The Development of a Proof-of-Concept Physician-Driven Informatics Consult System for the Individualized Treatment of Patients with Orthopaedic Conditions

doi:10.21203/rs.3.rs-2117565/v1

Download PDF

Research Article

The Development of a Proof-of-Concept Physician-Driven Informatics Consult System for the Individualized Treatment of Patients with Orthopaedic Conditions

https://doi.org/10.21203/rs.3.rs-2117565/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background: The diversity and variation of patient data now increasingly documented in Electronic Health Record (EHR) systems can be used to generate evidence about which treatments work and for whom they work. The objective of this work was to demonstrate proof-of-concept of a physician-driven informatics consult system which can drive improvements in the quality and safety of orthopaedic care. The informatics consult system allows for physicians to perform dynamic searches and receive immediate data summaries of similar, previously treated patients at the point of care. As a first use case, we apply our informatics consult prototype to the treatment of acute proximal humerus fractures (PHF).

Methods: We used an EHR-based, simulated patient data file to develop the informatics consult system prototype, ICSCEDIT (Informatics Consult System for Comparative Effectiveness Data for Individualized Treatment). We developed a user interface for physicians to define similar patient groups or reference classes, and we use comparative effectiveness research methodology and visual analytics to summarize outcomes for patients under treatment alternatives. Candidate wireframes were reviewed by the research and clinical team members in iterative user experience phases, each phase employing 1) prototype design and demonstration, 2) collection and documentation of expert/user feedback, and 3) prototype redesign based on feedback.

Results: ICSCEDIT was designed with a series of patient characteristic filters that can be applied to subset historical patient records into similar cohorts matching the index patient of interest. A total of 14 patient filters across demographic, fracture characteristics, and patient comorbidities are included. The evidence for patient reference classes, displayed by treatment group, appears in a series of three tabs. The first tab displays treatment rates of initial surgery and conservative management for the reference class. The second tab contains summaries of patient-reported outcome measures presented by treatment group and the third tab displays summaries of healthcare utilization over the year following the PHF injury.

Conclusion: We describe a prototype of an informatics consult system for orthopaedic medicine. Informatics consult systems that enable learning from historical treatment decisions and outcomes holds great potential for improving the quality of care for all orthopaedic patients.

orthopaedic treatment

clinical decision support

proximal humerus fracture

informatics consult

personalized medicine

Evidence-based medicine has achieved many successes since its popularization in the 1990s,^{(1, 2)} though many clinical questions remain unanswered.^{(1, 3)} Randomized clinical trials (RCTs) have stringent inclusion criteria which limit the applicability of results to the diverse patients treated in real-world practice.^(4–6) Furthermore, many clinical studies report average estimates which can mask valuable insights on treatment effect heterogeneity, which is also described as “patient diversity in responsiveness, vulnerability to adverse events, and utility for different outcomes” stemming from treatment.^{(7, 8)} The diversity and variation of patient and treatment data outside of classical study populations, now increasingly documented in Electronic Health Record (EHR) systems, represents an incredibly rich data source which can be used to generate evidence about which treatments work, and for whom those treatments work in the real-world.^(9–14) The United States Food and Drug Administration, among other agencies, is increasingly recognizing the role EHR data and real-world evidence can play in answering questions surrounding treatment choice.⁽¹⁵⁾ Medical specialties like orthopaedic medicine, which have variation in pathology and patient populations and lack treatment consensus and the funding necessary for large clinical trials,^{(16, 17)} are particularly ripe for using real-world data to inform individual treatment decisions.

Learning from historical patient data stored in EHR systems to assist current clinical decision-making was first conceptualized by researchers at Stanford and referred to as an informatics consult service.^{(13, 14, 18, 19)} Informatics consult services are designed to provide data summaries to help physicians and patients make data-informed treatment choices for and with their patients. Initial informatics consult service use, still early in development and applicability, have mostly focused on questions of prognosis and not treatment choice.⁽¹⁹⁾ Furthermore, informatics consult services currently employ a team of researchers and data scientists to field requests and return responses to physicians within a number of days. We believe that we can build upon the informatic consult service concept^{(19, 20)} to develop an informatics consult system to allow for physicians to perform dynamic searches and receive immediate data summaries at the point of care when treatment decisions are being made. Therefore, our objective is to demonstrate proof-of-concept of a physician-driven informatics consult system that has the key design and functional elements that will maximize usability, and drive improvements in the quality and safety of orthopaedic care. We refer to this system as the Informatics Consult System for Comparative Effectiveness Decisions for Individualized Treatment (ICSCEDIT, pronounced “I said it”).

