Importance of trend in modeling seasonal variation of hip fracture in Montreal, Canada

Hip fracture represents a major public health problem considering the high rates of morbidity and mortality and socioeconomic burden. A better understanding of the seasonal variation of hip fracture would help plan health care and other services for the ageing population in the future. This study examined and modeled seasonal variation of monthly hip fracture incidence (HF) and age-standardized hip fracture rate (HFr). Female and male patients aged 40-74 (F_40-74 and M_40-74) and 75+ (F₇₅₊ and M₇₅₊) of Montreal (Canada) for the period 1993-2004 were gathered. The seasonality of HF and HFr indicated sharp differences between winter and summer. Seasonal variation showed increasing rates of HF and HFr in the late autumn and winter, a decline through spring and remained constant throughout summer. The 12-year trend assessment showed significant trends in HF, increasing for F₇₅₊ and M₇₅₊ groups and decreasing for F_40-74. Significant decreasing trends were found in HFr for F_40-74, F₇₅₊ and M₇₅₊ groups. No significant trends were found in HF or HFr of the M_40-74 group. A seasonal ARIMA (SARIMA) model was applied for HF and HFr time series and a time trend ARIMA model (TT-ARIMA) was developed and fitted to all time series with a significant trend. A multi criteria comparison indicates that the performance of the TT-ARIMA model is two times better than the SARIMA model and the error has been reduced by half. This type of ARIMA model is therefore strongly suggested for time series modeling of HF and HFr with significant trends.