Volume 39, Issue 2 , Pages 180-185, February 2010
High Outpatient Pain Intensity Scores Predict Impending Hospital Admissions in Patients with Cancer
Article Outline
Abstract
Context
Pain intensity scores (PIS) are frequently collected in the outpatient setting. The implications for patients with high PIS have not been well-studied.
Objectives
This retrospective review was designed to determine whether high outpatient encounter PIS identify patients at risk of hospital admission.
Methods
Numerical PIS (0–10) were collected from all outpatient medical and radiation oncology encounters at the Johns Hopkins Comprehensive Cancer Center from 2004 to 2006. These were merged with an inpatient database to identify admissions occurring within 30 days of the outpatient encounter. PIS were categorized as 0–3 (mild), 4–6 (moderate), and 7–10 (severe). Odds ratios for hospital admission were calculated using generalized estimating equations.
Results
Of 119,069 encounters, 116,713 (98%) were evaluable, and 5,089 encounters (4.5%) had PIS of 7–10. Twenty-nine percent of these high PIS encounters had hospital admissions within 30 days. Encounters with PIS of 7–10 and 4–6 were 96% and 43%, respectively, more likely to result in hospital admission within 30 days compared with encounters with PIS
<4 (P<0.001). Hospital admission rates after encounters with PIS of 7–10 were highest in patients with melanoma (58%), sarcoma (42%), female genital cancer (39%), and upper aerodigestive (36%) cancer.
Conclusion
Outpatients with cancer and high PIS are at increased risk of hospital admission within 30 days. This high-risk group should be targeted for early supportive care interventions aimed at reducing hospitalizations and improving quality of life.
Key Words: Outpatient, cancer pain, hospitalizations, elderly
Introduction
Significant pain occurs in over two-thirds of patients with cancer and is more frequent and severe as the cancer progresses.1, 2, 3 During the past several decades, genuine advances have been made in the evaluation and therapy of cancer pain. As a result, the vast majority of pain in cancer patients can be well controlled with opioid analgesics and adjuvant medications, coupled with local radiation and anesthetic and surgical approaches when indicated.4 Acupuncture, guided imagery, and distraction can also be helpful in selected situations. Screening tools to detect clinically significant pain have now been widely adopted, and algorithms for treating cancer-related pain are based on the documented pain intensity.5, 6 Currently, the Joint Commission on Accreditation of Healthcare Organizations and guidelines from the National Comprehensive Cancer Network and American Pain Society have led to the routine assessment and documentation of quantitative pain measurements to ensure that patients receive optimal pain control.4, 6, 7
In the Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins, quantitative monitoring of pain intensity scores (PIS) are recorded for each outpatient visit in an electronic database and are used to identify patients with suboptimal analgesia. We demonstrated that this approach was as feasible in a busy community-based outpatient oncology practice as it was in our university-based cancer center.8
An earlier study that we conducted suggested that outpatients with high pain scores were more frequently admitted to the hospital than those with lower pain scores.9 This retrospective study was conducted to confirm these findings and explore how the roles of age, gender, or diagnosis might influence this relationship. The overall objective of this research was to determine if we could identify patients at high risk of hospital admission using outpatient PIS so that future efforts could be directed to provide additional pain and supportive care interventions to reduce hospitalizations and to improve quality of life in this patient population.
Methods
Numerical PIS were obtained with each outpatient visit at the Johns Hopkins Comprehensive Cancer Center. Patients were asked to rate their current level of pain on a scale of 0–10 at the time of the clinic visit, and scores were recorded by the medical assistants in an electronic database. PIS from all outpatient medical and radiation oncology encounters from January 1, 2004, to December 31, 2006 were merged with an inpatient database to screen for admissions that occurred within 30 days of the outpatient encounter. Unkept appointments, absent pain scores, and invalid pain scores were excluded. Pain scores were deemed invalid if the number recorded was greater than 10. If a patient had multiple encounters within the 30 days before a hospitalization, each of these encounters was coded as resulting in a hospital admission. Diagnoses from each encounter were obtained from the outpatient database and assigned to one of 30 categories as outlined by Clinical Classifications Software (Agency for Healthcare Research and Quality, Rockville, MD) for International Classification of Diseases-10 data and eventually collapsed into 15 broader categories for analysis. Institutional review board approval was obtained to conduct this retrospective review.
