Application of Forcing Functions to Electronic Health Records Is Associated With Improved Pain Control for Patients Undergoing Radiation Therapy for Bone Metastases

Abstract

Pain is a common problem for patients undergoing radiation therapy, exacerbated by inconsistent pain documentation. Free-form templates, pain score prompts, and forcing functions are a hierarchy of constraint systems that can be applied to data entry. This study assessed the impact of incorporating these models into electronic health records on pain documentation rates during 450 on-treatment visits and pain severity of 258 patients with bone metastases and breast and thoracic cancer during radiation therapy. Pain documentation is associated with more robust constraint systems: free form (0.11, 95% CI [0.07, 0.18]), pain score prompts (0.87, 95% CI [0.81, 0.92]), and forcing functions (0.97, 95% CI [0.93, 0.99]). Forcing functions also were associated with improved pain control over the course of radiation treatment for bone metastases compared with pain score prompts (P = .026, nonparametric Kruskal-Wallis). Use of forcing functions correlates with increased pain documentation rates, which contributes to improved pain management.

Keywords

pain documentation forcing functions radiation therapy quality

Pain is a common symptom experienced by cancer patients during the course of treatment. Factors that contribute to the experience of pain in a cancer patient include the type of cancer, pain management, and the cancer treatment being used. In radiation oncology, the prevalence of pain is commonplace. Radiation is often used to palliate cancer pain but also may precipitate pain by way of side effects induced by radiation therapy.¹ For instance, patients with head and neck, breast, and thoracic cancers are more likely to experience pain after initiation of curative treatment because of skin reaction or dysphagia.² Therefore, optimizing pain management is a critical part of providing high-quality care to patients receiving radiation therapy. One strategy for radiation-related pain follows the World Health Organization analgesic ladder, which provides a stepwise protocol for dosing and prescribing medications.³ Quality metrics that help monitor the use of these standards are provided by the National Quality Forum (NQF) and the American Society of Clinical Oncology’s (ASCO) Quality Oncology Practice Initiative. These metrics monitor documentation rates of pain as an indication of quality of care. However, methods of documentation vary widely in radiation oncology. This can be attributed in part to the diversity of electronic medical record (EMR) systems that give health care providers a large degree of freedom and discretion in charting their patient visits. Lack of a robust standardized method for pain documentation can lead to inconsistencies in pain management. Challenges that have been noted include need for EMR reminders and templates, documentation of pain believed to be unrelated to cancer, and ill-defined descriptors of pain severity.^4,5 As such, opportunities remain to develop a system that captures the experience of pain more consistently during radiation therapy.

Application of constraint systems, such as forcing functions, is one method to increase provider compliance with recording a patient’s pain response. Broadly, forcing functions—traditionally employed in human factors engineering or systems design—help prevent human error.⁶ In health care, forcing functions are used to reduce medical errors by ensuring that necessary actions are completed, as well as preventing them from being performed out of order.⁷ They are often seen in systems that have clinically important default settings, such as pharmacy script-writing software.⁸ In this study, constraint systems are installed as EMR templates for pain score documentation as a model for improving oncologic pain management. A response, qualitative or quantitative, is required before the physician note can be closed during the treatment visit. Forcing functions represent the most robust form of constraint system.

At the study institution, significant changes in the EMR system have occurred over the past several years. An early constraint model was adopted in the legacy information management software in 2014, which introduced prepopulated prompts, including one for pain documentation. In the past 2 years, the study institution underwent a hospital-wide transition to the Epic EMR system (Epic Systems Corporation, Verona, Wisconsin), which uses a more robust forcing function for pain documentation. Given these 2 major EMR system changes, the research team sought to analyze the results of these changes in regard to physician documentation of pain and whether improved documentation had an effect on pain scores over the course of radiation therapy. The team hypothesized that increasing prompts and forcing functions would improve documentation rates and that this improvement would lead to improved pain management for both patients undergoing radiation for pain palliation and those for whom radiation would be expected to induce pain secondary to treatment side effects.

