First, maintaining a list of people with accurate demographics including contact information is still very hard in 2015. In addition to simple names, addresses and ages--all of which can change over time even with perfect data fidelity, we have to add a status to that list. Are you a patient of Dr. Schutzbank? Do you have Masshealth insurance? Since when? When does or did that change?
Outside of health care, we do this well and we call it Facebook. Everyone ensures that their information is updated regularly. However, Facebook is voluntary and the information is necessarily interesting to the poster. Sharing pictures of vacation may motivate me to log in, find the photo, upload it, tag it, etc. Updating my primary care doctor selection is just not that interesting to me. Furthermore, this information must be kept perfectly private and secure, so crowdsourcing solutions are a challenge. Lists of names remain hard, but fortunately perfection is not required for progress. Instead, we can define populations for different players in health care in descending order of ease.
Government & Public Health: Geography-based populations
If your foray into health care is through government and/or public health methods, your population is likely defined geographically. Cities, states, zip code ranges, whole countries will be in or out of your jurisdiction. This will cause all sorts of problems when neighboring geographies have radically different services and people start to move. Short of that, it is easy to at least theoretically know who is in your population.
Having a criteria that can be measured makes thing easier. Furthermore, geographic constraint (which is true regardless of who you are as all health care is local), makes it a little bit easier to design constrained interventions. If obesity is a problem and there is no place to walk in your town, you can make places to walk in your town (without worrying about making places to walk in every town). Not easy, but easier.
Payors of health care: Customer-based populations
If you are a health plan, insurance company, union, self-insured employer or some other purchaser of health care services, then you too can define your population. Someone, somewhere, somehow, is paying for your services. And that is your list. You have all of the list problems, and that people die, they lose benefits, paperwork has a lag, etc. Fundamentally though, with good list hygiene, your population is knowable.
Since the relationship here is customary rather than geographic, the interventions will look differently. The tend to focus on remote, scalable solutions using technology. Once you are not in the same place as your population, you could be anywhere in the world. Although attractive, there is a growing recognition at the importance of relationships in care and therefore a shift in intervention design.
Additionally, Payors arguably have the most difficult regulatory challenges. Public health and government officials have plenty of rules with which to contend, but it is the payors that have to guess if certain behaviors break the rules or not, often post hoc. Fear of regulatory violations, especially as organizations grow large (which all regulated organizations must do), can chill innovations in the name of fraud and waste prevention. Once we cross into care providers the game changes quite a bit. Although mired in regulation, the truth is that a license to practice medicine confers incredibly wide latitude in the treatment of patients, especially outside of facilities.
Specialist Care Providers: Condition-based populations
When it comes to specialty care, populations may not be fixed, but they can be predicted. The very advantage of specialization is that by limiting what comes through the door, you can meet the needs of your population. It is why primary care doctors claim specialists have it so easy. Additionally, you often have some geographic limitation to your specialty, although "catchment areas" can span several states based on disease prevalence.
Your population is guided by your conditions, and therefore your interventions are condition specific. If you are an endocrinologist specializing in thyroid disease, you better be setting up your interventions to be convenient and effective for the treatment of thyroid disease. You have an ultrasound in the office. You can biopsy right there and have the ability to prepare and view tissue. However, true innovation is dependent on a deep understanding of the diagnosis, treatment and ongoing management of your conditions. Specialists ought to be leading the way in keeping patients out of the office, but fee for service keeps getting in the way.
Primary Care Providers: Geographic, customer, condition-based populations
Primary care may very well have the hardest problem in defining population. The typical method is empanelment--a list of patients who are supposed to under care of a given doctor. This sounds great, but it is a particularly challenging instance of the list problem. Patients may "establish care," but do they continue? Do they have to break up with their primary doctor formally? If they change insurance, change jobs or move away does the primary care physician ever get notified? Formal empanelment via HMOs has been tried, but is arguably less effective and often rejected as too controlling. Attribution methods are used (patient saw Dr. X. for 2 visits and Dr. Y for 1 visit, therefore they are a Dr. X patient, even when the last visit was a "switch" to Dr. Y).
Most Primary Care practices that successfully solve this problem do it through payment reform of some kind-- HMO attribution, concierge and/or Direct Primary care customer relationship, health plan style enrollment/dis-enrollment, limiting eligibility to membership in certain organizations (employers, Medicare, etc.).
Even without knowing your population for sure, primary care offices can go a long way to determine prevalence of disease and geographic needs. Preparing for common conditions and community engagement outside the practice may be the only way practices can manage their population. It may mean that the measurement end is weak, that it is hard to prove interventions work, but healthy patients do not need you to prove that you helped them, only that they are helped.
Do not give up! There is a way to get your arms around your population somehow and doing so will allow you to become far more effective in caring for your patients.