Statistical Considerations
Descriptive statistics (mean, standard deviation, frequency, and percents) were used to summarize age, gender, and admissions for patients and encounters. Pain scores were categorized into 0–3 (mild), 4–6 (moderate), and 7–10 (severe), and demographics were compared across pain categories. Generalized estimating equations to account for within-patient correlation were used to obtain odds ratios (ORs) and 95% confidence intervals (CIs) for hospital admission within 30 days for pain categories. Multivariable models, adjusting for age, gender, and diagnosis, were also constructed. Analysis was performed using SAS version 9.2 (SAS Institute Inc., Cary, NC). All reported P-values were two-sided.
Results
A total of 119,069 outpatient visits from 10,910 individual patients were assessed. Of these, 116,713 encounters (98.0%) contained recorded outpatient PIS and were considered evaluable (Fig. 1). Demographic characteristics and hospital admissions for the patient's first encounter and all encounters are presented in Table 1. PIS were found to be mild (PIS: 0–3) in 86.6% of these encounters, moderate (PIS: 4–6) in 8.9%, and high (PIS: 7–10) in 4.5%. A greater proportion of female patients were in the higher pain groups (P<0.001). There was a small, though highly significant (P<0.001), association between age and pain category, with younger patients having greater pain. Admission to Johns Hopkins Hospital (JHH) within 30 days occurred in 14.3% of encounters of patients with low PIS, 22.8% with moderate PIS, and 29.2% with high PIS.
Fig. 1.
Flow diagram for outpatient encounters. Med Onc
=medical oncology; Rad Onc=radiation oncology.
Table 1. Demographic and Clinical Characteristics of Outpatient Encounters (n=116,713)—Overall and Across Pain Categoriesa
| Characteristic | Pain Intensity Score | |||
|---|---|---|---|---|
| 0–3 | 4–6 | 7–10 | All | |
| By encounter | ||||
| Gender | ||||
| Male | 56,425 (87.69) | 5,405 (8.40) | 2,516 (3.91) | 64,346 (55.14) |
| Female | 44,663 (85.30) | 5,028 (9.60) | 2,667 (5.09) | 52,358 (44.86) |
| Age, in years | 58.48 (13.98) | 56.86 (13.29) | 57.01 (13.65) | 58.27 (13.92) |
| Hospital admission | ||||
| None within 30 days | 86,662 (88.08) | 8,054 (8.19) | 3,671 (3.73) | 98,387 (84.30) |
| Within 30 days | 14,435 (78.77) | 2,379 (12.98) | 1,512 (8.25) | 18,326 (15.70) |
aFrequencies (percents) or means (standard deviations) are presented in parentheses. |
Encounters with PIS of 4–6 were 43% more likely, and PIS of 7–10 were almost twice as likely to result in a hospital admission to JHH within 30 days compared with encounters with PIS less than 4 (P<0.001) (Table 2). Gender was not predictive of hospital admission but was retained in the multivariable model because of its association with pain score. Age was significantly associated with hospital admission; for each 10-year increase in age, patients were 0.5% less likely to be hospitalized after an encounter (Table 2). We tested the interaction of age and gender with pain score, but neither was significant. Pain remained significantly associated with admission in this multivariable model (Table 2).
Table 2. OR and 95% CI from Generalized Estimating Equations for Hospital Admission Within 30 Days of an Encounter
| Model | OR | 95% CI | P-value |
|---|---|---|---|
| Univariate model | |||
| Pain score | |||
| 0–3 | 1.000 | – | |
| 4–6 | 1.428 | 1.321–1.544 | <0.001 |
| 7–10 | 1.962 | 1.779–2.163 | <0.001 |
| Multivariate model | |||
| Gender | 0.36 | ||
| Male | 1.000 | — | |
| Female | 1.041 | 0.956–1.135 | |
| Age, decade increase | 0.995 | 0.992–0.998 | 0.003 |
| Pain score | |||
| 0–3 | 1.000 | — | |
| 4–6 | 1.429 | 1.322–1.544 | <0.001 |
| 7–10 | 1.963 | 1.780–2.164 | <0.001 |
To determine the role of tumor histology, all encounters were categorized into one of 15 disease categories (Table 3). Diagnoses were missing for 2,879 encounters, and these were not included in this analysis. Hematologic malignancies made up the largest category (34%) and included leukemia, myelodysplastic syndrome, lymphoma, and multiple myeloma. The second largest group was genitourinary cancer (10%), which included prostate, renal cell, testis, bladder, and other urinary organs. Melanoma and nonepithelial skin cancers formed the smallest disease category (0.5%), but 9% of the group had high pain scores, and admission rates were highest for those with high pain scores (58%) (Table 3). Head and neck cancers were evaluated as a separate entity from upper aerodigestive cancers, which included thyroid and esophageal cancer, and had a greater percentage of encounters with both a high pain score and subsequent admission (2.3% vs. 1.6%). Breast and genitourinary cancers had the lowest admission rates for those with high pain scores. In a subset analysis of the hematologic malignancies, multiple myeloma encounters were more likely to have high pain scores (5%) compared with leukemia (3.5%) and lymphoma (3.3%). However, controlling for pain and age, multiple myeloma encounter patients were much less likely to be admitted than lymphoma (OR: 0.50; 95% CI: 0.39–0.64) or leukemia (OR: 0.31; 95% CI: 0.24–0.39) encounter patients.
