Background: For population health management, it is important to have health indices that can monitor prevailing health trends in the population. Traditional health indices are generally measurable at different geographical levels with varied number of health dimensions. The aim of this work was to develop and validate a scalable and extendable multi-dimensional health index based on individual data.
Methods: We defined health to be made up of five different domains: Physical, Mental, Social, Risk, and Healthcare utilization. Item response theory was used to develop models to compute domain scores and a health index. These were normalized to represent an individual's health percentile relative to the population (0 = worst health, 100 = best health). Data for the models came from a longitudinal health survey on 1,942 participants. The health index was validated using age, frailty, post-survey one-year healthcare utilization and one-year mortality.
Results: The Spearman rho between the health index and age, frailty and post-survey one-year healthcare utilization were -0.571, -0.561 and -0.435, respectively, with all p<0.001. The area under the Receiver Operating Characteristic curve (AUROC) for post-survey one-year mortality was 0.930. An advantage of the health index is that it can be calculated using different sets of questions and the number of questions can be easily expanded.
Conclusion: The health index can be used at the individual, program, local, regional or national level to track the state of health of the population. When used together with the domain scores, it can identify regions with poor health and deficiencies within each of the five health domains.