Leveraging Large Language Models to Analyze Continuous Glucose Monitoring Data: A Case Study.

Continuous glucose monitors (CGM) provide patients and clinicians with valuable insights about glycemic control that aid in diabetes management. The advent of large language models (LLMs), such as GPT-4, has enabled real-time text generation and summarization of medical data. Further, recent advancements have enabled the integration of data analysis features in chatbots, such that raw data can be uploaded and analyzed when prompted. Studying both the accuracy and suitability of LLM-derived data analysis performed on medical time series data, such as CGM data, is an important area of research. The objective of this study was to assess the strengths and limitations of using an LLM to analyze raw CGM data and produce summaries of 14 days of data for patients with type 1 diabetes. This study used simulated CGM data from 10 different cases. We first evaluated the ability of GPT-4 to compute quantitative metrics specific to diabetes found in an Ambulatory Glucose Profile (AGP). Then, using two independent clinician graders, we evaluated the accuracy, completeness, safety, and suitability of qualitative descriptions produced by GPT-4 across five different CGM analysis tasks. We demonstrated that GPT-4 performs well across measures of accuracy, completeness, and safety when producing summaries of CGM data across all tasks. These results highlight the capabilities of using an LLM to produce accurate and safe narrative summaries of medical time series data. We highlight several limitations of the work, including concerns related to how GPT-4 may misprioritize highlighting instances of hypoglycemia and hyperglycemia. Our work serves as a preliminary study on how generative language models can be integrated into diabetes care through CGM analysis, and more broadly, the potential to leverage LLMs for streamlined medical time series analysis.

[COVID-19 and the predominant groups of preventive behaviors and associated factors: semi-urban area]. / La COVID-19 y los grupos predominantes de conductas preventivas y factores asociados: zona semiurbana.

Background: Non-pharmacological interventions (NFI) or preventive actions (PA) are the best alternatives to control future pandemics, especially in vulnerable populations, such as semi-urban areas. Objective: To describe the predominant health behavior groups (HBG) and associated factors during the third wave of COVID-19 in a semi-urban area. Material and methods: A survey which included characteristics, factors related to COVID-19, perceptions of health behavior and PA was applied in a probabilistic sample in a first-level hospital in the State of Mexico. People of both sexes over 18 years of age were included. Using a hierarchical cluster analysis, HBGs were obtained and characterized with a descriptive and multivariate statistical analysis. Results: In a probabilistic sample (n = 260), 4 HBGs were identified: 2 of high-risk (HRG) and 2 of low-risk (LRG), and the proportions were 43.5% y 56.5%, respectively. The sociodemographic characteristics of both groups were similar. Perceptions of severity and COVID-19-related barriers significantly influenced health behaviors in LRG. In HRG, low security played a significant role, highlighting the importance of comorbidities as a clinical factor. Conclusions: In a semi-urban area, 2 crucial health behaviors were identified: one associated with low risk and the other with high risk. In the HRG, the perception of insecurity was particularly relevant, emphasizing the importance of comorbidities as a clinical factor.

Introducción: las intervenciones no farmacológicas (INF) o acciones preventivas (AP) contra enfermedades son la mejor alternativa para controlar futuras pandemias, en especial en poblaciones vulnerables, como las zonas semiurbanas. Objetivo: describir los grupos predominantes de conductas de salud (GCS) y los factores asociados durante la tercera ola de la COVID-19 en una zona semiurbana. Material y métodos: se aplicó una encuesta que incluyó las características, los factores relacionados con COVID-19, las percepciones de conductas de salud y las AP, en una muestra probabilística en un hospital de primer nivel del Estado de México. Se incluyeron personas de ambos sexos, mayores de 18 años. Mediante un análisis de conglomerados se caracterizaron los GCS con un análisis estadístico descriptivo y multivariado. Resultados: en una muestra probabilistica (n = 260), se identificaron cuatro GCS: 2 de riesgo alto de contagio por la COVID-19 (GRA) y 2 de riesgo bajo (GRB) y las proporciones fueron 43.5% y 56.5%, respectivamente. Las características sociodemográficas de los grupos fueron similares. Para los GRB los factores significativos fueron las percepciones sobre la severidad y las barreras relacionadas con la COVID-19. En los GRA fue la seguridad baja y destacó la importancia de la comorbilidad como factor clínico. Conclusiones: en una zona semiurbana se identificaron 2 conductas de salud de importancia: una de bajo riesgo y otra de alto riesgo. En el GRA, la percepción de seguridad baja fue especialmente relevante, lo cual resalta la importancia de las comorbilidades como factor clínico.

