Editor’s note: This article is adapted from the author’s book, Roadmaps to Value-based Profitability: A Practice Transformation Guide.
The use of data to better inform care is at the core of population health. Managing population health starts with looking at the common characteristics of your patients. Developing a strong understanding of your patients is the first step to impacting overall care in terms of cost and quality — two goals of the Triple Aim. Every practice should be moving toward population health management as part of transformation efforts.
Population health is defined as the health outcomes of a group of individuals, including the distribution of such outcomes within the group.1 Health outcomes are defined as more than the absence of disease. Multiple influences impact population health,2 including:
- Access to care
- Understanding of health
- Prevention and management of chronic conditions
- Socioeconomic factors, such as income, availability of food and other social determinants of health (SDoH).3
Population health management involves a continuous cycle of six steps:
- Set goals.
- Identify interventions.
- Engage patients.
- Coordinate care.
- Measure and assess performance.
- Using information from Step 5, return to Step 1.
The goal of population health management is to improve overall health to influence the surrounding community. With a full ripple effect, health improvement spreads beyond the local community to a broad range of populations.
The starting point of population health is understanding the characteristics of the patients you serve. By analyzing various data points, you will develop an understanding of not only the medical conditions being experienced by your patients but also their issues and concerns, which could be impeding them from maintaining or improving their health. By understanding a broad picture of your patient panel, you can start to assess how your practice can meet the overall needs of your patients.
Start by looking at what your patients have in common. Understanding the bigger picture can help with directing resources to structure staffing, to identify what educational information to make available, and to decide which clinical guidelines to implement. By gearing your practice to address common concerns and medical conditions, your providers and clinical team will have a solid foundation for tailoring care to individual patient needs.
Getting to know your patient population
To prepare for accepting value-based payment models, you must understand whom you serve. When considering your patient panel, what characteristics come to mind? Maybe the first thoughts are based on demographics, such as age and race, or possibly certain diagnoses or conditions most commonly treated in the practice. Gathering ideas from multiple people in the practice and considering what data is currently available can inform what makes the most sense for your practice. How the patient panel is described may vary across practice sites and by individual provider.
However, with a common way of looking at patients, comparisons between providers and practice sites can lead to identifying variations in practice patterns. This type of analysis is not intended to create a cookie cutter approach in defining how to deliver care, but rather to allow for a deeper understanding of a standardized approach to care based on common characteristics. This generalized approach is a starting point which is then tailored appropriately. Before addressing available data, understanding some basics about the data stored in the information systems is helpful.
Crucial to any effort is ensuring you have structured data that uses standardized formats for defined purposes. In simple terms, structured data represents information that can be classified as yes/no, date or alphanumeric. Structured data allows for only the defined data to be entered into a field through either validation rules (i.e., not allowing a date to be entered if numbers are not used and the specific format is not followed) or the use of drop-down boxes to control potential answers.
Unstructured data is not easily organized in pre-defined structures and is more challenging for reporting. Within the EHR, a text box in which you can write a note is an example of unstructured data. To use unstructured data, the logic of the program to generate a report would need to interpret what is written in the note. While this can be done, the reliability of the data is limited.
To illustrate the difference in how structured and unstructured data are used in creating a report, imagine you are running a report assessing whether a copay was collected. If the copay information was written in a text box, you might be able to look for the word “copay,” in the hopes that everyone enters the word “copay” before adding the dollar amount. Depending on how the logic is created for that word, you may find information about copays, but creating a calculable report with dollar amounts would be difficult. The data from an unstructured field cannot be aggregated or tabulated, resulting in information that is hard to manipulate. However, if the copay is entered directly into a field labeled “copay” and requires entry as a number or dollar amount, a report can be created based on this structured data. The report can assess any formula based on the copay such as sum or average.
Practices may encounter difficulties when there are fields with similar descriptions and the source of the data is unclear. It can be common for IT or reporting staff to develop the report without understanding where data is captured by the end user. This can lead to frustration between both parties, as the point of reference for looking at the data is different. Those developing the reports are looking at the data tables where information is stored, while the end user will only know where the data appears on a specific screen. If the two are disconnected, you may run a report using a specific data field from a data table, but the data will not populate correctly. This can occur when there is more than one field for a data element, as a result of more than one screen where the end user enters the data and unclear direction about data entry. These sorts of discrepancies are often found when new reports are generated.
In an ideal situation, there should be only one location to record a discrete data point. Unfortunately, that is not always the case. For example, in the early stages of Meaningful Use (MU) data reporting, practices would find that the data did not match what the providers intuitively knew from entering data during office visits. To identify why the data was showing low results, the logic of the report had to be examined. The report was pulling from one field that had an area for recording the blood pressure (BP) reading. However, in practice, the BP reading was being recorded on a different screen, which was captured in a different data field.
To get the most accurate results, either everyone needed clear directions as to where the BP reading must be recorded so that data was collected in one location, or, the less desirable alternative, the report needed to be modified to look in the two different locations. The risk with the report looking in two locations is that data could be double counted or incorrectly represented. These types of issues should be kept in mind as you review what data is captured and in which screens. This will be valuable information to convey when creating reports to characterize your patient population and other reports needed for gaining deeper insights about whom you serve.
Starting with the easy stuff
Within the data you capture to bill for services is a host of information you can use to understand the basic characteristics of your patient population. At a minimum, to bill for a service, you need to verify insurance coverage if applicable, the billable procedure code, the diagnosis, and the date of service. Payers use this information to generate sophisticated analytics about members served. This same data is readily available within your practice — along with additional data not necessarily available to payers. Understanding how to use this information is important in practicing value-based care.
