Background: In the primary care setting, up to one third of patients are categorized as complex, while this proportion is not well known in the hospital setting. By definition, complex patients demand more effort and/or resources and required care processes are not routine or standard. The treating physician’s complexity assessment is the gold standard, which makes difficult the monitoring of patient complexity over time. The Charlson comorbidity index is probably one of the most frequently tool used to estimate patient complexity, but has not been yet compared to the gold standard in this purpose. We aimed to develop and validate a new score to assess inpatient complexity (Patient Complexity Assessment, PCA score) and to compare its performance with the Charlson comorbidity index.
Methods: All consecutive patients discharged from the department of medicine of a large tertiary care hospital were prospectively included into a derivation cohort from October, 2016 to February 16, 2017, and a validation cohort from February 17, 2017, to March, 2017. The residents in charge of the patient assessed complexity at patient discharge. Potential predictors comprised 52 parameters from the electronic health record (EHR), including demographics, health factors, and hospital care usage. We fit a logistic regression model with backward selection to create a scoring system. The score was then externally validated in the validation cohort (temporal validation). Performance was assessed using the Area Under the Receiver Operating Characteristics Curve (AUROC).
Results: Overall, 447 of the 1,407 patients (32%) in the derivation cohort, and 116 of 482 patients (24%) in the validation cohort were identified as complex. Twelve variables independently associated with complexity were included in the PCA score (table). The cut-off of 25 and more points predicting complexity was chosen to match the frequency of observed complex patients. In the derivation set specificity and sensitivity were 84% and 57%, respectively. Positive and negative predictive values were 62% and 81%, respectively. The PCA score showed a very good AUROC in the derivation cohort (0.78, 95% confidence interval [CI] 0.75-0.80), and in the validation cohort 0.78 (95%CI 0.73-0.82). In comparison, the Charlson comorbidity index showed a lower AUROC of 0.58 (95%CI 0.55-0.62) and 0.61 (95%CI 0.54-0.67), respectively.
Conclusions: We derived and externally validated a score that reflects patient complexity in the hospital setting, and performs better than the Charlson comorbidity index. This PCA score that uses variables from the EHR could help monitoring the proportion of complex patients in the hospital setting, and bechmarking patient complexity level between hospitals.