Article, Gastroenterology

Derivation of a clinical prediction rule for evaluating patients with abdominal pain and diarrhea

Original Contribution

Derivation of a clinical prediction rule for evaluating patients with abdominal pain and diarrhea

Esther H. Chen MD?, Frances S. Shofer PhD, Anthony J. Dean MD, Judd E. Hollander MD, Jennifer L. Robey RN, Keara L. Sease MaEd, Angela M. Mills MD

Department of Emergency Medicine, Hospital of the University of Pennsylvania, Philadelphia, PA 19104-4283, USA

Received 11 July 2007; revised 27 July 2007; accepted 30 July 2007

Abstract

Objective: The objective of the study was to develop a simple prediction rule to reliably identify abdominal pain patients with diarrhea who may require surgical intervention.

Methods: We performed a secondary analysis of a prospective cohort study of adults with acute nontraumatic abdominal pain and diarrhea in an urban emergency department (ED). Structured data collection included 109 historical and 28 physical examination items, laboratory and radiographic results, and final diagnosis. The main outcome was operative intervention.

Results: One thousand patients were enrolled; 174 patients with diarrhea were included in this analysis. Patients had a mean age of 39 +- 16 years and were likely to be female (64%) and black (60%). Fifteen (9%) patients received a surgical intervention from the ED. Clinical variables associated with the need for surgical intervention using univariate analysis were age older than 40 years, constant pain, and peritonitis on examination. Using recursive partitioning multivariate analysis, the derived prediction rule included 2 variables: age older than 40 years and constant pain. This rule had a sensitivity of 1.0 (95% confidence interval, 0.78-1.0) and specificity of 0.23 (95% confidence interval, 0.16-0.30).

Conclusion: Patients older than 40 years with constant abdominal pain and diarrhea are likely to have a surgical cause of their symptoms.

(C) 2008

Introduction

Acute abdominal pain, responsible for 6.8% of total visits, is the most common presenting symptom of emergency department (ED) patients [1] and has a broad differential diagnosis, including benign and life-threatening diseases. The classic teaching in surgical and emergency medicine textbooks is that abdominal pain with concurrent diarrhea

* Corresponding author. Tel.: +1 215 349 8506; fax: +1 215 662 3953.

E-mail address: [email protected] (E.H. Chen).

suggests a diagnosis of gastroenteritis, infectious etiology, or if chronic and recurrent, inflammatory bowel disease [2,3]. Diarrhea, however, does not entirely exclude a surgical abdomen. In a retrospective study of ED patients with appendicitis, 22% had abdominal pain and diarrhea [4]. In another study of ischemic colitis diagnosed by Surgical pathology, 38% of the patients presented with diarrhea, sometimes without any accompanying pain [5].

The variability of symptoms and clinical features of many serious causes of abdominal pain has led to many physicians relying heavily on diagnostic testing. In one ED study, 65% of patients with undifferentiated nontraumatic abdominal

0735-6757/$ – see front matter (C) 2008 doi:10.1016/j.ajem.2007.07.023

Derivation of a clinical prediction rule 451

pain received at least one Imaging study during their evaluation [6]. The increase in diagnostic testing not only increases ED resource use, financial costs, and length of stay, but also exacerbates ED crowding. The increasing use of Abdominal computed tomography (CT) exposes patients to the additional risks of contrast and radiation exposure. Currently, there is no clinical guideline for predicting patients at risk for a surgical or emergent cause of their symptoms who might require further testing. We attempted to derive a clinical prediction rule for patients with abdominal pain and diarrhea who are likely to have a surgical abdomen.

Methods

Study design

We conducted a secondary analysis of a prospective cohort study of ED patients with acute abdominal pain. The institutional committee on research involving human sub- jects approved the study. Informed consent was obtained from all subjects.

Study setting and population

From April 5, 2004, to January 4, 2005, we enrolled nongravid adult patients with acute abdominal pain that presented to an urban university hospital ED with an annual census of approximately 55000 patient visits.

Study protocol

Trained research assistants identified and enrolled patients 18 years or older with nontraumatic abdominal pain of less than 72 hours’ duration. Pregnancy, abdominal trauma, and an abdominal surgical procedure within the previous 7 days were criteria for exclusion. Per protocol, patients had blood samples drawn for the investigation of serum diagnostic markers.

Research assistants recorded patient demographics, Laboratory test results, medications administered, and final disposition on a standardized data collection sheet. Treating physicians recorded patients’ clinical information, including features of the history, physical examination, results of diagnostic tests, and final diagnoses. The clinical variables recorded on the study data sheets were determined a priori by the investigators. All decisions regarding patient workup and management were made by the treating physicians indepen- dent of the study or the study investigators.