Diabetes gestacional, hipotiroidismo y concentración urinaria de yodo en embarazadas. Yodurias en escolares en Paraguay: Exceso de yodo en la sal y riesgo de hiper e hipotiroidismo / Gestational diabetes, hypothyroidism and iodine urinary concentration in pregnant women. Urinary iodine in schoolage children in Paraguay: Iodine excess in salt and hyper and hypothyroidism risk

The main inclicator of iodine nutritional status of salt for human con-sumption is the urinary concentration of iodine, which is useful in monitoring universal levels in salt. Paraguay has suffered deficiency in salt with highprevalence of goiter in thepopulation reaching48.6%. In 2000, results of the ThyroidMobileproject in Latin America showed a goiter prevalence of 17% but 79.6% of the consumed iodized salt was adequate in paraguayan households (more than 15 ppm of iodine). The average values of iodine urinary concentration in 4487 scholars was 437 µg/L, 30% of the average value was between the ideal of 100 to 199 µg/L and 46.1% were above 300 mg/L, with risk of excess iodine. Urinary concentration of iodine during pregnancy was measure for the first time in Paraguay in 200 women; the average was 484 µg/L but diabetes and hypothyroidism during pregnancy was observe in 50% of them. Objective: Determine iodine concentration in urine of the school population in 17 departments of the country and measure the levels of iodine in urine in 200 pregnant women from 15-37 years of age and their levels of glycaemia, and thyroid TSH. Methods: During the years, 2006 and 2007 were evaluated 4487 school randomized, in a probabilistic sample, 3198 in rural areas and 1,289 in urban areas; casual urine samples was collected to determine iodine content. The urinary concentration of iodine in two hundredpregnant women, concentration of iodine in salt, and levels of glycaemia and TSH were also included in the present study All candidates agreed with the informed consent under the ethics rules. Results: These data showed by comparing the levels of iodine in urine concentration in more than adequate and excessive level but this last was markedly elevated (93.8%). The median level greater than 300 µg/L was observed in 100% of the 4,487 urine samples, while the median urinary iodine levels of300-500 µg/L was 91.3% and above 500 was 9.7%. These levels showed the risk of developing thyroid autoimmune diseases. In 200 pregnancy women the average of urine concentration of iodine was 498 pcg/L, subclinical and clinical hypothyroidism and gestational diabetes was 50% of both. Conclu-sion: To obtain normal median urinary iodine levels is necessary to decrease the amount of iodine in salt. The main recommendation is to insist on adequate monitoring of iodized salt consumption. Also continue surveillance and monitoring in sentinel sites reporting the importance of adequate iodine intake of the population especially in pregnant women that also need to be warm about the diabetes during pregnancy but also a regimen is necessary to keep them within the normal levels of glycemia. The subclinical hypothyroidism has been also studied in the same patients.

El principal indicador del impacto de la yodación de la sal de consumo humano es la concentración urinaria de yodo la cual es útil en el monitoreo de la sal. En la encuesta del año 1988 realizada en el Paraguay, se alcanzó una prevalencia de bocio de 48,6% en la población escolar con un déficit de yodo en la sal, pero el año 2000 en el estudio del proyecto de Tiroides Móvil, se redujo por el método ecográfico a 17%. Ese mismo año la mediana de los niveles urinarios en niños escolares de 6-12 años fue 258 µg/L, considerando que 30% se encontraba entre el valor ideal de 100 a 199 µg/L y 46,1% sobre 300 µg/L, implicando un riesgo de exceso de yodo en dicha población con las posibles consecuencias de aparición tanto de hipo como de hipertiroidismo. El 93% presentó exceso de yodo en la sal y la mediana urinaria fue 437 ug/ mL. Un grupo de embarazadas han sido estudiadas en el Hospital San Pablo por primera vez en el Paraguay para la determinación de la yoduria, yodo en sal, diabetes gestacional, hipotiroidismo en el embarazo y sus valores antropométricos resultando un hipotiroidismo subclínico del 50% y diabetes gestacional del 50%. Objetivo: Determinar las yodurias en la población escolar de 6-12 años pre púber de ambos sexos en 17 departamentos del país y en una submuestra en embarazadas y también niveles de glicemia en el embarazo en un hospital de Asunción. Sujetos y Métodos: En el periodo del 2006-2007 fueron evaluados 4.487 escolares randomizados, en una muestra probabilística, 3.198 en el área rural y 1.289 en el área urbana con muestras en orina casual. El año 2015, doscientas embarazadas fueron estudiadas con muestras de glicemia en ayunas y 2 horas post prandial, dosaje de TSH, yodo en sal de 100 gramos de la sal de consumo en sus hogares y de orina para ver la concentración de yodo en sal y urinaria de yodo. Todos los candidatos fueron estudiados bajo consentimiento autorizado y de acuerdo a normas éticas. Resultados: Al comparar los niveles de yodo en orina en el nivel más que adecuado y el excesivo de la yoduria se vió que éste último nivel estaba marcadamente elevado (93.8%), que la mediana de los niveles mayores a 300 ug/L era del 100% del total de las 4.487 muestras de orina, mientras que la mediana de los niveles de yoduria entre 300-500 µg/L fue de 91,3% y por encima de 500 fue 9.7% lo cual implicó el riesgo de desarrollar enfermedades tiroideas autoinmunes. Las embarazadas presentaron una mediana urinaria de yodo 484 µg/L, diabetes gestacional e hipotiroidismo subclínico en 50%. Conclusión: Para normalizar los niveles de yoduria, es necesario disminuir la cantidad de yodo en la sal. La recomendación es insistir en el monitoreo de la adecuada yodación de las sales de consumo familiar y continuar con la vigilancia y el monitoreo constante en sitios centinelas divulgando la importancia de la adecuada ingesta de yodo a la población. Un régimen dietético a las embarazadas es necesario implementar en el primer trimestre del embarazo.