Within ICSCEDIT, we develop a user interface for physicians to define similar patient reference classes, and we use comparative effectiveness research (CER)-based approaches to summarize outcomes for similar patients under treatment alternatives. With traditional CER approaches, average treatment effect estimates are generated for a population subgroup or a “reference class” based on a combination of measured baseline factors specified prior to estimation.^(21–23) Specifying appropriate reference classes is recognized as a significant barrier to translating treatment effects for individual patients.⁽²³⁾ Even with a small number of measured baseline factors, an individual patient could be placed in an indefinite number of reference classes,^(21–23) and it is often unclear which reference class is best aligned to each patient.^{(7, 13, 21, 23–25)} Physicians possess accumulated clinical wisdom on the patient characteristics that are relevant when making treatment decisions.^{19, 51, 52} Because of this unique knowledge, we believe physicians are best positioned to define reference classes of similar patients. ICSCEDIT takes a physician-driven approach to refence class selection. We believe a ICSCEDIT has the underpinnings of an innovative informatics consult system that will initiate a paradigm shift in how orthopaedic medicine is practiced in the future.

In this paper we demonstrate proof-of-concept for ICSCEDIT in the context of treating patients with acute proximal humerus fractures (PHF). The management of PHF continues to demonstrate significant variation within the orthopaedic community.^(26–30) There is a lack of evidence to inform the choice between initial surgical options or conservative care for many patients with PHF.^{(31, 32)} Recent RCTs suggest that for certain groups of PHF patients, there is no significant difference in pain and functional outcomes between surgical treatment compared to conservative care, which is almost certainly an instance of RCT masking treatment effect heterogeneity.^{(27, 28)} Further, meta-analyses across the trials seem to validate this concern, suggesting the comparative effect of surgery versus conservative care is likely heterogenous across patients.⁽²⁹⁾ Surgical treatment is more costly, increases the risk of surgical complications and mortality,⁽³³⁾ and has not been shown in clinical studies to produce superior pain reduction or functional outcomes for patients.^{(28, 34)} Yet, surgery continues to be used at high rates in many regions of the country.^{(35, 36)} It is unclear whether current surgery rates for many subsets of patients with PHF represent over or underutilization, with the key question not being whether surgery or conservative management is “the” best treatment for all patients with PHF, but rather which treatment is best-suited for each patient.^(37–43) Evidence on the effects of alternative treatments for similar patients or “patients like me”⁽¹⁸⁾ is needed to help answer how each unique patient with PHF should be treated.

ICSCEDIT Development Process

The development process for ICSCEDIT began by working closely with orthopaedic surgeons to identify the patient characteristics that were important for reference class identification and specifying the outcome measures that are important to physicians and patients when making treatment decisions. Patient characteristics theorized to modify treatment effectiveness were identified from literature review, clinical consultation, and preliminary data analysis and were used as input criteria for reference class selection. Treatment comparator groups were defined as patients undergoing initial conservative management or patients receiving initial surgical treatment. ICSCEDIT uses an episode-based analytic approach that defines an initial treatment window of 30-days, therefore patients receiving surgery in the 30 days following a new PHF injury would be classified as receiving initial surgery. Treatment outcomes included shoulder forward flexion, patient-reported pain scores, shoulder-specific patient-reported outcome measures (PROMs), and measures of treatment success generated from narrative clinical notes. Patient data needed to perform reference class selection, isolate treatment group comparisons, and provide data summaries of outcomes were mapped to routinely collected data elements or proxies found in EHR systems. Table 1 contains patient characteristics used for reference class identification, treatment group comparisons and outcomes, and the location and format of each data element in the EHR.

Table 1

Mapping of Patient-level Reference Class, Treatment and Outcome Concepts to Available EHR Data
Data Categories	Data Variables	EHR Location	Data Format
Patient demographics
	Age	Demographics section of clinical chart	Structured
	Sex	Demographics section of clinical chart	Structured
Fracture Characteristics
	Fracture of dominant shoulder	HPI section of clinical chart	Unstructured
	Displacement status	Imaging study result or HPI of clinical chart	Unstructured
	Location of fracture	Imaging study result or HPI of clinical chart	Unstructured
	Number of fracture parts	Imaging study result or HPI of clinical chart	Unstructured
	Amount of Displacement	Imaging study result or HPI of clinical chart	Unstructured
	Presence of concurrent fracture(s)	HPI section of clinical chart and/or concurrent diagnoses codes listed	Unstructured or structured
Surgical risk factors
	History of falls	HPI section of clinical chart, Problem list in EHR, or previous encounters	Unstructured or structured
	Diabetes	HPI section of clinical chart, Problem list in EHR, or previous encounters	Unstructured or structured
	Dementia	HPI section of clinical chart, Problem list in EHR, or previous encounters	Unstructured or structured
	Osteoporosis	HPI section of clinical chart, Problem list in EHR, or previous encounters	Unstructured or structured
	Lives at home	HPI section of clinical chart	Unstructured
	Medical Frailty	HPI section of clinical chart, Problem list in EHR, or previous encounters	Unstructured or structured
	Shoulder Osteoarthritis	HPI section of clinical chart, Problem list in EHR, or previous encounters	Unstructured or structured
Treatment Comparators
	Conservative management	Procedure Codes/encounters in EHR, or Treatment plan section of clinical chart	Structured
	Surgical treatment	Procedure Codes/encounters in EHR, or Treatment plan section of clinical chart	Structured
Treatment Outcomes
	Visual analog pain score	Clinical evaluation section of the clinical chart	Unstructured or structured
	Constant score	Various locations depending on data collection workflows and integration	Unstructured or structured
	SANE	Various locations depending on data collection workflows and integration	Unstructured or structured
	PROMIS physical function score	Various locations depending on data collection workflows and integration	Unstructured or structured
	Forward flexion	Clinical evaluation section of the clinical chart	Unstructured or structured
Healthcare utilization
	Number of follow-up orthopaedic visits	Procedure codes or encounters in EHR	Structured
	Number of Physical Therapy visits	Procedure codes or encounters in EHR	Structured
	Number of injections	Procedure codes or encounters in EHR	Structured
	Hospital admissions	Procedure codes or encounters in EHR	Structured
Treatment Success
	Treatment success indicators in clinical notes	Clinical note narrative from HPI section of follow-up orthopaedic visits	Unstructured

