Acute diverticulitis: demographic, clinical and laboratory features associated with computed tomography findings in 741 patients

Authors


Correspondence to:

Dr G. F. Longstreth, Department of Research and Evaluation, Kaiser Permanente Southern California 100 S. Los Robles, Pasadena, CA 91101, USA.

E-mail: george.f.longstreth@kp.org

Summary

Background

Computed tomography (CT) demonstrates diverticulitis severity.

Aim

To assess demographic, clinical and leucocyte features in association with severity.

Methods

We reviewed medical records of 741 emergency department cases and in-patients with diverticulitis. CT findings were: (i) nondiagnostic; (ii) moderate (peri-colic inflammation); and (iii) severe (abscess and/or extra-luminal gas and/or contrast).

Results

Patients with severe vs. nondiagnostic/moderate findings had fewer females (42.4% vs. 58.2%, P = .004), less lower abdominal pain only (74.7% vs. 83.7%, P = .042) and more constipation (24.4% vs. 12.5%, P = .002), fever (52.2% vs. 27.0%, P < .0001), leucocytosis (81.5% vs. 55.2%, P < .0001), neutrophilia (86.2% vs. 59.0%, P < .0001), ‘bandemia’ (18.5% vs. 5.5%, P < .0001) and the triad of abdominal pain, fever and leucocytosis (46.7% vs. 19.9%, P < .0001) respectively. Severe vs. nondiagnostic/moderate findings occurred in 4.8% vs. 95.2% without fever or leucocytosis, 7.0% vs. 93.0% with fever, 12.3% vs. 87.7% with leucocytosis and 25.1% vs. 74.9% with fever and leucocytosis respectively (P < .0001). The former group (odds ratio [95% CI]) included females less often (0.45 [0.26–0.76]) and had less lower abdominal pain only (0.54 [0.29–0.99]) and more constipation (2.32 [1.27–4.23]), fever (2.13 [1.27–3.57]) and leucocytosis (2.67 [1.43–4.99]).

Conclusions

Less than 50% of severe cases have the clinical/laboratory triad of abdominal pain, fever and leucocytosis, but only 1 of 20 with pain who lack fever and leucocytosis have severe diverticulitis. Male gender, pain not limited to the lower abdomen, constipation, fever and leucocytosis are independently associated with severe diverticulitis.

Introduction

Colonic diverticulosis affects up to 65% of Western populations by age 80.[1] In the US, acute diverticulitis (AD) accounts for a majority of the 1.5 million days of in-patient (IP) care annually for diverticular disease, and hospitalisation is increasing.[2, 3] Even more patients are treated without hospitalisation.[4-6]

Acute diverticulitis is often diagnosed on history and physical examination, but clinical evaluation can be inaccurate.[6-13] The American College of Radiology[14] and the American Society of Colon and Rectal Surgeons[15] recommend computed tomography (CT) as the preferred imaging procedure to diagnose AD and its complications,[15] and Ambrosetti[9, 10] and modified Hinchey[16] classifications of AD severity are based on CT findings. Although antibiotic therapy is standard,[17] Swedish investigators have used CT to help identify patients who can be successfully treated without antibiotics.[18] Measurement of abscess size can indicate whether antibiotic therapy alone or drainage is required,[17, 19] and CT findings can predict recurrence.[16]

The use of CT in US emergency departments for the evaluation of acute abdominal pain increased more than twofold between 2001 and 2005, but detection rates of AD did not increase.[20] In contrast to generally increasing CT use, some authors argue against any initial imaging in patients suspected to have AD without peritonitis and instead suggest reliance on clinical diagnosis.[21] Furthermore, routine use of CT should be reconsidered in light of recent evidence concerning the potential for radiation exposure from CT to cause cancer.[22, 23]

Physicians often suspect AD in patients with the triad of abdominal pain, fever and leucocytosis, and a modified Hinchey classification depends on these features to diagnose mild clinical AD without confirmation by imaging or surgery.[16] Abdominal tenderness is usually present.[4, 8, 11-13] Obesity increases the risk of diverticular disease.[24, 25] In the past few years, physicians have used demographic, clinical and laboratory features to more accurately diagnose AD,[8, 11-13] but there are few data on these factors in relation to disease severity, which is generally related to the severity of CT findings. Knowledge of such features that are associated with CT findings typical of severe AD could help physicians utilise this imaging procedure more judiciously and reduce cost and risk.

