Predictors of Concordance between Patient-Reported and Provider-Documented Symptoms in the Context of Cancer and Multimorbidity

Abstract

Background The integration of patient-reported outcomes (PROs) into clinical care, particularly in the context of cancer and multimorbidity, is crucial. While PROs have the potential to enhance patient-centered care and improve health outcomes through improved symptom assessment, they are not always adequately documented by the health care team.

Objectives This study aimed to explore the concordance between patient-reported symptom occurrence and symptoms documented in electronic health records (EHRs) in people undergoing treatment for cancer in the context of multimorbidity.

Methods We analyzed concordance between patient-reported symptom occurrence of 13 symptoms from the Memorial Symptom Assessment Scale and provider-documented symptoms extracted using NimbleMiner, a machine learning tool, from EHRs for 99 patients with various cancer diagnoses. Logistic regression guided with the Akaike Information Criterion was used to identify significant predictors of symptom concordance.

Results Our findings revealed discrepancies in patient and provider reports, with itching showing the highest concordance (66%) and swelling showing the lowest concordance (40%). There was no statistically significant association between multimorbidity and high concordance, while lower concordance was observed for women, patients with advanced cancer stages, individuals with lower education levels, those who had partners, and patients undergoing highly emetogenic chemotherapy.

Conclusion These results highlight the challenges in achieving accurate and complete symptom documentation in EHRs and the necessity for targeted interventions to improve the precision of clinical documentation. By addressing these gaps, health care providers can better understand and manage patient symptoms, ultimately contributing to more personalized and effective cancer care.

Keywords

cancer - concordance - multimorbidity - oncology - patient-reported outcomes

Background and Significance

Patient-reported outcomes (PROs) provide essential firsthand insights from patients about their health status and functionality, playing a crucial role in delivering high-quality, patient-centered care.[1] [2] Cancer and its treatment are associated with a large number of symptoms, making patient-reported symptoms vital PROs in the context of cancer care. Acknowledging symptoms as essential PROs allows for a more patient-centered approach to cancer care, resulting in improved outcomes such as improved survival and quality of life (QOL).[3] [4] [5] The documentation of patient-reported symptoms in electronic health records (EHRs) is a key step toward achieving a more integrated view of a patient's health status.[6] [7] Accurate and timely documentation not only provides a detailed record of a patient's symptoms but also acts as the foundation for informed decision-making and personalized symptom management.[8] For instance, evidence has shown that accurate documentation of PROs such as symptoms and side effects can facilitate better communication between health care providers (HCPs) from various clinical specialties, improve continuity of care, particularly during care transitions, and reduce delays and errors in patient treatment.[9] [10] However, other studies have identified challenges in achieving accurate and complete documentation, including discrepancies between patient-reported symptoms and those documented by providers.[11] [12] [13] These challenges include the factors that drive HCP documentation (e.g., billing, accreditation, time constraints, focus of the visit, provider specialty) as well as patient beliefs and attitudes (e.g., symptoms are expected, do not want to be a bother, only describing symptoms when asked about them).

The degree to which HCPs document patient-reported symptoms is instrumental in establishing the concordance between PROs and clinical assessments within the EHR. This complex relationship points out the need for precision in translating patients' narratives into their medical records.[14] However, a significant challenge lies in the potential discordance between patient-reported symptoms and those documented by HCPs in the EHR.[13] Concordance, or the agreement between patient-reported experiences and clinical evaluations, is fundamental for assessing and validating PROs[15] and can enhance patient survival and improve QOL.[15] [16] However, the concordance between patients and HCPs on symptom severity often varies, with patients reporting more severe symptoms than what their HCPs document.[17] [18] [19] [20] Accuracy of symptom documentation can be further impacted when patients have complex health conditions, such as having multimorbidity. Rather than experiencing and managing symptoms from one chronic condition, patients with multimorbidity experience a wide range of symptoms each related to the different chronic conditions they face. Needing to prioritize and attend to many different symptoms from multiple chronic conditions can result in more discrepancies between patient self-reports of symptoms and those documented in the EHR.[21] [22]

Objectives

The concordance between patient-reported symptoms and their documentation in EHRs remains underexplored in the context of cancer and multimorbidity. To our knowledge, there have been no published analyses reporting symptom concordance between patients and HCPs in the context of cancer and multimorbidity. Thorough and accurate documentation of PROs is especially important in this population as cancer care is already challenging and managing multiple chronic conditions will add to this complexity, putting these patients at risk for having under assessed and under managed symptoms. Therefore, the objective of this study is 2-fold: first, to determine the level of concordance between patient-reported symptoms and EHR documentation of symptoms in the context of cancer and multimorbidity, and second, to explore if multimorbidity influences this concordance. This study will help us better understand how patient-reported symptoms match up with EHR records, ultimately contributing to advancing patient-centered care in cancer and oncology.

Methods

Data

This retrospective, cross-sectional study integrated two datasets to assess concordance between symptom occurrence reports: patient-reported symptoms and provider-reported symptoms ([Fig. 1]). Participants included adults with breast, prostate, lung, renal, bladder, or skin (melanoma) cancer receiving chemotherapy at a large Midwestern cancer center. This study combines the data from two separate studies in which there were overlapping participants. Study 1 collected symptoms reported from patients undergoing treatment for cancer.[23] Study 2 extracted symptom documentation from free text notes as part of a large study to develop interpretable machine learning models that predict the development of cancer symptoms.[24]

Fig. 1 Study framework illustrating the use of patient-reported symptoms and provider-documented symptoms to measure symptom concordance. EHR, electronic health record.