The design and development process for ICSCEDIT started with the creation of user interface wireframes using AdobeXD. Candidate wireframes were reviewed by the research and clinical team members in iterative user experience phases, each phase employing 1) prototype design and demonstration, 2) collection and documentation of expert/user feedback, and 3) prototype redesign based on feedback. ICSCEDIT was designed to be intuitive for physicians to use and visualize data so that it can be efficiently, accurately, and meaningfully interpreted. The research and clinical team consisted of one health services researcher, two healthcare technology user experience designers, one health economist with expertise in comparative effectiveness research and five orthopaedic surgeons, including three fellowship trained shoulder specialists, one orthopaedic traumatologist, and one hand and upper extremity specialist. The iterative user experience phases spanned 4 months and consisted of weekly meetings and design revisions. The process resulted in mockups of a final prototype application.

Reference Class Selection

Informatics consult systems should be designed to leverage physician expertise in defining patient groups similar to each patient being treated. To achieve that desired functionality, ICSCEDIT was designed with a series of patient characteristic filters which can be applied to subset historical patient records into similar cohorts matching the index patient of interest. A total of 14 patient filters across demographic, fracture characteristics, and patient comorbidities are included. Patient characteristic filters allow for data stratification on important treatment effect modifiers and confounding factors and aim to create homogeneous groups of patients with PHF. The physician user can apply as many or as few filters as desired and immediately view treatment utilization and outcome data by treatment group, for the selected matched cohort. Ideally, the informatics system would produce dynamic cohort searches and on-demand data summaries in seconds, making the system relevant during the clinical encounter as treatment options are discussed with the patient. Figure 1 displays reference class filter criteria on the left-hand side of the image. Once patient characteristics filters are applied, the total reference class sample size would be displayed across the bottom of the left-hand pane. The physician user can quickly modify his or her search by adding or removing filters, which will adjust the reference class size.

In Fig. 1 the search criteria would reflect a target patient with the following profile: female patient, age 75, with a fracture on her non-dominate shoulder. The fracture is a 2-part fracture which is majorly displaced at the surgical neck. The patient did not sustain any concurrent fractures in other regions of her body. She does not have a history of falls, does not currently have dementia or shoulder osteoarthritis, she lives at home, and is evaluated by the physician user to be frail. The diabetes filter was not engaged in this particular reference class search; therefore the reference cohort contains patients with and without diabetes.

Reference Class Treatment Rates

ICSCEDIT is designed to provide comparative effectiveness evidence for the reference class of patients the physician defines as similar to the index patient being treated. The first tab in the treatment comparators results pane displays treatment rates of initial surgery and conservative management for the reference class. This data will help physicians evaluate how much consensus there is in the treatment choice for this specific subset of patients. The middle portion of the data pane contains the surgical treatment rate over time for the physician user and the average surgical treatment across all physicians. This information will be helpful in showing how treatment rates are trending over time and how the physician user’s practice patterns compare to the average across all other surgeons.

At the bottom of the middle pane are two pie charts. These charts are used to represent the historical treatment patterns of 1) the physician user and 2) high-volume physician users. High volume physicians are defined as the physicians that care for 10% or more of the reference class cohort. The comparison of the physician user data compared to high-volume physicians provides insight on an individual’s practice patterns compared to their most experienced peer physicians. This feature introduces a learning component that may serve to educate physician outliers, thereby decreasing variability in treatment and improving overall quality of care.