We studied patients in a large, integrated healthcare system who were diagnosed with AD after undergoing abdominopelvic CT and treated in an emergency department (ED) or after hospitalisation. We aimed to investigate their demographic and clinical features, including body mass index (BMI) and leucocyte counts and assess the association of these variables with CT results that ranged from nondiagnostic to those of severe AD.

Methods

Practice setting and data sources

The study was conducted in the Kaiser Permanente Medical Care Plan (KPMCP) of Southern California, a managed care organisation that serves 3.4 million members who resemble demographically the general population of southern California. A comprehensive electronic medical record system is used, and electronic databases summarise member-reported demographic features, International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) codes and Current Procedural Technology (CPT) codes.

In the IP and ED settings of this study, encounters are coded by professional coders using all the clinically relevant information. KPMCP electronic data have been used extensively in research studies, and validation of the codes has revealed high accuracy; for example, >95% positive predictive value in identifying atrial fibrillation[26] and gastro-oesophageal reflux disease.[27]

Patients studied and data acquisition

We identified samples of patients aged 18–100 who were assigned ICD-9-CM codes 562.11 or 562.13 for AD between 1 January 2008 and 31 August 2009. We randomly selected samples of patients from the entire KPMCP population, stratifying them 1:1 by care in an ED or hospital IP setting, yielding 1576 patients. IP cases included patients who had initially received ED care but who returned to the ED with persisting symptoms and required IP care. Because only about 10% of cases treated in out-patient offices without emergency department or hospital care undergo CT, we did not study these patients. Figure 1 details the disposition of the initially identified patients. A KPMCP database provided race/ethnicity data. Other demographic characteristics and clinical features were obtained from comprehensive medical record review according to detailed rules by a research associate who conferred with the primary author to resolve ambiguities. Using ICD-9-CM and CPT codes, we excluded patients with previous colorectal malignancy,[28] Crohn's disease[29] or colorectal resection. We also excluded patients without an abdominopelvic CT report, external-claim cases (transferred records tend to be less complete than KPMCP records), and patients whose ICD-9-CM code was not a final diagnosis or referred to a historical episode, elective surgery or major diverticular haemorrhage. When physician notes in the records did not confirm diagnosis and treatment of AD, cases were excluded as erroneous.[30] When physicians had considered another diagnosis, the primary author excluded patients who had another disease than AD that appeared responsible for the gastrointestinal symptoms. This was a clinical decision based on the preponderance of evidence, as AD cannot be diagnosed with 100% accuracy in some patients based on retrospective review. Finally, we excluded patients whose CT reports lacked a comment on at least one of four abnormalities: colonic wall thickening, peri-colic inflammation (phlegmon or ‘fat stranding’), abscess or extra-luminal gas or contrast, leaving 741 patients who had been diagnosed and treated for AD after undergoing CT.

Figure 1.

Derivation of the study population. CPT, current procedural technology; CT, computed tomography; ICD-9-CM, international classification of diseases, 9th revision, clinical modification.

Due to missing data, the number of cases in which factors were assessed varied. Missing symptom data were inferred as absent only if records indicated that all other symptoms except those recorded were absent; symptoms were never inferred as present. We analysed the highest oral temperature and maximum counts of total leucocytes, neutrophils and ‘band cells’ (immature neutrophils) during the illness, defined as elevated if >11 000/mm3, >7700/mm3 and >700/mm3, respectively,[31] calculating the latter two counts from the highest leucocyte count accompanied by a differential count. The BMI was calculated, preferably from data during the AD episode. Data unavailable at that time were imputed from the height recorded at any time and the most recent weight measurement during 1 year before presentation or, if unavailable then, up to one year afterward. The BMI classification (kg/m2) was: underweight (<18.5), normal (18.5–24.9), overweight (25.0–29.9) and obesity I (30.0–34.9), obesity II (35.0–39.9) and extreme obesity (≥40).[32]