Demographic information was obtained from participants and included age, gender, education, and partner status. Details about participant cancer diagnoses and treatment were obtained from medical record review and included cancer diagnosis and stage, and emetogenic potential of chemotherapy (e.g., the likelihood of chemotherapy causing nausea and vomiting). Self-reported gender was used in these analyses rather than biological sex to reflect the individual's self-representation.

Patient-reported symptom occurrence was collected using 13 items from the Memorial Symptom Assessment Scale (MSAS).[25] [26] The MSAS asks participants to review a list of 32 symptoms and indicate their presence in the last week, as well as symptom duration, severity, and distress. For this analysis, symptom occurrence reports of 13 symptoms common across multiple cancer primary sites were selected. Participants provided these symptom occurrence ratings on the day they received chemotherapy in the cancer clinic.

Provider-documented symptom occurrence of 13 symptoms was extracted from EHRs using NimbleMiner,[27] an advanced machine learning tool designed for symptom extraction from clinical notes. NimbleMiner combines machine and human intelligence to train the system to recognize symptoms mentioned in EHRs. NimbleMiner's output assigns a binary value (1 for presence, 0 for absence) to each symptom at the note level. The performance of NimbleMiner was validated against a gold-standard dataset of 1,112 notes, manually annotated by two independent reviewers with a high interannotator reliability score of 0.924.[28] NimbleMiner demonstrated a microaveraged precision of 0.878 and a recall of 0.876, indicating that the symptom extraction from EHR using NimbleMiner is reasonably accurate.

Chronic conditions were recorded through EHR review using a modified Charlson Comorbidity Index (CCI).[29] [30] The CCI records the presence of 17 chronic conditions such as congestive heart failure, myocardial infarction, and chronic obstructive pulmonary disease. In addition to these chronic conditions, we also included assessments for obesity (body mass index ≥ 30), hypertension, and hyperlipidemia. Finally, as all our participants had a cancer diagnosis, we did not include this chronic condition in the count of chronic conditions.

Evaluating Concordance

For this analysis, we focused on 13 common cancer symptoms collected on MSAS and documented as part of standard care in the oncology clinical notes and included anxiety, pain, loss of appetite, xerostomia (dry mouth), nausea, constipation, pruritus (itching), depressed mood, numbness, fatigue, shortness of breath, sleep disturbance, and swelling.

Concordance was determined by comparing the patient-reported symptoms on the day of completing the survey with the corresponding EHR documentation of these symptoms within a range of 1 week before and after the patient symptom survey, identifying them as either present or absent. All note types were included (e.g., outpatient clinic notes, inpatient notes, telephone calls, etc.). A symptom was considered concordant when patient-reported symptom and provider-documented symptom reports aligned for a symptom (i.e., both describe the symptom as present or both described the symptom as absent), whereas discordance was noted for mismatches. Overall concordance for each patient was calculated by tallying the number of concordant symptoms and converting this count into a percentage of total symptoms assessed, reflecting the frequency of alignment between patient and provider reports. We created a high concordance variable, which is assigned a value of 1 when the overall concordance is at least 7 symptoms (out of 13 possible) and 0 for 6 or fewer concordant symptoms.

Model Selection Using Akaike Information Criterion

We employed logistic regression, guided by the Akaike Information Criterion (AIC),[31] to determine the statistical significance of various predictors in relation to high concordance, which served as the target outcome. The set of explanatory variables used in the regression model were age, gender, education, partner status (self-reported from demographic questionnaire), and type of cancer, cancer stage, emetogenicity category, and total number of comorbid conditions (extracted from the EHR). We initiated our analysis with a comprehensive model that included all potential predictors and then applied AIC to refine this model by systematically removing less informative variables to identify the best model for predicting the likelihood of high concordance between patient-reported symptoms and provider-documented symptoms. AIC is an estimate of the distance between a candidate model and the “true model,” on a log-scale, based on the Kullback–Leibler divergence. Model distance to the “true model” can be compared based on their closeness to the “true model.” The best model is one that has the smallest distance (i.e., AIC value).[31] AIC provides a balance between complexity and interpretability, avoiding the biases of simpler statistical methods[32] [33] as well as the limitations of p-value reliance.[34] A further benefit of AIC is the ability to handle both nested and non-nested models,[35] which allows for consideration of all variable subsets, providing both a statistically sound and meaningful model. This approach ensures critical insights into symptom management and patient–provider communication are not missed. Previous research has reported that a decrease of 2 or more AIC units between models indicates a meaningful improvement in penalized fit, with models within 2 units considered similarly effective.[35] For our analysis, we prioritized minimizing the risk of overlooking significant findings (Type II errors) that could enhance patient care and accepting a higher likelihood of false positives (Type I errors).[36] Based on this, it is acceptable when using AIC to set the level of significance (α) at 0.10 for individual predictors in the final model, recognizing that AIC-based model selection is functionally equivalent to using a more liberal significance threshold of approximately 0.157.[37]

Results

Patient Characteristics

Our sample included a total of 99 participants ([Table 1]). The average age was 59 years, most were women (62%) and partnered (70%). Educational backgrounds varied, with a significant portion having completed high school, general educational development (GED), or some college (41%), and a smaller group holding advanced degrees (21%). Participants were diagnosed with various cancers, predominantly lung cancer (33%) and breast cancer (29%), with the majority in the advanced stage IV (42%). Most participants experienced low to moderate chemotherapy emetogenicity (78%). The comorbidity count ranged from 0 to 9, with an average comorbidity count of 2.3.

