) were reported. Educational intervention regarding to standardized precautions in the intervention group was significantly increased the mean score of knowledge constructs, perceived sensitivity, perceived severity, perceived benefits, perceived barriers and behaviors. However, no significant changes were observed in increasing the self-efficacy the score. Conclusion: The results indicate the effectiveness of educational intervention on standard precautionsamong healthcare provider based on health belief model. Educational program based on promotion behavioral pattern in relation to standard precautionsis recommended to the healthcare provider

Implementation of the new ILAE classification of epilepsies into clinical practice - A cohort study.

PURPOSE: Appropriate management of patients with epilepsy requires precise classification of their disease. Implementation of the recent International League Against Epilepsy (ILAE) classification of seizures and epilepsies may affect data on the relative proportions of specific types of seizures or epilepsies and should be tested in everyday practice. The aim of the study was to determine the prevalence of specific epilepsy types, syndromes, and etiologies, as defined by the new ILAE classification, in a large cohort of adult patients with epilepsy. MATERIAL AND METHODS: The single-center cohort study involved consecutive adult patients with epilepsy seen at the university epilepsy clinic. Information about medical history, neurological examination, neuroimaging, electroencephalography (EEG), genetic tests, epilepsy treatment, and other investigations was collected from medical records and prospectively updated if necessary. Epilepsy types and etiology, as well as epileptic syndromes, were classified according to the new ILAE classifications. RESULTS: We studied 653 patients (mean age: 37.2 years, 59.9% were women). Epilepsy was classified as focal in 458 cases (70.2%), generalized in 155 subjects (23.7%), or as combined focal and generalized in 11 patients (1.7%). The epilepsy type was labeled as unknown in 29 (4.4%) patients. A definite cause of epilepsy was identified in 59.4% of the cases, with a structural etiology (n = 179, 27.4%) and genetic or presumed genetic etiology (n = 169, 25.9%) being the most common. In 167 (25.5%) patients, specific epilepsy syndromes, mostly genetic generalized epilepsy syndromes, were diagnosed. CONCLUSION: The use of the recent ILAE classification of seizures and epilepsies in the cohort of patients with epilepsy seen in single epilepsy center enabled unequivocal characterization of epilepsy type in >95% of patients. A definite etiology of epilepsy could be established in about 60% of patients.

Residential scene classification for gridded population sampling in developing countries using deep convolutional neural networks on satellite imagery.

BACKGROUND: Conducting surveys in low- and middle-income countries is often challenging because many areas lack a complete sampling frame, have outdated census information, or have limited data available for designing and selecting a representative sample. Geosampling is a probability-based, gridded population sampling method that addresses some of these issues by using geographic information system (GIS) tools to create logistically manageable area units for sampling. GIS grid cells are overlaid to partition a country's existing administrative boundaries into area units that vary in size from 50 m × 50 m to 150 m × 150 m. To avoid sending interviewers to unoccupied areas, researchers manually classify grid cells as "residential" or "nonresidential" through visual inspection of aerial images. "Nonresidential" units are then excluded from sampling and data collection. This process of manually classifying sampling units has drawbacks since it is labor intensive, prone to human error, and creates the need for simplifying assumptions during calculation of design-based sampling weights. In this paper, we discuss the development of a deep learning classification model to predict whether aerial images are residential or nonresidential, thus reducing manual labor and eliminating the need for simplifying assumptions. RESULTS: On our test sets, the model performs comparable to a human-level baseline in both Nigeria (94.5% accuracy) and Guatemala (96.4% accuracy), and outperforms baseline machine learning models trained on crowdsourced or remote-sensed geospatial features. Additionally, our findings suggest that this approach can work well in new areas with relatively modest amounts of training data. CONCLUSIONS: Gridded population sampling methods like geosampling are becoming increasingly popular in countries with outdated or inaccurate census data because of their timeliness, flexibility, and cost. Using deep learning models directly on satellite images, we provide a novel method for sample frame construction that identifies residential gridded aerial units. In cases where manual classification of satellite images is used to (1) correct for errors in gridded population data sets or (2) classify grids where population estimates are unavailable, this methodology can help reduce annotation burden with comparable quality to human analysts.