The final diagnoses of study patients were made in one of several ways. For admitted patients, operative inter- ventions and final diagnoses were obtained from the medical record at discharge. Patients who received an ED CT scan that revealed no significant pathology and those

who were hospitalized without an ED CT scan and subsequently discharged with symptom resolution but without a specific diagnosis were categorized as having nonspecific abdominal pain not requiring a surgical intervention. They received no further follow-up. Patients who were discharged from the ED without a CT scan received a standardized phone follow-up at 7 days and were interviewed regarding symptom resolution, follow-up physician’s visits, further radiographic testing, surgical procedures, and final diagnosis. Patients with Persistent symptoms were recontacted at 21 days after presentation. If they had not received a diagnosis by 21 days, they were considered to have a nonspecific, nonsurgical cause of their abdominal pain.

Final diagnoses were grouped into the broad categories (Table 2) by the 2 principal investigators who independently reviewed each patient’s chart. In cases where there was disagreement, a third member of the study team reviewed the case and a consensus was reached.

Measurements

Diarrhea was defined as an increase in the frequency of liquid stools. Surgical intervention was defined as an operative procedure requiring general anesthesia for the treatment of the patient’s abdominal pain.

Statistical analysis

The primary outcome measure was the need for acute surgical intervention. Descriptive data are presented as means +- standard deviation, frequencies, and percentages, with 95% confidence interval (CI) where appropriate. To determine which variables were associated with surgical intervention, Fisher exact test and Student t test for categorical and continuous variables were used, respec- tively. A multiple logistic regression analysis was used to identify factors that were independently associated with surgical intervention. Model selection was determined by using simultaneous inclusion and then backward elimina- tion. Only variables that were significant at the 0.1 level on univariate analysis were included. In addition, these potential variables were combined using recursive parti- tioning with CART analysis software (Version 6.0, Salford Systems, San Diego, CA) to optimize sensitivity while achieving maximum specificity. All other analyses were performed using SAS statistical software (Version 9.1, SAS Institute, Cary, NC).

Results

Of the 1000 patients with acute abdominal pain who were enrolled, 174 (17%) patients also had symptoms of diarrhea. Patient characteristics are described in Table 1.

452 E.H. Chen et al.

Table 1 Patient characteristics

Characteristic Age (+-SD)

Sex

Female Male Race

Frequency (%)

39 (+-16) y

111 (64)

63 (36)

Table 3 Univariate analysis of variables associated with the need for surgical intervention

Black

104 (60)

White

60 (34)

Other

10 (6)

Medical history abdominal surgery

43 (25)

Diabetes

17 (10)

Surgical intervention

15

bowel resection

7

Appendectomy

4

Cholecystectomy

2

Basket stone extraction

1

Salpingectomy

1

Radiography

CT imaging

92 (53)

Ultrasound

16 (9)

Disposition

Admission

66 (38)

Phone follow-up

Successful follow-up 167 (96)

Follow-up surgery 2

Variables

Surgery (n = 150) (%)

No surgery (n = 159) (%)

P

Age N40 y

12

(80)

58 (36)

.002

Previous abdominal surgery

3

(20)

40 (25)

NS

History

RUQ

4

(27)

23 (15)

NS

RLQ

6

(40)

29 (18)

.08

LUQ

0

(0)

12 (8)

NS

LLQ

2

(13)

30 (19)

NS

Diffuse

4

(27)

62 (39)

NS

Constant pain

14

(93)

97 (63)

.02

Vomiting

9

(60)

93 (59)

NS

Fever

1

(7)

29 (18)

NS

Blood in the stools

2

(14)

23 (14)

NS

Physical examination

No tenderness

1

(7)

19 (12)

NS

RUQ tenderness

4

(27)

24 (15)

NS

RLQ tenderness

5

(33)

29 (18)

NS

LUQ tenderness

1

(7)

16 (10)

NS

LLQ tenderness

3

(20)

28 (18)

NS

Diffuse tenderness

2

(13)

51 (32)

NS

Rebound

3

(20)

6 (4)

.03

RUQ indicates right upper quadrant; RLQ, right lower quadrant; LUQ, left upper quadrant; LLQ, left lower quadrant; NS, not significant.

Diagnostically, 100 (57%) patients received a radiologic study, of which 92 had CT imaging, 16 had Ultrasonography , and 8 had both. Clinical evaluation and diagnostic testing identified 15 (9%) patients requiring surgery. Sixty- six patients had an abNormal CT scan, 11 of whom had a surgical etiology identified for their pain (Table 1). Of the 26 patients who had a normal CT scan, one was diagnosed with acute cholecystitis by US and treated with cholecystectomy. Of the remaining 3 patients that underwent surgery, one had acute cholecystitis identified by US, one had ischemic bowel diagnosed by upper gastrointestinal series, and the third had a clinically diagnosed strangulated hernia. Final diagnoses are presented in Table 2.

Table 2 Final diagnoses

Diagnostic categories

Frequency (%)

Appendicitis

4 (2)

Gastroenteritis

61 (35)

Genitourinary disease

13 (7)

Gynecologic disease

13 (7)

Hepatobiliary disease

10 (6)

Nonspecific abdominal pain

24 (14)

Other gastrointestinal disease (eg, diverticulitis,

37 (21)

inflammatory bowel disease)

Other

12 (7)

At the end of the ED evaluation, 105 (60%) patients were discharged from the ED, 66 (38%) were admitted to the hospital, and 3 (2%) left against medical advice. On follow- up, 2 patients subsequently underwent elective outpatient cholecystectomy and 7 patients could not be contacted. Two of those that were lost to follow-up had initially left against medical advice. There were no reported deaths.