CT severity classification

We ranked CT severity as: (i) nondiagnostic; (ii) moderate (inflammation of peri-colic fat); and (iii) severe (abscess and/or extra-luminal gas and/or extra-luminal contrast). Moderate or severe classification required unequivocal abnormalities; for example, ‘possible’ or ‘probable’ findings alone were insufficient. The nondiagnostic category comprised cases with only equivocal findings, colonic wall thickening without another ranking criterion, unequivocal but nonranking findings, and ‘diverticulitis’ with unspecified findings. The moderate category is identical to stage Ia of a modified Hinchey classification, and the severe category comprises the remaining stages, Ib through IV.[16] Also, the moderate and severe categories resemble the two-stage Ambrosetti criteria,[9, 10] except for lacking localised colonic wall thickening, which generally occurs when more advanced findings are present.

Data analysis

We analysed data with sas version 9.2 software (SAS Institute, Cary, NC, USA). Univariate analysis comprised Student's t-test for continuous variables and the chi-squared and Fisher's exact tests for categorical variables, accepting statistical significance as .05 by two-tailed testing. We performed multivariate logistic regression to evaluate the odds ratios for the association of CT severity with age, gender, non-Hispanic white race, pain limited to the lower abdomen, nausea, vomiting, diarrhoea, constipation, rectal bleeding, BMI ≥25 kg/m2, abdominal tenderness, oral temperature >37.5 °C and leucocyte count >11 000/mm.[3]

Results

Patient features

Comparison of the 741 studied patients with the 499 cases excluded due to lack of CT data revealed that included cases were younger (57.2 ± 16.1 vs. 63.5 ± 14.4 years, P < .0001) but similar in proportion of females (417 [56.3%] vs. 257 [51.5%], P = .10). Similar proportions had fever (218/724 [30.1%] vs. 95/327 [29.1%], P = .73), but included patients more often had leucocytosis (430/735 [58.5%] vs. 158/336 [47.0%], P = .0005) and the triad of abdominal pain, fever and leucocytosis (167/719 [23.2%] vs. 25/287 [8.7%], P < .0001).

Table 1 shows the patient features. The number of patients in each age cohort was: 18–45, 193 (26%); 46–55, 154 (20.8%); 56–65, 154 (20.8%) and >65, 240 (32.4%). Females predominated slightly. Race/ethnicity was: non-Hispanic white, 485 (65.6%); non-Hispanic African-American, 67 (9.0%); non-Hispanic Asian/Pacific islander, 37 (5.0%); Hispanic, 71 (9.6%); others 79 (10.7%) and unknown, 2 (0.3%).

Table 1. Demographic, clinical and leucocyte features of 741 patients with acute diverticulitis treated in an emergency department or after hospital admission
FeatureaTotalEmergency departmentHospitalP-value
  1. s.d., standard deviation; BMI, body mass index.

  2. a

    Categorical data expressed as n/n with data on feature (%).

  3. b

    Abdominal pain, oral temperature >37.5 °C and leucocytes >11 000/mm3.