Table 1 Demographic and clinical characteristics

Characteristics	N = 99
Age, mean (SD)	59.02 ± 12.86[a]
Gender, %
Women	61 (62)
Men	38 (38)
Marital[a], %
Married/partner	69 (70)
Single/widower/divorced	29 (30)
Education[a], %
Associate/Bachelor	37 (38)
High school/GED/some college	40 (41)
Some graduate/master/doctoral	21 (21)
Diagnosis, %
Lung cancer	33 (33)
Breast cancer	29 (29)
Melanoma	29 (29)
Bladder cancer	3 (3)
Prostate cancer	3 (3)
Kidney cancer	2 (2)
Stage[a], %
Stage I	11 (11)
Stage II	17 (18)
Stage III	28 (29)
Stage IV	41 (42)
Emetogenicity[b], %
High/very high	22 (22)
Low/moderate	77 (78)
Comorbidity
Range	0–9
Mean (SD)	2.27 ± 1.86[a]

Abbreviations: GED, general educational development; SD, standard deviation.

^a Missing data. Percents may not add to 100% due to rounding.

^b The likelihood of chemotherapy to cause nausea and vomiting.

Concordance

The comparison between patient-reported symptoms and provider-documented symptoms revealed a notable discrepancy in perception and reporting ([Fig. 2]). While patients identified fatigue (80%), pain (53%), and sleep disturbance (57%) as their most frequent symptoms, HCPs more commonly reported pain (74%), shortness of breath (54%), and swelling (62%). Concordance in the reporting of 13 symptoms by both parties was generally moderate, ranging from 40 to 66%. Specifically, swelling showed the lowest level of agreement at 40%, with HCPs reporting this symptom more than three times as often as patients. Sleep disturbance and fatigue were also among the symptoms with lower concordance rates, at 46 and 49%, respectively, but these unlike swelling were reported more by patients than by clinicians. Pruritis was the symptom exhibiting the highest concordance rate at 66%, indicating a closer agreement between patients and HCPs.

Fig. 2 Comparison of patient-reported and health care provider (HCP)-documented symptom prevalence with concordance rates. The proportion of symptoms reported by patients and HCPs, along with the percentage of concordance between these two sources for each symptom is listed.

Through a series of logistic regression models ([Table 2]), we refined the predictors for the high concordance indicator. Our comprehensive starting point, Model 1, incorporated variables including age, gender, education, partner status, cancer stage, cancer site, total comorbidity, and emetogenicity, with an AIC value of 138.81 across all the variables included in the model. As we streamlined the model by excluding diagnosis, then age, total comorbidity, and finally emetogenicity, the log-likelihood values modestly increased, and the complexity of the model decreased. This iterative process culminated in a final model with the most substantial reduction in AIC (132.01). Although the model with the lowest AIC is typically considered the best, the difference in AIC between the last (132.01) and the penultimate model (132.24) was less than 2. Given this minimal difference (0.23) and adhering to the principle that models within 2 AIC units of each other have similar penalized fits,[35] we opted for the penultimate model as our final choice.

Table 2 Stepwise logistics regression analysis for predicting high concordance

Model	−2log (likelihood)	k	ΔAIC[c]
Model 1[a]	55.41	14	6.80
Model 1—Diagnosis	58.81	9	3.60
Model 1—Diagnosis − Age	59.00	8	1.98
Model 1—Diagnosis – Age − Total Comorbidity	59.12	7	0.23
Model 1—Diagnosis − Age − Total Comorbidity − Emetogenicity	60.01	6	0.00[b]

Abbreviation: AIC, Akaike Information Criterion.

^a Model 1 = Gender + Stage + Education + Marital + Diagnosis + Age + Total Comorbidity + Emetogenicity.

^b Lowest AIC = 132.01

^c ΔAIC = AIC of current model − lowest AIC.

In our final logistic regression model ([Table 3]), we found that gender was a significant predictor of symptom concordance with women being less likely to have concordant symptom reports (odds ratio = 3.2, p = 0.021). Cancer stage was also a significant predictor of symptom concordance, with each additional cancer stage reducing concordance by 34% (odds ratio = 0.66, p = 0.067). Individuals with a high school diploma, GED, or some college education had significantly lower concordance compared with participants with an associate or bachelor's degree (odds ratio = 2.48, p = 0.092); however, postgraduate education did not significantly impact concordance (odds ratio = 0.86, p = 0.810). Participants who were partnered were less likely to show symptom concordance compared with participants who were single (odds ratio = 0.47, p = 0.16) and participants receiving high/very high emetogenic chemotherapy were less likely to show symptom concordance compared with participants receiving low/moderate emetogenic chemotherapy (odds ratio = 0.46, p = 0.19), but these did not reach statistical significance.