Table 3. Pain Scores and Admissions by Disease Category
| Disease Category | Patient Encounters | ||||
|---|---|---|---|---|---|
| Number | Pain Score 7–10 | Admissions Within 30 Days | Admission Rate for Those With Pain Score 7–10 | Pain Score 7–10 and Admitteda | |
| n (%) | n (%) | ||||
| Head and neck | 7,518 | 595 (7.9) | 1,531 (20.4) | 28.9 | 2.3 |
| Lung/bronchus | 11,186 | 691 (6.2) | 1,692 (15.1) | 29.7 | 1.8 |
| Upper aerodigestive | 3,834 | 166 (4.3) | 862 (22.5) | 36.1 | 1.6 |
| Gastrointestinal | 9,716 | 432 (4.4) | 1,194 (12.3) | 28.0 | 1.2 |
| Pancreas | 5,996 | 195 (3.2) | 873 (14.6) | 34.9 | 1.1 |
| Sarcoma | 1,899 | 65 (3.4) | 544 (28.6) | 41.5 | 1.4 |
| Melanoma/skin | 575 | 52 (9.0) | 137 (23.8) | 57.7 | 5.2 |
| Breast | 6,491 | 366 (5.6) | 487 (7.5) | 14.8 | 0.8 |
| Female genital | 4,516 | 195 (4.3) | 866 (19.2) | 38.5 | 1.7 |
| Genitourinary | 12,179 | 399 (3.3) | 1,013 (8.3) | 24.6 | 0.8 |
| Brain | 2,973 | 96 (3.2) | 415 (14.0) | 29.2 | 0.9 |
| Heme malignancies | 39,908 | 1469 (3.7) | 7,432 (18.6) | 30.4 | 1.1 |
| Benign heme | 2,091 | 72 (3.4) | 299 (14.3) | 34.7 | 1.2 |
| Benign disease | 1,839 | 95 (5.2) | 169 (9.2) | 26.3 | 1.4 |
| Cancer, unknown/other | 3,113 | 201 (6.5) | 513 (16.5) | 27.9 | 1.8 |
aPercent of patient encounters with both a pain score of 7–10 and admitted within 30 days. |
The location of the outpatient visit also was reviewed. Radiation oncology encounters were associated with higher pain scores compared with medical oncology encounters. Severe pain was reported in 7.5% of radiation oncology encounters compared with 4.1% of medical oncology encounters. After adjusting for pain in a logistic regression analysis, radiation oncology encounters were slightly more likely to be admitted (OR: 1.09; 95% CI: 0.99–1.19). The various disease types were not controlled for in this analysis.
Discussion
The findings from this study suggest that outpatient pain ratings are an easy and a feasible means to identify patients who might be at risk of early hospitalization and, thus, targeted for interventions. The strengths of the study lie in the impressive size of the database and the ability to evaluate the interactions of pain scores with age, gender, and cancer diagnosis on a large scale.
Our primary objective was to identify potentially high-risk groups that may benefit from palliative care interventions. We were particularly interested in the elderly, given described differences in the perception of pain and attitudes about how pain should be managed in this population.10, 11, 12 In a survey of over 13,000 elderly cancer patients, 29% had daily pain and 26% of these patients were not receiving any analgesics.13 Interestingly, our study did not demonstrate a difference in the severity of pain scores based on age, although patients who were older were slightly less likely to be admitted with severe pain. Our study suggests that underreporting of pain is not entirely implicated as one of the barriers for achieving adequate analgesia in the elderly. Many unmeasured factors are likely involved in these disparities, meriting further exploration.
Certain diseases were associated with higher risks of severe pain and hospitalizations. Although the strong association of severe pain and hospitalization was intuitive in particular cancers, such as head and neck, the significance with other malignancies was more perplexing. For example, melanoma and skin cancers that are not classically considered “painful” cancers had the highest rate of high pain scores and hospitalizations. Similarly, the treatment modality appeared to impact results, with radiation oncology encounters having higher rates of severe pain and admissions. This likely reflects the selective referral of patients to radiation oncology for palliation of painful lesions.