The display of treatment variation across physicians and time, as demonstrated in the run chart at the center of the middle pane and the pie charts at the bottom of the middle pane, is data that has never previously been readily available to orthopaedic physicians. Instead, physicians have relied upon their own historical treatment experiences when making new treatment recommendations, which can be subject to cognitive errors. Now, physicians can evaluate their own decision-making in reference to other practicing physicians. This feature may be especially useful for residents, fellows, or young physicians who do not yet have years of experience treating patients. The feature adds a learning or training component that could be highly valuable for physicians. Additionally, with advances in data interoperability, this feature could be beneficial for physicians in small, remote practices who do not routinely treat a high volume of PHF injuries and could benefit from peer comparison.⁽²⁰⁾

Figure 1 displays the Treatment Comparators data visualizations. For the reference class displayed in Fig. 1, we see that 80% of the entire reference class was initially treated with conservative management. The physician user had historical surgical rates that at times differed and at times were similar to the average surgery rate for other physicians. Currently in Q2, the physician user was using surgery at a lower rate than other physicians. In the bottom two charts, we see that the physician user used surgery at a lower rate compared to high-volume physicians for patients in the matched reference class. The physician user surgery rate was only 12.5% compared to 25% for high volume physicians.

Personalized Treatment Evidence - PROMs

The second tab is the Mean Outcome Scores pane which contains summaries of PROMs presented for patients in the reference class by treatment groups. PROMs are considered a gold standard outcome in orthopaedic medicine as they measure improvement or changes in outcome dimensions that are important to patients.⁽⁴⁴⁾The PROMs selected for inclusion in ICSCEDIT include the visual analog pain score (VAPS), the Constant score, the Single Alpha Numeric Evaluation (SANE), the PROMIS-physical function score, and degrees of shoulder range of motion. PROMs are reported for four time points across the outcome period at three months, six months, one year and two years after the index PHF injury. Results are stratified and color-coded by treatment group for easy comparison. Physicians can quickly glance across all outcomes and evaluate which treatment group achieved superior outcomes.

Figure 2 shows the data visualizations in the Mean Outcome Scores pane. The patients in the reference class that received initial surgery are color coded with purple and the initial conservative management patients have scores in blue. From Fig. 2, we see that patients treated with surgery had higher SANE and Constant scores at 3 and 6 months, but SANE and Constant scores were similar between surgical and conservative groups by 2 years. Similarly, pain was under control as early as 3 months for the surgical group, but conservatively managed patients reported higher pain scores through six months. In terms of range of motion, surgically-treated patients achieved better forward flexion of 170 degrees compared to only 110 degrees for conservatively-managed patients.

Personalized Treatment Evidence – Healthcare Utilization

We believe that providing summaries of healthcare utilization over the year following the PHF injury provides helpful insight for physicians and patients as they weigh treatment options. The mean number of follow-up orthopaedic visits, physical therapy visits, injections, and hospital admissions are reported in ICSCEDIT. Follow-up orthopaedic visits and care may be indicative of routine care, or if prolonged, could indicate suboptimal outcomes. Similarly, severe adverse events which are unintended consequences of orthopaedic treatment, are captured by hospital admissions. The anticipated health care utilization associated with a treatment pathway is important for patients to know prior to treatment choice, as it helps patients manage expectations in the outcome period. We also report on the proportion of the conservative management treatment group that converted to surgical treatment within 60 days. This is a highly relevant data point for the treatment of PHF and may help expedite the decision to opt for initial surgical treatment.

Lastly, we introduce a conceptual measure of orthopaedic treatment success derived from EHR clinical notes. Clinical notes document the degree of improvement or relief experienced and reported directly by patients, in addition to scenarios in which symptoms have not resolved, or when subsequent complications have arisen. We believe that deep learning natural language processing (NLP) models can use phrases found in the clinical notes in EHR systems to develop measures of orthopaedic treatment success. Figure 3 shows the data visualizations in the Healthcare Utilization Outcomes pane.

We describe the proof-of-concept ICSCEDIT, a comparative effectiveness-based informatics consult system for the treatment of PHF. The long-term goal for ICSCEDIT is to improve physicians’ and patients’ ability to make real-time, data-informed treatment decisions and select the treatment that is best-suited to patient goals and preferences. ICSCEDIT offers personalized evidence for the unique patient that is sitting in front of the physician, which provides more tailored information than systematic reviews or meta-analyses on narrow RCT groups of patients. Visualized data summaries of how similar patients have been treated brings a unique perspective to the clinical encounter that is not currently available. Historical data summarized by ICSCEDIT may reveal previously concealed data relationships and patterns, like higher rates of treatment success or adverse events for patient subsets and can overcome common physician cognitive pitfalls. Furthermore, ICSCEDIT delivers up-to-date evidence on rapidly evolving practice trends which is an advancement over the sluggish translation of published literature findings. Each patient that is treated in clinical practice is looped back into the ICSCEDIT data pool, which creates an ongoing learning cycle. Future patient decision-making benefits from the treatment of patients in practice today. Through ongoing development, systems like ICSCEDIT can supplement physician decision-making and make on-demand informatics consults during the clinical encounter a reality.