Table 3 Odds ratio and statistical significance (at 10% level of significance) of predictors for high concordance

Variable	Category	Odds ratio	p-Value
Gender	Woman
	Man	3.20	0.021
Stage[a]	Per 1 unit increase in stage	0.66	0.067
Education	High school/GED/some college
	Associate/Bachelor	2.48	0.092
	Some graduate/master/doctoral	0.86	0.810
Marital status	Single/widower/divorced
	Married/partner	0.47	0.156
Emetogenicity	Low/moderate
	High/very high	0.46	0.193

Abbreviations: GED, general educational development.

^a Stage is considered as a continuous variable for this analysis.

Concordance and Multimorbidity

Within the logistic regression analysis, total comorbidity count did not emerge as a significant predictor of concordance. As a result, this variable was not included in the final model selection as determined by the AIC, suggesting its limited impact on the prediction of high concordance in our study sample.

Discussion

Patient self-report of symptoms has long been the gold standard in cancer symptom science and clinical practice.[38] [39] The lack of concordance between patient self-report and HCP documentation in oncology has been an area of concern for nearly 20 years[40] with minimal improvement in the area despite multiple reports of its importance.[41] In addition, there is growing evidence suggesting that multimorbidity may negatively impact outcomes such as symptom burden and QOL in patients with cancer.[42] Therefore, this study sought to assess the concordance between patient-reported symptoms, as key PROs in cancer care, and EHR documentation of symptoms in the context of cancer and multimorbidity.

We found discrepancies in the reporting of symptoms, with concordance levels ranging from approximately 40 to 66%. HCP documentation of symptoms had worse concordance for women, those with advanced cancer, those with lower education levels, those who had partners, and those undergoing highly emetogenic chemotherapy. Surprisingly, multimorbidity did not emerge as a significant factor impacting concordance between patient and HCP reports of symptoms.

While the existing literature and our data ([Fig. 3]) indicate that women report symptoms at a higher frequency,[43] this heightened reporting does not translate into a greater concordance with HCP documentation. Instead, we found that men's self-reports exhibit a higher likelihood of alignment with HCP documentation ([Table 3]). One possible explanation is that while women may be more likely than men to self-report symptoms on a questionnaire, that difference may not translate to clinical discussions with their HCP. In fact, social desirability is a factor that has been associated with decreased reporting of symptoms if the patient perceives that it is more desirable for the provider to do so.[41] Similar disparities in reporting and documentation have been observed across different races, ethnicities, and language preferences.[44] This suggests that sociodemographic factors may influence the likelihood of patient self-reports being documented by HCPs. Therefore, it is crucial for HCPs to be aware of these potential disparities and actively work to ensure that patient self-reports and PROs, including symptoms, are accurately captured and addressed.

Fig. 3 Comparison of the prevalence of patient-reported and health care providers (HCPs)-reported symptoms by gender (women and men).

Our results also found that participants with a high school, GED, or some college education had lower symptom concordance compared with those with an associate or bachelor's degree. Decades of research have shown that higher levels of education attainment are associated with numerous health outcomes, including ability to navigate the health care system and communicate more effectively with their HCPs.[45] With regard to partnering status, while being single/unpartnered was retained in the final regression model it was not a statistically significant predictor of symptom concordance. The association of partnering status with patient-centered outcomes, such as functional status, symptoms, and QOL,[46] [47] is mixed thus limiting evidence regarding the impact of relationship status on concordance. Our study demonstrated that partnering status did not significantly influence symptoms reporting concordance in the context of cancer. Future research is needed to better understand the role of relationship status and patient–provider concordance of PROs such as symptoms.

In addition, we found that advanced stage cancer was associated with lower symptom concordance. One explanation for this result is that advanced stage cancer is frequently characterized by high symptom burden including the “constitutional” symptoms common in the SPPADE (i.e., sleep disturbance, pain, physical function impairment, anxiety, depression, and low energy/fatigue) symptom cluster[48] with patient reporting an average of 10 more concurrent symptoms[49] [50] suggesting that HCP may simply assume patients are experiencing multiple symptoms, which may not meet their threshold for documentation.[49] [50]

Finally, an area of particular interest with this paper was to understand the role of multimorbidity in symptom concordance. Our results did not find this relationship to be significant. Previous research has reported discrepancies in symptom prioritization and communication challenges when managing patients with multimorbidity;[51] [52] thus, the results of our analysis were surprising. This result may be due to the symptom assessment survey and the HCP assessment and documentation are both centered on cancer, rather than primary care where other chronic conditions and their associated symptoms are the focus. Future research is needed to evaluate concordance in the context of multimorbidity across medical specialties as well as in primary care.

