I am known in the disease management and wellness fields as a naysayer, critic, curmudgeon, and/or traitor…and those are only the nouns that are allowed to be blogged across state lines. This is because I am driven not by wishful thinking but rather by data. The data usually goes the wrong way, and all I do is write down what happened. Then the vendors blame me for being negative — sort of like blaming the thermometer because the room is too hot — because they can’t execute a program.
However, the nonprofit Iowa Chronic Care Consortium (ICCC) apparently can execute a program. They reduced total diabetes events by 6% in the rural counties they targeted. This success supports a hypothesis that in rural (presumably underserved) areas, disease management fulfills a critical clinical gap: it provides enough basic support that otherwise would not be provided even to those who actively seek it to reduce near-term complications and exacerbations.
This result will likely produce its own unanticipated consequence: because many people now believe (thanks, ironically, to some of my own past work) that disease management doesn’t produce savings, there will be widespread skepticism about the validity of this study. Quite the opposite: this “natural experiment” is as close to pristine as one could hope for in population health, for five reasons:
- There was no participation/self-selection bias because outcomes were measured on all Iowa Medicaid members.
- The program was offered in some Iowa counties but not others, so there was no eligibility or benefits design bias, Medicaid being a statewide program.
- The program encompassed only one chronic condition (diabetes) rather than all five common chronic conditions normally managed together (asthma, CAD, CHF, and COPD being the other four). Since all five conditions were tracked concurrently, whatever confounders affected the event rate in one of those conditions should have affected all of them. And event rates in the four other conditions did indeed move together in both the control and study counties. Just not diabetes.
- The data was collected exactly the same manner by the same (unaffiliated) analysts using exactly the same database so there is no inter-rater reliability issue.
- Both groups contained hundreds of thousands of person-years and thousands of events.
As one who has reviewed another high-profile “natural experiment,” North Carolina Medicaid, and found that the financial outcomes were the reverse of what the state’s consultants originally claimed (incorrectly, as they later acknowledged by changing their answer), I can also say that natural experiments in population health don’t harbor some as-yet-unidentified confounder that causes the study population to outperform the control population.