ICSCEDIT has a similar conceptual function and design to other informatics consult conceptualizations,^{(13, 45, 46)} though our system differs in a few meaningful ways. First, unlike other systems that depend upon computer-based algorithms to phenotype⁽⁴⁷⁾ or identify patient reference classes, we leverage a physician-driven reference class identification process. Machine learning or artificial intelligence-based solutions use statistical methods to group patients by closeness or fit, but these methods are void of physician acumen and intuition.^{(10, 48, 49)} Computer-based approaches lack transparency in how patient reference classes are defined and can lead to challenges in the ability to explain and interpret results ex post because physicians can’t clearly identify the defining characteristics that were used to group patients.⁽⁵⁰⁾ Furthermore, there is hesitancy among patients and physicians to use computer-based recommender systems that lack the art of human element of medicine and intend to base clinical decisions on computer algorithms.^{(51, 52)} Physicians possess accumulated clinical wisdom on the patient characteristics that are relevant when making treatment decisions.^{(19, 53, 54)} Because of this unique knowledge, physicians are better positioned to phenotype or define reference classes of patients. ICSCEDIT has the capability to track physician reference class search patterns over time and learn from physician behavior. This provides the best translation of physician expert knowledge into computerized patterns. The expertise that a physician “knows what to search for” can be used to train computerized models to, at some point, automate the search for the index patient in the future.

Second, we introduce a novel element in ICSCEDIT which is the visualization of peer physician practice patterns. This feature allows the physician user to observe how their own treatment decision making differs from high-volume and physician peers. This feature would be especially valuable for inexperienced physicians or low-volume physicians and those without a large similar-specialty peer group. Peer comparison data has been introduced in other medical specialties and has been shown to improve the quality of care.⁽⁵⁵⁾ Furthermore, this makes ICSCEDIT a valuable educational tool for medical students, residents, and fellows in orthopaedic medicine.

Lastly, we introduce a conceptual measure of orthopaedic treatment success derived from EHR clinical notes. This measure applies deep learning NLP models to phrases found in the clinical notes in EHR systems to develop measures of orthopaedic treatment success. For example, from preliminary work we found phrases of success to include, “on exam, patient has near complete full range of motion”, and “patient has no complaints whatsoever”. On the other hand, phrases which indicate treatment failure include “patient continues to have significant pain and limitations”, and “patient wants to proceed with hardware removal”. The availability of clinical notes in the EHR system makes measures of orthopaedic treatment success widely available, which overcomes challenges with PROM collection and integration.⁽⁵⁶⁾

Future Steps

We developed a prototype that we believe demonstrates the ideal characteristics of a valuable informatics consult system for orthopaedic medicine. Our prototype will likely undergo many iterations as our knowledge about physician-computer interaction patterns evolves, and as we aim to develop a system that works seamlessly within the clinical encounter. Some possible design changes include the refinement of patient characteristics as reference class filters. We know that clinical details on the complexity of the fracture has important bearing on treatment choice and outcomes,⁽⁵⁷⁾ as well as lifestyle-related patient data like activity level and socioeconomic factors;⁽⁵⁸⁾ though challenges exist surrounding the consistency of fracture diagnosing across physicians.⁽³¹⁾ Future work should focus on efforts to standardize diagnosing and the availability of fracture specific characteristics in EHR systems. Ensuring that complete clinical and lifestyle patient characteristic data is available for phenotyping is a lynchpin of this work. Our research group is currently exploring ways to ensure complete data exists in the EHR for patients with PHF.

We also believe that patient general health or frailty, opposed to or in addition to the presence of individual comorbidities, is relevant to treatment choice and outcomes. We currently have included individual comorbidities (i.e., diabetes, dementia, shoulder osteoarthritis) as reference class selection criteria, but it may prove better to include a general measure of patient frailty or well-being. Indexes like the Charlson Comorbidity Index⁽⁵⁹⁾ or the Function-Related indicators (FRIs)⁽⁶⁰⁾ may be potential options. Additionally, the treatment outcomes and PROMs which are selected for inclusion in ICSCEDIT may differ by physician or health system and may be dictated by data collection and availability. Although PROM data collection is challenging for orthopaedic practices and health systems, data collection efforts are on the rise due to value-based payment initiatives.⁽⁶¹⁾ As data collection efforts expand in years to come, outcome data including routine clinical indicators of flexion, abduction and external rotation could be expanded. Alternatively, it is possible that a system like ICSCEDIT could function with orthopaedic society registries, which would offer national samples of patients and practice variability.

Despite anticipated modifications to future iterations and data sources, there is enthusiasm among physicians for ICSCEDIT in its current form and tools with similar functionality.⁽²⁰⁾ Though, beyond just physicians, we believe that patients, health system administrators, and payers all would be interested users of a system like ICSCEDIT to guide individual and population health-level treatment decisions. ICSCEDIT and similar systems bring incredible value to assist physicians when faced with difficult treatment decisions. Unlike some other clinical decision support tools, we are deliberate in our efforts to develop a tool that complements, not supersedes, physician decision-making in a transparent, concise way.

We have developed mockups and conducted initial physician demonstrations and evaluations. Our next step for ICSCEDIT will be to use real patient data and develop a functioning prototype so that we may evaluate functionality of the system in practice.

