Concordance between an algorithm and clinical diagnosis in the severity of decompensation of chronic diseases in...

Telemonitoring of biological variables can facilitate the early detection of decompensation in chronic disease. Data processing through digital tools enables the delivery of care via automated decision-support systems, such as diagnostic algorithms or alerts in high-rink situations. These tools may improve health outcomes and optimize healthcare management.

A diagnostic accuracy study with a retrospective cross-sectional design was conducted. Agreement in classifying the severity of decompensation in patients with chronic diseases was compared between the algorithm used in the Tele-COMPARTE program and the consensus of three observers, which was considered the reference standard. The weighted Kappa index was used to assess agreement.

The final version of our algorithm achieved a Kappa index (95% CI) of 0,76 (0,7-0.82). By subgroups, the values were 0,84 (0,74-0,94) for alerts in patients with heart failure, 0,71 (0,57-0,85) in COPD, and 0,69 (0,6-0,79) in alerts involving both conditions, all with P <,001. The positive likelihood ratios were 6,3 in low-level and moderate alerts, 10 for high-level alerts, and 186,3 for critical alerts. Predictive values ​​exceeded 75% in all categories.

The level of agreement between the Tele-COMPARTE algorithm and clinical consensus in stratifying the severity of decompensation in chronic disease was "good", reaching a "very good" level in patients with heart failure. The application of the algorithm is a valid and safe procedure that supports its use within our remote monitoring program.