Adverse events following surgery of the hip

Abstract: Introduction An adverse event (AE) is when a patient is harmed in healthcare. There are many different definitions of AEs and internationally, many different definitions are used. It may seem obvious that the healthcare should not harm the patients, but nevertheless, over one million patients die each year because of AE following surgical care, globally. This thesis is dedicated to study AEs following acute and elective hip surgery. It consists of four different papers from two studies: The fast-track study and the VARA (Validation of Register Data After Hip Arthroplasty)-study. The overall aim of this thesis was to compare two fast-track systems for hip fracture patients, to validate an instrument for measuring AEs following hip arthroplasty surgery, to study AE claims and to create a new model for measuring AEs following hip arthroplasty surgery. Paper I, the fast-track study The aim of this paper was study if the implementation of a new fast-track system for hip fracture patients could reduce the time from arrival at the hospital to the commencement of surgery. We included 415 consecutive hip fracture patients participating in two parallel fast- track systems in this prospective cohort study. Main outcomes were time to surgery and the proportion of patients that underwent surgery within 24 hours, secondary outcomes were number of AEs and mortality. The patients in the improved fast-track group had in mean three hours shorter time to surgery and there were a 13-percentage difference in the proportion of patients operated within 24 hours. The VARA-study papers II to IV Paper II The aim of this paper was to validate a Swedish instrument for measuring AEs following hip arthroplasty surgery and to calculate the incidence of AEs. In this Swedish multicentre study we included 2,000 acute and elective hip arthroplasty patients and performed retrospective record review (RRR) on all medical records, on all admissions and unplanned out-patient visits within 90 days after surgery. The results were used for validation of the AE measure instrument. The instrument is based on diagnosis codes in the patient register. We also calculated the adjusted cumulative incidence of AEs. The 30-day sensitivity was 6% and specificity 95% for the AE measure instrument. The adjusted cumulative 30-day incidence was 28% for all patients, and 51% and 17% for the acute and the elective patients, respectively. Paper III The aim of this paper was to study the proportion of patients with an AE from paper II that also had an accepted claim from the mutual insurance company of the county councils (Löf). The patients in the VARA-study were matched against Löf’s records and the proportion of patients with a major preventable AE that had an accepted claim was calculated. The proportion was 7%. 94% of the claims were approved and received compensation. The proportion of accepted claims was higher for the elective patients compared with the acute patients. Paper IV The aim for this paper develop a new model to predict AEs following hip arthroplasty surgery. The dataset from the VARA-study was used to train and evaluate different statistical models for predicting AEs. Different machine learning models including neural networks were used. The best performing model was a logistic regression model including the variables age, length of stay for the primary admission, number of readmissions and accident and emergency department (A&E) visits. It was compared with the AE measure instrument from paper II, a model based on diagnosis codes. The new model had two to three times better sensitivity and the same specificity as the diagnosis code-based model. Conclusions Paper I An improved fast-track system that bypasses the A&E could reduce the time to hip fracture surgery by 3 hours and the proportion of patients who underwent surgery within 24 hours. The fast-track system could be performed in a safe way but did not affect mortality or the number of AEs. Paper II The cumulative incidence of AEs following hip arthroplasty surgery was high, and the instrument based on administrative data with diagnosis codes could not measure this incidence with any convincing accuracy. Furthermore, the incidence of AEs was much higher for the acute patients than the elective patients, and only approximately half of the identified AEs had a correct diagnosis code. Paper III The proportion of accepted claims for AEs following hip arthroplasty is very low in Sweden, even for obvious and serious AEs such as periprosthetic joint infection. The proportion of accepted claims is higher for elective than acute patients. Whether the healthcare system fails to inform patients about their rights to file a claim for compensation or the patients are informed but choose not to file a claim is unknown. Paper IV A prediction model for AEs following hip arthroplasty surgery based on administrative data without diagnosis codes is more accurate than a model based on diagnosis codes. In addition to the accuracy variables such as LOS, readmissions, gender and age are robust and objective and, therefore, not prone to bias in a manner similar to diagnosis codes.