Methods

Comparisons of Pain Documentation Rates Across Different EMR Constraint Systems

This is an Institutional Review Board-approved retrospective observational study. From 2014 to 2017, a total of 450 on-treatment visits (OTVs) were randomly selected from 2 EMR software systems using 3 distinct EMR pain documentation models. OTVs are performed at least weekly during radiation therapy as part of routine clinical practice. The EMR software studied included MOSAIQ (Elektra AB, Stockholm, Sweden), a care management software curated for radiation oncology practice, and Epic, which the study institution transitioned to in 2016. The pain documentation systems included a free-form template, a pain score prompt, and a forcing function system. The free-form template allowed documentation of pain at the physician’s discretion. The pain prompt template used a numerical or descriptive indicator of pain, but there was an option to leave the response unanswered. The forcing function system required a numerical or descriptive pain score to complete the patient EMR chart. A total of 150 OTVs were randomly selected from August 2014 to November 2014, which corresponded to a free-form MOSAIQ pain documentation system; 150 OTVs were randomly selected from January 2015 to March 2015, which corresponded to a pain prompt MOSAIQ constraint system; and 150 OTVs were randomly selected from November 2016 to December 2017, which corresponded to the Epic forcing function system. The random selection of OTVs for pain documentation rates was performed so that different OTVs from the same patient can be selected. This would account for the possibility that some patients can have pain scores recorded at some OTVs and not others during the course of their treatment. The 450 total OTVs were from 331 unique patients.

Rate of pain documentation was calculated from the completion or description of pain scores in the patient EMR chart during OTVs. Patients receiving treatment who did not have recorded OTVs were excluded from this study. Patients whose treatment course overlapped the implementation of the pain prompt template, between November 2014 and January 2015, were not excluded from the study. The change in constraint systems from free form to pain prompt did not affect the selection of the OTVs, which were chosen prior to November 2014. Descriptive indicators of pain from both MOSAIQ and Epic were converted to a numerical value for statistical analysis following a conversion chart native to the MOSAIQ program (Table 1).

Table 1.

MOSAIQ On-Treatment Visit Pain Assessment Conversion Chart.^a

Subjective Pain Score	Corresponding Numerical Pain Score
No pain	0
Between no pain and mild pain	1
Mild pain	2
Between mild and moderate pain	3
Moderate pain	4
Between moderate and severe pain	5
Severe pain	6
Between severe pain and very severe pain	7
Very severe pain	8
Between very severe pain and worse possible pain	9
Worst possible pain	10

A conversion chart native to the MOSAIQ electronic medical record (EMR) software that converts a subjective answer to pain during an on-treatment visit to a numerical pain score. This conversion formula is used in the study to convert subjective pain scores of patients who were documented in both MOSAIQ and Epic EMR systems.

Change in Pain Severity During Radiation Therapy for Specific Oncologic Diagnoses

From January 2016 to December 2016, 129 patients charted in MOSAIQ with specific cancer diagnoses were randomly generated. These diagnoses were breast (n = 39), thoracic (n = 54), and secondary bone metastatic (n = 36) cancers. From January 2017 to December 2017, 129 patients charted in Epic with the same cancer diagnoses and distribution of patients per diagnosis were randomly generated. Because of the flexibility of documenting diagnoses, including the addition of multiple diagnoses, it was difficult to design a query that could identify every patient who fit the inclusion criteria. An additional manual search was performed to capture patients who fit the criteria but were not selected by the query. For these 258 patients, pain scores from the first and last OTVs were recorded. Change in pain severity over the course of treatment was measured as the difference in pain score between the first and last OTVs. Patients without documented pain scores during the course of their radiation therapy were not included in the study of change in pain severity during treatment.

Patients were randomly selected using a structured query language code or were generated randomly using Excel software. Statistical analysis for this study was performed using R studio (R Foundation for Statistical Computing, Vienna, Austria).

Results

Comparisons of Pain Documentation Rates Across Different EMR Constraint Systems

Pain scores from 450 OTVs were sampled from 2014 to 2017 and were divided equally into 3 groups. Each sampling group (n = 150) represented pre implementation or post implementation of the pain score prompt in MOSAIQ and post implementation of a forcing function in Epic, respectively. Implementation of prompts and forcing functions for pain scores increased the likelihood that patients will be asked about their experience of pain during radiation therapy at the study institution (Figure 1). The rate of pain documentation before implementation of the prompt in MOSAIQ was 0.11 (95% CI: 0.07, 0.18). The rate of pain documentation afterward was 0.87 (95% CI: 0.81, 0.92). After transition to Epic, the rate of pain documentation was 0.97 (95% CI: 0.93, 0.99). There was a statistically significant different rate of documentation among the 3 groups (P < .001, Fisher exact test). Paired comparisons between groups showed differences in documentation rates as well (P < .001 for 2014 vs 2015; P = .002 for 2015 vs 2017; P < .001 for 2014 vs 2017; post hoc Fisher exact test). The rate of pain documentation refers to the percentage of OTVs during which pain was documented or described. There is a statistically significant increase in pain documentation over the study period, which corresponds to the adoption of increasingly robust constraint systems.