Age, years (mean ± s.d.)57.2 ± 16.155.4 ± 15.559.8 ± 16.60.0002
Female417/741 (56.3)237/431 (55.0)180/310 (58.1)0.41
Non-Hispanic white485/739 (65.6)273/430 (63.5)212/309 (68.6)0.15
Abdominal pain722/740 (97.6)422/430 (98.1)300/310 (96.8)0.23
Lower abdomen only586/709 (82.7)372/422 (88.2)214/287 (74.6) 
Not limited to lower123/709 (17.3)50/422 (11.8)73/287 (25.4)<0.0001
Nausea278/731 (38.0)143/427 (33.5)135/304 (44.4)0.0023
Vomiting118/730 (16.2)39/427 (9.1)79/303 (26.0)<0.0001
Diarrhoea170/732 (23.2)82/427 (19.2)88/305 (28.9)0.002
Constipation102/731 (14.0)50/426 (11.7)52/305 (17.1)0.04
Rectal bleeding50/732 (6.8)25/427 (5.9)25/305 (8.2)0.22
BMI ≥25 kg/m2592/736 (80.4)357/426 (83.8)235/310 (75.8)0.007
Abdominal tenderness655/734 (89.2)391/424 (92.2)264/310 (85.2)0.002
Oral temperature >37.5 °C218/724 (30.1)66/420 (15.7)152/304 (50.0)<0.0001
Leucocytes >11 000/mm3430/735 (58.5)202/426 (47.4)228/309 (73.8)<0.0001
Neutrophils >7700/mm3377/609 (61.9)200/384 (52.1)177/225 (78.7)<0.0001
‘Band cells’ >700/mm342/609 (6.9)13/384 (3.4)29/225 (12.9)<0.0001
Clinical/laboratory triadb167/719 (23.2)46/416 (11.1)121/303 (39.9)<0.0001

Nearly all patients had abdominal pain, usually in the lower abdomen only. A majority had leucocytosis, a minority had fever and nearly one-fourth had the clinical/laboratory triad of abdominal pain, fever and leucocytosis. Compared with ED patients, IP cases were older and more often had nausea, vomiting, diarrhoea, constipation, fever, leucocytosis, neutrophilia, ‘bandemia,’ and the triad. ED patients more often had lower abdominal pain only, overweight/obesity and abdominal tenderness. Gender, non-Hispanic white race/ethnicity and rectal bleeding were similar. BMI categories were: underweight, 7 (1.0%); normal, 137 (18.6%); overweight, 252 (34.2%); obesity I, 194 (26.4%); obesity II, 87 (11.8%) and extreme obesity, 59 (8.0%).

Nondiagnostic, moderate and severe CT findings occurred in 66 (15.3%), 350 (81.2%) and 15 (3.5%) ED patients and 54 (17.4%), 179 (57.7%) and 77 (24.8%) IP cases respectively (P < .0001). The colonic site of AD in the 621 (83.8%) moderate and severe cases was: left, 530 (85.3%); right, 47 (7.6%); transverse, 7 (1.1%); and indeterminate, 37 (6.0%). Moderate severity predominated in both groups, but severe AD was much more common in IP cases. The 120 (16.2%) nondiagnostic reports comprised: focal colonic wall thickening, 36 (4.9%); ‘diverticulitis’, 10 (1.3%) or equivocal ‘diverticulitis’, 9 (1.2%) without ranking criteria; and equivocally reported thickening, 10 (1.3%), peri-colic inflammation, 18 (2.4%), abscess, 3 (0.4%), extra-luminal gas, 1 (0.1%) or fistula 1 (0.1%). Other findings without ranking criteria were: diverticulosis, 17 (2.3%), no unequivocal finding, 11 (1.5%) and colon dilation 4 (0.5%). Severe cases comprised: abscess, 42 (45.7%) and extra-luminal gas without abscess, 50 (54.3%).

Relation of features to severity of CT findings

Table 2 reveals that severe cases had significantly more constipation, fever, leucocytosis, neutrophilia, ‘bandemia,’ and the triad of abdominal pain, fever and leucocytosis; however, less than 50% of severe cases had the complete triad. Severe cases were less often female and had less lower abdominal pain only. Defining the clinical/laboratory triad with pain only in the lower abdomen revealed the triad in 31 of 81 (38.3%) and 91 of 607 (15.0%) of severe and nondiagnostic/moderate cases, respectively (P < 0.0001), fewer patients than without this restriction.

Table 2. Demographic, clinical and leucocyte features of 741 patients with acute diverticulitis in relation to severity of computed tomography findings
FeatureaTotalNondiagnostic/moderateSevereP-value
  1. s.d., standard deviation; BMI, body mass index.

