By Terri Lewis, Ph.D.
Little talked about, the quality and social indicators of care are important factors to evaluate when determining whether care for complex illnesses is adequate. The public discussion is consumed by continuous attention placed on dose, days of dosing, and pill counting along with denial of care and increasing anxiety about increasing numbers of overdoses and suicide. But is this enough to understand the entire conversation in context? I think not.
So far, 3170 or so persons have contributed their experience to this national survey. I have taken two approaches to examination. First, to document the qualitative responses you have offered and second, to evaluate the contextual open-ended responses to each question. The first is easy. But to really understand and interpret the first set of questions, one has to understand the contextual elements offered by the open-ended questions. That is not so easy and it takes lots of time.
It also requires some framework building to find a way to construct this context so that it aligns with other data mechanisms that are used to communicate to the public about this public health issue of chronic pain. Accordingly, I have adopted a method for interpreting the open ended data so that it can be compared to the data sets that are captured by electronic health records and billing data for Medicare, Medicaid and other forms of insurance like VA, Tricare, and other private health information systems. All of these are built against a framework of the ICD-10, a universally accepted data coding system promoted for use by the World Health Organization. The ICD-10 is a component of the International Classification of Functioning (ICF); both are found on these links:
The ICD-10 is fully explained at this link:
All electronic health datasets are constructed on this framework but all electronic health records systems do not perform in exactly the same way – there are functional differences between the tools adopted by various physician offices and other health providers. The coding sets of the ICD-10 are universal to HHS-CMS and deployed through all contract systems so it makes sense to use them here, even though it is time consuming and complex. It is also offering a word-rich environment in which to evaluate the responses of those who have participated in this exercise. After all, there are only so many ways to describe some things, right?
Generally, respondents meet the criteria for multiple chronic comorbid conditions – that is they have two or more health care problems that represent diseases present from birth, acquired after birth or as the result of injury or traumatic events. Diseases or injuries failed to cure on their own, and symptoms became progressive with deterioration over time, often accompanied by the onset of additional symptoms. The ICD-10 codes against multiple body systems, so I am coding your responses based on diagnostic codes that closely match (as far as I can determine based on your information and descriptions) definitions of illness or injury that are known and described within the ICD-10 universal diagnosis body system code sets. Why is this helpful?
Everyone is consumed by the conversation about chronic pain and we argue about its contribution to impairment in terms that establish the value of disease or progression of illness against terms associated with chronicity – acute, chronic, intractable. Despite the fact that underlying real diseases are responsible for pain generation and the degree to which it lingers, we don’t give enough attention to the impact of disease and its course on body systems, adaptive functioning, pain generation, and maintenance or loss of capacity. Separating the use of ICD-10 from the data collection context of the ICF compounds this problem because it removes contextual information from the patient report. This looks like a problem to me as a person focused on the rehabilitative aspects of chronic care from a whole person/body perspective. After all, human interaction occurs within a system that includes the person, the family or friends and the community, each of which places demands that require responses.
If we are just focused on ‘pain’ (which is often what brings us into contact with care through the emergency room or the doctor’s office) it creates a problem because of something called ‘anchor effect’ and ‘confirmation biases.’ Anchor effect causes us to focus on a single aspect of disease management while interpreting all of the information available to us through a narrow lens. This is often the problem of specialized pain care. Another way of looking at this is ‘if one has a hammer then everything is a nail.’ Confirmation bias tends to focus our attention on whether or not our interventions work and which ones are most effective. Cure is measured by declining interactions.
In a world where opioids have become demonized and MMEs, days of dosing, and volume of prescriptions are our measures of effectiveness, we interpret the adequacy of our approaches through pain volume and frequency data – less is more, more is less, up is down, and down is up. Fewer prescriptions should yield fewer overdoses, and fewer MMEs should yield increased health. We know through our own experiences that this is not happening. Occam’s Razor is illustrative of confirmation bias – the simplest solution should generally be the best. Step therapies are built around the notion of these two heuristics working in tandem.
