Predicting health expenditure of not hospitalized patients

Background In the Italian NHS there is a systematic discrepancy between spending and financing, due to difficulties in defining and measuring health needs. Our aim was to construct a tool melting both objective and subjective health data, in order to estimate the related costs of not-hospitalized patients. Methods The study was carried out from 2009 to 2013. Clinical information were obtained from 887 patients who attended General Practitioner in Siena’s area (Italy); subjective health profiles were obtained using Short Form 36 Questionnaire(SF36) whose 8 scales originate 2 indexes: Physical Component Summary(PCS) and Mental Component Summary(MCS). Severity(SI) and Comorbidity Indexes were obtained from the Comorbidity Illness Rating Scale (CIRS). Body Mass Index(BMI) and Charlson Index(CI) were also included. Health costs per patient per year were obtained linking health expenditure (pharmaceutical and hospital discharges data) with patients profiles. Univariate and multivariate analysis were performed with Stata. Results Health expenditure was influenced by age, education, job, BMI, CI, SI, SF36 scales and PCS(p < 0.05). We obtained that SI and PCS were the best indexes to predict health expenditure; from these results, we categorized the PCS in percentiles, we included it in the CIRS list as a new parameter, and we obtained the Severity Index implemented with Perceived Health(SI-PH). In the multivariate analysis the regression coefficient of the SIPH, with health expenditure as an outcome, was strongly higher than that of PCS and slightly higher than that of SI. Conclusions Multidimensional indicators are best predictors of health spending than mono-dimensional ones. Limitations in the use of traditional objective health measures seem to be overcome with the new combined index, useful on out-of-hospital population.