As opening testimony to his vexation, he quoted directly from one actuarial analysis:
'….cannot state which trend assumptions reveal the true program effect. The choice of trend has a large impact on estimates of financial savings.' (bolding from the DMCB)
Al Lewis argues this ‘choice’ is a canard: savings are or they aren’t. He believes the business-as-usual disease management pre-post analyses being foisted on unsuspecting purchasers miserably fail to account for the persons with chronic illness that are missing in the baseline. Missing claims can be due to any number of problems, including not having any encounters but having disease, or having too few claims, or not having the right 'kind' of insurance claims. In Al’s experience, both of the competing approaches of a) 'prospective' identification (once chronic, always chronic and counted over the entire period of observation) and b) 'annual' identification (must be requalified with relevant claims every year) are vulnerable to this shortcoming. After showing some simplified examples in which both approaches led to inflated savings because of missed claims and regression to the mean, Al argued that the better method is to assess total event rates over time across a population while comparing them to valid national sample.
Al’s dyspepsia reached its zenith when he examined the issue of ‘innumeracy.’ He had examples of huge savings being touted with little understanding of the usual or reasonable baseline measures typical among persons with chronic illness, or the role of cost drivers like hospitalizations. Using coronary artery disease management programs as an example, Al bascially recommended that you keep it simple: measure the baseline heart attack rate over several years and compare it to a simultaneous national mean. If the DM program is successful, heart attack rates should comparatively decline over time.
Co-presenter Roberta Herman MD of Harvard Pilgrim Health Care agreed with Al. Sharing numbers from HPHC's disease management programs, she showed there was gratifying drop in asthma ER/hospital rates over the last 7 years. However, the DMPC method demonstrated that the program was far less effective when it was compared to national baseline. In contrast, while diabetes appeared to flatline, there was a concurrent decline in heart attack rates that bettered national statistics and seemed to contribute to Harvard Pilgrim’s NCQA success.
Impressions from the DMCB:
Maybe the fancy actuarially-focused disease management trend reconciliations are too complicated for a lot of folks in the disease management industry. Al’s method clearly resonated with the audience, who were skeptical about the vendors’ conflict of interest in running a disease management program and being simultaneously responsible for assessing the return on investment.
Roberta Hughes also pointed out that C suite leaders also need the mumbo-jumbo free simplicity of overall vs. national rates, especially if the skeptical actuaries over in the CFO's office are telling them that the disease management programs are having no impact on overall trend.
When asked, Al doesn’t have an issue with the impact of disease management programs. Rather, he thinks the DMPC measurement methodology is a better way to assess the degree of the benefit. The DMCB asked him about the option of using multiple measurement methodologies to “triangulate” on the return on investment from several vantage points. He agreed that the DPMC approach could complement other analyses – assuming they’re all in the same ball park.