The problem is that these biases, built into the algorithms of electronic health records and PDMPs or other systems, rely on metrics of frequency or size, separate the context from the transactions, and don’t address the outcomes associated with these assumptions and heuristics. And they don’t get tested for their influence on more distal indicators and outcomes important for evaluating patient quality of care. The data that is collected and reported through billing records is exactly that – billing and payment data, which reflects what physicians bill for, what they can get paid for, and which may not necessarily be reflective of your experience as a patient or whether your transactions with the health care system are yielding transformational outcomes that maintain or improve your overall care. This problem is magnified by the reduction of influence of primary care and the increased influence of specialized care which increases the influence of anchor effect.
The Use of Big Data Tools
Yet another problem in all of this is that the focus on billing codes and the increasingly limited time available to the physician/patient exchange in our increasingly strained system causes us to short change the collection of valuable contextual information that tells us some things we need to understand about the effectiveness of individual patient care protocols. Public health data is built (by and large) on billing datasets. These billing data sets do not represent anything more than the bell curve of the population that interacts with the health care system. It leaves out the uninsured, the underinsured, the infrequent user, and those who have become so frustrated that they have abandoned the health care system. In general, public health data isn’t about you or your personal health care outcomes.
With the approaching deadline for implementing the CMS Final Rule for 2019 physicians will have a requirement to facilitate disability utilization reviews for persons who receive opioids at more than 90 MME. Between 90 MME and 200 MME, insurers will be allowed to determine whether some user groups will get a ‘pass’ based on their combinations of illnesses, injuries, disabilities and risk levels. Above 200 MME, CMS has created an appeal process. NONE of these determinations can be constructed on billing data frameworks. Whether one is benefited by a particular dose cannot be determined by the length of time one has been receiving it, or the MMEs. It all requires the integration of diagnosis codes with prescribing and intervention history, together with information about quality of care, quality of life and patient choice. So I am working to pick out these details by using all of the features of the ICD-10 and ICF that are available.
The Survey Analysis Process to Date
I am coding reported diagnoses, and into each section, the diagnostic categories for the pain symptoms reported with each primary diagnosis code. Coding is based on the specificity of the patient description and where it is not possible to extract anything more than a global term, the referent is coded as ‘unspecified.’ Where appropriate, I am adding links to known rare disease categories as found in OrphaNet, National Organization for Rare Diseases, and NIH’s GARD. This is important contextual information. So far, it looks like about half of the reported diagnoses fall into the category of rare diseases for which there is no known cure, only symptom management. A number of these have pain as the leading symptom to be managed throughout the progressive course of the disease process. Links to these lookup tools are found here:
NIH Genetic and Rare Disease Database (GARD)
It is important to note that this data has limitations. It is self-reported and I have not examined medical records to verify the reports. It has not been submitted under controlled conditions but rather as a snowball survey in social media using precautions that prevent duplication or data sharing. It is being subjected to close examination and it will serve as a basis for future examinations under controlled conditions.
On the following chart you see the ICD-10 coding system reference, the patient reported codes, the rare disease link if applicable, and the number of patient comment records across the survey that are associated with specific diagnosis codes. Examples within each category look like this:
A summary of the entire categories of the ICD-10 as reported by patients and extracted looks like this so far:
How We Speak About Pain Matters
Clearly, the terms acute, chronic and intractable pain are nonspecific. Based on what I am seeing develop in this process, I am not certain that they properly describe patients, the implications of disease course, or inform treatment, and it may be that they foreclose our thinking due to anchor and confirmation biases. And, given the terms in the table that follows as Table 2, I wonder if their use may actually serve to distract the clinician from properly conceptualizing treatment based on questions asked of and reported by patients. How we talk about this matters. There’s a lot more to look at here.
So here’s a little something for you to think about: How does the FACES Pain Rating Scale interact with the range of terms as reported by patients and extracted from comments?
If you wish to contribute to this ongoing process, please complete the survey at this link:
And if you want more information about the kinds of demographics emerging through this collection process, check out the link at this location:
A huge thank you to all who are contributing!