Dimensionamento de amostras e o mito dos números mágicos: ponto de vista / Tamaño muestral y el mito de los números mágicos: punto de vista (AU) / Sample size and the myth of magical numbers: point of view

O presente texto dissertará sobre aspectos que devem ser cuidadosamente ponderados por pesquisadores com o propósito de se estabelecer, a priori, o tamanho amostral em estudos na área da saúde, aptidão física e desempenho humano. Entretanto, é importante salientar que o cálculo de amostras é uma ciência que exige seguras suposições realistas, anterior à sua estimativa. Por fim, observamos que, apesar dos grandes avanços tecnológicos que têm contribuído para a maior velocidade de produção científica, é de vital importância que o pesquisador tenha o compromisso em respeitar premissas que habilitam a qualidade do trabalho científico. Nesse sentido, a apropriada especificação do tamanho amostral confere validade interna ao estudo e deve ser requisito obrigatório para a aprovação de projetos de pesquisa (AU)

El presente texto versará sobre aspectos que deberían ser cuidadosamente ponderados por parte de los investigadores con el objetivo de establecer, a priori, el tamaño muestral de estudios en áreas de la salud, actividad física y rendimiento humano. Hay que resaltar que el cálculo del tamaño de la muestra no es una ciencia exacta y, por tanto, requiere suposiciones realistas previas a su estimación. Por último, observamos que, a pesar de los enormes avances biotecnológicos que han contribuido a una más rápida producción científica, es de vital importancia que los investigadores se comprometan a respetar premisas que permitan la calidad del trabajo científico. En este sentido, una adecuada especificación del tamaño muestral confiere validez interna al estudio y deben ser un requisito obligatorio para la aprobación de proyectos de investigación (AU)

The present text will deal with aspects that should be carefully taken into account by researchers with the purpose of establishing sample size in studies in the areas of health, physical capacities and human development. It is important to highlight that sample calculation is not an exact science, and, as such, it requires realistic suppositions before estimation. It is observed that, despite great biotechnological advances that have contributed to faster scientific production, it is of vital importance that researchers commit to respect premises that enable the quality of the scientific work. In this regard, appropriate specifications of sample size grant internal validity to the study and must be a mandatory requisite for research projects approva (AU)

Studying User Income through Language, Behaviour and Affect in Social Media.

Automatically inferring user demographics from social media posts is useful for both social science research and a range of downstream applications in marketing and politics. We present the first extensive study where user behaviour on Twitter is used to build a predictive model of income. We apply non-linear methods for regression, i.e. Gaussian Processes, achieving strong correlation between predicted and actual user income. This allows us to shed light on the factors that characterise income on Twitter and analyse their interplay with user emotions and sentiment, perceived psycho-demographics and language use expressed through the topics of their posts. Our analysis uncovers correlations between different feature categories and income, some of which reflect common belief e.g. higher perceived education and intelligence indicates higher earnings, known differences e.g. gender and age differences, however, others show novel findings e.g. higher income users express more fear and anger, whereas lower income users express more of the time emotion and opinions.

Taxonomy for methods of resource use measurement.

Resource use measures, including forms, diaries and questionnaires, are ubiquitous in trial-based economic evaluations in the UK. However, there are concerns about the accuracy of how they are described, which limits the transparency of reporting. We developed a simple and structured taxonomy for methods of resource use measurement by examining 94 resource use measures (RUMs) employed within clinical trials, conducting a descriptive synthesis of the extracted data and soliciting wider opinion during a period of consultation. The reporting of RUMs was found to be varied and inconsistent. Our new taxonomy, which considered the views of 20 consultees, requires that RUMs are reported with a description of the following: (i) the source of data (patient; patient proxy, e.g. carer, parent or guardian; observation of contemporary events; medical records; or other databases); (ii) who completes the RUM (patient or their proxy, and researcher or health care professional); (iii) how it is administered (to self [the patient], face to face or telephone); (iv) how it is recorded (form, questionnaire, log or diary); and (v) medium of recording (e.g. paper or electronically). Based on the present analysis, we have developed a taxonomy for RUMs that should result in data collection methods being described more accurately.

The assignment of scores procedure for ordinal categorical data.

Ordinal data are the most frequently encountered type of data in the social sciences. Many statistical methods can be used to process such data. One common method is to assign scores to the data, convert them into interval data, and further perform statistical analysis. There are several authors who have recently developed assigning score methods to assign scores to ordered categorical data. This paper proposes an approach that defines an assigning score system for an ordinal categorical variable based on underlying continuous latent distribution with interpretation by using three case study examples. The results show that the proposed score system is well for skewed ordinal categorical data.

Survey analysis with mixture Rasch models.