  2. a

    Categorical data expressed as n/n with data on feature (%).

  3. b

    Abdominal pain, oral temperature >37.5 °C and leucocytes >11 000/mm3.

Age, years (mean ± s.d.)57.2 ± 16.157.3 ± 16.156.6 ± 15.90.71
Female417/741 (56.3)378/649 (58.2)39/92 (42.4)0.004
Non-Hispanic white485/739 (65.6)426/647 (65.8)59/92 (64.1)0.75
Abdominal pain722/740 (97.6)633/648 (97.7)89/92 (96.7)0.48
Lower abdomen only586/709 (82.7)524/626 (83.7)62/83 (74.7) 
Not limited to lower123/709 (17.3)102/626 (16.3)21/83 (25.3)0.042
Nausea278/731 (38.0)243/641 (37.9)35/90(38.9)0.86
Vomiting118/730 (16.2)102/640 (15.9)16/90 (17.8)0.66
Diarrhoea170/732 (23.2)152/642 (23.7)18/90 (20.0)0.44
Constipation102/731 (14.0)80/641 (12.5)22/90 (24.4)0.002
Rectal bleeding50/732 (6.8)43/642 (6.7)7/90 (7.8)0.70
BMI ≥25 kg/m2592/736 (80.4)525/645 (81.4)67/91 (73.6)0.08
Abdominal tenderness655/734 (89.2)575/643 (89.4)80/91 (87.9)0.66
Oral temperature >37.5 °C218/724 (30.1)171/634 (27.0)47/90 (52.2)<0.0001
Leucocytes >11 000/mm3430/735 (58.5)355/643 (55.2)75/92 (81.5)<0.0001
Neutrophils >7700/mm3377/609 (61.9)321/544 (59.0)56/65 (86.2)<0.0001
‘Band cells’ >700/mm342/609 (6.9)30/544(5.5)12/65 (18.5)<0.0001
Clinical/laboratory triadb167/719 (23.2)125/629 (19.9)42/90 (46.7)<0.0001

As shown in Figure 2, only 1 of 20 patients with abdominal pain who lacked fever and leucocytosis had severe AD, but only 1 of 4 patients with both features had severe AD. Among the 570 cases with lower abdominal pain only and temperature and leucocyte data, severe vs. nondiagnostic/moderate CT findings occurred in 8 (3.8%) vs. 205 (96.2%) of 213 without fever or leucocytosis, 2 (6.1%) vs. 31 (93.9%) of 33 with fever, 19 (9.4%) vs. 183 (90.6%) of 202 with leucocytosis and 31 (25.4%) vs. 91 (74.6%) of 122 with fever and leucocytosis respectively (P < 0.0001). Thus, assessing cases with lower abdominal pain only revealed little difference.

Figure 2.

Histogram showing fever (oral temperature >37.5 °C) and leucocytosis (>11 000/mm3) in patients with nondiagnostic/moderate vs. severe computed tomography findings.

Multivariate regression analysis of features in relation to severity of CT findings

Regression analysis (Table 3) comparing patients with severe vs. nondiagnostic/moderate CT findings showed that with increased severity there were greater than twofold increases in constipation, fever and leucocytosis and negative associations of female gender and lower abdominal pain only.

Table 3. Odds ratio of demographic, clinical and laboratory features for severity of computed tomography in 741 patients with acute diverticulitis
FeatureOdds ratio95% Confidence ratioP-value
Nondiagnostic/moderate vs. severe
Age, years1.000.98–1.02088
Gender (Female = 1)0.450.26–0.76.003
Non-hispanic white race0.770.46–1.31.34
Lower abdominal pain only0.540.29–0.99.047
Nausea1.130.64–1.97.68
Vomiting1.260.61–2.60.54
Constipation2.321.27–4.23.006
Diarrhoea0.850.45–1.62.63
Rectal bleeding1.760.65–4.75.26
Body mass index, kg/m21.000.96–1.04.93
Abdominal tenderness1.540.50–4.71.45
Oral temperature >37.5 °C2.131.27–3.57.004
Leucocytes >11 000/mm32.671.43–4.99.002