There are limitations to this retrospective study. The increased hospitalization rates of patients with moderate and severe pain are striking; yet the database only captured hospitalizations to the Johns Hopkins Medical Institutions. As a result, hospital admission rates are likely underestimated. Although obtaining outside records would have increased the number of hospitalizations, it is unlikely that it would have changed our findings. In addition, more in-depth and prospective assessment of the patients' pain would have provided substantially more information. Unfortunately, this was not available in the database. As a result, it remains uncertain whether unbearable pain was what prompted the admission or whether the patients were clinically deteriorating in general, and pain was an associated finding. Nevertheless, our data clearly demonstrate that patients with moderate to severe pain ratings are much more likely to require hospitalization within one month. Finally, although additional information comparing admission and discharge pain scores may have provided information on the success of pain interventions, this information was not present in the database and was also beyond the scope of this research study.
Referrals to palliative care and pain specialists have been shown to significantly improve pain control.14, 15 These resources are limited; thus, proper utilization is essential. In reviewing our retrospective data, seeking expert consultation for encounters with moderate and severe pain would have resulted in nearly 20 consults per day. Palliative care experts have recommended implementing a classification system for cancer pain to help guide clinicians in anticipating the need for specialist input.16 The appropriate implementation of resources in these “at risk” populations could potentially improve quality of life and decrease hospitalizations. Our study strongly suggests that outpatients with cancer and severe pain are at high risk of hospitalization and, therefore, constitute an ideal population for future targeted intervention studies.
References
- Pain in ambulatory patients with lung or colon cancer. Prevalence, characteristics, and effect. Cancer. 1992;70:1616–1624
- . The epidemiology of cancer pain. Cancer Invest. 2005;23:182–190
- Pain and its treatment in outpatients with metastatic cancer. N Engl J Med. 1994;330:592–596
- National Comprehensive Cancer Network Adult Pain Guidelines. J Pain Palliat Care Pharmacother. 2006;20:94
- Members NCPP. NCCN Adult Pain Guidelines V.I, 2009. NCCN: Ft. Washington, PA: 2009.
- American Pain Society recommendations for improving the quality of acute and cancer pain management. American Pain Society Quality of Care Task Force. Arch Intern Med. 2005;165:1574–1580
- JCAHO. Standards, intents, examples and scoring questions for pain assessment and management—comprehensive accreditation manual for hospitals. Oakbrook Terrace, IL: JCAHO Department of Standards, 1999:1–11.
- . Feasibility of quantitative pain assessment in outpatient oncology practice. J Clin Oncol. 2001;19:501–508
- . High outpatient patient scores identify patients at high risk for inpatient hospital admission. [Abstract] Proc Am Soc Clin Oncol. 2003;22:737
- . Management of pain in the older person with cancer. Oncology (Williston Park). 2008;22:56–61
- . Age-related differences in endogenous pain modulation: a comparison of diffuse noxious inhibitory controls in healthy older and younger adults. Pain. 2003;101:155–165
- . Age-related differences in pain perception and report. Clin Geriatr Med. 2001;17:433–456v–vi
- Management of pain in elderly patients with cancer. SAGE study group. Systematic assessment of geriatric drug use via epidemiology. JAMA. 1998;279:1877–1882
- . An assessment of methods used to evaluate the adequacy of cancer pain management. J Pain Symptom Manage. 2006;32:581–588
- Impact of a half-day multidisciplinary symptom control and palliative care outpatient clinic in a comprehensive cancer center on recommendations, symptom intensity, and patient satisfaction: a retrospective descriptive study. J Pain Symptom Manage. 2004;27:481–491
- . Cancer pain assessment—can we predict the need for specialist input?. Eur J Cancer. 2008;44:1072–1077
This study was accepted as an abstract for publication at the 44th Annual American Society of Clinical Oncology (ASCO) Meeting in May 2008 and at the 4th Annual Chicago Supportive Oncology Conference in October 2008.
Kathryn A. Carson's work on this manuscript was supported by Grant Number UL1 RR 025005 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH) and NIH Roadmap for Medical Research.
The authors declare no conflicts of interest.
PII: S0885-3924(09)00845-8
doi:10.1016/j.jpainsymman.2009.06.012
© 2010 U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved.
Volume 39, Issue 2 , Pages 180-185, February 2010
