Sleep prediction algorithm based on machine learning technology

Abstract

Schizophrenia is one of the most expensive psychiatric illnesses. This study compared retrospectively health-care resources consumed 12 months before and 24 months after risperidone long-acting injection (RLAI) treatment in Hong Kong. A mirror-image analysis was conducted using data (N=191) from three public hospitals in Hong Kong from 2003 to 2007. The main outcome measure was hospitalisation cost. Other secondary outcomes such as hospitalisation episodes, outpatient visits and adverse events were also compared. A predictive model was established using linear regression based on generalised estimating equations. Analysis showed that RLAI was associated with a reduction in hospitalisation cost by HK$10,001,390 (24.7%) (HK$40,418,694 vs. HK$30,417,303; P-value <0.05). Days of hospitalisation were reduced by 1538 days (10.1%) (15,271 vs. 13,733; P-value <0.05). The predictive model estimated that the hospitalisation cost of patients using RLAI was only 11.1% (3.1–3.93%, 95% confidence interval (CI)) compared to those receiving conventional antipsychotics combined with oral risperidone. Cost of hospitalisation was significantly reduced after RLAI therapy. However, results should be considered as indicative or suggestive only, due to potential channelling bias where certain drug regimens are preferentially prescribed to patients with particular conditions. The findings from our study may be useful in health-care decision making considering treatment options for schizophrenia in resource-limited settings.