Limitations

While this work is among the first to explore predictors of high symptom reporting concordance between patients with cancer and their HCPs, we must acknowledge some limitations. First, this analysis used a small dataset (N = 99) from patients with various cancers seen at a single academic medical center, which may impact the generalizability of the findings. Next, while the CCI is a frequently used measure of comorbidity, it has a limited number of comorbid conditions that include conditions associated with rehospitalization and mortality. This sample of participants who were receiving outpatient chemotherapy was a relatively healthy group reporting relatively few chronic conditions assessed by the CCI. A measure of multimorbidity that is more expansive and focuses on the individual burden of multimorbidity[53] may provide a more nuanced understanding of the experience of symptoms in people with cancer and other chronic conditions. Additionally, our analysis did not include language as a potential predictor of concordance due to our sample predominantly consisting of English-speaking patients. Language can be a key factor influencing self-reports and PROs. Previous studies have shown significant discrepancies in the accuracy of EHR language data, emphasizing the need for regular quality assurance to ensure accurate documentation of patients' preferred languages.[54] It is possible that concordance between patient self-reports and HCP documentation may be affected by language differences. Therefore, future research should include language as a factor to better understand its impact on concordance. Finally, a potential area of bias in the HCP clinical notes is that we limited the notes to the clinic appointments one week before or after the date of the symptom self-report questionnaire. It is possible that HCPs documented symptoms at a time outside this window. Similarly, HCPs clinical assessments may have included discussion of symptoms that was not included in the clinical documentation. This limitation may have resulted in greater discordance between the patient–provider perspectives. Future research is needed that can provide a longitudinal assessment of both patient-reported and HCP documentation of symptoms to better understand concordance across time.

Conclusion

PROs are integral in capturing the subjective nature of cancer symptoms, which is critical to various aspects of comprehensive cancer care including reducing symptom burden and personalizing treatment plans. These outcomes have shown to positively influence overall survival rates.[16] [55] In the context of oncology, monitoring of symptoms and side-effects that are associated with treatment toxicities, indicators of the need for dose adjustments, and/or potentially life threatening are routinely assessed.[56] The results from this study suggest that symptoms that are more “constitutional” in nature (such as fatigue, shortness of breath, and sleep disturbance)[48] and in which effective symptom management strategies require more cognitive and behavioral interventions rather than medications are less frequently documented by HCPs. One explanation is that the symptom management care associated with “constitutional” symptoms falls to other members of the health care team such as nurses, social workers, and psychologists that have documentation standards specific to their discipline. It is likely that at times HCPs discuss symptoms during a visit that do not end up being documented in the clinical note. While this practice is may be common and understandable within the context of a busy clinic day, the assessment and management of a wide range of symptoms are considered part of standard clinical practice based on National Comprehensive Cancer Network clinical guidelines and are standard for cancer clinical trials.[57] A core component of assessment and management is documentation. In addition, the assessment and documentation of pain, depressed mood, distress/anxiety, nausea and vomiting, anorexia/loss of appetite, fatigue, diarrhea, shortness of breath, confusion and delirium, and rash have been identified as Cancer-Quality Indicators by the Agency for Healthcare Research and Quality.[58]

The critical importance of PROs in enhancing oncology care quality is evident and has been widely reported, with this study highlighting the challenges in aligning the continued variation between patient symptom self-report experiences and HCP clinical documentation. The moderate concordance rates between patient-reported symptoms and provider-documented symptoms, influenced by factors like gender, education, and cancer stage, points out the need for interventions specifically targeted on improving documentation accuracy. For example, standardized and routine symptom monitoring between visits or at the time of check-in has been shown to promote patient–provider communication and reduce symptom burden. In addition, customized approaches such as data-driven clinical decision support tools, that account for individual variation (e.g., patients with high emetogenic treatment regiments and high multimorbidity) and communication differences could promote HCP to assess for particular symptoms. In addition, it is imperative to pursue future research and develop innovative solutions that ensure patient voices, through PROs, are accurately represented in clinical decision-making, driving forward the agenda for truly patient-centered care, and improved health outcomes in oncology. One such innovative solution is the integration of computer-adaptive testing into EHR to reduce patient burden while maintaining measurement precision, thus enhancing the collection and utilization of PROs.[59]

Clinical Relevance Statement

While the critical role of PROs in enhancing the quality of oncology care is well-established, this study highlights the challenges in aligning patient experiences with clinical documentation, particularly in the context of multimorbidity. The observed discrepancies between patient-reported symptoms and HCP-documented symptoms, influenced by factors like gender, education, and cancer stage, emphasize the need for targeted interventions to enhance the accuracy of clinical documentation. There is clear evidence that when symptom self-report is collected routinely between visits through a PRO system that is integrated into the EHR, there is improved symptom management care.[16] Standard PRO monitoring provides an opportunity for focused assessment and communication between the patient and provider.[60] [61] The burden of gathering PROs and symptom information should not fall exclusively to HCPs. Many EHR systems have existing tools that can be implemented to facilitate patient self-reporting of their symptoms and subsequently making it easier for HCPs to quickly review, identify, and prioritize the critical clinical needs at a given appointment. Deploying these tools requires institutional buy-in resulting from multiple stakeholders seeing value in implementing these tools into their clinic workflow. Until EHR systems routinely include PROs including symptoms,[7] HCPs can use any number of standardized symptom assessment tools such as the MD Anderson Symptom Inventory, which take under 5 minutes to complete.[62] These assessment tools can be given to the patient when they check-in for an appointment to be completed using paper and pencil or on a clinic intake tablet. Routinely, collecting and discussing a range of common symptoms during oncology appointments communicates to the patient that understanding their symptom experience is important.[63] In addition, standardized, routine assessment and documentation of symptoms will also address variations between groups in terms of factors that are associated with higher or lower concordance. Finally, consistent documentation in the EHR can provide the HCP a longitudinal view of the total symptom burden their patients are experiencing thus providing an opportunity for more patient-centered and holistic care.[64]

Multiple-Choice Questions

1. What symptom showed the lowest level of concordance between patient-reported symptoms and provider-documented symptoms in the study?