Discussion

We studied a large cohort of ED and IP patients who were diagnosed and treated for AD after undergoing CT. The three most important findings were: (i) Fever and leucocytosis increase with severity, but the traditional clinical/laboratory triad of these features plus abdominal pain occurs in only about one-fourth of all patients; (ii) Few patients with abdominal pain who lack both fever and leucocytosis have severe AD; and (iii) Demographic, laboratory and other clinical characteristics also vary with severity, and some are independently associated with severe AD. These results extend previous publications and are potentially useful in practice.

In-patients had clinical and laboratory indicators of more severe illness than emergency department cases. Although pain was usually limited to the lower abdomen, this type of pain was negatively associated with severe AD. Only 30% of patients overall had fever, which occurred in barely a majority of patients with severe CT findings. Leucocytosis occurred in about 60% of all cases. Even in patients with severe disease, a little less than 50% had the clinical/laboratory triad. By comparison, the 1999 American College of Gastroenterology (ACG) practice guidelines state that a majority of patients with diverticulitis have fever,[33] and more recent reviews indicate that left-lower abdominal pain[7] or lower abdominal pain,[4] fever and leucocytosis are typically present[7] or common.[4] However, published data on these features are scarce, as reflected by the citation of only one supporting reference in these reviews,[33] a report by Ambrosetti et al. of rectal temperature >38 °C and leucocytes >11 000/mm3 in 40% and 53% of patients respectively.[9] In the modified Hinchey criteria proposed by Kaiser et al.,[16] the triad of left-lower abdominal pain, fever and leucocytosis can diagnose AD without imaging or surgical confirmation. However, The American Society of Colon and Rectal Surgeons assigned both the lowest level for source of evidence and lowest grade of recommendation to the usefulness of clinical and laboratory evaluation in diagnosing sigmoid diverticulitis.[15] Notably, the modified Hinchey criteria[16] could have included only a minority of our patients in the absence of CT even though we included abdominal pain at any site, which allowed more patients to fulfil the triad than when only lower abdominal pain was considered.

Since these major reviews, a Swedish report comprising out-patients and IPs revealed that only 29% and 25% had fever and leucocytosis, respectively, but these parameters were undefined.[11] Among Dutch hospital patients, 60% had a temperature (site unspecified) >37.5 °C, and the proportion with leucocytes >11 000/mm3 is unstated.[13] Other recent reports indicate that these and other features individually are diagnostically unreliable,[8, 12, 13] and Toorenvliet et al.[8] and Laurell et al.[11] reported sensitivities of only 68% and 64%, respectively, for the history, physical examination and laboratory indicators of inflammation. Andeweg et al.[13] proposed that the accuracy of clinical diagnosis of AD could be improved to 86% by scoring seven independent predictors of the diagnosis.[13] These studies did not assess patient features in relation to severity.

Absence of both fever and leucocytosis may be especially useful information to physicians, as such patients infrequently had severe AD, whether they had unrestricted abdominal pain or lower abdominal pain only. In an uncontrolled case series, 96% of 193 patients with CT findings corresponding to our moderate severity level were successfully treated without antibiotics, and none required emergency surgery.[18] The infrequency of severe AD in patients without fever or leucocytosis could guide physicians to manage some patients without CT if their main reason for obtaining it is to identify severe AD. However, the presence of the clinical/laboratory triad in less than 50% of patients with severe AD calls for careful follow-up in cases after such care.

In contrast to others,[34] we found no association between younger age and severe AD. The female predominance overall resembles that of US patients nationwide[2] and other series,[9-13, 34] and we found that female gender was negatively associated with severe disease. We are unaware of comparison data for our finding of an association of non-Hispanic white race with severe disease. Constipation was less common than diarrhoea overall, consistent with other observations,[9, 11] but it was independently associated with severe AD. Infrequent rectal bleeding is consistent with ACG practice guidelines.[33] Our results suggest that physicians considering CT to confirm AD take into account not only the traditional clinical/laboratory triad but also demographic features and constipation.

