Good days and bad days: Innovation in capturing data about the functional status of our patients

Authors


Introduction

Functional status represents an important component of patient assessment in arthritis care and research. Functional status data is relevant in making individual clinical decisions, judging the effectiveness of therapies, and justifying the importance of resource allocation to sustain an arthritis care infrastructure in our health care system. Arthritis impacts patients' lives in pervasive ways. Few if any aspects of an individual's functioning in daily life are immune from being affected by arthritis. Recognition of this clinically intuitive fact has motivated and shaped efforts to develop and implement functional status measurement in clinical and research settings (1, 2). Functional status measurement is now considered an essential and integral part of outcome measurement in rheumatology (3).

In a recent National Committee on Vital Health Statistics Report (4), functional status is viewed as a broad concept that encompasses an individual's capability to carry out various activities and participate in life situations and society. Consistent with this view, this article assumes the following definition of functional status: functional status is the state or capacity an individual possesses to perform, undertake, or participate in activities, simple or complex, that are regarded as essential to independent functioning. This definition could include basic physical activities of mobility, such as sitting, standing, or walking; cognitive activities, such as focusing attention and communicating; and more complex activities, such as utilizing transportation or maintaining a household.

Current methods of collecting data

For more than 2 decades, structured survey instruments have been the mainstay for collecting functional status data in arthritis care and research. Many of these instruments have been built around 2 core instruments, the Stanford Health Assessment Questionnaire (1) and the Arthritis Impact Measurement Scales (2). Others, such as the Western Ontario McMaster Osteoarthritis Questionnaire (5), were developed independently but cover many overlapping conceptual areas. Although data derived from patient self report have been shown to correlate with long range arthritis-related outcomes, the utility and usability of these instruments in detecting clinically meaningful change (6, 7) or their use to support decisions in routine clinical practice for individual patients is not clear. Although the necessity of capturing data about patient function is intuitively important in arthritis care, it is neither intuitive nor proven that structured questionnaires are the best, if not the only, way to collect functional status information. Research on the cognitive science of patient self report suggests that there may exist significant limitations encountered in the use of questionnaire instruments to collect data about functional status of our patients. Technological advances present new opportunities to capture data pertinent to the functional states of patients that are potentially richer, more valid representations of the patient's day-to-day experience, and thus are potentially more responsive to a broader range of clinical uses.

Time continuous signals

Our understanding and appreciation of a patient's state of function often comes from a variety of inputs. Hence, functional status can be regarded as a composite concept. In most patient self-report questionnaires, the composite notion of functional status is reflected in some combinatorial algorithm of the response items directed at the patient's ability to perform or participate in certain activities and tasks and the level of such constitutional symptoms as pain, fatigue, or depression. Because of the time variant nature of the factors that impact patient function, functional status fluctuates over time. Our patients often reveal the undulating nature of their functional status when they talk about “good days” and “bad days.” Actual changes in functional status from time point to time point and the relative numbers of discrete “good” days versus “bad” days may inform and motivate clinical decisions or judgments on the effectiveness of our therapeutic interventions differently than the aggregated or “averaged” recollections found in health assessment questionnaires. If we could assume there existed a perfect, comprehensive combinatorial algorithm for the various components and indicators of patient function that yielded a single universally applicable value of a patient's functional status, then we might find it useful to think of functional status as a time continuous biological signal, much like a blood pressure wave form or body temperature. Figure 1 models how the hypothesized time continuous tracing of this idealized functional status parameter might look. Modeling functional status as a single time continuous biologic signal like other vital signs (8) provides us a way to depict and reflect on the limits and the capabilities of methods that are available to us to capture data about functional status.

Figure 1.

True functional status time-continuous signal derived from multiple “momentary” assessments over small sampling interval and sampling frequency that preserves the inherent “truth” in the signal.

Time continuous signals are often referred to as analog signals (9). Speech, sound, and intraarterial pressure are examples of analog signals. The process of capturing a signal at a discrete point or interval in time and transforming it into a readable or computable form is called “analog to digital conversion” (10). This is usually a 3-step process. First, the signal must be sampled at discrete time intervals, called the sampling interval or the sampling frequency. Second, the signal at that sampling instant must be represented or translated into some quantifiable measurable phenomenon, a process called quantization. Finally, the quantized signal must be measured or digitized into a computable value. Electroencephalograms, electrocardiographs, and arterial pressure wave tracings are examples of analog signals that have been analyzed and processed into a form or representation that is easily assimilated into clinical decisions. The fidelity of an analog signal, or the ability of the 3-step process to preserve the essence of the underlying signal over time and represent it faithfully, is dependent on many factors. Errors can occur in the quantization process when a signal's true value is not reflected in the measurable quantity chosen over the range of possible values of the signal. Other frequent causes of error in the representation of a time continuous signal relate to the selected sampling intervals or sampling frequencies that result in portions of the signal to go unrecognized.

