Self-reported versus health administrative data: implications for assessing chronic illness burden in populations. A cross-sectional study