This research provides a demonstration of the utility of mixture Rasch models (MRMs) for the analysis of survey data. Specifically, a framework based on a mixture partial credit model (MPCM) will be presented. MRMs are able to provide information regarding latent classes (subpopulations without manifest grouping variables) and separate item parameter estimates for each of these latent classes. Analyses can provide insight into how a survey scale is functioning and how survey respondents differ from one another. The paper provides a detailed example with real survey data from a higher education survey administered to college seniors through all stages of model estimation and selection, description of model results, and follow-up analyses using the MRM results. The results found three distinct classes and discussed each class in terms of the pattern of item parameter estimates within class. The paper also investigated differences of class assignment based on the college the student belongs to on campus.

Comparison of characteristics of international and national databases for rheumatoid arthritis: a systematic literature review.

OBJECTIVES: To evaluate current (inter)national registers and observational cohorts in Europe, and to compare inclusion criteria, aims, collected data, and participation in the European League Against Rheumatism (EULAR) repository. METHOD: We performed a systematic search strategy in six literature databases for rheumatoid arthritis (RA). Publications reporting European (inter)national prospective registers/cohorts including > 200 RA patients with at least half a year of follow-up were selected. RESULTS: In total, 417 articles and abstracts were included from four international databases and 39 national databases/cohorts. International databases were of similar design, frequency of data collection and selection criteria and are mostly initiated to monitor and compare clinical patient care among countries. National databases/cohorts vary in aims and inclusion criteria. Half of the national registers are connected to the EULAR repository of databases. CONCLUSIONS: Our findings indicate that, among researchers, there is little awareness of guidelines to set up registers or cohorts and of the existence of the database collaboration network of EULAR.

Performance criteria for verbal autopsy-based systems to estimate national causes of death: development and application to the Indian Million Death Study.

BACKGROUND: Verbal autopsy (VA) has been proposed to determine the cause of death (COD) distributions in settings where most deaths occur without medical attention or certification. We develop performance criteria for VA-based COD systems and apply these to the Registrar General of India's ongoing, nationally-representative Indian Million Death Study (MDS). METHODS: Performance criteria include a low ill-defined proportion of deaths before old age; reproducibility, including consistency of COD distributions with independent resampling; differences in COD distribution of hospital, home, urban or rural deaths; age-, sex- and time-specific plausibility of specific diseases; stability and repeatability of dual physician coding; and the ability of the mortality classification system to capture a wide range of conditions. RESULTS: The introduction of the MDS in India reduced the proportion of ill-defined deaths before age 70 years from 13% to 4%. The cause-specific mortality fractions (CSMFs) at ages 5 to 69 years for independently resampled deaths and the MDS were very similar across 19 disease categories. By contrast, CSMFs at these ages differed between hospital and home deaths and between urban and rural deaths. Thus, reliance mostly on urban or hospital data can distort national estimates of CODs. Age-, sex- and time-specific patterns for various diseases were plausible. Initial physician agreement on COD occurred about two-thirds of the time. The MDS COD classification system was able to capture more eligible records than alternative classification systems. By these metrics, the Indian MDS performs well for deaths prior to age 70 years. The key implication for low- and middle-income countries where medical certification of death remains uncommon is to implement COD surveys that randomly sample all deaths, use simple but high-quality field work with built-in resampling, and use electronic rather than paper systems to expedite field work and coding. CONCLUSIONS: Simple criteria can evaluate the performance of VA-based COD systems. Despite the misclassification of VA, the MDS demonstrates that national surveys of CODs using VA are an order of magnitude better than the limited COD data previously available.

Prevalencia del dolor lumbar durante la infancia y la adolescencia. Una revisión sistemática / Prevalence of low back pain during childoohd and adolescence. A systematic review

Fundamentos: El dolor lumbar en niños y adolescentes es un problema de salud frecuente y su presencia aumenta el riesgo de sufrir dolor lumbar en la edad adulta. El objetivo de este estudio fue describir la prevalencia del dolor lumbar en niños y adolescentes y evaluar la calidad metodológica de los estudios. Metodos: Se efectuó una revisión sistemática de estudios epidemiológicos observacionales. Se realizaron búsquedas en las bases ISI Web of Knowledge, Medline, PEDro, IME, LILACS y CINAHL, rastreo manual y consulta con expertos. La selección de los estudios incluyó resultados con tasas de prevalencia de la muestra. Debían estar publicados o realizados entre los años 1980 y 2011, que el tamaño de la muestra fuera al menos de 50 sujetos, con 18 años de edad o menos, de ambos sexos y de cualquier etnia. Las variables moderadoras de los estudios fueron codificadas por dos de los autores. Se realizaron análisis descriptivos de distribución de frecuencias, porcentajes, recuento de casos y cálculos de medianas para las tasas de prevalencia. Resultados: Se incluyeron 59 artículos. La edad media de los sujetos estudiados fue de 13,56 años, y el 51,15% eran varones. Las medianas de las prevalencias de lumbalgia puntual, de período y de vida fueron respectivamente de 13,60%, 24,75% y 38,50% . Conclusiones: La prevalencia de la lumbalgia aumenta con la edad. El sexo influye en las tasas de prevalencia. La falta de definición y delimitación del dolor lumbar y de especificaciones como frecuencia, duración, e intensidad del dolor, son las deficiencias metodológicas más comunmente encontradas (AU)