The clinical variables that correlated with the need for surgical intervention in patients with abdominal pain and diarrhea are shown in Table 3. We performed recursive partitioning multivariate analysis using only those highly correlated variables and developed a sensitive, clinically relevant model. A sensible rule containing 2 variables was derived, with a sensitivity of 1.0 (95% CI, 0.78-1.0) and specificity of 0.23 (95% CI, 0.16-0.30). According to this rule, patients older than 40 years with constant abdominal pain and diarrhea may have a surgical cause of their pain. Patients younger than 40 years without constant abdominal pain were unlikely to require surgical intervention.

Discussion

The classic teaching in surgery and emergency medicine textbooks is that diarrhea is a symptom often associated with abdominal pain and suggests a nonacute, infectious diagnosis such as gastroenteritis [2,3]. However, studies have shown that the presence of diarrhea does not entirely

Derivation of a clinical prediction rule 453

exclude a surgical etiology. In a retrospective study of ED patients with appendicitis, 22% had abdominal pain and diarrhea [4]. In another study of ischemic colitis diagnosed by surgical pathology, 38% of the patients presented with diarrhea, sometimes without any accompanying abdominal pain [5]. The variability of symptoms and clinical features of many serious causes of abdominal pain has increased the use of diagnostic imaging during the evaluation of these patients. We therefore attempted to derive a sensitive rule for predicting patients with acute abdominal pain and diarrhea who may have a surgical diagnosis of their symptoms.

In our patient population, the clinical variables that strongly correlated with requiring a surgical intervention included age older than 40 years, symptoms of constant pain, and rebound tenderness on physical examination. Using these, we developed a rule with 2 clinical variables that was highly sensitive and reliable for predicting the presence of a surgical process in abdominal pain patients with diarrhea. Patients without these 2 clinical variables were likely to have a nonsurgical cause of their pain and therefore may not benefit from further diagnostic imaging.

Limitations

There are several limitations to this study. Most importantly, because we have not validated this prediction rule prospectively in a demographically different population, this rule is not ready to be implemented clinically. We are in the process of designing a validation study in a larger patient population.

Our study population was young, which may limit the applicability of our prediction rule in older patients. The general guideline for decision rules is to have 10 outcome events for each clinical variable [7]. Therefore, a limitation of our study, with only 15 patients requiring surgical interven- tion, is the small number of patients with the main outcome. Our inclusion criteria used a qualitative (ie, frequent liquid stools), not quantitative, definition of diarrhea. Although it lacks precision, this definition is widely accepted in clinical practice because patients rarely measure the volume of their diarrhea [8]. The lack of standardization may have biased our

results. Furthermore, we did not standardize patient assess- ment, although our resident trainees are taught to ask standard questions regarding abdominal pain and to perform an adequate Abdominal examination. We also did not conduct an interobserver assessment of the reliability of peritoneal signs on physical examination, which has been shown in one study to be variable among clinicians [9].

Conclusions

We derived a sensitive clinical rule to identify acute abdominal pain patients with diarrhea who have a high risk of having a serious surgical diagnosis. Patients older than 40 years with constant abdominal pain and diarrhea are likely to have a surgical cause of their symptoms.

References

  1. McCaig LF, Nawar EW. National Hospital Ambulatory Medical Care Survey: 2004 emergency department summary. Adv Data 2006 (372):1-29.
  2. Jones RS, Claridge JA. Chapter 43. Acute abdomen. In: Townsend CM, Beauchamp RD, Evers BM, Mattox KL, editors. Sabiston textbook of surgery. Philadelphia (Pa): Saunders; 2004. p. 1219-38.
  3. King KE, Wightman JM. Chapter 22. Abdominal pain. In: Marx JA, editor. Rosen’s emergency medicine: concepts and clinical practice. Philadelphia (Pa): Mosby; 2006. p. 209-20.
  4. Guss DA, Richards C. Comparison of men and women presenting to an ED with acute appendicitis. Am J Emerg Med 2000;18(4):372-5.
  5. Huguier M, Barrier A, Boelle PY, et al. Ischemic colitis. Am J Surg 2006;192(5):679-84.
  6. Nagurney JT, Brown DF, Chang Y, et al. Use of diagnostic testing in the emergency department for patients presenting with non-traumatic abdominal pain. J Emerg Med 2003;25(4):363-71.
  7. Laupacis A, Sekar N, Stiell IG. Clinical prediction rules. A review and suggested modifications of methodological standards. JAMA 1997; 277(6):488-94.
  8. Ahlquist DA, Camilleri M. In: Braunwald E, editor. Diarrhea and constipation. New York (NY): McGraw-Hill; 2001. p. 241-50.
  9. Pines J, Uscher Pines L, Hall A, et al. The interrater variation of ED abdominal examination findings in patients with acute abdominal pain. Am J Emerg Med 2005;23(4):483-7.