Figure 1.

Pain documentation rates: a comparison of pain documentation rates using different levels of constraint systems. Error bars show 95% CI.

Change in Pain Severity During Radiation Therapy for Specific Oncologic Diagnoses

Forcing functions also play a role in pain symptom control during palliative treatment of advanced or metastatic disease. A comparison of the distribution of pain scores by diagnosis before and after Epic implementation is shown in Figure 2. Each histogram plots the change in pain score after radiation therapy. Distribution of the difference in pain scores during treatment, correlating to either an increase or decrease in patient-reported pain, is shown in Figure 3. These box plots show outlier observations by cancer diagnosis and EMR constraint system used. Transition to the Epic EMR with an imbedded pain score forcing function was associated with a decrease in pain over the course of radiotherapy for bone metastases (P = .026, nonparametric Kruskal-Wallis test). Descriptive statistics, by cancer diagnosis and EMR constraint system used, are shown in Table 2. There was no statistical significance in the experience of pain for curative treatment of breast cancer (P = .657) or thoracic cancers (P = .095). Summary statistics for pre to post treatment differences in pain score are shown in Table 3.

Figure 2.

Change in pain score after radiation therapy: these histograms show the distribution of changes in pain after radiation therapy. Pain score prompts were used in 2016, and forcing functions were implemented in 2017. A positive change in pain score indicates an increase in pain after therapy. A negative change in pain score indicates a decrease in pain after therapy. Zero indicates no change in pain after therapy.

Figure 3.

Box plots by cancer diagnosis: these box plots show the distribution of pain scores by cancer diagnosis. A positive pain score difference correlates to a patient-reported increase in pain during treatment. A negative pain score difference correlates to a patient-reported decrease in pain during treatment. The x-axis represents the year of treatment, which corresponds to the electronic medical record constraint system used. Pain score prompts were used in 2016, and forcing functions were implemented in 2017.

Table 2.

Pain Score Documentation Rate by Electronic Medical Record Constraint System.^a

Constraint System Used	Number of On-Treatment Visit (OTVs)	Number of OTVs With Pain Score Documentation	Percentage of Pain Score Documentation (Mean [95% CI])	Standard Deviation
Free form (MOSAIQ)	150	17	0.11 (0.07, 0.18)	0.02
Pain score prompt (MOSAIQ)	150	131	0.87 (0.81, 0.92)	0.05
Forcing function (Epic)	150	145	0.97 (0.93, 0.99)	0.01

This table provides a comparison of the difference in pain documentation rates during on-treatment visits between 3 separate electronic medical record constraint systems.

Table 3.

Summary Statistics for Pre to Post Treatment Differences in Pain Score.^a

Year	Cancer Diagnosis	Minimum	Q1	Median	Mean	Q3	Maximum	SD
2016	Bone	−5	−1.25	0	−0.28	0	8	2.54
2017	Bone	−10	−4	−1.5	−2.19	0	2	2.92
2016	Breast	−5	0	0	1.10	2	10	2.78
2017	Breast	−9	0	0	0.92	3.5	8	3.28
2016	Thoracic	−5	0	0.5	1.26	2.75	7	2.49
2017	Thoracic	−8	0	0	0.48	2	10	3.29

This table provides descriptive statistics for pain score improvement by cancer diagnosis and year. Year 2016 data represent pain scores recorded when prompts were used in electronic medical records. Year 2017 data represent pain scores recorded when forcing functions were implemented.