Background: Low back pain in children and adolescents is a significant public health problem and its presence at this stage increases the risk of LBP in adulthood. The objective of this study was to describe the prevalence of back pain in children and adolescents and assess the methodological quality of the studies. Methods: A systematic review of observational epidemiology was conducted. The databases ISI Web of Knowledge, MEDLINE, PEDro, IME, LILACS and CINAHL, hand searching and consultation with experts were used. Selection of included studies: results of prevalence rates in the sample; published or performed between 1980 and 2011; sample size of at least 50 subjects; 18 years or less and of any ethnicity. Moderating variables of the studies were coded by two authors. Analysis descriptive of frequency distributions, percentages, numbers of cases, and medium estimates for prevalence rates were performed. Results: 59 articles were included. The mean age was 13.56, 51.15% males. The medians for point, period and lifetime LBP prevalence were respectively13.60%, 24.75%, and 38.50% . Conclusions: The prevalence of LBP increases with age. Sex influences the prevalence rates. The lack of an adequate definition and delimitation of low back pain and the absence of specifications such as frequency, duration, and intensity of pain were the most common methodological flaws in the studies (AU)

Reaching black men who have sex with men: a comparison between respondent-driven sampling and time-location sampling.

OBJECTIVES: The authors explored whether respondent-driven sampling (RDS) can generate a more diverse sample of black men who have sex with men (MSM) than time-location sampling (TLS) by comparing sample characteristics accrued by each method in two independent studies. METHODS: The first study exclusively recruited black MSM through RDS (N=256), while the second recruited MSM through TLS including a subsample of black MSM (N=69). Crude and adjusted point estimates and 95% CIs were calculated for socio-demographic and behavioural characteristics, HIV prevalence and prevalence of unrecognised infections, and were compared using the Z-test. RESULTS: The samples differed significantly regarding all socio-demographic and some behavioural characteristics. Compared with TLS, RDS estimated higher proportions of older, less educated, poorer, currently homeless and self-identified bisexual black MSM. Participants in RDS were less likely to have a main partner, had fewer male partners, were more likely to have a female partner and have both male and female partners, and reported greater methamphetamine, crack and heroin use. Prevalence of HIV and unrecognised infections were slightly higher among RDS participants. CONCLUSIONS: The RDS sample comprised black MSM who were more diverse with respect to socio-demographic characteristics and may also be at higher risk for HIV. Thus, RDS has advantages in reaching higher risk black MSM who are most hidden from intervention research and service delivery. Future studies of black MSM using RDS could use steering strategies to recruit younger participants and other subgroups of greatest interest to public health and prevention.

Improving data accuracy of commercial food outlet databases.

PURPOSE. Assessing food environments often requires using commercially available data. Disparate methods used for classifying food outlets in these databases call for creating a classification approach using common definitions. A systematic strategy for reclassifying food stores and restaurants, as they appear in commercial databases, into categories that differentiate the availability of healthy options is described here. DESIGN AND SETTING. Commercially available data on food outlets including names, addresses, North American Industry Classification System codes, and associated characteristics was obtained for five New Jersey communities. ANALYSIS. A reclassification methodology was developed using criteria and definitions from the literature to categorize food outlets based on availability of healthy options. Information in the database was supplemented by systematic Internet and key word searches, and from phone calls to food outlets. RESULTS. The methodology resulted in 622 supermarket/grocery stores, 183 convenience stores, and 148 specialty stores in the original data to be reclassified into 58 supermarkets, 30 grocery stores, 692 convenience stores, and 115 specialty stores. Outlets from the original list of 1485 full-service restaurants and 506 limited-service restaurants were reclassified as 563 full-service restaurants and 1247 limited-service restaurants. Reclassification resulted in less than one-seventh the number of supermarkets and grocery stores, more than three times the number of convenience stores, and twice as many limited-service restaurants-a much less healthy profile than the one generated by using exclusively the commercial databases. CONCLUSION. An explicit and replicable strategy is proposed for reclassifying food outlets in commercial databases into categories that differentiate on the basis of healthy food availability. The intent is to contribute towards building a consensus among researchers on definitions used in public health research for characterizing different types of food outlets.