   • Shortness of breath

   • Pain

   • Swelling

   • Anxiety

   Correct Answer: The correct answer is option c. Concordance in the reporting of 13 symptoms by patients and providers was generally moderate, ranging from 40.4 to 65.7%, with swelling showing the lowest level of agreement at 40.4%.

2. In this study, which factors were found to be significant predictors of high concordance between patient-reported symptoms and provider-documented symptoms in the final logistic regression model?

   • Socioeconomic status and emetogenicity

   • Gender and cancer site

   • Total comorbidity, education, and age

   • Education, gender, and cancer stage

   Correct Answer: The correct answer is option d. We found that gender was a significant predictor of symptom concordance, with women being less likely to have concordant symptom reports. The cancer stage was also a significant predictor of symptom concordance, with each additional cancer stage reducing concordance by 34%. Individuals with a high-school diploma, general education, or some college education had significantly lower concordance compared with participants with an associate or bachelor's degree; however, postgraduate education did not significantly impact concordance.

3. What is the primary takeaway regarding the significance of this study and its implications for cancer care?

   • The study emphasizes the limited impact of PROs on cancer care quality

   • The findings highlight the challenges in aligning patient experiences with clinical documentation, calling for targeted interventions and improved communication

   • The study suggests that effective patient–provider communication is not a significant factor in achieving symptom concordance

   • The study concludes that patient education levels do not play a crucial role in symptom concordance between patients and health care providers

   Correct Answer: The correct answer is option b. This research not only emphasizes the necessity of effective patient–provider communication but also suggests that customized approaches, acknowledging individual experience (patients with high emetogenic treatment regiments and high multimorbidity) and communication differences, could significantly enhance care quality.

Conflict of Interest

None declared.

Protection of Human and Animal Subjects

The study was performed in compliance with the World Medical Association Declaration of Helsinki on Ethical Principles for Medical Research Involving Human Subjects and was reviewed and approved by the Institutional Review Board (IRB approval number: 201805851) at the authors' institution.