We describe a prototype of an informatics consult system that is emerging and has been refined for orthopaedic medicine. Historical EHR data holds important answers about how we should treat current and future patients. Informatics consult systems that enable learning from historical treatment decisions and outcomes holds great potential for improving the quality of care for all orthopaedic patients. Given that orthopaedic medicine lacks treatment evidence for many conditions and current approaches to generate that evidence will likely take years, if ever accomplished, ICSCEDIT offers a novel paradigm to accelerate learning from historical practice and associated outcomes. We envision expanding ICSCEDIT to other conditions and injuries within orthopaedic medicine.

RCT

Randomized Clinical Trial

EHR

Electronic Health Record

ICSCEDIT

Informatics Consult System for Comparative Effectiveness Data for Individualized Treatment

CER

Comparative Effectiveness Research

PHF

Proximal Humerus Fracture

PROM

Patient-Reported Outcome Measure

VAPS

Visual Analog Pain Score

SANE

Single Assessment Numeric Evaluation

NLP

Natural Language Processing

Ethics approval and consent to participate: Not applicable

Consent for publication: Not applicable

Availability of data and materials: The datasets generated and/or analyzed during the current study are not publicly available due to the ongoing development of the clinical decision support tool but are available from the corresponding author on reasonable request.

Competing interests: The authors declare that they have no competing interests.

Funding: This work was supported by P20 GM121342. The funding source had no role in the design, collection, analysis, or interpretation of the data, or the decision to submit the manuscript for publication.

Authors’ contributions: AS, BS, SL, JB and SF all made substantial contributions to the conception of the work, and the analysis and interpretation of the data. MK, SP, GF, CD and KJ all made substantial contributions to the conception of the work. All authors read, substantially modified, and approved the final manuscript prior to submission.