Discussion

Cancer or cancer therapy-induced pain is typically responsive to treatment regimens, with up to 70% of patients well controlled on individualized pain management strategies.⁹ Similarly, palliative care programs have been validated for effective pain control of advanced cancer disease.¹⁰ Despite available treatment, significant oncologic pain is still reported in up to 50% of patients receiving care, with moderate to severe pain seen in 38% of all cancer patients.^5,11 From the patient perspective, the most common reason given for not taking analgesics for disease or treatment is health care providers not prescribing medications.^12,13 From the physician perspective, a majority of radiation oncologists acknowledged that poor pain assessment was a barrier to pain control.¹⁴ The discrepancy between the documented effectiveness of pain control regimens and patient experience of pain in the clinical setting are a reflection of the complexity of cancer pain management.

Patient-reported quality of care has been shown to be significantly correlated with severity of pain, which is independently associated with emotional distress.^15,16 The experience of extended severe pain also is related to a decline in the quality of life. In totality of these issues, NQF and ASCO adopted pain management documentation as an important quality metric in oncology. Even so, methods of pain management documentation currently vary widely from institution to institution. A review of 19 National Comprehensive Cancer Network (NCCN) centers indicated that documentation of pain also was associated with the type of provider; nurses and pain specialists were more likely to ask patients about their pain than physicians.¹⁷ A study of colorectal cancer patients at NCCN centers indicated that incomplete documentation played a significant role in cases of nonconcordance to ASCO quality guidelines.¹⁸ Lack of a standard of documentation may be affecting the ability of physicians and other health care providers to monitor patient pain experience consistently.

This study shows that constraint systems, specifically forcing functions, can be an effective method to improve pain documentation rates. Transition from the study institution’s legacy free-form EMR templates to prepopulated prompts yielded a 76% increase in documentation of pain score records. The subsequent transition to a forcing function in Epic resulted in an additional 10% increase in documentation rate. A forcing function simplifies the interface between the EMR and the physician while eliminating the ambiguity provided by a free-form template. This combination of factors ensures that a pain score will be documented consistently during a patient encounter; the health care provider would have to manually delete the prompt in order to complete the note.

The addition of a forcing function to evaluate pain also was associated with improved pain control for bone metastases over the course of palliative radiotherapy. Bone metastases cause moderate to severe pain more frequently compared to other cancer diagnoses.¹⁹ This study examined pain score outcomes from 4 years of treatment. During this time frame, there were no significant changes to treatment protocols or other aspects of care delivery. There also were no new initiatives that otherwise would significantly affect the institution’s standard of care for radiation therapy. In addition, box plot representation of the data does not suggest that the statistically significant difference is related to a small number of outlier observations (Figure 3). For these patients, improvement in pain control likely is related to consistent reporting of pain scores that allow for better titration of adjunct analgesia during palliative treatment. Improvement in the clinical picture of cancer patients where the primary goal is curative treatment is less clear. Increasing pain documentation rates was not associated with a direct decrease in prevalence of pain for breast and thoracic cancer patients. As symptoms such as skin irritation and dysphagia related to radiation are anticipated, there may already have been proactive practices in place to manage these side effects. Because irradiation therapy damages healthy tissue and causes skin inflammation and treatment site pain, control of pain in these patients requires a more complex organization of therapeutic strategies.²⁰ Increasing pain documentation alone was not enough to significantly affect the prevalence of pain in these patients.

This study is limited by its retrospective design and patient population. Lack of patient input on the experience of pain during radiation therapy restricts the scope of this study and should be assessed prospectively in future studies. In addition, some patients who received radiation treatment had incomplete pain documentation and subsequently were excluded from the study. Selecting against this group of patients, across all 3 methods of pain documentation, may result in over- or underestimation of documentation rates. This may present as a potential source of bias.

Although this study focuses on cancer patients receiving palliative and curative radiation therapy, forcing functions also are applicable to patients receiving other modalities of cancer treatment, including surgery and chemotherapy. In particular, for patients receiving combination therapy, proper pain documentation will provide physicians with the information to better develop a proper analgesic strategy for each patient.

Conclusions

Forcing functions represent a useful platform as a standard for pain documentation that can be incorporated into native EMR systems and designed to become a routine question during a cancer patient’s treatment visit for radiation therapy. It provides a simple and effective method to significantly increase the rate of pain documentation. When the primary goal of treatment is pain relief, such as in patients with bone metastases, implementation of a forcing function is significantly associated with a decrease in severity of pain. As patient experience of pain continues to garner attention as an important quality metric, a standardized system of pain documentation should be strongly considered as a part of institutional strategy.