• References

• 1 Weldring T, Smith SM. Patient-reported outcomes (PROs) and patient-reported outcome measures (PROMs). Health Serv Insights 2013; 6: 61-68
• 2 Campbell R, Ju A, King MT, Rutherford C. Perceived benefits and limitations of using patient-reported outcome measures in clinical practice with individual patients: a systematic review of qualitative studies. Qual Life Res 2022; 31 (06) 1597-1620
• 3 Graupner C, Kimman ML, Mul S. et al. Patient outcomes, patient experiences and process indicators associated with the routine use of patient-reported outcome measures (PROMs) in cancer care: a systematic review. Support Care Cancer 2021; 29 (02) 573-593
• 4 Mooney K, Berry DL, Whisenant M, Sjoberg D. Improving cancer care through the patient experience: how to use patient-reported outcomes in clinical practice. Am Soc Clin Oncol Educ Book 2017; 37 (37) 695-704
• 5 Caminiti C, Maglietta G, Diodati F. et al. The effects of patient-reported outcome screening on the survival of people with cancer: a systematic review and meta-analysis. Cancers (Basel) 2022; 14 (21) 5470
• 6 Zhang R, Burgess ER, Reddy MC. et al. Provider perspectives on the integration of patient-reported outcomes in an electronic health record. JAMIA Open 2019; 2 (01) 73-80
• 7 Payne TH, Lehmann CU, Zatzick AK. The voice of the patient and the electronic health record. Appl Clin Inform 2023; 14 (02) 254-257
• 8 Asan O, Nattinger AB, Gurses AP, Tyszka JT, Yen TWF. Oncologists' views regarding the role of electronic health records in care coordination. JCO Clin Cancer Inform 2018; 2: 1-12
• 9 Vos JFJ, Boonstra A, Kooistra A, Seelen M, van Offenbeek M. The influence of electronic health record use on collaboration among medical specialties. BMC Health Serv Res 2020; 20 (01) 676
• 10 Sheehan J, Laver K, Bhopti A. et al. Methods and effectiveness of communication between hospital allied health and primary care practitioners: a systematic narrative review. J Multidiscip Healthc 2021; 14: 493-511
• 11 Pakhomov SV, Jacobsen SJ, Chute CG, Roger VL. Agreement between patient-reported symptoms and their documentation in the medical record. Am J Manag Care 2008; 14 (08) 530-539
• 12 Valikodath NG, Newman-Casey PA, Lee PP, Musch DC, Niziol LM, Woodward MA. Agreement of ocular symptom reporting between patient-reported outcomes and medical records. JAMA Ophthalmol 2017; 135 (03) 225-231
• 13 Schwartz C, Winchester DE. Discrepancy between patient-reported and clinician-documented symptoms for myocardial perfusion imaging: initial findings from a prospective registry. Int J Qual Health Care 2021; 33 (02) mzab076
• 14 Lewis AE, Weiskopf N, Abrams ZB. et al. Electronic health record data quality assessment and tools: a systematic review. J Am Med Inform Assoc 2023; 30 (10) 1730-1740
• 15 Marino D, Baratelli C, Guida G. et al. Impact of adoption of patient-reported outcomes in clinical practice on the accuracy of symptom reporting in medical records of cancer patients. Recenti Prog Med 2020; 111 (12) 740-748
• 16 Basch E, Deal AM, Dueck AC. et al. Overall survival results of a trial assessing patient-reported outcomes for symptom monitoring during routine cancer treatment. JAMA 2017; 318 (02) 197-198
• 17 Atkinson TM, Ryan SJ, Bennett AV. et al. The association between clinician-based common terminology criteria for adverse events (CTCAE) and patient-reported outcomes (PRO): a systematic review. Support Care Cancer 2016; 24 (08) 3669-3676
• 18 Rammant E, Ost P, Swimberghe M. et al. Patient- versus physician-reported outcomes in prostate cancer patients receiving hypofractionated radiotherapy within a randomized controlled trial. Strahlenther Onkol 2019; 195 (05) 393-401
• 19 Behroozian T, Milton L, Zhang L. et al. How do patient-reported outcomes compare with clinician assessments? A prospective study of radiation dermatitis in breast cancer. Radiother Oncol 2021; 159: 98-105
• 20 Lam E, Yee C, Wong G. et al. A systematic review and meta-analysis of clinician-reported versus patient-reported outcomes of radiation dermatitis. Breast 2020; 50: 125-134
• 21 Bierbaum M, Rapport F, Arnolda G. et al. Clinicians' attitudes and perceived barriers and facilitators to cancer treatment clinical practice guideline adherence: a systematic review of qualitative and quantitative literature. Implement Sci 2020; 15 (01) 39
• 22 Lisy K, Kent J, Piper A, Jefford M. Facilitators and barriers to shared primary and specialist cancer care: a systematic review. Support Care Cancer 2021; 29 (01) 85-96
• 23 Saeidzadeh S, Perkhounkova Y, Gilbertson-White S, Cherwin CH. The influence of multiple chronic conditions on symptom clusters in people with solid tumor cancers. Cancer Nurs 2022; 45 (01) E279-E290
• 24 Zeinali N, Albashayreh A, Fan W, White SG. Symptom-BERT: enhancing cancer symptom detection in EHR clinical notes. J Pain Symptom Manage 2024; 68 (02) 190-198.e1
• 25 Chang VT, Hwang SS, Thaler HT, Kasimis BS, Portenoy RK. Memorial symptom assessment scale. Expert Rev Pharmacoecon Outcomes Res 2004; 4 (02) 171-178
• 26 Portenoy RK, Thaler HT, Kornblith AB. et al. The Memorial Symptom Assessment Scale: an instrument for the evaluation of symptom prevalence, characteristics and distress. Eur J Cancer 1994; 30A [phone]
• 27 Topaz M, Murga L, Bar-Bachar O, McDonald M, Bowles K. NimbleMiner: an open-source nursing-sensitive natural language processing system based on word embedding. Comput Inform Nurs 2019; 37 (11) 583-590
• 28 Albashayreh A, Bandyopadhyay A, Zeinali N, Zhang M, Fan W, Gilbertson White S. Natural language processing accurately differentiates cancer symptom information in electronic health record narratives. JCO Clin Cancer Inform 2024; 8: e2300235
• 29 Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987; 40 (05) 373-383
• 30 Chaudhry S, Jin L, Meltzer D. Use of a self-report-generated Charlson Comorbidity Index for predicting mortality. Med Care 2005; 43 (06) 607-615
• 31 Akaike H. Information theory and an extension of the maximum likelihood principle. In: Parzen E, Tanabe K, Kitagawa G. eds. Selected Papers of Hirotugu Akaike. Springer New York: 1998: 199-213
• 32 Kutner MH. Applied Linear Statistical Models. McGraw-Hill Irwin; 2005
• 33 Harrell FE. Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis. Springer: 2001