Acknowledgements: Not applicable

Sheridan DJ, Julian DG. Achievements and Limitations of Evidence-Based Medicine. J Am Coll Cardiol. 2016;68(2):204–13.
Sackett DL, Rosenberg WMC, Gray JAM, Haynes RB, Richardson WS. Evidence based medicine: what it is and what it isn't. BMJ. 1996;312(7023):71–2.
(US). IoM. Evidence-Based Medicine and the Changing Nature of Healthcare: 2007 IOM Annual Meeting Summary.. Washington (DC): National Academies Press 2008.
Greenhalgh T, Howick J, Maskrey N. Evidence based medicine: a movement in crisis? BMJ. BMJ. 2014;348:g3725.
Ostropolets A, Chen R, Zhang L, Hripcsak G. Characterizing physicians' information needs related to a gap in knowledge unmet by current evidence. JAMIA Open. 2020;3(2):281–9.
Du Vaure CBDA, Battin C, Ravaud P, Boutron I. Exclusion of patients with concomitant chronic conditions in ongoing randomised controlled trials targeting 10 common chronic conditions and registered at ClinicalTrials. gov: a systematic review of registration details. BMJ Open. 2016;e012265.
Kravitz RL, Duan N, Braslow J. Evidence-Based Medicine, Heterogeneity of Treatment Effects, and the Trouble with Averages. Millbank Q. 2004;82(4):661–87.
Gabler NB, Duan N, Liao D, Elmore JG, Ganiats TG, Kravitz RL. Dealing with heterogeneity of treatment effects: is the literature up to the challenge? Trials. 2009;10(1):43.
Ross MK, Wei W, Ohno-Machado L. "Big data" and the electronic health record. Yearb Med Inform. 2014;9(1):97–104.
Tang PC, Miller S, Stavropoulos H, Kartoun U, Zambrano J, Ng K. Precision population analytics: population management at the point-of-care. J Am Med Inform Assoc. 2021;28(3):588–95.
Gallego B, Dunn AG, Coiera E. Role of electronic health records in comparative effectiveness research. J Comp Eff Res. 2013;2(6):529–32.
Frankovich J, Longhurst C, Sutherland SM. Evidence-Based Medicine in the EMR Era. NEJM. 2011.
Gallego B, Walter SR, Day RO, Dunn AG, Sivaraman V, Shah N, et al. Bringing cohort studies to the bedside: framework for a "green button' to support clinical decision-making. J Comp Eff Res. 2015;4(3):191–7.
Longhurst CA, Harrington RA, Shah NH. A 'green button' for using aggregate patient data at the point of care. Health Aff (Millwood). 2014;33(7):1229–35.
Administration USFaD. Framework for FDA’S real-world evidence program.
Blom AW, Donovan RL, Beswick AD, Whitehouse MR, Kunutsor SK. Common elective orthopaedic procedures and their clinical effectiveness: umbrella review of level 1 evidence. BMJ. 2021;374:n1511.
Obremskey WT, Pappas N, Attallah-Wasif E, Tornetta P 3rd, Bhandari M. Level of evidence in orthopaedic journals. J Bone Joint Surg Am. 2005;87(12):2632–8.
Gombar S, Callahan A, Califf R, Harrington R. Shah, N. It is time to learn from patients like mine. Digit Med. 2019;2:16.
Schuler A, Callahan A, Jung K, Shah NH. Performing an Informatics Consult: Methods and Challenges. J Am Coll Radiol. 2018;15(3 Pt B):563–8.
Lai AG, Chang WH, Parisinos CA, Katsoulis M, Blackburn RM, Shah AD, et al. An informatics consult approach for generating clinical evidence for treatment decisions. BMC Med Inform Decis Mak. 2021;21(1):281.
Dahabreh IJ, Hayward R, Kent DM. Using group data to treat individuals: understanding heterogeneous treatment effects in the age of precision medicine and patient-centred evidence. Int J Epidemiol. 2016;45(6):2184–93.
Kent D, Paulus J, van Klaveren D, D'Agostino, Goodman S, et al. The Predictive Approaches to Treatment effect Heterogeneity (PATH) Statement. Ann Intern Med. 2020;172(1):35–45.
Kent DM, Steyerberg E, van Klaveren D. Personalized evidence based medicine: predictive approaches to heterogeneous treatment effects. BMJ. 2018;363:k4245.
Wendling T, Jung K, Callahan A, Schuler A, Shah NH, Gallego B. Comparing methods for estimation of heterogeneous treatment effects using observational data from health care databases. Stat Med. 2018;37(23):3309–24.
Brooks JM, Chapman CG, Schroeder MC. Understanding Treatment Effect Estimates When Treatment Effects Are Heterogeneous for More Than One Outcome. Appl Health Econ Health Policy. 2018;16(3):381–93.
Handoll H, Brealey S, Rangan A, Keding A, Corbacho B, Jefferson L, et al. The ProFHER (PROximal Fracture of the Humerus: Evaluation by Randomisation) trial - a pragmatic multicentre randomised controlled trial evaluating the clinical effectiveness and cost-effectiveness of surgical compared with non-surgical treatment for proximal fracture of the humerus in adults. Health Technol Assess (Winchester Eng). 2015;19(24):1–280.
Handoll HH, Keding A, Corbacho B, Brealey SD, Hewitt C, Rangan A. Five-year follow-up results of the PROFHER trial comparing operative and non-operative treatment of adults with a displaced fracture of the proximal humerus. Bone Joint J. 2017;99-b(3):383 – 92.
Rangan A, Handoll H, Brealey S, Jefferson L, Keding A, Martin BC, et al. Surgical vs nonsurgical treatment of adults with displaced fractures of the proximal humerus: the PROFHER randomized clinical trial. JAMA. 2015;313(10):1037–47.
Sabharwal S, Patel NK, Griffiths D, Athanasiou T, Gupte CM, Reilly P. Trials based on specific fracture configuration and surgical procedures likely to be more relevant for decision making in the management of fractures of the proximal humerus: Findings of a meta-analysis. Bone & joint research. 2016;5(10):470–80.
LaMartina J 2nd, Christmas KN, Simon P, Streit JJ, Allert JW, Clark J, et al. Difficulty in decision making in the treatment of displaced proximal humerus fractures: the effect of uncertainty on surgical outcomes. J Shoulder Elbow Surg. 2018;27(3):470–7.
Handoll HH, Brorson S. Interventions for treating proximal humeral fractures in adults. Cochrane Database Syst Rev. 2015(11):Cd000434.