Autobiographical memory

If we chose to view functional status as such a time continuous signal, then how well do patient self-report questionnaires capture and represent its inherent information? The success of self-report questionnaires in reproducing or tracing with fidelity the true time continuous functional status signal of a patient depends on how accurately patient experience can be encoded at a given point in time and how well that encoded information can be preserved and recalled. The encoding and storage process of self-perception information is called autobiographical memory. Research in the cognitive science of autobiographical memory suggests there exist significant limitations in the encoding and preservation of information of patients' self perceptions over time (11).

A patient's encoding or storage of information about their experiences is influenced by their mental and cognitive state at the time of the experience and situational factors that may be in play at the time (12, 13). Hence, the quantization of their functional state at a given moment in time is influenced by a person's subjective state, the perceived relative importance of events, or distraction by other stimuli. Likewise the accuracy of past experience recall is not simply the opening of a saved file of prior experience tracing, such as the opening of file of unaltered data stored on computer hard disc drive. Patients cannot arithmetically average symptom intensity or occurrence frequency over specific time intervals (14). Rather, research indicates that patients use a series of highly individualized cognitive steps to reconstruct their experience (15), much like we do when we reconstruct our activities when searching for a lost item of which we knew we had possession at 1 point in time. Patients use these strategies to attempt to rebuild information about past experience, or in the context of the biologic signal model, attempt to retrace or reproduce their time continuous tracing of momentary experiences of functioning.

Bias

Research indicates that the encoding and heuristic retracing processes used by patients is heavily influenced by several biasing factors. State bias describes a type of bias in which certain experiences are recalled differently based on the state of mind of the patient. Hence, state biases (16) would tend to influence a patient who is in an angry or negative mood to recall negative experiences more prominently in a retrospective aggregation of experience. Events that are more recent to the time of assessment and those that are more salient are weighed differently than those occurring earlier or which are subjectively judged as less salient in the retrospective interval over which patients are asked to report. Theses biases are termed recency bias and saliency bias, respectively (17). Patients often reconstruct or recollect their experiences to satisfy a strong belief, judgment, or expectation (18, 19). This process has been termed “effort after meaning.” After a sentinel event, such as hip replacement surgery, patients often retrospectively overestimate their actual preoperative pain and difficulties (20). It is postulated that patients recall their experience to make sense or to ascribe meaning to the sentinel event. Hence, a patient might reason, “If I needed this operation I must have been doing poorly.”

Structured questionnaire instruments often contain instruction sets that might be ambiguous or erroneously interpreted. For example, does the question (21) “Over the past week how many days did you feel well” mean how many days over the past week did one feel well for any part of the day, or for how many days did one feel good for the entire day? Research has indicated that patients often misinterpret time intervals (22, 23) contained in questionnaire instructions. Patients often judge phrases such as “during the past month” differently. Hence, references to the past month on the 15th day of a given month can be interpreted as the time to the 15th day of the previous month, the 30-day interval previous to the first day of the current month, or the past 6 week period (24). Figure 2 depicts how biased recall can result in loss of information in a functional status time continuous signal.

Figure 2.

When a patient is asked to provide an aggregated summary of overall functional status over a period, they can only do so over that interval available in autobiographical memory. A large portion of the functional status curve has been lost and therefore is subject to biases of recall and individual patient memory heuristics. Lost also is the “shape” of the functional status signal curve reflecting the undulations or change in functional status over unit time intervals (e.g., day to day).

Two general strategies can eliminate some of the biases and errors in the quantization of the idealized functional status signal using self-report questionnaires. One strategy is to narrow the time interval for self report such that recall biases are eliminated or minimized and the momentary value of the time dependent variable is captured. The second strategy is to capture data about functional status in such a way that unobtrusively eliminates the need for burden in the patient self report itself. Indeed, in the landmark article describing validation of the Health Assessment Questionnaire self reporting difficulties, nurses performed visual real time validation of what was described by patients by personally observing them perform specific activities in their homes (1).

New methods of capturing data

Diary-based methods are 1 approach to capture momentary self-report measurements close to real time or over shorter time intervals to minimize distortions inherent in autobiographical recall (25). In an effort to collect patient self-report data in real time, some studies have employed paper diaries. However, paper diaries have been demonstrated to suffer several limitations as well. First, patient entries in paper diaries can be incoherent, illegible, or out of specified data ranges, often creating significant data cleaning obstacles. Compliance can be a significant problem with paper diaries (26, 27). Patients often forget to perform scheduled entries, hoard data cards, and attempt to complete them in batches to remain compliant (28). Hence, the patient self report once again becomes prey to the very retrospective recall biases that the diary approach was intended to avoid. The time-dependent change and undulation in the functional status signal can once again become lost.