• 34 Wasserstein RL, Lazar NA. The ASA statement on p-values: context, process, and purpose. Am Stat 2016; 70 (02) 129-133
• 35 Burnham KP, Anderson DR. Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach. Springer New York: 2003
• 36 Shreffler J, Huecker MR. Type I and Type II Errors and Statistical Power. StatPearls; 2024
• 37 Portet S. A primer on model selection using the Akaike Information Criterion. Infect Dis Model 2020; 5: 111-128
• 38 Cleeland CS, Sloan JA. ASCPRO Organizing Group. Assessing the symptoms of cancer using patient-reported outcomes (ASCPRO): searching for standards. J Pain Symptom Manage 2010; 39 (06) 1077-1085
• 39 Mathew A, Doorenbos AZ, Vincent C. Symptom management theory: analysis, evaluation, and implications for caring for adults with cancer. ANS Adv Nurs Sci 2021; 44 (03) E93-E112
• 40 Basch E, Iasonos A, McDonough T. et al. Patient versus clinician symptom reporting using the National Cancer Institute Common Terminology Criteria for Adverse Events: results of a questionnaire-based study. Lancet Oncol 2006; 7 (11) 903-909
• 41 Deshields TL, Penalba V, Arroyo C. et al. The relationship between response style and symptom reporting in cancer patients. Support Care Cancer 2023; 31 (05) 312
• 42 Ritchie CS, Zhao F, Patel K. et al. Association between patients' perception of the comorbidity burden and symptoms in outpatients with common solid tumors. Cancer 2017; 123 (19) 3835-3842
• 43 Oertelt-Prigione S, de Rooij BH, Mols F. et al. Sex-differences in symptoms and functioning in >5000 cancer survivors: results from the PROFILES registry. Eur J Cancer 2021; 156: 24-34
• 44 Bourgeois FC, Hart NJ, Dong Z. et al. Partnering with patients and families to improve diagnostic safety through the OurDX tool: effects of race, ethnicity, and language preference. Appl Clin Inform 2023; 14 (05) 903-912
• 45 Aelbrecht K, Hanssens L, Detollenaere J, Willems S, Deveugele M, Pype P. Determinants of physician-patient communication: the role of language, education and ethnicity. Patient Educ Couns 2019; 102 (04) 776-781
• 46 Hutchings H, Behinaein P, Enofe N. et al. Association of social determinants with patient-reported outcomes in patients with cancer. Cancers (Basel) 2024; 16 (05) 1015
• 47 Zhu S, Lei C. Association between marital status and all-cause mortality of patients with metastatic breast cancer: a population-based study. Sci Rep 2023; 13 (01) 9067
• 48 Kroenke K, Lam V, Ruddy KJ. et al. Prevalence, severity, and co-occurrence of SPPADE symptoms in 31,866 patients with cancer. J Pain Symptom Manage 2023; 65 (05) 367-377
• 49 Esther Kim JE, Dodd MJ, Aouizerat BE, Jahan T, Miaskowski C. A review of the prevalence and impact of multiple symptoms in oncology patients. J Pain Symptom Manage 2009; 37 (04) 715-736
• 50 Gilbertson-White S, Aouizerat BE, Jahan T. et al. Determination of cutpoints for low and high number of symptoms in patients with advanced cancer. J Palliat Med 2012; 15 (09) 1027-1036
• 51 Neuner-Jehle S, Zechmann S, Grundmann Maissen D, Rosemann T, Senn O. Patient-provider concordance in the perception of illness and disease: a cross-sectional study among multimorbid patients and their general practitioners in Switzerland. Patient Prefer Adherence 2017; 11: 1451-1458
• 52 Zulman DM, Kerr EA, Hofer TP, Heisler M, Zikmund-Fisher BJ. Patient-provider concordance in the prioritization of health conditions among hypertensive diabetes patients. J Gen Intern Med 2010; 25 (05) 408-414
• 53 Calderón-Larrañaga A, Vetrano DL, Onder G. et al. Assessing and measuring chronic multimorbidity in the older population: a proposal for its operationalization. J Gerontol A Biol Sci Med Sci 2017; 72 (10) 1417-1423
• 54 Rajaram A, Thomas D, Sallam F, Verma AA, Rawal S. Accuracy of the preferred language field in the electronic health records of two Canadian hospitals. Appl Clin Inform 2020; 11 (04) 644-649
• 55 Gotay CC, Kawamoto CT, Bottomley A, Efficace F. The prognostic significance of patient-reported outcomes in cancer clinical trials. J Clin Oncol 2008; 26 (08) 1355-1363
• 56 Cleeland CS, Zhao F, Chang VT. et al. The symptom burden of cancer: evidence for a core set of cancer-related and treatment-related symptoms from the Eastern Cooperative Oncology Group Symptom Outcomes and Practice Patterns study. Cancer 2013; 119 (24) 4333-4340
• 57 Basch E, Reeve BB, Mitchell SA. et al. Development of the National Cancer Institute's patient-reported outcomes version of the common terminology criteria for adverse events (PRO-CTCAE). J Natl Cancer Inst 2014; 106 (09) dju244
• 58 Dy SM, Lorenz KA, O'Neill SM. et al. Cancer quality-ASSIST supportive oncology quality indicator set: feasibility, reliability, and validity testing. Cancer 2010; 116 (13) 3267-3275
• 59 Nolla K, Rasmussen LV, Rothrock NE. et al. Seamless integration of computer-adaptive patient reported outcomes into an electronic health record. Appl Clin Inform 2024; 15 (01) 145-154
• 60 Cella D, Garcia SF, Cahue S. et al. Implementation and evaluation of an expanded electronic health record-integrated bilingual electronic symptom management program across a multi-site Comprehensive Cancer Center: the NU IMPACT protocol. Contemp Clin Trials 2023; 128: 107171
• 61 Mooney K, Dumas K, Fausett A, Littledike M, Ulrich CM, Titchener K. Oncology hospital at home in rural communities: the Huntsman at Home rural experience. J Clin Oncol 2022; 40 (16, Suppl): 1535
• 62 Cleeland CS, Mendoza TR, Wang XS. et al. Assessing symptom distress in cancer patients: the M.D. Anderson Symptom Inventory. Cancer 2000; 89 (07) 1634-1646
• 63 Conley CC, Otto AK, McDonnell GA, Tercyak KP. Multiple approaches to enhancing cancer communication in the next decade: translating research into practice and policy. Transl Behav Med 2021; 11 (11) 2018-2032
• 64 Kang D, Kim S, Kim H. et al. Surveillance of symptom burden using the patient-reported outcome version of the common terminology criteria for adverse events in patients with various types of cancers during chemoradiation therapy: real-world study. JMIR Public Health Surveill 2023; 9: e44105

Publikationsverlauf

Eingereicht: 02. April 2024

Angenommen: 16. September 2024

Artikel online veröffentlicht:
25. Dezember 2024