Handoll HH, Ollivere BJ, Rollins KE. Interventions for treating proximal humeral fractures in adults. Cochrane Database Syst Rev. 2012;12:Cd000434.
Floyd S, Thigpen C, Kissenberth M, Brooks J. Association of Surgical Treatment With Adverse Events and Mortality Among Medicare Beneficiaries With Proximal Humerus Fracture. JAMA Open. 2020;3(1):1–14.
Handoll HH, Keding A, Corbacho B, Brealey SD, Hewitt C, Rangan A. Five-year follow-up results of the PROFHER trial comparing operative and non-operative treatment of adults with a displaced fracture of the proximal humerus. The bone & joint journal. 2017;99-B(3):383–92.
Floyd S, Campbell J, Chapman C, Thigpen C, Kissenberth M, Brooks J. Geographic variation in the treatment of proximal humerus fracture: an update on surgery rates and treatment consensus. Journal of Orthopaedic Surgery and Research. 2019;14.
Bell LB, Spratt K, Koval K, Weinstein J, Goodman D, et al. Trends and Variation in Incidence, Surgical Treatment, and Repeat Surgery of Proximal Humeral Fractures in the Elderly. J Bone Joint Surgery-American Volume. 2011;93A(2):121–31.
Wennberg JE. Which Rate is Right? N Engl J Med. 1986;315(13):810–5.
McCulloch P, Nagendran M, Campbell WB, Price A, Jani A, Birkmeyer JD, et al. Strategies to reduce variation in the use of surgery. Lancet. 2013;382(9898):1130–9.
Bernstein J. Not the Last Word: Choosing Wisely. Clin Orthop Relat Res. 2015;473(10):3091–7.
Bernstein J. Not the Last Word: Bhandari's Paradox. Clin Orthop Relat Res. 2018;476(4):674–7.
Lurie JD, Bell JE, Weinstein J. What rate of utilization is appropriate in musculoskeletal care? Clin Orthop Relat Res. 2009;467(10):2506–11.
Cram P, Landon BE, Matelski J, Ling V, Perruccio AV, Paterson JM, et al. Utilization and Outcomes for Spine Surgery in the United States and Canada. Spine (Phila Pa 1976). 2019;44(19):1371–80.
Austin DC, Torchia MT, Lurie JD, Jevsevar DS, Bell JE. Identifying regional characteristics influencing variation in the utilization of rotator cuff repair in the United States. J Shoulder Elbow Surg. 2019;28(8):1568–77.
Gagnier J. Patient reported outcomes in orthopaedics. J Orthop Res. 2017;35(10):2098–108.
Mane KK, Bizon C, Schmitt C, Owen P, Burchett B, Pietrobon R, et al. VisualDecisionLinc: a visual analytics approach for comparative effectiveness-based clinical decision support in psychiatry. J Biomed Inform. 2012;45(1):101–6.
Rogers J, Spina N, Neese A, Hess R, Brodke D, Lex A. Composer-Visual Cohort Analysis of Patient Outcomes. Appl Clin Inform. 2019;10(2):278–85.
Newton KM, Peissig PL, Kho AN, Bielinski SJ, Berg RL, Choudhary V, et al. Validation of electronic medical record-based phenotyping algorithms: results and lessons learned from the eMERGE network. J Am Med Inform Assoc. 2013;20(e1):e147-54.
Ng K, Kartoun U, Stavropoulos H, Zambrano JA, Tang PC. Personalized treatment options for chronic diseases using precision cohort analytics. Sci Rep. 2021;11(1):1139.
Lee J, Maslove DM, Dubin JA. Personalized mortality prediction driven by electronic medical data and a patient similarity metric. PLoS ONE. 2015;10(5):e0127428.
Fang HSA, Tan NC, Tan WY, Oei RW, Lee ML, Hsu W. Patient similarity analytics for explainable clinical risk prediction. BMC Med Inf Decis Mak. 2021;21(1):207.
Richardson JP, Smith C, Curtis S, Watson S, Zhu X, Barry B, et al. Patient apprehensions about the use of artificial intelligence in healthcare. npj Digit Med. 2021;4(1):140.
Soellner M, Koenigstorfer J. Compliance with medical recommendations depending on the use of artificial intelligence as a diagnostic method. BMC Med Inform Decis Mak. 2021;21(1):236.
Pearson H. Science and intuition: do both have a place in clinical decision making? Br J Nurs. 2013;22(4):212–5.
Jayakumar P, Moore MLG, Bozic KJ. Value-based Healthcare: Can Artificial Intelligence Provide Value in Orthopaedic Surgery? Clin Orthop Relat Res. 2019;477(8):1777–80.
Navathe AS, Volpp KG, Bond AM, Linn KA, Caldarella KL, Troxel AB, et al. Assessing The Effectiveness Of Peer Comparisons As A Way To Improve Health Care Quality. Health Aff (Millwood). 2020;39(5):852–61.
Lizzio VA, Dekhne MS, Makhni EC. Electronic Patient-Reported Outcome Collection Systems in Orthopaedic Clinical Practice. JBJS Reviews. 2019;7(7).
Okike K, Lee OC, Makanji H, Harris MB, Vrahas MS. Factors associated with the decision for operative versus non-operative treatment of displaced proximal humerus fractures in the elderly. Injury. 2013;44(4):448–55.
Miquel J, Elisa C, Fernando S, Alba R, Torrens C. Non-medical patient-related factor influence in proximal humeral fracture outcomes: a multicentric study. Arch Orthop Trauma Surg. 2021;141(11):1919–26.
Charlson M, Pompei P, Ales K, Mac Kenzie C. A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Disease. 1987;40(5):373–83.
Chrischilles E, Schneider K, Wilwert J, Lessman G, O'Donnell B, Gryzlak B, et al. Beyond comorbidity: expanding the definition and measurement of complexity among older adults using administrative claims data. Med Care. 2014;52(Suppl 3):75–84.
Baumhauer JF, Bozic KJ. Value-based Healthcare: Patient-reported Outcomes in Clinical Decision Making. Clin Orthop Relat Res. 2016;474(6):1375–8.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

The Development of a Proof-of-Concept Physician-Driven Informatics Consult System for the Individualized Treatment of Patients with Orthopaedic Conditions

Status:

Version 1

Abstract

Figures

Background

Methods

ICSCEDIT Development Process

Results - Icscedit Overview

Personalized Treatment Evidence - PROMs

Personalized Treatment Evidence – Healthcare Utilization

Discussion

Future Steps

Conclusion

Abbreviations

Declarations

References

Additional Declarations

Status:

Version 1