Computerized diaries using handheld computer devices offer the potential to overcome many limitations of paper diaries (29). An individual can unobtrusively carry handheld devices throughout the course of the day with little interference in their routine. Data entered can be instantaneously validated and data entry constraints can be readily enforced to eliminate out of range data or back-filled data. By using electronic prompting, the opportunity to capture desired information at the instant or in the time frame that it is desired or relevant can be presented. Data entered as such can be immediately locked and time stamped and can be considered to most accurately reflect the patient's momentary state without being vulnerable to retrospective distortion. Data can be stored, transferred, or forwarded at desired or convenient times automatically into data warehouses that can perform analysis as data arrives. Such momentary data that is captured in the person's natural environment avoids influences on measurement from artificial reporting contexts, such as occurs in a doctor's office with blood pressure measurement (30). With data acquired and linked to multiple specific time points, a true discrete time continuous signal can be constructed based on real time and real events.

Programmatic branching or response-conditional nested assessments can be implemented in electronic hand-held devices to gather more information in certain specific situations or identified states. Such information may be inherently richer than time aggregated or averaged data and could be the key to recognizing subtle but real therapeutic efficacy trends that are not apparent in lumped experience assessment (31). This additional information may provide unique insights into a patient's clinical course or therapeutic response that are otherwise lost in aggregated retrospective data (32). Data about real time patient experience captured using hand-held computers could be easily coupled to other digitally acquired physiologic data (e.g., glucometer, blood pressure monitors, automated pulmonary flow meters, biomechanical information) to present functional status data integrated with various physiological parameters.

Methodologic approaches emphasizing real time, real life momentary patient experience have been used to gain unique insights into an array of health behavior phenomena, such as mood (33), chronic pain (34), rheumatic syndromes and conditions (35, 36), and sleep (37), and have been applied in young (38) and old (39) populations alike. Collectively, these methodologies aimed at real time data capture about the near immediate functional states of patients in their real environments have been coined under the term Ecological Momentary Assessment (40). Data obtained from momentary assessments by patients can have many relevant applications in rheumatology. Episodes of Raynaud's phenomenon, duration of morning stiffness, and time dosing relationships of symptoms suspected to be side effects of medication administration are a few clinical phenomena of potential interest to rheumatologists, of which momentary assessments may provide data that is more clinically valid and insightful.

Other technologies provide innovative opportunities to capture rich functional status data. Patient function and performance in activities and tasks can be captured using digital video. Digital cameras are rapidly becoming widely prevalent in the world of consumer electronics (41). Friends or family members could assist in video capture of patients performing activities of clinical interest or relevance. Such digital glances into patient function offer the advantage of acquiring a sample of functional status data that is rich and arguably of greater validity with respect to time and performance in the patient's actual living environment than aggregated retrospective recall. Such data can potentially be viewed not only in real time but stored, forwarded, and analyzed in data batches to assess the time course of improvement or deterioration. Digital video encoding of patient activity performance offers the opportunity for automated data manipulation, comparison, and analysis, such as time image sequence analysis (42), to obtain objective measures of patient performance.

Mobility is an important component of many functional status assessments pertinent to rheumatology. Passive devices exist that can unobtrusively collect real time data about the movement of patients in their home environments. Many devices, such as motion detectors and pedometers, are simple, inexpensive, and available in home hardware stores (43). More sophisticated devices, such as accelerometers and tilt meters, have shown potential to detect and electronically record mobility-related events, such as falls, rate of position change, and time periods during which a patient is active (44, 45). Data from such devices can be directly captured by transceivers, digitally encoded, collected, and downloaded into a computer database for analysis over any time interval of interest. As high bandwidth Internet connections increasingly are used in households, there will be innovative opportunities for electronic capture and integration of rich functional status data into a multimedia medical record (46).

Conclusion

In reality, functional status is a complex concept with many potential inputs (47), but modeling that complexity as a single time continuous biologic signal enables one to appreciate the limitations of established paper-based methods for functional status assessment, and the potential in developing, expanding, and applying technologically innovative methods for sampling over time data relevant to functional status. Do relative numbers of discrete “good days” versus “bad days” not captured in aggregated retrospective recall have clinical meaning? What are the clinically relevant events, factors, or occurrences that make a “good day” good or a “bad day” bad? Although retrospective patient self reports can continue to contribute to functional status assessment, their limitations and the emerging opportunities to capture richer continuous time discrete data provided by new technology should stimulate exploration of new applications and methods for functional status assessment in arthritis care and research.

Ancillary