Footnotes

Authors’ Note

This article was presented as an abstract at the American Society for Radiation Oncology’s 60th Annual Meeting in San Antonio, Texas, from October 21 to October 24, 2018.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The statistical analysis of the manuscript was funded by a National Institutes of Health/National Cancer Institute core grant (P30 CA0 056036-10).

ORCID iD

Denis Huang

References

Pignon

Fernandez

Ayasso

Durand

Badinand

Cowen

Impact of radiation oncology practice on pain: a cross-sectional survey. Int J Radiat Oncol Biol Phys. 2004;60:1204-1210.

van den Beuken-van Everdingen

MHJ

de Rijke

Kessels

Schouten

van Kleef

Patijn

High prevalence of pain in patients with cancer in a large population-based study in The Netherlands. Pain. 2007;132:312-320.

Prommer

EE.

Pharmacological management of cancer-related pain. Cancer Control. 2015;22:412-425.

Ranpura

Agrawal

Chokshi

, et al. Improving documentation of pain management at MedStar Washington Cancer Institute. J Oncol Pract. 2015;11:155-157.

van den Beuken-van Everdingen

MHJ

Hochstenbach

LMJ

Joosten

EAJ

Tjan-Heijnen

VCG

Janssen

DJA.

Update on prevalence of pain in patients with cancer: systematic review and meta-analysis. J Pain Symptom Manage. 2016;51:1070.e9-1090.e9.

Siewert

Hochman

MG.

Improving safety through human factors engineering. RadioGraphics. 2015;35:1694-1705.

Woodward

Mytton

Lemer

, et al. What have we learned about interventions to reduce medical errors? Annu Rev Public Health. 2010;31:479-497.

Kohn

Corrigan

Donaldson

, eds. To Err Is Human: Building a Safer Health System. Washington, DC: National Academies Press; 2000.

Goisis

Gorini

Ratti

Luliri

Application of a WHO protocol on medical therapy for oncologic pain in an internal medicine hospital. Tumori. 1989;75:470-472.

10.

Zech

DFJ

Grond

Lynch

Hertel

Lehmann

. Validation of World Health Organization Guidelines for cancer pain relief: a 10-year prospective study. Pain. 1995;63:65-76.

11.

Neufeld

Elnahal

Alvarez

RH.

Cancer pain: a review of epidemiology, clinical quality and value impact. Future Oncol. 2017;13:833-841.

12.

Simone

Vapiwala

Hampshire

Metz

JM.

Cancer patient attitudes toward analgesic usage and pain intervention. Clin J Pain. 2012;28:157-162.

13.

Tse

MMY

Wong

ACF

Lee

Chong

Leung

. The effect of a pain management program on patients with cancer pain. Cancer Nurs. 2012;35:438-446.

14.

Cleeland

Janjan

Scott

Seiferheld

Curran

WJ.

Cancer pain management by radiotherapists: a survey of radiation therapy oncology group physicians. Int J Radiat Oncol Biol Phys. 2000;47:203-208.

15.

Martinez

Snyder

Malin

SM.

Patient-reported quality of care and pain severity in cancer. Palliat Support Care. 2015;13:875-884.

16.

O’Connor

Weir

Butcher

, et al. Pain in patients attending a specialist cancer service: prevalence and association with emotional distress. J Pain Symptom Manage. 2012;43:29-38.

17.

Weinstein

Romanus

Lepisto

, et al. Documentation of pain in comprehensive cancer centers in the United States: a preliminary analysis. J Natl Compr Cancer Netw. 2004;2:173-180.

18.

Romanus

Weiser

Skibber

, et al. Concordance with NCCN colorectal cancer guidelines and ASCO/NCCN quality measures: an NCCN institutional analysis. J Natl Compr Canc Netw. 2009;7:895-904.

19.

Di Maio

Gridelli

Gallo

, et al. Prevalence and management of pain in Italian patients with advanced non-small-cell lung cancer. Br J Cancer. 2004;90:2288-2296.

20.

Gewandter

Walker

Heckler

Morrow

Ryan

JL.

Characterization of skin reactions and pain reported by patients receiving radiation therapy for cancer at different sites. J Support Oncol. 2013;11:183-189.