As part of enrolling a new patient into a practice — or updating existing records — the practice gathers demographics such as age, gender, insurance and contact information. With the introduction of MU, many practices expanded the data collection to gather even more information. Additionally, within each patient record is current data, such as vitals and lab results, information that may not be accessible to the payer. The lab tests and vitals can be leading indicators for individuals at risk of developing a chronic condition or experiencing progression or regression of an existing condition. Using this data to identify individuals who need more proactive management for prevention is a significant way to avoid future costs. Disrupting the rate of healthcare spending is essential for population health management, which is tied to more advanced forms of value-based payments such as gain and risk-sharing arrangements.
By its definition, demographic data includes statistical data that describes some socioeconomic characteristics of a given population. Previously, within the medical community, the demographic data collected often did not include information about income level and education, which are two socioeconomic factors. With the emphasis on addressing how SDoH affect overall health and outcomes, collecting and using this data helps to refine your understanding of your patients. The importance of collecting information about socioeconomic factors is gaining traction. Expanding the collection of basic information at patient registration, with regular verification and updates, can assist your practice to gain deeper knowledge of your patients and better anticipate and meet their needs.
With the advent of EHRs, capturing basic demographic information has become centralized. Many providers took advantage of incentives to implement EHRs after the Centers for Medicare & Medicaid Services (CMS) introduced the Electronic Health Records Incentive Program in 2011.
Examples of those early demographic data capture areas include:
- Patient date of birth, collected in MM/DD/YYYY format
- Sex or gender
- Race, defined by categories published by the Office of Management and Budget (OMB)4
- Ethnicity, also defined by OMB-published categories
- Preferred language.
Other general data collected include home address, telephone number and email address. Rules within EHRs often prevent users from entering the wrong number of characters for certain fields (e.g., entering “18” instead of “2018” for year); such data validation techniques help prevent invalid data from being entered. Another example is preventing the entry of a character in a field where only numerals are used, such as entering a ZIP code. These rules help promote the integrity of the data.
For reporting and data analysis, using structured data fields that have specific data validation requirements provides the most flexibility. This basic demographic data allows for segmenting the patient population and is particularly useful when assessing for disparities in care.
Diagnosis and procedure codes
In capturing information about services delivered, at least one diagnosis is identified, as well as at least one procedure code. With the introduction of ICD-10, the diagnosis code can have a level of granular specificity not previously captured. The diagnosis code can be used as a key data element to examine how many patients have a certain diagnosis — or aggregate by diagnostic group. By identifying multiple diagnosis codes within a single visit, multiple issues can be documented.
Further analysis based on the procedure code billed and the frequency of visits provides additional visibility on the complexity of the patient’s needs. Insurance companies use this same type of data to analyze provider performance and routinely review diagnoses and procedure codes submitted on claims or encounters. Being able to think like a payer, including understanding whom you serve and trends in utilization and cost, will help maximize a practice’s profit. When analyzed, this information can provide insight on how to impact utilization, cost and health outcomes. Understanding and implementing these controls are important for succeeding in negotiating additional payments or higher reimbursement under value-based payment methodologies.
Extracting the data from your billing system is a great starting place for identifying patient characteristics. If you can further combine the billing data with clinical data such as blood pressure readings, A1c levels, or other relevant tests for a specific diagnosis, you start to create a richer picture of your patient population. This involves taking information you may currently have set up as a registry or report and taking it one step further. For example, you may already have a report that shows patients who have poor control of blood sugar with an A1c level greater than 9. If you were able to understand more about the patients who have these higher levels, by analyzing additional data points, you may see if there are patterns that suggest a need for a new approach to engaging patients or understanding additional risks.
Information that you should have readily available includes:
- Type of insurance
- How often patients are seen
- Types of visits (procedure code)
- Frequency of reschedules or no-shows.
Additionally, if you can capture information about emergency department visits and inpatient admissions, you can further identify patterns that may indicate the need for a different approach to engaging the patient or potentially a need to recommend additional support.
General health outcomes and factors
Another source for gathering information about what your patient may be experiencing is to look at general health outcomes and health factors. By knowing generalities of the local community, you may find that certain educational information or referral community support are needed, even if the patient does not discuss these needs. General characteristics for every county can be found at countyhealthrankings.org. This data is collected from multiple sources, including the Centers for Disease Control and Prevention (CDC). Each county within a state is ranked for health outcomes based on two measures: how long people live and how healthy people feel while alive.
For the ranking by health factors, measures include health behaviors; clinical care; and social, economic, and physical environment factors. This also provides a potential benchmark for comparing statistics and characteristics of your patient population. For example, if you are tracking the number of adult patients with a BMI over 30 and classified as obese, you can compare the percentage of patients in your practice to the overall number within your county. If it is significantly different, then you may have an underlying issue within your practice. By focusing on weight control, you may have a positive impact on the patient and the surrounding community as desired by the end goal of population health.
Finding out what you know
Whether you are tracking BMI, A1c levels, or any other measure, when evaluating the characteristics of your patient population, you need to know what information is available, where it is stored in the system, how to retrieve relevant data and have the ability to interpret it. Gathering information about what data is collected in your EHR and practice management systems is a good starting point. This information combined with data about socioeconomic factors can provide a deeper understanding of your patient population as your practice moves toward managing population health.
To hear more from Jennifer Ternay about making the shift to value-based payment models, listen to her MGMA podcast.
- Kindig D, Stoddart G. “What is Population Health?” American Journal of Public Health, October 10, 2011, ajph.aphapublications.org/doi/10.2105/AJPH.93.3.380.
- OMB. “Directive No. 15, Race and Ethnic Standards for Federal Statistics and Administrative Reporting (as adopted on May 12, 1977).” Available from: bit.ly/38Trh1F.