A frustrating problem in controlling health spending and improving health care management is identifying those people most in need of such management. A number of strategies have been tried and whole companies have been set up around the premise that they have the magic algorithm for identifying high-cost patients. Anyone can obviously do that in hindsight, but retrospective analyses haven’t always yielded clear correlations to use for anticipating high spending in future years. And the more it costs to identify and manage supposed high-cost patients, the more money you need to save to justify those management costs. A new analysis in the journal Health Services Research attempts to advance the cause. (HSR Article) Looking at a Medicare fee-for-service population of 68 years old and older, the researchers examined number of chronic diseases, or morbidities, and the length of time a patient had the disease and attempted to identify relationships with spending. It is widely believed that chronic diseases are associated with high spending. They started with the full population of Medicare fee-for-service beneficiaries in 2015 and looked back as far as 1999 to identify all chronic diseases and the duration of the disease. The final study group was over 20 million people who had continuous enrollment in that 1999 to 2015 time period. 19 largely chronic conditions were tracked. Each patient was grouped by number of conditions and by duration of each condition.
The average age of the study cohort was 77, 56% were female and 11% were dual-eligibles in 2015. The most prevalent conditions were hypertension, 61.4% of the population, hyperlipidemia, 50.4%, arthritis, 33.4%, ischemic heart disease, 30.9% and diabetes 27.7%. 71.7% of the study group had two or more of the conditions and 17.3% had six or more. The cumulative average duration of all conditions was 23.6 years per person. There was a wide range of duration. Note that this duration is only for the time the person was enrolled in Medicare; the actual disease duration is probably much longer. Age, sex, race/ethnicity and Medicaid status collectively had little correlation with spending, less than 6% explanatory power. Number of chronic conditions predicted about 46% of spending, while cumulative duration predicted only 27%. This suggests that knowing how many diseases a person has tells you more about their likely future health spending than knowing how long they have had the diseases, but still isn’t hugely predictive, and frankly, is only consistent with common sense. The real trick to me is how to identify persons who are likely to have an acute exacerbation of a disease. Most chronic diseases can be managed with a somewhat high, but consistent, level of spending. Hypertension, for example, can be treated with generic drugs that are inexpensive. But the acute exacerbations can result in sudden episodes of very high spending. I haven’t seen a good technique to predicting those events.