The frequent finding of overweight or obese patients in our diverse cohort supports a small case series in which patients with perforated diverticulitis had higher BMIs than controls[35] and reports that men had increasing risk of hospitalisation for diverticular disease as their BMI increased[36] and that self-reported AD was higher in men with a BMI ≥30 kg/m2 than in those with a BMI <21 kg/m2.[24] We found only a statistically insignificant trend of increased weight/obesity with severity, taking into account potential confounding factors. This result contrasts with a recent study of women that revealed a BMI ≥25 kg/m2 compared with a lower BMI independently increased the risk of hospitalisation for diverticular disease, especially for perforation or abscess.[25] In contrast to that report, we studied both men and women and excluded patients without AD confirmed by record review, including those with diverticular haemorrhage, and we assessed different potential confounders, including symptoms, fever and leucocyte count. As obesity can interfere with ultrasonography, its high prevalence portends limited accuracy in some patients for this imaging procedure, which is favoured by some physicians.[8, 37]

Limitations of our study include its retrospective design. We could not assess severity of abdominal pain or the various meanings to patients of the terms, ‘diarrhoea’ and ‘constipation’.[38] Lack of prospective standardised data collection also complicated the interpretation of CT reports. We minimised this weakness by requiring specified and unequivocal radiologist statements to rank the severity of CT findings. This strict rule probably downgraded the severity of some cases whose CT findings could have been reported more definitively by other radiologists. However, the main effect of downgrading would minimise the findings related to severity. The estimated sensitivity of CT for AD of 87–97% based on peri-colic inflammation[37] implies that some patients have CT abnormalities less advanced than peri-colic inflammation, and some studies of ultrasonography accept colon thickening alone as diagnostic,[37] as was reported in 5% of our cases. In view of our potential downgrading of severity, the imperfect sensitivity of CT and our desire to include a broad range of disease severity, we included the 16% of patients with nondiagnostic CT findings as a milder level of severity than included in other classifications.[9, 10, 16] This procedure could have included patients who did not have AD, but we preferred not to exclude these patients whose illness was diagnosed and treated as AD, some of which had CT reports of ‘diverticulitis’ without specific comment on a minimum criterion required for higher severity ranking. However, we did not study some extremely ill patients, such as those who did not undergo CT because they went directly from the ED to urgent surgery, or those who received only terminal comfort care. Also, patients who did not undergo CT due to physician or patient preference were not studied. In patients excluded due to lack of CT data, the lower frequency of leucocytosis and the complete clinical/laboratory triad suggests they were less ill than the studied patients.

Strengths of the study include the large patient sample from a general population who were managed in an ED or after hospital admission. Out-patient treatment of AD following evaluation in an ED or clinic is common,[4-6] but we did not study patients seen only in the out-patient clinic because few undergo CT. Through use of electronic databases and comprehensive record review, we excluded cases whose diagnosis could be questioned by incomplete records, those that lacked physician documentation of diagnosis and therapy of AD, those with a history of disease or subsequent disease that could confound diagnosis,[28, 29] or miscoded cases[30] or those with another likely primary diagnosis. When there were multiple measurements of oral temperature and leucocyte count, we assessed their maximum levels. As all patients had prepaid care, variable use of CT related to type of insurance[22] was unlikely. The CT criteria for moderate and severe classification resembled standard classifications.[9, 10, 16]

Optimal use of CT in patients suspected to have AD based on clinical and laboratory features could reduce cost and radiation hazard. We identified demographic, clinical and laboratory features that are related to the severity of CT findings. Prospective study of the use of these features in deciding on CT and treatment could lead to firmer evidence of their usefulness.

Acknowledgements

Declaration of personal interests: Yen and Hodgkins are employees of Shire and hold stock and/or stock options in Shire. Declaration of funding interests: Grant support from Shire Development LLC. Longstreth received funding from Shire Development LLC in connection with the research.

Ancillary