Revista Clínica Española (English Edition) Revista Clínica Española (English Edition)
Rev Clin Esp 2018;218:271-8 - Vol. 218 Num.6 DOI: 10.1016/j.rceng.2018.03.018
Original article
Derivation and validation of a predictive model for the readmission of patients with diabetes mellitus treated in internal medicine departments
Derivación y validación de un modelo predictivo de reingreso en pacientes con diabetes mellitus atendidos en servicios de medicina interna
J. Enaa,, , R. Gómez-Huelgasb, B.C. Gracia-Telloc, P. Vázquez-Rodríguezd, J.N. Alcalá-Pedrajase, F.J. Carrasco-Sánchezf, B. Murcia-Casasg, M. Romero-Sánchezh, J.V. Segura-Herasi, J. Carreteroj, Diabetes, Obesity and Nutrition Group of the Spanish Society of Internal Medicine
a Servicio de Medicina Interna, Hospital Marina Baixa, Villajoyosa, Alicante, Spain
b Servicio de Medicina Interna, Hospital Regional Universitario de Málaga, Málaga, Spain
c Servicio de Medicina Interna, Hospital Clínico Universitario Lozano Blesa, Zaragoza, Spain
d Servicio de Medicina Interna, Complexo Hospitalario Universitario A Coruña, A Coruña, Spain
e Servicio de Medicina Interna, Hospital Comarcal de Pozoblanco, Pozoblanco, Córdoba, Spain
f Servicio de Medicina Interna, Hospital Juan Ramón Jiménez, Huelva, Spain
g Servicio de Medicina Interna, Hospital Clínico Universitario Virgen de la Victoria, Málaga, Spain
h Servicio de Medicina Interna, Hospital de Fuenlabrada, Fuenlabrada, Madrid, Spain
i Centro de Investigación Operativa, Universidad Miguel Hernández, Elche, Alicante, Spain
j Servicio de Medicina Interna, Hospital Comarcal de Zafra, Zafra, Badajoz, Spain
Received 11 February 2018, Accepted 16 March 2018
Abstract
Objectives

We developed a predictive model for the hospital readmission of patients with diabetes. The objective was to identify the frail population that requires additional strategies to prevent readmissions at 90 days.

Methods

Using data collected from 1977 patients in 3 studies on the national prevalence of diabetes (2015–2017), we developed and validated a predictive model of readmission at 90 days for patients with diabetes.

Results

A total of 704 (36%) readmissions were recorded. There were no differences in the readmission rates over the course of the 3 studies. The hospitals with more than 500 beds showed significantly (p=.02) higher readmission rates than those with fewer beds. The main reasons for readmission were infectious diseases (29%), cardiovascular diseases (24) and respiratory diseases (14%). Readmissions directly related to diabetic decompensations accounted for only 2% of all readmissions. The independent variables associated with hospital readmission were patient's age, degree of comorbidity, estimated glomerular filtration rate, degree of disability, presence of previous episodes of hypoglycaemia, use of insulin in treating diabetes and the use of systemic glucocorticoids. The predictive model showed an area under the ROC curve (AUC) of 0.676 (95% confidence interval [95% CI] 0.642–0.709; p=.001) in the referral cohort. In the validation cohort, the model showed an AUC of 0.661 (95% CI 0.612–0.710; p=.001).

Conclusion

The model we developed for predicting readmissions for hospitalised patients with type 2 diabetes helps identify a subgroup of frail patients with a high risk of readmission.

Resumen
Objetivos

Hemos desarrollado un modelo predictivo de reingreso hospitalario en pacientes con diabetes. El objetivo es identificar aquella población frágil que requiera estrategias adicionales para evitar reingresos a 90 días.

Métodos

Utilizando datos recogidos en 3 estudios de prevalencia nacionales (2015-2017) que incluyeron un total de 1.977 pacientes hemos desarrollado y validado un modelo predictivo de reingreso a 90 días en pacientes con diabetes.

Resultados

Se registraron un total de 704 (36%) reingresos. No hubo diferencias en la tasa de reingreso a lo largo de los 3 periodos de estudio. Los hospitales de más de 500 camas mostraron de forma estadísticamente significativa (p=0,02) mayores tasas de reingreso que los de menor tamaño. Los motivos principales de reingreso fueron enfermedades infecciosas (29%), enfermedades cardiovasculares (24%) y enfermedades respiratorias (14%). Los reingresos directamente relacionados con descompensaciones diabéticas fueron solo un 2%. Las variables independientes asociadas con reingresos hospitalarios fueron la edad del paciente, el grado de cormobilidad, el filtrado glomerular estimado, el grado de discapacidad, la presencia de episodios previos de hipoglucemia, el uso de insulina en el tratamiento de la diabetes y el uso de glucocorticoides sistémicos. El modelo predictivo mostró en la cohorte de derivación un área bajo de curva ROC: 0,676 (intervalo de confianza al 95% [IC 95%]: 0,642-0,709; p=0,001). En la cohorte de validación el modelo mostró un área bajo la curva: 0,661 (IC 95%: 0,612-0,710; p=0,001).

Conclusión

El modelo de predicción de reingresos para pacientes con diabetes tipo 2 hospitalizados que hemos desarrollado permite identificar un subgrupo de pacientes frágiles con alto riesgo de reingreso.

Keywords
Diabetes mellitus, Hospital medicine, Patient readmission, Epidemiological methods
Palabras clave
Diabetes mellitus, Medicina hospitalaria, Reingreso de pacientes, Métodos epidemiológicos

Article

To access the full text of the publication Revista Clínica Española (English Edition) you need to be a subscriber or member of the Sociedad Española de Medicina Interna.

  • If you are a member of the Sociedad Española de Medicina Interna::
    1.- You must go to the following address http://www.fesemi.org/publicaciones/revista-clinica/ultimo/view
    2.- Authenticate as a member of the SEMI. You must use your login details
    3.- Access to the contents of Revista Clínica Española

  • If you are a subscriber:
    Can you recover your data in the "I Forgot my Password" button on the right sidebar of this website.

Rev Clin Esp 2018;218:271-8 - Vol. 218 Num.6 DOI: 10.1016/j.rceng.2018.03.018
Cookies Policy
x
To improve our services and products, we use cookies (own or third parties authorized) to show advertising related to client preferences through the analyses of navigation customer behavior. Continuing navigation will be considered as acceptance of this use. You can change the settings or obtain more information by clicking here.