Several variables obtained from routine electronic health records (EHRs), including prescriptions in primary care and laboratory blood test results, are predictive of 90-day mortality following hospitalization for chronic obstructive pulmonary disease (COPD). These were among results of an analysis published in Pharmacological Research.

The researchers for the current study theorized that EHRs offer an opportunity for the development of a prognostic model representative of patients who are hospitalized with COPD, but that EHRs may not record some variables that are used in currently existing models. Although prognostic models based on EHRs can be built upon data from large numbers of patients who are treated by a broad range of health care professionals, the EHR rarely includes potentially important COPD-related information, such as patient questionnaires data related to smoking habits, severity of patients’ symptoms, and quality of life. The EHR also may not record physical examination results.

To explore the value of EHR data in predicting 90-day COPD mortality post hospitalization, the researchers conducted a retrospective cohort study among patients with an unplanned hospitalization for COPD in the NHS Greater Glasgow and Clyde area in Scotland between 2011 and 2017. The prespecified outcome of interest in the current study was all-cause mortality within 90 days of hospital discharge.


Continue Reading

A total of 17,973 patients with an unplanned hospitalization for COPD were included in the present study. Overall, 10,502 of the participants were women, 3703 were <60 years of age, 4231 were ≥80 years of age, and 9923 were in the most deprived quintile of Scottish Index of Multiple Deprivation (SIMD).

Comorbid conditions were usually reported during secondary care health encounters in the 90 days prior to the index admission, including hypertension (16%), diabetes (10%), atrial fibrillation (9%), prior myocardial infarction (7%), cerebrovascular disease (6%), heart failure (5%), chronic kidney injury or disease (5%) acute kidney injury (4%), peripheral vascular disease (4%), dementia (4%), and cancers other than lung cancer (4%).

Bronchodilators had been prescribed at least once in the prior year for 74% of the participants, inhaled corticosteroids for 55% of the patients, loop diuretics for 22% of the participants, and positive inotropic agents for 3% of the patients. The median duration of hospital admission was 5 days (range, 2 to 8 days), with 16% of the participants spending more than 14 days in the hospital and 7% spending more than 28 days in the hospital.

Within 90 days of hospital discharge, a total of 1003 patients died. Mortality increased linearly with age — from less than 2% in those less than 60 years of age to more than 9% in those more than 80 years of age. Additionally, mortality was higher among men than among women (6.7% vs 4.8%, respectively). Per univariate analysis, mortality was higher among those who were most affluent (7.5%) compared with patients with the most severe levels of socio-economic deprivation (5.2%) — most likely reflective of the older age of the most affluent participants.

Mortality following hospital discharge also increased linearly with length of stay. In the basic multivariable model, age, sex, and length of hospital stay were predictive of mortality, but not of social deprivation, even after adjusting for age. The multivariable model was well calibrated, with an area under the curve (AUC) of 0.702 (95% CI, 0.686-0.718).

Overall, 12 variables were identified as being strongly associated with prognosis, including age, sex, length of index hospitalization stay, previous diagnosis of cancer (excluding lung cancer) or dementia, prescription of oxygen or digoxin, neutrophil/ lymphocyte ratio, and serum chloride, urea, creatinine, and albumin, which maintained calibration with only a slight loss of discrimination (AUC, 0.806; 95% CI, 0.792-0.820).

The investigators concluded that the risk-calculator described herein might prove useful for evaluating and auditing clinical practice, thus guiding clinical management and risk-stratifying/-selecting patients to be invited to participate in clinical research.

Disclosure: One of the study authors has declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of the author’s disclosures. 

Reference 

Pellicori P, McConnachie A, Carlin C, Wales A, Cleland JG. Predicting mortality after hospitalisation for COPD using electronic health records. Pharmacol Res. Published online April 2, 2022. doi:10.1016/j.phrs.2022.106199