New Clinical Model Predicts Long-Term Post-COVID-19 Condition Risk Two Years Post-Infection

Post-COVID-19 condition can be predicted two years after initial infection using a twelve-variable clinical model developed by Dutch researchers, bioengineer.org reported.

The investigative team evaluated 904 former patients through the CORona Follow-Up research initiative. An optimism-adjusted area under the receiver operating characteristic curve of 71.2 percent was reached using biological sex, body mass index, acute care admission status, and eight pre-existing chronic conditions.

The condition imposes substantial burdens on healthcare systems and working populations.

Affected individuals report profound fatigue, cognitive dysfunction, reduced attention span, and persistent challenges with daily problem solving.

Just 27 percent of those impacted indicate they maintain their pre-infection working hours two full years after their acute illness. Researchers combined data across seven pre-existing Dutch COVID-19 retrospective cohort studies to capture these extended outcomes systematically.

Demographics and clinical settings of the 904 participants

The analytical sample comprised 904 adult participants aged 18 years and older. Data collection occurred between October 1, 2021, and December 31, 2022, capturing symptoms approximately two years following initial SARS-CoV-2 infection. Outpatient management accounted for 16.2 percent of the study cohort, while 56.5 percent required admission to a standard hospital ward. Intensive care units treated 27.3 percent of the analyzed group.

Multivariable Risk Model Development

Researchers winnowed an initial panel of 20 candidate predictors down to twelve final components using multivariable logistic regression. Surviving comorbidity indicators included pre-existing arrhythmia, chronic bronchitis, asthma, angina pectoris, previous stroke, severe back problems, osteoarthritis, and chronic rheumatoid arthritis.

Internal validation using 1,000 bootstrap resamples produced an optimism estimate of 1.5 percent. Calibration plots demonstrated strong agreement between predicted probabilities and observed event frequencies across the complete spectrum of calculated risk.

Strategies for risk stratification and prevention

Early prognostic risk stratification assists physicians in tailoring post-acute monitoring and evaluating targeted therapies. Preventing subsequent viral reinfections is a key secondary prevention objective because reinfections generate cumulative increases in baseline risk.

Recommended booster vaccinations demonstrate effectiveness in lowering post-COVID probabilities even when administered subsequent to an initial infection. Clinicians are encouraged to connect identified high-risk patients with regional specialized rehabilitation and outpatient support networks.

What data sources contributed to the CORFU initiative?

What specific data sources contributed to the CORFU initiative?

The research network combined patient records from five academic medical centers, 18 teaching hospitals, 13 non-teaching secondary hospitals, and community settings across the Netherlands. The Maastricht Intensive Care COVID study, the Bernhoven early detection of vascular damage after COVID-19 initiative, the ZuydErLand COVID-19 registry, the Cardiac complications in patients with COVID-19 consortium, and the POPCORN cohort were among the contributing groups.

How were missing values handled in the study dataset?

Missing data affecting educational entries, comorbidity questions, and body mass index values were handled using stochastic regression imputation with fully conditional specification. This method allowed 33.4 percent of participants with incomplete predictor data to contribute to multivariable analysis.

What performance metrics did prior models achieve compared to this study?

While Antony and co-authors created earlier models reaching higher area under the curve scores of 75 to 76 percent, those frameworks required 86 candidate variables and depended on diagnostic billing codes that flagged a condition prevalence of merely 0.3 percent.

More on this story: New Study Suggests Long COVID Symptoms May Be Less Severe and Frequent Than Previously Thought